]>
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> | |
ebe7e575 | 59 | #include <sys/zfs_ctldir.h> |
34dc7c2f BB |
60 | #include <sys/zfs_fuid.h> |
61 | #include <sys/bootconf.h> | |
62 | #include <sys/sunddi.h> | |
63 | #include <sys/dnlc.h> | |
64 | #include <sys/dmu_objset.h> | |
65 | #include <sys/spa_boot.h> | |
428870ff | 66 | #include <sys/sa.h> |
3558fd73 | 67 | #include <sys/zpl.h> |
428870ff | 68 | #include "zfs_comutil.h" |
34dc7c2f | 69 | |
9babb374 | 70 | |
34dc7c2f BB |
71 | /*ARGSUSED*/ |
72 | int | |
03f9ba9d | 73 | zfs_sync(struct super_block *sb, int wait, cred_t *cr) |
34dc7c2f | 74 | { |
03f9ba9d BB |
75 | zfs_sb_t *zsb = sb->s_fs_info; |
76 | ||
34dc7c2f BB |
77 | /* |
78 | * Data integrity is job one. We don't want a compromised kernel | |
79 | * writing to the storage pool, so we never sync during panic. | |
80 | */ | |
d5e53f9d | 81 | if (unlikely(oops_in_progress)) |
34dc7c2f BB |
82 | return (0); |
83 | ||
03f9ba9d BB |
84 | /* |
85 | * Semantically, the only requirement is that the sync be initiated. | |
86 | * The DMU syncs out txgs frequently, so there's nothing to do. | |
87 | */ | |
88 | if (!wait) | |
89 | return (0); | |
90 | ||
3558fd73 | 91 | if (zsb != NULL) { |
34dc7c2f BB |
92 | /* |
93 | * Sync a specific filesystem. | |
94 | */ | |
9babb374 | 95 | dsl_pool_t *dp; |
34dc7c2f | 96 | |
3558fd73 BB |
97 | ZFS_ENTER(zsb); |
98 | dp = dmu_objset_pool(zsb->z_os); | |
9babb374 BB |
99 | |
100 | /* | |
101 | * If the system is shutting down, then skip any | |
102 | * filesystems which may exist on a suspended pool. | |
103 | */ | |
03f9ba9d | 104 | if (spa_suspended(dp->dp_spa)) { |
3558fd73 | 105 | ZFS_EXIT(zsb); |
9babb374 BB |
106 | return (0); |
107 | } | |
108 | ||
3558fd73 BB |
109 | if (zsb->z_log != NULL) |
110 | zil_commit(zsb->z_log, 0); | |
428870ff | 111 | |
3558fd73 | 112 | ZFS_EXIT(zsb); |
34dc7c2f BB |
113 | } else { |
114 | /* | |
115 | * Sync all ZFS filesystems. This is what happens when you | |
116 | * run sync(1M). Unlike other filesystems, ZFS honors the | |
117 | * request by waiting for all pools to commit all dirty data. | |
118 | */ | |
119 | spa_sync_allpools(); | |
120 | } | |
121 | ||
122 | return (0); | |
123 | } | |
e5c39b95 | 124 | EXPORT_SYMBOL(zfs_sync); |
34dc7c2f | 125 | |
2cf7f52b BB |
126 | boolean_t |
127 | zfs_is_readonly(zfs_sb_t *zsb) | |
128 | { | |
129 | return (!!(zsb->z_sb->s_flags & MS_RDONLY)); | |
130 | } | |
131 | EXPORT_SYMBOL(zfs_is_readonly); | |
132 | ||
34dc7c2f BB |
133 | static void |
134 | atime_changed_cb(void *arg, uint64_t newval) | |
135 | { | |
2cf7f52b | 136 | ((zfs_sb_t *)arg)->z_atime = newval; |
34dc7c2f BB |
137 | } |
138 | ||
139 | static void | |
140 | xattr_changed_cb(void *arg, uint64_t newval) | |
141 | { | |
3558fd73 | 142 | zfs_sb_t *zsb = arg; |
34dc7c2f | 143 | |
82a37189 | 144 | if (newval == ZFS_XATTR_OFF) { |
2cf7f52b | 145 | zsb->z_flags &= ~ZSB_XATTR; |
82a37189 BB |
146 | } else { |
147 | zsb->z_flags |= ZSB_XATTR; | |
148 | ||
149 | if (newval == ZFS_XATTR_SA) | |
150 | zsb->z_xattr_sa = B_TRUE; | |
151 | else | |
152 | zsb->z_xattr_sa = B_FALSE; | |
153 | } | |
34dc7c2f BB |
154 | } |
155 | ||
156 | static void | |
157 | blksz_changed_cb(void *arg, uint64_t newval) | |
158 | { | |
3558fd73 | 159 | zfs_sb_t *zsb = arg; |
34dc7c2f BB |
160 | |
161 | if (newval < SPA_MINBLOCKSIZE || | |
162 | newval > SPA_MAXBLOCKSIZE || !ISP2(newval)) | |
163 | newval = SPA_MAXBLOCKSIZE; | |
164 | ||
3558fd73 | 165 | zsb->z_max_blksz = newval; |
34dc7c2f BB |
166 | } |
167 | ||
168 | static void | |
169 | readonly_changed_cb(void *arg, uint64_t newval) | |
170 | { | |
3558fd73 BB |
171 | zfs_sb_t *zsb = arg; |
172 | struct super_block *sb = zsb->z_sb; | |
34dc7c2f | 173 | |
2cf7f52b BB |
174 | if (sb == NULL) |
175 | return; | |
176 | ||
177 | if (newval) | |
3558fd73 | 178 | sb->s_flags |= MS_RDONLY; |
2cf7f52b | 179 | else |
3558fd73 | 180 | sb->s_flags &= ~MS_RDONLY; |
34dc7c2f BB |
181 | } |
182 | ||
183 | static void | |
184 | devices_changed_cb(void *arg, uint64_t newval) | |
185 | { | |
34dc7c2f BB |
186 | } |
187 | ||
188 | static void | |
189 | setuid_changed_cb(void *arg, uint64_t newval) | |
190 | { | |
34dc7c2f BB |
191 | } |
192 | ||
193 | static void | |
194 | exec_changed_cb(void *arg, uint64_t newval) | |
195 | { | |
34dc7c2f BB |
196 | } |
197 | ||
34dc7c2f BB |
198 | static void |
199 | nbmand_changed_cb(void *arg, uint64_t newval) | |
200 | { | |
3558fd73 BB |
201 | zfs_sb_t *zsb = arg; |
202 | struct super_block *sb = zsb->z_sb; | |
203 | ||
2cf7f52b BB |
204 | if (sb == NULL) |
205 | return; | |
206 | ||
207 | if (newval == TRUE) | |
3558fd73 | 208 | sb->s_flags |= MS_MANDLOCK; |
2cf7f52b | 209 | else |
3558fd73 | 210 | sb->s_flags &= ~MS_MANDLOCK; |
34dc7c2f BB |
211 | } |
212 | ||
213 | static void | |
214 | snapdir_changed_cb(void *arg, uint64_t newval) | |
215 | { | |
3558fd73 | 216 | ((zfs_sb_t *)arg)->z_show_ctldir = newval; |
34dc7c2f BB |
217 | } |
218 | ||
219 | static void | |
220 | vscan_changed_cb(void *arg, uint64_t newval) | |
221 | { | |
3558fd73 | 222 | ((zfs_sb_t *)arg)->z_vscan = newval; |
34dc7c2f BB |
223 | } |
224 | ||
34dc7c2f BB |
225 | static void |
226 | acl_inherit_changed_cb(void *arg, uint64_t newval) | |
227 | { | |
3558fd73 | 228 | ((zfs_sb_t *)arg)->z_acl_inherit = newval; |
34dc7c2f BB |
229 | } |
230 | ||
e5c39b95 | 231 | int |
3558fd73 | 232 | zfs_register_callbacks(zfs_sb_t *zsb) |
34dc7c2f BB |
233 | { |
234 | struct dsl_dataset *ds = NULL; | |
3558fd73 | 235 | objset_t *os = zsb->z_os; |
34dc7c2f BB |
236 | int error = 0; |
237 | ||
2cf7f52b BB |
238 | if (zfs_is_readonly(zsb) || !spa_writeable(dmu_objset_spa(os))) |
239 | readonly_changed_cb(zsb, B_TRUE); | |
34dc7c2f BB |
240 | |
241 | /* | |
242 | * Register property callbacks. | |
243 | * | |
244 | * It would probably be fine to just check for i/o error from | |
245 | * the first prop_register(), but I guess I like to go | |
246 | * overboard... | |
247 | */ | |
248 | ds = dmu_objset_ds(os); | |
3558fd73 BB |
249 | error = dsl_prop_register(ds, |
250 | "atime", atime_changed_cb, zsb); | |
34dc7c2f | 251 | error = error ? error : dsl_prop_register(ds, |
3558fd73 | 252 | "xattr", xattr_changed_cb, zsb); |
34dc7c2f | 253 | error = error ? error : dsl_prop_register(ds, |
3558fd73 | 254 | "recordsize", blksz_changed_cb, zsb); |
34dc7c2f | 255 | error = error ? error : dsl_prop_register(ds, |
3558fd73 | 256 | "readonly", readonly_changed_cb, zsb); |
34dc7c2f | 257 | error = error ? error : dsl_prop_register(ds, |
3558fd73 | 258 | "devices", devices_changed_cb, zsb); |
34dc7c2f | 259 | error = error ? error : dsl_prop_register(ds, |
3558fd73 | 260 | "setuid", setuid_changed_cb, zsb); |
34dc7c2f | 261 | error = error ? error : dsl_prop_register(ds, |
3558fd73 | 262 | "exec", exec_changed_cb, zsb); |
34dc7c2f | 263 | error = error ? error : dsl_prop_register(ds, |
3558fd73 | 264 | "snapdir", snapdir_changed_cb, zsb); |
34dc7c2f | 265 | error = error ? error : dsl_prop_register(ds, |
3558fd73 | 266 | "aclinherit", acl_inherit_changed_cb, zsb); |
34dc7c2f | 267 | error = error ? error : dsl_prop_register(ds, |
3558fd73 | 268 | "vscan", vscan_changed_cb, zsb); |
2cf7f52b BB |
269 | error = error ? error : dsl_prop_register(ds, |
270 | "nbmand", nbmand_changed_cb, zsb); | |
34dc7c2f BB |
271 | if (error) |
272 | goto unregister; | |
273 | ||
34dc7c2f BB |
274 | return (0); |
275 | ||
276 | unregister: | |
277 | /* | |
278 | * We may attempt to unregister some callbacks that are not | |
279 | * registered, but this is OK; it will simply return ENOMSG, | |
280 | * which we will ignore. | |
281 | */ | |
3558fd73 BB |
282 | (void) dsl_prop_unregister(ds, "atime", atime_changed_cb, zsb); |
283 | (void) dsl_prop_unregister(ds, "xattr", xattr_changed_cb, zsb); | |
284 | (void) dsl_prop_unregister(ds, "recordsize", blksz_changed_cb, zsb); | |
285 | (void) dsl_prop_unregister(ds, "readonly", readonly_changed_cb, zsb); | |
286 | (void) dsl_prop_unregister(ds, "devices", devices_changed_cb, zsb); | |
287 | (void) dsl_prop_unregister(ds, "setuid", setuid_changed_cb, zsb); | |
288 | (void) dsl_prop_unregister(ds, "exec", exec_changed_cb, zsb); | |
289 | (void) dsl_prop_unregister(ds, "snapdir", snapdir_changed_cb, zsb); | |
34dc7c2f | 290 | (void) dsl_prop_unregister(ds, "aclinherit", acl_inherit_changed_cb, |
3558fd73 BB |
291 | zsb); |
292 | (void) dsl_prop_unregister(ds, "vscan", vscan_changed_cb, zsb); | |
2cf7f52b | 293 | (void) dsl_prop_unregister(ds, "nbmand", nbmand_changed_cb, zsb); |
34dc7c2f | 294 | |
3558fd73 | 295 | return (error); |
34dc7c2f | 296 | } |
e5c39b95 | 297 | EXPORT_SYMBOL(zfs_register_callbacks); |
34dc7c2f | 298 | |
428870ff BB |
299 | static int |
300 | zfs_space_delta_cb(dmu_object_type_t bonustype, void *data, | |
301 | uint64_t *userp, uint64_t *groupp) | |
9babb374 | 302 | { |
428870ff BB |
303 | znode_phys_t *znp = data; |
304 | int error = 0; | |
9babb374 | 305 | |
428870ff BB |
306 | /* |
307 | * Is it a valid type of object to track? | |
308 | */ | |
309 | if (bonustype != DMU_OT_ZNODE && bonustype != DMU_OT_SA) | |
310 | return (ENOENT); | |
9babb374 | 311 | |
428870ff BB |
312 | /* |
313 | * If we have a NULL data pointer | |
314 | * then assume the id's aren't changing and | |
315 | * return EEXIST to the dmu to let it know to | |
316 | * use the same ids | |
317 | */ | |
318 | if (data == NULL) | |
319 | return (EEXIST); | |
9babb374 | 320 | |
428870ff BB |
321 | if (bonustype == DMU_OT_ZNODE) { |
322 | *userp = znp->zp_uid; | |
323 | *groupp = znp->zp_gid; | |
9babb374 | 324 | } else { |
428870ff | 325 | int hdrsize; |
9babb374 | 326 | |
428870ff BB |
327 | ASSERT(bonustype == DMU_OT_SA); |
328 | hdrsize = sa_hdrsize(data); | |
329 | ||
330 | if (hdrsize != 0) { | |
331 | *userp = *((uint64_t *)((uintptr_t)data + hdrsize + | |
332 | SA_UID_OFFSET)); | |
333 | *groupp = *((uint64_t *)((uintptr_t)data + hdrsize + | |
334 | SA_GID_OFFSET)); | |
335 | } else { | |
336 | /* | |
337 | * This should only happen for newly created | |
338 | * files that haven't had the znode data filled | |
339 | * in yet. | |
340 | */ | |
341 | *userp = 0; | |
342 | *groupp = 0; | |
343 | } | |
9babb374 | 344 | } |
428870ff | 345 | return (error); |
9babb374 BB |
346 | } |
347 | ||
348 | static void | |
3558fd73 | 349 | fuidstr_to_sid(zfs_sb_t *zsb, const char *fuidstr, |
9babb374 BB |
350 | char *domainbuf, int buflen, uid_t *ridp) |
351 | { | |
9babb374 BB |
352 | uint64_t fuid; |
353 | const char *domain; | |
354 | ||
355 | fuid = strtonum(fuidstr, NULL); | |
356 | ||
3558fd73 | 357 | domain = zfs_fuid_find_by_idx(zsb, FUID_INDEX(fuid)); |
9babb374 BB |
358 | if (domain) |
359 | (void) strlcpy(domainbuf, domain, buflen); | |
360 | else | |
361 | domainbuf[0] = '\0'; | |
362 | *ridp = FUID_RID(fuid); | |
363 | } | |
364 | ||
365 | static uint64_t | |
3558fd73 | 366 | zfs_userquota_prop_to_obj(zfs_sb_t *zsb, zfs_userquota_prop_t type) |
9babb374 BB |
367 | { |
368 | switch (type) { | |
369 | case ZFS_PROP_USERUSED: | |
370 | return (DMU_USERUSED_OBJECT); | |
371 | case ZFS_PROP_GROUPUSED: | |
372 | return (DMU_GROUPUSED_OBJECT); | |
373 | case ZFS_PROP_USERQUOTA: | |
3558fd73 | 374 | return (zsb->z_userquota_obj); |
9babb374 | 375 | case ZFS_PROP_GROUPQUOTA: |
3558fd73 | 376 | return (zsb->z_groupquota_obj); |
149e873a BB |
377 | default: |
378 | return (ENOTSUP); | |
9babb374 BB |
379 | } |
380 | return (0); | |
381 | } | |
382 | ||
383 | int | |
3558fd73 | 384 | zfs_userspace_many(zfs_sb_t *zsb, zfs_userquota_prop_t type, |
9babb374 BB |
385 | uint64_t *cookiep, void *vbuf, uint64_t *bufsizep) |
386 | { | |
387 | int error; | |
388 | zap_cursor_t zc; | |
389 | zap_attribute_t za; | |
390 | zfs_useracct_t *buf = vbuf; | |
391 | uint64_t obj; | |
392 | ||
3558fd73 | 393 | if (!dmu_objset_userspace_present(zsb->z_os)) |
9babb374 BB |
394 | return (ENOTSUP); |
395 | ||
3558fd73 | 396 | obj = zfs_userquota_prop_to_obj(zsb, type); |
9babb374 BB |
397 | if (obj == 0) { |
398 | *bufsizep = 0; | |
399 | return (0); | |
400 | } | |
401 | ||
3558fd73 | 402 | for (zap_cursor_init_serialized(&zc, zsb->z_os, obj, *cookiep); |
9babb374 BB |
403 | (error = zap_cursor_retrieve(&zc, &za)) == 0; |
404 | zap_cursor_advance(&zc)) { | |
405 | if ((uintptr_t)buf - (uintptr_t)vbuf + sizeof (zfs_useracct_t) > | |
406 | *bufsizep) | |
407 | break; | |
408 | ||
3558fd73 | 409 | fuidstr_to_sid(zsb, za.za_name, |
9babb374 BB |
410 | buf->zu_domain, sizeof (buf->zu_domain), &buf->zu_rid); |
411 | ||
412 | buf->zu_space = za.za_first_integer; | |
413 | buf++; | |
414 | } | |
415 | if (error == ENOENT) | |
416 | error = 0; | |
417 | ||
418 | ASSERT3U((uintptr_t)buf - (uintptr_t)vbuf, <=, *bufsizep); | |
419 | *bufsizep = (uintptr_t)buf - (uintptr_t)vbuf; | |
420 | *cookiep = zap_cursor_serialize(&zc); | |
421 | zap_cursor_fini(&zc); | |
422 | return (error); | |
423 | } | |
e5c39b95 | 424 | EXPORT_SYMBOL(zfs_userspace_many); |
9babb374 BB |
425 | |
426 | /* | |
427 | * buf must be big enough (eg, 32 bytes) | |
428 | */ | |
429 | static int | |
3558fd73 | 430 | id_to_fuidstr(zfs_sb_t *zsb, const char *domain, uid_t rid, |
9babb374 BB |
431 | char *buf, boolean_t addok) |
432 | { | |
433 | uint64_t fuid; | |
434 | int domainid = 0; | |
435 | ||
436 | if (domain && domain[0]) { | |
3558fd73 | 437 | domainid = zfs_fuid_find_by_domain(zsb, domain, NULL, addok); |
9babb374 BB |
438 | if (domainid == -1) |
439 | return (ENOENT); | |
440 | } | |
441 | fuid = FUID_ENCODE(domainid, rid); | |
442 | (void) sprintf(buf, "%llx", (longlong_t)fuid); | |
443 | return (0); | |
444 | } | |
445 | ||
446 | int | |
3558fd73 | 447 | zfs_userspace_one(zfs_sb_t *zsb, zfs_userquota_prop_t type, |
9babb374 BB |
448 | const char *domain, uint64_t rid, uint64_t *valp) |
449 | { | |
450 | char buf[32]; | |
451 | int err; | |
452 | uint64_t obj; | |
453 | ||
454 | *valp = 0; | |
455 | ||
3558fd73 | 456 | if (!dmu_objset_userspace_present(zsb->z_os)) |
9babb374 BB |
457 | return (ENOTSUP); |
458 | ||
3558fd73 | 459 | obj = zfs_userquota_prop_to_obj(zsb, type); |
9babb374 BB |
460 | if (obj == 0) |
461 | return (0); | |
462 | ||
3558fd73 | 463 | err = id_to_fuidstr(zsb, domain, rid, buf, B_FALSE); |
9babb374 BB |
464 | if (err) |
465 | return (err); | |
466 | ||
3558fd73 | 467 | err = zap_lookup(zsb->z_os, obj, buf, 8, 1, valp); |
9babb374 BB |
468 | if (err == ENOENT) |
469 | err = 0; | |
470 | return (err); | |
471 | } | |
e5c39b95 | 472 | EXPORT_SYMBOL(zfs_userspace_one); |
9babb374 BB |
473 | |
474 | int | |
3558fd73 | 475 | zfs_set_userquota(zfs_sb_t *zsb, zfs_userquota_prop_t type, |
9babb374 BB |
476 | const char *domain, uint64_t rid, uint64_t quota) |
477 | { | |
478 | char buf[32]; | |
479 | int err; | |
480 | dmu_tx_t *tx; | |
481 | uint64_t *objp; | |
482 | boolean_t fuid_dirtied; | |
483 | ||
484 | if (type != ZFS_PROP_USERQUOTA && type != ZFS_PROP_GROUPQUOTA) | |
485 | return (EINVAL); | |
486 | ||
3558fd73 | 487 | if (zsb->z_version < ZPL_VERSION_USERSPACE) |
9babb374 BB |
488 | return (ENOTSUP); |
489 | ||
3558fd73 BB |
490 | objp = (type == ZFS_PROP_USERQUOTA) ? &zsb->z_userquota_obj : |
491 | &zsb->z_groupquota_obj; | |
9babb374 | 492 | |
3558fd73 | 493 | err = id_to_fuidstr(zsb, domain, rid, buf, B_TRUE); |
9babb374 BB |
494 | if (err) |
495 | return (err); | |
3558fd73 | 496 | fuid_dirtied = zsb->z_fuid_dirty; |
9babb374 | 497 | |
3558fd73 | 498 | tx = dmu_tx_create(zsb->z_os); |
9babb374 BB |
499 | dmu_tx_hold_zap(tx, *objp ? *objp : DMU_NEW_OBJECT, B_TRUE, NULL); |
500 | if (*objp == 0) { | |
501 | dmu_tx_hold_zap(tx, MASTER_NODE_OBJ, B_TRUE, | |
502 | zfs_userquota_prop_prefixes[type]); | |
503 | } | |
504 | if (fuid_dirtied) | |
3558fd73 | 505 | zfs_fuid_txhold(zsb, tx); |
9babb374 BB |
506 | err = dmu_tx_assign(tx, TXG_WAIT); |
507 | if (err) { | |
508 | dmu_tx_abort(tx); | |
509 | return (err); | |
510 | } | |
511 | ||
3558fd73 | 512 | mutex_enter(&zsb->z_lock); |
9babb374 | 513 | if (*objp == 0) { |
3558fd73 | 514 | *objp = zap_create(zsb->z_os, DMU_OT_USERGROUP_QUOTA, |
9babb374 | 515 | DMU_OT_NONE, 0, tx); |
3558fd73 | 516 | VERIFY(0 == zap_add(zsb->z_os, MASTER_NODE_OBJ, |
9babb374 BB |
517 | zfs_userquota_prop_prefixes[type], 8, 1, objp, tx)); |
518 | } | |
3558fd73 | 519 | mutex_exit(&zsb->z_lock); |
9babb374 BB |
520 | |
521 | if (quota == 0) { | |
3558fd73 | 522 | err = zap_remove(zsb->z_os, *objp, buf, tx); |
9babb374 BB |
523 | if (err == ENOENT) |
524 | err = 0; | |
525 | } else { | |
3558fd73 | 526 | err = zap_update(zsb->z_os, *objp, buf, 8, 1, "a, tx); |
9babb374 BB |
527 | } |
528 | ASSERT(err == 0); | |
529 | if (fuid_dirtied) | |
3558fd73 | 530 | zfs_fuid_sync(zsb, tx); |
9babb374 BB |
531 | dmu_tx_commit(tx); |
532 | return (err); | |
533 | } | |
e5c39b95 | 534 | EXPORT_SYMBOL(zfs_set_userquota); |
9babb374 BB |
535 | |
536 | boolean_t | |
3558fd73 | 537 | zfs_fuid_overquota(zfs_sb_t *zsb, boolean_t isgroup, uint64_t fuid) |
9babb374 BB |
538 | { |
539 | char buf[32]; | |
540 | uint64_t used, quota, usedobj, quotaobj; | |
541 | int err; | |
542 | ||
543 | usedobj = isgroup ? DMU_GROUPUSED_OBJECT : DMU_USERUSED_OBJECT; | |
3558fd73 | 544 | quotaobj = isgroup ? zsb->z_groupquota_obj : zsb->z_userquota_obj; |
9babb374 | 545 | |
3558fd73 | 546 | if (quotaobj == 0 || zsb->z_replay) |
9babb374 BB |
547 | return (B_FALSE); |
548 | ||
549 | (void) sprintf(buf, "%llx", (longlong_t)fuid); | |
3558fd73 | 550 | err = zap_lookup(zsb->z_os, quotaobj, buf, 8, 1, "a); |
9babb374 BB |
551 | if (err != 0) |
552 | return (B_FALSE); | |
553 | ||
3558fd73 | 554 | err = zap_lookup(zsb->z_os, usedobj, buf, 8, 1, &used); |
9babb374 BB |
555 | if (err != 0) |
556 | return (B_FALSE); | |
557 | return (used >= quota); | |
558 | } | |
e5c39b95 | 559 | EXPORT_SYMBOL(zfs_fuid_overquota); |
9babb374 | 560 | |
428870ff | 561 | boolean_t |
3558fd73 | 562 | zfs_owner_overquota(zfs_sb_t *zsb, znode_t *zp, boolean_t isgroup) |
428870ff BB |
563 | { |
564 | uint64_t fuid; | |
565 | uint64_t quotaobj; | |
428870ff | 566 | |
3558fd73 | 567 | quotaobj = isgroup ? zsb->z_groupquota_obj : zsb->z_userquota_obj; |
428870ff | 568 | |
572e2857 | 569 | fuid = isgroup ? zp->z_gid : zp->z_uid; |
428870ff | 570 | |
3558fd73 | 571 | if (quotaobj == 0 || zsb->z_replay) |
428870ff BB |
572 | return (B_FALSE); |
573 | ||
3558fd73 | 574 | return (zfs_fuid_overquota(zsb, isgroup, fuid)); |
428870ff | 575 | } |
e5c39b95 | 576 | EXPORT_SYMBOL(zfs_owner_overquota); |
428870ff | 577 | |
9babb374 | 578 | int |
3558fd73 | 579 | zfs_sb_create(const char *osname, zfs_sb_t **zsbp) |
9babb374 BB |
580 | { |
581 | objset_t *os; | |
3558fd73 | 582 | zfs_sb_t *zsb; |
9babb374 BB |
583 | uint64_t zval; |
584 | int i, error; | |
428870ff | 585 | uint64_t sa_obj; |
9babb374 | 586 | |
bafc4e9e | 587 | zsb = kmem_zalloc(sizeof (zfs_sb_t), KM_SLEEP | KM_NODEBUG); |
9babb374 | 588 | |
428870ff BB |
589 | /* |
590 | * We claim to always be readonly so we can open snapshots; | |
591 | * other ZPL code will prevent us from writing to snapshots. | |
592 | */ | |
3558fd73 | 593 | error = dmu_objset_own(osname, DMU_OST_ZFS, B_TRUE, zsb, &os); |
428870ff | 594 | if (error) { |
3558fd73 | 595 | kmem_free(zsb, sizeof (zfs_sb_t)); |
9babb374 | 596 | return (error); |
428870ff | 597 | } |
9babb374 BB |
598 | |
599 | /* | |
600 | * Initialize the zfs-specific filesystem structure. | |
601 | * Should probably make this a kmem cache, shuffle fields, | |
602 | * and just bzero up to z_hold_mtx[]. | |
603 | */ | |
2cf7f52b | 604 | zsb->z_sb = NULL; |
3558fd73 BB |
605 | zsb->z_parent = zsb; |
606 | zsb->z_max_blksz = SPA_MAXBLOCKSIZE; | |
607 | zsb->z_show_ctldir = ZFS_SNAPDIR_VISIBLE; | |
608 | zsb->z_os = os; | |
9babb374 | 609 | |
3558fd73 | 610 | error = zfs_get_zplprop(os, ZFS_PROP_VERSION, &zsb->z_version); |
9babb374 BB |
611 | if (error) { |
612 | goto out; | |
3558fd73 | 613 | } else if (zsb->z_version > |
428870ff | 614 | zfs_zpl_version_map(spa_version(dmu_objset_spa(os)))) { |
149e873a | 615 | (void) printk("Can't mount a version %lld file system " |
428870ff | 616 | "on a version %lld pool\n. Pool must be upgraded to mount " |
3558fd73 | 617 | "this file system.", (u_longlong_t)zsb->z_version, |
428870ff | 618 | (u_longlong_t)spa_version(dmu_objset_spa(os))); |
9babb374 BB |
619 | error = ENOTSUP; |
620 | goto out; | |
621 | } | |
9babb374 BB |
622 | if ((error = zfs_get_zplprop(os, ZFS_PROP_NORMALIZE, &zval)) != 0) |
623 | goto out; | |
3558fd73 | 624 | zsb->z_norm = (int)zval; |
9babb374 BB |
625 | |
626 | if ((error = zfs_get_zplprop(os, ZFS_PROP_UTF8ONLY, &zval)) != 0) | |
627 | goto out; | |
3558fd73 | 628 | zsb->z_utf8 = (zval != 0); |
9babb374 BB |
629 | |
630 | if ((error = zfs_get_zplprop(os, ZFS_PROP_CASE, &zval)) != 0) | |
631 | goto out; | |
3558fd73 | 632 | zsb->z_case = (uint_t)zval; |
9babb374 BB |
633 | |
634 | /* | |
635 | * Fold case on file systems that are always or sometimes case | |
636 | * insensitive. | |
637 | */ | |
3558fd73 BB |
638 | if (zsb->z_case == ZFS_CASE_INSENSITIVE || |
639 | zsb->z_case == ZFS_CASE_MIXED) | |
640 | zsb->z_norm |= U8_TEXTPREP_TOUPPER; | |
9babb374 | 641 | |
3558fd73 BB |
642 | zsb->z_use_fuids = USE_FUIDS(zsb->z_version, zsb->z_os); |
643 | zsb->z_use_sa = USE_SA(zsb->z_version, zsb->z_os); | |
428870ff | 644 | |
3558fd73 | 645 | if (zsb->z_use_sa) { |
428870ff BB |
646 | /* should either have both of these objects or none */ |
647 | error = zap_lookup(os, MASTER_NODE_OBJ, ZFS_SA_ATTRS, 8, 1, | |
648 | &sa_obj); | |
649 | if (error) | |
591fb62f | 650 | goto out; |
82a37189 BB |
651 | |
652 | error = zfs_get_zplprop(os, ZFS_PROP_XATTR, &zval); | |
653 | if ((error == 0) && (zval == ZFS_XATTR_SA)) | |
654 | zsb->z_xattr_sa = B_TRUE; | |
428870ff BB |
655 | } else { |
656 | /* | |
657 | * Pre SA versions file systems should never touch | |
658 | * either the attribute registration or layout objects. | |
659 | */ | |
660 | sa_obj = 0; | |
661 | } | |
662 | ||
572e2857 | 663 | error = sa_setup(os, sa_obj, zfs_attr_table, ZPL_END, |
3558fd73 | 664 | &zsb->z_attr_table); |
572e2857 BB |
665 | if (error) |
666 | goto out; | |
428870ff | 667 | |
3558fd73 | 668 | if (zsb->z_version >= ZPL_VERSION_SA) |
428870ff | 669 | sa_register_update_callback(os, zfs_sa_upgrade); |
9babb374 BB |
670 | |
671 | error = zap_lookup(os, MASTER_NODE_OBJ, ZFS_ROOT_OBJ, 8, 1, | |
3558fd73 | 672 | &zsb->z_root); |
9babb374 BB |
673 | if (error) |
674 | goto out; | |
3558fd73 | 675 | ASSERT(zsb->z_root != 0); |
9babb374 BB |
676 | |
677 | error = zap_lookup(os, MASTER_NODE_OBJ, ZFS_UNLINKED_SET, 8, 1, | |
3558fd73 | 678 | &zsb->z_unlinkedobj); |
9babb374 BB |
679 | if (error) |
680 | goto out; | |
681 | ||
682 | error = zap_lookup(os, MASTER_NODE_OBJ, | |
683 | zfs_userquota_prop_prefixes[ZFS_PROP_USERQUOTA], | |
3558fd73 | 684 | 8, 1, &zsb->z_userquota_obj); |
9babb374 BB |
685 | if (error && error != ENOENT) |
686 | goto out; | |
687 | ||
688 | error = zap_lookup(os, MASTER_NODE_OBJ, | |
689 | zfs_userquota_prop_prefixes[ZFS_PROP_GROUPQUOTA], | |
3558fd73 | 690 | 8, 1, &zsb->z_groupquota_obj); |
9babb374 BB |
691 | if (error && error != ENOENT) |
692 | goto out; | |
693 | ||
694 | error = zap_lookup(os, MASTER_NODE_OBJ, ZFS_FUID_TABLES, 8, 1, | |
3558fd73 | 695 | &zsb->z_fuid_obj); |
9babb374 BB |
696 | if (error && error != ENOENT) |
697 | goto out; | |
698 | ||
699 | error = zap_lookup(os, MASTER_NODE_OBJ, ZFS_SHARES_DIR, 8, 1, | |
3558fd73 | 700 | &zsb->z_shares_dir); |
9babb374 BB |
701 | if (error && error != ENOENT) |
702 | goto out; | |
703 | ||
3558fd73 BB |
704 | mutex_init(&zsb->z_znodes_lock, NULL, MUTEX_DEFAULT, NULL); |
705 | mutex_init(&zsb->z_lock, NULL, MUTEX_DEFAULT, NULL); | |
706 | list_create(&zsb->z_all_znodes, sizeof (znode_t), | |
9babb374 | 707 | offsetof(znode_t, z_link_node)); |
3558fd73 BB |
708 | rrw_init(&zsb->z_teardown_lock); |
709 | rw_init(&zsb->z_teardown_inactive_lock, NULL, RW_DEFAULT, NULL); | |
710 | rw_init(&zsb->z_fuid_lock, NULL, RW_DEFAULT, NULL); | |
9babb374 | 711 | for (i = 0; i != ZFS_OBJ_MTX_SZ; i++) |
3558fd73 | 712 | mutex_init(&zsb->z_hold_mtx[i], NULL, MUTEX_DEFAULT, NULL); |
9babb374 | 713 | |
ebe7e575 BB |
714 | avl_create(&zsb->z_ctldir_snaps, snapentry_compare, |
715 | sizeof (zfs_snapentry_t), offsetof(zfs_snapentry_t, se_node)); | |
716 | mutex_init(&zsb->z_ctldir_lock, NULL, MUTEX_DEFAULT, NULL); | |
717 | ||
3558fd73 | 718 | *zsbp = zsb; |
9babb374 BB |
719 | return (0); |
720 | ||
721 | out: | |
3558fd73 BB |
722 | dmu_objset_disown(os, zsb); |
723 | *zsbp = NULL; | |
724 | kmem_free(zsb, sizeof (zfs_sb_t)); | |
9babb374 BB |
725 | return (error); |
726 | } | |
86f35f34 | 727 | EXPORT_SYMBOL(zfs_sb_create); |
9babb374 | 728 | |
86f35f34 | 729 | int |
3558fd73 | 730 | zfs_sb_setup(zfs_sb_t *zsb, boolean_t mounting) |
34dc7c2f | 731 | { |
34dc7c2f BB |
732 | int error; |
733 | ||
3558fd73 | 734 | error = zfs_register_callbacks(zsb); |
34dc7c2f BB |
735 | if (error) |
736 | return (error); | |
737 | ||
738 | /* | |
3558fd73 | 739 | * Set the objset user_ptr to track its zsb. |
34dc7c2f | 740 | */ |
3558fd73 BB |
741 | mutex_enter(&zsb->z_os->os_user_ptr_lock); |
742 | dmu_objset_set_user(zsb->z_os, zsb); | |
743 | mutex_exit(&zsb->z_os->os_user_ptr_lock); | |
34dc7c2f | 744 | |
3558fd73 | 745 | zsb->z_log = zil_open(zsb->z_os, zfs_get_data); |
9babb374 | 746 | |
34dc7c2f BB |
747 | /* |
748 | * If we are not mounting (ie: online recv), then we don't | |
749 | * have to worry about replaying the log as we blocked all | |
750 | * operations out since we closed the ZIL. | |
751 | */ | |
752 | if (mounting) { | |
b128c09f BB |
753 | boolean_t readonly; |
754 | ||
34dc7c2f BB |
755 | /* |
756 | * During replay we remove the read only flag to | |
757 | * allow replays to succeed. | |
758 | */ | |
2cf7f52b | 759 | readonly = zfs_is_readonly(zsb); |
fb5f0bc8 | 760 | if (readonly != 0) |
2cf7f52b | 761 | readonly_changed_cb(zsb, B_FALSE); |
fb5f0bc8 | 762 | else |
3558fd73 | 763 | zfs_unlinked_drain(zsb); |
34dc7c2f | 764 | |
428870ff BB |
765 | /* |
766 | * Parse and replay the intent log. | |
767 | * | |
768 | * Because of ziltest, this must be done after | |
769 | * zfs_unlinked_drain(). (Further note: ziltest | |
770 | * doesn't use readonly mounts, where | |
771 | * zfs_unlinked_drain() isn't called.) This is because | |
772 | * ziltest causes spa_sync() to think it's committed, | |
773 | * but actually it is not, so the intent log contains | |
774 | * many txg's worth of changes. | |
775 | * | |
776 | * In particular, if object N is in the unlinked set in | |
777 | * the last txg to actually sync, then it could be | |
778 | * actually freed in a later txg and then reallocated | |
779 | * in a yet later txg. This would write a "create | |
780 | * object N" record to the intent log. Normally, this | |
781 | * would be fine because the spa_sync() would have | |
782 | * written out the fact that object N is free, before | |
783 | * we could write the "create object N" intent log | |
784 | * record. | |
785 | * | |
786 | * But when we are in ziltest mode, we advance the "open | |
787 | * txg" without actually spa_sync()-ing the changes to | |
788 | * disk. So we would see that object N is still | |
789 | * allocated and in the unlinked set, and there is an | |
790 | * intent log record saying to allocate it. | |
791 | */ | |
3558fd73 | 792 | if (spa_writeable(dmu_objset_spa(zsb->z_os))) { |
572e2857 | 793 | if (zil_replay_disable) { |
3558fd73 | 794 | zil_destroy(zsb->z_log, B_FALSE); |
572e2857 | 795 | } else { |
3558fd73 BB |
796 | zsb->z_replay = B_TRUE; |
797 | zil_replay(zsb->z_os, zsb, | |
572e2857 | 798 | zfs_replay_vector); |
3558fd73 | 799 | zsb->z_replay = B_FALSE; |
572e2857 | 800 | } |
fb5f0bc8 | 801 | } |
2cf7f52b BB |
802 | |
803 | /* restore readonly bit */ | |
804 | if (readonly != 0) | |
805 | readonly_changed_cb(zsb, B_TRUE); | |
34dc7c2f BB |
806 | } |
807 | ||
34dc7c2f BB |
808 | return (0); |
809 | } | |
86f35f34 | 810 | EXPORT_SYMBOL(zfs_sb_setup); |
34dc7c2f | 811 | |
9babb374 | 812 | void |
3558fd73 | 813 | zfs_sb_free(zfs_sb_t *zsb) |
34dc7c2f | 814 | { |
9babb374 | 815 | int i; |
9babb374 | 816 | |
3558fd73 | 817 | zfs_fuid_destroy(zsb); |
9babb374 | 818 | |
3558fd73 BB |
819 | mutex_destroy(&zsb->z_znodes_lock); |
820 | mutex_destroy(&zsb->z_lock); | |
821 | list_destroy(&zsb->z_all_znodes); | |
822 | rrw_destroy(&zsb->z_teardown_lock); | |
823 | rw_destroy(&zsb->z_teardown_inactive_lock); | |
824 | rw_destroy(&zsb->z_fuid_lock); | |
9babb374 | 825 | for (i = 0; i != ZFS_OBJ_MTX_SZ; i++) |
3558fd73 | 826 | mutex_destroy(&zsb->z_hold_mtx[i]); |
ebe7e575 BB |
827 | mutex_destroy(&zsb->z_ctldir_lock); |
828 | avl_destroy(&zsb->z_ctldir_snaps); | |
3558fd73 | 829 | kmem_free(zsb, sizeof (zfs_sb_t)); |
34dc7c2f | 830 | } |
86f35f34 | 831 | EXPORT_SYMBOL(zfs_sb_free); |
34dc7c2f | 832 | |
9babb374 | 833 | static void |
3558fd73 | 834 | zfs_set_fuid_feature(zfs_sb_t *zsb) |
9babb374 | 835 | { |
3558fd73 BB |
836 | zsb->z_use_fuids = USE_FUIDS(zsb->z_version, zsb->z_os); |
837 | zsb->z_use_sa = USE_SA(zsb->z_version, zsb->z_os); | |
9babb374 | 838 | } |
34dc7c2f BB |
839 | |
840 | void | |
3558fd73 | 841 | zfs_unregister_callbacks(zfs_sb_t *zsb) |
34dc7c2f | 842 | { |
3558fd73 | 843 | objset_t *os = zsb->z_os; |
34dc7c2f BB |
844 | struct dsl_dataset *ds; |
845 | ||
846 | /* | |
847 | * Unregister properties. | |
848 | */ | |
849 | if (!dmu_objset_is_snapshot(os)) { | |
850 | ds = dmu_objset_ds(os); | |
851 | VERIFY(dsl_prop_unregister(ds, "atime", atime_changed_cb, | |
3558fd73 | 852 | zsb) == 0); |
34dc7c2f BB |
853 | |
854 | VERIFY(dsl_prop_unregister(ds, "xattr", xattr_changed_cb, | |
3558fd73 | 855 | zsb) == 0); |
34dc7c2f BB |
856 | |
857 | VERIFY(dsl_prop_unregister(ds, "recordsize", blksz_changed_cb, | |
3558fd73 | 858 | zsb) == 0); |
34dc7c2f BB |
859 | |
860 | VERIFY(dsl_prop_unregister(ds, "readonly", readonly_changed_cb, | |
3558fd73 | 861 | zsb) == 0); |
34dc7c2f BB |
862 | |
863 | VERIFY(dsl_prop_unregister(ds, "devices", devices_changed_cb, | |
3558fd73 | 864 | zsb) == 0); |
34dc7c2f BB |
865 | |
866 | VERIFY(dsl_prop_unregister(ds, "setuid", setuid_changed_cb, | |
3558fd73 | 867 | zsb) == 0); |
34dc7c2f BB |
868 | |
869 | VERIFY(dsl_prop_unregister(ds, "exec", exec_changed_cb, | |
3558fd73 | 870 | zsb) == 0); |
34dc7c2f BB |
871 | |
872 | VERIFY(dsl_prop_unregister(ds, "snapdir", snapdir_changed_cb, | |
3558fd73 | 873 | zsb) == 0); |
34dc7c2f | 874 | |
34dc7c2f | 875 | VERIFY(dsl_prop_unregister(ds, "aclinherit", |
3558fd73 | 876 | acl_inherit_changed_cb, zsb) == 0); |
34dc7c2f BB |
877 | |
878 | VERIFY(dsl_prop_unregister(ds, "vscan", | |
3558fd73 | 879 | vscan_changed_cb, zsb) == 0); |
2cf7f52b BB |
880 | |
881 | VERIFY(dsl_prop_unregister(ds, "nbmand", | |
882 | nbmand_changed_cb, zsb) == 0); | |
34dc7c2f BB |
883 | } |
884 | } | |
e5c39b95 | 885 | EXPORT_SYMBOL(zfs_unregister_callbacks); |
34dc7c2f | 886 | |
bc3e15e3 | 887 | #ifdef HAVE_MLSLABEL |
428870ff BB |
888 | /* |
889 | * zfs_check_global_label: | |
890 | * Check that the hex label string is appropriate for the dataset | |
891 | * being mounted into the global_zone proper. | |
892 | * | |
893 | * Return an error if the hex label string is not default or | |
894 | * admin_low/admin_high. For admin_low labels, the corresponding | |
895 | * dataset must be readonly. | |
896 | */ | |
897 | int | |
898 | zfs_check_global_label(const char *dsname, const char *hexsl) | |
899 | { | |
900 | if (strcasecmp(hexsl, ZFS_MLSLABEL_DEFAULT) == 0) | |
901 | return (0); | |
902 | if (strcasecmp(hexsl, ADMIN_HIGH) == 0) | |
903 | return (0); | |
904 | if (strcasecmp(hexsl, ADMIN_LOW) == 0) { | |
905 | /* must be readonly */ | |
906 | uint64_t rdonly; | |
907 | ||
908 | if (dsl_prop_get_integer(dsname, | |
909 | zfs_prop_to_name(ZFS_PROP_READONLY), &rdonly, NULL)) | |
910 | return (EACCES); | |
911 | return (rdonly ? 0 : EACCES); | |
912 | } | |
913 | return (EACCES); | |
914 | } | |
86f35f34 | 915 | EXPORT_SYMBOL(zfs_check_global_label); |
bc3e15e3 | 916 | #endif /* HAVE_MLSLABEL */ |
428870ff | 917 | |
e5c39b95 | 918 | int |
3558fd73 | 919 | zfs_statvfs(struct dentry *dentry, struct kstatfs *statp) |
34dc7c2f | 920 | { |
3558fd73 | 921 | zfs_sb_t *zsb = dentry->d_sb->s_fs_info; |
34dc7c2f | 922 | uint64_t refdbytes, availbytes, usedobjs, availobjs; |
04f9432d | 923 | uint64_t fsid; |
3558fd73 | 924 | uint32_t bshift; |
34dc7c2f | 925 | |
3558fd73 | 926 | ZFS_ENTER(zsb); |
34dc7c2f | 927 | |
3558fd73 | 928 | dmu_objset_space(zsb->z_os, |
34dc7c2f BB |
929 | &refdbytes, &availbytes, &usedobjs, &availobjs); |
930 | ||
04f9432d | 931 | fsid = dmu_objset_fsid_guid(zsb->z_os); |
34dc7c2f | 932 | /* |
05ff35c6 BB |
933 | * The underlying storage pool actually uses multiple block |
934 | * size. Under Solaris frsize (fragment size) is reported as | |
935 | * the smallest block size we support, and bsize (block size) | |
936 | * as the filesystem's maximum block size. Unfortunately, | |
937 | * under Linux the fragment size and block size are often used | |
938 | * interchangeably. Thus we are forced to report both of them | |
939 | * as the filesystem's maximum block size. | |
34dc7c2f | 940 | */ |
05ff35c6 | 941 | statp->f_frsize = zsb->z_max_blksz; |
3558fd73 BB |
942 | statp->f_bsize = zsb->z_max_blksz; |
943 | bshift = fls(statp->f_bsize) - 1; | |
34dc7c2f BB |
944 | |
945 | /* | |
3558fd73 BB |
946 | * The following report "total" blocks of various kinds in |
947 | * the file system, but reported in terms of f_bsize - the | |
948 | * "preferred" size. | |
34dc7c2f BB |
949 | */ |
950 | ||
3558fd73 BB |
951 | statp->f_blocks = (refdbytes + availbytes) >> bshift; |
952 | statp->f_bfree = availbytes >> bshift; | |
34dc7c2f BB |
953 | statp->f_bavail = statp->f_bfree; /* no root reservation */ |
954 | ||
955 | /* | |
956 | * statvfs() should really be called statufs(), because it assumes | |
957 | * static metadata. ZFS doesn't preallocate files, so the best | |
958 | * we can do is report the max that could possibly fit in f_files, | |
959 | * and that minus the number actually used in f_ffree. | |
960 | * For f_ffree, report the smaller of the number of object available | |
961 | * and the number of blocks (each object will take at least a block). | |
962 | */ | |
baab0630 | 963 | statp->f_ffree = MIN(availobjs, availbytes >> DNODE_SHIFT); |
34dc7c2f | 964 | statp->f_files = statp->f_ffree + usedobjs; |
04f9432d CP |
965 | statp->f_fsid.val[0] = (uint32_t)fsid; |
966 | statp->f_fsid.val[1] = (uint32_t)(fsid >> 32); | |
3558fd73 BB |
967 | statp->f_type = ZFS_SUPER_MAGIC; |
968 | statp->f_namelen = ZFS_MAXNAMELEN; | |
34dc7c2f BB |
969 | |
970 | /* | |
3558fd73 | 971 | * We have all of 40 characters to stuff a string here. |
34dc7c2f BB |
972 | * Is there anything useful we could/should provide? |
973 | */ | |
3558fd73 | 974 | bzero(statp->f_spare, sizeof (statp->f_spare)); |
34dc7c2f | 975 | |
3558fd73 | 976 | ZFS_EXIT(zsb); |
34dc7c2f BB |
977 | return (0); |
978 | } | |
e5c39b95 | 979 | EXPORT_SYMBOL(zfs_statvfs); |
34dc7c2f | 980 | |
e5c39b95 | 981 | int |
3558fd73 | 982 | zfs_root(zfs_sb_t *zsb, struct inode **ipp) |
34dc7c2f | 983 | { |
34dc7c2f BB |
984 | znode_t *rootzp; |
985 | int error; | |
986 | ||
3558fd73 | 987 | ZFS_ENTER(zsb); |
34dc7c2f | 988 | |
3558fd73 | 989 | error = zfs_zget(zsb, zsb->z_root, &rootzp); |
34dc7c2f | 990 | if (error == 0) |
3558fd73 | 991 | *ipp = ZTOI(rootzp); |
34dc7c2f | 992 | |
3558fd73 | 993 | ZFS_EXIT(zsb); |
34dc7c2f BB |
994 | return (error); |
995 | } | |
e5c39b95 | 996 | EXPORT_SYMBOL(zfs_root); |
34dc7c2f | 997 | |
ab26409d BB |
998 | #ifdef HAVE_SHRINK |
999 | int | |
1000 | zfs_sb_prune(struct super_block *sb, unsigned long nr_to_scan, int *objects) | |
1001 | { | |
1002 | zfs_sb_t *zsb = sb->s_fs_info; | |
1003 | struct shrinker *shrinker = &sb->s_shrink; | |
1004 | struct shrink_control sc = { | |
1005 | .nr_to_scan = nr_to_scan, | |
1006 | .gfp_mask = GFP_KERNEL, | |
1007 | }; | |
1008 | ||
1009 | ZFS_ENTER(zsb); | |
1010 | *objects = (*shrinker->shrink)(shrinker, &sc); | |
1011 | ZFS_EXIT(zsb); | |
1012 | ||
1013 | return (0); | |
1014 | } | |
1015 | EXPORT_SYMBOL(zfs_sb_prune); | |
1016 | #endif /* HAVE_SHRINK */ | |
1017 | ||
34dc7c2f | 1018 | /* |
3558fd73 | 1019 | * Teardown the zfs_sb_t::z_os. |
34dc7c2f BB |
1020 | * |
1021 | * Note, if 'unmounting' if FALSE, we return with the 'z_teardown_lock' | |
1022 | * and 'z_teardown_inactive_lock' held. | |
1023 | */ | |
3558fd73 | 1024 | int |
86f35f34 | 1025 | zfs_sb_teardown(zfs_sb_t *zsb, boolean_t unmounting) |
34dc7c2f BB |
1026 | { |
1027 | znode_t *zp; | |
1028 | ||
3558fd73 | 1029 | rrw_enter(&zsb->z_teardown_lock, RW_WRITER, FTAG); |
34dc7c2f BB |
1030 | |
1031 | if (!unmounting) { | |
1032 | /* | |
ceb43b93 BB |
1033 | * We purge the parent filesystem's super block as the |
1034 | * parent filesystem and all of its snapshots have their | |
1035 | * inode's super block set to the parent's filesystem's | |
1036 | * super block. Note, 'z_parent' is self referential | |
1037 | * for non-snapshots. | |
34dc7c2f | 1038 | */ |
ceb43b93 | 1039 | shrink_dcache_sb(zsb->z_parent->z_sb); |
e30c0ada | 1040 | (void) spl_invalidate_inodes(zsb->z_parent->z_sb, 0); |
34dc7c2f BB |
1041 | } |
1042 | ||
b00131d4 GB |
1043 | /* |
1044 | * Drain the iput_taskq to ensure all active references to the | |
1045 | * zfs_sb_t have been handled only then can it be safely destroyed. | |
1046 | */ | |
1047 | taskq_wait(dsl_pool_iput_taskq(dmu_objset_pool(zsb->z_os))); | |
1048 | ||
34dc7c2f BB |
1049 | /* |
1050 | * Close the zil. NB: Can't close the zil while zfs_inactive | |
1051 | * threads are blocked as zil_close can call zfs_inactive. | |
1052 | */ | |
3558fd73 BB |
1053 | if (zsb->z_log) { |
1054 | zil_close(zsb->z_log); | |
1055 | zsb->z_log = NULL; | |
34dc7c2f BB |
1056 | } |
1057 | ||
3558fd73 | 1058 | rw_enter(&zsb->z_teardown_inactive_lock, RW_WRITER); |
34dc7c2f BB |
1059 | |
1060 | /* | |
1061 | * If we are not unmounting (ie: online recv) and someone already | |
1062 | * unmounted this file system while we were doing the switcheroo, | |
1063 | * or a reopen of z_os failed then just bail out now. | |
1064 | */ | |
3558fd73 BB |
1065 | if (!unmounting && (zsb->z_unmounted || zsb->z_os == NULL)) { |
1066 | rw_exit(&zsb->z_teardown_inactive_lock); | |
1067 | rrw_exit(&zsb->z_teardown_lock, FTAG); | |
34dc7c2f BB |
1068 | return (EIO); |
1069 | } | |
1070 | ||
1071 | /* | |
1072 | * At this point there are no vops active, and any new vops will | |
1073 | * fail with EIO since we have z_teardown_lock for writer (only | |
1074 | * relavent for forced unmount). | |
1075 | * | |
1076 | * Release all holds on dbufs. | |
1077 | */ | |
3558fd73 BB |
1078 | mutex_enter(&zsb->z_znodes_lock); |
1079 | for (zp = list_head(&zsb->z_all_znodes); zp != NULL; | |
1080 | zp = list_next(&zsb->z_all_znodes, zp)) | |
428870ff | 1081 | if (zp->z_sa_hdl) { |
3558fd73 | 1082 | ASSERT(atomic_read(&ZTOI(zp)->i_count) > 0); |
34dc7c2f BB |
1083 | zfs_znode_dmu_fini(zp); |
1084 | } | |
3558fd73 | 1085 | mutex_exit(&zsb->z_znodes_lock); |
34dc7c2f BB |
1086 | |
1087 | /* | |
1088 | * If we are unmounting, set the unmounted flag and let new vops | |
1089 | * unblock. zfs_inactive will have the unmounted behavior, and all | |
1090 | * other vops will fail with EIO. | |
1091 | */ | |
1092 | if (unmounting) { | |
3558fd73 BB |
1093 | zsb->z_unmounted = B_TRUE; |
1094 | rrw_exit(&zsb->z_teardown_lock, FTAG); | |
1095 | rw_exit(&zsb->z_teardown_inactive_lock); | |
34dc7c2f BB |
1096 | } |
1097 | ||
1098 | /* | |
1099 | * z_os will be NULL if there was an error in attempting to reopen | |
3558fd73 BB |
1100 | * zsb, so just return as the properties had already been |
1101 | * | |
34dc7c2f BB |
1102 | * unregistered and cached data had been evicted before. |
1103 | */ | |
3558fd73 | 1104 | if (zsb->z_os == NULL) |
34dc7c2f BB |
1105 | return (0); |
1106 | ||
1107 | /* | |
1108 | * Unregister properties. | |
1109 | */ | |
3558fd73 | 1110 | zfs_unregister_callbacks(zsb); |
34dc7c2f BB |
1111 | |
1112 | /* | |
1113 | * Evict cached data | |
1114 | */ | |
3558fd73 | 1115 | if (dmu_objset_is_dirty_anywhere(zsb->z_os)) |
2cf7f52b | 1116 | if (!zfs_is_readonly(zsb)) |
3558fd73 BB |
1117 | txg_wait_synced(dmu_objset_pool(zsb->z_os), 0); |
1118 | (void) dmu_objset_evict_dbufs(zsb->z_os); | |
34dc7c2f BB |
1119 | |
1120 | return (0); | |
1121 | } | |
86f35f34 | 1122 | EXPORT_SYMBOL(zfs_sb_teardown); |
34dc7c2f | 1123 | |
5547c2f1 BB |
1124 | #if defined(HAVE_BDI) && !defined(HAVE_BDI_SETUP_AND_REGISTER) |
1125 | atomic_long_t zfs_bdi_seq = ATOMIC_LONG_INIT(0); | |
1126 | #endif /* HAVE_BDI && !HAVE_BDI_SETUP_AND_REGISTER */ | |
76659dc1 | 1127 | |
e5c39b95 | 1128 | int |
3558fd73 | 1129 | zfs_domount(struct super_block *sb, void *data, int silent) |
34dc7c2f | 1130 | { |
3558fd73 BB |
1131 | zpl_mount_data_t *zmd = data; |
1132 | const char *osname = zmd->z_osname; | |
1133 | zfs_sb_t *zsb; | |
1134 | struct inode *root_inode; | |
1135 | uint64_t recordsize; | |
1136 | int error; | |
34dc7c2f | 1137 | |
3558fd73 BB |
1138 | error = zfs_sb_create(osname, &zsb); |
1139 | if (error) | |
1140 | return (error); | |
1141 | ||
1142 | if ((error = dsl_prop_get_integer(osname, "recordsize", | |
1143 | &recordsize, NULL))) | |
1144 | goto out; | |
1145 | ||
1146 | zsb->z_sb = sb; | |
3558fd73 BB |
1147 | sb->s_fs_info = zsb; |
1148 | sb->s_magic = ZFS_SUPER_MAGIC; | |
1149 | sb->s_maxbytes = MAX_LFS_FILESIZE; | |
1150 | sb->s_time_gran = 1; | |
1151 | sb->s_blocksize = recordsize; | |
1152 | sb->s_blocksize_bits = ilog2(recordsize); | |
5547c2f1 BB |
1153 | |
1154 | #ifdef HAVE_BDI | |
1155 | /* | |
1156 | * 2.6.32 API change, | |
1157 | * Added backing_device_info (BDI) per super block interfaces. A BDI | |
1158 | * must be configured when using a non-device backed filesystem for | |
1159 | * proper writeback. This is not required for older pdflush kernels. | |
1160 | * | |
1161 | * NOTE: Linux read-ahead is disabled in favor of zfs read-ahead. | |
1162 | */ | |
1163 | zsb->z_bdi.ra_pages = 0; | |
1164 | sb->s_bdi = &zsb->z_bdi; | |
1165 | ||
1166 | error = -bdi_setup_and_register(&zsb->z_bdi, "zfs", BDI_CAP_MAP_COPY); | |
1167 | if (error) | |
1168 | goto out; | |
1169 | #endif /* HAVE_BDI */ | |
3558fd73 BB |
1170 | |
1171 | /* Set callback operations for the file system. */ | |
1172 | sb->s_op = &zpl_super_operations; | |
1173 | sb->s_xattr = zpl_xattr_handlers; | |
055656d4 | 1174 | sb->s_export_op = &zpl_export_operations; |
3558fd73 BB |
1175 | |
1176 | /* Set features for file system. */ | |
1177 | zfs_set_fuid_feature(zsb); | |
1178 | ||
1179 | if (dmu_objset_is_snapshot(zsb->z_os)) { | |
1180 | uint64_t pval; | |
1181 | ||
1182 | atime_changed_cb(zsb, B_FALSE); | |
1183 | readonly_changed_cb(zsb, B_TRUE); | |
1184 | if ((error = dsl_prop_get_integer(osname,"xattr",&pval,NULL))) | |
1185 | goto out; | |
1186 | xattr_changed_cb(zsb, pval); | |
1187 | zsb->z_issnap = B_TRUE; | |
1188 | zsb->z_os->os_sync = ZFS_SYNC_DISABLED; | |
1189 | ||
1190 | mutex_enter(&zsb->z_os->os_user_ptr_lock); | |
1191 | dmu_objset_set_user(zsb->z_os, zsb); | |
1192 | mutex_exit(&zsb->z_os->os_user_ptr_lock); | |
1193 | } else { | |
1194 | error = zfs_sb_setup(zsb, B_TRUE); | |
34dc7c2f BB |
1195 | } |
1196 | ||
3558fd73 BB |
1197 | /* Allocate a root inode for the filesystem. */ |
1198 | error = zfs_root(zsb, &root_inode); | |
1199 | if (error) { | |
1200 | (void) zfs_umount(sb); | |
1201 | goto out; | |
34dc7c2f BB |
1202 | } |
1203 | ||
3558fd73 | 1204 | /* Allocate a root dentry for the filesystem */ |
6a0936ba | 1205 | sb->s_root = d_make_root(root_inode); |
3558fd73 BB |
1206 | if (sb->s_root == NULL) { |
1207 | (void) zfs_umount(sb); | |
1208 | error = ENOMEM; | |
1209 | goto out; | |
1210 | } | |
ebe7e575 BB |
1211 | |
1212 | if (!zsb->z_issnap) | |
1213 | zfsctl_create(zsb); | |
3558fd73 BB |
1214 | out: |
1215 | if (error) { | |
1216 | dmu_objset_disown(zsb->z_os, zsb); | |
1217 | zfs_sb_free(zsb); | |
1218 | } | |
34dc7c2f | 1219 | |
3558fd73 BB |
1220 | return (error); |
1221 | } | |
1222 | EXPORT_SYMBOL(zfs_domount); | |
1223 | ||
ebe7e575 BB |
1224 | /* |
1225 | * Called when an unmount is requested and certain sanity checks have | |
1226 | * already passed. At this point no dentries or inodes have been reclaimed | |
1227 | * from their respective caches. We drop the extra reference on the .zfs | |
1228 | * control directory to allow everything to be reclaimed. All snapshots | |
1229 | * must already have been unmounted to reach this point. | |
1230 | */ | |
1231 | void | |
1232 | zfs_preumount(struct super_block *sb) | |
1233 | { | |
1234 | zfs_sb_t *zsb = sb->s_fs_info; | |
1235 | ||
1f0d8a56 | 1236 | if (zsb != NULL && zsb->z_ctldir != NULL) |
ebe7e575 BB |
1237 | zfsctl_destroy(zsb); |
1238 | } | |
1239 | EXPORT_SYMBOL(zfs_preumount); | |
1240 | ||
1241 | /* | |
1242 | * Called once all other unmount released tear down has occurred. | |
1243 | * It is our responsibility to release any remaining infrastructure. | |
1244 | */ | |
3558fd73 BB |
1245 | /*ARGSUSED*/ |
1246 | int | |
1247 | zfs_umount(struct super_block *sb) | |
1248 | { | |
1249 | zfs_sb_t *zsb = sb->s_fs_info; | |
1250 | objset_t *os; | |
1251 | ||
86f35f34 | 1252 | VERIFY(zfs_sb_teardown(zsb, B_TRUE) == 0); |
3558fd73 | 1253 | os = zsb->z_os; |
34dc7c2f | 1254 | |
5547c2f1 BB |
1255 | #ifdef HAVE_BDI |
1256 | bdi_destroy(sb->s_bdi); | |
1257 | #endif /* HAVE_BDI */ | |
76659dc1 | 1258 | |
34dc7c2f BB |
1259 | /* |
1260 | * z_os will be NULL if there was an error in | |
3558fd73 | 1261 | * attempting to reopen zsb. |
34dc7c2f BB |
1262 | */ |
1263 | if (os != NULL) { | |
1264 | /* | |
1265 | * Unset the objset user_ptr. | |
1266 | */ | |
428870ff | 1267 | mutex_enter(&os->os_user_ptr_lock); |
34dc7c2f | 1268 | dmu_objset_set_user(os, NULL); |
428870ff | 1269 | mutex_exit(&os->os_user_ptr_lock); |
34dc7c2f BB |
1270 | |
1271 | /* | |
b128c09f | 1272 | * Finally release the objset |
34dc7c2f | 1273 | */ |
3558fd73 | 1274 | dmu_objset_disown(os, zsb); |
34dc7c2f BB |
1275 | } |
1276 | ||
3558fd73 | 1277 | zfs_sb_free(zsb); |
34dc7c2f BB |
1278 | return (0); |
1279 | } | |
e5c39b95 | 1280 | EXPORT_SYMBOL(zfs_umount); |
34dc7c2f | 1281 | |
0de19dad BB |
1282 | int |
1283 | zfs_remount(struct super_block *sb, int *flags, char *data) | |
1284 | { | |
0de19dad | 1285 | /* |
2cf7f52b BB |
1286 | * All namespace flags (MNT_*) and super block flags (MS_*) will |
1287 | * be handled by the Linux VFS. Only handle custom options here. | |
0de19dad | 1288 | */ |
0de19dad BB |
1289 | return (0); |
1290 | } | |
1291 | EXPORT_SYMBOL(zfs_remount); | |
1292 | ||
e5c39b95 | 1293 | int |
2cf7f52b | 1294 | zfs_vget(struct super_block *sb, struct inode **ipp, fid_t *fidp) |
34dc7c2f | 1295 | { |
2cf7f52b | 1296 | zfs_sb_t *zsb = sb->s_fs_info; |
34dc7c2f BB |
1297 | znode_t *zp; |
1298 | uint64_t object = 0; | |
1299 | uint64_t fid_gen = 0; | |
1300 | uint64_t gen_mask; | |
1301 | uint64_t zp_gen; | |
3558fd73 | 1302 | int i, err; |
34dc7c2f | 1303 | |
3558fd73 | 1304 | *ipp = NULL; |
34dc7c2f | 1305 | |
3558fd73 | 1306 | ZFS_ENTER(zsb); |
34dc7c2f BB |
1307 | |
1308 | if (fidp->fid_len == LONG_FID_LEN) { | |
1309 | zfid_long_t *zlfid = (zfid_long_t *)fidp; | |
1310 | uint64_t objsetid = 0; | |
1311 | uint64_t setgen = 0; | |
1312 | ||
1313 | for (i = 0; i < sizeof (zlfid->zf_setid); i++) | |
1314 | objsetid |= ((uint64_t)zlfid->zf_setid[i]) << (8 * i); | |
1315 | ||
1316 | for (i = 0; i < sizeof (zlfid->zf_setgen); i++) | |
1317 | setgen |= ((uint64_t)zlfid->zf_setgen[i]) << (8 * i); | |
1318 | ||
3558fd73 | 1319 | ZFS_EXIT(zsb); |
34dc7c2f | 1320 | |
ebe7e575 | 1321 | err = zfsctl_lookup_objset(sb, objsetid, &zsb); |
34dc7c2f BB |
1322 | if (err) |
1323 | return (EINVAL); | |
ebe7e575 | 1324 | |
3558fd73 | 1325 | ZFS_ENTER(zsb); |
34dc7c2f BB |
1326 | } |
1327 | ||
1328 | if (fidp->fid_len == SHORT_FID_LEN || fidp->fid_len == LONG_FID_LEN) { | |
1329 | zfid_short_t *zfid = (zfid_short_t *)fidp; | |
1330 | ||
1331 | for (i = 0; i < sizeof (zfid->zf_object); i++) | |
1332 | object |= ((uint64_t)zfid->zf_object[i]) << (8 * i); | |
1333 | ||
1334 | for (i = 0; i < sizeof (zfid->zf_gen); i++) | |
1335 | fid_gen |= ((uint64_t)zfid->zf_gen[i]) << (8 * i); | |
1336 | } else { | |
3558fd73 | 1337 | ZFS_EXIT(zsb); |
34dc7c2f BB |
1338 | return (EINVAL); |
1339 | } | |
1340 | ||
1341 | /* A zero fid_gen means we are in the .zfs control directories */ | |
1342 | if (fid_gen == 0 && | |
1343 | (object == ZFSCTL_INO_ROOT || object == ZFSCTL_INO_SNAPDIR)) { | |
3558fd73 BB |
1344 | *ipp = zsb->z_ctldir; |
1345 | ASSERT(*ipp != NULL); | |
34dc7c2f | 1346 | if (object == ZFSCTL_INO_SNAPDIR) { |
ebe7e575 BB |
1347 | VERIFY(zfsctl_root_lookup(*ipp, "snapshot", ipp, |
1348 | 0, kcred, NULL, NULL) == 0); | |
34dc7c2f | 1349 | } else { |
3558fd73 | 1350 | igrab(*ipp); |
34dc7c2f | 1351 | } |
3558fd73 | 1352 | ZFS_EXIT(zsb); |
34dc7c2f BB |
1353 | return (0); |
1354 | } | |
1355 | ||
1356 | gen_mask = -1ULL >> (64 - 8 * i); | |
1357 | ||
1358 | dprintf("getting %llu [%u mask %llx]\n", object, fid_gen, gen_mask); | |
3558fd73 BB |
1359 | if ((err = zfs_zget(zsb, object, &zp))) { |
1360 | ZFS_EXIT(zsb); | |
34dc7c2f BB |
1361 | return (err); |
1362 | } | |
3558fd73 | 1363 | (void) sa_lookup(zp->z_sa_hdl, SA_ZPL_GEN(zsb), &zp_gen, |
428870ff BB |
1364 | sizeof (uint64_t)); |
1365 | zp_gen = zp_gen & gen_mask; | |
34dc7c2f BB |
1366 | if (zp_gen == 0) |
1367 | zp_gen = 1; | |
1368 | if (zp->z_unlinked || zp_gen != fid_gen) { | |
1369 | dprintf("znode gen (%u) != fid gen (%u)\n", zp_gen, fid_gen); | |
3558fd73 BB |
1370 | iput(ZTOI(zp)); |
1371 | ZFS_EXIT(zsb); | |
34dc7c2f BB |
1372 | return (EINVAL); |
1373 | } | |
1374 | ||
3558fd73 BB |
1375 | *ipp = ZTOI(zp); |
1376 | if (*ipp) | |
1377 | zfs_inode_update(ITOZ(*ipp)); | |
960e08fe | 1378 | |
3558fd73 | 1379 | ZFS_EXIT(zsb); |
34dc7c2f BB |
1380 | return (0); |
1381 | } | |
e5c39b95 | 1382 | EXPORT_SYMBOL(zfs_vget); |
34dc7c2f BB |
1383 | |
1384 | /* | |
3558fd73 | 1385 | * Block out VOPs and close zfs_sb_t::z_os |
34dc7c2f BB |
1386 | * |
1387 | * Note, if successful, then we return with the 'z_teardown_lock' and | |
1388 | * 'z_teardown_inactive_lock' write held. | |
1389 | */ | |
1390 | int | |
3558fd73 | 1391 | zfs_suspend_fs(zfs_sb_t *zsb) |
34dc7c2f BB |
1392 | { |
1393 | int error; | |
1394 | ||
86f35f34 | 1395 | if ((error = zfs_sb_teardown(zsb, B_FALSE)) != 0) |
34dc7c2f | 1396 | return (error); |
3558fd73 | 1397 | dmu_objset_disown(zsb->z_os, zsb); |
34dc7c2f BB |
1398 | |
1399 | return (0); | |
1400 | } | |
e5c39b95 | 1401 | EXPORT_SYMBOL(zfs_suspend_fs); |
34dc7c2f BB |
1402 | |
1403 | /* | |
3558fd73 | 1404 | * Reopen zfs_sb_t::z_os and release VOPs. |
34dc7c2f BB |
1405 | */ |
1406 | int | |
3558fd73 | 1407 | zfs_resume_fs(zfs_sb_t *zsb, const char *osname) |
34dc7c2f | 1408 | { |
428870ff | 1409 | int err, err2; |
34dc7c2f | 1410 | |
3558fd73 BB |
1411 | ASSERT(RRW_WRITE_HELD(&zsb->z_teardown_lock)); |
1412 | ASSERT(RW_WRITE_HELD(&zsb->z_teardown_inactive_lock)); | |
34dc7c2f | 1413 | |
3558fd73 | 1414 | err = dmu_objset_own(osname, DMU_OST_ZFS, B_FALSE, zsb, &zsb->z_os); |
34dc7c2f | 1415 | if (err) { |
3558fd73 | 1416 | zsb->z_os = NULL; |
34dc7c2f BB |
1417 | } else { |
1418 | znode_t *zp; | |
428870ff BB |
1419 | uint64_t sa_obj = 0; |
1420 | ||
3558fd73 | 1421 | err2 = zap_lookup(zsb->z_os, MASTER_NODE_OBJ, |
428870ff BB |
1422 | ZFS_SA_ATTRS, 8, 1, &sa_obj); |
1423 | ||
3558fd73 | 1424 | if ((err || err2) && zsb->z_version >= ZPL_VERSION_SA) |
428870ff BB |
1425 | goto bail; |
1426 | ||
1427 | ||
3558fd73 BB |
1428 | if ((err = sa_setup(zsb->z_os, sa_obj, |
1429 | zfs_attr_table, ZPL_END, &zsb->z_attr_table)) != 0) | |
572e2857 | 1430 | goto bail; |
34dc7c2f | 1431 | |
3558fd73 | 1432 | VERIFY(zfs_sb_setup(zsb, B_FALSE) == 0); |
34dc7c2f BB |
1433 | |
1434 | /* | |
1435 | * Attempt to re-establish all the active znodes with | |
1436 | * their dbufs. If a zfs_rezget() fails, then we'll let | |
1437 | * any potential callers discover that via ZFS_ENTER_VERIFY_VP | |
1438 | * when they try to use their znode. | |
1439 | */ | |
3558fd73 BB |
1440 | mutex_enter(&zsb->z_znodes_lock); |
1441 | for (zp = list_head(&zsb->z_all_znodes); zp; | |
1442 | zp = list_next(&zsb->z_all_znodes, zp)) { | |
34dc7c2f BB |
1443 | (void) zfs_rezget(zp); |
1444 | } | |
3558fd73 | 1445 | mutex_exit(&zsb->z_znodes_lock); |
34dc7c2f BB |
1446 | |
1447 | } | |
1448 | ||
428870ff | 1449 | bail: |
34dc7c2f | 1450 | /* release the VOPs */ |
3558fd73 BB |
1451 | rw_exit(&zsb->z_teardown_inactive_lock); |
1452 | rrw_exit(&zsb->z_teardown_lock, FTAG); | |
34dc7c2f BB |
1453 | |
1454 | if (err) { | |
1455 | /* | |
3558fd73 | 1456 | * Since we couldn't reopen zfs_sb_t::z_os, force |
34dc7c2f BB |
1457 | * unmount this file system. |
1458 | */ | |
3558fd73 | 1459 | (void) zfs_umount(zsb->z_sb); |
34dc7c2f BB |
1460 | } |
1461 | return (err); | |
1462 | } | |
e5c39b95 | 1463 | EXPORT_SYMBOL(zfs_resume_fs); |
34dc7c2f | 1464 | |
34dc7c2f | 1465 | int |
3558fd73 | 1466 | zfs_set_version(zfs_sb_t *zsb, uint64_t newvers) |
34dc7c2f BB |
1467 | { |
1468 | int error; | |
3558fd73 | 1469 | objset_t *os = zsb->z_os; |
34dc7c2f | 1470 | dmu_tx_t *tx; |
34dc7c2f BB |
1471 | |
1472 | if (newvers < ZPL_VERSION_INITIAL || newvers > ZPL_VERSION) | |
1473 | return (EINVAL); | |
1474 | ||
3558fd73 | 1475 | if (newvers < zsb->z_version) |
9babb374 | 1476 | return (EINVAL); |
34dc7c2f | 1477 | |
428870ff | 1478 | if (zfs_spa_version_map(newvers) > |
3558fd73 | 1479 | spa_version(dmu_objset_spa(zsb->z_os))) |
428870ff BB |
1480 | return (ENOTSUP); |
1481 | ||
34dc7c2f | 1482 | tx = dmu_tx_create(os); |
9babb374 | 1483 | dmu_tx_hold_zap(tx, MASTER_NODE_OBJ, B_FALSE, ZPL_VERSION_STR); |
3558fd73 | 1484 | if (newvers >= ZPL_VERSION_SA && !zsb->z_use_sa) { |
428870ff BB |
1485 | dmu_tx_hold_zap(tx, MASTER_NODE_OBJ, B_TRUE, |
1486 | ZFS_SA_ATTRS); | |
1487 | dmu_tx_hold_zap(tx, DMU_NEW_OBJECT, FALSE, NULL); | |
1488 | } | |
34dc7c2f BB |
1489 | error = dmu_tx_assign(tx, TXG_WAIT); |
1490 | if (error) { | |
1491 | dmu_tx_abort(tx); | |
9babb374 BB |
1492 | return (error); |
1493 | } | |
428870ff | 1494 | |
9babb374 BB |
1495 | error = zap_update(os, MASTER_NODE_OBJ, ZPL_VERSION_STR, |
1496 | 8, 1, &newvers, tx); | |
1497 | ||
1498 | if (error) { | |
1499 | dmu_tx_commit(tx); | |
1500 | return (error); | |
34dc7c2f | 1501 | } |
34dc7c2f | 1502 | |
3558fd73 | 1503 | if (newvers >= ZPL_VERSION_SA && !zsb->z_use_sa) { |
428870ff BB |
1504 | uint64_t sa_obj; |
1505 | ||
3558fd73 | 1506 | ASSERT3U(spa_version(dmu_objset_spa(zsb->z_os)), >=, |
428870ff BB |
1507 | SPA_VERSION_SA); |
1508 | sa_obj = zap_create(os, DMU_OT_SA_MASTER_NODE, | |
1509 | DMU_OT_NONE, 0, tx); | |
1510 | ||
1511 | error = zap_add(os, MASTER_NODE_OBJ, | |
1512 | ZFS_SA_ATTRS, 8, 1, &sa_obj, tx); | |
1513 | ASSERT3U(error, ==, 0); | |
1514 | ||
1515 | VERIFY(0 == sa_set_sa_object(os, sa_obj)); | |
1516 | sa_register_update_callback(os, zfs_sa_upgrade); | |
1517 | } | |
1518 | ||
1519 | spa_history_log_internal(LOG_DS_UPGRADE, | |
1520 | dmu_objset_spa(os), tx, "oldver=%llu newver=%llu dataset = %llu", | |
3558fd73 | 1521 | zsb->z_version, newvers, dmu_objset_id(os)); |
9babb374 | 1522 | |
34dc7c2f BB |
1523 | dmu_tx_commit(tx); |
1524 | ||
3558fd73 | 1525 | zsb->z_version = newvers; |
9babb374 | 1526 | |
3558fd73 BB |
1527 | if (zsb->z_version >= ZPL_VERSION_FUID) |
1528 | zfs_set_fuid_feature(zsb); | |
9babb374 BB |
1529 | |
1530 | return (0); | |
34dc7c2f | 1531 | } |
e5c39b95 | 1532 | EXPORT_SYMBOL(zfs_set_version); |
34dc7c2f BB |
1533 | |
1534 | /* | |
1535 | * Read a property stored within the master node. | |
1536 | */ | |
1537 | int | |
1538 | zfs_get_zplprop(objset_t *os, zfs_prop_t prop, uint64_t *value) | |
1539 | { | |
1540 | const char *pname; | |
b128c09f | 1541 | int error = ENOENT; |
34dc7c2f BB |
1542 | |
1543 | /* | |
1544 | * Look up the file system's value for the property. For the | |
1545 | * version property, we look up a slightly different string. | |
1546 | */ | |
1547 | if (prop == ZFS_PROP_VERSION) | |
1548 | pname = ZPL_VERSION_STR; | |
1549 | else | |
1550 | pname = zfs_prop_to_name(prop); | |
1551 | ||
b128c09f BB |
1552 | if (os != NULL) |
1553 | error = zap_lookup(os, MASTER_NODE_OBJ, pname, 8, 1, value); | |
34dc7c2f BB |
1554 | |
1555 | if (error == ENOENT) { | |
1556 | /* No value set, use the default value */ | |
1557 | switch (prop) { | |
1558 | case ZFS_PROP_VERSION: | |
1559 | *value = ZPL_VERSION; | |
1560 | break; | |
1561 | case ZFS_PROP_NORMALIZE: | |
1562 | case ZFS_PROP_UTF8ONLY: | |
1563 | *value = 0; | |
1564 | break; | |
1565 | case ZFS_PROP_CASE: | |
1566 | *value = ZFS_CASE_SENSITIVE; | |
1567 | break; | |
1568 | default: | |
1569 | return (error); | |
1570 | } | |
1571 | error = 0; | |
1572 | } | |
1573 | return (error); | |
1574 | } | |
86f35f34 | 1575 | EXPORT_SYMBOL(zfs_get_zplprop); |
3558fd73 BB |
1576 | |
1577 | void | |
1578 | zfs_init(void) | |
1579 | { | |
ebe7e575 | 1580 | zfsctl_init(); |
3558fd73 BB |
1581 | zfs_znode_init(); |
1582 | dmu_objset_register_type(DMU_OST_ZFS, zfs_space_delta_cb); | |
1583 | register_filesystem(&zpl_fs_type); | |
ab26409d | 1584 | (void) arc_add_prune_callback(zpl_prune_sbs, NULL); |
3558fd73 BB |
1585 | } |
1586 | ||
1587 | void | |
1588 | zfs_fini(void) | |
1589 | { | |
1590 | unregister_filesystem(&zpl_fs_type); | |
1591 | zfs_znode_fini(); | |
ebe7e575 | 1592 | zfsctl_fini(); |
3558fd73 | 1593 | } |