]>
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 | |
82a37189 | 143 | if (newval == ZFS_XATTR_OFF) { |
2cf7f52b | 144 | zsb->z_flags &= ~ZSB_XATTR; |
82a37189 BB |
145 | } else { |
146 | zsb->z_flags |= ZSB_XATTR; | |
147 | ||
148 | if (newval == ZFS_XATTR_SA) | |
149 | zsb->z_xattr_sa = B_TRUE; | |
150 | else | |
151 | zsb->z_xattr_sa = B_FALSE; | |
152 | } | |
34dc7c2f BB |
153 | } |
154 | ||
155 | static void | |
156 | blksz_changed_cb(void *arg, uint64_t newval) | |
157 | { | |
3558fd73 | 158 | zfs_sb_t *zsb = arg; |
34dc7c2f BB |
159 | |
160 | if (newval < SPA_MINBLOCKSIZE || | |
161 | newval > SPA_MAXBLOCKSIZE || !ISP2(newval)) | |
162 | newval = SPA_MAXBLOCKSIZE; | |
163 | ||
3558fd73 | 164 | zsb->z_max_blksz = newval; |
34dc7c2f BB |
165 | } |
166 | ||
167 | static void | |
168 | readonly_changed_cb(void *arg, uint64_t newval) | |
169 | { | |
3558fd73 BB |
170 | zfs_sb_t *zsb = arg; |
171 | struct super_block *sb = zsb->z_sb; | |
34dc7c2f | 172 | |
2cf7f52b BB |
173 | if (sb == NULL) |
174 | return; | |
175 | ||
176 | if (newval) | |
3558fd73 | 177 | sb->s_flags |= MS_RDONLY; |
2cf7f52b | 178 | else |
3558fd73 | 179 | sb->s_flags &= ~MS_RDONLY; |
34dc7c2f BB |
180 | } |
181 | ||
182 | static void | |
183 | devices_changed_cb(void *arg, uint64_t newval) | |
184 | { | |
34dc7c2f BB |
185 | } |
186 | ||
187 | static void | |
188 | setuid_changed_cb(void *arg, uint64_t newval) | |
189 | { | |
34dc7c2f BB |
190 | } |
191 | ||
192 | static void | |
193 | exec_changed_cb(void *arg, uint64_t newval) | |
194 | { | |
34dc7c2f BB |
195 | } |
196 | ||
34dc7c2f BB |
197 | static void |
198 | nbmand_changed_cb(void *arg, uint64_t newval) | |
199 | { | |
3558fd73 BB |
200 | zfs_sb_t *zsb = arg; |
201 | struct super_block *sb = zsb->z_sb; | |
202 | ||
2cf7f52b BB |
203 | if (sb == NULL) |
204 | return; | |
205 | ||
206 | if (newval == TRUE) | |
3558fd73 | 207 | sb->s_flags |= MS_MANDLOCK; |
2cf7f52b | 208 | else |
3558fd73 | 209 | sb->s_flags &= ~MS_MANDLOCK; |
34dc7c2f BB |
210 | } |
211 | ||
212 | static void | |
213 | snapdir_changed_cb(void *arg, uint64_t newval) | |
214 | { | |
3558fd73 | 215 | ((zfs_sb_t *)arg)->z_show_ctldir = newval; |
34dc7c2f BB |
216 | } |
217 | ||
218 | static void | |
219 | vscan_changed_cb(void *arg, uint64_t newval) | |
220 | { | |
3558fd73 | 221 | ((zfs_sb_t *)arg)->z_vscan = newval; |
34dc7c2f BB |
222 | } |
223 | ||
34dc7c2f BB |
224 | static void |
225 | acl_inherit_changed_cb(void *arg, uint64_t newval) | |
226 | { | |
3558fd73 | 227 | ((zfs_sb_t *)arg)->z_acl_inherit = newval; |
34dc7c2f BB |
228 | } |
229 | ||
e5c39b95 | 230 | int |
3558fd73 | 231 | zfs_register_callbacks(zfs_sb_t *zsb) |
34dc7c2f BB |
232 | { |
233 | struct dsl_dataset *ds = NULL; | |
3558fd73 | 234 | objset_t *os = zsb->z_os; |
34dc7c2f BB |
235 | int error = 0; |
236 | ||
2cf7f52b BB |
237 | if (zfs_is_readonly(zsb) || !spa_writeable(dmu_objset_spa(os))) |
238 | readonly_changed_cb(zsb, B_TRUE); | |
34dc7c2f BB |
239 | |
240 | /* | |
241 | * Register property callbacks. | |
242 | * | |
243 | * It would probably be fine to just check for i/o error from | |
244 | * the first prop_register(), but I guess I like to go | |
245 | * overboard... | |
246 | */ | |
247 | ds = dmu_objset_ds(os); | |
3558fd73 BB |
248 | error = dsl_prop_register(ds, |
249 | "atime", atime_changed_cb, zsb); | |
34dc7c2f | 250 | error = error ? error : dsl_prop_register(ds, |
3558fd73 | 251 | "xattr", xattr_changed_cb, zsb); |
34dc7c2f | 252 | error = error ? error : dsl_prop_register(ds, |
3558fd73 | 253 | "recordsize", blksz_changed_cb, zsb); |
34dc7c2f | 254 | error = error ? error : dsl_prop_register(ds, |
3558fd73 | 255 | "readonly", readonly_changed_cb, zsb); |
34dc7c2f | 256 | error = error ? error : dsl_prop_register(ds, |
3558fd73 | 257 | "devices", devices_changed_cb, zsb); |
34dc7c2f | 258 | error = error ? error : dsl_prop_register(ds, |
3558fd73 | 259 | "setuid", setuid_changed_cb, zsb); |
34dc7c2f | 260 | error = error ? error : dsl_prop_register(ds, |
3558fd73 | 261 | "exec", exec_changed_cb, zsb); |
34dc7c2f | 262 | error = error ? error : dsl_prop_register(ds, |
3558fd73 | 263 | "snapdir", snapdir_changed_cb, zsb); |
34dc7c2f | 264 | error = error ? error : dsl_prop_register(ds, |
3558fd73 | 265 | "aclinherit", acl_inherit_changed_cb, zsb); |
34dc7c2f | 266 | error = error ? error : dsl_prop_register(ds, |
3558fd73 | 267 | "vscan", vscan_changed_cb, zsb); |
2cf7f52b BB |
268 | error = error ? error : dsl_prop_register(ds, |
269 | "nbmand", nbmand_changed_cb, zsb); | |
34dc7c2f BB |
270 | if (error) |
271 | goto unregister; | |
272 | ||
34dc7c2f BB |
273 | return (0); |
274 | ||
275 | unregister: | |
276 | /* | |
277 | * We may attempt to unregister some callbacks that are not | |
278 | * registered, but this is OK; it will simply return ENOMSG, | |
279 | * which we will ignore. | |
280 | */ | |
3558fd73 BB |
281 | (void) dsl_prop_unregister(ds, "atime", atime_changed_cb, zsb); |
282 | (void) dsl_prop_unregister(ds, "xattr", xattr_changed_cb, zsb); | |
283 | (void) dsl_prop_unregister(ds, "recordsize", blksz_changed_cb, zsb); | |
284 | (void) dsl_prop_unregister(ds, "readonly", readonly_changed_cb, zsb); | |
285 | (void) dsl_prop_unregister(ds, "devices", devices_changed_cb, zsb); | |
286 | (void) dsl_prop_unregister(ds, "setuid", setuid_changed_cb, zsb); | |
287 | (void) dsl_prop_unregister(ds, "exec", exec_changed_cb, zsb); | |
288 | (void) dsl_prop_unregister(ds, "snapdir", snapdir_changed_cb, zsb); | |
34dc7c2f | 289 | (void) dsl_prop_unregister(ds, "aclinherit", acl_inherit_changed_cb, |
3558fd73 BB |
290 | zsb); |
291 | (void) dsl_prop_unregister(ds, "vscan", vscan_changed_cb, zsb); | |
2cf7f52b | 292 | (void) dsl_prop_unregister(ds, "nbmand", nbmand_changed_cb, zsb); |
34dc7c2f | 293 | |
3558fd73 | 294 | return (error); |
34dc7c2f | 295 | } |
e5c39b95 | 296 | EXPORT_SYMBOL(zfs_register_callbacks); |
34dc7c2f | 297 | |
428870ff BB |
298 | static int |
299 | zfs_space_delta_cb(dmu_object_type_t bonustype, void *data, | |
300 | uint64_t *userp, uint64_t *groupp) | |
9babb374 | 301 | { |
428870ff BB |
302 | znode_phys_t *znp = data; |
303 | int error = 0; | |
9babb374 | 304 | |
428870ff BB |
305 | /* |
306 | * Is it a valid type of object to track? | |
307 | */ | |
308 | if (bonustype != DMU_OT_ZNODE && bonustype != DMU_OT_SA) | |
309 | return (ENOENT); | |
9babb374 | 310 | |
428870ff BB |
311 | /* |
312 | * If we have a NULL data pointer | |
313 | * then assume the id's aren't changing and | |
314 | * return EEXIST to the dmu to let it know to | |
315 | * use the same ids | |
316 | */ | |
317 | if (data == NULL) | |
318 | return (EEXIST); | |
9babb374 | 319 | |
428870ff BB |
320 | if (bonustype == DMU_OT_ZNODE) { |
321 | *userp = znp->zp_uid; | |
322 | *groupp = znp->zp_gid; | |
9babb374 | 323 | } else { |
428870ff | 324 | int hdrsize; |
9babb374 | 325 | |
428870ff BB |
326 | ASSERT(bonustype == DMU_OT_SA); |
327 | hdrsize = sa_hdrsize(data); | |
328 | ||
329 | if (hdrsize != 0) { | |
330 | *userp = *((uint64_t *)((uintptr_t)data + hdrsize + | |
331 | SA_UID_OFFSET)); | |
332 | *groupp = *((uint64_t *)((uintptr_t)data + hdrsize + | |
333 | SA_GID_OFFSET)); | |
334 | } else { | |
335 | /* | |
336 | * This should only happen for newly created | |
337 | * files that haven't had the znode data filled | |
338 | * in yet. | |
339 | */ | |
340 | *userp = 0; | |
341 | *groupp = 0; | |
342 | } | |
9babb374 | 343 | } |
428870ff | 344 | return (error); |
9babb374 BB |
345 | } |
346 | ||
347 | static void | |
3558fd73 | 348 | fuidstr_to_sid(zfs_sb_t *zsb, const char *fuidstr, |
9babb374 BB |
349 | char *domainbuf, int buflen, uid_t *ridp) |
350 | { | |
9babb374 BB |
351 | uint64_t fuid; |
352 | const char *domain; | |
353 | ||
354 | fuid = strtonum(fuidstr, NULL); | |
355 | ||
3558fd73 | 356 | domain = zfs_fuid_find_by_idx(zsb, FUID_INDEX(fuid)); |
9babb374 BB |
357 | if (domain) |
358 | (void) strlcpy(domainbuf, domain, buflen); | |
359 | else | |
360 | domainbuf[0] = '\0'; | |
361 | *ridp = FUID_RID(fuid); | |
362 | } | |
363 | ||
364 | static uint64_t | |
3558fd73 | 365 | zfs_userquota_prop_to_obj(zfs_sb_t *zsb, zfs_userquota_prop_t type) |
9babb374 BB |
366 | { |
367 | switch (type) { | |
368 | case ZFS_PROP_USERUSED: | |
369 | return (DMU_USERUSED_OBJECT); | |
370 | case ZFS_PROP_GROUPUSED: | |
371 | return (DMU_GROUPUSED_OBJECT); | |
372 | case ZFS_PROP_USERQUOTA: | |
3558fd73 | 373 | return (zsb->z_userquota_obj); |
9babb374 | 374 | case ZFS_PROP_GROUPQUOTA: |
3558fd73 | 375 | return (zsb->z_groupquota_obj); |
149e873a BB |
376 | default: |
377 | return (ENOTSUP); | |
9babb374 BB |
378 | } |
379 | return (0); | |
380 | } | |
381 | ||
382 | int | |
3558fd73 | 383 | zfs_userspace_many(zfs_sb_t *zsb, zfs_userquota_prop_t type, |
9babb374 BB |
384 | uint64_t *cookiep, void *vbuf, uint64_t *bufsizep) |
385 | { | |
386 | int error; | |
387 | zap_cursor_t zc; | |
388 | zap_attribute_t za; | |
389 | zfs_useracct_t *buf = vbuf; | |
390 | uint64_t obj; | |
391 | ||
3558fd73 | 392 | if (!dmu_objset_userspace_present(zsb->z_os)) |
9babb374 BB |
393 | return (ENOTSUP); |
394 | ||
3558fd73 | 395 | obj = zfs_userquota_prop_to_obj(zsb, type); |
9babb374 BB |
396 | if (obj == 0) { |
397 | *bufsizep = 0; | |
398 | return (0); | |
399 | } | |
400 | ||
3558fd73 | 401 | for (zap_cursor_init_serialized(&zc, zsb->z_os, obj, *cookiep); |
9babb374 BB |
402 | (error = zap_cursor_retrieve(&zc, &za)) == 0; |
403 | zap_cursor_advance(&zc)) { | |
404 | if ((uintptr_t)buf - (uintptr_t)vbuf + sizeof (zfs_useracct_t) > | |
405 | *bufsizep) | |
406 | break; | |
407 | ||
3558fd73 | 408 | fuidstr_to_sid(zsb, za.za_name, |
9babb374 BB |
409 | buf->zu_domain, sizeof (buf->zu_domain), &buf->zu_rid); |
410 | ||
411 | buf->zu_space = za.za_first_integer; | |
412 | buf++; | |
413 | } | |
414 | if (error == ENOENT) | |
415 | error = 0; | |
416 | ||
417 | ASSERT3U((uintptr_t)buf - (uintptr_t)vbuf, <=, *bufsizep); | |
418 | *bufsizep = (uintptr_t)buf - (uintptr_t)vbuf; | |
419 | *cookiep = zap_cursor_serialize(&zc); | |
420 | zap_cursor_fini(&zc); | |
421 | return (error); | |
422 | } | |
e5c39b95 | 423 | EXPORT_SYMBOL(zfs_userspace_many); |
9babb374 BB |
424 | |
425 | /* | |
426 | * buf must be big enough (eg, 32 bytes) | |
427 | */ | |
428 | static int | |
3558fd73 | 429 | id_to_fuidstr(zfs_sb_t *zsb, const char *domain, uid_t rid, |
9babb374 BB |
430 | char *buf, boolean_t addok) |
431 | { | |
432 | uint64_t fuid; | |
433 | int domainid = 0; | |
434 | ||
435 | if (domain && domain[0]) { | |
3558fd73 | 436 | domainid = zfs_fuid_find_by_domain(zsb, domain, NULL, addok); |
9babb374 BB |
437 | if (domainid == -1) |
438 | return (ENOENT); | |
439 | } | |
440 | fuid = FUID_ENCODE(domainid, rid); | |
441 | (void) sprintf(buf, "%llx", (longlong_t)fuid); | |
442 | return (0); | |
443 | } | |
444 | ||
445 | int | |
3558fd73 | 446 | zfs_userspace_one(zfs_sb_t *zsb, zfs_userquota_prop_t type, |
9babb374 BB |
447 | const char *domain, uint64_t rid, uint64_t *valp) |
448 | { | |
449 | char buf[32]; | |
450 | int err; | |
451 | uint64_t obj; | |
452 | ||
453 | *valp = 0; | |
454 | ||
3558fd73 | 455 | if (!dmu_objset_userspace_present(zsb->z_os)) |
9babb374 BB |
456 | return (ENOTSUP); |
457 | ||
3558fd73 | 458 | obj = zfs_userquota_prop_to_obj(zsb, type); |
9babb374 BB |
459 | if (obj == 0) |
460 | return (0); | |
461 | ||
3558fd73 | 462 | err = id_to_fuidstr(zsb, domain, rid, buf, B_FALSE); |
9babb374 BB |
463 | if (err) |
464 | return (err); | |
465 | ||
3558fd73 | 466 | err = zap_lookup(zsb->z_os, obj, buf, 8, 1, valp); |
9babb374 BB |
467 | if (err == ENOENT) |
468 | err = 0; | |
469 | return (err); | |
470 | } | |
e5c39b95 | 471 | EXPORT_SYMBOL(zfs_userspace_one); |
9babb374 BB |
472 | |
473 | int | |
3558fd73 | 474 | zfs_set_userquota(zfs_sb_t *zsb, zfs_userquota_prop_t type, |
9babb374 BB |
475 | const char *domain, uint64_t rid, uint64_t quota) |
476 | { | |
477 | char buf[32]; | |
478 | int err; | |
479 | dmu_tx_t *tx; | |
480 | uint64_t *objp; | |
481 | boolean_t fuid_dirtied; | |
482 | ||
483 | if (type != ZFS_PROP_USERQUOTA && type != ZFS_PROP_GROUPQUOTA) | |
484 | return (EINVAL); | |
485 | ||
3558fd73 | 486 | if (zsb->z_version < ZPL_VERSION_USERSPACE) |
9babb374 BB |
487 | return (ENOTSUP); |
488 | ||
3558fd73 BB |
489 | objp = (type == ZFS_PROP_USERQUOTA) ? &zsb->z_userquota_obj : |
490 | &zsb->z_groupquota_obj; | |
9babb374 | 491 | |
3558fd73 | 492 | err = id_to_fuidstr(zsb, domain, rid, buf, B_TRUE); |
9babb374 BB |
493 | if (err) |
494 | return (err); | |
3558fd73 | 495 | fuid_dirtied = zsb->z_fuid_dirty; |
9babb374 | 496 | |
3558fd73 | 497 | tx = dmu_tx_create(zsb->z_os); |
9babb374 BB |
498 | dmu_tx_hold_zap(tx, *objp ? *objp : DMU_NEW_OBJECT, B_TRUE, NULL); |
499 | if (*objp == 0) { | |
500 | dmu_tx_hold_zap(tx, MASTER_NODE_OBJ, B_TRUE, | |
501 | zfs_userquota_prop_prefixes[type]); | |
502 | } | |
503 | if (fuid_dirtied) | |
3558fd73 | 504 | zfs_fuid_txhold(zsb, tx); |
9babb374 BB |
505 | err = dmu_tx_assign(tx, TXG_WAIT); |
506 | if (err) { | |
507 | dmu_tx_abort(tx); | |
508 | return (err); | |
509 | } | |
510 | ||
3558fd73 | 511 | mutex_enter(&zsb->z_lock); |
9babb374 | 512 | if (*objp == 0) { |
3558fd73 | 513 | *objp = zap_create(zsb->z_os, DMU_OT_USERGROUP_QUOTA, |
9babb374 | 514 | DMU_OT_NONE, 0, tx); |
3558fd73 | 515 | VERIFY(0 == zap_add(zsb->z_os, MASTER_NODE_OBJ, |
9babb374 BB |
516 | zfs_userquota_prop_prefixes[type], 8, 1, objp, tx)); |
517 | } | |
3558fd73 | 518 | mutex_exit(&zsb->z_lock); |
9babb374 BB |
519 | |
520 | if (quota == 0) { | |
3558fd73 | 521 | err = zap_remove(zsb->z_os, *objp, buf, tx); |
9babb374 BB |
522 | if (err == ENOENT) |
523 | err = 0; | |
524 | } else { | |
3558fd73 | 525 | err = zap_update(zsb->z_os, *objp, buf, 8, 1, "a, tx); |
9babb374 BB |
526 | } |
527 | ASSERT(err == 0); | |
528 | if (fuid_dirtied) | |
3558fd73 | 529 | zfs_fuid_sync(zsb, tx); |
9babb374 BB |
530 | dmu_tx_commit(tx); |
531 | return (err); | |
532 | } | |
e5c39b95 | 533 | EXPORT_SYMBOL(zfs_set_userquota); |
9babb374 BB |
534 | |
535 | boolean_t | |
3558fd73 | 536 | zfs_fuid_overquota(zfs_sb_t *zsb, boolean_t isgroup, uint64_t fuid) |
9babb374 BB |
537 | { |
538 | char buf[32]; | |
539 | uint64_t used, quota, usedobj, quotaobj; | |
540 | int err; | |
541 | ||
542 | usedobj = isgroup ? DMU_GROUPUSED_OBJECT : DMU_USERUSED_OBJECT; | |
3558fd73 | 543 | quotaobj = isgroup ? zsb->z_groupquota_obj : zsb->z_userquota_obj; |
9babb374 | 544 | |
3558fd73 | 545 | if (quotaobj == 0 || zsb->z_replay) |
9babb374 BB |
546 | return (B_FALSE); |
547 | ||
548 | (void) sprintf(buf, "%llx", (longlong_t)fuid); | |
3558fd73 | 549 | err = zap_lookup(zsb->z_os, quotaobj, buf, 8, 1, "a); |
9babb374 BB |
550 | if (err != 0) |
551 | return (B_FALSE); | |
552 | ||
3558fd73 | 553 | err = zap_lookup(zsb->z_os, usedobj, buf, 8, 1, &used); |
9babb374 BB |
554 | if (err != 0) |
555 | return (B_FALSE); | |
556 | return (used >= quota); | |
557 | } | |
e5c39b95 | 558 | EXPORT_SYMBOL(zfs_fuid_overquota); |
9babb374 | 559 | |
428870ff | 560 | boolean_t |
3558fd73 | 561 | zfs_owner_overquota(zfs_sb_t *zsb, znode_t *zp, boolean_t isgroup) |
428870ff BB |
562 | { |
563 | uint64_t fuid; | |
564 | uint64_t quotaobj; | |
428870ff | 565 | |
3558fd73 | 566 | quotaobj = isgroup ? zsb->z_groupquota_obj : zsb->z_userquota_obj; |
428870ff | 567 | |
572e2857 | 568 | fuid = isgroup ? zp->z_gid : zp->z_uid; |
428870ff | 569 | |
3558fd73 | 570 | if (quotaobj == 0 || zsb->z_replay) |
428870ff BB |
571 | return (B_FALSE); |
572 | ||
3558fd73 | 573 | return (zfs_fuid_overquota(zsb, isgroup, fuid)); |
428870ff | 574 | } |
e5c39b95 | 575 | EXPORT_SYMBOL(zfs_owner_overquota); |
428870ff | 576 | |
9babb374 | 577 | int |
3558fd73 | 578 | zfs_sb_create(const char *osname, zfs_sb_t **zsbp) |
9babb374 BB |
579 | { |
580 | objset_t *os; | |
3558fd73 | 581 | zfs_sb_t *zsb; |
9babb374 BB |
582 | uint64_t zval; |
583 | int i, error; | |
428870ff | 584 | uint64_t sa_obj; |
9babb374 | 585 | |
3558fd73 | 586 | zsb = kmem_zalloc(sizeof (zfs_sb_t), KM_SLEEP); |
9babb374 | 587 | |
428870ff BB |
588 | /* |
589 | * We claim to always be readonly so we can open snapshots; | |
590 | * other ZPL code will prevent us from writing to snapshots. | |
591 | */ | |
3558fd73 | 592 | error = dmu_objset_own(osname, DMU_OST_ZFS, B_TRUE, zsb, &os); |
428870ff | 593 | if (error) { |
3558fd73 | 594 | kmem_free(zsb, sizeof (zfs_sb_t)); |
9babb374 | 595 | return (error); |
428870ff | 596 | } |
9babb374 BB |
597 | |
598 | /* | |
599 | * Initialize the zfs-specific filesystem structure. | |
600 | * Should probably make this a kmem cache, shuffle fields, | |
601 | * and just bzero up to z_hold_mtx[]. | |
602 | */ | |
2cf7f52b | 603 | zsb->z_sb = NULL; |
3558fd73 BB |
604 | zsb->z_parent = zsb; |
605 | zsb->z_max_blksz = SPA_MAXBLOCKSIZE; | |
606 | zsb->z_show_ctldir = ZFS_SNAPDIR_VISIBLE; | |
607 | zsb->z_os = os; | |
9babb374 | 608 | |
3558fd73 | 609 | error = zfs_get_zplprop(os, ZFS_PROP_VERSION, &zsb->z_version); |
9babb374 BB |
610 | if (error) { |
611 | goto out; | |
3558fd73 | 612 | } else if (zsb->z_version > |
428870ff | 613 | zfs_zpl_version_map(spa_version(dmu_objset_spa(os)))) { |
149e873a | 614 | (void) printk("Can't mount a version %lld file system " |
428870ff | 615 | "on a version %lld pool\n. Pool must be upgraded to mount " |
3558fd73 | 616 | "this file system.", (u_longlong_t)zsb->z_version, |
428870ff | 617 | (u_longlong_t)spa_version(dmu_objset_spa(os))); |
9babb374 BB |
618 | error = ENOTSUP; |
619 | goto out; | |
620 | } | |
9babb374 BB |
621 | if ((error = zfs_get_zplprop(os, ZFS_PROP_NORMALIZE, &zval)) != 0) |
622 | goto out; | |
3558fd73 | 623 | zsb->z_norm = (int)zval; |
9babb374 BB |
624 | |
625 | if ((error = zfs_get_zplprop(os, ZFS_PROP_UTF8ONLY, &zval)) != 0) | |
626 | goto out; | |
3558fd73 | 627 | zsb->z_utf8 = (zval != 0); |
9babb374 BB |
628 | |
629 | if ((error = zfs_get_zplprop(os, ZFS_PROP_CASE, &zval)) != 0) | |
630 | goto out; | |
3558fd73 | 631 | zsb->z_case = (uint_t)zval; |
9babb374 BB |
632 | |
633 | /* | |
634 | * Fold case on file systems that are always or sometimes case | |
635 | * insensitive. | |
636 | */ | |
3558fd73 BB |
637 | if (zsb->z_case == ZFS_CASE_INSENSITIVE || |
638 | zsb->z_case == ZFS_CASE_MIXED) | |
639 | zsb->z_norm |= U8_TEXTPREP_TOUPPER; | |
9babb374 | 640 | |
3558fd73 BB |
641 | zsb->z_use_fuids = USE_FUIDS(zsb->z_version, zsb->z_os); |
642 | zsb->z_use_sa = USE_SA(zsb->z_version, zsb->z_os); | |
428870ff | 643 | |
3558fd73 | 644 | if (zsb->z_use_sa) { |
428870ff BB |
645 | /* should either have both of these objects or none */ |
646 | error = zap_lookup(os, MASTER_NODE_OBJ, ZFS_SA_ATTRS, 8, 1, | |
647 | &sa_obj); | |
648 | if (error) | |
591fb62f | 649 | goto out; |
82a37189 BB |
650 | |
651 | error = zfs_get_zplprop(os, ZFS_PROP_XATTR, &zval); | |
652 | if ((error == 0) && (zval == ZFS_XATTR_SA)) | |
653 | zsb->z_xattr_sa = B_TRUE; | |
428870ff BB |
654 | } else { |
655 | /* | |
656 | * Pre SA versions file systems should never touch | |
657 | * either the attribute registration or layout objects. | |
658 | */ | |
659 | sa_obj = 0; | |
660 | } | |
661 | ||
572e2857 | 662 | error = sa_setup(os, sa_obj, zfs_attr_table, ZPL_END, |
3558fd73 | 663 | &zsb->z_attr_table); |
572e2857 BB |
664 | if (error) |
665 | goto out; | |
428870ff | 666 | |
3558fd73 | 667 | if (zsb->z_version >= ZPL_VERSION_SA) |
428870ff | 668 | sa_register_update_callback(os, zfs_sa_upgrade); |
9babb374 BB |
669 | |
670 | error = zap_lookup(os, MASTER_NODE_OBJ, ZFS_ROOT_OBJ, 8, 1, | |
3558fd73 | 671 | &zsb->z_root); |
9babb374 BB |
672 | if (error) |
673 | goto out; | |
3558fd73 | 674 | ASSERT(zsb->z_root != 0); |
9babb374 BB |
675 | |
676 | error = zap_lookup(os, MASTER_NODE_OBJ, ZFS_UNLINKED_SET, 8, 1, | |
3558fd73 | 677 | &zsb->z_unlinkedobj); |
9babb374 BB |
678 | if (error) |
679 | goto out; | |
680 | ||
681 | error = zap_lookup(os, MASTER_NODE_OBJ, | |
682 | zfs_userquota_prop_prefixes[ZFS_PROP_USERQUOTA], | |
3558fd73 | 683 | 8, 1, &zsb->z_userquota_obj); |
9babb374 BB |
684 | if (error && error != ENOENT) |
685 | goto out; | |
686 | ||
687 | error = zap_lookup(os, MASTER_NODE_OBJ, | |
688 | zfs_userquota_prop_prefixes[ZFS_PROP_GROUPQUOTA], | |
3558fd73 | 689 | 8, 1, &zsb->z_groupquota_obj); |
9babb374 BB |
690 | if (error && error != ENOENT) |
691 | goto out; | |
692 | ||
693 | error = zap_lookup(os, MASTER_NODE_OBJ, ZFS_FUID_TABLES, 8, 1, | |
3558fd73 | 694 | &zsb->z_fuid_obj); |
9babb374 BB |
695 | if (error && error != ENOENT) |
696 | goto out; | |
697 | ||
698 | error = zap_lookup(os, MASTER_NODE_OBJ, ZFS_SHARES_DIR, 8, 1, | |
3558fd73 | 699 | &zsb->z_shares_dir); |
9babb374 BB |
700 | if (error && error != ENOENT) |
701 | goto out; | |
702 | ||
3558fd73 BB |
703 | mutex_init(&zsb->z_znodes_lock, NULL, MUTEX_DEFAULT, NULL); |
704 | mutex_init(&zsb->z_lock, NULL, MUTEX_DEFAULT, NULL); | |
705 | list_create(&zsb->z_all_znodes, sizeof (znode_t), | |
9babb374 | 706 | offsetof(znode_t, z_link_node)); |
3558fd73 BB |
707 | rrw_init(&zsb->z_teardown_lock); |
708 | rw_init(&zsb->z_teardown_inactive_lock, NULL, RW_DEFAULT, NULL); | |
709 | rw_init(&zsb->z_fuid_lock, NULL, RW_DEFAULT, NULL); | |
9babb374 | 710 | for (i = 0; i != ZFS_OBJ_MTX_SZ; i++) |
3558fd73 | 711 | mutex_init(&zsb->z_hold_mtx[i], NULL, MUTEX_DEFAULT, NULL); |
9babb374 | 712 | |
3558fd73 | 713 | *zsbp = zsb; |
9babb374 BB |
714 | return (0); |
715 | ||
716 | out: | |
3558fd73 BB |
717 | dmu_objset_disown(os, zsb); |
718 | *zsbp = NULL; | |
719 | kmem_free(zsb, sizeof (zfs_sb_t)); | |
9babb374 BB |
720 | return (error); |
721 | } | |
86f35f34 | 722 | EXPORT_SYMBOL(zfs_sb_create); |
9babb374 | 723 | |
86f35f34 | 724 | int |
3558fd73 | 725 | zfs_sb_setup(zfs_sb_t *zsb, boolean_t mounting) |
34dc7c2f | 726 | { |
34dc7c2f BB |
727 | int error; |
728 | ||
3558fd73 | 729 | error = zfs_register_callbacks(zsb); |
34dc7c2f BB |
730 | if (error) |
731 | return (error); | |
732 | ||
733 | /* | |
3558fd73 | 734 | * Set the objset user_ptr to track its zsb. |
34dc7c2f | 735 | */ |
3558fd73 BB |
736 | mutex_enter(&zsb->z_os->os_user_ptr_lock); |
737 | dmu_objset_set_user(zsb->z_os, zsb); | |
738 | mutex_exit(&zsb->z_os->os_user_ptr_lock); | |
34dc7c2f | 739 | |
3558fd73 | 740 | zsb->z_log = zil_open(zsb->z_os, zfs_get_data); |
9babb374 | 741 | |
34dc7c2f BB |
742 | /* |
743 | * If we are not mounting (ie: online recv), then we don't | |
744 | * have to worry about replaying the log as we blocked all | |
745 | * operations out since we closed the ZIL. | |
746 | */ | |
747 | if (mounting) { | |
b128c09f BB |
748 | boolean_t readonly; |
749 | ||
34dc7c2f BB |
750 | /* |
751 | * During replay we remove the read only flag to | |
752 | * allow replays to succeed. | |
753 | */ | |
2cf7f52b | 754 | readonly = zfs_is_readonly(zsb); |
fb5f0bc8 | 755 | if (readonly != 0) |
2cf7f52b | 756 | readonly_changed_cb(zsb, B_FALSE); |
fb5f0bc8 | 757 | else |
3558fd73 | 758 | zfs_unlinked_drain(zsb); |
34dc7c2f | 759 | |
428870ff BB |
760 | /* |
761 | * Parse and replay the intent log. | |
762 | * | |
763 | * Because of ziltest, this must be done after | |
764 | * zfs_unlinked_drain(). (Further note: ziltest | |
765 | * doesn't use readonly mounts, where | |
766 | * zfs_unlinked_drain() isn't called.) This is because | |
767 | * ziltest causes spa_sync() to think it's committed, | |
768 | * but actually it is not, so the intent log contains | |
769 | * many txg's worth of changes. | |
770 | * | |
771 | * In particular, if object N is in the unlinked set in | |
772 | * the last txg to actually sync, then it could be | |
773 | * actually freed in a later txg and then reallocated | |
774 | * in a yet later txg. This would write a "create | |
775 | * object N" record to the intent log. Normally, this | |
776 | * would be fine because the spa_sync() would have | |
777 | * written out the fact that object N is free, before | |
778 | * we could write the "create object N" intent log | |
779 | * record. | |
780 | * | |
781 | * But when we are in ziltest mode, we advance the "open | |
782 | * txg" without actually spa_sync()-ing the changes to | |
783 | * disk. So we would see that object N is still | |
784 | * allocated and in the unlinked set, and there is an | |
785 | * intent log record saying to allocate it. | |
786 | */ | |
3558fd73 | 787 | if (spa_writeable(dmu_objset_spa(zsb->z_os))) { |
572e2857 | 788 | if (zil_replay_disable) { |
3558fd73 | 789 | zil_destroy(zsb->z_log, B_FALSE); |
572e2857 | 790 | } else { |
3558fd73 BB |
791 | zsb->z_replay = B_TRUE; |
792 | zil_replay(zsb->z_os, zsb, | |
572e2857 | 793 | zfs_replay_vector); |
3558fd73 | 794 | zsb->z_replay = B_FALSE; |
572e2857 | 795 | } |
fb5f0bc8 | 796 | } |
2cf7f52b BB |
797 | |
798 | /* restore readonly bit */ | |
799 | if (readonly != 0) | |
800 | readonly_changed_cb(zsb, B_TRUE); | |
34dc7c2f BB |
801 | } |
802 | ||
34dc7c2f BB |
803 | return (0); |
804 | } | |
86f35f34 | 805 | EXPORT_SYMBOL(zfs_sb_setup); |
34dc7c2f | 806 | |
9babb374 | 807 | void |
3558fd73 | 808 | zfs_sb_free(zfs_sb_t *zsb) |
34dc7c2f | 809 | { |
9babb374 | 810 | int i; |
9babb374 | 811 | |
3558fd73 | 812 | zfs_fuid_destroy(zsb); |
9babb374 | 813 | |
3558fd73 BB |
814 | mutex_destroy(&zsb->z_znodes_lock); |
815 | mutex_destroy(&zsb->z_lock); | |
816 | list_destroy(&zsb->z_all_znodes); | |
817 | rrw_destroy(&zsb->z_teardown_lock); | |
818 | rw_destroy(&zsb->z_teardown_inactive_lock); | |
819 | rw_destroy(&zsb->z_fuid_lock); | |
9babb374 | 820 | for (i = 0; i != ZFS_OBJ_MTX_SZ; i++) |
3558fd73 BB |
821 | mutex_destroy(&zsb->z_hold_mtx[i]); |
822 | kmem_free(zsb, sizeof (zfs_sb_t)); | |
34dc7c2f | 823 | } |
86f35f34 | 824 | EXPORT_SYMBOL(zfs_sb_free); |
34dc7c2f | 825 | |
9babb374 | 826 | static void |
3558fd73 | 827 | zfs_set_fuid_feature(zfs_sb_t *zsb) |
9babb374 | 828 | { |
3558fd73 BB |
829 | zsb->z_use_fuids = USE_FUIDS(zsb->z_version, zsb->z_os); |
830 | zsb->z_use_sa = USE_SA(zsb->z_version, zsb->z_os); | |
9babb374 | 831 | } |
34dc7c2f BB |
832 | |
833 | void | |
3558fd73 | 834 | zfs_unregister_callbacks(zfs_sb_t *zsb) |
34dc7c2f | 835 | { |
3558fd73 | 836 | objset_t *os = zsb->z_os; |
34dc7c2f BB |
837 | struct dsl_dataset *ds; |
838 | ||
839 | /* | |
840 | * Unregister properties. | |
841 | */ | |
842 | if (!dmu_objset_is_snapshot(os)) { | |
843 | ds = dmu_objset_ds(os); | |
844 | VERIFY(dsl_prop_unregister(ds, "atime", atime_changed_cb, | |
3558fd73 | 845 | zsb) == 0); |
34dc7c2f BB |
846 | |
847 | VERIFY(dsl_prop_unregister(ds, "xattr", xattr_changed_cb, | |
3558fd73 | 848 | zsb) == 0); |
34dc7c2f BB |
849 | |
850 | VERIFY(dsl_prop_unregister(ds, "recordsize", blksz_changed_cb, | |
3558fd73 | 851 | zsb) == 0); |
34dc7c2f BB |
852 | |
853 | VERIFY(dsl_prop_unregister(ds, "readonly", readonly_changed_cb, | |
3558fd73 | 854 | zsb) == 0); |
34dc7c2f BB |
855 | |
856 | VERIFY(dsl_prop_unregister(ds, "devices", devices_changed_cb, | |
3558fd73 | 857 | zsb) == 0); |
34dc7c2f BB |
858 | |
859 | VERIFY(dsl_prop_unregister(ds, "setuid", setuid_changed_cb, | |
3558fd73 | 860 | zsb) == 0); |
34dc7c2f BB |
861 | |
862 | VERIFY(dsl_prop_unregister(ds, "exec", exec_changed_cb, | |
3558fd73 | 863 | zsb) == 0); |
34dc7c2f BB |
864 | |
865 | VERIFY(dsl_prop_unregister(ds, "snapdir", snapdir_changed_cb, | |
3558fd73 | 866 | zsb) == 0); |
34dc7c2f | 867 | |
34dc7c2f | 868 | VERIFY(dsl_prop_unregister(ds, "aclinherit", |
3558fd73 | 869 | acl_inherit_changed_cb, zsb) == 0); |
34dc7c2f BB |
870 | |
871 | VERIFY(dsl_prop_unregister(ds, "vscan", | |
3558fd73 | 872 | vscan_changed_cb, zsb) == 0); |
2cf7f52b BB |
873 | |
874 | VERIFY(dsl_prop_unregister(ds, "nbmand", | |
875 | nbmand_changed_cb, zsb) == 0); | |
34dc7c2f BB |
876 | } |
877 | } | |
e5c39b95 | 878 | EXPORT_SYMBOL(zfs_unregister_callbacks); |
34dc7c2f | 879 | |
bc3e15e3 | 880 | #ifdef HAVE_MLSLABEL |
428870ff BB |
881 | /* |
882 | * zfs_check_global_label: | |
883 | * Check that the hex label string is appropriate for the dataset | |
884 | * being mounted into the global_zone proper. | |
885 | * | |
886 | * Return an error if the hex label string is not default or | |
887 | * admin_low/admin_high. For admin_low labels, the corresponding | |
888 | * dataset must be readonly. | |
889 | */ | |
890 | int | |
891 | zfs_check_global_label(const char *dsname, const char *hexsl) | |
892 | { | |
893 | if (strcasecmp(hexsl, ZFS_MLSLABEL_DEFAULT) == 0) | |
894 | return (0); | |
895 | if (strcasecmp(hexsl, ADMIN_HIGH) == 0) | |
896 | return (0); | |
897 | if (strcasecmp(hexsl, ADMIN_LOW) == 0) { | |
898 | /* must be readonly */ | |
899 | uint64_t rdonly; | |
900 | ||
901 | if (dsl_prop_get_integer(dsname, | |
902 | zfs_prop_to_name(ZFS_PROP_READONLY), &rdonly, NULL)) | |
903 | return (EACCES); | |
904 | return (rdonly ? 0 : EACCES); | |
905 | } | |
906 | return (EACCES); | |
907 | } | |
86f35f34 | 908 | EXPORT_SYMBOL(zfs_check_global_label); |
bc3e15e3 | 909 | #endif /* HAVE_MLSLABEL */ |
428870ff | 910 | |
e5c39b95 | 911 | int |
3558fd73 | 912 | zfs_statvfs(struct dentry *dentry, struct kstatfs *statp) |
34dc7c2f | 913 | { |
3558fd73 | 914 | zfs_sb_t *zsb = dentry->d_sb->s_fs_info; |
34dc7c2f | 915 | uint64_t refdbytes, availbytes, usedobjs, availobjs; |
3558fd73 | 916 | uint32_t bshift; |
34dc7c2f | 917 | |
3558fd73 | 918 | ZFS_ENTER(zsb); |
34dc7c2f | 919 | |
3558fd73 | 920 | dmu_objset_space(zsb->z_os, |
34dc7c2f BB |
921 | &refdbytes, &availbytes, &usedobjs, &availobjs); |
922 | ||
923 | /* | |
05ff35c6 BB |
924 | * The underlying storage pool actually uses multiple block |
925 | * size. Under Solaris frsize (fragment size) is reported as | |
926 | * the smallest block size we support, and bsize (block size) | |
927 | * as the filesystem's maximum block size. Unfortunately, | |
928 | * under Linux the fragment size and block size are often used | |
929 | * interchangeably. Thus we are forced to report both of them | |
930 | * as the filesystem's maximum block size. | |
34dc7c2f | 931 | */ |
05ff35c6 | 932 | statp->f_frsize = zsb->z_max_blksz; |
3558fd73 BB |
933 | statp->f_bsize = zsb->z_max_blksz; |
934 | bshift = fls(statp->f_bsize) - 1; | |
34dc7c2f BB |
935 | |
936 | /* | |
3558fd73 BB |
937 | * The following report "total" blocks of various kinds in |
938 | * the file system, but reported in terms of f_bsize - the | |
939 | * "preferred" size. | |
34dc7c2f BB |
940 | */ |
941 | ||
3558fd73 BB |
942 | statp->f_blocks = (refdbytes + availbytes) >> bshift; |
943 | statp->f_bfree = availbytes >> bshift; | |
34dc7c2f BB |
944 | statp->f_bavail = statp->f_bfree; /* no root reservation */ |
945 | ||
946 | /* | |
947 | * statvfs() should really be called statufs(), because it assumes | |
948 | * static metadata. ZFS doesn't preallocate files, so the best | |
949 | * we can do is report the max that could possibly fit in f_files, | |
950 | * and that minus the number actually used in f_ffree. | |
951 | * For f_ffree, report the smaller of the number of object available | |
952 | * and the number of blocks (each object will take at least a block). | |
953 | */ | |
baab0630 | 954 | statp->f_ffree = MIN(availobjs, availbytes >> DNODE_SHIFT); |
34dc7c2f | 955 | statp->f_files = statp->f_ffree + usedobjs; |
53cf50e0 | 956 | statp->f_fsid.val[0] = dentry->d_sb->s_dev; |
3558fd73 BB |
957 | statp->f_fsid.val[1] = 0; |
958 | statp->f_type = ZFS_SUPER_MAGIC; | |
959 | statp->f_namelen = ZFS_MAXNAMELEN; | |
34dc7c2f BB |
960 | |
961 | /* | |
3558fd73 | 962 | * We have all of 40 characters to stuff a string here. |
34dc7c2f BB |
963 | * Is there anything useful we could/should provide? |
964 | */ | |
3558fd73 | 965 | bzero(statp->f_spare, sizeof (statp->f_spare)); |
34dc7c2f | 966 | |
3558fd73 | 967 | ZFS_EXIT(zsb); |
34dc7c2f BB |
968 | return (0); |
969 | } | |
e5c39b95 | 970 | EXPORT_SYMBOL(zfs_statvfs); |
34dc7c2f | 971 | |
e5c39b95 | 972 | int |
3558fd73 | 973 | zfs_root(zfs_sb_t *zsb, struct inode **ipp) |
34dc7c2f | 974 | { |
34dc7c2f BB |
975 | znode_t *rootzp; |
976 | int error; | |
977 | ||
3558fd73 | 978 | ZFS_ENTER(zsb); |
34dc7c2f | 979 | |
3558fd73 | 980 | error = zfs_zget(zsb, zsb->z_root, &rootzp); |
34dc7c2f | 981 | if (error == 0) |
3558fd73 | 982 | *ipp = ZTOI(rootzp); |
34dc7c2f | 983 | |
3558fd73 | 984 | ZFS_EXIT(zsb); |
34dc7c2f BB |
985 | return (error); |
986 | } | |
e5c39b95 | 987 | EXPORT_SYMBOL(zfs_root); |
34dc7c2f | 988 | |
ab26409d BB |
989 | #ifdef HAVE_SHRINK |
990 | int | |
991 | zfs_sb_prune(struct super_block *sb, unsigned long nr_to_scan, int *objects) | |
992 | { | |
993 | zfs_sb_t *zsb = sb->s_fs_info; | |
994 | struct shrinker *shrinker = &sb->s_shrink; | |
995 | struct shrink_control sc = { | |
996 | .nr_to_scan = nr_to_scan, | |
997 | .gfp_mask = GFP_KERNEL, | |
998 | }; | |
999 | ||
1000 | ZFS_ENTER(zsb); | |
1001 | *objects = (*shrinker->shrink)(shrinker, &sc); | |
1002 | ZFS_EXIT(zsb); | |
1003 | ||
1004 | return (0); | |
1005 | } | |
1006 | EXPORT_SYMBOL(zfs_sb_prune); | |
1007 | #endif /* HAVE_SHRINK */ | |
1008 | ||
34dc7c2f | 1009 | /* |
3558fd73 | 1010 | * Teardown the zfs_sb_t::z_os. |
34dc7c2f BB |
1011 | * |
1012 | * Note, if 'unmounting' if FALSE, we return with the 'z_teardown_lock' | |
1013 | * and 'z_teardown_inactive_lock' held. | |
1014 | */ | |
3558fd73 | 1015 | int |
86f35f34 | 1016 | zfs_sb_teardown(zfs_sb_t *zsb, boolean_t unmounting) |
34dc7c2f BB |
1017 | { |
1018 | znode_t *zp; | |
1019 | ||
3558fd73 | 1020 | rrw_enter(&zsb->z_teardown_lock, RW_WRITER, FTAG); |
34dc7c2f BB |
1021 | |
1022 | if (!unmounting) { | |
1023 | /* | |
ceb43b93 BB |
1024 | * We purge the parent filesystem's super block as the |
1025 | * parent filesystem and all of its snapshots have their | |
1026 | * inode's super block set to the parent's filesystem's | |
1027 | * super block. Note, 'z_parent' is self referential | |
1028 | * for non-snapshots. | |
34dc7c2f | 1029 | */ |
ceb43b93 | 1030 | shrink_dcache_sb(zsb->z_parent->z_sb); |
e30c0ada | 1031 | (void) spl_invalidate_inodes(zsb->z_parent->z_sb, 0); |
34dc7c2f BB |
1032 | } |
1033 | ||
b00131d4 GB |
1034 | /* |
1035 | * Drain the iput_taskq to ensure all active references to the | |
1036 | * zfs_sb_t have been handled only then can it be safely destroyed. | |
1037 | */ | |
1038 | taskq_wait(dsl_pool_iput_taskq(dmu_objset_pool(zsb->z_os))); | |
1039 | ||
34dc7c2f BB |
1040 | /* |
1041 | * Close the zil. NB: Can't close the zil while zfs_inactive | |
1042 | * threads are blocked as zil_close can call zfs_inactive. | |
1043 | */ | |
3558fd73 BB |
1044 | if (zsb->z_log) { |
1045 | zil_close(zsb->z_log); | |
1046 | zsb->z_log = NULL; | |
34dc7c2f BB |
1047 | } |
1048 | ||
3558fd73 | 1049 | rw_enter(&zsb->z_teardown_inactive_lock, RW_WRITER); |
34dc7c2f BB |
1050 | |
1051 | /* | |
1052 | * If we are not unmounting (ie: online recv) and someone already | |
1053 | * unmounted this file system while we were doing the switcheroo, | |
1054 | * or a reopen of z_os failed then just bail out now. | |
1055 | */ | |
3558fd73 BB |
1056 | if (!unmounting && (zsb->z_unmounted || zsb->z_os == NULL)) { |
1057 | rw_exit(&zsb->z_teardown_inactive_lock); | |
1058 | rrw_exit(&zsb->z_teardown_lock, FTAG); | |
34dc7c2f BB |
1059 | return (EIO); |
1060 | } | |
1061 | ||
1062 | /* | |
1063 | * At this point there are no vops active, and any new vops will | |
1064 | * fail with EIO since we have z_teardown_lock for writer (only | |
1065 | * relavent for forced unmount). | |
1066 | * | |
1067 | * Release all holds on dbufs. | |
1068 | */ | |
3558fd73 BB |
1069 | mutex_enter(&zsb->z_znodes_lock); |
1070 | for (zp = list_head(&zsb->z_all_znodes); zp != NULL; | |
1071 | zp = list_next(&zsb->z_all_znodes, zp)) | |
428870ff | 1072 | if (zp->z_sa_hdl) { |
3558fd73 | 1073 | ASSERT(atomic_read(&ZTOI(zp)->i_count) > 0); |
34dc7c2f BB |
1074 | zfs_znode_dmu_fini(zp); |
1075 | } | |
3558fd73 | 1076 | mutex_exit(&zsb->z_znodes_lock); |
34dc7c2f BB |
1077 | |
1078 | /* | |
1079 | * If we are unmounting, set the unmounted flag and let new vops | |
1080 | * unblock. zfs_inactive will have the unmounted behavior, and all | |
1081 | * other vops will fail with EIO. | |
1082 | */ | |
1083 | if (unmounting) { | |
3558fd73 BB |
1084 | zsb->z_unmounted = B_TRUE; |
1085 | rrw_exit(&zsb->z_teardown_lock, FTAG); | |
1086 | rw_exit(&zsb->z_teardown_inactive_lock); | |
34dc7c2f BB |
1087 | } |
1088 | ||
1089 | /* | |
1090 | * z_os will be NULL if there was an error in attempting to reopen | |
3558fd73 BB |
1091 | * zsb, so just return as the properties had already been |
1092 | * | |
34dc7c2f BB |
1093 | * unregistered and cached data had been evicted before. |
1094 | */ | |
3558fd73 | 1095 | if (zsb->z_os == NULL) |
34dc7c2f BB |
1096 | return (0); |
1097 | ||
1098 | /* | |
1099 | * Unregister properties. | |
1100 | */ | |
3558fd73 | 1101 | zfs_unregister_callbacks(zsb); |
34dc7c2f BB |
1102 | |
1103 | /* | |
1104 | * Evict cached data | |
1105 | */ | |
3558fd73 | 1106 | if (dmu_objset_is_dirty_anywhere(zsb->z_os)) |
2cf7f52b | 1107 | if (!zfs_is_readonly(zsb)) |
3558fd73 BB |
1108 | txg_wait_synced(dmu_objset_pool(zsb->z_os), 0); |
1109 | (void) dmu_objset_evict_dbufs(zsb->z_os); | |
34dc7c2f BB |
1110 | |
1111 | return (0); | |
1112 | } | |
86f35f34 | 1113 | EXPORT_SYMBOL(zfs_sb_teardown); |
34dc7c2f | 1114 | |
5547c2f1 BB |
1115 | #if defined(HAVE_BDI) && !defined(HAVE_BDI_SETUP_AND_REGISTER) |
1116 | atomic_long_t zfs_bdi_seq = ATOMIC_LONG_INIT(0); | |
1117 | #endif /* HAVE_BDI && !HAVE_BDI_SETUP_AND_REGISTER */ | |
76659dc1 | 1118 | |
e5c39b95 | 1119 | int |
3558fd73 | 1120 | zfs_domount(struct super_block *sb, void *data, int silent) |
34dc7c2f | 1121 | { |
3558fd73 BB |
1122 | zpl_mount_data_t *zmd = data; |
1123 | const char *osname = zmd->z_osname; | |
1124 | zfs_sb_t *zsb; | |
1125 | struct inode *root_inode; | |
1126 | uint64_t recordsize; | |
1127 | int error; | |
34dc7c2f | 1128 | |
3558fd73 BB |
1129 | error = zfs_sb_create(osname, &zsb); |
1130 | if (error) | |
1131 | return (error); | |
1132 | ||
1133 | if ((error = dsl_prop_get_integer(osname, "recordsize", | |
1134 | &recordsize, NULL))) | |
1135 | goto out; | |
1136 | ||
1137 | zsb->z_sb = sb; | |
3558fd73 BB |
1138 | sb->s_fs_info = zsb; |
1139 | sb->s_magic = ZFS_SUPER_MAGIC; | |
1140 | sb->s_maxbytes = MAX_LFS_FILESIZE; | |
1141 | sb->s_time_gran = 1; | |
1142 | sb->s_blocksize = recordsize; | |
1143 | sb->s_blocksize_bits = ilog2(recordsize); | |
5547c2f1 BB |
1144 | |
1145 | #ifdef HAVE_BDI | |
1146 | /* | |
1147 | * 2.6.32 API change, | |
1148 | * Added backing_device_info (BDI) per super block interfaces. A BDI | |
1149 | * must be configured when using a non-device backed filesystem for | |
1150 | * proper writeback. This is not required for older pdflush kernels. | |
1151 | * | |
1152 | * NOTE: Linux read-ahead is disabled in favor of zfs read-ahead. | |
1153 | */ | |
1154 | zsb->z_bdi.ra_pages = 0; | |
1155 | sb->s_bdi = &zsb->z_bdi; | |
1156 | ||
1157 | error = -bdi_setup_and_register(&zsb->z_bdi, "zfs", BDI_CAP_MAP_COPY); | |
1158 | if (error) | |
1159 | goto out; | |
1160 | #endif /* HAVE_BDI */ | |
3558fd73 BB |
1161 | |
1162 | /* Set callback operations for the file system. */ | |
1163 | sb->s_op = &zpl_super_operations; | |
1164 | sb->s_xattr = zpl_xattr_handlers; | |
055656d4 | 1165 | sb->s_export_op = &zpl_export_operations; |
3558fd73 BB |
1166 | |
1167 | /* Set features for file system. */ | |
1168 | zfs_set_fuid_feature(zsb); | |
1169 | ||
1170 | if (dmu_objset_is_snapshot(zsb->z_os)) { | |
1171 | uint64_t pval; | |
1172 | ||
1173 | atime_changed_cb(zsb, B_FALSE); | |
1174 | readonly_changed_cb(zsb, B_TRUE); | |
1175 | if ((error = dsl_prop_get_integer(osname,"xattr",&pval,NULL))) | |
1176 | goto out; | |
1177 | xattr_changed_cb(zsb, pval); | |
1178 | zsb->z_issnap = B_TRUE; | |
1179 | zsb->z_os->os_sync = ZFS_SYNC_DISABLED; | |
1180 | ||
1181 | mutex_enter(&zsb->z_os->os_user_ptr_lock); | |
1182 | dmu_objset_set_user(zsb->z_os, zsb); | |
1183 | mutex_exit(&zsb->z_os->os_user_ptr_lock); | |
1184 | } else { | |
1185 | error = zfs_sb_setup(zsb, B_TRUE); | |
1186 | #ifdef HAVE_SNAPSHOT | |
1187 | (void) zfs_snap_create(zsb); | |
1188 | #endif /* HAVE_SNAPSHOT */ | |
34dc7c2f BB |
1189 | } |
1190 | ||
3558fd73 BB |
1191 | /* Allocate a root inode for the filesystem. */ |
1192 | error = zfs_root(zsb, &root_inode); | |
1193 | if (error) { | |
1194 | (void) zfs_umount(sb); | |
1195 | goto out; | |
34dc7c2f BB |
1196 | } |
1197 | ||
3558fd73 BB |
1198 | /* Allocate a root dentry for the filesystem */ |
1199 | sb->s_root = d_alloc_root(root_inode); | |
1200 | if (sb->s_root == NULL) { | |
1201 | (void) zfs_umount(sb); | |
1202 | error = ENOMEM; | |
1203 | goto out; | |
1204 | } | |
1205 | out: | |
1206 | if (error) { | |
1207 | dmu_objset_disown(zsb->z_os, zsb); | |
1208 | zfs_sb_free(zsb); | |
1209 | } | |
34dc7c2f | 1210 | |
3558fd73 BB |
1211 | return (error); |
1212 | } | |
1213 | EXPORT_SYMBOL(zfs_domount); | |
1214 | ||
1215 | /*ARGSUSED*/ | |
1216 | int | |
1217 | zfs_umount(struct super_block *sb) | |
1218 | { | |
1219 | zfs_sb_t *zsb = sb->s_fs_info; | |
1220 | objset_t *os; | |
1221 | ||
86f35f34 | 1222 | VERIFY(zfs_sb_teardown(zsb, B_TRUE) == 0); |
3558fd73 | 1223 | os = zsb->z_os; |
34dc7c2f | 1224 | |
5547c2f1 BB |
1225 | #ifdef HAVE_BDI |
1226 | bdi_destroy(sb->s_bdi); | |
1227 | #endif /* HAVE_BDI */ | |
76659dc1 | 1228 | |
34dc7c2f BB |
1229 | /* |
1230 | * z_os will be NULL if there was an error in | |
3558fd73 | 1231 | * attempting to reopen zsb. |
34dc7c2f BB |
1232 | */ |
1233 | if (os != NULL) { | |
1234 | /* | |
1235 | * Unset the objset user_ptr. | |
1236 | */ | |
428870ff | 1237 | mutex_enter(&os->os_user_ptr_lock); |
34dc7c2f | 1238 | dmu_objset_set_user(os, NULL); |
428870ff | 1239 | mutex_exit(&os->os_user_ptr_lock); |
34dc7c2f BB |
1240 | |
1241 | /* | |
b128c09f | 1242 | * Finally release the objset |
34dc7c2f | 1243 | */ |
3558fd73 | 1244 | dmu_objset_disown(os, zsb); |
34dc7c2f BB |
1245 | } |
1246 | ||
3558fd73 | 1247 | zfs_sb_free(zsb); |
34dc7c2f BB |
1248 | return (0); |
1249 | } | |
e5c39b95 | 1250 | EXPORT_SYMBOL(zfs_umount); |
34dc7c2f | 1251 | |
0de19dad BB |
1252 | int |
1253 | zfs_remount(struct super_block *sb, int *flags, char *data) | |
1254 | { | |
0de19dad | 1255 | /* |
2cf7f52b BB |
1256 | * All namespace flags (MNT_*) and super block flags (MS_*) will |
1257 | * be handled by the Linux VFS. Only handle custom options here. | |
0de19dad | 1258 | */ |
0de19dad BB |
1259 | return (0); |
1260 | } | |
1261 | EXPORT_SYMBOL(zfs_remount); | |
1262 | ||
e5c39b95 | 1263 | int |
2cf7f52b | 1264 | zfs_vget(struct super_block *sb, struct inode **ipp, fid_t *fidp) |
34dc7c2f | 1265 | { |
2cf7f52b | 1266 | zfs_sb_t *zsb = sb->s_fs_info; |
34dc7c2f BB |
1267 | znode_t *zp; |
1268 | uint64_t object = 0; | |
1269 | uint64_t fid_gen = 0; | |
1270 | uint64_t gen_mask; | |
1271 | uint64_t zp_gen; | |
3558fd73 | 1272 | int i, err; |
34dc7c2f | 1273 | |
3558fd73 | 1274 | *ipp = NULL; |
34dc7c2f | 1275 | |
3558fd73 | 1276 | ZFS_ENTER(zsb); |
34dc7c2f BB |
1277 | |
1278 | if (fidp->fid_len == LONG_FID_LEN) { | |
1279 | zfid_long_t *zlfid = (zfid_long_t *)fidp; | |
1280 | uint64_t objsetid = 0; | |
1281 | uint64_t setgen = 0; | |
1282 | ||
1283 | for (i = 0; i < sizeof (zlfid->zf_setid); i++) | |
1284 | objsetid |= ((uint64_t)zlfid->zf_setid[i]) << (8 * i); | |
1285 | ||
1286 | for (i = 0; i < sizeof (zlfid->zf_setgen); i++) | |
1287 | setgen |= ((uint64_t)zlfid->zf_setgen[i]) << (8 * i); | |
1288 | ||
3558fd73 | 1289 | ZFS_EXIT(zsb); |
34dc7c2f | 1290 | |
3558fd73 BB |
1291 | #ifdef HAVE_SNAPSHOT |
1292 | err = zfsctl_lookup_objset(vfsp, objsetid, &zsb); | |
34dc7c2f BB |
1293 | if (err) |
1294 | return (EINVAL); | |
3558fd73 BB |
1295 | #endif /* HAVE_SNAPSHOT */ |
1296 | ZFS_ENTER(zsb); | |
34dc7c2f BB |
1297 | } |
1298 | ||
1299 | if (fidp->fid_len == SHORT_FID_LEN || fidp->fid_len == LONG_FID_LEN) { | |
1300 | zfid_short_t *zfid = (zfid_short_t *)fidp; | |
1301 | ||
1302 | for (i = 0; i < sizeof (zfid->zf_object); i++) | |
1303 | object |= ((uint64_t)zfid->zf_object[i]) << (8 * i); | |
1304 | ||
1305 | for (i = 0; i < sizeof (zfid->zf_gen); i++) | |
1306 | fid_gen |= ((uint64_t)zfid->zf_gen[i]) << (8 * i); | |
1307 | } else { | |
3558fd73 | 1308 | ZFS_EXIT(zsb); |
34dc7c2f BB |
1309 | return (EINVAL); |
1310 | } | |
1311 | ||
3558fd73 | 1312 | #ifdef HAVE_SNAPSHOT |
34dc7c2f BB |
1313 | /* A zero fid_gen means we are in the .zfs control directories */ |
1314 | if (fid_gen == 0 && | |
1315 | (object == ZFSCTL_INO_ROOT || object == ZFSCTL_INO_SNAPDIR)) { | |
3558fd73 BB |
1316 | *ipp = zsb->z_ctldir; |
1317 | ASSERT(*ipp != NULL); | |
34dc7c2f | 1318 | if (object == ZFSCTL_INO_SNAPDIR) { |
3558fd73 | 1319 | VERIFY(zfsctl_root_lookup(*ipp, "snapshot", ipp, NULL, |
34dc7c2f BB |
1320 | 0, NULL, NULL, NULL, NULL, NULL) == 0); |
1321 | } else { | |
3558fd73 | 1322 | igrab(*ipp); |
34dc7c2f | 1323 | } |
3558fd73 | 1324 | ZFS_EXIT(zsb); |
34dc7c2f BB |
1325 | return (0); |
1326 | } | |
3558fd73 | 1327 | #endif /* HAVE_SNAPSHOT */ |
34dc7c2f BB |
1328 | |
1329 | gen_mask = -1ULL >> (64 - 8 * i); | |
1330 | ||
1331 | dprintf("getting %llu [%u mask %llx]\n", object, fid_gen, gen_mask); | |
3558fd73 BB |
1332 | if ((err = zfs_zget(zsb, object, &zp))) { |
1333 | ZFS_EXIT(zsb); | |
34dc7c2f BB |
1334 | return (err); |
1335 | } | |
3558fd73 | 1336 | (void) sa_lookup(zp->z_sa_hdl, SA_ZPL_GEN(zsb), &zp_gen, |
428870ff BB |
1337 | sizeof (uint64_t)); |
1338 | zp_gen = zp_gen & gen_mask; | |
34dc7c2f BB |
1339 | if (zp_gen == 0) |
1340 | zp_gen = 1; | |
1341 | if (zp->z_unlinked || zp_gen != fid_gen) { | |
1342 | dprintf("znode gen (%u) != fid gen (%u)\n", zp_gen, fid_gen); | |
3558fd73 BB |
1343 | iput(ZTOI(zp)); |
1344 | ZFS_EXIT(zsb); | |
34dc7c2f BB |
1345 | return (EINVAL); |
1346 | } | |
1347 | ||
3558fd73 BB |
1348 | *ipp = ZTOI(zp); |
1349 | if (*ipp) | |
1350 | zfs_inode_update(ITOZ(*ipp)); | |
960e08fe | 1351 | |
3558fd73 | 1352 | ZFS_EXIT(zsb); |
34dc7c2f BB |
1353 | return (0); |
1354 | } | |
e5c39b95 | 1355 | EXPORT_SYMBOL(zfs_vget); |
34dc7c2f BB |
1356 | |
1357 | /* | |
3558fd73 | 1358 | * Block out VOPs and close zfs_sb_t::z_os |
34dc7c2f BB |
1359 | * |
1360 | * Note, if successful, then we return with the 'z_teardown_lock' and | |
1361 | * 'z_teardown_inactive_lock' write held. | |
1362 | */ | |
1363 | int | |
3558fd73 | 1364 | zfs_suspend_fs(zfs_sb_t *zsb) |
34dc7c2f BB |
1365 | { |
1366 | int error; | |
1367 | ||
86f35f34 | 1368 | if ((error = zfs_sb_teardown(zsb, B_FALSE)) != 0) |
34dc7c2f | 1369 | return (error); |
3558fd73 | 1370 | dmu_objset_disown(zsb->z_os, zsb); |
34dc7c2f BB |
1371 | |
1372 | return (0); | |
1373 | } | |
e5c39b95 | 1374 | EXPORT_SYMBOL(zfs_suspend_fs); |
34dc7c2f BB |
1375 | |
1376 | /* | |
3558fd73 | 1377 | * Reopen zfs_sb_t::z_os and release VOPs. |
34dc7c2f BB |
1378 | */ |
1379 | int | |
3558fd73 | 1380 | zfs_resume_fs(zfs_sb_t *zsb, const char *osname) |
34dc7c2f | 1381 | { |
428870ff | 1382 | int err, err2; |
34dc7c2f | 1383 | |
3558fd73 BB |
1384 | ASSERT(RRW_WRITE_HELD(&zsb->z_teardown_lock)); |
1385 | ASSERT(RW_WRITE_HELD(&zsb->z_teardown_inactive_lock)); | |
34dc7c2f | 1386 | |
3558fd73 | 1387 | err = dmu_objset_own(osname, DMU_OST_ZFS, B_FALSE, zsb, &zsb->z_os); |
34dc7c2f | 1388 | if (err) { |
3558fd73 | 1389 | zsb->z_os = NULL; |
34dc7c2f BB |
1390 | } else { |
1391 | znode_t *zp; | |
428870ff BB |
1392 | uint64_t sa_obj = 0; |
1393 | ||
3558fd73 | 1394 | err2 = zap_lookup(zsb->z_os, MASTER_NODE_OBJ, |
428870ff BB |
1395 | ZFS_SA_ATTRS, 8, 1, &sa_obj); |
1396 | ||
3558fd73 | 1397 | if ((err || err2) && zsb->z_version >= ZPL_VERSION_SA) |
428870ff BB |
1398 | goto bail; |
1399 | ||
1400 | ||
3558fd73 BB |
1401 | if ((err = sa_setup(zsb->z_os, sa_obj, |
1402 | zfs_attr_table, ZPL_END, &zsb->z_attr_table)) != 0) | |
572e2857 | 1403 | goto bail; |
34dc7c2f | 1404 | |
3558fd73 | 1405 | VERIFY(zfs_sb_setup(zsb, B_FALSE) == 0); |
34dc7c2f BB |
1406 | |
1407 | /* | |
1408 | * Attempt to re-establish all the active znodes with | |
1409 | * their dbufs. If a zfs_rezget() fails, then we'll let | |
1410 | * any potential callers discover that via ZFS_ENTER_VERIFY_VP | |
1411 | * when they try to use their znode. | |
1412 | */ | |
3558fd73 BB |
1413 | mutex_enter(&zsb->z_znodes_lock); |
1414 | for (zp = list_head(&zsb->z_all_znodes); zp; | |
1415 | zp = list_next(&zsb->z_all_znodes, zp)) { | |
34dc7c2f BB |
1416 | (void) zfs_rezget(zp); |
1417 | } | |
3558fd73 | 1418 | mutex_exit(&zsb->z_znodes_lock); |
34dc7c2f BB |
1419 | |
1420 | } | |
1421 | ||
428870ff | 1422 | bail: |
34dc7c2f | 1423 | /* release the VOPs */ |
3558fd73 BB |
1424 | rw_exit(&zsb->z_teardown_inactive_lock); |
1425 | rrw_exit(&zsb->z_teardown_lock, FTAG); | |
34dc7c2f BB |
1426 | |
1427 | if (err) { | |
1428 | /* | |
3558fd73 | 1429 | * Since we couldn't reopen zfs_sb_t::z_os, force |
34dc7c2f BB |
1430 | * unmount this file system. |
1431 | */ | |
3558fd73 | 1432 | (void) zfs_umount(zsb->z_sb); |
34dc7c2f BB |
1433 | } |
1434 | return (err); | |
1435 | } | |
e5c39b95 | 1436 | EXPORT_SYMBOL(zfs_resume_fs); |
34dc7c2f | 1437 | |
34dc7c2f | 1438 | int |
3558fd73 | 1439 | zfs_set_version(zfs_sb_t *zsb, uint64_t newvers) |
34dc7c2f BB |
1440 | { |
1441 | int error; | |
3558fd73 | 1442 | objset_t *os = zsb->z_os; |
34dc7c2f | 1443 | dmu_tx_t *tx; |
34dc7c2f BB |
1444 | |
1445 | if (newvers < ZPL_VERSION_INITIAL || newvers > ZPL_VERSION) | |
1446 | return (EINVAL); | |
1447 | ||
3558fd73 | 1448 | if (newvers < zsb->z_version) |
9babb374 | 1449 | return (EINVAL); |
34dc7c2f | 1450 | |
428870ff | 1451 | if (zfs_spa_version_map(newvers) > |
3558fd73 | 1452 | spa_version(dmu_objset_spa(zsb->z_os))) |
428870ff BB |
1453 | return (ENOTSUP); |
1454 | ||
34dc7c2f | 1455 | tx = dmu_tx_create(os); |
9babb374 | 1456 | dmu_tx_hold_zap(tx, MASTER_NODE_OBJ, B_FALSE, ZPL_VERSION_STR); |
3558fd73 | 1457 | if (newvers >= ZPL_VERSION_SA && !zsb->z_use_sa) { |
428870ff BB |
1458 | dmu_tx_hold_zap(tx, MASTER_NODE_OBJ, B_TRUE, |
1459 | ZFS_SA_ATTRS); | |
1460 | dmu_tx_hold_zap(tx, DMU_NEW_OBJECT, FALSE, NULL); | |
1461 | } | |
34dc7c2f BB |
1462 | error = dmu_tx_assign(tx, TXG_WAIT); |
1463 | if (error) { | |
1464 | dmu_tx_abort(tx); | |
9babb374 BB |
1465 | return (error); |
1466 | } | |
428870ff | 1467 | |
9babb374 BB |
1468 | error = zap_update(os, MASTER_NODE_OBJ, ZPL_VERSION_STR, |
1469 | 8, 1, &newvers, tx); | |
1470 | ||
1471 | if (error) { | |
1472 | dmu_tx_commit(tx); | |
1473 | return (error); | |
34dc7c2f | 1474 | } |
34dc7c2f | 1475 | |
3558fd73 | 1476 | if (newvers >= ZPL_VERSION_SA && !zsb->z_use_sa) { |
428870ff BB |
1477 | uint64_t sa_obj; |
1478 | ||
3558fd73 | 1479 | ASSERT3U(spa_version(dmu_objset_spa(zsb->z_os)), >=, |
428870ff BB |
1480 | SPA_VERSION_SA); |
1481 | sa_obj = zap_create(os, DMU_OT_SA_MASTER_NODE, | |
1482 | DMU_OT_NONE, 0, tx); | |
1483 | ||
1484 | error = zap_add(os, MASTER_NODE_OBJ, | |
1485 | ZFS_SA_ATTRS, 8, 1, &sa_obj, tx); | |
1486 | ASSERT3U(error, ==, 0); | |
1487 | ||
1488 | VERIFY(0 == sa_set_sa_object(os, sa_obj)); | |
1489 | sa_register_update_callback(os, zfs_sa_upgrade); | |
1490 | } | |
1491 | ||
1492 | spa_history_log_internal(LOG_DS_UPGRADE, | |
1493 | dmu_objset_spa(os), tx, "oldver=%llu newver=%llu dataset = %llu", | |
3558fd73 | 1494 | zsb->z_version, newvers, dmu_objset_id(os)); |
9babb374 | 1495 | |
34dc7c2f BB |
1496 | dmu_tx_commit(tx); |
1497 | ||
3558fd73 | 1498 | zsb->z_version = newvers; |
9babb374 | 1499 | |
3558fd73 BB |
1500 | if (zsb->z_version >= ZPL_VERSION_FUID) |
1501 | zfs_set_fuid_feature(zsb); | |
9babb374 BB |
1502 | |
1503 | return (0); | |
34dc7c2f | 1504 | } |
e5c39b95 | 1505 | EXPORT_SYMBOL(zfs_set_version); |
34dc7c2f BB |
1506 | |
1507 | /* | |
1508 | * Read a property stored within the master node. | |
1509 | */ | |
1510 | int | |
1511 | zfs_get_zplprop(objset_t *os, zfs_prop_t prop, uint64_t *value) | |
1512 | { | |
1513 | const char *pname; | |
b128c09f | 1514 | int error = ENOENT; |
34dc7c2f BB |
1515 | |
1516 | /* | |
1517 | * Look up the file system's value for the property. For the | |
1518 | * version property, we look up a slightly different string. | |
1519 | */ | |
1520 | if (prop == ZFS_PROP_VERSION) | |
1521 | pname = ZPL_VERSION_STR; | |
1522 | else | |
1523 | pname = zfs_prop_to_name(prop); | |
1524 | ||
b128c09f BB |
1525 | if (os != NULL) |
1526 | error = zap_lookup(os, MASTER_NODE_OBJ, pname, 8, 1, value); | |
34dc7c2f BB |
1527 | |
1528 | if (error == ENOENT) { | |
1529 | /* No value set, use the default value */ | |
1530 | switch (prop) { | |
1531 | case ZFS_PROP_VERSION: | |
1532 | *value = ZPL_VERSION; | |
1533 | break; | |
1534 | case ZFS_PROP_NORMALIZE: | |
1535 | case ZFS_PROP_UTF8ONLY: | |
1536 | *value = 0; | |
1537 | break; | |
1538 | case ZFS_PROP_CASE: | |
1539 | *value = ZFS_CASE_SENSITIVE; | |
1540 | break; | |
1541 | default: | |
1542 | return (error); | |
1543 | } | |
1544 | error = 0; | |
1545 | } | |
1546 | return (error); | |
1547 | } | |
86f35f34 | 1548 | EXPORT_SYMBOL(zfs_get_zplprop); |
3558fd73 BB |
1549 | |
1550 | void | |
1551 | zfs_init(void) | |
1552 | { | |
1553 | zfs_znode_init(); | |
1554 | dmu_objset_register_type(DMU_OST_ZFS, zfs_space_delta_cb); | |
1555 | register_filesystem(&zpl_fs_type); | |
ab26409d | 1556 | (void) arc_add_prune_callback(zpl_prune_sbs, NULL); |
3558fd73 BB |
1557 | } |
1558 | ||
1559 | void | |
1560 | zfs_fini(void) | |
1561 | { | |
1562 | unregister_filesystem(&zpl_fs_type); | |
1563 | zfs_znode_fini(); | |
1564 | } |