]>
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> |
a94addd9 | 52 | #include <sys/sa_impl.h> |
34dc7c2f BB |
53 | #include <sys/varargs.h> |
54 | #include <sys/policy.h> | |
55 | #include <sys/atomic.h> | |
56 | #include <sys/mkdev.h> | |
57 | #include <sys/modctl.h> | |
58 | #include <sys/refstr.h> | |
59 | #include <sys/zfs_ioctl.h> | |
ebe7e575 | 60 | #include <sys/zfs_ctldir.h> |
34dc7c2f BB |
61 | #include <sys/zfs_fuid.h> |
62 | #include <sys/bootconf.h> | |
63 | #include <sys/sunddi.h> | |
64 | #include <sys/dnlc.h> | |
65 | #include <sys/dmu_objset.h> | |
66 | #include <sys/spa_boot.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 | 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 | 320 | if (bonustype == DMU_OT_ZNODE) { |
a94addd9 | 321 | znode_phys_t *znp = data; |
428870ff BB |
322 | *userp = znp->zp_uid; |
323 | *groupp = znp->zp_gid; | |
9babb374 | 324 | } else { |
428870ff | 325 | int hdrsize; |
a94addd9 MA |
326 | sa_hdr_phys_t *sap = data; |
327 | sa_hdr_phys_t sa = *sap; | |
328 | boolean_t swap = B_FALSE; | |
9babb374 | 329 | |
428870ff | 330 | ASSERT(bonustype == DMU_OT_SA); |
428870ff | 331 | |
a94addd9 | 332 | if (sa.sa_magic == 0) { |
428870ff BB |
333 | /* |
334 | * This should only happen for newly created | |
335 | * files that haven't had the znode data filled | |
336 | * in yet. | |
337 | */ | |
338 | *userp = 0; | |
339 | *groupp = 0; | |
a94addd9 MA |
340 | return (0); |
341 | } | |
342 | if (sa.sa_magic == BSWAP_32(SA_MAGIC)) { | |
343 | sa.sa_magic = SA_MAGIC; | |
344 | sa.sa_layout_info = BSWAP_16(sa.sa_layout_info); | |
345 | swap = B_TRUE; | |
346 | } else { | |
347 | VERIFY3U(sa.sa_magic, ==, SA_MAGIC); | |
348 | } | |
349 | ||
350 | hdrsize = sa_hdrsize(&sa); | |
351 | VERIFY3U(hdrsize, >=, sizeof (sa_hdr_phys_t)); | |
352 | *userp = *((uint64_t *)((uintptr_t)data + hdrsize + | |
353 | SA_UID_OFFSET)); | |
354 | *groupp = *((uint64_t *)((uintptr_t)data + hdrsize + | |
355 | SA_GID_OFFSET)); | |
356 | if (swap) { | |
357 | *userp = BSWAP_64(*userp); | |
358 | *groupp = BSWAP_64(*groupp); | |
428870ff | 359 | } |
9babb374 | 360 | } |
428870ff | 361 | return (error); |
9babb374 BB |
362 | } |
363 | ||
364 | static void | |
3558fd73 | 365 | fuidstr_to_sid(zfs_sb_t *zsb, const char *fuidstr, |
9babb374 BB |
366 | char *domainbuf, int buflen, uid_t *ridp) |
367 | { | |
9babb374 BB |
368 | uint64_t fuid; |
369 | const char *domain; | |
370 | ||
371 | fuid = strtonum(fuidstr, NULL); | |
372 | ||
3558fd73 | 373 | domain = zfs_fuid_find_by_idx(zsb, FUID_INDEX(fuid)); |
9babb374 BB |
374 | if (domain) |
375 | (void) strlcpy(domainbuf, domain, buflen); | |
376 | else | |
377 | domainbuf[0] = '\0'; | |
378 | *ridp = FUID_RID(fuid); | |
379 | } | |
380 | ||
381 | static uint64_t | |
3558fd73 | 382 | zfs_userquota_prop_to_obj(zfs_sb_t *zsb, zfs_userquota_prop_t type) |
9babb374 BB |
383 | { |
384 | switch (type) { | |
385 | case ZFS_PROP_USERUSED: | |
386 | return (DMU_USERUSED_OBJECT); | |
387 | case ZFS_PROP_GROUPUSED: | |
388 | return (DMU_GROUPUSED_OBJECT); | |
389 | case ZFS_PROP_USERQUOTA: | |
3558fd73 | 390 | return (zsb->z_userquota_obj); |
9babb374 | 391 | case ZFS_PROP_GROUPQUOTA: |
3558fd73 | 392 | return (zsb->z_groupquota_obj); |
149e873a BB |
393 | default: |
394 | return (ENOTSUP); | |
9babb374 BB |
395 | } |
396 | return (0); | |
397 | } | |
398 | ||
399 | int | |
3558fd73 | 400 | zfs_userspace_many(zfs_sb_t *zsb, zfs_userquota_prop_t type, |
9babb374 BB |
401 | uint64_t *cookiep, void *vbuf, uint64_t *bufsizep) |
402 | { | |
403 | int error; | |
404 | zap_cursor_t zc; | |
405 | zap_attribute_t za; | |
406 | zfs_useracct_t *buf = vbuf; | |
407 | uint64_t obj; | |
408 | ||
3558fd73 | 409 | if (!dmu_objset_userspace_present(zsb->z_os)) |
9babb374 BB |
410 | return (ENOTSUP); |
411 | ||
3558fd73 | 412 | obj = zfs_userquota_prop_to_obj(zsb, type); |
9babb374 BB |
413 | if (obj == 0) { |
414 | *bufsizep = 0; | |
415 | return (0); | |
416 | } | |
417 | ||
3558fd73 | 418 | for (zap_cursor_init_serialized(&zc, zsb->z_os, obj, *cookiep); |
9babb374 BB |
419 | (error = zap_cursor_retrieve(&zc, &za)) == 0; |
420 | zap_cursor_advance(&zc)) { | |
421 | if ((uintptr_t)buf - (uintptr_t)vbuf + sizeof (zfs_useracct_t) > | |
422 | *bufsizep) | |
423 | break; | |
424 | ||
3558fd73 | 425 | fuidstr_to_sid(zsb, za.za_name, |
9babb374 BB |
426 | buf->zu_domain, sizeof (buf->zu_domain), &buf->zu_rid); |
427 | ||
428 | buf->zu_space = za.za_first_integer; | |
429 | buf++; | |
430 | } | |
431 | if (error == ENOENT) | |
432 | error = 0; | |
433 | ||
434 | ASSERT3U((uintptr_t)buf - (uintptr_t)vbuf, <=, *bufsizep); | |
435 | *bufsizep = (uintptr_t)buf - (uintptr_t)vbuf; | |
436 | *cookiep = zap_cursor_serialize(&zc); | |
437 | zap_cursor_fini(&zc); | |
438 | return (error); | |
439 | } | |
e5c39b95 | 440 | EXPORT_SYMBOL(zfs_userspace_many); |
9babb374 BB |
441 | |
442 | /* | |
443 | * buf must be big enough (eg, 32 bytes) | |
444 | */ | |
445 | static int | |
3558fd73 | 446 | id_to_fuidstr(zfs_sb_t *zsb, const char *domain, uid_t rid, |
9babb374 BB |
447 | char *buf, boolean_t addok) |
448 | { | |
449 | uint64_t fuid; | |
450 | int domainid = 0; | |
451 | ||
452 | if (domain && domain[0]) { | |
3558fd73 | 453 | domainid = zfs_fuid_find_by_domain(zsb, domain, NULL, addok); |
9babb374 BB |
454 | if (domainid == -1) |
455 | return (ENOENT); | |
456 | } | |
457 | fuid = FUID_ENCODE(domainid, rid); | |
458 | (void) sprintf(buf, "%llx", (longlong_t)fuid); | |
459 | return (0); | |
460 | } | |
461 | ||
462 | int | |
3558fd73 | 463 | zfs_userspace_one(zfs_sb_t *zsb, zfs_userquota_prop_t type, |
9babb374 BB |
464 | const char *domain, uint64_t rid, uint64_t *valp) |
465 | { | |
466 | char buf[32]; | |
467 | int err; | |
468 | uint64_t obj; | |
469 | ||
470 | *valp = 0; | |
471 | ||
3558fd73 | 472 | if (!dmu_objset_userspace_present(zsb->z_os)) |
9babb374 BB |
473 | return (ENOTSUP); |
474 | ||
3558fd73 | 475 | obj = zfs_userquota_prop_to_obj(zsb, type); |
9babb374 BB |
476 | if (obj == 0) |
477 | return (0); | |
478 | ||
3558fd73 | 479 | err = id_to_fuidstr(zsb, domain, rid, buf, B_FALSE); |
9babb374 BB |
480 | if (err) |
481 | return (err); | |
482 | ||
3558fd73 | 483 | err = zap_lookup(zsb->z_os, obj, buf, 8, 1, valp); |
9babb374 BB |
484 | if (err == ENOENT) |
485 | err = 0; | |
486 | return (err); | |
487 | } | |
e5c39b95 | 488 | EXPORT_SYMBOL(zfs_userspace_one); |
9babb374 BB |
489 | |
490 | int | |
3558fd73 | 491 | zfs_set_userquota(zfs_sb_t *zsb, zfs_userquota_prop_t type, |
9babb374 BB |
492 | const char *domain, uint64_t rid, uint64_t quota) |
493 | { | |
494 | char buf[32]; | |
495 | int err; | |
496 | dmu_tx_t *tx; | |
497 | uint64_t *objp; | |
498 | boolean_t fuid_dirtied; | |
499 | ||
500 | if (type != ZFS_PROP_USERQUOTA && type != ZFS_PROP_GROUPQUOTA) | |
501 | return (EINVAL); | |
502 | ||
3558fd73 | 503 | if (zsb->z_version < ZPL_VERSION_USERSPACE) |
9babb374 BB |
504 | return (ENOTSUP); |
505 | ||
3558fd73 BB |
506 | objp = (type == ZFS_PROP_USERQUOTA) ? &zsb->z_userquota_obj : |
507 | &zsb->z_groupquota_obj; | |
9babb374 | 508 | |
3558fd73 | 509 | err = id_to_fuidstr(zsb, domain, rid, buf, B_TRUE); |
9babb374 BB |
510 | if (err) |
511 | return (err); | |
3558fd73 | 512 | fuid_dirtied = zsb->z_fuid_dirty; |
9babb374 | 513 | |
3558fd73 | 514 | tx = dmu_tx_create(zsb->z_os); |
9babb374 BB |
515 | dmu_tx_hold_zap(tx, *objp ? *objp : DMU_NEW_OBJECT, B_TRUE, NULL); |
516 | if (*objp == 0) { | |
517 | dmu_tx_hold_zap(tx, MASTER_NODE_OBJ, B_TRUE, | |
518 | zfs_userquota_prop_prefixes[type]); | |
519 | } | |
520 | if (fuid_dirtied) | |
3558fd73 | 521 | zfs_fuid_txhold(zsb, tx); |
9babb374 BB |
522 | err = dmu_tx_assign(tx, TXG_WAIT); |
523 | if (err) { | |
524 | dmu_tx_abort(tx); | |
525 | return (err); | |
526 | } | |
527 | ||
3558fd73 | 528 | mutex_enter(&zsb->z_lock); |
9babb374 | 529 | if (*objp == 0) { |
3558fd73 | 530 | *objp = zap_create(zsb->z_os, DMU_OT_USERGROUP_QUOTA, |
9babb374 | 531 | DMU_OT_NONE, 0, tx); |
3558fd73 | 532 | VERIFY(0 == zap_add(zsb->z_os, MASTER_NODE_OBJ, |
9babb374 BB |
533 | zfs_userquota_prop_prefixes[type], 8, 1, objp, tx)); |
534 | } | |
3558fd73 | 535 | mutex_exit(&zsb->z_lock); |
9babb374 BB |
536 | |
537 | if (quota == 0) { | |
3558fd73 | 538 | err = zap_remove(zsb->z_os, *objp, buf, tx); |
9babb374 BB |
539 | if (err == ENOENT) |
540 | err = 0; | |
541 | } else { | |
3558fd73 | 542 | err = zap_update(zsb->z_os, *objp, buf, 8, 1, "a, tx); |
9babb374 BB |
543 | } |
544 | ASSERT(err == 0); | |
545 | if (fuid_dirtied) | |
3558fd73 | 546 | zfs_fuid_sync(zsb, tx); |
9babb374 BB |
547 | dmu_tx_commit(tx); |
548 | return (err); | |
549 | } | |
e5c39b95 | 550 | EXPORT_SYMBOL(zfs_set_userquota); |
9babb374 BB |
551 | |
552 | boolean_t | |
3558fd73 | 553 | zfs_fuid_overquota(zfs_sb_t *zsb, boolean_t isgroup, uint64_t fuid) |
9babb374 BB |
554 | { |
555 | char buf[32]; | |
556 | uint64_t used, quota, usedobj, quotaobj; | |
557 | int err; | |
558 | ||
559 | usedobj = isgroup ? DMU_GROUPUSED_OBJECT : DMU_USERUSED_OBJECT; | |
3558fd73 | 560 | quotaobj = isgroup ? zsb->z_groupquota_obj : zsb->z_userquota_obj; |
9babb374 | 561 | |
3558fd73 | 562 | if (quotaobj == 0 || zsb->z_replay) |
9babb374 BB |
563 | return (B_FALSE); |
564 | ||
565 | (void) sprintf(buf, "%llx", (longlong_t)fuid); | |
3558fd73 | 566 | err = zap_lookup(zsb->z_os, quotaobj, buf, 8, 1, "a); |
9babb374 BB |
567 | if (err != 0) |
568 | return (B_FALSE); | |
569 | ||
3558fd73 | 570 | err = zap_lookup(zsb->z_os, usedobj, buf, 8, 1, &used); |
9babb374 BB |
571 | if (err != 0) |
572 | return (B_FALSE); | |
573 | return (used >= quota); | |
574 | } | |
e5c39b95 | 575 | EXPORT_SYMBOL(zfs_fuid_overquota); |
9babb374 | 576 | |
428870ff | 577 | boolean_t |
3558fd73 | 578 | zfs_owner_overquota(zfs_sb_t *zsb, znode_t *zp, boolean_t isgroup) |
428870ff BB |
579 | { |
580 | uint64_t fuid; | |
581 | uint64_t quotaobj; | |
428870ff | 582 | |
3558fd73 | 583 | quotaobj = isgroup ? zsb->z_groupquota_obj : zsb->z_userquota_obj; |
428870ff | 584 | |
572e2857 | 585 | fuid = isgroup ? zp->z_gid : zp->z_uid; |
428870ff | 586 | |
3558fd73 | 587 | if (quotaobj == 0 || zsb->z_replay) |
428870ff BB |
588 | return (B_FALSE); |
589 | ||
3558fd73 | 590 | return (zfs_fuid_overquota(zsb, isgroup, fuid)); |
428870ff | 591 | } |
e5c39b95 | 592 | EXPORT_SYMBOL(zfs_owner_overquota); |
428870ff | 593 | |
9babb374 | 594 | int |
3558fd73 | 595 | zfs_sb_create(const char *osname, zfs_sb_t **zsbp) |
9babb374 BB |
596 | { |
597 | objset_t *os; | |
3558fd73 | 598 | zfs_sb_t *zsb; |
9babb374 BB |
599 | uint64_t zval; |
600 | int i, error; | |
428870ff | 601 | uint64_t sa_obj; |
9babb374 | 602 | |
bafc4e9e | 603 | zsb = kmem_zalloc(sizeof (zfs_sb_t), KM_SLEEP | KM_NODEBUG); |
9babb374 | 604 | |
428870ff BB |
605 | /* |
606 | * We claim to always be readonly so we can open snapshots; | |
607 | * other ZPL code will prevent us from writing to snapshots. | |
608 | */ | |
3558fd73 | 609 | error = dmu_objset_own(osname, DMU_OST_ZFS, B_TRUE, zsb, &os); |
428870ff | 610 | if (error) { |
3558fd73 | 611 | kmem_free(zsb, sizeof (zfs_sb_t)); |
9babb374 | 612 | return (error); |
428870ff | 613 | } |
9babb374 BB |
614 | |
615 | /* | |
616 | * Initialize the zfs-specific filesystem structure. | |
617 | * Should probably make this a kmem cache, shuffle fields, | |
618 | * and just bzero up to z_hold_mtx[]. | |
619 | */ | |
2cf7f52b | 620 | zsb->z_sb = NULL; |
3558fd73 BB |
621 | zsb->z_parent = zsb; |
622 | zsb->z_max_blksz = SPA_MAXBLOCKSIZE; | |
623 | zsb->z_show_ctldir = ZFS_SNAPDIR_VISIBLE; | |
624 | zsb->z_os = os; | |
9babb374 | 625 | |
3558fd73 | 626 | error = zfs_get_zplprop(os, ZFS_PROP_VERSION, &zsb->z_version); |
9babb374 BB |
627 | if (error) { |
628 | goto out; | |
3558fd73 | 629 | } else if (zsb->z_version > |
428870ff | 630 | zfs_zpl_version_map(spa_version(dmu_objset_spa(os)))) { |
149e873a | 631 | (void) printk("Can't mount a version %lld file system " |
428870ff | 632 | "on a version %lld pool\n. Pool must be upgraded to mount " |
3558fd73 | 633 | "this file system.", (u_longlong_t)zsb->z_version, |
428870ff | 634 | (u_longlong_t)spa_version(dmu_objset_spa(os))); |
9babb374 BB |
635 | error = ENOTSUP; |
636 | goto out; | |
637 | } | |
9babb374 BB |
638 | if ((error = zfs_get_zplprop(os, ZFS_PROP_NORMALIZE, &zval)) != 0) |
639 | goto out; | |
3558fd73 | 640 | zsb->z_norm = (int)zval; |
9babb374 BB |
641 | |
642 | if ((error = zfs_get_zplprop(os, ZFS_PROP_UTF8ONLY, &zval)) != 0) | |
643 | goto out; | |
3558fd73 | 644 | zsb->z_utf8 = (zval != 0); |
9babb374 BB |
645 | |
646 | if ((error = zfs_get_zplprop(os, ZFS_PROP_CASE, &zval)) != 0) | |
647 | goto out; | |
3558fd73 | 648 | zsb->z_case = (uint_t)zval; |
9babb374 BB |
649 | |
650 | /* | |
651 | * Fold case on file systems that are always or sometimes case | |
652 | * insensitive. | |
653 | */ | |
3558fd73 BB |
654 | if (zsb->z_case == ZFS_CASE_INSENSITIVE || |
655 | zsb->z_case == ZFS_CASE_MIXED) | |
656 | zsb->z_norm |= U8_TEXTPREP_TOUPPER; | |
9babb374 | 657 | |
3558fd73 BB |
658 | zsb->z_use_fuids = USE_FUIDS(zsb->z_version, zsb->z_os); |
659 | zsb->z_use_sa = USE_SA(zsb->z_version, zsb->z_os); | |
428870ff | 660 | |
3558fd73 | 661 | if (zsb->z_use_sa) { |
428870ff BB |
662 | /* should either have both of these objects or none */ |
663 | error = zap_lookup(os, MASTER_NODE_OBJ, ZFS_SA_ATTRS, 8, 1, | |
664 | &sa_obj); | |
665 | if (error) | |
591fb62f | 666 | goto out; |
82a37189 BB |
667 | |
668 | error = zfs_get_zplprop(os, ZFS_PROP_XATTR, &zval); | |
669 | if ((error == 0) && (zval == ZFS_XATTR_SA)) | |
670 | zsb->z_xattr_sa = B_TRUE; | |
428870ff BB |
671 | } else { |
672 | /* | |
673 | * Pre SA versions file systems should never touch | |
674 | * either the attribute registration or layout objects. | |
675 | */ | |
676 | sa_obj = 0; | |
677 | } | |
678 | ||
572e2857 | 679 | error = sa_setup(os, sa_obj, zfs_attr_table, ZPL_END, |
3558fd73 | 680 | &zsb->z_attr_table); |
572e2857 BB |
681 | if (error) |
682 | goto out; | |
428870ff | 683 | |
3558fd73 | 684 | if (zsb->z_version >= ZPL_VERSION_SA) |
428870ff | 685 | sa_register_update_callback(os, zfs_sa_upgrade); |
9babb374 BB |
686 | |
687 | error = zap_lookup(os, MASTER_NODE_OBJ, ZFS_ROOT_OBJ, 8, 1, | |
3558fd73 | 688 | &zsb->z_root); |
9babb374 BB |
689 | if (error) |
690 | goto out; | |
3558fd73 | 691 | ASSERT(zsb->z_root != 0); |
9babb374 BB |
692 | |
693 | error = zap_lookup(os, MASTER_NODE_OBJ, ZFS_UNLINKED_SET, 8, 1, | |
3558fd73 | 694 | &zsb->z_unlinkedobj); |
9babb374 BB |
695 | if (error) |
696 | goto out; | |
697 | ||
698 | error = zap_lookup(os, MASTER_NODE_OBJ, | |
699 | zfs_userquota_prop_prefixes[ZFS_PROP_USERQUOTA], | |
3558fd73 | 700 | 8, 1, &zsb->z_userquota_obj); |
9babb374 BB |
701 | if (error && error != ENOENT) |
702 | goto out; | |
703 | ||
704 | error = zap_lookup(os, MASTER_NODE_OBJ, | |
705 | zfs_userquota_prop_prefixes[ZFS_PROP_GROUPQUOTA], | |
3558fd73 | 706 | 8, 1, &zsb->z_groupquota_obj); |
9babb374 BB |
707 | if (error && error != ENOENT) |
708 | goto out; | |
709 | ||
710 | error = zap_lookup(os, MASTER_NODE_OBJ, ZFS_FUID_TABLES, 8, 1, | |
3558fd73 | 711 | &zsb->z_fuid_obj); |
9babb374 BB |
712 | if (error && error != ENOENT) |
713 | goto out; | |
714 | ||
715 | error = zap_lookup(os, MASTER_NODE_OBJ, ZFS_SHARES_DIR, 8, 1, | |
3558fd73 | 716 | &zsb->z_shares_dir); |
9babb374 BB |
717 | if (error && error != ENOENT) |
718 | goto out; | |
719 | ||
3558fd73 BB |
720 | mutex_init(&zsb->z_znodes_lock, NULL, MUTEX_DEFAULT, NULL); |
721 | mutex_init(&zsb->z_lock, NULL, MUTEX_DEFAULT, NULL); | |
722 | list_create(&zsb->z_all_znodes, sizeof (znode_t), | |
9babb374 | 723 | offsetof(znode_t, z_link_node)); |
3558fd73 BB |
724 | rrw_init(&zsb->z_teardown_lock); |
725 | rw_init(&zsb->z_teardown_inactive_lock, NULL, RW_DEFAULT, NULL); | |
726 | rw_init(&zsb->z_fuid_lock, NULL, RW_DEFAULT, NULL); | |
9babb374 | 727 | for (i = 0; i != ZFS_OBJ_MTX_SZ; i++) |
3558fd73 | 728 | mutex_init(&zsb->z_hold_mtx[i], NULL, MUTEX_DEFAULT, NULL); |
9babb374 | 729 | |
ebe7e575 BB |
730 | avl_create(&zsb->z_ctldir_snaps, snapentry_compare, |
731 | sizeof (zfs_snapentry_t), offsetof(zfs_snapentry_t, se_node)); | |
732 | mutex_init(&zsb->z_ctldir_lock, NULL, MUTEX_DEFAULT, NULL); | |
733 | ||
3558fd73 | 734 | *zsbp = zsb; |
9babb374 BB |
735 | return (0); |
736 | ||
737 | out: | |
3558fd73 BB |
738 | dmu_objset_disown(os, zsb); |
739 | *zsbp = NULL; | |
740 | kmem_free(zsb, sizeof (zfs_sb_t)); | |
9babb374 BB |
741 | return (error); |
742 | } | |
86f35f34 | 743 | EXPORT_SYMBOL(zfs_sb_create); |
9babb374 | 744 | |
86f35f34 | 745 | int |
3558fd73 | 746 | zfs_sb_setup(zfs_sb_t *zsb, boolean_t mounting) |
34dc7c2f | 747 | { |
34dc7c2f BB |
748 | int error; |
749 | ||
3558fd73 | 750 | error = zfs_register_callbacks(zsb); |
34dc7c2f BB |
751 | if (error) |
752 | return (error); | |
753 | ||
754 | /* | |
3558fd73 | 755 | * Set the objset user_ptr to track its zsb. |
34dc7c2f | 756 | */ |
3558fd73 BB |
757 | mutex_enter(&zsb->z_os->os_user_ptr_lock); |
758 | dmu_objset_set_user(zsb->z_os, zsb); | |
759 | mutex_exit(&zsb->z_os->os_user_ptr_lock); | |
34dc7c2f | 760 | |
3558fd73 | 761 | zsb->z_log = zil_open(zsb->z_os, zfs_get_data); |
9babb374 | 762 | |
34dc7c2f BB |
763 | /* |
764 | * If we are not mounting (ie: online recv), then we don't | |
765 | * have to worry about replaying the log as we blocked all | |
766 | * operations out since we closed the ZIL. | |
767 | */ | |
768 | if (mounting) { | |
b128c09f BB |
769 | boolean_t readonly; |
770 | ||
34dc7c2f BB |
771 | /* |
772 | * During replay we remove the read only flag to | |
773 | * allow replays to succeed. | |
774 | */ | |
2cf7f52b | 775 | readonly = zfs_is_readonly(zsb); |
fb5f0bc8 | 776 | if (readonly != 0) |
2cf7f52b | 777 | readonly_changed_cb(zsb, B_FALSE); |
fb5f0bc8 | 778 | else |
3558fd73 | 779 | zfs_unlinked_drain(zsb); |
34dc7c2f | 780 | |
428870ff BB |
781 | /* |
782 | * Parse and replay the intent log. | |
783 | * | |
784 | * Because of ziltest, this must be done after | |
785 | * zfs_unlinked_drain(). (Further note: ziltest | |
786 | * doesn't use readonly mounts, where | |
787 | * zfs_unlinked_drain() isn't called.) This is because | |
788 | * ziltest causes spa_sync() to think it's committed, | |
789 | * but actually it is not, so the intent log contains | |
790 | * many txg's worth of changes. | |
791 | * | |
792 | * In particular, if object N is in the unlinked set in | |
793 | * the last txg to actually sync, then it could be | |
794 | * actually freed in a later txg and then reallocated | |
795 | * in a yet later txg. This would write a "create | |
796 | * object N" record to the intent log. Normally, this | |
797 | * would be fine because the spa_sync() would have | |
798 | * written out the fact that object N is free, before | |
799 | * we could write the "create object N" intent log | |
800 | * record. | |
801 | * | |
802 | * But when we are in ziltest mode, we advance the "open | |
803 | * txg" without actually spa_sync()-ing the changes to | |
804 | * disk. So we would see that object N is still | |
805 | * allocated and in the unlinked set, and there is an | |
806 | * intent log record saying to allocate it. | |
807 | */ | |
3558fd73 | 808 | if (spa_writeable(dmu_objset_spa(zsb->z_os))) { |
572e2857 | 809 | if (zil_replay_disable) { |
3558fd73 | 810 | zil_destroy(zsb->z_log, B_FALSE); |
572e2857 | 811 | } else { |
3558fd73 BB |
812 | zsb->z_replay = B_TRUE; |
813 | zil_replay(zsb->z_os, zsb, | |
572e2857 | 814 | zfs_replay_vector); |
3558fd73 | 815 | zsb->z_replay = B_FALSE; |
572e2857 | 816 | } |
fb5f0bc8 | 817 | } |
2cf7f52b BB |
818 | |
819 | /* restore readonly bit */ | |
820 | if (readonly != 0) | |
821 | readonly_changed_cb(zsb, B_TRUE); | |
34dc7c2f BB |
822 | } |
823 | ||
34dc7c2f BB |
824 | return (0); |
825 | } | |
86f35f34 | 826 | EXPORT_SYMBOL(zfs_sb_setup); |
34dc7c2f | 827 | |
9babb374 | 828 | void |
3558fd73 | 829 | zfs_sb_free(zfs_sb_t *zsb) |
34dc7c2f | 830 | { |
9babb374 | 831 | int i; |
9babb374 | 832 | |
3558fd73 | 833 | zfs_fuid_destroy(zsb); |
9babb374 | 834 | |
3558fd73 BB |
835 | mutex_destroy(&zsb->z_znodes_lock); |
836 | mutex_destroy(&zsb->z_lock); | |
837 | list_destroy(&zsb->z_all_znodes); | |
838 | rrw_destroy(&zsb->z_teardown_lock); | |
839 | rw_destroy(&zsb->z_teardown_inactive_lock); | |
840 | rw_destroy(&zsb->z_fuid_lock); | |
9babb374 | 841 | for (i = 0; i != ZFS_OBJ_MTX_SZ; i++) |
3558fd73 | 842 | mutex_destroy(&zsb->z_hold_mtx[i]); |
ebe7e575 BB |
843 | mutex_destroy(&zsb->z_ctldir_lock); |
844 | avl_destroy(&zsb->z_ctldir_snaps); | |
3558fd73 | 845 | kmem_free(zsb, sizeof (zfs_sb_t)); |
34dc7c2f | 846 | } |
86f35f34 | 847 | EXPORT_SYMBOL(zfs_sb_free); |
34dc7c2f | 848 | |
9babb374 | 849 | static void |
3558fd73 | 850 | zfs_set_fuid_feature(zfs_sb_t *zsb) |
9babb374 | 851 | { |
3558fd73 BB |
852 | zsb->z_use_fuids = USE_FUIDS(zsb->z_version, zsb->z_os); |
853 | zsb->z_use_sa = USE_SA(zsb->z_version, zsb->z_os); | |
9babb374 | 854 | } |
34dc7c2f BB |
855 | |
856 | void | |
3558fd73 | 857 | zfs_unregister_callbacks(zfs_sb_t *zsb) |
34dc7c2f | 858 | { |
3558fd73 | 859 | objset_t *os = zsb->z_os; |
34dc7c2f BB |
860 | struct dsl_dataset *ds; |
861 | ||
862 | /* | |
863 | * Unregister properties. | |
864 | */ | |
865 | if (!dmu_objset_is_snapshot(os)) { | |
866 | ds = dmu_objset_ds(os); | |
867 | VERIFY(dsl_prop_unregister(ds, "atime", atime_changed_cb, | |
3558fd73 | 868 | zsb) == 0); |
34dc7c2f BB |
869 | |
870 | VERIFY(dsl_prop_unregister(ds, "xattr", xattr_changed_cb, | |
3558fd73 | 871 | zsb) == 0); |
34dc7c2f BB |
872 | |
873 | VERIFY(dsl_prop_unregister(ds, "recordsize", blksz_changed_cb, | |
3558fd73 | 874 | zsb) == 0); |
34dc7c2f BB |
875 | |
876 | VERIFY(dsl_prop_unregister(ds, "readonly", readonly_changed_cb, | |
3558fd73 | 877 | zsb) == 0); |
34dc7c2f BB |
878 | |
879 | VERIFY(dsl_prop_unregister(ds, "devices", devices_changed_cb, | |
3558fd73 | 880 | zsb) == 0); |
34dc7c2f BB |
881 | |
882 | VERIFY(dsl_prop_unregister(ds, "setuid", setuid_changed_cb, | |
3558fd73 | 883 | zsb) == 0); |
34dc7c2f BB |
884 | |
885 | VERIFY(dsl_prop_unregister(ds, "exec", exec_changed_cb, | |
3558fd73 | 886 | zsb) == 0); |
34dc7c2f BB |
887 | |
888 | VERIFY(dsl_prop_unregister(ds, "snapdir", snapdir_changed_cb, | |
3558fd73 | 889 | zsb) == 0); |
34dc7c2f | 890 | |
34dc7c2f | 891 | VERIFY(dsl_prop_unregister(ds, "aclinherit", |
3558fd73 | 892 | acl_inherit_changed_cb, zsb) == 0); |
34dc7c2f BB |
893 | |
894 | VERIFY(dsl_prop_unregister(ds, "vscan", | |
3558fd73 | 895 | vscan_changed_cb, zsb) == 0); |
2cf7f52b BB |
896 | |
897 | VERIFY(dsl_prop_unregister(ds, "nbmand", | |
898 | nbmand_changed_cb, zsb) == 0); | |
34dc7c2f BB |
899 | } |
900 | } | |
e5c39b95 | 901 | EXPORT_SYMBOL(zfs_unregister_callbacks); |
34dc7c2f | 902 | |
bc3e15e3 | 903 | #ifdef HAVE_MLSLABEL |
428870ff BB |
904 | /* |
905 | * zfs_check_global_label: | |
906 | * Check that the hex label string is appropriate for the dataset | |
907 | * being mounted into the global_zone proper. | |
908 | * | |
909 | * Return an error if the hex label string is not default or | |
910 | * admin_low/admin_high. For admin_low labels, the corresponding | |
911 | * dataset must be readonly. | |
912 | */ | |
913 | int | |
914 | zfs_check_global_label(const char *dsname, const char *hexsl) | |
915 | { | |
916 | if (strcasecmp(hexsl, ZFS_MLSLABEL_DEFAULT) == 0) | |
917 | return (0); | |
918 | if (strcasecmp(hexsl, ADMIN_HIGH) == 0) | |
919 | return (0); | |
920 | if (strcasecmp(hexsl, ADMIN_LOW) == 0) { | |
921 | /* must be readonly */ | |
922 | uint64_t rdonly; | |
923 | ||
924 | if (dsl_prop_get_integer(dsname, | |
925 | zfs_prop_to_name(ZFS_PROP_READONLY), &rdonly, NULL)) | |
926 | return (EACCES); | |
927 | return (rdonly ? 0 : EACCES); | |
928 | } | |
929 | return (EACCES); | |
930 | } | |
86f35f34 | 931 | EXPORT_SYMBOL(zfs_check_global_label); |
bc3e15e3 | 932 | #endif /* HAVE_MLSLABEL */ |
428870ff | 933 | |
e5c39b95 | 934 | int |
3558fd73 | 935 | zfs_statvfs(struct dentry *dentry, struct kstatfs *statp) |
34dc7c2f | 936 | { |
3558fd73 | 937 | zfs_sb_t *zsb = dentry->d_sb->s_fs_info; |
34dc7c2f | 938 | uint64_t refdbytes, availbytes, usedobjs, availobjs; |
04f9432d | 939 | uint64_t fsid; |
3558fd73 | 940 | uint32_t bshift; |
34dc7c2f | 941 | |
3558fd73 | 942 | ZFS_ENTER(zsb); |
34dc7c2f | 943 | |
3558fd73 | 944 | dmu_objset_space(zsb->z_os, |
34dc7c2f BB |
945 | &refdbytes, &availbytes, &usedobjs, &availobjs); |
946 | ||
04f9432d | 947 | fsid = dmu_objset_fsid_guid(zsb->z_os); |
34dc7c2f | 948 | /* |
05ff35c6 BB |
949 | * The underlying storage pool actually uses multiple block |
950 | * size. Under Solaris frsize (fragment size) is reported as | |
951 | * the smallest block size we support, and bsize (block size) | |
952 | * as the filesystem's maximum block size. Unfortunately, | |
953 | * under Linux the fragment size and block size are often used | |
954 | * interchangeably. Thus we are forced to report both of them | |
955 | * as the filesystem's maximum block size. | |
34dc7c2f | 956 | */ |
05ff35c6 | 957 | statp->f_frsize = zsb->z_max_blksz; |
3558fd73 BB |
958 | statp->f_bsize = zsb->z_max_blksz; |
959 | bshift = fls(statp->f_bsize) - 1; | |
34dc7c2f BB |
960 | |
961 | /* | |
3558fd73 BB |
962 | * The following report "total" blocks of various kinds in |
963 | * the file system, but reported in terms of f_bsize - the | |
964 | * "preferred" size. | |
34dc7c2f BB |
965 | */ |
966 | ||
3558fd73 BB |
967 | statp->f_blocks = (refdbytes + availbytes) >> bshift; |
968 | statp->f_bfree = availbytes >> bshift; | |
34dc7c2f BB |
969 | statp->f_bavail = statp->f_bfree; /* no root reservation */ |
970 | ||
971 | /* | |
972 | * statvfs() should really be called statufs(), because it assumes | |
973 | * static metadata. ZFS doesn't preallocate files, so the best | |
974 | * we can do is report the max that could possibly fit in f_files, | |
975 | * and that minus the number actually used in f_ffree. | |
976 | * For f_ffree, report the smaller of the number of object available | |
977 | * and the number of blocks (each object will take at least a block). | |
978 | */ | |
baab0630 | 979 | statp->f_ffree = MIN(availobjs, availbytes >> DNODE_SHIFT); |
34dc7c2f | 980 | statp->f_files = statp->f_ffree + usedobjs; |
04f9432d CP |
981 | statp->f_fsid.val[0] = (uint32_t)fsid; |
982 | statp->f_fsid.val[1] = (uint32_t)(fsid >> 32); | |
3558fd73 BB |
983 | statp->f_type = ZFS_SUPER_MAGIC; |
984 | statp->f_namelen = ZFS_MAXNAMELEN; | |
34dc7c2f BB |
985 | |
986 | /* | |
3558fd73 | 987 | * We have all of 40 characters to stuff a string here. |
34dc7c2f BB |
988 | * Is there anything useful we could/should provide? |
989 | */ | |
3558fd73 | 990 | bzero(statp->f_spare, sizeof (statp->f_spare)); |
34dc7c2f | 991 | |
3558fd73 | 992 | ZFS_EXIT(zsb); |
34dc7c2f BB |
993 | return (0); |
994 | } | |
e5c39b95 | 995 | EXPORT_SYMBOL(zfs_statvfs); |
34dc7c2f | 996 | |
e5c39b95 | 997 | int |
3558fd73 | 998 | zfs_root(zfs_sb_t *zsb, struct inode **ipp) |
34dc7c2f | 999 | { |
34dc7c2f BB |
1000 | znode_t *rootzp; |
1001 | int error; | |
1002 | ||
3558fd73 | 1003 | ZFS_ENTER(zsb); |
34dc7c2f | 1004 | |
3558fd73 | 1005 | error = zfs_zget(zsb, zsb->z_root, &rootzp); |
34dc7c2f | 1006 | if (error == 0) |
3558fd73 | 1007 | *ipp = ZTOI(rootzp); |
34dc7c2f | 1008 | |
3558fd73 | 1009 | ZFS_EXIT(zsb); |
34dc7c2f BB |
1010 | return (error); |
1011 | } | |
e5c39b95 | 1012 | EXPORT_SYMBOL(zfs_root); |
34dc7c2f | 1013 | |
ab26409d BB |
1014 | #ifdef HAVE_SHRINK |
1015 | int | |
1016 | zfs_sb_prune(struct super_block *sb, unsigned long nr_to_scan, int *objects) | |
1017 | { | |
1018 | zfs_sb_t *zsb = sb->s_fs_info; | |
1019 | struct shrinker *shrinker = &sb->s_shrink; | |
1020 | struct shrink_control sc = { | |
1021 | .nr_to_scan = nr_to_scan, | |
1022 | .gfp_mask = GFP_KERNEL, | |
1023 | }; | |
1024 | ||
1025 | ZFS_ENTER(zsb); | |
1026 | *objects = (*shrinker->shrink)(shrinker, &sc); | |
1027 | ZFS_EXIT(zsb); | |
1028 | ||
1029 | return (0); | |
1030 | } | |
1031 | EXPORT_SYMBOL(zfs_sb_prune); | |
1032 | #endif /* HAVE_SHRINK */ | |
1033 | ||
34dc7c2f | 1034 | /* |
7b3e34ba | 1035 | * Teardown the zfs_sb_t. |
34dc7c2f BB |
1036 | * |
1037 | * Note, if 'unmounting' if FALSE, we return with the 'z_teardown_lock' | |
1038 | * and 'z_teardown_inactive_lock' held. | |
1039 | */ | |
3558fd73 | 1040 | int |
86f35f34 | 1041 | zfs_sb_teardown(zfs_sb_t *zsb, boolean_t unmounting) |
34dc7c2f BB |
1042 | { |
1043 | znode_t *zp; | |
1044 | ||
3558fd73 | 1045 | rrw_enter(&zsb->z_teardown_lock, RW_WRITER, FTAG); |
34dc7c2f BB |
1046 | |
1047 | if (!unmounting) { | |
1048 | /* | |
ceb43b93 BB |
1049 | * We purge the parent filesystem's super block as the |
1050 | * parent filesystem and all of its snapshots have their | |
1051 | * inode's super block set to the parent's filesystem's | |
1052 | * super block. Note, 'z_parent' is self referential | |
1053 | * for non-snapshots. | |
34dc7c2f | 1054 | */ |
ceb43b93 | 1055 | shrink_dcache_sb(zsb->z_parent->z_sb); |
34dc7c2f BB |
1056 | } |
1057 | ||
7973e464 BB |
1058 | /* |
1059 | * Drain the iput_taskq to ensure all active references to the | |
1060 | * zfs_sb_t have been handled only then can it be safely destroyed. | |
1061 | */ | |
1062 | taskq_wait(dsl_pool_iput_taskq(dmu_objset_pool(zsb->z_os))); | |
1063 | ||
34dc7c2f BB |
1064 | /* |
1065 | * Close the zil. NB: Can't close the zil while zfs_inactive | |
1066 | * threads are blocked as zil_close can call zfs_inactive. | |
1067 | */ | |
3558fd73 BB |
1068 | if (zsb->z_log) { |
1069 | zil_close(zsb->z_log); | |
1070 | zsb->z_log = NULL; | |
34dc7c2f BB |
1071 | } |
1072 | ||
3558fd73 | 1073 | rw_enter(&zsb->z_teardown_inactive_lock, RW_WRITER); |
34dc7c2f BB |
1074 | |
1075 | /* | |
1076 | * If we are not unmounting (ie: online recv) and someone already | |
1077 | * unmounted this file system while we were doing the switcheroo, | |
1078 | * or a reopen of z_os failed then just bail out now. | |
1079 | */ | |
3558fd73 BB |
1080 | if (!unmounting && (zsb->z_unmounted || zsb->z_os == NULL)) { |
1081 | rw_exit(&zsb->z_teardown_inactive_lock); | |
1082 | rrw_exit(&zsb->z_teardown_lock, FTAG); | |
34dc7c2f BB |
1083 | return (EIO); |
1084 | } | |
1085 | ||
1086 | /* | |
7b3e34ba BB |
1087 | * At this point there are no VFS ops active, and any new VFS ops |
1088 | * will fail with EIO since we have z_teardown_lock for writer (only | |
1089 | * relevant for forced unmount). | |
34dc7c2f BB |
1090 | * |
1091 | * Release all holds on dbufs. | |
1092 | */ | |
3558fd73 BB |
1093 | mutex_enter(&zsb->z_znodes_lock); |
1094 | for (zp = list_head(&zsb->z_all_znodes); zp != NULL; | |
7b3e34ba | 1095 | zp = list_next(&zsb->z_all_znodes, zp)) { |
428870ff | 1096 | if (zp->z_sa_hdl) { |
3558fd73 | 1097 | ASSERT(atomic_read(&ZTOI(zp)->i_count) > 0); |
34dc7c2f BB |
1098 | zfs_znode_dmu_fini(zp); |
1099 | } | |
7b3e34ba | 1100 | } |
3558fd73 | 1101 | mutex_exit(&zsb->z_znodes_lock); |
34dc7c2f BB |
1102 | |
1103 | /* | |
7b3e34ba | 1104 | * If we are unmounting, set the unmounted flag and let new VFS ops |
34dc7c2f | 1105 | * unblock. zfs_inactive will have the unmounted behavior, and all |
7b3e34ba | 1106 | * other VFS ops will fail with EIO. |
34dc7c2f BB |
1107 | */ |
1108 | if (unmounting) { | |
3558fd73 BB |
1109 | zsb->z_unmounted = B_TRUE; |
1110 | rrw_exit(&zsb->z_teardown_lock, FTAG); | |
1111 | rw_exit(&zsb->z_teardown_inactive_lock); | |
34dc7c2f BB |
1112 | } |
1113 | ||
1114 | /* | |
1115 | * z_os will be NULL if there was an error in attempting to reopen | |
3558fd73 BB |
1116 | * zsb, so just return as the properties had already been |
1117 | * | |
34dc7c2f BB |
1118 | * unregistered and cached data had been evicted before. |
1119 | */ | |
3558fd73 | 1120 | if (zsb->z_os == NULL) |
34dc7c2f BB |
1121 | return (0); |
1122 | ||
1123 | /* | |
1124 | * Unregister properties. | |
1125 | */ | |
3558fd73 | 1126 | zfs_unregister_callbacks(zsb); |
34dc7c2f BB |
1127 | |
1128 | /* | |
1129 | * Evict cached data | |
1130 | */ | |
04434775 MA |
1131 | if (dsl_dataset_is_dirty(dmu_objset_ds(zsb->z_os)) && |
1132 | !zfs_is_readonly(zsb)) | |
1133 | txg_wait_synced(dmu_objset_pool(zsb->z_os), 0); | |
3558fd73 | 1134 | (void) dmu_objset_evict_dbufs(zsb->z_os); |
34dc7c2f BB |
1135 | |
1136 | return (0); | |
1137 | } | |
86f35f34 | 1138 | EXPORT_SYMBOL(zfs_sb_teardown); |
34dc7c2f | 1139 | |
5547c2f1 BB |
1140 | #if defined(HAVE_BDI) && !defined(HAVE_BDI_SETUP_AND_REGISTER) |
1141 | atomic_long_t zfs_bdi_seq = ATOMIC_LONG_INIT(0); | |
1142 | #endif /* HAVE_BDI && !HAVE_BDI_SETUP_AND_REGISTER */ | |
76659dc1 | 1143 | |
e5c39b95 | 1144 | int |
3558fd73 | 1145 | zfs_domount(struct super_block *sb, void *data, int silent) |
34dc7c2f | 1146 | { |
3558fd73 BB |
1147 | zpl_mount_data_t *zmd = data; |
1148 | const char *osname = zmd->z_osname; | |
1149 | zfs_sb_t *zsb; | |
1150 | struct inode *root_inode; | |
1151 | uint64_t recordsize; | |
1152 | int error; | |
34dc7c2f | 1153 | |
3558fd73 BB |
1154 | error = zfs_sb_create(osname, &zsb); |
1155 | if (error) | |
1156 | return (error); | |
1157 | ||
1158 | if ((error = dsl_prop_get_integer(osname, "recordsize", | |
1159 | &recordsize, NULL))) | |
1160 | goto out; | |
1161 | ||
1162 | zsb->z_sb = sb; | |
3558fd73 BB |
1163 | sb->s_fs_info = zsb; |
1164 | sb->s_magic = ZFS_SUPER_MAGIC; | |
1165 | sb->s_maxbytes = MAX_LFS_FILESIZE; | |
1166 | sb->s_time_gran = 1; | |
1167 | sb->s_blocksize = recordsize; | |
1168 | sb->s_blocksize_bits = ilog2(recordsize); | |
5547c2f1 BB |
1169 | |
1170 | #ifdef HAVE_BDI | |
1171 | /* | |
1172 | * 2.6.32 API change, | |
1173 | * Added backing_device_info (BDI) per super block interfaces. A BDI | |
1174 | * must be configured when using a non-device backed filesystem for | |
1175 | * proper writeback. This is not required for older pdflush kernels. | |
1176 | * | |
1177 | * NOTE: Linux read-ahead is disabled in favor of zfs read-ahead. | |
1178 | */ | |
1179 | zsb->z_bdi.ra_pages = 0; | |
1180 | sb->s_bdi = &zsb->z_bdi; | |
1181 | ||
1182 | error = -bdi_setup_and_register(&zsb->z_bdi, "zfs", BDI_CAP_MAP_COPY); | |
1183 | if (error) | |
1184 | goto out; | |
1185 | #endif /* HAVE_BDI */ | |
3558fd73 BB |
1186 | |
1187 | /* Set callback operations for the file system. */ | |
1188 | sb->s_op = &zpl_super_operations; | |
1189 | sb->s_xattr = zpl_xattr_handlers; | |
055656d4 | 1190 | sb->s_export_op = &zpl_export_operations; |
ee930353 BB |
1191 | #ifdef HAVE_S_D_OP |
1192 | sb->s_d_op = &zpl_dentry_operations; | |
1193 | #endif /* HAVE_S_D_OP */ | |
3558fd73 BB |
1194 | |
1195 | /* Set features for file system. */ | |
1196 | zfs_set_fuid_feature(zsb); | |
1197 | ||
1198 | if (dmu_objset_is_snapshot(zsb->z_os)) { | |
1199 | uint64_t pval; | |
1200 | ||
1201 | atime_changed_cb(zsb, B_FALSE); | |
1202 | readonly_changed_cb(zsb, B_TRUE); | |
1203 | if ((error = dsl_prop_get_integer(osname,"xattr",&pval,NULL))) | |
1204 | goto out; | |
1205 | xattr_changed_cb(zsb, pval); | |
1206 | zsb->z_issnap = B_TRUE; | |
1207 | zsb->z_os->os_sync = ZFS_SYNC_DISABLED; | |
1208 | ||
1209 | mutex_enter(&zsb->z_os->os_user_ptr_lock); | |
1210 | dmu_objset_set_user(zsb->z_os, zsb); | |
1211 | mutex_exit(&zsb->z_os->os_user_ptr_lock); | |
1212 | } else { | |
1213 | error = zfs_sb_setup(zsb, B_TRUE); | |
34dc7c2f BB |
1214 | } |
1215 | ||
3558fd73 BB |
1216 | /* Allocate a root inode for the filesystem. */ |
1217 | error = zfs_root(zsb, &root_inode); | |
1218 | if (error) { | |
1219 | (void) zfs_umount(sb); | |
1220 | goto out; | |
34dc7c2f BB |
1221 | } |
1222 | ||
3558fd73 | 1223 | /* Allocate a root dentry for the filesystem */ |
6a0936ba | 1224 | sb->s_root = d_make_root(root_inode); |
3558fd73 BB |
1225 | if (sb->s_root == NULL) { |
1226 | (void) zfs_umount(sb); | |
1227 | error = ENOMEM; | |
1228 | goto out; | |
1229 | } | |
ebe7e575 BB |
1230 | |
1231 | if (!zsb->z_issnap) | |
1232 | zfsctl_create(zsb); | |
3558fd73 BB |
1233 | out: |
1234 | if (error) { | |
1235 | dmu_objset_disown(zsb->z_os, zsb); | |
1236 | zfs_sb_free(zsb); | |
1237 | } | |
34dc7c2f | 1238 | |
3558fd73 BB |
1239 | return (error); |
1240 | } | |
1241 | EXPORT_SYMBOL(zfs_domount); | |
1242 | ||
ebe7e575 BB |
1243 | /* |
1244 | * Called when an unmount is requested and certain sanity checks have | |
1245 | * already passed. At this point no dentries or inodes have been reclaimed | |
1246 | * from their respective caches. We drop the extra reference on the .zfs | |
1247 | * control directory to allow everything to be reclaimed. All snapshots | |
1248 | * must already have been unmounted to reach this point. | |
1249 | */ | |
1250 | void | |
1251 | zfs_preumount(struct super_block *sb) | |
1252 | { | |
1253 | zfs_sb_t *zsb = sb->s_fs_info; | |
1254 | ||
1f0d8a56 | 1255 | if (zsb != NULL && zsb->z_ctldir != NULL) |
ebe7e575 BB |
1256 | zfsctl_destroy(zsb); |
1257 | } | |
1258 | EXPORT_SYMBOL(zfs_preumount); | |
1259 | ||
1260 | /* | |
1261 | * Called once all other unmount released tear down has occurred. | |
1262 | * It is our responsibility to release any remaining infrastructure. | |
1263 | */ | |
3558fd73 BB |
1264 | /*ARGSUSED*/ |
1265 | int | |
1266 | zfs_umount(struct super_block *sb) | |
1267 | { | |
1268 | zfs_sb_t *zsb = sb->s_fs_info; | |
1269 | objset_t *os; | |
1270 | ||
86f35f34 | 1271 | VERIFY(zfs_sb_teardown(zsb, B_TRUE) == 0); |
3558fd73 | 1272 | os = zsb->z_os; |
34dc7c2f | 1273 | |
5547c2f1 BB |
1274 | #ifdef HAVE_BDI |
1275 | bdi_destroy(sb->s_bdi); | |
1276 | #endif /* HAVE_BDI */ | |
76659dc1 | 1277 | |
34dc7c2f BB |
1278 | /* |
1279 | * z_os will be NULL if there was an error in | |
3558fd73 | 1280 | * attempting to reopen zsb. |
34dc7c2f BB |
1281 | */ |
1282 | if (os != NULL) { | |
1283 | /* | |
1284 | * Unset the objset user_ptr. | |
1285 | */ | |
428870ff | 1286 | mutex_enter(&os->os_user_ptr_lock); |
34dc7c2f | 1287 | dmu_objset_set_user(os, NULL); |
428870ff | 1288 | mutex_exit(&os->os_user_ptr_lock); |
34dc7c2f BB |
1289 | |
1290 | /* | |
b128c09f | 1291 | * Finally release the objset |
34dc7c2f | 1292 | */ |
3558fd73 | 1293 | dmu_objset_disown(os, zsb); |
34dc7c2f BB |
1294 | } |
1295 | ||
3558fd73 | 1296 | zfs_sb_free(zsb); |
34dc7c2f BB |
1297 | return (0); |
1298 | } | |
e5c39b95 | 1299 | EXPORT_SYMBOL(zfs_umount); |
34dc7c2f | 1300 | |
0de19dad BB |
1301 | int |
1302 | zfs_remount(struct super_block *sb, int *flags, char *data) | |
1303 | { | |
0de19dad | 1304 | /* |
2cf7f52b BB |
1305 | * All namespace flags (MNT_*) and super block flags (MS_*) will |
1306 | * be handled by the Linux VFS. Only handle custom options here. | |
0de19dad | 1307 | */ |
0de19dad BB |
1308 | return (0); |
1309 | } | |
1310 | EXPORT_SYMBOL(zfs_remount); | |
1311 | ||
e5c39b95 | 1312 | int |
2cf7f52b | 1313 | zfs_vget(struct super_block *sb, struct inode **ipp, fid_t *fidp) |
34dc7c2f | 1314 | { |
2cf7f52b | 1315 | zfs_sb_t *zsb = sb->s_fs_info; |
34dc7c2f BB |
1316 | znode_t *zp; |
1317 | uint64_t object = 0; | |
1318 | uint64_t fid_gen = 0; | |
1319 | uint64_t gen_mask; | |
1320 | uint64_t zp_gen; | |
3558fd73 | 1321 | int i, err; |
34dc7c2f | 1322 | |
3558fd73 | 1323 | *ipp = NULL; |
34dc7c2f | 1324 | |
3558fd73 | 1325 | ZFS_ENTER(zsb); |
34dc7c2f BB |
1326 | |
1327 | if (fidp->fid_len == LONG_FID_LEN) { | |
1328 | zfid_long_t *zlfid = (zfid_long_t *)fidp; | |
1329 | uint64_t objsetid = 0; | |
1330 | uint64_t setgen = 0; | |
1331 | ||
1332 | for (i = 0; i < sizeof (zlfid->zf_setid); i++) | |
1333 | objsetid |= ((uint64_t)zlfid->zf_setid[i]) << (8 * i); | |
1334 | ||
1335 | for (i = 0; i < sizeof (zlfid->zf_setgen); i++) | |
1336 | setgen |= ((uint64_t)zlfid->zf_setgen[i]) << (8 * i); | |
1337 | ||
3558fd73 | 1338 | ZFS_EXIT(zsb); |
34dc7c2f | 1339 | |
ebe7e575 | 1340 | err = zfsctl_lookup_objset(sb, objsetid, &zsb); |
34dc7c2f BB |
1341 | if (err) |
1342 | return (EINVAL); | |
ebe7e575 | 1343 | |
3558fd73 | 1344 | ZFS_ENTER(zsb); |
34dc7c2f BB |
1345 | } |
1346 | ||
1347 | if (fidp->fid_len == SHORT_FID_LEN || fidp->fid_len == LONG_FID_LEN) { | |
1348 | zfid_short_t *zfid = (zfid_short_t *)fidp; | |
1349 | ||
1350 | for (i = 0; i < sizeof (zfid->zf_object); i++) | |
1351 | object |= ((uint64_t)zfid->zf_object[i]) << (8 * i); | |
1352 | ||
1353 | for (i = 0; i < sizeof (zfid->zf_gen); i++) | |
1354 | fid_gen |= ((uint64_t)zfid->zf_gen[i]) << (8 * i); | |
1355 | } else { | |
3558fd73 | 1356 | ZFS_EXIT(zsb); |
34dc7c2f BB |
1357 | return (EINVAL); |
1358 | } | |
1359 | ||
1360 | /* A zero fid_gen means we are in the .zfs control directories */ | |
1361 | if (fid_gen == 0 && | |
1362 | (object == ZFSCTL_INO_ROOT || object == ZFSCTL_INO_SNAPDIR)) { | |
3558fd73 BB |
1363 | *ipp = zsb->z_ctldir; |
1364 | ASSERT(*ipp != NULL); | |
34dc7c2f | 1365 | if (object == ZFSCTL_INO_SNAPDIR) { |
ebe7e575 BB |
1366 | VERIFY(zfsctl_root_lookup(*ipp, "snapshot", ipp, |
1367 | 0, kcred, NULL, NULL) == 0); | |
34dc7c2f | 1368 | } else { |
3558fd73 | 1369 | igrab(*ipp); |
34dc7c2f | 1370 | } |
3558fd73 | 1371 | ZFS_EXIT(zsb); |
34dc7c2f BB |
1372 | return (0); |
1373 | } | |
1374 | ||
1375 | gen_mask = -1ULL >> (64 - 8 * i); | |
1376 | ||
1377 | dprintf("getting %llu [%u mask %llx]\n", object, fid_gen, gen_mask); | |
3558fd73 BB |
1378 | if ((err = zfs_zget(zsb, object, &zp))) { |
1379 | ZFS_EXIT(zsb); | |
34dc7c2f BB |
1380 | return (err); |
1381 | } | |
3558fd73 | 1382 | (void) sa_lookup(zp->z_sa_hdl, SA_ZPL_GEN(zsb), &zp_gen, |
428870ff BB |
1383 | sizeof (uint64_t)); |
1384 | zp_gen = zp_gen & gen_mask; | |
34dc7c2f BB |
1385 | if (zp_gen == 0) |
1386 | zp_gen = 1; | |
1387 | if (zp->z_unlinked || zp_gen != fid_gen) { | |
1388 | dprintf("znode gen (%u) != fid gen (%u)\n", zp_gen, fid_gen); | |
3558fd73 BB |
1389 | iput(ZTOI(zp)); |
1390 | ZFS_EXIT(zsb); | |
34dc7c2f BB |
1391 | return (EINVAL); |
1392 | } | |
1393 | ||
3558fd73 BB |
1394 | *ipp = ZTOI(zp); |
1395 | if (*ipp) | |
1396 | zfs_inode_update(ITOZ(*ipp)); | |
960e08fe | 1397 | |
3558fd73 | 1398 | ZFS_EXIT(zsb); |
34dc7c2f BB |
1399 | return (0); |
1400 | } | |
e5c39b95 | 1401 | EXPORT_SYMBOL(zfs_vget); |
34dc7c2f BB |
1402 | |
1403 | /* | |
7b3e34ba | 1404 | * Block out VFS ops and close zfs_sb_t |
34dc7c2f BB |
1405 | * |
1406 | * Note, if successful, then we return with the 'z_teardown_lock' and | |
1407 | * 'z_teardown_inactive_lock' write held. | |
1408 | */ | |
1409 | int | |
3558fd73 | 1410 | zfs_suspend_fs(zfs_sb_t *zsb) |
34dc7c2f BB |
1411 | { |
1412 | int error; | |
1413 | ||
86f35f34 | 1414 | if ((error = zfs_sb_teardown(zsb, B_FALSE)) != 0) |
34dc7c2f | 1415 | return (error); |
7b3e34ba | 1416 | |
3558fd73 | 1417 | dmu_objset_disown(zsb->z_os, zsb); |
34dc7c2f BB |
1418 | |
1419 | return (0); | |
1420 | } | |
e5c39b95 | 1421 | EXPORT_SYMBOL(zfs_suspend_fs); |
34dc7c2f BB |
1422 | |
1423 | /* | |
7b3e34ba | 1424 | * Reopen zfs_sb_t and release VFS ops. |
34dc7c2f BB |
1425 | */ |
1426 | int | |
3558fd73 | 1427 | zfs_resume_fs(zfs_sb_t *zsb, const char *osname) |
34dc7c2f | 1428 | { |
428870ff | 1429 | int err, err2; |
34dc7c2f | 1430 | |
3558fd73 BB |
1431 | ASSERT(RRW_WRITE_HELD(&zsb->z_teardown_lock)); |
1432 | ASSERT(RW_WRITE_HELD(&zsb->z_teardown_inactive_lock)); | |
34dc7c2f | 1433 | |
3558fd73 | 1434 | err = dmu_objset_own(osname, DMU_OST_ZFS, B_FALSE, zsb, &zsb->z_os); |
34dc7c2f | 1435 | if (err) { |
3558fd73 | 1436 | zsb->z_os = NULL; |
34dc7c2f BB |
1437 | } else { |
1438 | znode_t *zp; | |
428870ff BB |
1439 | uint64_t sa_obj = 0; |
1440 | ||
3558fd73 | 1441 | err2 = zap_lookup(zsb->z_os, MASTER_NODE_OBJ, |
428870ff BB |
1442 | ZFS_SA_ATTRS, 8, 1, &sa_obj); |
1443 | ||
3558fd73 | 1444 | if ((err || err2) && zsb->z_version >= ZPL_VERSION_SA) |
428870ff BB |
1445 | goto bail; |
1446 | ||
1447 | ||
3558fd73 BB |
1448 | if ((err = sa_setup(zsb->z_os, sa_obj, |
1449 | zfs_attr_table, ZPL_END, &zsb->z_attr_table)) != 0) | |
572e2857 | 1450 | goto bail; |
34dc7c2f | 1451 | |
3558fd73 | 1452 | VERIFY(zfs_sb_setup(zsb, B_FALSE) == 0); |
7b3e34ba | 1453 | zsb->z_rollback_time = jiffies; |
34dc7c2f BB |
1454 | |
1455 | /* | |
7b3e34ba BB |
1456 | * Attempt to re-establish all the active inodes with their |
1457 | * dbufs. If a zfs_rezget() fails, then we unhash the inode | |
1458 | * and mark it stale. This prevents a collision if a new | |
1459 | * inode/object is created which must use the same inode | |
1460 | * number. The stale inode will be be released when the | |
1461 | * VFS prunes the dentry holding the remaining references | |
1462 | * on the stale inode. | |
34dc7c2f | 1463 | */ |
3558fd73 BB |
1464 | mutex_enter(&zsb->z_znodes_lock); |
1465 | for (zp = list_head(&zsb->z_all_znodes); zp; | |
1466 | zp = list_next(&zsb->z_all_znodes, zp)) { | |
7b3e34ba BB |
1467 | err2 = zfs_rezget(zp); |
1468 | if (err2) { | |
1469 | remove_inode_hash(ZTOI(zp)); | |
1470 | zp->z_is_stale = B_TRUE; | |
1471 | } | |
34dc7c2f | 1472 | } |
3558fd73 | 1473 | mutex_exit(&zsb->z_znodes_lock); |
34dc7c2f BB |
1474 | } |
1475 | ||
428870ff | 1476 | bail: |
7b3e34ba | 1477 | /* release the VFS ops */ |
3558fd73 BB |
1478 | rw_exit(&zsb->z_teardown_inactive_lock); |
1479 | rrw_exit(&zsb->z_teardown_lock, FTAG); | |
34dc7c2f BB |
1480 | |
1481 | if (err) { | |
1482 | /* | |
7b3e34ba | 1483 | * Since we couldn't reopen zfs_sb_t, force |
34dc7c2f BB |
1484 | * unmount this file system. |
1485 | */ | |
3558fd73 | 1486 | (void) zfs_umount(zsb->z_sb); |
34dc7c2f BB |
1487 | } |
1488 | return (err); | |
1489 | } | |
e5c39b95 | 1490 | EXPORT_SYMBOL(zfs_resume_fs); |
34dc7c2f | 1491 | |
34dc7c2f | 1492 | int |
3558fd73 | 1493 | zfs_set_version(zfs_sb_t *zsb, uint64_t newvers) |
34dc7c2f BB |
1494 | { |
1495 | int error; | |
3558fd73 | 1496 | objset_t *os = zsb->z_os; |
34dc7c2f | 1497 | dmu_tx_t *tx; |
34dc7c2f BB |
1498 | |
1499 | if (newvers < ZPL_VERSION_INITIAL || newvers > ZPL_VERSION) | |
1500 | return (EINVAL); | |
1501 | ||
3558fd73 | 1502 | if (newvers < zsb->z_version) |
9babb374 | 1503 | return (EINVAL); |
34dc7c2f | 1504 | |
428870ff | 1505 | if (zfs_spa_version_map(newvers) > |
3558fd73 | 1506 | spa_version(dmu_objset_spa(zsb->z_os))) |
428870ff BB |
1507 | return (ENOTSUP); |
1508 | ||
34dc7c2f | 1509 | tx = dmu_tx_create(os); |
9babb374 | 1510 | dmu_tx_hold_zap(tx, MASTER_NODE_OBJ, B_FALSE, ZPL_VERSION_STR); |
3558fd73 | 1511 | if (newvers >= ZPL_VERSION_SA && !zsb->z_use_sa) { |
428870ff BB |
1512 | dmu_tx_hold_zap(tx, MASTER_NODE_OBJ, B_TRUE, |
1513 | ZFS_SA_ATTRS); | |
1514 | dmu_tx_hold_zap(tx, DMU_NEW_OBJECT, FALSE, NULL); | |
1515 | } | |
34dc7c2f BB |
1516 | error = dmu_tx_assign(tx, TXG_WAIT); |
1517 | if (error) { | |
1518 | dmu_tx_abort(tx); | |
9babb374 BB |
1519 | return (error); |
1520 | } | |
428870ff | 1521 | |
9babb374 BB |
1522 | error = zap_update(os, MASTER_NODE_OBJ, ZPL_VERSION_STR, |
1523 | 8, 1, &newvers, tx); | |
1524 | ||
1525 | if (error) { | |
1526 | dmu_tx_commit(tx); | |
1527 | return (error); | |
34dc7c2f | 1528 | } |
34dc7c2f | 1529 | |
3558fd73 | 1530 | if (newvers >= ZPL_VERSION_SA && !zsb->z_use_sa) { |
428870ff BB |
1531 | uint64_t sa_obj; |
1532 | ||
3558fd73 | 1533 | ASSERT3U(spa_version(dmu_objset_spa(zsb->z_os)), >=, |
428870ff BB |
1534 | SPA_VERSION_SA); |
1535 | sa_obj = zap_create(os, DMU_OT_SA_MASTER_NODE, | |
1536 | DMU_OT_NONE, 0, tx); | |
1537 | ||
1538 | error = zap_add(os, MASTER_NODE_OBJ, | |
1539 | ZFS_SA_ATTRS, 8, 1, &sa_obj, tx); | |
c99c9001 | 1540 | ASSERT0(error); |
428870ff BB |
1541 | |
1542 | VERIFY(0 == sa_set_sa_object(os, sa_obj)); | |
1543 | sa_register_update_callback(os, zfs_sa_upgrade); | |
1544 | } | |
1545 | ||
1546 | spa_history_log_internal(LOG_DS_UPGRADE, | |
1547 | dmu_objset_spa(os), tx, "oldver=%llu newver=%llu dataset = %llu", | |
3558fd73 | 1548 | zsb->z_version, newvers, dmu_objset_id(os)); |
9babb374 | 1549 | |
34dc7c2f BB |
1550 | dmu_tx_commit(tx); |
1551 | ||
3558fd73 | 1552 | zsb->z_version = newvers; |
9babb374 | 1553 | |
3558fd73 BB |
1554 | if (zsb->z_version >= ZPL_VERSION_FUID) |
1555 | zfs_set_fuid_feature(zsb); | |
9babb374 BB |
1556 | |
1557 | return (0); | |
34dc7c2f | 1558 | } |
e5c39b95 | 1559 | EXPORT_SYMBOL(zfs_set_version); |
34dc7c2f BB |
1560 | |
1561 | /* | |
1562 | * Read a property stored within the master node. | |
1563 | */ | |
1564 | int | |
1565 | zfs_get_zplprop(objset_t *os, zfs_prop_t prop, uint64_t *value) | |
1566 | { | |
1567 | const char *pname; | |
b128c09f | 1568 | int error = ENOENT; |
34dc7c2f BB |
1569 | |
1570 | /* | |
1571 | * Look up the file system's value for the property. For the | |
1572 | * version property, we look up a slightly different string. | |
1573 | */ | |
1574 | if (prop == ZFS_PROP_VERSION) | |
1575 | pname = ZPL_VERSION_STR; | |
1576 | else | |
1577 | pname = zfs_prop_to_name(prop); | |
1578 | ||
b128c09f BB |
1579 | if (os != NULL) |
1580 | error = zap_lookup(os, MASTER_NODE_OBJ, pname, 8, 1, value); | |
34dc7c2f BB |
1581 | |
1582 | if (error == ENOENT) { | |
1583 | /* No value set, use the default value */ | |
1584 | switch (prop) { | |
1585 | case ZFS_PROP_VERSION: | |
1586 | *value = ZPL_VERSION; | |
1587 | break; | |
1588 | case ZFS_PROP_NORMALIZE: | |
1589 | case ZFS_PROP_UTF8ONLY: | |
1590 | *value = 0; | |
1591 | break; | |
1592 | case ZFS_PROP_CASE: | |
1593 | *value = ZFS_CASE_SENSITIVE; | |
1594 | break; | |
1595 | default: | |
1596 | return (error); | |
1597 | } | |
1598 | error = 0; | |
1599 | } | |
1600 | return (error); | |
1601 | } | |
86f35f34 | 1602 | EXPORT_SYMBOL(zfs_get_zplprop); |
3558fd73 BB |
1603 | |
1604 | void | |
1605 | zfs_init(void) | |
1606 | { | |
ebe7e575 | 1607 | zfsctl_init(); |
3558fd73 BB |
1608 | zfs_znode_init(); |
1609 | dmu_objset_register_type(DMU_OST_ZFS, zfs_space_delta_cb); | |
1610 | register_filesystem(&zpl_fs_type); | |
ab26409d | 1611 | (void) arc_add_prune_callback(zpl_prune_sbs, NULL); |
3558fd73 BB |
1612 | } |
1613 | ||
1614 | void | |
1615 | zfs_fini(void) | |
1616 | { | |
1617 | unregister_filesystem(&zpl_fs_type); | |
1618 | zfs_znode_fini(); | |
ebe7e575 | 1619 | zfsctl_fini(); |
3558fd73 | 1620 | } |