]> git.proxmox.com Git - mirror_zfs.git/blame - module/zfs/zfs_vfsops.c
Change delimiter for ZED email scripts
[mirror_zfs.git] / module / zfs / zfs_vfsops.c
CommitLineData
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.
2e528b49 23 * Copyright (c) 2013 by Delphix. All rights reserved.
34dc7c2f
BB
24 */
25
428870ff
BB
26/* Portions Copyright 2010 Robert Milkowski */
27
34dc7c2f
BB
28#include <sys/types.h>
29#include <sys/param.h>
30#include <sys/systm.h>
31#include <sys/sysmacros.h>
32#include <sys/kmem.h>
33#include <sys/pathname.h>
34#include <sys/vnode.h>
35#include <sys/vfs.h>
36#include <sys/vfs_opreg.h>
37#include <sys/mntent.h>
38#include <sys/mount.h>
39#include <sys/cmn_err.h>
40#include "fs/fs_subr.h"
41#include <sys/zfs_znode.h>
3558fd73 42#include <sys/zfs_vnops.h>
34dc7c2f
BB
43#include <sys/zfs_dir.h>
44#include <sys/zil.h>
45#include <sys/fs/zfs.h>
46#include <sys/dmu.h>
47#include <sys/dsl_prop.h>
48#include <sys/dsl_dataset.h>
49#include <sys/dsl_deleg.h>
50#include <sys/spa.h>
51#include <sys/zap.h>
428870ff 52#include <sys/sa.h>
a94addd9 53#include <sys/sa_impl.h>
34dc7c2f
BB
54#include <sys/varargs.h>
55#include <sys/policy.h>
56#include <sys/atomic.h>
57#include <sys/mkdev.h>
58#include <sys/modctl.h>
59#include <sys/refstr.h>
60#include <sys/zfs_ioctl.h>
ebe7e575 61#include <sys/zfs_ctldir.h>
34dc7c2f
BB
62#include <sys/zfs_fuid.h>
63#include <sys/bootconf.h>
64#include <sys/sunddi.h>
65#include <sys/dnlc.h>
66#include <sys/dmu_objset.h>
67#include <sys/spa_boot.h>
3558fd73 68#include <sys/zpl.h>
428870ff 69#include "zfs_comutil.h"
34dc7c2f 70
9babb374 71
34dc7c2f
BB
72/*ARGSUSED*/
73int
03f9ba9d 74zfs_sync(struct super_block *sb, int wait, cred_t *cr)
34dc7c2f 75{
03f9ba9d
BB
76 zfs_sb_t *zsb = sb->s_fs_info;
77
34dc7c2f
BB
78 /*
79 * Data integrity is job one. We don't want a compromised kernel
80 * writing to the storage pool, so we never sync during panic.
81 */
d5e53f9d 82 if (unlikely(oops_in_progress))
34dc7c2f
BB
83 return (0);
84
03f9ba9d
BB
85 /*
86 * Semantically, the only requirement is that the sync be initiated.
87 * The DMU syncs out txgs frequently, so there's nothing to do.
88 */
89 if (!wait)
90 return (0);
91
3558fd73 92 if (zsb != NULL) {
34dc7c2f
BB
93 /*
94 * Sync a specific filesystem.
95 */
9babb374 96 dsl_pool_t *dp;
34dc7c2f 97
3558fd73
BB
98 ZFS_ENTER(zsb);
99 dp = dmu_objset_pool(zsb->z_os);
9babb374
BB
100
101 /*
102 * If the system is shutting down, then skip any
103 * filesystems which may exist on a suspended pool.
104 */
03f9ba9d 105 if (spa_suspended(dp->dp_spa)) {
3558fd73 106 ZFS_EXIT(zsb);
9babb374
BB
107 return (0);
108 }
109
3558fd73
BB
110 if (zsb->z_log != NULL)
111 zil_commit(zsb->z_log, 0);
428870ff 112
3558fd73 113 ZFS_EXIT(zsb);
34dc7c2f
BB
114 } else {
115 /*
116 * Sync all ZFS filesystems. This is what happens when you
117 * run sync(1M). Unlike other filesystems, ZFS honors the
118 * request by waiting for all pools to commit all dirty data.
119 */
120 spa_sync_allpools();
121 }
122
123 return (0);
124}
e5c39b95 125EXPORT_SYMBOL(zfs_sync);
34dc7c2f 126
2cf7f52b
BB
127boolean_t
128zfs_is_readonly(zfs_sb_t *zsb)
129{
130 return (!!(zsb->z_sb->s_flags & MS_RDONLY));
131}
132EXPORT_SYMBOL(zfs_is_readonly);
133
34dc7c2f
BB
134static void
135atime_changed_cb(void *arg, uint64_t newval)
136{
2cf7f52b 137 ((zfs_sb_t *)arg)->z_atime = newval;
34dc7c2f
BB
138}
139
6d111134
TC
140static void
141relatime_changed_cb(void *arg, uint64_t newval)
142{
143 ((zfs_sb_t *)arg)->z_relatime = newval;
144}
145
34dc7c2f
BB
146static void
147xattr_changed_cb(void *arg, uint64_t newval)
148{
3558fd73 149 zfs_sb_t *zsb = arg;
34dc7c2f 150
82a37189 151 if (newval == ZFS_XATTR_OFF) {
2cf7f52b 152 zsb->z_flags &= ~ZSB_XATTR;
82a37189
BB
153 } else {
154 zsb->z_flags |= ZSB_XATTR;
155
156 if (newval == ZFS_XATTR_SA)
157 zsb->z_xattr_sa = B_TRUE;
158 else
159 zsb->z_xattr_sa = B_FALSE;
160 }
34dc7c2f
BB
161}
162
023699cd
MM
163static void
164acltype_changed_cb(void *arg, uint64_t newval)
165{
166 zfs_sb_t *zsb = arg;
167
168 switch (newval) {
169 case ZFS_ACLTYPE_OFF:
170 zsb->z_acl_type = ZFS_ACLTYPE_OFF;
171 zsb->z_sb->s_flags &= ~MS_POSIXACL;
172 break;
173 case ZFS_ACLTYPE_POSIXACL:
b695c34e 174#ifdef CONFIG_FS_POSIX_ACL
023699cd
MM
175 zsb->z_acl_type = ZFS_ACLTYPE_POSIXACL;
176 zsb->z_sb->s_flags |= MS_POSIXACL;
b695c34e
MM
177#else
178 zsb->z_acl_type = ZFS_ACLTYPE_OFF;
179 zsb->z_sb->s_flags &= ~MS_POSIXACL;
180#endif /* CONFIG_FS_POSIX_ACL */
023699cd
MM
181 break;
182 default:
183 break;
184 }
185}
186
34dc7c2f
BB
187static void
188blksz_changed_cb(void *arg, uint64_t newval)
189{
3558fd73 190 zfs_sb_t *zsb = arg;
34dc7c2f
BB
191
192 if (newval < SPA_MINBLOCKSIZE ||
193 newval > SPA_MAXBLOCKSIZE || !ISP2(newval))
194 newval = SPA_MAXBLOCKSIZE;
195
3558fd73 196 zsb->z_max_blksz = newval;
34dc7c2f
BB
197}
198
199static void
200readonly_changed_cb(void *arg, uint64_t newval)
201{
3558fd73
BB
202 zfs_sb_t *zsb = arg;
203 struct super_block *sb = zsb->z_sb;
34dc7c2f 204
2cf7f52b
BB
205 if (sb == NULL)
206 return;
207
208 if (newval)
3558fd73 209 sb->s_flags |= MS_RDONLY;
2cf7f52b 210 else
3558fd73 211 sb->s_flags &= ~MS_RDONLY;
34dc7c2f
BB
212}
213
214static void
215devices_changed_cb(void *arg, uint64_t newval)
216{
34dc7c2f
BB
217}
218
219static void
220setuid_changed_cb(void *arg, uint64_t newval)
221{
34dc7c2f
BB
222}
223
224static void
225exec_changed_cb(void *arg, uint64_t newval)
226{
34dc7c2f
BB
227}
228
34dc7c2f
BB
229static void
230nbmand_changed_cb(void *arg, uint64_t newval)
231{
3558fd73
BB
232 zfs_sb_t *zsb = arg;
233 struct super_block *sb = zsb->z_sb;
234
2cf7f52b
BB
235 if (sb == NULL)
236 return;
237
238 if (newval == TRUE)
3558fd73 239 sb->s_flags |= MS_MANDLOCK;
2cf7f52b 240 else
3558fd73 241 sb->s_flags &= ~MS_MANDLOCK;
34dc7c2f
BB
242}
243
244static void
245snapdir_changed_cb(void *arg, uint64_t newval)
246{
3558fd73 247 ((zfs_sb_t *)arg)->z_show_ctldir = newval;
34dc7c2f
BB
248}
249
250static void
251vscan_changed_cb(void *arg, uint64_t newval)
252{
3558fd73 253 ((zfs_sb_t *)arg)->z_vscan = newval;
34dc7c2f
BB
254}
255
34dc7c2f
BB
256static void
257acl_inherit_changed_cb(void *arg, uint64_t newval)
258{
3558fd73 259 ((zfs_sb_t *)arg)->z_acl_inherit = newval;
34dc7c2f
BB
260}
261
e5c39b95 262int
3558fd73 263zfs_register_callbacks(zfs_sb_t *zsb)
34dc7c2f
BB
264{
265 struct dsl_dataset *ds = NULL;
3558fd73 266 objset_t *os = zsb->z_os;
c9ada6d5 267 boolean_t do_readonly = B_FALSE;
34dc7c2f
BB
268 int error = 0;
269
2cf7f52b 270 if (zfs_is_readonly(zsb) || !spa_writeable(dmu_objset_spa(os)))
c9ada6d5 271 do_readonly = B_TRUE;
34dc7c2f
BB
272
273 /*
274 * Register property callbacks.
275 *
276 * It would probably be fine to just check for i/o error from
277 * the first prop_register(), but I guess I like to go
278 * overboard...
279 */
280 ds = dmu_objset_ds(os);
13fe0198 281 dsl_pool_config_enter(dmu_objset_pool(os), FTAG);
3558fd73 282 error = dsl_prop_register(ds,
13fe0198 283 zfs_prop_to_name(ZFS_PROP_ATIME), atime_changed_cb, zsb);
98fad862 284 error = error ? error : dsl_prop_register(ds,
6d111134 285 zfs_prop_to_name(ZFS_PROP_RELATIME), relatime_changed_cb, zsb);
34dc7c2f 286 error = error ? error : dsl_prop_register(ds,
13fe0198 287 zfs_prop_to_name(ZFS_PROP_XATTR), xattr_changed_cb, zsb);
34dc7c2f 288 error = error ? error : dsl_prop_register(ds,
13fe0198 289 zfs_prop_to_name(ZFS_PROP_RECORDSIZE), blksz_changed_cb, zsb);
34dc7c2f 290 error = error ? error : dsl_prop_register(ds,
13fe0198 291 zfs_prop_to_name(ZFS_PROP_READONLY), readonly_changed_cb, zsb);
34dc7c2f 292 error = error ? error : dsl_prop_register(ds,
13fe0198 293 zfs_prop_to_name(ZFS_PROP_DEVICES), devices_changed_cb, zsb);
34dc7c2f 294 error = error ? error : dsl_prop_register(ds,
13fe0198 295 zfs_prop_to_name(ZFS_PROP_SETUID), setuid_changed_cb, zsb);
34dc7c2f 296 error = error ? error : dsl_prop_register(ds,
13fe0198 297 zfs_prop_to_name(ZFS_PROP_EXEC), exec_changed_cb, zsb);
34dc7c2f 298 error = error ? error : dsl_prop_register(ds,
13fe0198 299 zfs_prop_to_name(ZFS_PROP_SNAPDIR), snapdir_changed_cb, zsb);
34dc7c2f 300 error = error ? error : dsl_prop_register(ds,
023699cd
MM
301 zfs_prop_to_name(ZFS_PROP_ACLTYPE), acltype_changed_cb, zsb);
302 error = error ? error : dsl_prop_register(ds,
303 zfs_prop_to_name(ZFS_PROP_ACLINHERIT), acl_inherit_changed_cb, zsb);
34dc7c2f 304 error = error ? error : dsl_prop_register(ds,
13fe0198 305 zfs_prop_to_name(ZFS_PROP_VSCAN), vscan_changed_cb, zsb);
2cf7f52b 306 error = error ? error : dsl_prop_register(ds,
13fe0198
MA
307 zfs_prop_to_name(ZFS_PROP_NBMAND), nbmand_changed_cb, zsb);
308 dsl_pool_config_exit(dmu_objset_pool(os), FTAG);
34dc7c2f
BB
309 if (error)
310 goto unregister;
311
c9ada6d5
BB
312 if (do_readonly)
313 readonly_changed_cb(zsb, B_TRUE);
314
34dc7c2f
BB
315 return (0);
316
317unregister:
318 /*
319 * We may attempt to unregister some callbacks that are not
320 * registered, but this is OK; it will simply return ENOMSG,
321 * which we will ignore.
322 */
13fe0198
MA
323 (void) dsl_prop_unregister(ds, zfs_prop_to_name(ZFS_PROP_ATIME),
324 atime_changed_cb, zsb);
6d111134
TC
325 (void) dsl_prop_unregister(ds, zfs_prop_to_name(ZFS_PROP_RELATIME),
326 relatime_changed_cb, zsb);
13fe0198
MA
327 (void) dsl_prop_unregister(ds, zfs_prop_to_name(ZFS_PROP_XATTR),
328 xattr_changed_cb, zsb);
329 (void) dsl_prop_unregister(ds, zfs_prop_to_name(ZFS_PROP_RECORDSIZE),
330 blksz_changed_cb, zsb);
331 (void) dsl_prop_unregister(ds, zfs_prop_to_name(ZFS_PROP_READONLY),
332 readonly_changed_cb, zsb);
333 (void) dsl_prop_unregister(ds, zfs_prop_to_name(ZFS_PROP_DEVICES),
334 devices_changed_cb, zsb);
335 (void) dsl_prop_unregister(ds, zfs_prop_to_name(ZFS_PROP_SETUID),
336 setuid_changed_cb, zsb);
337 (void) dsl_prop_unregister(ds, zfs_prop_to_name(ZFS_PROP_EXEC),
338 exec_changed_cb, zsb);
339 (void) dsl_prop_unregister(ds, zfs_prop_to_name(ZFS_PROP_SNAPDIR),
340 snapdir_changed_cb, zsb);
023699cd
MM
341 (void) dsl_prop_unregister(ds, zfs_prop_to_name(ZFS_PROP_ACLTYPE),
342 acltype_changed_cb, zsb);
13fe0198
MA
343 (void) dsl_prop_unregister(ds, zfs_prop_to_name(ZFS_PROP_ACLINHERIT),
344 acl_inherit_changed_cb, zsb);
345 (void) dsl_prop_unregister(ds, zfs_prop_to_name(ZFS_PROP_VSCAN),
346 vscan_changed_cb, zsb);
347 (void) dsl_prop_unregister(ds, zfs_prop_to_name(ZFS_PROP_NBMAND),
348 nbmand_changed_cb, zsb);
34dc7c2f 349
3558fd73 350 return (error);
34dc7c2f 351}
e5c39b95 352EXPORT_SYMBOL(zfs_register_callbacks);
34dc7c2f 353
428870ff
BB
354static int
355zfs_space_delta_cb(dmu_object_type_t bonustype, void *data,
356 uint64_t *userp, uint64_t *groupp)
9babb374 357{
428870ff
BB
358 /*
359 * Is it a valid type of object to track?
360 */
361 if (bonustype != DMU_OT_ZNODE && bonustype != DMU_OT_SA)
2e528b49 362 return (SET_ERROR(ENOENT));
9babb374 363
428870ff
BB
364 /*
365 * If we have a NULL data pointer
366 * then assume the id's aren't changing and
367 * return EEXIST to the dmu to let it know to
368 * use the same ids
369 */
370 if (data == NULL)
2e528b49 371 return (SET_ERROR(EEXIST));
9babb374 372
428870ff 373 if (bonustype == DMU_OT_ZNODE) {
a94addd9 374 znode_phys_t *znp = data;
428870ff
BB
375 *userp = znp->zp_uid;
376 *groupp = znp->zp_gid;
9babb374 377 } else {
428870ff 378 int hdrsize;
a94addd9
MA
379 sa_hdr_phys_t *sap = data;
380 sa_hdr_phys_t sa = *sap;
381 boolean_t swap = B_FALSE;
9babb374 382
428870ff 383 ASSERT(bonustype == DMU_OT_SA);
428870ff 384
a94addd9 385 if (sa.sa_magic == 0) {
428870ff
BB
386 /*
387 * This should only happen for newly created
388 * files that haven't had the znode data filled
389 * in yet.
390 */
391 *userp = 0;
392 *groupp = 0;
a94addd9
MA
393 return (0);
394 }
395 if (sa.sa_magic == BSWAP_32(SA_MAGIC)) {
396 sa.sa_magic = SA_MAGIC;
397 sa.sa_layout_info = BSWAP_16(sa.sa_layout_info);
398 swap = B_TRUE;
399 } else {
400 VERIFY3U(sa.sa_magic, ==, SA_MAGIC);
401 }
402
403 hdrsize = sa_hdrsize(&sa);
404 VERIFY3U(hdrsize, >=, sizeof (sa_hdr_phys_t));
405 *userp = *((uint64_t *)((uintptr_t)data + hdrsize +
406 SA_UID_OFFSET));
407 *groupp = *((uint64_t *)((uintptr_t)data + hdrsize +
408 SA_GID_OFFSET));
409 if (swap) {
410 *userp = BSWAP_64(*userp);
411 *groupp = BSWAP_64(*groupp);
428870ff 412 }
9babb374 413 }
13fe0198 414 return (0);
9babb374
BB
415}
416
417static void
3558fd73 418fuidstr_to_sid(zfs_sb_t *zsb, const char *fuidstr,
9babb374
BB
419 char *domainbuf, int buflen, uid_t *ridp)
420{
9babb374
BB
421 uint64_t fuid;
422 const char *domain;
423
424 fuid = strtonum(fuidstr, NULL);
425
3558fd73 426 domain = zfs_fuid_find_by_idx(zsb, FUID_INDEX(fuid));
9babb374
BB
427 if (domain)
428 (void) strlcpy(domainbuf, domain, buflen);
429 else
430 domainbuf[0] = '\0';
431 *ridp = FUID_RID(fuid);
432}
433
434static uint64_t
3558fd73 435zfs_userquota_prop_to_obj(zfs_sb_t *zsb, zfs_userquota_prop_t type)
9babb374
BB
436{
437 switch (type) {
438 case ZFS_PROP_USERUSED:
439 return (DMU_USERUSED_OBJECT);
440 case ZFS_PROP_GROUPUSED:
441 return (DMU_GROUPUSED_OBJECT);
442 case ZFS_PROP_USERQUOTA:
3558fd73 443 return (zsb->z_userquota_obj);
9babb374 444 case ZFS_PROP_GROUPQUOTA:
3558fd73 445 return (zsb->z_groupquota_obj);
149e873a 446 default:
2e528b49 447 return (SET_ERROR(ENOTSUP));
9babb374
BB
448 }
449 return (0);
450}
451
452int
3558fd73 453zfs_userspace_many(zfs_sb_t *zsb, zfs_userquota_prop_t type,
9babb374
BB
454 uint64_t *cookiep, void *vbuf, uint64_t *bufsizep)
455{
456 int error;
457 zap_cursor_t zc;
458 zap_attribute_t za;
459 zfs_useracct_t *buf = vbuf;
460 uint64_t obj;
461
3558fd73 462 if (!dmu_objset_userspace_present(zsb->z_os))
2e528b49 463 return (SET_ERROR(ENOTSUP));
9babb374 464
3558fd73 465 obj = zfs_userquota_prop_to_obj(zsb, type);
9babb374
BB
466 if (obj == 0) {
467 *bufsizep = 0;
468 return (0);
469 }
470
3558fd73 471 for (zap_cursor_init_serialized(&zc, zsb->z_os, obj, *cookiep);
9babb374
BB
472 (error = zap_cursor_retrieve(&zc, &za)) == 0;
473 zap_cursor_advance(&zc)) {
474 if ((uintptr_t)buf - (uintptr_t)vbuf + sizeof (zfs_useracct_t) >
475 *bufsizep)
476 break;
477
3558fd73 478 fuidstr_to_sid(zsb, za.za_name,
9babb374
BB
479 buf->zu_domain, sizeof (buf->zu_domain), &buf->zu_rid);
480
481 buf->zu_space = za.za_first_integer;
482 buf++;
483 }
484 if (error == ENOENT)
485 error = 0;
486
487 ASSERT3U((uintptr_t)buf - (uintptr_t)vbuf, <=, *bufsizep);
488 *bufsizep = (uintptr_t)buf - (uintptr_t)vbuf;
489 *cookiep = zap_cursor_serialize(&zc);
490 zap_cursor_fini(&zc);
491 return (error);
492}
e5c39b95 493EXPORT_SYMBOL(zfs_userspace_many);
9babb374
BB
494
495/*
496 * buf must be big enough (eg, 32 bytes)
497 */
498static int
3558fd73 499id_to_fuidstr(zfs_sb_t *zsb, const char *domain, uid_t rid,
9babb374
BB
500 char *buf, boolean_t addok)
501{
502 uint64_t fuid;
503 int domainid = 0;
504
505 if (domain && domain[0]) {
3558fd73 506 domainid = zfs_fuid_find_by_domain(zsb, domain, NULL, addok);
9babb374 507 if (domainid == -1)
2e528b49 508 return (SET_ERROR(ENOENT));
9babb374
BB
509 }
510 fuid = FUID_ENCODE(domainid, rid);
511 (void) sprintf(buf, "%llx", (longlong_t)fuid);
512 return (0);
513}
514
515int
3558fd73 516zfs_userspace_one(zfs_sb_t *zsb, zfs_userquota_prop_t type,
9babb374
BB
517 const char *domain, uint64_t rid, uint64_t *valp)
518{
519 char buf[32];
520 int err;
521 uint64_t obj;
522
523 *valp = 0;
524
3558fd73 525 if (!dmu_objset_userspace_present(zsb->z_os))
2e528b49 526 return (SET_ERROR(ENOTSUP));
9babb374 527
3558fd73 528 obj = zfs_userquota_prop_to_obj(zsb, type);
9babb374
BB
529 if (obj == 0)
530 return (0);
531
3558fd73 532 err = id_to_fuidstr(zsb, domain, rid, buf, B_FALSE);
9babb374
BB
533 if (err)
534 return (err);
535
3558fd73 536 err = zap_lookup(zsb->z_os, obj, buf, 8, 1, valp);
9babb374
BB
537 if (err == ENOENT)
538 err = 0;
539 return (err);
540}
e5c39b95 541EXPORT_SYMBOL(zfs_userspace_one);
9babb374
BB
542
543int
3558fd73 544zfs_set_userquota(zfs_sb_t *zsb, zfs_userquota_prop_t type,
9babb374
BB
545 const char *domain, uint64_t rid, uint64_t quota)
546{
547 char buf[32];
548 int err;
549 dmu_tx_t *tx;
550 uint64_t *objp;
551 boolean_t fuid_dirtied;
552
553 if (type != ZFS_PROP_USERQUOTA && type != ZFS_PROP_GROUPQUOTA)
2e528b49 554 return (SET_ERROR(EINVAL));
9babb374 555
3558fd73 556 if (zsb->z_version < ZPL_VERSION_USERSPACE)
2e528b49 557 return (SET_ERROR(ENOTSUP));
9babb374 558
3558fd73
BB
559 objp = (type == ZFS_PROP_USERQUOTA) ? &zsb->z_userquota_obj :
560 &zsb->z_groupquota_obj;
9babb374 561
3558fd73 562 err = id_to_fuidstr(zsb, domain, rid, buf, B_TRUE);
9babb374
BB
563 if (err)
564 return (err);
3558fd73 565 fuid_dirtied = zsb->z_fuid_dirty;
9babb374 566
3558fd73 567 tx = dmu_tx_create(zsb->z_os);
9babb374
BB
568 dmu_tx_hold_zap(tx, *objp ? *objp : DMU_NEW_OBJECT, B_TRUE, NULL);
569 if (*objp == 0) {
570 dmu_tx_hold_zap(tx, MASTER_NODE_OBJ, B_TRUE,
571 zfs_userquota_prop_prefixes[type]);
572 }
573 if (fuid_dirtied)
3558fd73 574 zfs_fuid_txhold(zsb, tx);
9babb374
BB
575 err = dmu_tx_assign(tx, TXG_WAIT);
576 if (err) {
577 dmu_tx_abort(tx);
578 return (err);
579 }
580
3558fd73 581 mutex_enter(&zsb->z_lock);
9babb374 582 if (*objp == 0) {
3558fd73 583 *objp = zap_create(zsb->z_os, DMU_OT_USERGROUP_QUOTA,
9babb374 584 DMU_OT_NONE, 0, tx);
3558fd73 585 VERIFY(0 == zap_add(zsb->z_os, MASTER_NODE_OBJ,
9babb374
BB
586 zfs_userquota_prop_prefixes[type], 8, 1, objp, tx));
587 }
3558fd73 588 mutex_exit(&zsb->z_lock);
9babb374
BB
589
590 if (quota == 0) {
3558fd73 591 err = zap_remove(zsb->z_os, *objp, buf, tx);
9babb374
BB
592 if (err == ENOENT)
593 err = 0;
594 } else {
3558fd73 595 err = zap_update(zsb->z_os, *objp, buf, 8, 1, &quota, tx);
9babb374
BB
596 }
597 ASSERT(err == 0);
598 if (fuid_dirtied)
3558fd73 599 zfs_fuid_sync(zsb, tx);
9babb374
BB
600 dmu_tx_commit(tx);
601 return (err);
602}
e5c39b95 603EXPORT_SYMBOL(zfs_set_userquota);
9babb374
BB
604
605boolean_t
3558fd73 606zfs_fuid_overquota(zfs_sb_t *zsb, boolean_t isgroup, uint64_t fuid)
9babb374
BB
607{
608 char buf[32];
609 uint64_t used, quota, usedobj, quotaobj;
610 int err;
611
612 usedobj = isgroup ? DMU_GROUPUSED_OBJECT : DMU_USERUSED_OBJECT;
3558fd73 613 quotaobj = isgroup ? zsb->z_groupquota_obj : zsb->z_userquota_obj;
9babb374 614
3558fd73 615 if (quotaobj == 0 || zsb->z_replay)
9babb374
BB
616 return (B_FALSE);
617
618 (void) sprintf(buf, "%llx", (longlong_t)fuid);
3558fd73 619 err = zap_lookup(zsb->z_os, quotaobj, buf, 8, 1, &quota);
9babb374
BB
620 if (err != 0)
621 return (B_FALSE);
622
3558fd73 623 err = zap_lookup(zsb->z_os, usedobj, buf, 8, 1, &used);
9babb374
BB
624 if (err != 0)
625 return (B_FALSE);
626 return (used >= quota);
627}
e5c39b95 628EXPORT_SYMBOL(zfs_fuid_overquota);
9babb374 629
428870ff 630boolean_t
3558fd73 631zfs_owner_overquota(zfs_sb_t *zsb, znode_t *zp, boolean_t isgroup)
428870ff
BB
632{
633 uint64_t fuid;
634 uint64_t quotaobj;
428870ff 635
3558fd73 636 quotaobj = isgroup ? zsb->z_groupquota_obj : zsb->z_userquota_obj;
428870ff 637
572e2857 638 fuid = isgroup ? zp->z_gid : zp->z_uid;
428870ff 639
3558fd73 640 if (quotaobj == 0 || zsb->z_replay)
428870ff
BB
641 return (B_FALSE);
642
3558fd73 643 return (zfs_fuid_overquota(zsb, isgroup, fuid));
428870ff 644}
e5c39b95 645EXPORT_SYMBOL(zfs_owner_overquota);
428870ff 646
9babb374 647int
3558fd73 648zfs_sb_create(const char *osname, zfs_sb_t **zsbp)
9babb374
BB
649{
650 objset_t *os;
3558fd73 651 zfs_sb_t *zsb;
9babb374
BB
652 uint64_t zval;
653 int i, error;
428870ff 654 uint64_t sa_obj;
9babb374 655
bafc4e9e 656 zsb = kmem_zalloc(sizeof (zfs_sb_t), KM_SLEEP | KM_NODEBUG);
9babb374 657
428870ff
BB
658 /*
659 * We claim to always be readonly so we can open snapshots;
660 * other ZPL code will prevent us from writing to snapshots.
661 */
3558fd73 662 error = dmu_objset_own(osname, DMU_OST_ZFS, B_TRUE, zsb, &os);
428870ff 663 if (error) {
3558fd73 664 kmem_free(zsb, sizeof (zfs_sb_t));
9babb374 665 return (error);
428870ff 666 }
9babb374
BB
667
668 /*
669 * Initialize the zfs-specific filesystem structure.
670 * Should probably make this a kmem cache, shuffle fields,
671 * and just bzero up to z_hold_mtx[].
672 */
2cf7f52b 673 zsb->z_sb = NULL;
3558fd73
BB
674 zsb->z_parent = zsb;
675 zsb->z_max_blksz = SPA_MAXBLOCKSIZE;
676 zsb->z_show_ctldir = ZFS_SNAPDIR_VISIBLE;
677 zsb->z_os = os;
9babb374 678
3558fd73 679 error = zfs_get_zplprop(os, ZFS_PROP_VERSION, &zsb->z_version);
9babb374
BB
680 if (error) {
681 goto out;
3558fd73 682 } else if (zsb->z_version >
428870ff 683 zfs_zpl_version_map(spa_version(dmu_objset_spa(os)))) {
149e873a 684 (void) printk("Can't mount a version %lld file system "
428870ff 685 "on a version %lld pool\n. Pool must be upgraded to mount "
3558fd73 686 "this file system.", (u_longlong_t)zsb->z_version,
428870ff 687 (u_longlong_t)spa_version(dmu_objset_spa(os)));
2e528b49 688 error = SET_ERROR(ENOTSUP);
9babb374
BB
689 goto out;
690 }
9babb374
BB
691 if ((error = zfs_get_zplprop(os, ZFS_PROP_NORMALIZE, &zval)) != 0)
692 goto out;
3558fd73 693 zsb->z_norm = (int)zval;
9babb374
BB
694
695 if ((error = zfs_get_zplprop(os, ZFS_PROP_UTF8ONLY, &zval)) != 0)
696 goto out;
3558fd73 697 zsb->z_utf8 = (zval != 0);
9babb374
BB
698
699 if ((error = zfs_get_zplprop(os, ZFS_PROP_CASE, &zval)) != 0)
700 goto out;
3558fd73 701 zsb->z_case = (uint_t)zval;
9babb374 702
023699cd
MM
703 if ((error = zfs_get_zplprop(os, ZFS_PROP_ACLTYPE, &zval)) != 0)
704 goto out;
705 zsb->z_acl_type = (uint_t)zval;
706
9babb374
BB
707 /*
708 * Fold case on file systems that are always or sometimes case
709 * insensitive.
710 */
3558fd73
BB
711 if (zsb->z_case == ZFS_CASE_INSENSITIVE ||
712 zsb->z_case == ZFS_CASE_MIXED)
713 zsb->z_norm |= U8_TEXTPREP_TOUPPER;
9babb374 714
3558fd73
BB
715 zsb->z_use_fuids = USE_FUIDS(zsb->z_version, zsb->z_os);
716 zsb->z_use_sa = USE_SA(zsb->z_version, zsb->z_os);
428870ff 717
3558fd73 718 if (zsb->z_use_sa) {
428870ff
BB
719 /* should either have both of these objects or none */
720 error = zap_lookup(os, MASTER_NODE_OBJ, ZFS_SA_ATTRS, 8, 1,
721 &sa_obj);
722 if (error)
591fb62f 723 goto out;
82a37189
BB
724
725 error = zfs_get_zplprop(os, ZFS_PROP_XATTR, &zval);
726 if ((error == 0) && (zval == ZFS_XATTR_SA))
727 zsb->z_xattr_sa = B_TRUE;
428870ff
BB
728 } else {
729 /*
730 * Pre SA versions file systems should never touch
731 * either the attribute registration or layout objects.
732 */
733 sa_obj = 0;
734 }
735
572e2857 736 error = sa_setup(os, sa_obj, zfs_attr_table, ZPL_END,
3558fd73 737 &zsb->z_attr_table);
572e2857
BB
738 if (error)
739 goto out;
428870ff 740
3558fd73 741 if (zsb->z_version >= ZPL_VERSION_SA)
428870ff 742 sa_register_update_callback(os, zfs_sa_upgrade);
9babb374
BB
743
744 error = zap_lookup(os, MASTER_NODE_OBJ, ZFS_ROOT_OBJ, 8, 1,
3558fd73 745 &zsb->z_root);
9babb374
BB
746 if (error)
747 goto out;
3558fd73 748 ASSERT(zsb->z_root != 0);
9babb374
BB
749
750 error = zap_lookup(os, MASTER_NODE_OBJ, ZFS_UNLINKED_SET, 8, 1,
3558fd73 751 &zsb->z_unlinkedobj);
9babb374
BB
752 if (error)
753 goto out;
754
755 error = zap_lookup(os, MASTER_NODE_OBJ,
756 zfs_userquota_prop_prefixes[ZFS_PROP_USERQUOTA],
3558fd73 757 8, 1, &zsb->z_userquota_obj);
9babb374
BB
758 if (error && error != ENOENT)
759 goto out;
760
761 error = zap_lookup(os, MASTER_NODE_OBJ,
762 zfs_userquota_prop_prefixes[ZFS_PROP_GROUPQUOTA],
3558fd73 763 8, 1, &zsb->z_groupquota_obj);
9babb374
BB
764 if (error && error != ENOENT)
765 goto out;
766
767 error = zap_lookup(os, MASTER_NODE_OBJ, ZFS_FUID_TABLES, 8, 1,
3558fd73 768 &zsb->z_fuid_obj);
9babb374
BB
769 if (error && error != ENOENT)
770 goto out;
771
772 error = zap_lookup(os, MASTER_NODE_OBJ, ZFS_SHARES_DIR, 8, 1,
3558fd73 773 &zsb->z_shares_dir);
9babb374
BB
774 if (error && error != ENOENT)
775 goto out;
776
3558fd73
BB
777 mutex_init(&zsb->z_znodes_lock, NULL, MUTEX_DEFAULT, NULL);
778 mutex_init(&zsb->z_lock, NULL, MUTEX_DEFAULT, NULL);
779 list_create(&zsb->z_all_znodes, sizeof (znode_t),
9babb374 780 offsetof(znode_t, z_link_node));
13fe0198 781 rrw_init(&zsb->z_teardown_lock, B_FALSE);
3558fd73
BB
782 rw_init(&zsb->z_teardown_inactive_lock, NULL, RW_DEFAULT, NULL);
783 rw_init(&zsb->z_fuid_lock, NULL, RW_DEFAULT, NULL);
9babb374 784 for (i = 0; i != ZFS_OBJ_MTX_SZ; i++)
3558fd73 785 mutex_init(&zsb->z_hold_mtx[i], NULL, MUTEX_DEFAULT, NULL);
9babb374 786
ebe7e575
BB
787 avl_create(&zsb->z_ctldir_snaps, snapentry_compare,
788 sizeof (zfs_snapentry_t), offsetof(zfs_snapentry_t, se_node));
789 mutex_init(&zsb->z_ctldir_lock, NULL, MUTEX_DEFAULT, NULL);
790
3558fd73 791 *zsbp = zsb;
9babb374
BB
792 return (0);
793
794out:
3558fd73
BB
795 dmu_objset_disown(os, zsb);
796 *zsbp = NULL;
797 kmem_free(zsb, sizeof (zfs_sb_t));
9babb374
BB
798 return (error);
799}
86f35f34 800EXPORT_SYMBOL(zfs_sb_create);
9babb374 801
86f35f34 802int
3558fd73 803zfs_sb_setup(zfs_sb_t *zsb, boolean_t mounting)
34dc7c2f 804{
34dc7c2f
BB
805 int error;
806
3558fd73 807 error = zfs_register_callbacks(zsb);
34dc7c2f
BB
808 if (error)
809 return (error);
810
811 /*
3558fd73 812 * Set the objset user_ptr to track its zsb.
34dc7c2f 813 */
3558fd73
BB
814 mutex_enter(&zsb->z_os->os_user_ptr_lock);
815 dmu_objset_set_user(zsb->z_os, zsb);
816 mutex_exit(&zsb->z_os->os_user_ptr_lock);
34dc7c2f 817
3558fd73 818 zsb->z_log = zil_open(zsb->z_os, zfs_get_data);
9babb374 819
34dc7c2f
BB
820 /*
821 * If we are not mounting (ie: online recv), then we don't
822 * have to worry about replaying the log as we blocked all
823 * operations out since we closed the ZIL.
824 */
825 if (mounting) {
b128c09f
BB
826 boolean_t readonly;
827
34dc7c2f
BB
828 /*
829 * During replay we remove the read only flag to
830 * allow replays to succeed.
831 */
2cf7f52b 832 readonly = zfs_is_readonly(zsb);
fb5f0bc8 833 if (readonly != 0)
2cf7f52b 834 readonly_changed_cb(zsb, B_FALSE);
fb5f0bc8 835 else
3558fd73 836 zfs_unlinked_drain(zsb);
34dc7c2f 837
428870ff
BB
838 /*
839 * Parse and replay the intent log.
840 *
841 * Because of ziltest, this must be done after
842 * zfs_unlinked_drain(). (Further note: ziltest
843 * doesn't use readonly mounts, where
844 * zfs_unlinked_drain() isn't called.) This is because
845 * ziltest causes spa_sync() to think it's committed,
846 * but actually it is not, so the intent log contains
847 * many txg's worth of changes.
848 *
849 * In particular, if object N is in the unlinked set in
850 * the last txg to actually sync, then it could be
851 * actually freed in a later txg and then reallocated
852 * in a yet later txg. This would write a "create
853 * object N" record to the intent log. Normally, this
854 * would be fine because the spa_sync() would have
855 * written out the fact that object N is free, before
856 * we could write the "create object N" intent log
857 * record.
858 *
859 * But when we are in ziltest mode, we advance the "open
860 * txg" without actually spa_sync()-ing the changes to
861 * disk. So we would see that object N is still
862 * allocated and in the unlinked set, and there is an
863 * intent log record saying to allocate it.
864 */
3558fd73 865 if (spa_writeable(dmu_objset_spa(zsb->z_os))) {
572e2857 866 if (zil_replay_disable) {
3558fd73 867 zil_destroy(zsb->z_log, B_FALSE);
572e2857 868 } else {
3558fd73
BB
869 zsb->z_replay = B_TRUE;
870 zil_replay(zsb->z_os, zsb,
572e2857 871 zfs_replay_vector);
3558fd73 872 zsb->z_replay = B_FALSE;
572e2857 873 }
fb5f0bc8 874 }
2cf7f52b
BB
875
876 /* restore readonly bit */
877 if (readonly != 0)
878 readonly_changed_cb(zsb, B_TRUE);
34dc7c2f
BB
879 }
880
34dc7c2f
BB
881 return (0);
882}
86f35f34 883EXPORT_SYMBOL(zfs_sb_setup);
34dc7c2f 884
9babb374 885void
3558fd73 886zfs_sb_free(zfs_sb_t *zsb)
34dc7c2f 887{
9babb374 888 int i;
9babb374 889
3558fd73 890 zfs_fuid_destroy(zsb);
9babb374 891
3558fd73
BB
892 mutex_destroy(&zsb->z_znodes_lock);
893 mutex_destroy(&zsb->z_lock);
894 list_destroy(&zsb->z_all_znodes);
895 rrw_destroy(&zsb->z_teardown_lock);
896 rw_destroy(&zsb->z_teardown_inactive_lock);
897 rw_destroy(&zsb->z_fuid_lock);
9babb374 898 for (i = 0; i != ZFS_OBJ_MTX_SZ; i++)
3558fd73 899 mutex_destroy(&zsb->z_hold_mtx[i]);
ebe7e575
BB
900 mutex_destroy(&zsb->z_ctldir_lock);
901 avl_destroy(&zsb->z_ctldir_snaps);
3558fd73 902 kmem_free(zsb, sizeof (zfs_sb_t));
34dc7c2f 903}
86f35f34 904EXPORT_SYMBOL(zfs_sb_free);
34dc7c2f 905
9babb374 906static void
3558fd73 907zfs_set_fuid_feature(zfs_sb_t *zsb)
9babb374 908{
3558fd73
BB
909 zsb->z_use_fuids = USE_FUIDS(zsb->z_version, zsb->z_os);
910 zsb->z_use_sa = USE_SA(zsb->z_version, zsb->z_os);
9babb374 911}
34dc7c2f
BB
912
913void
3558fd73 914zfs_unregister_callbacks(zfs_sb_t *zsb)
34dc7c2f 915{
3558fd73 916 objset_t *os = zsb->z_os;
34dc7c2f
BB
917 struct dsl_dataset *ds;
918
919 /*
920 * Unregister properties.
921 */
922 if (!dmu_objset_is_snapshot(os)) {
923 ds = dmu_objset_ds(os);
924 VERIFY(dsl_prop_unregister(ds, "atime", atime_changed_cb,
3558fd73 925 zsb) == 0);
34dc7c2f 926
6d111134
TC
927 VERIFY(dsl_prop_unregister(ds, "relatime", relatime_changed_cb,
928 zsb) == 0);
929
34dc7c2f 930 VERIFY(dsl_prop_unregister(ds, "xattr", xattr_changed_cb,
3558fd73 931 zsb) == 0);
34dc7c2f
BB
932
933 VERIFY(dsl_prop_unregister(ds, "recordsize", blksz_changed_cb,
3558fd73 934 zsb) == 0);
34dc7c2f
BB
935
936 VERIFY(dsl_prop_unregister(ds, "readonly", readonly_changed_cb,
3558fd73 937 zsb) == 0);
34dc7c2f
BB
938
939 VERIFY(dsl_prop_unregister(ds, "devices", devices_changed_cb,
3558fd73 940 zsb) == 0);
34dc7c2f
BB
941
942 VERIFY(dsl_prop_unregister(ds, "setuid", setuid_changed_cb,
3558fd73 943 zsb) == 0);
34dc7c2f
BB
944
945 VERIFY(dsl_prop_unregister(ds, "exec", exec_changed_cb,
3558fd73 946 zsb) == 0);
34dc7c2f
BB
947
948 VERIFY(dsl_prop_unregister(ds, "snapdir", snapdir_changed_cb,
3558fd73 949 zsb) == 0);
34dc7c2f 950
023699cd
MM
951 VERIFY(dsl_prop_unregister(ds, "acltype", acltype_changed_cb,
952 zsb) == 0);
953
34dc7c2f 954 VERIFY(dsl_prop_unregister(ds, "aclinherit",
3558fd73 955 acl_inherit_changed_cb, zsb) == 0);
34dc7c2f
BB
956
957 VERIFY(dsl_prop_unregister(ds, "vscan",
3558fd73 958 vscan_changed_cb, zsb) == 0);
2cf7f52b
BB
959
960 VERIFY(dsl_prop_unregister(ds, "nbmand",
961 nbmand_changed_cb, zsb) == 0);
34dc7c2f
BB
962 }
963}
e5c39b95 964EXPORT_SYMBOL(zfs_unregister_callbacks);
34dc7c2f 965
bc3e15e3 966#ifdef HAVE_MLSLABEL
428870ff 967/*
d3cc8b15
WA
968 * Check that the hex label string is appropriate for the dataset being
969 * mounted into the global_zone proper.
428870ff 970 *
d3cc8b15
WA
971 * Return an error if the hex label string is not default or
972 * admin_low/admin_high. For admin_low labels, the corresponding
973 * dataset must be readonly.
428870ff
BB
974 */
975int
976zfs_check_global_label(const char *dsname, const char *hexsl)
977{
978 if (strcasecmp(hexsl, ZFS_MLSLABEL_DEFAULT) == 0)
979 return (0);
980 if (strcasecmp(hexsl, ADMIN_HIGH) == 0)
981 return (0);
982 if (strcasecmp(hexsl, ADMIN_LOW) == 0) {
983 /* must be readonly */
984 uint64_t rdonly;
985
986 if (dsl_prop_get_integer(dsname,
987 zfs_prop_to_name(ZFS_PROP_READONLY), &rdonly, NULL))
2e528b49 988 return (SET_ERROR(EACCES));
428870ff
BB
989 return (rdonly ? 0 : EACCES);
990 }
2e528b49 991 return (SET_ERROR(EACCES));
428870ff 992}
86f35f34 993EXPORT_SYMBOL(zfs_check_global_label);
bc3e15e3 994#endif /* HAVE_MLSLABEL */
428870ff 995
e5c39b95 996int
3558fd73 997zfs_statvfs(struct dentry *dentry, struct kstatfs *statp)
34dc7c2f 998{
3558fd73 999 zfs_sb_t *zsb = dentry->d_sb->s_fs_info;
34dc7c2f 1000 uint64_t refdbytes, availbytes, usedobjs, availobjs;
04f9432d 1001 uint64_t fsid;
3558fd73 1002 uint32_t bshift;
34dc7c2f 1003
3558fd73 1004 ZFS_ENTER(zsb);
34dc7c2f 1005
3558fd73 1006 dmu_objset_space(zsb->z_os,
34dc7c2f
BB
1007 &refdbytes, &availbytes, &usedobjs, &availobjs);
1008
04f9432d 1009 fsid = dmu_objset_fsid_guid(zsb->z_os);
34dc7c2f 1010 /*
05ff35c6
BB
1011 * The underlying storage pool actually uses multiple block
1012 * size. Under Solaris frsize (fragment size) is reported as
1013 * the smallest block size we support, and bsize (block size)
1014 * as the filesystem's maximum block size. Unfortunately,
1015 * under Linux the fragment size and block size are often used
1016 * interchangeably. Thus we are forced to report both of them
1017 * as the filesystem's maximum block size.
34dc7c2f 1018 */
05ff35c6 1019 statp->f_frsize = zsb->z_max_blksz;
3558fd73
BB
1020 statp->f_bsize = zsb->z_max_blksz;
1021 bshift = fls(statp->f_bsize) - 1;
34dc7c2f
BB
1022
1023 /*
3558fd73
BB
1024 * The following report "total" blocks of various kinds in
1025 * the file system, but reported in terms of f_bsize - the
1026 * "preferred" size.
34dc7c2f
BB
1027 */
1028
3558fd73
BB
1029 statp->f_blocks = (refdbytes + availbytes) >> bshift;
1030 statp->f_bfree = availbytes >> bshift;
34dc7c2f
BB
1031 statp->f_bavail = statp->f_bfree; /* no root reservation */
1032
1033 /*
1034 * statvfs() should really be called statufs(), because it assumes
1035 * static metadata. ZFS doesn't preallocate files, so the best
1036 * we can do is report the max that could possibly fit in f_files,
1037 * and that minus the number actually used in f_ffree.
1038 * For f_ffree, report the smaller of the number of object available
1039 * and the number of blocks (each object will take at least a block).
1040 */
baab0630 1041 statp->f_ffree = MIN(availobjs, availbytes >> DNODE_SHIFT);
34dc7c2f 1042 statp->f_files = statp->f_ffree + usedobjs;
04f9432d
CP
1043 statp->f_fsid.val[0] = (uint32_t)fsid;
1044 statp->f_fsid.val[1] = (uint32_t)(fsid >> 32);
3558fd73
BB
1045 statp->f_type = ZFS_SUPER_MAGIC;
1046 statp->f_namelen = ZFS_MAXNAMELEN;
34dc7c2f
BB
1047
1048 /*
3558fd73 1049 * We have all of 40 characters to stuff a string here.
34dc7c2f
BB
1050 * Is there anything useful we could/should provide?
1051 */
3558fd73 1052 bzero(statp->f_spare, sizeof (statp->f_spare));
34dc7c2f 1053
3558fd73 1054 ZFS_EXIT(zsb);
34dc7c2f
BB
1055 return (0);
1056}
e5c39b95 1057EXPORT_SYMBOL(zfs_statvfs);
34dc7c2f 1058
e5c39b95 1059int
3558fd73 1060zfs_root(zfs_sb_t *zsb, struct inode **ipp)
34dc7c2f 1061{
34dc7c2f
BB
1062 znode_t *rootzp;
1063 int error;
1064
3558fd73 1065 ZFS_ENTER(zsb);
34dc7c2f 1066
3558fd73 1067 error = zfs_zget(zsb, zsb->z_root, &rootzp);
34dc7c2f 1068 if (error == 0)
3558fd73 1069 *ipp = ZTOI(rootzp);
34dc7c2f 1070
3558fd73 1071 ZFS_EXIT(zsb);
34dc7c2f
BB
1072 return (error);
1073}
e5c39b95 1074EXPORT_SYMBOL(zfs_root);
34dc7c2f 1075
ab26409d
BB
1076#ifdef HAVE_SHRINK
1077int
1078zfs_sb_prune(struct super_block *sb, unsigned long nr_to_scan, int *objects)
1079{
1080 zfs_sb_t *zsb = sb->s_fs_info;
1081 struct shrinker *shrinker = &sb->s_shrink;
1082 struct shrink_control sc = {
1083 .nr_to_scan = nr_to_scan,
1084 .gfp_mask = GFP_KERNEL,
1085 };
1086
1087 ZFS_ENTER(zsb);
1088 *objects = (*shrinker->shrink)(shrinker, &sc);
1089 ZFS_EXIT(zsb);
1090
1091 return (0);
1092}
1093EXPORT_SYMBOL(zfs_sb_prune);
1094#endif /* HAVE_SHRINK */
1095
34dc7c2f 1096/*
7b3e34ba 1097 * Teardown the zfs_sb_t.
34dc7c2f
BB
1098 *
1099 * Note, if 'unmounting' if FALSE, we return with the 'z_teardown_lock'
1100 * and 'z_teardown_inactive_lock' held.
1101 */
3558fd73 1102int
86f35f34 1103zfs_sb_teardown(zfs_sb_t *zsb, boolean_t unmounting)
34dc7c2f
BB
1104{
1105 znode_t *zp;
1106
fd23720a
BB
1107 /*
1108 * If someone has not already unmounted this file system,
1109 * drain the iput_taskq to ensure all active references to the
1110 * zfs_sb_t have been handled only then can it be safely destroyed.
1111 */
1112 if (zsb->z_os)
1113 taskq_wait(dsl_pool_iput_taskq(dmu_objset_pool(zsb->z_os)));
1114
3558fd73 1115 rrw_enter(&zsb->z_teardown_lock, RW_WRITER, FTAG);
34dc7c2f
BB
1116
1117 if (!unmounting) {
1118 /*
ceb43b93
BB
1119 * We purge the parent filesystem's super block as the
1120 * parent filesystem and all of its snapshots have their
1121 * inode's super block set to the parent's filesystem's
1122 * super block. Note, 'z_parent' is self referential
1123 * for non-snapshots.
34dc7c2f 1124 */
ceb43b93 1125 shrink_dcache_sb(zsb->z_parent->z_sb);
34dc7c2f
BB
1126 }
1127
1128 /*
1129 * Close the zil. NB: Can't close the zil while zfs_inactive
1130 * threads are blocked as zil_close can call zfs_inactive.
1131 */
3558fd73
BB
1132 if (zsb->z_log) {
1133 zil_close(zsb->z_log);
1134 zsb->z_log = NULL;
34dc7c2f
BB
1135 }
1136
3558fd73 1137 rw_enter(&zsb->z_teardown_inactive_lock, RW_WRITER);
34dc7c2f
BB
1138
1139 /*
1140 * If we are not unmounting (ie: online recv) and someone already
1141 * unmounted this file system while we were doing the switcheroo,
1142 * or a reopen of z_os failed then just bail out now.
1143 */
3558fd73
BB
1144 if (!unmounting && (zsb->z_unmounted || zsb->z_os == NULL)) {
1145 rw_exit(&zsb->z_teardown_inactive_lock);
1146 rrw_exit(&zsb->z_teardown_lock, FTAG);
2e528b49 1147 return (SET_ERROR(EIO));
34dc7c2f
BB
1148 }
1149
1150 /*
7b3e34ba
BB
1151 * At this point there are no VFS ops active, and any new VFS ops
1152 * will fail with EIO since we have z_teardown_lock for writer (only
1153 * relevant for forced unmount).
34dc7c2f
BB
1154 *
1155 * Release all holds on dbufs.
1156 */
3558fd73
BB
1157 mutex_enter(&zsb->z_znodes_lock);
1158 for (zp = list_head(&zsb->z_all_znodes); zp != NULL;
7b3e34ba 1159 zp = list_next(&zsb->z_all_znodes, zp)) {
2e40f094 1160 if (zp->z_sa_hdl)
34dc7c2f 1161 zfs_znode_dmu_fini(zp);
7b3e34ba 1162 }
3558fd73 1163 mutex_exit(&zsb->z_znodes_lock);
34dc7c2f
BB
1164
1165 /*
7b3e34ba 1166 * If we are unmounting, set the unmounted flag and let new VFS ops
34dc7c2f 1167 * unblock. zfs_inactive will have the unmounted behavior, and all
7b3e34ba 1168 * other VFS ops will fail with EIO.
34dc7c2f
BB
1169 */
1170 if (unmounting) {
3558fd73
BB
1171 zsb->z_unmounted = B_TRUE;
1172 rrw_exit(&zsb->z_teardown_lock, FTAG);
1173 rw_exit(&zsb->z_teardown_inactive_lock);
34dc7c2f
BB
1174 }
1175
1176 /*
1177 * z_os will be NULL if there was an error in attempting to reopen
3558fd73
BB
1178 * zsb, so just return as the properties had already been
1179 *
34dc7c2f
BB
1180 * unregistered and cached data had been evicted before.
1181 */
3558fd73 1182 if (zsb->z_os == NULL)
34dc7c2f
BB
1183 return (0);
1184
1185 /*
1186 * Unregister properties.
1187 */
3558fd73 1188 zfs_unregister_callbacks(zsb);
34dc7c2f
BB
1189
1190 /*
1191 * Evict cached data
1192 */
04434775
MA
1193 if (dsl_dataset_is_dirty(dmu_objset_ds(zsb->z_os)) &&
1194 !zfs_is_readonly(zsb))
1195 txg_wait_synced(dmu_objset_pool(zsb->z_os), 0);
13fe0198 1196 dmu_objset_evict_dbufs(zsb->z_os);
34dc7c2f
BB
1197
1198 return (0);
1199}
86f35f34 1200EXPORT_SYMBOL(zfs_sb_teardown);
34dc7c2f 1201
5547c2f1
BB
1202#if defined(HAVE_BDI) && !defined(HAVE_BDI_SETUP_AND_REGISTER)
1203atomic_long_t zfs_bdi_seq = ATOMIC_LONG_INIT(0);
1204#endif /* HAVE_BDI && !HAVE_BDI_SETUP_AND_REGISTER */
76659dc1 1205
e5c39b95 1206int
3558fd73 1207zfs_domount(struct super_block *sb, void *data, int silent)
34dc7c2f 1208{
3558fd73
BB
1209 zpl_mount_data_t *zmd = data;
1210 const char *osname = zmd->z_osname;
1211 zfs_sb_t *zsb;
1212 struct inode *root_inode;
1213 uint64_t recordsize;
1214 int error;
34dc7c2f 1215
3558fd73
BB
1216 error = zfs_sb_create(osname, &zsb);
1217 if (error)
1218 return (error);
1219
1220 if ((error = dsl_prop_get_integer(osname, "recordsize",
1221 &recordsize, NULL)))
1222 goto out;
1223
1224 zsb->z_sb = sb;
3558fd73
BB
1225 sb->s_fs_info = zsb;
1226 sb->s_magic = ZFS_SUPER_MAGIC;
1227 sb->s_maxbytes = MAX_LFS_FILESIZE;
1228 sb->s_time_gran = 1;
1229 sb->s_blocksize = recordsize;
1230 sb->s_blocksize_bits = ilog2(recordsize);
5547c2f1
BB
1231
1232#ifdef HAVE_BDI
1233 /*
1234 * 2.6.32 API change,
1235 * Added backing_device_info (BDI) per super block interfaces. A BDI
1236 * must be configured when using a non-device backed filesystem for
1237 * proper writeback. This is not required for older pdflush kernels.
1238 *
1239 * NOTE: Linux read-ahead is disabled in favor of zfs read-ahead.
1240 */
1241 zsb->z_bdi.ra_pages = 0;
1242 sb->s_bdi = &zsb->z_bdi;
1243
1244 error = -bdi_setup_and_register(&zsb->z_bdi, "zfs", BDI_CAP_MAP_COPY);
1245 if (error)
1246 goto out;
1247#endif /* HAVE_BDI */
3558fd73
BB
1248
1249 /* Set callback operations for the file system. */
1250 sb->s_op = &zpl_super_operations;
1251 sb->s_xattr = zpl_xattr_handlers;
055656d4 1252 sb->s_export_op = &zpl_export_operations;
ee930353
BB
1253#ifdef HAVE_S_D_OP
1254 sb->s_d_op = &zpl_dentry_operations;
1255#endif /* HAVE_S_D_OP */
3558fd73
BB
1256
1257 /* Set features for file system. */
1258 zfs_set_fuid_feature(zsb);
1259
1260 if (dmu_objset_is_snapshot(zsb->z_os)) {
1261 uint64_t pval;
1262
1263 atime_changed_cb(zsb, B_FALSE);
1264 readonly_changed_cb(zsb, B_TRUE);
d1d7e268
MK
1265 if ((error = dsl_prop_get_integer(osname,
1266 "xattr", &pval, NULL)))
3558fd73
BB
1267 goto out;
1268 xattr_changed_cb(zsb, pval);
d1d7e268
MK
1269 if ((error = dsl_prop_get_integer(osname,
1270 "acltype", &pval, NULL)))
023699cd
MM
1271 goto out;
1272 acltype_changed_cb(zsb, pval);
3558fd73
BB
1273 zsb->z_issnap = B_TRUE;
1274 zsb->z_os->os_sync = ZFS_SYNC_DISABLED;
1275
1276 mutex_enter(&zsb->z_os->os_user_ptr_lock);
1277 dmu_objset_set_user(zsb->z_os, zsb);
1278 mutex_exit(&zsb->z_os->os_user_ptr_lock);
1279 } else {
1280 error = zfs_sb_setup(zsb, B_TRUE);
34dc7c2f
BB
1281 }
1282
3558fd73
BB
1283 /* Allocate a root inode for the filesystem. */
1284 error = zfs_root(zsb, &root_inode);
1285 if (error) {
1286 (void) zfs_umount(sb);
1287 goto out;
34dc7c2f
BB
1288 }
1289
3558fd73 1290 /* Allocate a root dentry for the filesystem */
6a0936ba 1291 sb->s_root = d_make_root(root_inode);
3558fd73
BB
1292 if (sb->s_root == NULL) {
1293 (void) zfs_umount(sb);
2e528b49 1294 error = SET_ERROR(ENOMEM);
3558fd73
BB
1295 goto out;
1296 }
ebe7e575
BB
1297
1298 if (!zsb->z_issnap)
1299 zfsctl_create(zsb);
3558fd73
BB
1300out:
1301 if (error) {
1302 dmu_objset_disown(zsb->z_os, zsb);
1303 zfs_sb_free(zsb);
1304 }
34dc7c2f 1305
3558fd73
BB
1306 return (error);
1307}
1308EXPORT_SYMBOL(zfs_domount);
1309
ebe7e575
BB
1310/*
1311 * Called when an unmount is requested and certain sanity checks have
1312 * already passed. At this point no dentries or inodes have been reclaimed
1313 * from their respective caches. We drop the extra reference on the .zfs
1314 * control directory to allow everything to be reclaimed. All snapshots
1315 * must already have been unmounted to reach this point.
1316 */
1317void
1318zfs_preumount(struct super_block *sb)
1319{
1320 zfs_sb_t *zsb = sb->s_fs_info;
1321
1f0d8a56 1322 if (zsb != NULL && zsb->z_ctldir != NULL)
ebe7e575
BB
1323 zfsctl_destroy(zsb);
1324}
1325EXPORT_SYMBOL(zfs_preumount);
1326
1327/*
1328 * Called once all other unmount released tear down has occurred.
1329 * It is our responsibility to release any remaining infrastructure.
1330 */
3558fd73
BB
1331/*ARGSUSED*/
1332int
1333zfs_umount(struct super_block *sb)
1334{
1335 zfs_sb_t *zsb = sb->s_fs_info;
1336 objset_t *os;
1337
86f35f34 1338 VERIFY(zfs_sb_teardown(zsb, B_TRUE) == 0);
3558fd73 1339 os = zsb->z_os;
34dc7c2f 1340
5547c2f1
BB
1341#ifdef HAVE_BDI
1342 bdi_destroy(sb->s_bdi);
1343#endif /* HAVE_BDI */
76659dc1 1344
34dc7c2f
BB
1345 /*
1346 * z_os will be NULL if there was an error in
3558fd73 1347 * attempting to reopen zsb.
34dc7c2f
BB
1348 */
1349 if (os != NULL) {
1350 /*
1351 * Unset the objset user_ptr.
1352 */
428870ff 1353 mutex_enter(&os->os_user_ptr_lock);
34dc7c2f 1354 dmu_objset_set_user(os, NULL);
428870ff 1355 mutex_exit(&os->os_user_ptr_lock);
34dc7c2f
BB
1356
1357 /*
b128c09f 1358 * Finally release the objset
34dc7c2f 1359 */
3558fd73 1360 dmu_objset_disown(os, zsb);
34dc7c2f
BB
1361 }
1362
3558fd73 1363 zfs_sb_free(zsb);
34dc7c2f
BB
1364 return (0);
1365}
e5c39b95 1366EXPORT_SYMBOL(zfs_umount);
34dc7c2f 1367
0de19dad
BB
1368int
1369zfs_remount(struct super_block *sb, int *flags, char *data)
1370{
0de19dad 1371 /*
2cf7f52b
BB
1372 * All namespace flags (MNT_*) and super block flags (MS_*) will
1373 * be handled by the Linux VFS. Only handle custom options here.
0de19dad 1374 */
0de19dad
BB
1375 return (0);
1376}
1377EXPORT_SYMBOL(zfs_remount);
1378
e5c39b95 1379int
2cf7f52b 1380zfs_vget(struct super_block *sb, struct inode **ipp, fid_t *fidp)
34dc7c2f 1381{
2cf7f52b 1382 zfs_sb_t *zsb = sb->s_fs_info;
34dc7c2f
BB
1383 znode_t *zp;
1384 uint64_t object = 0;
1385 uint64_t fid_gen = 0;
1386 uint64_t gen_mask;
1387 uint64_t zp_gen;
3558fd73 1388 int i, err;
34dc7c2f 1389
3558fd73 1390 *ipp = NULL;
34dc7c2f 1391
3558fd73 1392 ZFS_ENTER(zsb);
34dc7c2f
BB
1393
1394 if (fidp->fid_len == LONG_FID_LEN) {
1395 zfid_long_t *zlfid = (zfid_long_t *)fidp;
1396 uint64_t objsetid = 0;
1397 uint64_t setgen = 0;
1398
1399 for (i = 0; i < sizeof (zlfid->zf_setid); i++)
1400 objsetid |= ((uint64_t)zlfid->zf_setid[i]) << (8 * i);
1401
1402 for (i = 0; i < sizeof (zlfid->zf_setgen); i++)
1403 setgen |= ((uint64_t)zlfid->zf_setgen[i]) << (8 * i);
1404
3558fd73 1405 ZFS_EXIT(zsb);
34dc7c2f 1406
ebe7e575 1407 err = zfsctl_lookup_objset(sb, objsetid, &zsb);
34dc7c2f 1408 if (err)
2e528b49 1409 return (SET_ERROR(EINVAL));
ebe7e575 1410
3558fd73 1411 ZFS_ENTER(zsb);
34dc7c2f
BB
1412 }
1413
1414 if (fidp->fid_len == SHORT_FID_LEN || fidp->fid_len == LONG_FID_LEN) {
1415 zfid_short_t *zfid = (zfid_short_t *)fidp;
1416
1417 for (i = 0; i < sizeof (zfid->zf_object); i++)
1418 object |= ((uint64_t)zfid->zf_object[i]) << (8 * i);
1419
1420 for (i = 0; i < sizeof (zfid->zf_gen); i++)
1421 fid_gen |= ((uint64_t)zfid->zf_gen[i]) << (8 * i);
1422 } else {
3558fd73 1423 ZFS_EXIT(zsb);
2e528b49 1424 return (SET_ERROR(EINVAL));
34dc7c2f
BB
1425 }
1426
1427 /* A zero fid_gen means we are in the .zfs control directories */
1428 if (fid_gen == 0 &&
1429 (object == ZFSCTL_INO_ROOT || object == ZFSCTL_INO_SNAPDIR)) {
3558fd73
BB
1430 *ipp = zsb->z_ctldir;
1431 ASSERT(*ipp != NULL);
34dc7c2f 1432 if (object == ZFSCTL_INO_SNAPDIR) {
ebe7e575
BB
1433 VERIFY(zfsctl_root_lookup(*ipp, "snapshot", ipp,
1434 0, kcred, NULL, NULL) == 0);
34dc7c2f 1435 } else {
3558fd73 1436 igrab(*ipp);
34dc7c2f 1437 }
3558fd73 1438 ZFS_EXIT(zsb);
34dc7c2f
BB
1439 return (0);
1440 }
1441
1442 gen_mask = -1ULL >> (64 - 8 * i);
1443
1444 dprintf("getting %llu [%u mask %llx]\n", object, fid_gen, gen_mask);
3558fd73
BB
1445 if ((err = zfs_zget(zsb, object, &zp))) {
1446 ZFS_EXIT(zsb);
34dc7c2f
BB
1447 return (err);
1448 }
3558fd73 1449 (void) sa_lookup(zp->z_sa_hdl, SA_ZPL_GEN(zsb), &zp_gen,
428870ff
BB
1450 sizeof (uint64_t));
1451 zp_gen = zp_gen & gen_mask;
34dc7c2f
BB
1452 if (zp_gen == 0)
1453 zp_gen = 1;
1454 if (zp->z_unlinked || zp_gen != fid_gen) {
1455 dprintf("znode gen (%u) != fid gen (%u)\n", zp_gen, fid_gen);
3558fd73
BB
1456 iput(ZTOI(zp));
1457 ZFS_EXIT(zsb);
2e528b49 1458 return (SET_ERROR(EINVAL));
34dc7c2f
BB
1459 }
1460
3558fd73
BB
1461 *ipp = ZTOI(zp);
1462 if (*ipp)
1463 zfs_inode_update(ITOZ(*ipp));
960e08fe 1464
3558fd73 1465 ZFS_EXIT(zsb);
34dc7c2f
BB
1466 return (0);
1467}
e5c39b95 1468EXPORT_SYMBOL(zfs_vget);
34dc7c2f
BB
1469
1470/*
7b3e34ba 1471 * Block out VFS ops and close zfs_sb_t
34dc7c2f
BB
1472 *
1473 * Note, if successful, then we return with the 'z_teardown_lock' and
831baf06
KW
1474 * 'z_teardown_inactive_lock' write held. We leave ownership of the underlying
1475 * dataset and objset intact so that they can be atomically handed off during
1476 * a subsequent rollback or recv operation and the resume thereafter.
34dc7c2f
BB
1477 */
1478int
3558fd73 1479zfs_suspend_fs(zfs_sb_t *zsb)
34dc7c2f
BB
1480{
1481 int error;
1482
86f35f34 1483 if ((error = zfs_sb_teardown(zsb, B_FALSE)) != 0)
34dc7c2f 1484 return (error);
7b3e34ba 1485
34dc7c2f
BB
1486 return (0);
1487}
e5c39b95 1488EXPORT_SYMBOL(zfs_suspend_fs);
34dc7c2f
BB
1489
1490/*
7b3e34ba 1491 * Reopen zfs_sb_t and release VFS ops.
34dc7c2f
BB
1492 */
1493int
3558fd73 1494zfs_resume_fs(zfs_sb_t *zsb, const char *osname)
34dc7c2f 1495{
2008e920 1496 int err, err2;
831baf06
KW
1497 znode_t *zp;
1498 uint64_t sa_obj = 0;
34dc7c2f 1499
3558fd73
BB
1500 ASSERT(RRW_WRITE_HELD(&zsb->z_teardown_lock));
1501 ASSERT(RW_WRITE_HELD(&zsb->z_teardown_inactive_lock));
34dc7c2f 1502
831baf06
KW
1503 /*
1504 * We already own this, so just hold and rele it to update the
1505 * objset_t, as the one we had before may have been evicted.
1506 */
1507 VERIFY0(dmu_objset_hold(osname, zsb, &zsb->z_os));
1508 VERIFY3P(zsb->z_os->os_dsl_dataset->ds_owner, ==, zsb);
1509 VERIFY(dsl_dataset_long_held(zsb->z_os->os_dsl_dataset));
1510 dmu_objset_rele(zsb->z_os, zsb);
428870ff 1511
831baf06
KW
1512 /*
1513 * Make sure version hasn't changed
1514 */
428870ff 1515
831baf06
KW
1516 err = zfs_get_zplprop(zsb->z_os, ZFS_PROP_VERSION,
1517 &zsb->z_version);
c1fabe79 1518
831baf06
KW
1519 if (err)
1520 goto bail;
428870ff 1521
831baf06
KW
1522 err = zap_lookup(zsb->z_os, MASTER_NODE_OBJ,
1523 ZFS_SA_ATTRS, 8, 1, &sa_obj);
c1fabe79 1524
831baf06
KW
1525 if (err && zsb->z_version >= ZPL_VERSION_SA)
1526 goto bail;
428870ff 1527
831baf06
KW
1528 if ((err = sa_setup(zsb->z_os, sa_obj,
1529 zfs_attr_table, ZPL_END, &zsb->z_attr_table)) != 0)
1530 goto bail;
34dc7c2f 1531
831baf06
KW
1532 if (zsb->z_version >= ZPL_VERSION_SA)
1533 sa_register_update_callback(zsb->z_os,
1534 zfs_sa_upgrade);
c1fabe79 1535
831baf06 1536 VERIFY(zfs_sb_setup(zsb, B_FALSE) == 0);
c1fabe79 1537
831baf06
KW
1538 zfs_set_fuid_feature(zsb);
1539 zsb->z_rollback_time = jiffies;
34dc7c2f 1540
831baf06
KW
1541 /*
1542 * Attempt to re-establish all the active inodes with their
1543 * dbufs. If a zfs_rezget() fails, then we unhash the inode
1544 * and mark it stale. This prevents a collision if a new
1545 * inode/object is created which must use the same inode
1546 * number. The stale inode will be be released when the
1547 * VFS prunes the dentry holding the remaining references
1548 * on the stale inode.
1549 */
1550 mutex_enter(&zsb->z_znodes_lock);
1551 for (zp = list_head(&zsb->z_all_znodes); zp;
1552 zp = list_next(&zsb->z_all_znodes, zp)) {
2008e920
TC
1553 err2 = zfs_rezget(zp);
1554 if (err2) {
831baf06
KW
1555 remove_inode_hash(ZTOI(zp));
1556 zp->z_is_stale = B_TRUE;
34dc7c2f 1557 }
34dc7c2f 1558 }
831baf06 1559 mutex_exit(&zsb->z_znodes_lock);
34dc7c2f 1560
428870ff 1561bail:
7b3e34ba 1562 /* release the VFS ops */
3558fd73
BB
1563 rw_exit(&zsb->z_teardown_inactive_lock);
1564 rrw_exit(&zsb->z_teardown_lock, FTAG);
34dc7c2f
BB
1565
1566 if (err) {
1567 /*
831baf06
KW
1568 * Since we couldn't setup the sa framework, try to force
1569 * unmount this file system.
34dc7c2f 1570 */
ab68b6e5
TC
1571 if (zsb->z_os)
1572 (void) zfs_umount(zsb->z_sb);
34dc7c2f
BB
1573 }
1574 return (err);
1575}
e5c39b95 1576EXPORT_SYMBOL(zfs_resume_fs);
34dc7c2f 1577
34dc7c2f 1578int
3558fd73 1579zfs_set_version(zfs_sb_t *zsb, uint64_t newvers)
34dc7c2f
BB
1580{
1581 int error;
3558fd73 1582 objset_t *os = zsb->z_os;
34dc7c2f 1583 dmu_tx_t *tx;
34dc7c2f
BB
1584
1585 if (newvers < ZPL_VERSION_INITIAL || newvers > ZPL_VERSION)
2e528b49 1586 return (SET_ERROR(EINVAL));
34dc7c2f 1587
3558fd73 1588 if (newvers < zsb->z_version)
2e528b49 1589 return (SET_ERROR(EINVAL));
34dc7c2f 1590
428870ff 1591 if (zfs_spa_version_map(newvers) >
3558fd73 1592 spa_version(dmu_objset_spa(zsb->z_os)))
2e528b49 1593 return (SET_ERROR(ENOTSUP));
428870ff 1594
34dc7c2f 1595 tx = dmu_tx_create(os);
9babb374 1596 dmu_tx_hold_zap(tx, MASTER_NODE_OBJ, B_FALSE, ZPL_VERSION_STR);
3558fd73 1597 if (newvers >= ZPL_VERSION_SA && !zsb->z_use_sa) {
428870ff
BB
1598 dmu_tx_hold_zap(tx, MASTER_NODE_OBJ, B_TRUE,
1599 ZFS_SA_ATTRS);
1600 dmu_tx_hold_zap(tx, DMU_NEW_OBJECT, FALSE, NULL);
1601 }
34dc7c2f
BB
1602 error = dmu_tx_assign(tx, TXG_WAIT);
1603 if (error) {
1604 dmu_tx_abort(tx);
9babb374
BB
1605 return (error);
1606 }
428870ff 1607
9babb374
BB
1608 error = zap_update(os, MASTER_NODE_OBJ, ZPL_VERSION_STR,
1609 8, 1, &newvers, tx);
1610
1611 if (error) {
1612 dmu_tx_commit(tx);
1613 return (error);
34dc7c2f 1614 }
34dc7c2f 1615
3558fd73 1616 if (newvers >= ZPL_VERSION_SA && !zsb->z_use_sa) {
428870ff
BB
1617 uint64_t sa_obj;
1618
3558fd73 1619 ASSERT3U(spa_version(dmu_objset_spa(zsb->z_os)), >=,
428870ff
BB
1620 SPA_VERSION_SA);
1621 sa_obj = zap_create(os, DMU_OT_SA_MASTER_NODE,
1622 DMU_OT_NONE, 0, tx);
1623
1624 error = zap_add(os, MASTER_NODE_OBJ,
1625 ZFS_SA_ATTRS, 8, 1, &sa_obj, tx);
c99c9001 1626 ASSERT0(error);
428870ff
BB
1627
1628 VERIFY(0 == sa_set_sa_object(os, sa_obj));
1629 sa_register_update_callback(os, zfs_sa_upgrade);
1630 }
1631
6f1ffb06
MA
1632 spa_history_log_internal_ds(dmu_objset_ds(os), "upgrade", tx,
1633 "from %llu to %llu", zsb->z_version, newvers);
9babb374 1634
34dc7c2f
BB
1635 dmu_tx_commit(tx);
1636
3558fd73 1637 zsb->z_version = newvers;
9babb374 1638
c1fabe79 1639 zfs_set_fuid_feature(zsb);
9babb374
BB
1640
1641 return (0);
34dc7c2f 1642}
e5c39b95 1643EXPORT_SYMBOL(zfs_set_version);
34dc7c2f
BB
1644
1645/*
1646 * Read a property stored within the master node.
1647 */
1648int
1649zfs_get_zplprop(objset_t *os, zfs_prop_t prop, uint64_t *value)
1650{
1651 const char *pname;
2e528b49 1652 int error = SET_ERROR(ENOENT);
34dc7c2f
BB
1653
1654 /*
1655 * Look up the file system's value for the property. For the
1656 * version property, we look up a slightly different string.
1657 */
1658 if (prop == ZFS_PROP_VERSION)
1659 pname = ZPL_VERSION_STR;
1660 else
1661 pname = zfs_prop_to_name(prop);
1662
b128c09f
BB
1663 if (os != NULL)
1664 error = zap_lookup(os, MASTER_NODE_OBJ, pname, 8, 1, value);
34dc7c2f
BB
1665
1666 if (error == ENOENT) {
1667 /* No value set, use the default value */
1668 switch (prop) {
1669 case ZFS_PROP_VERSION:
1670 *value = ZPL_VERSION;
1671 break;
1672 case ZFS_PROP_NORMALIZE:
1673 case ZFS_PROP_UTF8ONLY:
1674 *value = 0;
1675 break;
1676 case ZFS_PROP_CASE:
1677 *value = ZFS_CASE_SENSITIVE;
1678 break;
023699cd
MM
1679 case ZFS_PROP_ACLTYPE:
1680 *value = ZFS_ACLTYPE_OFF;
1681 break;
34dc7c2f
BB
1682 default:
1683 return (error);
1684 }
1685 error = 0;
1686 }
1687 return (error);
1688}
86f35f34 1689EXPORT_SYMBOL(zfs_get_zplprop);
3558fd73
BB
1690
1691void
1692zfs_init(void)
1693{
ebe7e575 1694 zfsctl_init();
3558fd73
BB
1695 zfs_znode_init();
1696 dmu_objset_register_type(DMU_OST_ZFS, zfs_space_delta_cb);
1697 register_filesystem(&zpl_fs_type);
ab26409d 1698 (void) arc_add_prune_callback(zpl_prune_sbs, NULL);
3558fd73
BB
1699}
1700
1701void
1702zfs_fini(void)
1703{
023699cd 1704 taskq_wait(system_taskq);
3558fd73
BB
1705 unregister_filesystem(&zpl_fs_type);
1706 zfs_znode_fini();
ebe7e575 1707 zfsctl_fini();
3558fd73 1708}