*/
/*
* Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
+ * Copyright (c) 2013 by Delphix. All rights reserved.
*/
/* Portions Copyright 2010 Robert Milkowski */
#include <sys/zpl.h>
#include "zfs_comutil.h"
-
/*ARGSUSED*/
int
zfs_sync(struct super_block *sb, int wait, cred_t *cr)
((zfs_sb_t *)arg)->z_atime = newval;
}
+static void
+relatime_changed_cb(void *arg, uint64_t newval)
+{
+ ((zfs_sb_t *)arg)->z_relatime = newval;
+}
+
static void
xattr_changed_cb(void *arg, uint64_t newval)
{
}
static void
-blksz_changed_cb(void *arg, uint64_t newval)
+acltype_changed_cb(void *arg, uint64_t newval)
{
zfs_sb_t *zsb = arg;
- if (newval < SPA_MINBLOCKSIZE ||
- newval > SPA_MAXBLOCKSIZE || !ISP2(newval))
- newval = SPA_MAXBLOCKSIZE;
+ switch (newval) {
+ case ZFS_ACLTYPE_OFF:
+ zsb->z_acl_type = ZFS_ACLTYPE_OFF;
+ zsb->z_sb->s_flags &= ~MS_POSIXACL;
+ break;
+ case ZFS_ACLTYPE_POSIXACL:
+#ifdef CONFIG_FS_POSIX_ACL
+ zsb->z_acl_type = ZFS_ACLTYPE_POSIXACL;
+ zsb->z_sb->s_flags |= MS_POSIXACL;
+#else
+ zsb->z_acl_type = ZFS_ACLTYPE_OFF;
+ zsb->z_sb->s_flags &= ~MS_POSIXACL;
+#endif /* CONFIG_FS_POSIX_ACL */
+ break;
+ default:
+ break;
+ }
+}
+
+static void
+blksz_changed_cb(void *arg, uint64_t newval)
+{
+ zfs_sb_t *zsb = arg;
+ ASSERT3U(newval, <=, spa_maxblocksize(dmu_objset_spa(zsb->z_os)));
+ ASSERT3U(newval, >=, SPA_MINBLOCKSIZE);
+ ASSERT(ISP2(newval));
zsb->z_max_blksz = newval;
}
{
struct dsl_dataset *ds = NULL;
objset_t *os = zsb->z_os;
- boolean_t do_readonly = B_FALSE;
+ zfs_mntopts_t *zmo = zsb->z_mntopts;
int error = 0;
- if (zfs_is_readonly(zsb) || !spa_writeable(dmu_objset_spa(os)))
- do_readonly = B_TRUE;
+ ASSERT(zsb);
+ ASSERT(zmo);
+
+ /*
+ * The act of registering our callbacks will destroy any mount
+ * options we may have. In order to enable temporary overrides
+ * of mount options, we stash away the current values and
+ * restore them after we register the callbacks.
+ */
+ if (zfs_is_readonly(zsb) || !spa_writeable(dmu_objset_spa(os))) {
+ zmo->z_do_readonly = B_TRUE;
+ zmo->z_readonly = B_TRUE;
+ }
/*
* Register property callbacks.
* overboard...
*/
ds = dmu_objset_ds(os);
+ dsl_pool_config_enter(dmu_objset_pool(os), FTAG);
error = dsl_prop_register(ds,
- "atime", atime_changed_cb, zsb);
+ zfs_prop_to_name(ZFS_PROP_ATIME), atime_changed_cb, zsb);
+ error = error ? error : dsl_prop_register(ds,
+ zfs_prop_to_name(ZFS_PROP_RELATIME), relatime_changed_cb, zsb);
error = error ? error : dsl_prop_register(ds,
- "xattr", xattr_changed_cb, zsb);
+ zfs_prop_to_name(ZFS_PROP_XATTR), xattr_changed_cb, zsb);
error = error ? error : dsl_prop_register(ds,
- "recordsize", blksz_changed_cb, zsb);
+ zfs_prop_to_name(ZFS_PROP_RECORDSIZE), blksz_changed_cb, zsb);
error = error ? error : dsl_prop_register(ds,
- "readonly", readonly_changed_cb, zsb);
+ zfs_prop_to_name(ZFS_PROP_READONLY), readonly_changed_cb, zsb);
error = error ? error : dsl_prop_register(ds,
- "devices", devices_changed_cb, zsb);
+ zfs_prop_to_name(ZFS_PROP_DEVICES), devices_changed_cb, zsb);
error = error ? error : dsl_prop_register(ds,
- "setuid", setuid_changed_cb, zsb);
+ zfs_prop_to_name(ZFS_PROP_SETUID), setuid_changed_cb, zsb);
error = error ? error : dsl_prop_register(ds,
- "exec", exec_changed_cb, zsb);
+ zfs_prop_to_name(ZFS_PROP_EXEC), exec_changed_cb, zsb);
error = error ? error : dsl_prop_register(ds,
- "snapdir", snapdir_changed_cb, zsb);
+ zfs_prop_to_name(ZFS_PROP_SNAPDIR), snapdir_changed_cb, zsb);
error = error ? error : dsl_prop_register(ds,
- "aclinherit", acl_inherit_changed_cb, zsb);
+ zfs_prop_to_name(ZFS_PROP_ACLTYPE), acltype_changed_cb, zsb);
error = error ? error : dsl_prop_register(ds,
- "vscan", vscan_changed_cb, zsb);
+ zfs_prop_to_name(ZFS_PROP_ACLINHERIT), acl_inherit_changed_cb, zsb);
error = error ? error : dsl_prop_register(ds,
- "nbmand", nbmand_changed_cb, zsb);
+ zfs_prop_to_name(ZFS_PROP_VSCAN), vscan_changed_cb, zsb);
+ error = error ? error : dsl_prop_register(ds,
+ zfs_prop_to_name(ZFS_PROP_NBMAND), nbmand_changed_cb, zsb);
+ dsl_pool_config_exit(dmu_objset_pool(os), FTAG);
if (error)
goto unregister;
- if (do_readonly)
- readonly_changed_cb(zsb, B_TRUE);
+ /*
+ * Invoke our callbacks to restore temporary mount options.
+ */
+ if (zmo->z_do_readonly)
+ readonly_changed_cb(zsb, zmo->z_readonly);
+ if (zmo->z_do_setuid)
+ setuid_changed_cb(zsb, zmo->z_setuid);
+ if (zmo->z_do_exec)
+ exec_changed_cb(zsb, zmo->z_exec);
+ if (zmo->z_do_devices)
+ devices_changed_cb(zsb, zmo->z_devices);
+ if (zmo->z_do_xattr)
+ xattr_changed_cb(zsb, zmo->z_xattr);
+ if (zmo->z_do_atime)
+ atime_changed_cb(zsb, zmo->z_atime);
+ if (zmo->z_do_relatime)
+ relatime_changed_cb(zsb, zmo->z_relatime);
+ if (zmo->z_do_nbmand)
+ nbmand_changed_cb(zsb, zmo->z_nbmand);
return (0);
* registered, but this is OK; it will simply return ENOMSG,
* which we will ignore.
*/
- (void) dsl_prop_unregister(ds, "atime", atime_changed_cb, zsb);
- (void) dsl_prop_unregister(ds, "xattr", xattr_changed_cb, zsb);
- (void) dsl_prop_unregister(ds, "recordsize", blksz_changed_cb, zsb);
- (void) dsl_prop_unregister(ds, "readonly", readonly_changed_cb, zsb);
- (void) dsl_prop_unregister(ds, "devices", devices_changed_cb, zsb);
- (void) dsl_prop_unregister(ds, "setuid", setuid_changed_cb, zsb);
- (void) dsl_prop_unregister(ds, "exec", exec_changed_cb, zsb);
- (void) dsl_prop_unregister(ds, "snapdir", snapdir_changed_cb, zsb);
- (void) dsl_prop_unregister(ds, "aclinherit", acl_inherit_changed_cb,
- zsb);
- (void) dsl_prop_unregister(ds, "vscan", vscan_changed_cb, zsb);
- (void) dsl_prop_unregister(ds, "nbmand", nbmand_changed_cb, zsb);
+ (void) dsl_prop_unregister(ds, zfs_prop_to_name(ZFS_PROP_ATIME),
+ atime_changed_cb, zsb);
+ (void) dsl_prop_unregister(ds, zfs_prop_to_name(ZFS_PROP_RELATIME),
+ relatime_changed_cb, zsb);
+ (void) dsl_prop_unregister(ds, zfs_prop_to_name(ZFS_PROP_XATTR),
+ xattr_changed_cb, zsb);
+ (void) dsl_prop_unregister(ds, zfs_prop_to_name(ZFS_PROP_RECORDSIZE),
+ blksz_changed_cb, zsb);
+ (void) dsl_prop_unregister(ds, zfs_prop_to_name(ZFS_PROP_READONLY),
+ readonly_changed_cb, zsb);
+ (void) dsl_prop_unregister(ds, zfs_prop_to_name(ZFS_PROP_DEVICES),
+ devices_changed_cb, zsb);
+ (void) dsl_prop_unregister(ds, zfs_prop_to_name(ZFS_PROP_SETUID),
+ setuid_changed_cb, zsb);
+ (void) dsl_prop_unregister(ds, zfs_prop_to_name(ZFS_PROP_EXEC),
+ exec_changed_cb, zsb);
+ (void) dsl_prop_unregister(ds, zfs_prop_to_name(ZFS_PROP_SNAPDIR),
+ snapdir_changed_cb, zsb);
+ (void) dsl_prop_unregister(ds, zfs_prop_to_name(ZFS_PROP_ACLTYPE),
+ acltype_changed_cb, zsb);
+ (void) dsl_prop_unregister(ds, zfs_prop_to_name(ZFS_PROP_ACLINHERIT),
+ acl_inherit_changed_cb, zsb);
+ (void) dsl_prop_unregister(ds, zfs_prop_to_name(ZFS_PROP_VSCAN),
+ vscan_changed_cb, zsb);
+ (void) dsl_prop_unregister(ds, zfs_prop_to_name(ZFS_PROP_NBMAND),
+ nbmand_changed_cb, zsb);
return (error);
}
zfs_space_delta_cb(dmu_object_type_t bonustype, void *data,
uint64_t *userp, uint64_t *groupp)
{
- int error = 0;
-
/*
* Is it a valid type of object to track?
*/
if (bonustype != DMU_OT_ZNODE && bonustype != DMU_OT_SA)
- return (ENOENT);
+ return (SET_ERROR(ENOENT));
/*
* If we have a NULL data pointer
* use the same ids
*/
if (data == NULL)
- return (EEXIST);
+ return (SET_ERROR(EEXIST));
if (bonustype == DMU_OT_ZNODE) {
znode_phys_t *znp = data;
*groupp = BSWAP_64(*groupp);
}
}
- return (error);
+ return (0);
}
static void
case ZFS_PROP_GROUPQUOTA:
return (zsb->z_groupquota_obj);
default:
- return (ENOTSUP);
+ return (SET_ERROR(ENOTSUP));
}
return (0);
}
uint64_t obj;
if (!dmu_objset_userspace_present(zsb->z_os))
- return (ENOTSUP);
+ return (SET_ERROR(ENOTSUP));
obj = zfs_userquota_prop_to_obj(zsb, type);
if (obj == 0) {
if (domain && domain[0]) {
domainid = zfs_fuid_find_by_domain(zsb, domain, NULL, addok);
if (domainid == -1)
- return (ENOENT);
+ return (SET_ERROR(ENOENT));
}
fuid = FUID_ENCODE(domainid, rid);
(void) sprintf(buf, "%llx", (longlong_t)fuid);
*valp = 0;
if (!dmu_objset_userspace_present(zsb->z_os))
- return (ENOTSUP);
+ return (SET_ERROR(ENOTSUP));
obj = zfs_userquota_prop_to_obj(zsb, type);
if (obj == 0)
boolean_t fuid_dirtied;
if (type != ZFS_PROP_USERQUOTA && type != ZFS_PROP_GROUPQUOTA)
- return (EINVAL);
+ return (SET_ERROR(EINVAL));
if (zsb->z_version < ZPL_VERSION_USERSPACE)
- return (ENOTSUP);
+ return (SET_ERROR(ENOTSUP));
objp = (type == ZFS_PROP_USERQUOTA) ? &zsb->z_userquota_obj :
&zsb->z_groupquota_obj;
}
EXPORT_SYMBOL(zfs_owner_overquota);
+zfs_mntopts_t *
+zfs_mntopts_alloc(void)
+{
+ return (kmem_zalloc(sizeof (zfs_mntopts_t), KM_SLEEP));
+}
+
+void
+zfs_mntopts_free(zfs_mntopts_t *zmo)
+{
+ if (zmo->z_osname)
+ strfree(zmo->z_osname);
+
+ if (zmo->z_mntpoint)
+ strfree(zmo->z_mntpoint);
+
+ kmem_free(zmo, sizeof (zfs_mntopts_t));
+}
+
int
-zfs_sb_create(const char *osname, zfs_sb_t **zsbp)
+zfs_sb_create(const char *osname, zfs_mntopts_t *zmo, zfs_sb_t **zsbp)
{
objset_t *os;
zfs_sb_t *zsb;
int i, error;
uint64_t sa_obj;
- zsb = kmem_zalloc(sizeof (zfs_sb_t), KM_SLEEP | KM_NODEBUG);
+ zsb = kmem_zalloc(sizeof (zfs_sb_t), KM_SLEEP);
/*
* We claim to always be readonly so we can open snapshots;
return (error);
}
+ /*
+ * Optional temporary mount options, free'd in zfs_sb_free().
+ */
+ zsb->z_mntopts = (zmo ? zmo : zfs_mntopts_alloc());
+
/*
* Initialize the zfs-specific filesystem structure.
* Should probably make this a kmem cache, shuffle fields,
*/
zsb->z_sb = NULL;
zsb->z_parent = zsb;
- zsb->z_max_blksz = SPA_MAXBLOCKSIZE;
+ zsb->z_max_blksz = SPA_OLD_MAXBLOCKSIZE;
zsb->z_show_ctldir = ZFS_SNAPDIR_VISIBLE;
zsb->z_os = os;
error = zfs_get_zplprop(os, ZFS_PROP_VERSION, &zsb->z_version);
if (error) {
goto out;
- } else if (zsb->z_version >
- zfs_zpl_version_map(spa_version(dmu_objset_spa(os)))) {
- (void) printk("Can't mount a version %lld file system "
- "on a version %lld pool\n. Pool must be upgraded to mount "
- "this file system.", (u_longlong_t)zsb->z_version,
- (u_longlong_t)spa_version(dmu_objset_spa(os)));
- error = ENOTSUP;
+ } else if (zsb->z_version > ZPL_VERSION) {
+ error = SET_ERROR(ENOTSUP);
goto out;
}
if ((error = zfs_get_zplprop(os, ZFS_PROP_NORMALIZE, &zval)) != 0)
goto out;
zsb->z_case = (uint_t)zval;
+ if ((error = zfs_get_zplprop(os, ZFS_PROP_ACLTYPE, &zval)) != 0)
+ goto out;
+ zsb->z_acl_type = (uint_t)zval;
+
/*
* Fold case on file systems that are always or sometimes case
* insensitive.
mutex_init(&zsb->z_lock, NULL, MUTEX_DEFAULT, NULL);
list_create(&zsb->z_all_znodes, sizeof (znode_t),
offsetof(znode_t, z_link_node));
- rrw_init(&zsb->z_teardown_lock);
+ rrm_init(&zsb->z_teardown_lock, B_FALSE);
rw_init(&zsb->z_teardown_inactive_lock, NULL, RW_DEFAULT, NULL);
rw_init(&zsb->z_fuid_lock, NULL, RW_DEFAULT, NULL);
+
+ zsb->z_hold_mtx = vmem_zalloc(sizeof (kmutex_t) * ZFS_OBJ_MTX_SZ,
+ KM_SLEEP);
for (i = 0; i != ZFS_OBJ_MTX_SZ; i++)
mutex_init(&zsb->z_hold_mtx[i], NULL, MUTEX_DEFAULT, NULL);
- avl_create(&zsb->z_ctldir_snaps, snapentry_compare,
- sizeof (zfs_snapentry_t), offsetof(zfs_snapentry_t, se_node));
- mutex_init(&zsb->z_ctldir_lock, NULL, MUTEX_DEFAULT, NULL);
-
*zsbp = zsb;
return (0);
out:
dmu_objset_disown(os, zsb);
*zsbp = NULL;
+
+ vmem_free(zsb->z_hold_mtx, sizeof (kmutex_t) * ZFS_OBJ_MTX_SZ);
kmem_free(zsb, sizeof (zfs_sb_t));
return (error);
}
mutex_destroy(&zsb->z_znodes_lock);
mutex_destroy(&zsb->z_lock);
list_destroy(&zsb->z_all_znodes);
- rrw_destroy(&zsb->z_teardown_lock);
+ rrm_destroy(&zsb->z_teardown_lock);
rw_destroy(&zsb->z_teardown_inactive_lock);
rw_destroy(&zsb->z_fuid_lock);
for (i = 0; i != ZFS_OBJ_MTX_SZ; i++)
mutex_destroy(&zsb->z_hold_mtx[i]);
- mutex_destroy(&zsb->z_ctldir_lock);
- avl_destroy(&zsb->z_ctldir_snaps);
+ vmem_free(zsb->z_hold_mtx, sizeof (kmutex_t) * ZFS_OBJ_MTX_SZ);
+ zfs_mntopts_free(zsb->z_mntopts);
kmem_free(zsb, sizeof (zfs_sb_t));
}
EXPORT_SYMBOL(zfs_sb_free);
VERIFY(dsl_prop_unregister(ds, "atime", atime_changed_cb,
zsb) == 0);
+ VERIFY(dsl_prop_unregister(ds, "relatime", relatime_changed_cb,
+ zsb) == 0);
+
VERIFY(dsl_prop_unregister(ds, "xattr", xattr_changed_cb,
zsb) == 0);
VERIFY(dsl_prop_unregister(ds, "snapdir", snapdir_changed_cb,
zsb) == 0);
+ VERIFY(dsl_prop_unregister(ds, "acltype", acltype_changed_cb,
+ zsb) == 0);
+
VERIFY(dsl_prop_unregister(ds, "aclinherit",
acl_inherit_changed_cb, zsb) == 0);
#ifdef HAVE_MLSLABEL
/*
- * zfs_check_global_label:
- * Check that the hex label string is appropriate for the dataset
- * being mounted into the global_zone proper.
+ * Check that the hex label string is appropriate for the dataset being
+ * mounted into the global_zone proper.
*
- * Return an error if the hex label string is not default or
- * admin_low/admin_high. For admin_low labels, the corresponding
- * dataset must be readonly.
+ * Return an error if the hex label string is not default or
+ * admin_low/admin_high. For admin_low labels, the corresponding
+ * dataset must be readonly.
*/
int
zfs_check_global_label(const char *dsname, const char *hexsl)
if (dsl_prop_get_integer(dsname,
zfs_prop_to_name(ZFS_PROP_READONLY), &rdonly, NULL))
- return (EACCES);
+ return (SET_ERROR(EACCES));
return (rdonly ? 0 : EACCES);
}
- return (EACCES);
+ return (SET_ERROR(EACCES));
}
EXPORT_SYMBOL(zfs_check_global_label);
#endif /* HAVE_MLSLABEL */
}
EXPORT_SYMBOL(zfs_root);
-#ifdef HAVE_SHRINK
+#if !defined(HAVE_SPLIT_SHRINKER_CALLBACK) && !defined(HAVE_SHRINK) && \
+ defined(HAVE_D_PRUNE_ALIASES)
+/*
+ * Linux kernels older than 3.1 do not support a per-filesystem shrinker.
+ * To accommodate this we must improvise and manually walk the list of znodes
+ * attempting to prune dentries in order to be able to drop the inodes.
+ *
+ * To avoid scanning the same znodes multiple times they are always rotated
+ * to the end of the z_all_znodes list. New znodes are inserted at the
+ * end of the list so we're always scanning the oldest znodes first.
+ */
+static int
+zfs_sb_prune_aliases(zfs_sb_t *zsb, unsigned long nr_to_scan)
+{
+ znode_t **zp_array, *zp;
+ int max_array = MIN(nr_to_scan, PAGE_SIZE * 8 / sizeof (znode_t *));
+ int objects = 0;
+ int i = 0, j = 0;
+
+ zp_array = kmem_zalloc(max_array * sizeof (znode_t *), KM_SLEEP);
+
+ mutex_enter(&zsb->z_znodes_lock);
+ while ((zp = list_head(&zsb->z_all_znodes)) != NULL) {
+
+ if ((i++ > nr_to_scan) || (j >= max_array))
+ break;
+
+ ASSERT(list_link_active(&zp->z_link_node));
+ list_remove(&zsb->z_all_znodes, zp);
+ list_insert_tail(&zsb->z_all_znodes, zp);
+
+ /* Skip active znodes and .zfs entries */
+ if (MUTEX_HELD(&zp->z_lock) || zp->z_is_ctldir)
+ continue;
+
+ if (igrab(ZTOI(zp)) == NULL)
+ continue;
+
+ zp_array[j] = zp;
+ j++;
+ }
+ mutex_exit(&zsb->z_znodes_lock);
+
+ for (i = 0; i < j; i++) {
+ zp = zp_array[i];
+
+ ASSERT3P(zp, !=, NULL);
+ d_prune_aliases(ZTOI(zp));
+
+ if (atomic_read(&ZTOI(zp)->i_count) == 1)
+ objects++;
+
+ iput(ZTOI(zp));
+ }
+
+ kmem_free(zp_array, max_array * sizeof (znode_t *));
+
+ return (objects);
+}
+#endif /* HAVE_D_PRUNE_ALIASES */
+
+/*
+ * The ARC has requested that the filesystem drop entries from the dentry
+ * and inode caches. This can occur when the ARC needs to free meta data
+ * blocks but can't because they are all pinned by entries in these caches.
+ */
int
zfs_sb_prune(struct super_block *sb, unsigned long nr_to_scan, int *objects)
{
zfs_sb_t *zsb = sb->s_fs_info;
+ int error = 0;
+#if defined(HAVE_SHRINK) || defined(HAVE_SPLIT_SHRINKER_CALLBACK)
struct shrinker *shrinker = &sb->s_shrink;
struct shrink_control sc = {
.nr_to_scan = nr_to_scan,
.gfp_mask = GFP_KERNEL,
};
+#endif
ZFS_ENTER(zsb);
+
+#if defined(HAVE_SPLIT_SHRINKER_CALLBACK) && \
+ defined(SHRINK_CONTROL_HAS_NID) && \
+ defined(SHRINKER_NUMA_AWARE)
+ if (sb->s_shrink.flags & SHRINKER_NUMA_AWARE) {
+ *objects = 0;
+ for_each_online_node(sc.nid)
+ *objects += (*shrinker->scan_objects)(shrinker, &sc);
+ } else {
+ *objects = (*shrinker->scan_objects)(shrinker, &sc);
+ }
+#elif defined(HAVE_SPLIT_SHRINKER_CALLBACK)
+ *objects = (*shrinker->scan_objects)(shrinker, &sc);
+#elif defined(HAVE_SHRINK)
*objects = (*shrinker->shrink)(shrinker, &sc);
+#elif defined(HAVE_D_PRUNE_ALIASES)
+ *objects = zfs_sb_prune_aliases(zsb, nr_to_scan);
+#else
+#error "No available dentry and inode cache pruning mechanism."
+#endif
ZFS_EXIT(zsb);
- return (0);
+ dprintf_ds(zsb->z_os->os_dsl_dataset,
+ "pruning, nr_to_scan=%lu objects=%d error=%d\n",
+ nr_to_scan, *objects, error);
+
+ return (error);
}
EXPORT_SYMBOL(zfs_sb_prune);
-#endif /* HAVE_SHRINK */
/*
* Teardown the zfs_sb_t.
{
znode_t *zp;
- rrw_enter(&zsb->z_teardown_lock, RW_WRITER, FTAG);
+ /*
+ * If someone has not already unmounted this file system,
+ * drain the iput_taskq to ensure all active references to the
+ * zfs_sb_t have been handled only then can it be safely destroyed.
+ */
+ if (zsb->z_os) {
+ /*
+ * If we're unmounting we have to wait for the list to
+ * drain completely.
+ *
+ * If we're not unmounting there's no guarantee the list
+ * will drain completely, but iputs run from the taskq
+ * may add the parents of dir-based xattrs to the taskq
+ * so we want to wait for these.
+ *
+ * We can safely read z_nr_znodes without locking because the
+ * VFS has already blocked operations which add to the
+ * z_all_znodes list and thus increment z_nr_znodes.
+ */
+ int round = 0;
+ while (zsb->z_nr_znodes > 0) {
+ taskq_wait_outstanding(dsl_pool_iput_taskq(
+ dmu_objset_pool(zsb->z_os)), 0);
+ if (++round > 1 && !unmounting)
+ break;
+ }
+ }
+
+ rrm_enter(&zsb->z_teardown_lock, RW_WRITER, FTAG);
if (!unmounting) {
/*
shrink_dcache_sb(zsb->z_parent->z_sb);
}
- /*
- * If someone has not already unmounted this file system,
- * drain the iput_taskq to ensure all active references to the
- * zfs_sb_t have been handled only then can it be safely destroyed.
- */
- if (zsb->z_os)
- taskq_wait(dsl_pool_iput_taskq(dmu_objset_pool(zsb->z_os)));
-
/*
* Close the zil. NB: Can't close the zil while zfs_inactive
* threads are blocked as zil_close can call zfs_inactive.
*/
if (!unmounting && (zsb->z_unmounted || zsb->z_os == NULL)) {
rw_exit(&zsb->z_teardown_inactive_lock);
- rrw_exit(&zsb->z_teardown_lock, FTAG);
- return (EIO);
+ rrm_exit(&zsb->z_teardown_lock, FTAG);
+ return (SET_ERROR(EIO));
}
/*
*
* Release all holds on dbufs.
*/
- mutex_enter(&zsb->z_znodes_lock);
- for (zp = list_head(&zsb->z_all_znodes); zp != NULL;
- zp = list_next(&zsb->z_all_znodes, zp)) {
- if (zp->z_sa_hdl) {
- ASSERT(atomic_read(&ZTOI(zp)->i_count) > 0);
- zfs_znode_dmu_fini(zp);
+ if (!unmounting) {
+ mutex_enter(&zsb->z_znodes_lock);
+ for (zp = list_head(&zsb->z_all_znodes); zp != NULL;
+ zp = list_next(&zsb->z_all_znodes, zp)) {
+ if (zp->z_sa_hdl)
+ zfs_znode_dmu_fini(zp);
}
+ mutex_exit(&zsb->z_znodes_lock);
}
- mutex_exit(&zsb->z_znodes_lock);
/*
* If we are unmounting, set the unmounted flag and let new VFS ops
*/
if (unmounting) {
zsb->z_unmounted = B_TRUE;
- rrw_exit(&zsb->z_teardown_lock, FTAG);
+ rrm_exit(&zsb->z_teardown_lock, FTAG);
rw_exit(&zsb->z_teardown_inactive_lock);
}
if (dsl_dataset_is_dirty(dmu_objset_ds(zsb->z_os)) &&
!zfs_is_readonly(zsb))
txg_wait_synced(dmu_objset_pool(zsb->z_os), 0);
- (void) dmu_objset_evict_dbufs(zsb->z_os);
+ dmu_objset_evict_dbufs(zsb->z_os);
return (0);
}
EXPORT_SYMBOL(zfs_sb_teardown);
-#if defined(HAVE_BDI) && !defined(HAVE_BDI_SETUP_AND_REGISTER)
+#if !defined(HAVE_2ARGS_BDI_SETUP_AND_REGISTER) && \
+ !defined(HAVE_3ARGS_BDI_SETUP_AND_REGISTER)
atomic_long_t zfs_bdi_seq = ATOMIC_LONG_INIT(0);
-#endif /* HAVE_BDI && !HAVE_BDI_SETUP_AND_REGISTER */
+#endif
int
-zfs_domount(struct super_block *sb, void *data, int silent)
+zfs_domount(struct super_block *sb, zfs_mntopts_t *zmo, int silent)
{
- zpl_mount_data_t *zmd = data;
- const char *osname = zmd->z_osname;
+ const char *osname = zmo->z_osname;
zfs_sb_t *zsb;
struct inode *root_inode;
uint64_t recordsize;
int error;
- error = zfs_sb_create(osname, &zsb);
+ error = zfs_sb_create(osname, zmo, &zsb);
if (error)
return (error);
sb->s_time_gran = 1;
sb->s_blocksize = recordsize;
sb->s_blocksize_bits = ilog2(recordsize);
-
-#ifdef HAVE_BDI
- /*
- * 2.6.32 API change,
- * Added backing_device_info (BDI) per super block interfaces. A BDI
- * must be configured when using a non-device backed filesystem for
- * proper writeback. This is not required for older pdflush kernels.
- *
- * NOTE: Linux read-ahead is disabled in favor of zfs read-ahead.
- */
zsb->z_bdi.ra_pages = 0;
sb->s_bdi = &zsb->z_bdi;
- error = -bdi_setup_and_register(&zsb->z_bdi, "zfs", BDI_CAP_MAP_COPY);
+ error = -zpl_bdi_setup_and_register(&zsb->z_bdi, "zfs");
if (error)
goto out;
-#endif /* HAVE_BDI */
/* Set callback operations for the file system. */
sb->s_op = &zpl_super_operations;
atime_changed_cb(zsb, B_FALSE);
readonly_changed_cb(zsb, B_TRUE);
- if ((error = dsl_prop_get_integer(osname,"xattr",&pval,NULL)))
+ if ((error = dsl_prop_get_integer(osname,
+ "xattr", &pval, NULL)))
goto out;
xattr_changed_cb(zsb, pval);
+ if ((error = dsl_prop_get_integer(osname,
+ "acltype", &pval, NULL)))
+ goto out;
+ acltype_changed_cb(zsb, pval);
zsb->z_issnap = B_TRUE;
zsb->z_os->os_sync = ZFS_SYNC_DISABLED;
+ zsb->z_snap_defer_time = jiffies;
mutex_enter(&zsb->z_os->os_user_ptr_lock);
dmu_objset_set_user(zsb->z_os, zsb);
sb->s_root = d_make_root(root_inode);
if (sb->s_root == NULL) {
(void) zfs_umount(sb);
- error = ENOMEM;
+ error = SET_ERROR(ENOMEM);
goto out;
}
if (!zsb->z_issnap)
zfsctl_create(zsb);
+
+ zsb->z_arc_prune = arc_add_prune_callback(zpl_prune_sb, sb);
out:
if (error) {
dmu_objset_disown(zsb->z_os, zsb);
{
zfs_sb_t *zsb = sb->s_fs_info;
- if (zsb != NULL && zsb->z_ctldir != NULL)
- zfsctl_destroy(zsb);
+ if (zsb)
+ zfsctl_destroy(sb->s_fs_info);
}
EXPORT_SYMBOL(zfs_preumount);
zfs_sb_t *zsb = sb->s_fs_info;
objset_t *os;
+ arc_remove_prune_callback(zsb->z_arc_prune);
VERIFY(zfs_sb_teardown(zsb, B_TRUE) == 0);
os = zsb->z_os;
-
-#ifdef HAVE_BDI
bdi_destroy(sb->s_bdi);
-#endif /* HAVE_BDI */
/*
* z_os will be NULL if there was an error in
EXPORT_SYMBOL(zfs_umount);
int
-zfs_remount(struct super_block *sb, int *flags, char *data)
+zfs_remount(struct super_block *sb, int *flags, zfs_mntopts_t *zmo)
{
- /*
- * All namespace flags (MNT_*) and super block flags (MS_*) will
- * be handled by the Linux VFS. Only handle custom options here.
- */
- return (0);
+ zfs_sb_t *zsb = sb->s_fs_info;
+ int error;
+
+ zfs_unregister_callbacks(zsb);
+ error = zfs_register_callbacks(zsb);
+
+ return (error);
}
EXPORT_SYMBOL(zfs_remount);
err = zfsctl_lookup_objset(sb, objsetid, &zsb);
if (err)
- return (EINVAL);
+ return (SET_ERROR(EINVAL));
ZFS_ENTER(zsb);
}
fid_gen |= ((uint64_t)zfid->zf_gen[i]) << (8 * i);
} else {
ZFS_EXIT(zsb);
- return (EINVAL);
+ return (SET_ERROR(EINVAL));
}
/* A zero fid_gen means we are in the .zfs control directories */
gen_mask = -1ULL >> (64 - 8 * i);
- dprintf("getting %llu [%u mask %llx]\n", object, fid_gen, gen_mask);
+ dprintf("getting %llu [%llu mask %llx]\n", object, fid_gen, gen_mask);
if ((err = zfs_zget(zsb, object, &zp))) {
ZFS_EXIT(zsb);
return (err);
zp_gen = zp_gen & gen_mask;
if (zp_gen == 0)
zp_gen = 1;
+ if ((fid_gen == 0) && (zsb->z_root == object))
+ fid_gen = zp_gen;
if (zp->z_unlinked || zp_gen != fid_gen) {
- dprintf("znode gen (%u) != fid gen (%u)\n", zp_gen, fid_gen);
+ dprintf("znode gen (%llu) != fid gen (%llu)\n", zp_gen,
+ fid_gen);
iput(ZTOI(zp));
ZFS_EXIT(zsb);
- return (EINVAL);
+ return (SET_ERROR(EINVAL));
}
*ipp = ZTOI(zp);
* Block out VFS ops and close zfs_sb_t
*
* Note, if successful, then we return with the 'z_teardown_lock' and
- * 'z_teardown_inactive_lock' write held.
+ * 'z_teardown_inactive_lock' write held. We leave ownership of the underlying
+ * dataset and objset intact so that they can be atomically handed off during
+ * a subsequent rollback or recv operation and the resume thereafter.
*/
int
zfs_suspend_fs(zfs_sb_t *zsb)
if ((error = zfs_sb_teardown(zsb, B_FALSE)) != 0)
return (error);
- dmu_objset_disown(zsb->z_os, zsb);
-
return (0);
}
EXPORT_SYMBOL(zfs_suspend_fs);
zfs_resume_fs(zfs_sb_t *zsb, const char *osname)
{
int err, err2;
+ znode_t *zp;
+ uint64_t sa_obj = 0;
- ASSERT(RRW_WRITE_HELD(&zsb->z_teardown_lock));
+ ASSERT(RRM_WRITE_HELD(&zsb->z_teardown_lock));
ASSERT(RW_WRITE_HELD(&zsb->z_teardown_inactive_lock));
- err = dmu_objset_own(osname, DMU_OST_ZFS, B_FALSE, zsb, &zsb->z_os);
- if (err) {
- zsb->z_os = NULL;
- } else {
- znode_t *zp;
- uint64_t sa_obj = 0;
+ /*
+ * We already own this, so just hold and rele it to update the
+ * objset_t, as the one we had before may have been evicted.
+ */
+ VERIFY0(dmu_objset_hold(osname, zsb, &zsb->z_os));
+ VERIFY3P(zsb->z_os->os_dsl_dataset->ds_owner, ==, zsb);
+ VERIFY(dsl_dataset_long_held(zsb->z_os->os_dsl_dataset));
+ dmu_objset_rele(zsb->z_os, zsb);
- err2 = zap_lookup(zsb->z_os, MASTER_NODE_OBJ,
- ZFS_SA_ATTRS, 8, 1, &sa_obj);
+ /*
+ * Make sure version hasn't changed
+ */
- if ((err || err2) && zsb->z_version >= ZPL_VERSION_SA)
- goto bail;
+ err = zfs_get_zplprop(zsb->z_os, ZFS_PROP_VERSION,
+ &zsb->z_version);
+ if (err)
+ goto bail;
- if ((err = sa_setup(zsb->z_os, sa_obj,
- zfs_attr_table, ZPL_END, &zsb->z_attr_table)) != 0)
- goto bail;
+ err = zap_lookup(zsb->z_os, MASTER_NODE_OBJ,
+ ZFS_SA_ATTRS, 8, 1, &sa_obj);
- VERIFY(zfs_sb_setup(zsb, B_FALSE) == 0);
- zsb->z_rollback_time = jiffies;
+ if (err && zsb->z_version >= ZPL_VERSION_SA)
+ goto bail;
- /*
- * Attempt to re-establish all the active inodes with their
- * dbufs. If a zfs_rezget() fails, then we unhash the inode
- * and mark it stale. This prevents a collision if a new
- * inode/object is created which must use the same inode
- * number. The stale inode will be be released when the
- * VFS prunes the dentry holding the remaining references
- * on the stale inode.
- */
- mutex_enter(&zsb->z_znodes_lock);
- for (zp = list_head(&zsb->z_all_znodes); zp;
- zp = list_next(&zsb->z_all_znodes, zp)) {
- err2 = zfs_rezget(zp);
- if (err2) {
- remove_inode_hash(ZTOI(zp));
- zp->z_is_stale = B_TRUE;
- }
+ if ((err = sa_setup(zsb->z_os, sa_obj,
+ zfs_attr_table, ZPL_END, &zsb->z_attr_table)) != 0)
+ goto bail;
+
+ if (zsb->z_version >= ZPL_VERSION_SA)
+ sa_register_update_callback(zsb->z_os,
+ zfs_sa_upgrade);
+
+ VERIFY(zfs_sb_setup(zsb, B_FALSE) == 0);
+
+ zfs_set_fuid_feature(zsb);
+ zsb->z_rollback_time = jiffies;
+
+ /*
+ * Attempt to re-establish all the active inodes with their
+ * dbufs. If a zfs_rezget() fails, then we unhash the inode
+ * and mark it stale. This prevents a collision if a new
+ * inode/object is created which must use the same inode
+ * number. The stale inode will be be released when the
+ * VFS prunes the dentry holding the remaining references
+ * on the stale inode.
+ */
+ mutex_enter(&zsb->z_znodes_lock);
+ for (zp = list_head(&zsb->z_all_znodes); zp;
+ zp = list_next(&zsb->z_all_znodes, zp)) {
+ err2 = zfs_rezget(zp);
+ if (err2) {
+ remove_inode_hash(ZTOI(zp));
+ zp->z_is_stale = B_TRUE;
}
- mutex_exit(&zsb->z_znodes_lock);
}
+ mutex_exit(&zsb->z_znodes_lock);
bail:
/* release the VFS ops */
rw_exit(&zsb->z_teardown_inactive_lock);
- rrw_exit(&zsb->z_teardown_lock, FTAG);
+ rrm_exit(&zsb->z_teardown_lock, FTAG);
if (err) {
/*
- * Since we couldn't reopen zfs_sb_t or, setup the
- * sa framework, force unmount this file system.
+ * Since we couldn't setup the sa framework, try to force
+ * unmount this file system.
*/
if (zsb->z_os)
(void) zfs_umount(zsb->z_sb);
dmu_tx_t *tx;
if (newvers < ZPL_VERSION_INITIAL || newvers > ZPL_VERSION)
- return (EINVAL);
+ return (SET_ERROR(EINVAL));
if (newvers < zsb->z_version)
- return (EINVAL);
+ return (SET_ERROR(EINVAL));
if (zfs_spa_version_map(newvers) >
spa_version(dmu_objset_spa(zsb->z_os)))
- return (ENOTSUP);
+ return (SET_ERROR(ENOTSUP));
tx = dmu_tx_create(os);
dmu_tx_hold_zap(tx, MASTER_NODE_OBJ, B_FALSE, ZPL_VERSION_STR);
sa_register_update_callback(os, zfs_sa_upgrade);
}
- spa_history_log_internal(LOG_DS_UPGRADE,
- dmu_objset_spa(os), tx, "oldver=%llu newver=%llu dataset = %llu",
- zsb->z_version, newvers, dmu_objset_id(os));
+ spa_history_log_internal_ds(dmu_objset_ds(os), "upgrade", tx,
+ "from %llu to %llu", zsb->z_version, newvers);
dmu_tx_commit(tx);
zsb->z_version = newvers;
- if (zsb->z_version >= ZPL_VERSION_FUID)
- zfs_set_fuid_feature(zsb);
+ zfs_set_fuid_feature(zsb);
return (0);
}
zfs_get_zplprop(objset_t *os, zfs_prop_t prop, uint64_t *value)
{
const char *pname;
- int error = ENOENT;
+ int error = SET_ERROR(ENOENT);
/*
* Look up the file system's value for the property. For the
case ZFS_PROP_CASE:
*value = ZFS_CASE_SENSITIVE;
break;
+ case ZFS_PROP_ACLTYPE:
+ *value = ZFS_ACLTYPE_OFF;
+ break;
default:
return (error);
}
zfs_znode_init();
dmu_objset_register_type(DMU_OST_ZFS, zfs_space_delta_cb);
register_filesystem(&zpl_fs_type);
- (void) arc_add_prune_callback(zpl_prune_sbs, NULL);
}
void
zfs_fini(void)
{
+ taskq_wait_outstanding(system_taskq, 0);
unregister_filesystem(&zpl_fs_type);
zfs_znode_fini();
zfsctl_fini();