/*
* Called when an unmount is requested and certain sanity checks have
* already passed. At this point no dentries or inodes have been reclaimed
* from their respective caches. We drop the extra reference on the .zfs
* control directory to allow everything to be reclaimed. All snapshots
* must already have been unmounted to reach this point.
*/
void
zfs_preumount(struct super_block *sb)
{
zfs_sb_t *zsb = sb->s_fs_info;
/* zsb is NULL when zfs_domount fails during mount */
if (zsb) {
zfsctl_destroy(sb->s_fs_info);
/*
* Wait for iput_async before entering evict_inodes in
* generic_shutdown_super. The reason we must finish before
* evict_inodes is when lazytime is on, or when zfs_purgedir
* calls zfs_zget, iput would bump i_count from 0 to 1. This
* would race with the i_count check in evict_inodes. This means
* it could destroy the inode while we are still using it.
*
* We wait for two passes. xattr directories in the first pass
* may add xattr entries in zfs_purgedir, so in the second pass
* we wait for them. We don't use taskq_wait here because it is
* a pool wide taskq. Other mounted filesystems can constantly
* do iput_async and there's no guarantee when taskq will be
* empty.
*/
taskq_wait_outstanding(dsl_pool_iput_taskq(
dmu_objset_pool(zsb->z_os)), 0);
taskq_wait_outstanding(dsl_pool_iput_taskq(
dmu_objset_pool(zsb->z_os)), 0);
}
}
static int
splat_taskq_test7_impl(struct file *file, void *arg, boolean_t prealloc)
{
taskq_t *tq;
splat_taskq_arg_t *tq_arg;
taskq_ent_t *tqe;
int error;
splat_vprint(file, SPLAT_TASKQ_TEST7_NAME,
"Taskq '%s' creating (%s dispatch)\n",
SPLAT_TASKQ_TEST7_NAME,
prealloc ? "prealloc" : "dynamic");
if ((tq = taskq_create(SPLAT_TASKQ_TEST7_NAME, 1, maxclsyspri,
50, INT_MAX, TASKQ_PREPOPULATE)) == NULL) {
splat_vprint(file, SPLAT_TASKQ_TEST7_NAME,
"Taskq '%s' create failed\n",
SPLAT_TASKQ_TEST7_NAME);
return -EINVAL;
}
tq_arg = kmem_alloc(sizeof (splat_taskq_arg_t), KM_SLEEP);
tqe = kmem_alloc(sizeof (taskq_ent_t), KM_SLEEP);
tq_arg->depth = 0;
tq_arg->flag = 0;
tq_arg->id = 0;
tq_arg->file = file;
tq_arg->name = SPLAT_TASKQ_TEST7_NAME;
tq_arg->tq = tq;
if (prealloc) {
taskq_init_ent(tqe);
tq_arg->tqe = tqe;
} else {
tq_arg->tqe = NULL;
}
splat_taskq_test7_func(tq_arg);
if (tq_arg->flag == 0) {
splat_vprint(file, SPLAT_TASKQ_TEST7_NAME,
"Taskq '%s' waiting\n", tq_arg->name);
taskq_wait_outstanding(tq, SPLAT_TASKQ_DEPTH_MAX);
}
error = (tq_arg->depth == SPLAT_TASKQ_DEPTH_MAX ? 0 : -EINVAL);
kmem_free(tqe, sizeof (taskq_ent_t));
kmem_free(tq_arg, sizeof (splat_taskq_arg_t));
splat_vprint(file, SPLAT_TASKQ_TEST7_NAME,
"Taskq '%s' destroying\n", tq_arg->name);
taskq_destroy(tq);
return (error);
}
/*
* Teardown the zfs_sb_t.
*
* Note, if 'unmounting' if FALSE, we return with the 'z_teardown_lock'
* and 'z_teardown_inactive_lock' held.
*/
int
zfs_sb_teardown(zfs_sb_t *zsb, boolean_t unmounting)
{
znode_t *zp;
/*
* 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) {
/*
* We purge the parent filesystem's super block as the
* parent filesystem and all of its snapshots have their
* inode's super block set to the parent's filesystem's
* super block. Note, 'z_parent' is self referential
* for non-snapshots.
*/
shrink_dcache_sb(zsb->z_parent->z_sb);
}
/*
* Close the zil. NB: Can't close the zil while zfs_inactive
* threads are blocked as zil_close can call zfs_inactive.
*/
if (zsb->z_log) {
zil_close(zsb->z_log);
zsb->z_log = NULL;
}
rw_enter(&zsb->z_teardown_inactive_lock, RW_WRITER);
/*
* If we are not unmounting (ie: online recv) and someone already
* unmounted this file system while we were doing the switcheroo,
* or a reopen of z_os failed then just bail out now.
*/
if (!unmounting && (zsb->z_unmounted || zsb->z_os == NULL)) {
rw_exit(&zsb->z_teardown_inactive_lock);
rrm_exit(&zsb->z_teardown_lock, FTAG);
return (SET_ERROR(EIO));
}
/*
* At this point there are no VFS ops active, and any new VFS ops
* will fail with EIO since we have z_teardown_lock for writer (only
* relevant for forced unmount).
*
* Release all holds on dbufs.
*/
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);
}
/*
* If we are unmounting, set the unmounted flag and let new VFS ops
* unblock. zfs_inactive will have the unmounted behavior, and all
* other VFS ops will fail with EIO.
*/
if (unmounting) {
zsb->z_unmounted = B_TRUE;
rrm_exit(&zsb->z_teardown_lock, FTAG);
rw_exit(&zsb->z_teardown_inactive_lock);
}
/*
* z_os will be NULL if there was an error in attempting to reopen
* zsb, so just return as the properties had already been
*
* unregistered and cached data had been evicted before.
*/
if (zsb->z_os == NULL)
return (0);
/*
* Unregister properties.
*/
zfs_unregister_callbacks(zsb);
/*
* Evict cached data
*/
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);
dmu_objset_evict_dbufs(zsb->z_os);
return (0);
}
static int
splat_taskq_test6_impl(struct file *file, void *arg, boolean_t prealloc)
{
taskq_t *tq;
taskqid_t id;
splat_taskq_id_t tq_id[SPLAT_TASKQ_ORDER_MAX];
splat_taskq_arg_t tq_arg;
int order[SPLAT_TASKQ_ORDER_MAX] = { 1,2,3,6,7,8,4,5 };
taskq_ent_t *tqes;
int i, rc = 0;
uint_t tflags;
tqes = kmem_alloc(sizeof(*tqes) * SPLAT_TASKQ_ORDER_MAX, KM_SLEEP);
memset(tqes, 0, sizeof(*tqes) * SPLAT_TASKQ_ORDER_MAX);
splat_vprint(file, SPLAT_TASKQ_TEST6_NAME,
"Taskq '%s' creating (%s dispatch)\n",
SPLAT_TASKQ_TEST6_NAME,
prealloc ? "prealloc" : "dynamic");
if ((tq = taskq_create(SPLAT_TASKQ_TEST6_NAME, 3, maxclsyspri,
50, INT_MAX, TASKQ_PREPOPULATE)) == NULL) {
splat_vprint(file, SPLAT_TASKQ_TEST6_NAME,
"Taskq '%s' create failed\n",
SPLAT_TASKQ_TEST6_NAME);
return -EINVAL;
}
tq_arg.flag = 0;
memset(&tq_arg.order, 0, sizeof(int) * SPLAT_TASKQ_ORDER_MAX);
spin_lock_init(&tq_arg.lock);
tq_arg.file = file;
tq_arg.name = SPLAT_TASKQ_TEST6_NAME;
for (i = 0; i < SPLAT_TASKQ_ORDER_MAX; i++) {
taskq_init_ent(&tqes[i]);
tq_id[i].id = i + 1;
tq_id[i].arg = &tq_arg;
tflags = TQ_SLEEP;
if (i > 4)
tflags |= TQ_FRONT;
if (prealloc) {
taskq_dispatch_ent(tq, splat_taskq_test6_func,
&tq_id[i], tflags, &tqes[i]);
id = tqes[i].tqent_id;
} else {
id = taskq_dispatch(tq, splat_taskq_test6_func,
&tq_id[i], tflags);
}
if (id == 0) {
splat_vprint(file, SPLAT_TASKQ_TEST6_NAME,
"Taskq '%s' function '%s' dispatch failed\n",
tq_arg.name, sym2str(splat_taskq_test6_func));
rc = -EINVAL;
goto out;
}
if (tq_id[i].id != id) {
splat_vprint(file, SPLAT_TASKQ_TEST6_NAME,
"Taskq '%s' expected taskqid %d got %d\n",
tq_arg.name, (int)tq_id[i].id, (int)id);
rc = -EINVAL;
goto out;
}
/* Sleep to let tasks 1-3 start executing. */
if ( i == 2 )
msleep(100);
}
splat_vprint(file, SPLAT_TASKQ_TEST6_NAME, "Taskq '%s' "
"waiting for taskqid %d completion\n", tq_arg.name,
SPLAT_TASKQ_ORDER_MAX);
taskq_wait_outstanding(tq, SPLAT_TASKQ_ORDER_MAX);
rc = splat_taskq_test_order(&tq_arg, order);
out:
splat_vprint(file, SPLAT_TASKQ_TEST6_NAME,
"Taskq '%s' destroying\n", tq_arg.name);
taskq_destroy(tq);
kmem_free(tqes, sizeof(*tqes) * SPLAT_TASKQ_ORDER_MAX);
return rc;
}
