*/
#define DMF_BLOCK_IO 0
#define DMF_SUSPENDED 1
-#define DMF_FS_LOCKED 2
+#define DMF_FROZEN 2
struct mapped_device {
- struct rw_semaphore lock;
+ struct rw_semaphore io_lock;
+ struct semaphore suspend_lock;
rwlock_t map_lock;
atomic_t holders;
* freeze/thaw support require holding onto a super block
*/
struct super_block *frozen_sb;
- struct block_device *frozen_bdev;
+ struct block_device *suspended_bdev;
};
#define MIN_IOS 256
*/
static int queue_io(struct mapped_device *md, struct bio *bio)
{
- down_write(&md->lock);
+ down_write(&md->io_lock);
if (!test_bit(DMF_BLOCK_IO, &md->flags)) {
- up_write(&md->lock);
+ up_write(&md->io_lock);
return 1;
}
bio_list_add(&md->deferred, bio);
- up_write(&md->lock);
+ up_write(&md->io_lock);
return 0; /* deferred successfully */
}
/* error the io and bail out */
struct dm_io *io = tio->io;
free_tio(tio->io->md, tio);
- dec_pending(io, -EIO);
+ dec_pending(io, r);
bio_put(clone);
}
}
unsigned short idx;
};
+static void dm_bio_destructor(struct bio *bio)
+{
+ bio_free(bio, dm_set);
+}
+
/*
* Creates a little bio that is just does part of a bvec.
*/
struct bio_vec *bv = bio->bi_io_vec + idx;
clone = bio_alloc_bioset(GFP_NOIO, 1, dm_set);
+ clone->bi_destructor = dm_bio_destructor;
*clone->bi_io_vec = *bv;
clone->bi_sector = sector;
int r;
struct mapped_device *md = q->queuedata;
- down_read(&md->lock);
+ down_read(&md->io_lock);
/*
* If we're suspended we have to queue
* this io for later.
*/
while (test_bit(DMF_BLOCK_IO, &md->flags)) {
- up_read(&md->lock);
+ up_read(&md->io_lock);
if (bio_rw(bio) == READA) {
bio_io_error(bio, bio->bi_size);
* We're in a while loop, because someone could suspend
* before we get to the following read lock.
*/
- down_read(&md->lock);
+ down_read(&md->io_lock);
}
__split_bio(md, bio);
- up_read(&md->lock);
+ up_read(&md->io_lock);
return 0;
}
int ret = -ENXIO;
if (map) {
- ret = dm_table_flush_all(md->map);
+ ret = dm_table_flush_all(map);
dm_table_put(map);
}
goto bad1;
memset(md, 0, sizeof(*md));
- init_rwsem(&md->lock);
+ init_rwsem(&md->io_lock);
+ init_MUTEX(&md->suspend_lock);
rwlock_init(&md->map_lock);
atomic_set(&md->holders, 1);
atomic_set(&md->event_nr, 0);
wake_up(&md->eventq);
}
-static void __set_size(struct gendisk *disk, sector_t size)
+static void __set_size(struct mapped_device *md, sector_t size)
{
- struct block_device *bdev;
-
- set_capacity(disk, size);
- bdev = bdget_disk(disk, 0);
- if (bdev) {
- down(&bdev->bd_inode->i_sem);
- i_size_write(bdev->bd_inode, (loff_t)size << SECTOR_SHIFT);
- up(&bdev->bd_inode->i_sem);
- bdput(bdev);
- }
+ set_capacity(md->disk, size);
+
+ down(&md->suspended_bdev->bd_inode->i_sem);
+ i_size_write(md->suspended_bdev->bd_inode, (loff_t)size << SECTOR_SHIFT);
+ up(&md->suspended_bdev->bd_inode->i_sem);
}
static int __bind(struct mapped_device *md, struct dm_table *t)
sector_t size;
size = dm_table_get_size(t);
- __set_size(md->disk, size);
+ __set_size(md, size);
if (size == 0)
return 0;
+ dm_table_get(t);
+ dm_table_event_callback(t, event_callback, md);
+
write_lock(&md->map_lock);
md->map = t;
+ dm_table_set_restrictions(t, q);
write_unlock(&md->map_lock);
- dm_table_get(t);
- dm_table_event_callback(md->map, event_callback, md);
- dm_table_set_restrictions(t, q);
return 0;
}
return create_aux(minor, 1, result);
}
-void *dm_get_mdptr(dev_t dev)
+static struct mapped_device *dm_find_md(dev_t dev)
{
struct mapped_device *md;
- void *mdptr = NULL;
unsigned minor = MINOR(dev);
if (MAJOR(dev) != _major || minor >= (1 << MINORBITS))
down(&_minor_lock);
md = idr_find(&_minor_idr, minor);
-
- if (md && (dm_disk(md)->first_minor == minor))
- mdptr = md->interface_ptr;
+ if (!md || (dm_disk(md)->first_minor != minor))
+ md = NULL;
up(&_minor_lock);
+ return md;
+}
+
+struct mapped_device *dm_get_md(dev_t dev)
+{
+ struct mapped_device *md = dm_find_md(dev);
+
+ if (md)
+ dm_get(md);
+
+ return md;
+}
+
+void *dm_get_mdptr(dev_t dev)
+{
+ struct mapped_device *md;
+ void *mdptr = NULL;
+
+ md = dm_find_md(dev);
+ if (md)
+ mdptr = md->interface_ptr;
return mdptr;
}
struct dm_table *map = dm_get_table(md);
if (atomic_dec_and_test(&md->holders)) {
- if (!test_bit(DMF_SUSPENDED, &md->flags) && map) {
+ if (!dm_suspended(md)) {
dm_table_presuspend_targets(map);
dm_table_postsuspend_targets(map);
}
*/
int dm_swap_table(struct mapped_device *md, struct dm_table *table)
{
- int r;
+ int r = -EINVAL;
- down_write(&md->lock);
+ down(&md->suspend_lock);
/* device must be suspended */
- if (!test_bit(DMF_SUSPENDED, &md->flags)) {
- up_write(&md->lock);
- return -EPERM;
- }
+ if (!dm_suspended(md))
+ goto out;
__unbind(md);
r = __bind(md, table);
- if (r)
- return r;
- up_write(&md->lock);
- return 0;
+out:
+ up(&md->suspend_lock);
+ return r;
}
/*
* Functions to lock and unlock any filesystem running on the
* device.
*/
-static int __lock_fs(struct mapped_device *md)
+static int lock_fs(struct mapped_device *md)
{
- int error = -ENOMEM;
-
- if (test_and_set_bit(DMF_FS_LOCKED, &md->flags))
- return 0;
-
- md->frozen_bdev = bdget_disk(md->disk, 0);
- if (!md->frozen_bdev) {
- DMWARN("bdget failed in __lock_fs");
- goto out;
- }
+ int r;
WARN_ON(md->frozen_sb);
- md->frozen_sb = freeze_bdev(md->frozen_bdev);
+ md->frozen_sb = freeze_bdev(md->suspended_bdev);
if (IS_ERR(md->frozen_sb)) {
- error = PTR_ERR(md->frozen_sb);
- goto out_bdput;
+ r = PTR_ERR(md->frozen_sb);
+ md->frozen_sb = NULL;
+ return r;
}
+ set_bit(DMF_FROZEN, &md->flags);
+
/* don't bdput right now, we don't want the bdev
- * to go away while it is locked. We'll bdput
- * in __unlock_fs
+ * to go away while it is locked.
*/
return 0;
-
-out_bdput:
- bdput(md->frozen_bdev);
- md->frozen_sb = NULL;
- md->frozen_bdev = NULL;
-out:
- clear_bit(DMF_FS_LOCKED, &md->flags);
- return error;
}
-static void __unlock_fs(struct mapped_device *md)
+static void unlock_fs(struct mapped_device *md)
{
- if (!test_and_clear_bit(DMF_FS_LOCKED, &md->flags))
+ if (!test_bit(DMF_FROZEN, &md->flags))
return;
- thaw_bdev(md->frozen_bdev, md->frozen_sb);
- bdput(md->frozen_bdev);
-
+ thaw_bdev(md->suspended_bdev, md->frozen_sb);
md->frozen_sb = NULL;
- md->frozen_bdev = NULL;
+ clear_bit(DMF_FROZEN, &md->flags);
}
/*
* dm_bind_table, dm_suspend must be called to flush any in
* flight bios and ensure that any further io gets deferred.
*/
-int dm_suspend(struct mapped_device *md)
+int dm_suspend(struct mapped_device *md, int do_lockfs)
{
- struct dm_table *map;
+ struct dm_table *map = NULL;
DECLARE_WAITQUEUE(wait, current);
+ int r = -EINVAL;
- /* Flush I/O to the device. */
- down_read(&md->lock);
- if (test_bit(DMF_BLOCK_IO, &md->flags)) {
- up_read(&md->lock);
- return -EINVAL;
- }
+ down(&md->suspend_lock);
+
+ if (dm_suspended(md))
+ goto out;
map = dm_get_table(md);
- if (map)
- dm_table_presuspend_targets(map);
- __lock_fs(md);
- up_read(&md->lock);
+ /* This does not get reverted if there's an error later. */
+ dm_table_presuspend_targets(map);
- /*
- * First we set the BLOCK_IO flag so no more ios will be
- * mapped.
- */
- down_write(&md->lock);
- if (test_bit(DMF_BLOCK_IO, &md->flags)) {
- /*
- * If we get here we know another thread is
- * trying to suspend as well, so we leave the fs
- * locked for this thread.
- */
- up_write(&md->lock);
- return -EINVAL;
+ md->suspended_bdev = bdget_disk(md->disk, 0);
+ if (!md->suspended_bdev) {
+ DMWARN("bdget failed in dm_suspend");
+ r = -ENOMEM;
+ goto out;
+ }
+
+ /* Flush I/O to the device. */
+ if (do_lockfs) {
+ r = lock_fs(md);
+ if (r)
+ goto out;
}
+ /*
+ * First we set the BLOCK_IO flag so no more ios will be mapped.
+ */
+ down_write(&md->io_lock);
set_bit(DMF_BLOCK_IO, &md->flags);
+
add_wait_queue(&md->wait, &wait);
- up_write(&md->lock);
+ up_write(&md->io_lock);
/* unplug */
- if (map) {
+ if (map)
dm_table_unplug_all(map);
- dm_table_put(map);
- }
/*
* Then we wait for the already mapped ios to
}
set_current_state(TASK_RUNNING);
- down_write(&md->lock);
+ down_write(&md->io_lock);
remove_wait_queue(&md->wait, &wait);
/* were we interrupted ? */
+ r = -EINTR;
if (atomic_read(&md->pending)) {
- __unlock_fs(md);
+ up_write(&md->io_lock);
+ unlock_fs(md);
clear_bit(DMF_BLOCK_IO, &md->flags);
- up_write(&md->lock);
- return -EINTR;
+ goto out;
}
+ up_write(&md->io_lock);
+
+ dm_table_postsuspend_targets(map);
set_bit(DMF_SUSPENDED, &md->flags);
- map = dm_get_table(md);
- if (map)
- dm_table_postsuspend_targets(map);
- dm_table_put(map);
- up_write(&md->lock);
+ r = 0;
- return 0;
+out:
+ if (r && md->suspended_bdev) {
+ bdput(md->suspended_bdev);
+ md->suspended_bdev = NULL;
+ }
+
+ dm_table_put(map);
+ up(&md->suspend_lock);
+ return r;
}
int dm_resume(struct mapped_device *md)
{
+ int r = -EINVAL;
struct bio *def;
- struct dm_table *map = dm_get_table(md);
+ struct dm_table *map = NULL;
- down_write(&md->lock);
- if (!map ||
- !test_bit(DMF_SUSPENDED, &md->flags) ||
- !dm_table_get_size(map)) {
- up_write(&md->lock);
- dm_table_put(map);
- return -EINVAL;
- }
+ down(&md->suspend_lock);
+ if (!dm_suspended(md))
+ goto out;
+
+ map = dm_get_table(md);
+ if (!map || !dm_table_get_size(map))
+ goto out;
dm_table_resume_targets(map);
- clear_bit(DMF_SUSPENDED, &md->flags);
+
+ down_write(&md->io_lock);
clear_bit(DMF_BLOCK_IO, &md->flags);
def = bio_list_get(&md->deferred);
__flush_deferred_io(md, def);
- up_write(&md->lock);
- __unlock_fs(md);
+ up_write(&md->io_lock);
+
+ unlock_fs(md);
+
+ bdput(md->suspended_bdev);
+ md->suspended_bdev = NULL;
+
+ clear_bit(DMF_SUSPENDED, &md->flags);
+
dm_table_unplug_all(map);
+
+ r = 0;
+
+out:
dm_table_put(map);
+ up(&md->suspend_lock);
- return 0;
+ return r;
}
/*-----------------------------------------------------------------