[PATCH] make dm-mirror not issue invalid resync requests
[pandora-kernel.git] / drivers / md / dm.c
index ea180af..0e48151 100644 (file)
@@ -55,10 +55,11 @@ union map_info *dm_get_mapinfo(struct bio *bio)
  */
 #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;
 
@@ -97,7 +98,7 @@ struct mapped_device {
         * 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
@@ -248,16 +249,16 @@ static inline void free_tio(struct mapped_device *md, struct target_io *tio)
  */
 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 */
 }
 
@@ -384,7 +385,7 @@ static void __map_bio(struct dm_target *ti, struct bio *clone,
                /* 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);
        }
 }
@@ -399,6 +400,11 @@ struct clone_info {
        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.
  */
@@ -410,6 +416,7 @@ static struct bio *split_bvec(struct bio *bio, sector_t sector,
        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;
@@ -568,14 +575,14 @@ static int dm_request(request_queue_t *q, struct bio *bio)
        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);
@@ -594,11 +601,11 @@ static int dm_request(request_queue_t *q, struct bio *bio)
                 * 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;
 }
 
@@ -610,7 +617,7 @@ static int dm_flush_all(request_queue_t *q, struct gendisk *disk,
        int ret = -ENXIO;
 
        if (map) {
-               ret = dm_table_flush_all(md->map);
+               ret = dm_table_flush_all(map);
                dm_table_put(map);
        }
 
@@ -747,7 +754,8 @@ static struct mapped_device *alloc_dev(unsigned int minor, int persistent)
                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);
@@ -825,18 +833,13 @@ static void event_callback(void *context)
        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)
@@ -845,17 +848,18 @@ 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;
 }
 
@@ -899,10 +903,9 @@ int dm_create_with_minor(unsigned int minor, struct mapped_device **result)
        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))
@@ -911,12 +914,32 @@ void *dm_get_mdptr(dev_t dev)
        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;
 }
 
@@ -935,7 +958,7 @@ void dm_put(struct mapped_device *md)
        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);
                }
@@ -966,61 +989,55 @@ static void __flush_deferred_io(struct mapped_device *md, struct bio *c)
  */
 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)
 {
-       if (test_and_set_bit(DMF_FS_LOCKED, &md->flags))
-               return 0;
+       int r;
 
-       md->frozen_bdev = bdget_disk(md->disk, 0);
-       if (!md->frozen_bdev) {
-               DMWARN("bdget failed in __lock_fs");
-               return -ENOMEM;
+       WARN_ON(md->frozen_sb);
+
+       md->frozen_sb = freeze_bdev(md->suspended_bdev);
+       if (IS_ERR(md->frozen_sb)) {
+               r = PTR_ERR(md->frozen_sb);
+               md->frozen_sb = NULL;
+               return r;
        }
 
-       WARN_ON(md->frozen_sb);
-       md->frozen_sb = freeze_bdev(md->frozen_bdev);
+       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;
 }
 
-static int __unlock_fs(struct mapped_device *md)
+static void unlock_fs(struct mapped_device *md)
 {
-       if (!test_and_clear_bit(DMF_FS_LOCKED, &md->flags))
-               return 0;
-
-       thaw_bdev(md->frozen_bdev, md->frozen_sb);
-       bdput(md->frozen_bdev);
+       if (!test_bit(DMF_FROZEN, &md->flags))
+               return;
 
+       thaw_bdev(md->suspended_bdev, md->frozen_sb);
        md->frozen_sb = NULL;
-       md->frozen_bdev = NULL;
-
-       return 0;
+       clear_bit(DMF_FROZEN, &md->flags);
 }
 
 /*
@@ -1030,49 +1047,48 @@ static int __unlock_fs(struct mapped_device *md)
  * 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
@@ -1088,54 +1104,75 @@ int dm_suspend(struct mapped_device *md)
        }
        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;
 }
 
 /*-----------------------------------------------------------------