btrfs: try to reclaim some space when chunk allocation fails
authorMiao Xie <miaox@cn.fujitsu.com>
Wed, 5 Jan 2011 10:07:18 +0000 (10:07 +0000)
committerChris Mason <chris.mason@oracle.com>
Sun, 16 Jan 2011 16:30:19 +0000 (11:30 -0500)
commitd52a5b5f1fa40804f681cf9868d4a8f90661bdf3
tree93ae0faf84163748353d1edb6c5247483762f2df
parent299a08b1c34f9397797946a0fa215c5fd145c5cf
btrfs: try to reclaim some space when chunk allocation fails

We cannot write data into files when when there is tiny space in the filesystem.

Reproduce steps:
 # mkfs.btrfs /dev/sda1
 # mount /dev/sda1 /mnt
 # dd if=/dev/zero of=/mnt/tmpfile0 bs=4K count=1
 # dd if=/dev/zero of=/mnt/tmpfile1 bs=4K count=99999999999999
   (fill the filesystem)
 # umount /mnt
 # mount /dev/sda1 /mnt
 # rm -f /mnt/tmpfile0
 # dd if=/dev/zero of=/mnt/tmpfile0 bs=4K count=1
   (failed with nospec)

But if we do the last step again, we can write data successfully. The reason of
the problem is that btrfs didn't try to commit the current transaction and
reclaim some space when chunk allocation failed.

This patch fixes it by committing the current transaction to reclaim some
space when chunk allocation fails.

Signed-off-by: Miao Xie <miaox@cn.fujitsu.com>
Reviewed-by: Josef Bacik <josef@redhat.com>
Signed-off-by: Chris Mason <chris.mason@oracle.com>
fs/btrfs/extent-tree.c