btrfs: account for non-CoW'd blocks in btrfs_abort_transaction
authorJeff Mahoney <jeffm@suse.com>
Wed, 8 Jun 2016 04:36:38 +0000 (00:36 -0400)
committerGreg Kroah-Hartman <gregkh@linuxfoundation.org>
Wed, 27 Jul 2016 16:47:33 +0000 (09:47 -0700)
commit13226e1ed87182883f877f6021c2b4a2f019c363
tree0af0de996c4a31ad316128369de417cb38966cdf
parent3bb1138e7760404fb801f0b656f85d2fcfc2c39d
btrfs: account for non-CoW'd blocks in btrfs_abort_transaction

commit 64c12921e11b3a0c10d088606e328c58e29274d8 upstream.

The test for !trans->blocks_used in btrfs_abort_transaction is
insufficient to determine whether it's safe to drop the transaction
handle on the floor.  btrfs_cow_block, informed by should_cow_block,
can return blocks that have already been CoW'd in the current
transaction.  trans->blocks_used is only incremented for new block
allocations. If an operation overlaps the blocks in the current
transaction entirely and must abort the transaction, we'll happily
let it clean up the trans handle even though it may have modified
the blocks and will commit an incomplete operation.

In the long-term, I'd like to do closer tracking of when the fs
is actually modified so we can still recover as gracefully as possible,
but that approach will need some discussion.  In the short term,
since this is the only code using trans->blocks_used, let's just
switch it to a bool indicating whether any blocks were used and set
it when should_cow_block returns false.

Signed-off-by: Jeff Mahoney <jeffm@suse.com>
Reviewed-by: Filipe Manana <fdmanana@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
fs/btrfs/ctree.c
fs/btrfs/extent-tree.c
fs/btrfs/super.c
fs/btrfs/transaction.h