Re: btrfs: qgroup: account shared subtrees during snapshot delete

[Josef Bacik <jbacik@fb.com>]

On 06/19/2014 04:16 PM, Mark Fasheh wrote:
> Thanks for the review Josef, I will implement everything you mentioned. I
> have one question below though:
>
> On Thu, Jun 19, 2014 at 03:25:12PM -0700, Josef Bacik wrote:
>> On 06/19/2014 02:49 PM, Mark Fasheh wrote:
>>> diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c
>>> index 46f39bf..672d2a4 100644
>>> --- a/fs/btrfs/extent-tree.c
>>> +++ b/fs/btrfs/extent-tree.c
>>> @@ -7472,6 +7703,9 @@ static noinline int do_walk_down(struct btrfs_trans_handle *trans,
>>>
>>>    	if (wc->stage == DROP_REFERENCE) {
>>>    		if (wc->refs[level - 1] > 1) {
>>> +			account_shared_subtree(trans, root, next, generation,
>>> +					       level - 1);
>>> +
>>
>> We don't pay attention to the return value, we should probably abort the
>> transaction if there is a problem.
>
> Abort or log an error and continue? I ask because technically we could
> continue with the subvolume drop but obviously qgroup state will need to be
> fixed via a future rescan. I guess the question is which is more 'friendly'
> to the user.

I'd be ok with log an error and tell the user to rescan.  Thanks,

Josef