Re: [PATCH] Btrfs: implement support for fallocate collapse range

[Chandan Rajendra <chandan@linux.vnet.ibm.com>]

On Monday 23 Jun 2014 11:25:47 Filipe David Borba Manana wrote:
> diff --git a/fs/btrfs/ctree.c b/fs/btrfs/ctree.c
> index aeab453..8f1a371 100644
> --- a/fs/btrfs/ctree.c
> +++ b/fs/btrfs/ctree.c
> @@ -2825,12 +2825,12 @@ cow_done:
>  		 * It is safe to drop the lock on our parent before we
>  		 * go through the expensive btree search on b.
>  		 *
> -		 * If we're inserting or deleting (ins_len != 0), then we might
> -		 * be changing slot zero, which may require changing the parent.
> -		 * So, we can't drop the lock until after we know which slot
> -		 * we're operating on.
> +		 * If we're inserting, deleting or updating a key (cow != 0),
> +		 * then we might be changing slot zero, which may require
> +		 * changing the parent. So, we can't drop the lock until after
> +		 * we know which slot we're operating on.
>  		 */
> -		if (!ins_len && !p->keep_locks) {
> +		if (!cow && !p->keep_locks) {
>  			int u = level + 1;

For the "key update" case (i.e. (ins_len == 0) and (cow == 1)), maybe
we could optimize by having a variable to hold the return value of
should_cow_block(), i.e. it keeps track of whether the current
metadata block was COWed. If the variable indicates that the block was
*not* COWed, then we could release the lock on the parent block since
the update operation (even on slot 0) isn't going to change the
corresponding entry in the parent block.

Thanks,
chandan