On 1.02.20 г. 0:36 ч., Josef Bacik wrote:
> We can end up with free'd extents in the delayed refs, and thus
> may_commit_transaction() may not think we have enough pinned space to
> commit the transaction and we'll ENOSPC early. Handle this by running
> the delayed refs in order to make sure pinned is uptodate before we try
> to commit the transaction.
>
> Signed-off-by: Josef Bacik <josef@xxxxxxxxxxxxxx>
Actually wouldn't it be better if the stages are structured:
FLUSH_DELALLOC which potentially creates a bunch of delayed refs,
FLUSH_DELAYED_REFS which in turn could free some reservations. Then IPUT
to process anything from delayed refs running and finally committing
transaction to reclaim pinned space?
Codewise this is ok.
Reviewed-by: Nikolay Borisov <nborisov@xxxxxxxx>
> ---
> fs/btrfs/space-info.c | 1 +
> 1 file changed, 1 insertion(+)
>
> diff --git a/fs/btrfs/space-info.c b/fs/btrfs/space-info.c
> index c86fad4174f1..5b0dc1046daa 100644
> --- a/fs/btrfs/space-info.c
> +++ b/fs/btrfs/space-info.c
> @@ -805,6 +805,7 @@ static const enum btrfs_flush_state evict_flush_states[] = {
> static const enum btrfs_flush_state data_flush_states[] = {
> FLUSH_DELALLOC_WAIT,
> RUN_DELAYED_IPUTS,
> + FLUSH_DELAYED_REFS,
> COMMIT_TRANS,
> };
>
>
