Re: [PATCH] Btrfs: avoid deadlock in btrfs_async_run_delayed_root

[Liu Bo <bo.li.liu@oracle.com>]

Hi,

Please ignore this patch, I misunderstood the lock order.

thanks,

-liubo

On Fri, Oct 20, 2017 at 05:40:12PM -0600, Liu Bo wrote:
> Here %path is set with ->leave_spinning=1, thus it'll hold the rwlock
> after btrfs_search_slot(), however, when processing delayed items in
> __btrfs_commit_inode_delayed_items(), it goes to acquire node->mutex
> which might sleep, that is, it may sleep under spin_lock.
> 
> Given node->mutex is necessary to avoid race between async works, this
> remove the ->leave_spining setting.
> 
> Signed-off-by: Liu Bo <bo.li.liu@oracle.com>
> ---
>  fs/btrfs/delayed-inode.c | 1 -
>  1 file changed, 1 deletion(-)
> 
> diff --git a/fs/btrfs/delayed-inode.c b/fs/btrfs/delayed-inode.c
> index 19e4ad2..a857a2f 100644
> --- a/fs/btrfs/delayed-inode.c
> +++ b/fs/btrfs/delayed-inode.c
> @@ -1331,7 +1331,6 @@ static void btrfs_async_run_delayed_root(struct btrfs_work *work)
>  	if (!delayed_node)
>  		goto free_path;
>  
> -	path->leave_spinning = 1;
>  	root = delayed_node->root;
>  
>  	trans = btrfs_join_transaction(root);
> -- 
> 2.9.4
> 
> --
> To unsubscribe from this list: send the line "unsubscribe linux-btrfs" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html