Re: [PATCH 1/5] xfs: bump up reserved blocks in xfs_alloc_set_aside

[Brian Foster <bfoster@redhat.com>]

On Thu, Dec 22, 2016 at 09:00:18PM +0100, Christoph Hellwig wrote:
> Setting aside 4 blocks globally for bmbt splits isn't all that useful,
> as different threads can allocate space in parallel.  Bump it to 4
> blocks per AG to allow each thread that is currently doing an
> allocation to dip into it.
> 
> Signed-off-by: Christoph Hellwig <hch@lst.de>
> ---

Presumably this patch addresses the potential deadlock issues from the
previous version, but the commit log description makes no mention of it
whatsoever. While the code seems fine, I think the commit log
description needs more information wrt to that situation and the
relationship/dependency with minleft.

>  fs/xfs/libxfs/xfs_alloc.c | 5 +----
>  1 file changed, 1 insertion(+), 4 deletions(-)
> 
> diff --git a/fs/xfs/libxfs/xfs_alloc.c b/fs/xfs/libxfs/xfs_alloc.c
> index 5050056..0a46f84 100644
> --- a/fs/xfs/libxfs/xfs_alloc.c
> +++ b/fs/xfs/libxfs/xfs_alloc.c
> @@ -95,10 +95,7 @@ unsigned int
>  xfs_alloc_set_aside(
>  	struct xfs_mount	*mp)
>  {
> -	unsigned int		blocks;
> -
> -	blocks = 4 + (mp->m_sb.sb_agcount * XFS_ALLOC_AGFL_RESERVE);
> -	return blocks;
> +	return mp->m_sb.sb_agcount * (XFS_ALLOC_AGFL_RESERVE + 4);

The comment above xfs_alloc_set_aside() already touches on the writeback
situation, but why 4 blocks per ag? Wasn't the intent to use
worst_indlen() since that's the base for minleft?

Also, it looks like this causes a regression in xfs/004. On a quick
look, we might just need a test update however...

Brian

>  }
>  
>  /*
> -- 
> 2.1.4
> 
> --
> To unsubscribe from this list: send the line "unsubscribe linux-xfs" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html