Re: [PATCH v2 2/2] xfs: use sync buffer I/O for sync delwri queue submission

[Brian Foster <bfoster@redhat.com>]

On Mon, Jun 18, 2018 at 04:21:30AM -0700, Christoph Hellwig wrote:
> On Fri, Jun 15, 2018 at 08:39:56AM -0400, Brian Foster wrote:
> > Ok, codewise I don't have much of a preference, but I don't think it's
> > worth redoing the regression testing and lowmem testing and whatnot just
> > to change how the guts are refactored here. What's the endgame?
> 
> Avoid the nasty duplication as much as possible.  I think your patch
> already goes a long way towards that, but I'd rather go all the way.
> 

That's what I figured, thanks.

> > I came
> > up with the following on top of patch 2. Compile tested only, and I can
> > refold the _common() helper back into the caller and invert the
> > nowait logic or whatnot..
> 
> Mostly looks fine, except that it seems pointless to have
> __xfs_buf_submit_common around instead of merging it into the only
> caller.
> 

Yes..

> > +	if (!(bp->b_flags & XBF_ASYNC) && !sync_nowait)
> 
> Also I'd rather pass in a
> 
> 	'bool wait' parameter.
> 
> xfs_buf_submit() would set it based on XBF_ASYNC, and the delwri
> code would just set it to false explicitly.

A patch with the above changes is essentially what I put through some
testing over the weekend. It doesn't look like anything exploded, so
I'll try to get it cleaned up and posted later today.

Brian

> --
> 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