Re: Stable regression with 'tcp: allow splice() to build full TSO packets'

[Willy Tarreau <w@1wt.eu>]

Hi David,

On Thu, May 17, 2012 at 03:55:03PM -0400, David Miller wrote:
> From: Willy Tarreau <w@1wt.eu>
> Date: Thu, 17 May 2012 17:01:57 +0200
> 
> >>From 6da6a21798d0156e647a993c31782eec739fa5df Mon Sep 17 00:00:00 2001
> > From: Willy Tarreau <w@1wt.eu>
> > Date: Thu, 17 May 2012 16:48:56 +0200
> > Subject: [PATCH] tcp: force push data out when buffers are missing
> > 
> > Commit 2f533844242 (tcp: allow splice() to build full TSO packets)
> > significantly improved splice() performance for some workloads but
> > caused stalls when pipe buffers were larger than socket buffers.
> > 
> > The issue seems to happen when no data can be copied at all due to
> > lack of buffers, which results in pending data never being pushed.
> > 
> > This change checks if all pending data has been pushed or not and
> > pushes them when waiting for send buffers.
> 
> Eric, please indicate whether we need Willy's patch here.
> 
> I want to propagate this fix as fast as possible if so.

I think you should hold off for now, because it's possible that my patch
hides another issue instead of fixing it.

I'm having the same stall issue again since I applied Eric's build_skb
patch, but not for all data sizes. So if the same issue is still there,
it's possible that we're playing hide-and-seek with it. That's rather
strange.

Thanks,
Willy