Re: 2.6.0 NFS-server low to 0 performance

[Pavel Machek]

Hi!

> > > It is not just a problem of 2.6 with those specific network configurations
> > > - ftp / http / tftp transfers work fine. E.g. wget of the same file on the
> > > PXA with 2.6.0 from the PC1 with 2.4.21 over http takes about 2s. So, it
> > > is 2.6 + NFS.
> > >
> > > Is it fixed somewhere (2.6.1-rcx?), or what should I try / what further
> > > information is required?
> >
> > You will probably need to look at some tcpdump output to debug the problem...
> 
> Yep, just have done that - well, they differ... First obvious thing that I
> noticed is that 2.6 is trying to read bigger blocks (32K instead of 8K),
> but then - so far I cannot interpret what happens after the start of the

I've seen slow machine (386sx with ne1000) that could not receive 7 full-sized packets
back-to-back. You are sending 22 full packets back-to-back.
I'd expect some of them to be (almost deterministicaly) lost,
and no progress ever made.

In same scenario, TCP detects "congestion" and works mostly okay.
On ne1000 machine, TCP was still able to do 200KB/sec on
10Mbps network. Check if your slow machines are seeing all the packets you
send.
				Pavel

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/