Re: Anyone seeing serious network breakage in current Rawhide?

[Dan Williams <dcbw redhat com>]

On Thu, 2008-07-31 at 17:44 +0100, Richard W.M. Jones wrote:
> Just rebooted into a fully up to date Rawhide, and I'm experiencing
> really odd network-related problems.
> 
> For example, ssh & scp work fine.
> 
> HTTP is totally broken: most web pages time out (either in DNS or
> during page load), but occasionally I can get a whole webpage.  Tested
> this in all of Firefox, Seamonkey and w3m, so I'm fairly sure it's not
> just a broken Firefox.  I seem to have excluded local factors, such as
> the router and network connection itself, since other machines in my
> office work fine.
> 
> ping is OK with small and large packet sizes.
> 
> DNS lookups work fine from command line tools (eg. ping, ssh), but
> fail quite a lot in browsers.
> 
> Linux thinkpad 2.6.27-0.186.rc0.git15.fc10.i686 #1 SMP Sun Jul 27 05:22:35 EDT 2008 i686 i686 i386 GNU/Linux
> 
> The network is wireless, using iwl3945 driver, configured by
> NetworkManager.  As far as I'm aware the kernel is untainted and there
> is no non-free software installed on the machine.
> 
> Nothing very interesting in dmesg except some oopses in
> sched_mc_power_savings_store which appear to be unrelated.

There's breakage upstream in wireless-land for all mac80211-based
drivers (b43, b43-legacy, iwl3945, iwl4965, ath5k, rt2x00, etc) while
fallout from Dave Miller's multiqueue patches is fixed up.  mac80211 was
incorrectly using the skb->cb field that the multiqueue patches are now
also using, and that conflict is causing a lot of random weirdness.

Dan