Re: atl1: WARNING at net/sched/sch_generic.c:221

[Stephen Hemminger <shemminger@vyatta.com>]

On Sun, 14 Sep 2008 14:26:54 -0500
Jay Cliburn <jacliburn@bellsouth.net> wrote:

> On Thu, 21 Aug 2008 05:08:57 -0700 (PDT)
> David Miller <davem@davemloft.net> wrote:
> 
> > From: adobriyan@gmail.com
> > Date: Thu, 21 Aug 2008 16:04:09 +0400
> > 
> > > On Thu, Aug 21, 2008 at 04:59:10AM -0700, David Miller wrote:
> > > > From: adobriyan@gmail.com
> > > > Date: Thu, 21 Aug 2008 15:58:49 +0400
> > > > 
> > > > > This message happens more or less every reboot, sometimes cable
> > > > > unplug/plug is needed to restore connectivity, otherwise card
> > > > > is working fine.
> > > > 
> > > > What kernel version.... no, I can figure it out via osmosis never
> > > > mind! :-)
> > > 
> > > WARN already prints kernel version. ;-)
> > > 
> > > [ 2086.052503] Pid: 0, comm: swapper Not tainted
> > > 2.6.27-rc4-netns-nf #4 ^^^^^^^^^^^^^^^^^^^
> > > 
> > > And before you ask, "-netns-nf" part doesn't matter, it happens with
> > > strictly mainline kernels too and started around multiqueue TX
> > > changes, IIRC.
> > 
> > It's a simple transmit timeout error.
> > 
> > Perhaps atl1 doesn't call netif_carrier_off() in all the places that
> > it should.
> 
> For reference, the original report from Alexey on this matter is here:
> 
> http://marc.info/?l=linux-netdev&m=121931988219314&w=2
> 
> To which Dave responded above, "It's a simple transmit timeout error."
> 
> Should a netdev driver be coded such that a watchdog transmit timeout
> never occurs?
> 
> [ 2086.049998] NETDEV WATCHDOG: eth0 (atl1): transmit timed out
> 
> Or is a watchdog timeout an expected occurrence if a cable is
> unplugged/plugged?

Any transmit timeout is driver or hardware bug. The driver should be
shutting down transmit correctly on cable pull.