Panic in the udp_input() under heavy load

[Robert Watson]

On Fri, 30 Dec 2011, Maxim Sobolev wrote:

> On 12/30/2011 4:46 PM, Maxim Sobolev wrote:
>> I see. Would you guys mind if I put that NULL pointer check into the code 
>> for the time being and turn it into some kind of big nasty warning in 
>> 8-stable branch only?
>
> I could also open a ticket, put all debug information collected to date in 
> there. And encourage people to report to it once they see this warning on 
> their system. Then it would provide more information about the exposure. It 
> is definitely looks like locking issue somewhere, not just bad luck or flaky 
> hardware, as we see it happening consistently on top 4 most UDP-loaded 
> systems here, and it correlates well with the load. With my small NULL catch 
> the machines have been running happily for a month now, so there is no 
> visible side-effects.

Please do file the PR so that all the information is in one place -- this is a 
network stack hacking week for me, so I should be able to take a closer look.

Could you characterise the traffic load on these boxes a bit more?  Also, is 
there regular monitoring using netstat/bsnmp/etc going on?  I'd like to try 
and identify ways in which this workload differs from other common high-UDP 
workloads being used on 8.x that aren't seeing this problem...

Robert