apport-kerneloops reporting

[Tim Gardner]

On 08/03/2011 09:55 AM, Brian Murray wrote:
> I've recently made some changes to kerneloops and apport which
> modify the way that kerneloops are reported.
>
> In kerneloops there is now a filter so that any Oops with the word
> WARNING in it is not sent to apport for reporting this should
> significantly reduce the quantity of apport-kerneloops bug reports
> you receive.
>
> The linux package hook for apport will now also tag
> apport-kerneloops bug reports using a format of
> 'kernel-driver-$drivername' if it detects a driver name in the RIP /
> IP line.  For example, http://launchpad.net/bugs/814460 is tagged
> kernel-driver-i915.  I also went through and tagged all existing Oops
> reports the same way and sent Leann a list of the tags that were
> used.
>
> Looking at the current list of apport-kerneloops reports from
> Oneiric[1] I have one suggestion and one question.
>
> Suggestion:
>
> The importance of these bug reports should be automatically set to
> at least High by whatever script does automatic triage.
>
> A question or two:
>
> I noticed there are a few NETDEV WATCHDOG bug reports in the list.
>

These appear to be triggered by one of several network drivers that have 
various issues, some of which are being worked on. Its a failure to 
transmit a packet when the network stack thinks its should have been 
able to do so. As such its largely harmless and in the same class as WARN.

> 1) Are these worth reporting?
>
> 2) If so it seems to me the driver appears in a regular format so
> that we could tag them kernel-driver-atl1c.  Would that help?
>

I'm pretty sure this driver is getting some upstream love, so I'm not 
sure I'd bother.

rtg
-- 
Tim Gardner tim.gardner at canonical.com