[Ksummit-2013-discuss] [ATTEND] kernel testing

Nicholas A. Bellinger nab at linux-iscsi.org
Wed Jul 24 23:44:55 UTC 2013


On Mon, 2013-07-22 at 21:39 +0000, Myklebust, Trond wrote:
> On Sun, 2013-07-21 at 14:06 +0800, Fengguang Wu wrote:
> > Dear all,
> > 
> > I would like to attend this year's Kernel Summit.
> > 
> > During the past year I've been working on the 0day kernel testing
> > framework, which provides instant test service for 300+ kernel git
> > trees. The system currently runs
> > 
> > - build tests covering 200+ configs from 30 archs
> > - static checkers: sparse, smatch, coccinelle, checkpatch.pl
> > - randconfig boot tests in KVM
> > - trinity, cpu hotplug, xfstests etc. runtime tests
> > 
> > The system features auto bisecting and reporting. The high confident
> > error/warnings are directly reported to the author/committer/mailing
> > list, CC kbuild-all at 01.org for achieve purpose. The low confident ones
> > are sent to  kbuild at 01.org for manual checks.
> > 
> > https://lists.01.org/pipermail/kbuild
> > https://lists.01.org/pipermail/kbuild-all
> > 
> > 0day performance testing is coming soon. Hopefully you'll see more and
> > more performance reports in the coming months before kernel summit. :)
> > Basically we'll be monitoring performance changes in the 300+ public
> > developer/maintainer/upstream git trees and report any changed vmstat,
> > iostat, nfsstat, lock_stat, perf etc. numbers in a variety of test
> > cases, configurations and hardware.
> > 
> > I'm interested in any topic on kernel testing.
> 
> I second this. I think that it would be great to have a discussion on
> performance testing, and specifically on what maintainers can do to help
> Fengguang in this effort. Should we be supplying tests, performance
> measurement tools, other? If so, what would be the preferred approach to
> get these included in Fengguang's setup.

Hi Fengguang & Co,

I'd be very interested in including tests for drivers/target/ code,
if/when the framework is able to handle subsystem specific tests.

Also, any automated stress + performance testing that can be done for
the scsi-mq prototype would be extremely helpful too.

That said, the 0 day kernel testing framework is awesome, and has been
hugely useful so far.

I think the KS t-shirts should include something along the lines of:

    "All your kernel trees are belong to Fengguang bot"

Thanks!

--nab



More information about the Ksummit-2013-discuss mailing list