[dpdk-ci] DPDK Community Lab Meeting Minutes, Tuesday May 7th

Jeremy Plsek jplsek at iol.unh.edu
Wed May 8 22:07:10 CEST 2019


Hmmm, I am unable to reproduce the issue now. Maybe that was the case.
I thought I made sure to stop the OVS processes and VMs before trying
to unbind VFIO. I'll let you know if I experience it again.

Thanks



On Tue, May 7, 2019 at 10:23 PM Yao, Lei A <lei.a.yao at intel.com> wrote:
>
> Hi, Jeremy
>
> Could you have a check if you kill all OVS related process before
> unbind NIC from vfio-pci driver?
> If OVS process is still active, this action will fail.
>
> BRs
> Lei
>
> > -----Original Message-----
> > From: Xu, Qian Q
> > Sent: Wednesday, May 8, 2019 10:08 AM
> > To: O'Driscoll, Tim <tim.odriscoll at intel.com>; ci at dpdk.org; Yao, Lei A
> > <lei.a.yao at intel.com>
> > Cc: Xu, Qian Q <qian.q.xu at intel.com>
> > Subject: RE: DPDK Community Lab Meeting Minutes, Tuesday May 7th
> >
> > > - For Intel, after binding drivers to VFIO for OVS tests, can't bind back to
> > > I40E/IXGBE. Jeremy will send details in email.
> >
> > + Lei in the loop. Lei, any comments for it? Thx.
> >
> > > -----Original Message-----
> > > From: ci [mailto:ci-bounces at dpdk.org] On Behalf Of O'Driscoll, Tim
> > > Sent: Tuesday, May 7, 2019 9:27 PM
> > > To: ci at dpdk.org
> > > Subject: [dpdk-ci] DPDK Community Lab Meeting Minutes, Tuesday May
> > 7th
> > >
> > > Jeremy is working on setting up the OVS tests on the Intel and Mellanox
> > > machines. Issues have been encountered with both setups.
> > > - For Intel, after binding drivers to VFIO for OVS tests, can't bind back to
> > > I40E/IXGBE. Jeremy will send details in email.
> > > - For Mellanox, having issues trying to set up using VFIO. Jeremy will follow
> > up
> > > with Ali and Thomas.
> > >
> > > Testing is currently paused while Ali updates the Mellanox firmware. This
> > will be
> > > done today and tests should restart after that.
> > >
> > > There's a security issue if patches contains malicious code. Tests run on all
> > > servers with root permissions. Possible solutions include: stop running tests
> > as
> > > root, run tests in containers, implement static analysis of patches prior to
> > > running. Ali will investigate and make a recommendation.



--
Jeremy Plsek
UNH InterOperability Laboratory


More information about the ci mailing list