Re: PCI/e1000 BUG: unable to handle kernel paging request at 0ffff163

On Wed, Sep 05, 2012 at 11:41:04AM -0700, Yinghai Lu wrote:

On Tue, Sep 4, 2012 at 11:51 PM, Fengguang Wu [off-list ref] wrote:

quoted

Yinghai,

There are many kernel paging errors showing up in tree:

  git://git.kernel.org/pub/scm/linux/kernel/git/yinghai/linux-yinghai.git for-pci-for-each-res-addon-v2

The below summary shows that

1) it's a reliably reproducible bug
2) all paging fault happens at address 0ffff163 and in some e1000 functions

I'll try to bisect if the root cause is not obvious to you.  (Cannot
do so for now because there are 3 bisections on the way and I cannot
afford more..)

thanks, will check that...

the 32bit kernel as kvm gust, and host is doing pci passthrough to
export the VF to the KVM guest?

No I didn't do anything special for PCI passthrough. I just used KVM's
emulated e1000 device.

btw, I'll be able to start bisection after several hours.

Thanks,
Fengguang