[Xen-users] xm pci-detach lockup

["Andrew Lyon" <andrew.lyon@xxxxxxxxx>]

Hi,

I have a NEC USB2 PCIx card which I assign to a Vista HVM by adding
pci = [ '0000:02:01.0', '0000:02:01.1', '0000:02:01.2' ] to the config
file, the card works well in Vista but if I try to detach it, destroy
the domain without shutting it down, or if the domain crashes, then
the entire machine locks up, I have a serial console setup but it is
completely unresponsive.

Combined with the issue that when the machine is under heavy load
Vista sometimes BSOD with "A clock interrupt was not received on a
secondary processor" this causes the system to lockup frequently :(.

I made sure to trigger safe removal of the device before detaching it.

Perhaps all 3 functions of the card need to be detached at the same
time? or am I wrong in assigning all 3 in the first place? they are
all listed when I run xm pci-list-assignable-devices and it seems to
be necessary to assign all 3.

Assigning the device by adding the setting to the config file, or
using xm pci-attach after the domain has booted does work, only
detaching causes the lockup.

This happens on my main workstation which is a Supermicro X7DWA-N
running Xen 3.3.1 rc4 (also tried 3.3 with fix for "ValueError:
unsupported format character ':' (0x3a) at index 6" and
fix_find_parent.patch), I usually use my own Gentoo ebuild for Xen
dom0 kernel which is 2.6.27.8 with the openSUSE Xen patches rebased to
apply to vanilla without any of the other patches that openSUSE kernel
usually applies, but I have also tried using the interim 2.6.27 kernel
from http://xenbits.xensource.com/ext/linux-2.6.27-xen.hg with the
same result.

I replicated the problem on a Dell Optiplex 755 using the same
versions of Xen and kernel, but using a dlink network card which is
assigned as a single function (0000:02:01.0), it behaved exactly the
same, lockup on xm-detach.

I seem to recall a commit recently for "hypervisor panic on domain
destory with passthru", but I cannot find it, anybody know where i can
find it and if it may help?

Thanks
Andy

Andy

_______________________________________________
Xen-users mailing list
Xen-users@xxxxxxxxxxxxxxxxxxx
http://lists.xensource.com/xen-users