| |
|
 |
|
 |
|
|
|
| |
|
|
xen-devel
RE: [Xen-devel] Possible BUG: starting more than 1 domain with same pci=
|
Yes, this is a known issue (please see the attached
mail).
I have no time to fix it for now. :-( Hope somebody
can kindly help.
Thanks,
-- Dexuan
Xen
3.4-rc1
Dom0: 2.6.18.8-xen0
(pciback.hide=”03:00.0”)
DomU: Linux pv 2.6.26-1 (Debian) (
pci=[’03:00.0’] )
DomU: Windows XP SP3 HVM (
pci=[’03:00.0’] )
Starting Linux PV with passthrough
graphics OK, then start Windows HVM with passthrough also (happened by accident)
and xentools proceeds to start without warning or error. The passthrough to
Windows HVM is not working anyway.
Switching off the Windows HVM then
caused a complete lock up on Dom0, had to hit the reset
switch!
|
--- Begin Message ---
|
To: |
"xen-devel@xxxxxxxxxxxxxxxxxxx" <xen-devel@xxxxxxxxxxxxxxxxxxx> |
|
Subject: |
xc.test_assign_device() doesn't work for devices assigned via the traditional PCI passthrough(without VT-d). |
|
From: |
"Cui, Dexuan" <dexuan.cui@xxxxxxxxx> |
|
Date: |
Wed, 25 Feb 2009 19:27:41 +0800 |
|
Thread-index: |
AcmXPBIY3DpbmkMhSWGCchryUGOYHQ== |
|
Thread-topic: |
xc.test_assign_device() doesn't work for devices assigned via the traditional PCI passthrough(without VT-d). |
e.g., with xen parameter 'iommu=on,no-pv', after we assign a device to pv
guest1, we can still assign the device to hvm guest2 because when xend creates
guest2, it's wrongly told by xc.test_assign() the device is not assigned.
To judge if a device has been assigned, xc.test_assign() is not enough; maybe
we also have to scan xenstore?
Thanks,
-- Dexuan
--- End Message ---
_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxxxxxxxx
http://lists.xensource.com/xen-devel
|
|
|
| |
|
Home