[Xen-users] Can't boot PV from qcow image

[Matthias Blankenhaus <matthiasblankenhaus@xxxxxxxxx>]

Howdy !

I have been trying for more than 2 days to work this out without success. While I am able to identify a qcow image as a block device within the
guest, so far I can not use a qcow or qcow2 images as a block device due to IO errors.  As a reference, I am able to mount raw images via loopback device through DOM0.

Here is my setting:
DOM0
- Xen 4.2.0
- Debian wheezy with 3.1.0-rc8 / 64bit [without blktap2 patch]
- nfs mounted guest images

DOM-U
- PV
- CentOS 6.2 with 2.6.33.20.x86_64 /, no initrd
- no grub
- Xen guest disk config:
disk = ['format=raw,vdev=xvda,target=/dev/loop0','tap:qcow:/var/lib/xen-images/21339039-df03-43e2-b94f-2b600e2492b2/rootfs.qcow,xvdb,w'

When attempting to access the qcow backed device (xvdb) I get these errors:

# less /var/log/messages
...
Sep 19 01:26:15 kernel: [  189.596610] Buffer I/O error on device xvdb, logical block 163967
Sep 19 01:26:15 kernel: [  189.936239] end_request: I/O error, dev xvdb, sector 1311744
Sep 19 01:26:15 kernel: [  189.940237] Buffer I/O error on device xvdb, logical block 163968
Sep 19 01:26:16 kernel: [  190.276228] end_request: I/O error, dev xvdb, sector 1572864
...

Meanwhile, on DOM0 I see the following output in the process tree:
...
root       911     2  0 01:49 ?        00:00:00 [blkback.59.xvda]
root       848     1  0 01:49 ?        00:00:10 /usr/lib/xen/bin/qemu-system-i386 -xen-domid 59 -chardev socket,id=lib

The block back-end kernel thread is backing my rootfs which is mapped to a loopback device on DOM0.  I think qemu was launched to
back the qcow device.  Thus, this matches my conceptional understanding of how things should work.

Ultimately, I am looking for a way to boot PVs from disk image using Xen 4.2.  I was able to do this under Xen 4.1.x using a raw disk image in conjunction with the tap:aio option for the guests' disk spec and without using a loopback mount on DOM0.

Your help is much appreciated,
Matthias

_______________________________________________
Xen-users mailing list
Xen-users@xxxxxxxxxxxxx
http://lists.xen.org/xen-users