RE: [Xen-devel] Daily Xen-HVM Builds: cs9226

["Yu, Ping Y" <ping.y.yu@xxxxxxxxx>]

>
>On Thu, 2006-03-16 at 13:25 +0800, Yu, Ping Y wrote:
>> Daniel,
>>
>> Currently, HVM support multiple disks in QEMU configure, and you can
add
>> extra disks by configuring options in "disk", for example,
>> disk = [ 'file:/var/images/min-el3-i386.img,ioemu:hda,w',
>'file:/var/images/min-el3-i386_2.img,ioemu:hdb,w' ]
>> Does it meet your requirement?
>
>
>My requirement for what? I know HVM domains can support more than one
>disk image, the idea is to get xm-test to automate creating disk images
>for testing HVM domains. My plan is to eventually is to use
>device-mapper to present a read only root image, that all the xm-test
>HVM test domains will share and then add writable partitions as needed
>to test domains.
Maybe my idea is a little different from yours, for I notice that there
are an
"r/w" control bit for image, thus we can make use of it. As you said, we
can 
present a read only root image and add writeable partition as needed.
:-)
>
>
>> Currently problem is that strict check is added in VBD and forbid one
>> image for multiple HVM and all those test cases in xm-test failed,
see
>information
>>  below:
>>
>> [dom0] Running `xm create /tmp/xm-test.conf'
>> Using config file "/tmp/xm-test.conf".
>> Error: Device 768 (vbd) could not be connected.
>> File /opt/vmm/control_panel/xm-test/ramdisk/disk.img is
loopback-mounted
>through /dev/loop0,
>> which is mounted in a guest domain,
>> and so cannot be mounted now.
>> Failed to create test domain because:
>> Using config file "/tmp/xm-test.conf".
>> Error: Device 768 (vbd) could not be connected.
>> File /opt/vmm/control_panel/xm-test/ramdisk/disk.img is
loopback-mounted
>through /dev/loop0,
>> which is mounted in a guest domain,
>> and so cannot be mounted now.
>>
>> REASON: Failed to create domain
>
>
>The vbd issue wasn't that only one image could be loaded for one HVM
>domain, if that's what you're saying. The issue was exceeding the
number
>of loopback devices on the system. Qemu-dm loads disk images using
>loopback devices - so you are therefore the number of disk images able
>to be mounted is limited to the number of configured loopback devices.
>
>There was a bug in 11_create_concurrent_pos.py in xm-test because it
>goes and creates as many concurrent domains as possible based on memory
>and a cutoff of 50. This is fine for para virt, but broke for HVM
>because of the loopback device limit. I have patched the test and it
>should work for you. I have run 11_create_concurrent_pos.py on my x366
>where I've changed the kernel option max_loop=256 and been able to load
>50 disk images all using the same disk image.

Daniel, from my observation, I found that it is not caused by shortage
of
loop device number, but VBD protection. If I modify the control bit "w"
to "r", xm-test works quite well, and based on that, I will send out a 
patch to fix xm-test's current problem. Hope you can review it.

>
>Thanks,
>
>Dan

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