Re: libvirt, libxl and QDISKs

[Jim Fehlig <jfehlig@suse.com>]

Marek Marczykowski wrote:
> On 26.04.2013 13:40, Ian Campbell wrote:
>   
>> On Fri, 2013-04-26 at 12:31 +0100, Marek Marczykowski wrote:
>>     
>>> What about old good loop+phy based backend for file disk images? I don't want
>>> whole qemu in dom0 for PV domains, only for handling simple disk backend.
>>> Additionally sparse images + loop + phy + mount -o discard in domU makes the
>>> images "auto shrinking". Don't know if qemu is able to do this.
>>>       
>> IIRC the problem with loop+phy is that loop doesn't do O_DIRECT and
>> therefore your data isn't actually on the disk when you might think it
>> is, which can lead to filesystem corruption even if the f/s is doing
>> correct barriers.
>>
>>     
>>> Attached patch, which I currently use for that. If it is close to something
>>> that would be accepted, I will send it in new thread.
>>>       
>> I think you can use a block script for this (i.e. it does the loop
>> mount) and avoid patching libxl at all. That's what xend did at least...
>>     
>
> This also was solution I've evaluated, but libvirt developers don't want to
> hear about disk->script setting in official libvirt, so I still need to patch
> some library.

Good old loop+phy is supported by the legacy libvirt xen driver without
libvirt needing <disk><script /></disk>, generally by using <driver
name='file' />.

>  Choosing libxl seems to be more universal. Setting loop in libxl
> is also much faster than calling script for that.
> Perhaps I can rewrite this patch to allow the choice (LIBXL_DISK_BACKEND_LOOP)?
>   

With such a patch, libvirt could set the backend to
LIBXL_DISK_BACKEND_LOOP when <driver name='file'>, preserving loop+phy
backend setup done by the old toolstack.

Regards,
Jim