Xen devel - Jan 2008 - blktap for HVM guests without PV-on-HVM drivers

Hi,

Is blktap interposing supported for HVM guests not using PV-on-HVM
drivers? When using tap:aio with an HVM domain, we see that td_open
and td_close are getting called, but none of td_queue_read,
td_queue_write and td_submit. Is this supposed to work?

Cheers,
Muli


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

Currently qemu cannot access the PV backend driver data path. So PV backends
are only accessible using PV-on-HVM drivers within the HVM guest.

 -- Keir


On 28/1/08 13:17, "Muli Ben-Yehuda" <muli@il.ibm.com> wrote:
> Hi,
> 
> Is blktap interposing supported for HVM guests not using PV-on-HVM
> drivers? When using tap:aio with an HVM domain, we see that td_open
> and td_close are getting called, but none of td_queue_read,
> td_queue_write and td_submit. Is this supposed to work?
> 
> Cheers,
> Muli
> 
> 
> _______________________________________________
> Xen-devel mailing list
> Xen-devel@lists.xensource.com
> http://lists.xensource.com/xen-devel


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

On Mon, Jan 28, 2008 at 01:53:06PM +0000, Keir Fraser wrote:
> Currently qemu cannot access the PV backend driver data path. So PV
> backends are only accessible using PV-on-HVM drivers within the HVM
> guest.
Thanks, good to know I wasn''t missing something obvious. Are there any
particular obstacles to fixing it, or is it a simple matter of
programming?

Cheers,
Muli

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

On 28/1/08 13:57, "Muli Ben-Yehuda" <muli@il.ibm.com> wrote:
> On Mon, Jan 28, 2008 at 01:53:06PM +0000, Keir Fraser wrote:
> 
>> Currently qemu cannot access the PV backend driver data path. So PV
>> backends are only accessible using PV-on-HVM drivers within the HVM
>> guest.
> 
> Thanks, good to know I wasn''t missing something obvious. Are there
any
> particular obstacles to fixing it, or is it a simple matter of
> programming?
It''s a small matter of plumbing. There''s currently no local
interface to
allow dom0 userspace to get at the dom0 backend driver datapath.

Currently the only way to allow booting from a virtual disk that will later
be accessed via PV-on-HVM is to give qemu-dm direct access to the underlying
virtual-disk image (e.g., qcow or whatever) which obviously has
synchronisation issues when block requests start flying in on a completely
different data path once PV-on-HVM drivers are installed. So getting things
plumbed properly would be very nice indeed!

 -- Keir



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