Hi Kevin,
I agree with you, and if you would be so kind to send the patches, even
untested, they would be much appreciated. Anthony or I will make sure to
test them appropriately and fix them, if they turn out to be incomplete
or partially broken. Would that be OK?
Cheers,
Stefano
P.S.
FYI Xen works on top of QEMU with PV guests. So theoretically you could
install Xen inside QEMU and use Xen to lunch a PV guest which uses
xen_block as Xen backend. Fun!! :-D
On Thu, 29 Sep 2016, Kevin Wolf wrote:
> Hi Stefano and all,
>
> while working on some part of the QEMU block layer infrastructure that
> requires going from a BlockBackend to the qdev DeviceState, I noticed
> that xen_disk is still not qdevified after all the years. It's the last
> device, and has been for a while, that is blocking the necessary changes
> in the block layer.
>
> Specifically, I'm talking about the blk_attach_dev_nofail() and
> blk_detach_dev() functions (and related ones, which don't seem to be
> used by xen_disk though), which must be converted from accepting a void*
> for the device to DeviceState*.
>
> I think in theory it shouldn't be too hard to build a minimal qdev
> device between xen_disk and the block layer. If you want me to, I can
> try to do that - however, I don't have a Xen setup to actually test the
> result, so if things break, you get to keep the pieces. If someone else
> would like to look into this in the next few days, that might be the
> better option.
>
> In any case, we need to do something to make xen_disk compatible with
> the modern (well, not _that_ modern any more) infrastructure that all
> other devices use.
>
> Kevin
>
___
Xen-devel mailing list
Xen-devel@lists.xen.org
https://lists.xen.org/xen-devel
