Re: [Xen-devel] [DRAFT v3] XenSock protocol design document

[Stefano Stabellini <sstabellini@xxxxxxxxxx>]

On Thu, 28 Jul 2016, Sander Eikelenboom wrote:
> Thursday, July 28, 2016, 8:11:53 PM, you wrote:
> 
> > ping
> 
> Hi Stefano,
> 
> JFYI:
> Since this doesn't seem to be checked with the upstream kernel yet,
> I don't know if you are aware of the opinions expressed upstream 
> about the proposed Hyper-V socket patches:
> http://lkml.iu.edu/hypermail/linux/kernel/1607.3/01748.html
> 
> (and if that should either influence your design or design process)

Thanks Sander, I am aware of that conversation going on. However the
problem they have is that hv_sock is similar to vsock (at least in
purpose), and the kernel guys would like to see only one option for
VM-hypervisor communications.  That is understandable.  The Xen
community had a similar discussion when v4v was proposed (we already had
vchan).

This is not an inter-VM or VM-hypervisor communication protocol. It
cannot be replaced with vsock. They might still dislike xensock and even
nack it, but I think it will be for different reasons.

On a related topic, I am thinking of renaming xensock to something more
like "PVCalls".

XenSock is confusing. It encourages comparisons with vsock. xensock
sounds like vsock or hv_sock for xen, which is not. In fact in the
future there might be a virtio version of this protocol, and still it
wouldn't be able to replace virtio-vsock.

_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxx
https://lists.xen.org/xen-devel