Re: [PATCH v4 3/6] virtio/vhost-vsock: avoid potential deadlock

[Claudio Imbrenda <imbrenda@linux.vnet.ibm.com>]

On 11/04/16 17:19, Stefan Hajnoczi wrote:
> > +		/* Since we are now asynchronous, we ignore the return values.
> > +		 * After all we don't even know who we should report them to.
>
> I'm a little concerned about adding kmalloc and ignoring return values
> because virtio-vsock packets must never be lost.  Unlike TCP there is no
> retransmission so the transport must be reliable.

I understand your concern. I wrote this solution with the workqueue in
order to overcome a potential deadlock, if you find a different way to
fix the deadlock, then you can just ignore this patch.

Also, the workqueue only deals with reset packets, not with actual traffic.