Re: [PATCH] hvc_xen: introduce HVC_XEN_FRONTEND

[Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>]

On Tue, Mar 13, 2012 at 06:30:44PM +0000, Stefano Stabellini wrote:
> On Tue, 13 Mar 2012, Konrad Rzeszutek Wilk wrote:
> > > Even a 'feature-pv-on-hvm-console' flag  wouldn't help.
> > > 
> > > Maybe we need to introduce an explicit check in xenbus_probe_device_type
> > > to avoid calling bus->probe if type == "console" and dir[i] == "0", what
> > > do you think?
> > 
> > If that works..?
> 
> Yes, it does: the appended patch might be ugly but fixes the problem for
> me (tested xen 4.1, xm/xend, vfb enabled and disabled so both qemu and
> xenconsoled as console backends).

Yup. Seems to work with PV and HVM in both Xen 4.1 (with patch) and Xen 4.0.

> 
> ---
> 
> xenbus: ignore console/0
> 
> Unfortunately xend creates a bogus console/0 frotend/backend entry pair
> on xenstore that console backends cannot properly cope with.
> Any guest behavior that is not completely ignoring console/0 is going
> to either cause problems with xenconsoled or qemu.
> Returning 0 or -ENODEV from xencons_probe is not enough because it is
> going to cause the frontend state to become 4 or 6 respectively.
> The best possible thing we can do here is just ignore the entry from
> xenbus_probe_frontend.
> 
> Signed-off-by: Stefano Stabellini <stefano.stabellini@eu.citrix.com>
> 
> diff --git a/drivers/xen/xenbus/xenbus_probe_frontend.c b/drivers/xen/xenbus/xenbus_probe_frontend.c
> index 9c57819..f20c5f1 100644
> --- a/drivers/xen/xenbus/xenbus_probe_frontend.c
> +++ b/drivers/xen/xenbus/xenbus_probe_frontend.c
> @@ -53,6 +53,12 @@ static int xenbus_probe_frontend(struct xen_bus_type *bus, const char *type,
>  	char *nodename;
>  	int err;
>  
> +	/* ignore console/0 */
> +	if (!strncmp(type, "console", 7) && !strncmp(name, "0", 1)) {
> +		DPRINTK("Ignoring buggy device entry console/0");
> +		return 0;
> +	}
> +
>  	nodename = kasprintf(GFP_KERNEL, "%s/%s/%s", bus->root, type, name);
>  	if (!nodename)
>  		return -ENOMEM;