hw/xen: prevent guest from binding loopback event channel to itself

Fuzzing showed that a guest could bind an interdomain port to itself, by
guessing the next port to be allocated and putting that as the 'remote'
port number. By chance, that works because the newly-allocated port has
type EVTCHNSTAT_unbound. It shouldn't.

Signed-off-by: David Woodhouse <dwmw@amazon.co.uk>
Reviewed-by: Paul Durrant <paul@xen.org>
Message-Id: <20230801175747.145906-4-dwmw2@infradead.org>
Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>

diff --git a/hw/i386/kvm/xen_evtchn.c b/hw/i386/kvm/xen_evtchn.c
index 0e9c108614890caa4337f63e166ab79bea849e03..a731738411b48761bec771e30895e189e7f274ce 100644 (file)
--- a/hw/i386/kvm/xen_evtchn.c
+++ b/hw/i386/kvm/xen_evtchn.c
@@ -1408,8 +1408,15 @@ int xen_evtchn_bind_interdomain_op(struct evtchn_bind_interdomain *interdomain)
         XenEvtchnPort *rp = &s->port_table[interdomain->remote_port];
         XenEvtchnPort *lp = &s->port_table[interdomain->local_port];
 
-        if (rp->type == EVTCHNSTAT_unbound && rp->type_val == 0) {
-            /* It's a match! */
+        /*
+         * The 'remote' port for loopback must be an unbound port allocated for
+         * communication with the local domain (as indicated by rp->type_val
+         * being zero, not PORT_INFO_TYPEVAL_REMOTE_QEMU), and must *not* be
+         * the port that was just allocated for the local end.
+         */
+        if (interdomain->local_port != interdomain->remote_port &&
+            rp->type == EVTCHNSTAT_unbound && rp->type_val == 0) {
+
             rp->type = EVTCHNSTAT_interdomain;
             rp->type_val = interdomain->local_port;