xen/evtchn: Add missing barriers when accessing/allocating an event channel

While the allocation of a bucket is always performed with the per-domain
lock, the bucket may be accessed without the lock taken (for instance, see
evtchn_send()).

Instead such sites relies on port_is_valid() to return a non-zero value
when the port has a struct evtchn associated to it. The function will
mostly check whether the port is less than d->valid_evtchns as all the
buckets/event channels should be allocated up to that point.

Unfortunately a compiler is free to re-order the assignment in
evtchn_allocate_port() so it would be possible to have d->valid_evtchns
updated before the new bucket has finish to allocate.

Additionally on Arm, even if this was compiled "correctly", the
processor can still re-order the memory access.

Add a write memory barrier in the allocation side and a read memory
barrier when the port is valid to prevent any re-ordering issue.

This is XSA-340.

Reported-by: Julien Grall <jgrall@amazon.com>
Signed-off-by: Julien Grall <jgrall@amazon.com>
Reviewed-by: Stefano Stabellini <sstabellini@kernel.org>

xen/common/event_channel.c

@@ -178,6 +178,13 @@ int evtchn_allocate_port(struct domain *d, evtchn_port_t port)
             return -ENOMEM;
         bucket_from_port(d, port) = chn;
 
+        /*
+         * d->valid_evtchns is used to check whether the bucket can be
+         * accessed without the per-domain lock. Therefore,
+         * d->valid_evtchns should be seen *after* the new bucket has
+         * been setup.
+         */
+        smp_wmb();
         write_atomic(&d->valid_evtchns, d->valid_evtchns + EVTCHNS_PER_BUCKET);
     }
 

xen/include/xen/event.h

@@ -107,7 +107,17 @@ void notify_via_xen_event_channel(struct domain *ld, int lport);
 
 static inline bool_t port_is_valid(struct domain *d, unsigned int p)
 {
-    return p < read_atomic(&d->valid_evtchns);
+    if ( p >= read_atomic(&d->valid_evtchns) )
+        return false;
+
+    /*
+     * The caller will usually access the event channel afterwards and
+     * may be done without taking the per-domain lock. The barrier is
+     * going in pair the smp_wmb() barrier in evtchn_allocate_port().
+     */
+    smp_rmb();
+
+    return true;
 }
 
 static inline struct evtchn *evtchn_from_port(struct domain *d, unsigned int p)