drivers: hv: switch to use mb() instead of smp_mb()
authorJason Wang <jasowang@redhat.com>
Tue, 18 Jun 2013 05:04:23 +0000 (13:04 +0800)
committerBen Hutchings <ben@decadent.org.uk>
Sat, 27 Jul 2013 04:34:12 +0000 (05:34 +0100)
commit 35848f68b07df3f917cb13fc3c134718669f569b upstream.

Even if guest were compiled without SMP support, it could not assume that host
wasn't. So switch to use mb() instead of smp_mb() to force memory barriers for
UP guest.

Signed-off-by: Jason Wang <jasowang@redhat.com>
Cc: Haiyang Zhang <haiyangz@microsoft.com>
Signed-off-by: K. Y. Srinivasan <kys@microsoft.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
[bwh: Backported to 3.2:
 - Drop changes to functions that don't exist here
 - hv_ringbuffer_write() has only a write memory barrier]
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
drivers/hv/ring_buffer.c
drivers/hv/vmbus_drv.c

index 8af25a0..810658e 100644 (file)
@@ -383,7 +383,7 @@ int hv_ringbuffer_write(struct hv_ring_buffer_info *outring_info,
                                             sizeof(u64));
 
        /* Make sure we flush all writes before updating the writeIndex */
-       smp_wmb();
+       wmb();
 
        /* Now, update the write location */
        hv_set_next_write_location(outring_info, next_write_location);
index d2d0a2a..44442d5 100644 (file)
@@ -466,7 +466,7 @@ static void vmbus_on_msg_dpc(unsigned long data)
                 * will not deliver any more messages since there is
                 * no empty slot
                 */
-               smp_mb();
+               mb();
 
                if (msg->header.message_flags.msg_pending) {
                        /*