bool notified = false;
        struct controlvm_message_header *pmsg_hdr = NULL;
 
+       down(¬ifier_lock);
+
        if (!bus_info) {
                /* relying on a valid passed in response code */
                /* be lazy and re-use msg_hdr for this failure, is this ok?? */
                pmsg_hdr = msg_hdr;
-               goto away;
+               goto out_respond_and_unlock;
        }
 
        if (bus_info->pending_msg_hdr) {
                /* only non-NULL if dev is still waiting on a response */
                response = -CONTROLVM_RESP_ERROR_MESSAGE_ID_INVALID_FOR_CLIENT;
                pmsg_hdr = bus_info->pending_msg_hdr;
-               goto away;
+               goto out_respond_and_unlock;
        }
 
        if (need_response) {
                pmsg_hdr = kzalloc(sizeof(*pmsg_hdr), GFP_KERNEL);
                if (!pmsg_hdr) {
-                       response = -CONTROLVM_RESP_ERROR_KMALLOC_FAILED;
-                       goto away;
+                       POSTCODE_LINUX_4(MALLOC_FAILURE_PC, cmd,
+                                        bus_info->chipset_bus_no,
+                                        POSTCODE_SEVERITY_ERR);
+                       goto out_unlock;
                }
 
                memcpy(pmsg_hdr, msg_hdr,
                bus_info->pending_msg_hdr = pmsg_hdr;
        }
 
-       down(¬ifier_lock);
        if (response == CONTROLVM_RESP_SUCCESS) {
                switch (cmd) {
                case CONTROLVM_BUS_CREATE:
                        break;
                }
        }
-away:
+
+out_respond_and_unlock:
        if (notified)
                /* The callback function just called above is responsible
                 * for calling the appropriate visorchipset_busdev_responders
                 * directly and kfree() there.
                 */
                bus_responder(cmd, pmsg_hdr, response);
+
+out_unlock:
        up(¬ifier_lock);
 }