Reliable way to find out if a given uevent is proceeded ?

["Xinwei Hu" <hxinwei@xxxxxxxxx>]

Hi all:

  I'm looking for a reliable way to find out if a given seqnum of
uevent is proceeded.

  The best way I found was

     . read /dev/.udev/uevent_seqnum, make sure the uevent is enqueued already
     . look for /dev/.udev/queue/$seqnum
        -> if exist, not finished yet
        -> if not, finished or failed

  However, this is not a reliable way due to the fact that udevd write
seqnum before
export state into .udev/queue.

  Are there any other solution to this problem ?
  Or can we swap the order of writing seqnum and export state, like
the attached patch do ?

  Thanks.

diff --git a/udevd.c b/udevd.c
index c895c1d..d6977e9 100644
--- a/udevd.c
+++ b/udevd.c
@@ -269,6 +269,9 @@ static void msg_queue_insert(struct udevd_uevent_msg *msg)
 
 	msg->queue_time = time(NULL);
 
+	export_event_state(msg, EVENT_QUEUED);
+	info("seq %llu queued, '%s' '%s'\n", msg->seqnum, msg->action, msg->subsystem);
+
 	strlcpy(filename, udev_root, sizeof(filename));
 	strlcat(filename, "/" EVENT_SEQNUM, sizeof(filename));
 	fd = open(filename, O_WRONLY|O_TRUNC|O_CREAT, 0644);
@@ -281,9 +284,6 @@ static void msg_queue_insert(struct udevd_uevent_msg *msg)
 		close(fd);
 	}
 
-	export_event_state(msg, EVENT_QUEUED);
-	info("seq %llu queued, '%s' '%s'\n", msg->seqnum, msg->action, msg->subsystem);
-
 	/* run one event after the other in debug mode */
 	if (debug_trace) {
 		list_add_tail(&msg->node, &running_list);