PERFORCE change 138103 for review

[Sam Leffler]

http://perforce.freebsd.org/chv.cgi?CH=138103

Change 138103 by sam at sam_ebb on 2008/03/19 16:04:54

	remove timeout on disassociate event; this causes us to
	fallback on the normal event handler+state machine to
	clear the lease which is what you'd get in wired case

Affected files ...

.. //depot/projects/vap/sbin/dhclient/dhclient.c#6 edit

Differences ...

==== //depot/projects/vap/sbin/dhclient/dhclient.c#6 (text+ko) ====

@@ -278,7 +278,11 @@
 		switch (ifan->ifan_what) {
 		case RTM_IEEE80211_ASSOC:
 		case RTM_IEEE80211_REASSOC:
-			cancel_timeout(disassoc, ifi);
+			/*
+			 * Use assoc/reassoc event to kick state machine
+			 * in case we roam.  Otherwise fall back to the
+			 * normal state machine just like a wired network.
+			 */
 			jev = (struct ieee80211_join_event *) &ifan[1];
 			if (memcmp(curbssid, jev->iev_addr, 6)) {
 				disassoc(ifi);
@@ -286,22 +290,6 @@
 			}
 			memcpy(curbssid, jev->iev_addr, 6);
 			break;
-		case RTM_IEEE80211_DISASSOC:
-			/*
-			 * Defer state change in case we are re-associating
-			 * to the same ap after a deauth (e.g. due to being
-			 * idle).  We'll get an associate event and if we
-			 * roamed we'll handle it (see above).  Otherwise we'll
-			 * eventually clear state.  The only question is how
-			 * long to wait before doing this.  Give it 45 seconds
-			 * as this should be plenty to scan multiple bands
-			 * and complete a WPA handshake.  If this event is a
-			 * precursor to our being marked down or the interface
-			 * going away that'll be handled by the link state
-			 * change.
-			 */
-			add_timeout(t+45, disassoc, ifi);
-			break;
 		}
 		break;
 	default: