PERFORCE change 113190 for review

[Matt Jacob]

http://perforce.freebsd.org/chv.cgi?CH=113190

Change 113190 by mjacob at mjexp on 2007/01/20 03:55:24

	Ah, fixed a stupidville bug: when we're looking for a new
	handle to allocate for an arriving device, don't pick one
	that's already in use (pending valid) as the firmware will
	be quite content to let you remap that handle to the new
	device. Then you trip over a bunch of other code which
	ends up nuking one device and putting the new device in
	*two* different positions in the virtual map. Gawd.

Affected files ...

.. //depot/projects/mjexp/sys/dev/isp/isp.c#9 edit

Differences ...

==== //depot/projects/mjexp/sys/dev/isp/isp.c#9 (text+ko) ====

@@ -3956,6 +3956,7 @@
 static uint16_t
 isp_nxt_handle(ispsoftc_t *isp, uint16_t handle)
 {
+	int i;
 	if (handle == NIL_HANDLE) {
 		if (FCPARAM(isp)->isp_topo == TOPO_F_PORT) {
 			handle = 0;
@@ -3982,9 +3983,16 @@
 	}
 	if (handle == FCPARAM(isp)->isp_loopid) {
 		return (isp_nxt_handle(isp, handle));
-	} else {
-		return (handle);
+	}
+	for (i = 0; i < MAX_FC_TARG; i++) {
+		if (FCPARAM(isp)->portdb[i].state == FC_PORTDB_STATE_NIL) {
+			continue;
+		}
+		if (FCPARAM(isp)->portdb[i].handle == handle) {
+			return (isp_nxt_handle(isp, handle));
+		}
 	}
+	return (handle);
 }
 
 /*