USB: EHCI: don't lose events during a scan

This patch (as1584) fixes a minor bug that has been present in
ehci-hcd since the beginning.

Scanning the schedules for URB completions is single-threaded.  If a
completion interrupt occurs while an URB is being given back, the
interrupt handler realizes that a scan is in progress on another CPU
and avoids starting a new one.

This means that completion events can be lost.  If an URB completes
after it has been scanned but while a scan is still in progress, the
driver won't notice and won't rescan the completed URB.

The patch fixes the problem by adding a new flag to indicate that
another scan is needed after the current scan is done.  The flag gets
set whenever a completion interrupt occurs while a scan is in
progress.  The rescan will see the completion, thus preventing it from
getting lost.
Signed-off-by: Alan Stern <stern@rowland.harvard.edu>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

drivers/usb/host/ehci-hcd.c

@@ -400,14 +400,21 @@ static void ehci_work (struct ehci_hcd *ehci)
 	 * it reports urb completions.  this flag guards against bogus
 	 * attempts at re-entrant schedule scanning.
 	 */
-	if (ehci->scanning)
+	if (ehci->scanning) {
+		ehci->need_rescan = true;
 		return;
-	ehci->scanning = 1;
+	}
+	ehci->scanning = true;
+
+ rescan:
+	ehci->need_rescan = false;
 	if (ehci->async_count)
 		scan_async(ehci);
 	if (ehci->next_uframe != -1)
 		scan_periodic (ehci);
-	ehci->scanning = 0;
+	if (ehci->need_rescan)
+		goto rescan;
+	ehci->scanning = false;
 
 	/* the IO watchdog guards against hardware or driver bugs that
 	 * misplace IRQs, and should let us run completely without IRQs.

drivers/usb/host/ehci.h

@@ -113,7 +113,8 @@ struct ehci_hcd {			/* one per controller */
 	enum ehci_rh_state	rh_state;
 
 	/* general schedule support */
-	unsigned		scanning:1;
+	bool			scanning:1;
+	bool			need_rescan:1;
 	bool			intr_unlinking:1;
 	bool			async_unlinking:1;