USB: EHCI: ensure all watchdog timer events are deleted when suspending usb

This patch was previously discussed in the following thread:
http://thread.gmane.org/gmane.linux.usb.general/19472/focus=19484

On the OMAP3 device the usbhost controller is in a separate internal
power-domain. So when the usbhost is inactive or suspend is called,
we can disable clocks and power-down the usbhost to save power.

Recently we found that after calling ehci_bus_suspend() and disabling
the usbhost clocks we would see the ehci watchdog timer event fire. This
was causing a kernel panic because the usbhost controllers clocks were
disabled and inside the watchdog timer function the clocks were not
being re-enabled, so when the ehci registers were accessed this resulted
in a CPU data-abort.

To avoid this panic, per recommendation from Alan Stern (see above thread), we
make sure any pending timer events (that may have been scheduled by calling
ehci_work within the ehci_bus_suspend() function) are deleted before returning.

Signed-off-by: Fei Yang <fei.yang@motorola.com>
Signed-off-by: Jon Hunter <jon-hunter@ti.com>
Acked-by: Alan Stern <stern@rowland.harvard.edu>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>

diff --git a/drivers/usb/host/ehci-hub.c b/drivers/usb/host/ehci-hub.c
index 6fef1ee7d1010e9cc223290e73054cc0f60ccdc0..818647c33da889eced7bb700bb3ba7500f4f2030 100644 (file)
--- a/drivers/usb/host/ehci-hub.c
+++ b/drivers/usb/host/ehci-hub.c
@@ -209,6 +209,11 @@ static int ehci_bus_suspend (struct usb_hcd *hcd)
 
        ehci->next_statechange = jiffies + msecs_to_jiffies(10);
        spin_unlock_irq (&ehci->lock);
+
+       /* ehci_work() may have re-enabled the watchdog timer, which we do not
+        * want, and so we must delete any pending watchdog timer events.
+        */
+       del_timer_sync(&ehci->watchdog);
        return 0;
 }