This patch (as1652) fixes a long-standing bug in ehci-hcd. The driver relies on status polls to know when to stop port-resume signalling. It uses the root-hub status timer to schedule these status polls. But when the driver for the root hub is resumed, the timer is rescheduled to go off immediately -- before the port is ready. When this happens the timer does not get re-enabled, which prevents the port resume from finishing until some other event occurs.
The symptom is that when a new device is plugged in, it doesn't get recognized or enumerated until lsusb is run or something else happens. The solution is to re-enable the root-hub status timer after every status poll while a port resume is in progress. This bug hasn't surfaced before now because we never used to try to suspend the root hub in the middle of a port resume (except by coincidence). Signed-off-by: Alan Stern <[email protected]> Reported-and-tested-by: Norbert Preining <[email protected]> Tested-by: Ming Lei <[email protected]> CC: <[email protected]> --- Even though this bug is present in all kernels, its effects haven't shown up because hubs used to have a 2-second autosuspend delay. As of 3.8-rc1 that is no longer true, and the bug definitely shows up in the current -rc kernels. It would be good to merge this patch before 3.8-final. drivers/usb/host/ehci-hub.c | 6 +++++- 1 file changed, 5 insertions(+), 1 deletion(-) Index: usb-3.8/drivers/usb/host/ehci-hub.c =================================================================== --- usb-3.8.orig/drivers/usb/host/ehci-hub.c +++ usb-3.8/drivers/usb/host/ehci-hub.c @@ -649,7 +649,11 @@ ehci_hub_status_data (struct usb_hcd *hc status = STS_PCD; } } - /* FIXME autosuspend idle root hubs */ + + /* If a resume is in progress, make sure it can finish */ + if (ehci->resuming_ports) + mod_timer(&hcd->rh_timer, jiffies + msecs_to_jiffies(25)); + spin_unlock_irqrestore (&ehci->lock, flags); return status ? retval : 0; } -- To unsubscribe from this list: send the line "unsubscribe linux-usb" in the body of a message to [email protected] More majordomo info at http://vger.kernel.org/majordomo-info.html
