Re: [PATCH 1/2] USB: reset-resume the device when PORT_SUSPEND is set but timeout

[Alan Stern]

On Fri, May 07, 2021 at 05:33:28PM +0800, chris.chiu@xxxxxxxxxxxxx wrote:
> From: Chris Chiu <chris.chiu@xxxxxxxxxxxxx>
> 
> On the Realtek hub of Dell Dock WD19, the port which has wakeup
> enabled descendants will sometimes timeout when setting PORT_SUSPEND
> feature. After checking the PORT_SUSPEND bit in wPortStatus, it is
> already set. However, the hub will fail to activate because the
> PORT_SUSPEND feature of that port is not cleared during resume. All
> devices connecting via the port are lost after resume.
> 
> This commit force reset-resume the device connected to the timeout
> but suspended port so that the hub will have chance to clear the
> PORT_SUSPEND feature during resume.
> 
> Signed-off-by: Chris Chiu <chris.chiu@xxxxxxxxxxxxx>
> ---
>  drivers/usb/core/hub.c | 15 +++++++++++++++
>  1 file changed, 15 insertions(+)
> 
> diff --git a/drivers/usb/core/hub.c b/drivers/usb/core/hub.c
> index b2bc4b7c4289..18603949a8de 100644
> --- a/drivers/usb/core/hub.c
> +++ b/drivers/usb/core/hub.c
> @@ -3385,6 +3385,21 @@ int usb_port_suspend(struct usb_device *udev, pm_message_t msg)
>  		status = 0;
>  	}
>  	if (status) {
> +		if (status == -ETIMEDOUT) {
> +			u16 portstatus, portchange;
> +
> +			status = hub_port_status(hub, port1, &portstatus,
> +					&portchange);

Don't reuse status like this.  There will be a problem if this line sets 
status to 0 but the port isn't actually suspended.  Use a different 
variable instead.

Alan Stern

> +
> +			dev_dbg(&port_dev->dev,
> +				"suspend timeout, status %04x\n", portstatus);
> +
> +			if (status == 0 && port_is_suspended(hub, portstatus)) {
> +				udev->reset_resume = 1;
> +				goto err_wakeup;
> +			}
> +		}
> +
>  		dev_dbg(&port_dev->dev, "can't suspend, status %d\n", status);
>  
>  		/* Try to enable USB3 LTM again */
> -- 
> 2.20.1
>