RE: [PATCH] usb: xhci: Workaround for erratum-A010129

[Sriram Dash]

>From: Alan Stern [mailto:stern@xxxxxxxxxxxxxxxxxxx]
>On Fri, 21 Oct 2016, Sriram Dash wrote:
>
>> For the USB3.0 controller, USB 2.0 reset not driven while port is in
>> Resume state. So, do not program the USB 2.0 reset
>> (PORTSC[PR]=1) while in Resume state.
>>
>> Signed-off-by: Rajat Srivastava <rajat.srivastava@xxxxxxx>
>> Signed-off-by: Sriram Dash <sriram.dash@xxxxxxx>
>> Signed-off-by: Rajesh Bhagat <rajesh.bhagat@xxxxxxx>
>> ---
>>  drivers/usb/host/xhci-hub.c | 28 +++++++++++++++++++++++++++-
>>  drivers/usb/host/xhci.h     |  1 +
>>  2 files changed, 28 insertions(+), 1 deletion(-)
>>
>> diff --git a/drivers/usb/host/xhci-hub.c b/drivers/usb/host/xhci-hub.c
>> index 730b9fd..3def0dd 100644
>> --- a/drivers/usb/host/xhci-hub.c
>> +++ b/drivers/usb/host/xhci-hub.c
>> @@ -878,7 +878,7 @@ int xhci_hub_control(struct usb_hcd *hcd, u16 typeReq,
>u16 wValue,
>>  	struct xhci_hcd	*xhci = hcd_to_xhci(hcd);
>>  	int max_ports;
>>  	unsigned long flags;
>> -	u32 temp, status;
>> +	u32 temp, status, tmp_status = 0;
>>  	int retval = 0;
>>  	__le32 __iomem **port_array;
>>  	int slot_id;
>> @@ -1098,6 +1098,32 @@ int xhci_hub_control(struct usb_hcd *hcd, u16
>typeReq, u16 wValue,
>>  			spin_lock_irqsave(&xhci->lock, flags);
>>  			break;
>>  		case USB_PORT_FEAT_RESET:
>> +			/*
>> +			 * Erratum : A010129
>> +			 * Synopsys STAR 9000962562.
>> +			 * USB 2.0 Reset Not Driven While Port in Resume
>> +			 * While in USB 2.0 resume state (the PORTSC.PLS
>> +			 * register bit is set to 4'd15), if the xHCI driver
>> +			 * programs the PORTSC.PR register bit to 1, the
>> +			 * controller does not drive a USB 2.0 reset
>> +			 * and it does not generate a PORTSC.PRC=1 interrupt.
>> +			 * So, The xHCI driver should not program a USB
>> +			 * 2.0 reset (PORTSC.PR=1) while in resume
>> +			 * (PORTSC.PLS=4'd15).
>> +			 */
>> +			if (xhci->quirks & XHCI_PORT_RST_ON_RESUME) {
>
>It's always a bad idea to drive a reset signal while a port is resuming.  We don't
>need a quirk flag for this.  

Hi Alan,

Ok. Then I will make it generic for xhci in the next version.

>The HCD should never do it, and the USB core should
>never tell the HCD to do it.
>
>Alan Stern