Re: [RFC] usb: dwc3: core: Fix RAM interface getting stuck during enumeration

[Krishna Kurapati PSSNV <quic_kriskura@quicinc.com>]

On 10/13/2023 12:55 AM, Wesley Cheng wrote:
> Hi Krishna,
> 
> On 10/11/2023 3:02 AM, Krishna Kurapati wrote:
>> This implementation is to fix RAM interface getting stuck during

>> Synopsys confirmed that the issue is present on all USB3 devices and
>> as a workaround, suggested to re-initialize device mode.
>>
>> Signed-off-by: Krishna Kurapati <quic_kriskura@quicinc.com>
>> ---
>>   drivers/usb/dwc3/core.c   | 20 ++++++++++++++++++++
>>   drivers/usb/dwc3/core.h   |  4 ++++
>>   drivers/usb/dwc3/drd.c    |  5 +++++
>>   drivers/usb/dwc3/gadget.c |  6 ++++--
>>   4 files changed, 33 insertions(+), 2 deletions(-)
>>
>> diff --git a/drivers/usb/dwc3/core.c b/drivers/usb/dwc3/core.c
>> index 44ee8526dc28..d18b81cccdc5 100644
>> --- a/drivers/usb/dwc3/core.c
>> +++ b/drivers/usb/dwc3/core.c
>> @@ -122,6 +122,7 @@ static void __dwc3_set_mode(struct work_struct *work)
>>       unsigned long flags;
>>       int ret;
>>       u32 reg;
>> +    u8 timeout = 100;
>>       u32 desired_dr_role;
>>       mutex_lock(&dwc->mutex);
>> @@ -137,6 +138,25 @@ static void __dwc3_set_mode(struct work_struct 
>> *work)
>>       if (!desired_dr_role)
>>           goto out;
>> +    /*
>> +     * STAR 5001544 - If cable disconnect doesn't generate
>> +     * disconnect event in device mode, then re-initialize the
>> +     * controller.
>> +     */
>> +    if ((dwc->cable_disconnected == true) &&
>> +        (dwc->current_dr_role == DWC3_GCTL_PRTCAP_DEVICE)) {
>> +        while (dwc->connected == true && timeout != 0) {
>> +            mdelay(10);
>> +            timeout--;
>> +        }
>> +
>> +        if (timeout == 0) {
>> +            dwc3_gadget_soft_disconnect(dwc);
>> +            udelay(100);
>> +            dwc3_gadget_soft_connect(dwc);
>> +        }
>> +    }
>> +
>>       if (desired_dr_role == dwc->current_dr_role)
>>           goto out;
>> diff --git a/drivers/usb/dwc3/core.h b/drivers/usb/dwc3/core.h
>> index c6c87acbd376..7642330cf608 100644
>> --- a/drivers/usb/dwc3/core.h
>> +++ b/drivers/usb/dwc3/core.h
>> @@ -1355,6 +1355,7 @@ struct dwc3 {
>>       int            last_fifo_depth;
>>       int            num_ep_resized;
>>       struct dentry        *debug_root;
>> +    bool            cable_disconnected;
>>   };
>>   #define INCRX_BURST_MODE 0
>> @@ -1568,6 +1569,9 @@ void dwc3_event_buffers_cleanup(struct dwc3 *dwc);
>>   int dwc3_core_soft_reset(struct dwc3 *dwc);
>> +int dwc3_gadget_soft_disconnect(struct dwc3 *dwc);
>> +int dwc3_gadget_soft_connect(struct dwc3 *dwc);
>> +
>>   #if IS_ENABLED(CONFIG_USB_DWC3_HOST) || 
>> IS_ENABLED(CONFIG_USB_DWC3_DUAL_ROLE)
>>   int dwc3_host_init(struct dwc3 *dwc);
>>   void dwc3_host_exit(struct dwc3 *dwc);
>> diff --git a/drivers/usb/dwc3/drd.c b/drivers/usb/dwc3/drd.c
>> index 039bf241769a..593c023fc39a 100644
>> --- a/drivers/usb/dwc3/drd.c
>> +++ b/drivers/usb/dwc3/drd.c
>> @@ -446,6 +446,8 @@ static int dwc3_usb_role_switch_set(struct 
>> usb_role_switch *sw,
>>       struct dwc3 *dwc = usb_role_switch_get_drvdata(sw);
>>       u32 mode;
>> +    dwc->cable_disconnected = false;
>> +
>>       switch (role) {
>>       case USB_ROLE_HOST:
>>           mode = DWC3_GCTL_PRTCAP_HOST;
>> @@ -454,6 +456,9 @@ static int dwc3_usb_role_switch_set(struct 
>> usb_role_switch *sw,
>>           mode = DWC3_GCTL_PRTCAP_DEVICE;
>>           break;
>>       default:
>> +        if (role == USB_ROLE_NONE)
>> +            dwc->cable_disconnected = true;
>> +
> 
> How do we handle cases where role switch isn't used? (ie extcon or maybe 
> no cable connection notification at all)
> 

Hi Wesley,

  Since I was considering fixing it during disconnect, I made it in role 
switch. So no cable connection notification case has been ruled out in 
this patch. But yes, extcon is a valid case. Will need to account for it.

Regards,
Krishna,