[Bug] usb: dwc2: Add functions to set and clear force mode

[Caesar Wang]

Hi John,

在 2016年02月02日 07:31, John Youn 写道:
> On 2/1/2016 3:13 AM, Caesar Wang wrote:
>>
>> 在 2016年02月01日 19:12, Caesar Wang 写道:
>>> Doug
>>>
>>> 在 2016年01月30日 03:11, Doug Anderson 写道:
>>>> Caesar,
>>>>
>>>> On Tue, Jan 26, 2016 at 6:12 PM, Caesar Wang
>>>> <caesar.upstream at gmail.com> wrote:
>>>>> Thanks Doug's reply.
>>>>>
>>>>> Cc: Wulf
>>>>> 在 2016年01月27日 00:05, Doug Anderson 写道:
>>>>>> Hi,
>>>>>>
>>>>>> On Tue, Jan 26, 2016 at 4:02 AM, Caesar Wang <wxt at rock-chips.com>
>>>>>> wrote:
>>>>>>> Hi  John, Felipe
>>>>>>>
>>>>>>>
>>>>>>> I'm no familiar with usb stuff.
>>>>>>> then I found this patch will break usb working for rk3036 SoCs, maybe
>>>>>>> more
>>>>>>> SoCs.
>>>>>>> Says, U disk can't work on usb host.
>>>>>>>
>>>>>>> The failure log:
>>>>>>>
>>>>>>>       32.645481] usb usb2-port1: connect-debounce failed
>>>>>>>
>>>>>>>
>>>>>>> Tested by following branch:
>>>>>>> https://github.com/Caesar-github/rockchip/tree/kylin/next (kernel:
>>>>>>> 4.5-rc1)
>>>>>>>
>>>>>>> Revert "usb: dwc2: Add functions to set and clear force mode" will
>>>>>>> work
>>>>>>> for
>>>>>>> it.
>>>>>>>
>>>>>>>
>>>>>>> Maybe, someone have some suggestions or ideas?
>>>>>> Can you check if this series helps you?
>>>>>>
>>>>>> http://marc.info/?l=linux-usb&m=145255851516121&w=2
>>>>> Unluckily, this series patches can't fix it on rk3036 SoCs.
>>>>> I revert this CL ("usb: dwc2: Add functions to set and clear force
>>>>> mode") to
>>>>> work firstly.
>>>> You're 100% positive?  In particular you made sure you had this patch
>>>> (one of the two in John's series I pointed at), which explicitly
>>>> mentions fixing a problem with the patch you mention?
>>> I'm 100% positive the John's patches can fix this issue
>> Sorry, John's patches can't fix this issue.
>>
>>
>>> As the following verified on
>>> https://github.com/Caesar-github/rockchip/commits/for-usb-tests
>>>
>>> Anyway, Meanwhile if we add the folllowing patch can work for me. I
>>> will track it on tomorrow if you have other suggestions.
>>>
>>> diff --git a/drivers/usb/dwc2/core.c b/drivers/usb/dwc2/core.c
>>>
>>> index e991d55..90f4abf 100644
>>>
>>> --- a/drivers/usb/dwc2/core.c
>>>
>>> +++ b/drivers/usb/dwc2/core.c
>>>
>>> @@ -637,6 +637,13 @@ int dwc2_core_reset_and_force_dr_mode(struct
>>> dwc2_hsotg *hsotg)
>>>
>>>                  return retval;
>>>
>>>
>>>
>>>          dwc2_force_dr_mode(hsotg);
>>>
>>> +
>>>
>>> +       /*
>>>
>>> +        * NOTE: This long sleep is_very_  important, otherwise the core
>>> will
>>>
>>> +        * not stay in host mode after a connector ID change!
>>>
>>> +        */
>>>
>>> +        usleep_range(150000, 160000);
>>>
>>> +
>>>
>>>          return 0;
>>>
>>>   }
>>>
> Hi Cesar,
>
> On your platform, is the dwc2 controller OTG? If so, what mode is the
> driver running in (dr_mode=HOST, PERIPHERAL, or OTG)?


It's the Host mode. ( USB_DR_MODE_HOST)

>
> The force mode only requires 25ms delay but maybe the sleep needs to
> be longer on your system for some other reason. Which is unfortunate
> because 150ms is very long.
>
> Under what condition do you see the problem. Is it on connector ID
> change? If so we might put the long delay in a more appropriate
> location so as not to affect delay in probe.
>
> Also, could you try *only* reverting the following:
>
> commit 97e463886b873f62bea2293e7edf81fdb884b84f ("usb: dwc2: Reduce
> delay when forcing mode in reset")

I'm sure can't work if we revert it.
log
...:
"[    5.705971] usb usb2-port1: connect-debounce failed"

d4b3ff1 Revert "usb: dwc2: Reduce delay when forcing mode in reset"
c69654f usb: dwc2: Fix probe problem on bcm283
b9da0ff Revert "usb: dwc2: Move reset into dwc2_get_hwparams()

--

Then, add the following patch can work.

diff --git a/drivers/usb/dwc2/core.c b/drivers/usb/dwc2/core.c
index e991d55..90f4abf 100644
--- a/drivers/usb/dwc2/core.c
+++ b/drivers/usb/dwc2/core.c
@@ -637,6 +637,13 @@ int dwc2_core_reset_and_force_dr_mode(struct 
dwc2_hsotg *hsotg)
                 return retval;

         dwc2_force_dr_mode(hsotg);
+
+       /*
+        * NOTE: This long sleep is _very_ important, otherwise the core 
will
+        * not stay in host mode after a connector ID change!
+        */
+        usleep_range(150000, 160000);
+
         return 0;
  }

> The original patch you referred to preserves the long delay at the
> point it is used for OTG but then this later patch (97e46388)
> decreased the value. So maybe we can just revert this later patch.
>
> Regards,
> John
> _______________________________________________
> Linux-rockchip mailing list
> Linux-rockchip at lists.infradead.org
> http://lists.infradead.org/mailman/listinfo/linux-rockchip