On 2/2/2016 5:10 PM, Caesar Wang wrote: > > > ? 2016?02?03? 09:08, John Youn ??: >> On 2/2/2016 4:36 PM, Caesar Wang wrote: >>> Hi John, >>> >>> ? 2016?02?03? 08:03, John Youn ??: >>>> On 2/1/2016 6:53 PM, Caesar Wang wrote: >>>>> John, >>>>> >>>>> I will suggest the msleep(25) delay should put in >>>>> 'dwc2_force_dr_mode()' instead of the 'dwc2_clear_force_mode()' >>>>> >>>> Hi Caesar, >>>> >>>> Are you saying that just msleep(25) in that function like this solves >>>> your issue? >>>> >>>> diff --git a/drivers/usb/dwc2/core.c b/drivers/usb/dwc2/core.c >>>> index 39a0fa8..e8a9688 100644 >>>> --- a/drivers/usb/dwc2/core.c >>>> +++ b/drivers/usb/dwc2/core.c >>>> @@ -625,6 +625,8 @@ void dwc2_force_dr_mode(struct dwc2_hsotg *hsotg) >>>> __func__, hsotg->dr_mode); >>>> break; >>>> } >>>> + >>>> + msleep(25); >>> Yep, that will solve this issue. >>> >> Ok thanks. Though that still seems strange since you should be getting >> the msleep(25) from either the dwc2_force_mode() or >> dwc2_clear_force_mode(). >> >> Can you check if the following patch works? >> >> If so, I think we can go with that to fix the regression for >> now. Otherwise I'll just add the extra msleep(25) and revisit this >> later. >> >> Regards, >> John >> >> ---->8---- >> >> diff --git a/drivers/usb/dwc2/core.c b/drivers/usb/dwc2/core.c >> index e991d55..13c060c 100644 >> --- a/drivers/usb/dwc2/core.c >> +++ b/drivers/usb/dwc2/core.c >> @@ -576,7 +576,6 @@ static bool dwc2_force_mode(struct dwc2_hsotg *hsotg, bool host) >> gusbcfg |= set; >> dwc2_writel(gusbcfg, hsotg->regs + GUSBCFG); >> >> - msleep(25); >> return true; >> } >> >> @@ -596,7 +595,6 @@ static void dwc2_clear_force_mode(struct dwc2_hsotg *hsotg) >> * NOTE: This long sleep is _very_ important, otherwise the core will >> * not stay in host mode after a connector ID change! >> */ >> - msleep(25); >> } >> >> /* >> @@ -619,6 +617,8 @@ void dwc2_force_dr_mode(struct dwc2_hsotg *hsotg) >> __func__, hsotg->dr_mode); >> break; >> } >> + >> + msleep(25); >> } > > Right, that works for me. > Ok, interesting. Actually this won't work either since we need the separate msleeps when those functions are called from other contexts. Is your controller by chance *not* OTG capable, i.e. dwc2_hw_is_otg(hsotg) == false Then the msleep in dwc2_force_mode() is skipped. Then, for some reason, skipping it (even though nothing was forced) causes the failure on your platform. Regards, John
