> On Mon, 2020-07-20 at 03:44 +0000, Peter Chen wrote: > > > > > On Wed, 2020-07-15 at 00:51 +0000, Peter Chen wrote: > > > > > The Toradex Colibri iMX6ULL board has a special USB hardware > > > > > design. > > > > > With runtime-pm enabled USB reset itself continuously. > > > > > Furthermore > > > > > the OTG port is also not enumerating devices if the Chipidea IP > > > > > is in runtime sleep mode and a device or host gets plugged in. > > > > > > > > > > > > > Hi Philippe, > > > > > > > > You may describe the detail what's the special USB hardware design > > > > for your board, > > > > > > If I only knew the root-cause of that problem - unfortunately I > > > don't. > > > That's also why I have such a hard time to describe it. > > > > > > > and why it causes the problem, and why disable runtime pm could > > > > fix this issue, then, > > > > > > I cannot provide the 'why' part yet. I'll try something more and > > > hope I can provide you guys with the exact description. > > > > > > > the other users could know if it could apply to their platforms or > > > > not in future. > > > > > > I only found out about it because you were pointing me in that > > > direction. I debugged for hours now and didn't came to the > > > root-cause of the issue. I think to really understand it I would > > > need to know much more about the Chipidea IP. > > > > > > I'll get back to you guys with a proposal for a new description. > > > > > > > Philippe, is it possible to share your USB hardware design at 6ULL? > > It's actually pretty simple: We have on USB_OTG1_VBUS a 1uF capacitor and > +3.0V on VDD_USB_CAP together with 100n and 10u bypass caps. Now the big > problem is that the driver can not detect the 5V on VBUS signal. > Could you confirm it does not see VBUS at register OTGSC? If it is, how can it work with runtime disabled, the USBCMD.RS setting (ci_hdrc_gadget_connect is called) depends on VBUS. Peter > I tried to 'inject' 5V to that pin last week and things got really better with runtime-pm. > But I still thinks disabling it for our board would make sense. > > I'll send a new description today where I try to point to VBUS signal not connected. > > Philippe > > > And how ci_hdrc_gadget_connect is called when the runtime pm is > > disabled? > > > > Thanks, > > Peter > >