Re: usb: dwc2: Detecting cable disconnection in OTG mode?

[Pavel Hofman <pavel.hofman@ivitera.com>]

Dne 11. 12. 21 v 10:59 Pavel Hofman napsal(a):
> Dne 11. 12. 21 v 10:55 Pavel Hofman napsal(a):
>> Hi Minas,
>>
>> I am trying to find if dwc2 in OTG mode can pass information to gadget 
>> functions that the host has been disconnected. I am testing on RPi4. 
>> In a datasheet for a different dwc2 implementation 
>> https://www.mouser.cn/datasheet/2/196/Infineon-xmc4500_rm_v1.6_2016-UM-v01_06-EN-598157.pdf 
>> - chapter "16.8.3 Device Disconnection" I found:
>>
>>
>> ===========
>> The device session ends when the USB cable is disconnected or if the 
>> VBUS is switched off by the host.
>> The device disconnect flow is as follows:
>>
>> 1. When the USB cable is unplugged or when the VBUS is switched off by 
>> the host, the device core triggers GINTSTS.OTGInt [bit 2] interrupt bit
>> ============
>>
>> I put a printk to core_intr.c:dwc2_handle_otg_intr() which is called 
>> only from handling the OTGInt interrupt 
>> https://elixir.bootlin.com/linux/v5.1.9/source/drivers/usb/dwc2/core_intr.c#L803 
>> . But this method is not called at all when disconnecting the USB 
>> cable on RPi4 in gadget mode.
>>
> 
> To add, the method dwc2_hsotg_disconnect is called, but after 
> reconnecting the cable, as part of the gadget reset  when handling reset 
> interrupts (gintsts & (GINTSTS_USBRST | GINTSTS_RESETDET)) here 
> https://elixir.bootlin.com/linux/v5.1.9/source/drivers/usb/dwc2/gadget.c#L3653 
> 


Maybe the reason for the OTGInt not being thrown at cable disconnect is 
the fact that RPi4 most likely does not handle USB_OTG_ID signal. I 
asked about USB_OTG_ID wiring on RPi4 at their forum.

Nevertheless it turns out that DWC2 throws Suspend interrupt 
GINTSTS_USBSUSP (11) at cable disconnection on the RPi4, and the gadget 
composite driver calls suspend hook of struct usb_function. I wonder if 
implementing the suspend handler would be a solution for other dwc2 
implementations, or if the audio function should implement also some 
other handler in order to correctly serve the cable disconnection for 
all dwc2 implementations.

Thanks a lot,

Pavel.