Re: [PATCH v3 2/4] extcon: usb-gpio: add support for VBUS detection

[Robert Baldyga]

On 04/09/2015 11:59 AM, Roger Quadros wrote:
> Hi,
> 
> On 09/04/15 12:24, Robert Baldyga wrote:
>> Hi Chanwoo,
>>
>> On 04/09/2015 11:07 AM, Chanwoo Choi wrote:
>>> Hi Robert,
>>>
>>> On 04/09/2015 04:57 PM, Robert Baldyga wrote:
>>>> Hi Chanwoo,
>>>>
>>>> On 04/09/2015 04:12 AM, Chanwoo Choi wrote:
>>>>> Hi Robert,
>>>>>
>>>
>>> [snip]
>>>
>>>>> But, I have one question about case[3]
>>>>>
>>>>> If id is low and vbus is high, this patch will update the state of both USB and USB-HOST cable as attached state.
>>>>> Is it possible that two different cables (both USB and USB-HOST) are connected to one port simultaneously?
>>>>>
>>>>
>>>> It's because state of single USB cable connection cannot be completely
>>>> described using single extcon cable. USB cable state has two bits (VBUS
>>>> and ID), so we need to use two cables for single cable connection. We
>>>> use following convention:
>>>> cable "USB" = VBUS
>>>> cable "USB-HOST" = !ID.
>>>
>>> I think that extcon provider driver have to update the only one cable state
>>> of either USB or USB-HOST because USB and USB-HOST feature can not be used
>>> at the same time through one h/w port.
> 
> At least for the kernel users [1] we are treating USB-HOST as !ID and USB as VBUS.
> So it is not an issue for these kernel users if both USB and USB-HOST are attached.
> This is a valid USB state.
> If we don't do so then extcon with 3 cable states is not sufficient to capture the
> entire USB scenario. (we need 4 states for 2 pins).
> 
> [1]
> - drivers/usb/phy/phy-omap-otg.c
> - drivers/usb/dwc3/dwc3-omap.c
> 
> 
>>>
>>> If extcon-usb-gpio.c update two connected event of both USB and USB-HOST cable
>>> at the same time, the extcon consumer driver can not decide what handle either USB or USB-HOST.
>>>
>>
>> It can. USB OTG allows for that. Moreover device can be host even if
>> ID=1 (so detected cable type is USB device), or peripheral when ID=0 (so
>> detected cable type is USB host). Devices would need to have complete
>> information about USB cable connection, because OTG state machine needs
>> that. As I wrote, current USB cable names are misleading. It would be
>> better to have "USB-VBUS" and "USB-ID".
> 
> We need to first understand how user space is using "USB" and "USB-HOST" events
> and does it cause an issue if both USB and USB-HOST become attached.
> 
> What is the "ABI" explanation for "USB" and "USB-HOST" cable states?
> 

We can also leave USB and USB-HOST as "dummy cable detection states",
like they currently are, and add new USB-VBUS and USB-ID cables without
removing the old ones. It will cause some redundancy, but will make us
sure, that no ABI break can have a place.

Thanks,
Robert Baldyga
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/