Re: [PATCH] usb: dwc3: add quirk to handle DWC_USB3_NUM == DWC_USB3_NUM_IN_EPS

[Felipe Balbi <balbi-DgEjT+Ai2ygdnm+yROfE0A@public.gmane.org>]

[-- Attachment #1: Type: text/plain, Size: 1484 bytes --]


Hi,

Bryan O'Donoghue <pure.logic-SyKdqv6vbfZdzvEItQ6vdLNAH6kLmebB@public.gmane.org> writes:
> On 19/01/17 22:49, John Youn wrote:
>
>> So it is valid to have say, DWC_USB3_NUM_EPS=8 and
>> DWC_USB3_NUM_IN_EPS=8. Even though it is not possible to use all 8 EPs
>> as IN since you need at least one of them to be a control OUT. So you
>> could have a configuration of EP0 IN and OUT plus 6 IN EPs (total 1
>> OUT, 7 IN). Or EP0 IN and OUT plus any other combination of IN/OUT for
>> the remaining 6 EPs.
>> 
>> With the above in mind, you can probably just redo the endpoint number
>> logic in dwc3 to handle all cases without any quirks at all.
>
> I thought of that.
>
>> If it is disabled, it will fix the physical and logical EP number and
>> direction of the available EPs to meet timings for FPGA designs. This
>> shouldn't be used in final designs for ASICS but we don't know for
>> sure whether it has made it to any final designs or not.
>> 
>> And unfortunately, whether this is set or not is not visible to the
>> software so it will require a quirk.
>
> but arrived at this conclusion because I couldn't think of a reasonable
> guess value for IN/OUT endpoint numbers that would work if
> DWC_USB3_NUM_EPS == DWC_USB3_NUM_IN_EPS

Well, we can, for now at least, take the simple approach of half IN,
half OUT. If DWC3_USB3_NUM_EPS is odd, then OUT should take the extra
endpoint. If anybody complains, we can fix it later ;-)

-- 
balbi

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 832 bytes --]