Making INTRNG a requirement on armv6

[Warner Losh]

On Thu, Mar 16, 2017 at 8:44 PM, John Wehle <john at feith.com> wrote:
>> As for Amlogic ones, iirc, John did some tests on his devices with INTRNG
>> in December last year.
>
> The primary issue is regarding the older interrupt controller used on
> the aml8726-m3 and also present on the aml8726-m6.  I did the original
> porting work using the aml8726-m6 with the older interrupt controller
> which seemed to work fine ... there didn't seem to be any interrupts
> being lost.  It was only towards the end of the porting work that I
> switched the aml8726-m6 (which happens to have both the older controller
> and a GIC) to using the GIC.
>
> Back in December 2016 the aml8726-m6 (and -m8b) passed a smoke test using
> INTRNG with the GIC.  Disabling INTRNG caused the aml8726-m3 and the
> aml8726-m6 (configured for the older interrupt controller) to "lose"
> clock interrupts.
>
> I did at that time see about updating the older Amlogic interrupt code
> to use INTRNG which was a pretty straight forward effort, though even
> with the updated code the older interrupt controller still suffered
> from "lost" clock interrupts.  Unfortunately my attentions (as is
> probably evident) have been elsewhere so the "lost" clock interrupts
> issue is unresolved.
>
> I can see about providing the INTRNG version of the older Amlogic
> interrupt driver it that helps, however I'm probably not finding
> time to debug the "lost" clock interrupts issue anytime soon.

Almost working code that has an issue or two, especially if those are
documented, is much better than having to dump the port because we
want to require INTRNG... So yes, if it isn't a huge hassle, can you
dig that up?

Warner