Re: [PATCH v2 2/2] tests/qtest: Check that EXTI fan-in irqs are correctly connected

[Peter Maydell <peter.maydell@linaro.org>]

On Thu, 22 Feb 2024 at 15:09, Peter Maydell <peter.maydell@linaro.org> wrote:
>
> On Tue, 20 Feb 2024 at 18:41, Inès Varhol <ines.varhol@telecom-paris.fr> wrote:
> >
> > This commit adds a QTest that verifies each input line of a specific
> > EXTI OR gate can influence the output line.
> >
> > Signed-off-by: Inès Varhol <ines.varhol@telecom-paris.fr>
> > Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
> > ---
> >
> > Hello,
> >
> > I expected this test to fail after switching the two patch commits,
> > but it didn't.
> > I'm mentionning it in case it reveals a problem with the test I didn't notice.
>
> The specific thing that goes wrong if you don't have the OR
> gate handling is that the NVIC input will see every up
> and down transition from each input separately. (This happens
> because a GPIO/irq 'input' in QEMU is basically a function,
> and wiring up an 'output' to an 'input' is setting "this
> is the function pointer you should call when the output
> changes". Nothing syntactically stops you passing the
> same function pointer to multiple outputs.)
>
> So if you have for instance
>  raise A; raise B; drop B; drop A
> where A and B are ORed together into an NVIC input,
> the NVIC input is supposed to see the line go high
> at "raise A" and only drop at the last "drop B". Without

...at the last "drop *A*", I mean.

> the OR gate, it will see it go high at "raise A", and then
> drop at "drop B". (Well, it sees "level is 1", "level is 1",
> "level is 0", "level is 0", but inputs expect to sometimes see
> calls for "level happens to be the same thing it was
> previously", so it doesn't cause the NVIC to change state.)

-- PMM