Re: [PATCH 07/10] irqchip/cs42l43: Add support for the cs42l43 IRQs

[Charles Keepax]

On Tue, May 16, 2023 at 11:09:36AM +0100, Lee Jones wrote:
> On Tue, 16 May 2023, Marc Zyngier wrote:
> > On Mon, 15 May 2023 12:25:54 +0100,
> > Lee Jones <lee@xxxxxxxxxx> wrote:
> > > On Fri, 12 May 2023, Marc Zyngier wrote:
> > > > On Fri, 12 May 2023 16:39:33 +0100,
> > > > Charles Keepax <ckeepax@xxxxxxxxxxxxxxxxxxxxx> wrote:
> > > I'm not aware of another subsystem that deals with !IRQChip level IRQ
> > > controllers.  Where do simple or "second class" interrupt controllers
> > > go?
> > 
> > This isn't an interrupt controller. This is internal signalling, local
> > to a single component that has been artificially broken into discrete
> > bits, including an interrupt controller. The only *real* interrupts
> > here are the GPIOs.
> > 

I would question this statement a little, they are fixed function
IRQs sure but they are still real interrupts. These are lines which
receive a signal and on an edge they set a stick status bit, which
causes another signal to generate an edge, they have registers
which let you mask events, if it walks like a duck and all. The
only difference between this and a "real" interrupt is whether the
chip designer or the board designer was the person who decided
where the wire was connected.

> > I'm happy to see an interrupt controller for the GPIOs. But the rest
> > is just internal muck that doesn't really belong here. Where should it

Internal-ish, granted many of them are primarily useful to the
device itself. But it is very easy to construct situations where
say knowing the speaker thermals are high, or that a jack has
been inserted are useful outside of the CODEC driver itself.

> > go? Together with the rest of the stuff that manages the block as a
> > whole. Which looks like the MFD subsystem to me.
> 
> Very well.  Let's see this "muck" in a patch please!

Groovy I will do a re-spin moving the IRQ stuff to the MFD and
lets see where we get to.

Thank you all for your help in reviewing this so far.

Thanks,
Charles