[PATCH 48/74] GIC: Added dummy handlers for Power Management Suspend Resume

[Russell King - ARM Linux]

On Mon, Sep 20, 2010 at 11:55:50AM -0500, Rob Herring wrote:
> There is no programming distinction between an irq and an wakeup if the  
> GIC is powered on in suspend which is platform dependent. It is only a  
> s/w distinction of what interrupts you allow to wake you up. If GIC  
> remains powered, then it can assert nirq to the core and wake-up. If the  
> core is powered down, the external power controller can monitor the NIRQ  
> line to wake-up the core. So the powergating of the core has no impact  
> on the prvgramming of the GIC for wake-up.

Well, not everyone does this - and I know of no platforms where it has
been reported for the GIC to be able to wake the system up.  TI OMAP
doesn't allow it, neither does the one being discussed in this thread.

The system being discussed here has an external power controller which
needs to be appropriately programmed.

You can see this because the proposed patch doesn't add code like that
in vic.c, but adds a no-operation function so that set_wake() becomes
a no-op returning zero.

> How about adding a set_wake function ptr as a parameter to  
> gic_dist_init? If a platform needs more than the basic operation, they  
> can override it.

As I can see, we're not talking about wake-up being able to be done from
the IRQ controller here, but something entirely external and unrelated
to it.