Development Plan Proposal for Unifying Interrupt and PCI APIs

[Bogdan Vacaliuc]

Steve Holle wrote:
> Isn't interrupt prioritization very hardware specific?  Add a
> priority parameter and let the bsp handle it.  Should the
> differences between an 8259 and Coldfire interrupt priority scheme
> be incorporated into RTEMS? 

If the (new) RTEMS interrupt API provided a complete virtualization of interrupts, then the ideas of priority, sharing, etc. could
be incorporated.

In a private conversation I had with Thomas, I wrote that in a virtual interrupt scheme, each (virtual) interrupt level is unique
and defines its priority.  Thus prioritization is handled by the common interrupt API and sharing of interrupts is a BSP managed
item.

Eric Valette wrote:
> priorities should not be defined individually but as a whole for the
> BSP : how can you choose a single IRQ priority when you do not know the
> other irq priorities? So an API with individual priorities at connect
> is just plain wrong I think.

I wonder why?  The priorities are known and statically configured by the BSP designer.  The code to deal with it can be common,
which is what I think we are trying to achieve.

> is just plain wrong I think. Side comment : the notion of priority is
> less rich than the notion of irq masking set...

So then the common API should provide a means to do this as well.

-bogdan