MPC8xx interrupts

[Artemii S. Ivanov]

Ok.
Just another question:

Should one move MPC8xx specifc interrupt processing/dispatching from
lib/libbsp/powerpc/mbx8xx/irq (or lib/libbsp/powerpc/eth_comm/irq)
to the lib/libcpu/powerpc/mpc8xx/ directory? Will it be look rather
generic (it seems that this code could be employed in another MPC8xx
bsps)?


On Wed, 24 Oct 2001, VALETTE Eric wrote:

> Artemii S. Ivanov wrote:
>
> > Hello, all MPC8xx users!
> >
> > I have a question concerning interrupts processing -
> > as I see interrupt processing for MPC8xx is concentrated in
> > 'irq' subdirectory of bsp directory (e.g. lib/libbsp/powerpc/mbx8xx/irq)
> > I think this code could be shared between differnet MPC8xx bsps (eth_comm
> > bsp use the same interrup processing code as mbx8xx bsp) and placed into
> > libcpu/powerpc/m8xx/ subdirectory.
> >
> > Some usefull thing could be added to this interrupt dispatching code:
> > It seems that interrupt handler routine should have input parameter -
> > void interrupt_handler(void *arg),
>
>
> This has already been discussed several time on the list. I'm not in
> favor of that because the actual API
> 	1) enables to do what you want using different trampoline handler,
> 	2) Increase the number of instruction in the generic code even if you do
> not need an argument,
>
> if you connect two handlers :
>
> trampoline_irq1 and trampoline_irq2 and do
>
> trampoline_irq1() { generic_handler(table_args[irq1]);}
> trampoline_irq2() { generic_handler(table_args[irq2]);}
>
> You have exactly what you want just the code is inside the application
> and not inside the generic irq processing and thus do not cause extra
> processing for irq that do not need multiplexing.
>
> Otherwyse, the code is easy to modify to do what you want. NB : I even
> submitted a patch for ix86 to show that it was easy to implement...
>
>