undefined IRQs in request_irq

undefined IRQs in request_irq

[Mark Chambers]

I wrote a driver for our custom MPC860 platform that we've ported linux 2.4.19
to.  When I insmod'd the driver it produced a kernel panic.  I traced it to
request_irq() in ppc8xx_pic.c, where the code calls panic() if the requested
irq is anything other than IDEx_INTERRUPT.  I fixed my problem by just adding
another case.

First, it seems extreme to me to crash the system instead of just refusing the
request.  Am I missing something?

Second, a suggested improvement:  How about having the individual board.h
files include #defines like ENABLE_SIU_LEVEL1, etc. that could be used like
the 'well known' IDEx_INTERRUPTs in request_irq()?

Mark Chambers
Microfirst, Inc.
mchambers@microfirst.com

** Sent via the linuxppc-dev mail list. See http://lists.linuxppc.org/

Re: undefined IRQs in request_irq

[Paul Mackerras]

Mark Chambers writes:

> I wrote a driver for our custom MPC860 platform that we've ported linux 2.4.19
> to.  When I insmod'd the driver it produced a kernel panic.  I traced it to
> request_irq() in ppc8xx_pic.c, where the code calls panic() if the requested
> irq is anything other than IDEx_INTERRUPT.  I fixed my problem by just adding
> another case.
>
> First, it seems extreme to me to crash the system instead of just refusing the
> request.  Am I missing something?

No.  IMO the current request_8xxirq / request_irq stuff is completely
bogus.  I can understand it as a temporary construction while
debugging stuff, but it has long outlived its usefulness.  We should
just use request_irq for everything, request_8xxirq should die.

Paul.

** Sent via the linuxppc-dev mail list. See http://lists.linuxppc.org/

RE: undefined IRQs in request_irq

[Allen Curtis]

> No.  IMO the current request_8xxirq / request_irq stuff is completely
> bogus.  I can understand it as a temporary construction while
> debugging stuff, but it has long outlived its usefulness.  We should
> just use request_irq for everything, request_8xxirq should die.

Multiple people have submitted patches to fix this but they have not made it
into the release code. What can be done?


** Sent via the linuxppc-dev mail list. See http://lists.linuxppc.org/