Re: __do_IRQ() going away

[Manuel Lauss <mano@roarinelk.homelinux.net>]

On Thu, Mar 12, 2009 at 10:28 AM, Ralf Baechle <ralf@linux-mips.org> wrote:
> On Thu, Mar 12, 2009 at 08:26:18AM +0100, Manuel Lauss wrote:
>
>>
>> On Wed, Mar 11, 2009 at 12:28:06PM +0100, Ralf Baechle wrote:
>> > __do_IRQ() is deprecated since a long time and there are plans to remove
>> > it for 2.6.30.  The MIPS platforms seem to fall into three classes:
>>
>> >  o Platforms that still seem to rely on __do_IRQ():
>> >      o All Alchemy platforms:
>> >     db1000_defconfig, db1100_defconfig, db1200_defconfig, db1500_defconfig,
>> >     db1550_defconfig, mtx1_defconfig, pb1100_defconfig, pb1500_defconfig
>> >     and pb1550_defconfig
>>
>> I believe that the defconfigs just need to be updated.  There are no
>> __do_IRQ invocations in the alchemy/ tree anymore, and generic hardirqs are
>> enabled by CONFIG_SOC_AU1X00.
>
> __do_IRQ will be called from the generic code if irq_desc->handle_irq is
> not set for an interrupt and handle_irq will be left NULL if a platform
> only calls set_irq_chip or even does a homebrew initialization.  Fix is
> to call set_irq_chip_and_handler or better set_irq_chip_and_handler_name.

Alchemy does all that...


> Iow, now with CONFIG_GENERIC_HARDIRQS_NO__DO_IRQ always set half the
> platforms will blow up because the function pointer irq_desc->handle_irq
> is unset.

...and it works fine so far on the DB1200 and another 2 boards I have.
(I.e. your patch didn't break anything).  Unless I'm missing something
very big.

Manuel Lauss