[PATCH] ARM: OMAP2+: irq: Increase no of supported interrupts to 128

[Kevin Hilman]

"Hiremath, Vaibhav" <hvaibhav at ti.com> writes:

> On Wed, May 09, 2012 at 00:09:34, Hilman, Kevin wrote:
>> Vaibhav Hiremath <hvaibhav at ti.com> writes:
>> 
>> > With addition to TI81XX, AM33XX family of devices, the number
>> > of interrupts supported has increased to 128, compared to 96.
>> > The current implementation for irq handling is hardcoded to use
>> > 96 interrupts (with 3 register-sets to handle), this patch cleanups
>> > the code, to increase maximum number of interrupts support
>> > to 128, with dynamic detection of no of registers required for
>> > handling all interrupts.
>> >
>> >
>> > Signed-off-by: Vaibhav Hiremath <hvaibhav at ti.com>
>> > Signed-off-by: Afzal Mohammed <afzal at ti.com>
>> > Cc: Tony Lindgren <tony at atomide.com>
>> > Cc: Kevin Hilman <khilman at ti.com>
>> > Cc: Paul Walmsley <paul at pwsan.com>
>> > ---
>> > Ideally, we should use dynamic allocation to allocate memory
>> > for registers/arrays, 
>> 
>> Yes.
>> 
>
> Thanks Kevin, I will put this activity in my TODO list.
>
>> > may be too much cleanup for this patch,
>> 
>> There is no such thing as too much cleanup.  ;)  
>> And the INTC is in need of it, IMO.
>> 
>
> Indeed it is in need of cleanup...
>
>
>> > so as of now restricting to minimal changes to fit devices
>> > like, am33xx/ti81xx.
>> 
>> Then someone else will have to do the cleanup later.  It would be
>> greatly appreciated if you could do the necessary cleanup in order to
>> cleanly add support for more SoCs.  Yes, we probably should've insisted
>> when support for TI81xx was added, but that one slipped in under the
>> radar.
>> 
>
> Yeah, I understand. As I said I will put this activity in my TODO list.
>
>> For starters, the notion of a banks this code is a rather messed up and
>> needs a cleanup.  A bank is supposed to be a group of 32 interrupts, 
>> and the INTC is made up of 3 (or 4) banks.   However, the current
>> code creates a single "bank" of 96 (or 128) interrupts.  
>> 
>> It also confuses what registers are part of the bank and what are global
>> to the INTC.  This confusion is both in init and in context save/restore.
>> 
>> IMO, to clean this up, first the notion of banks needs to be fixed in
>> that code there is a distinction between what acts on banks and what
>> works on the whole INTC.
>> 
>> Then, the init/alloc should be done dynamically based on the number of
>> interrupts passed to omap_init_irq()
>> 
>
> Kevin,
> Let me finish up with am33xx baseport upstream activity which is currently 
> going on at full swing, then next thing I will pick up is this code cleanup.
>
> I still feel, this is still a valid cleanup patch, and can be merged, as it 
> is required/used when we do major cleanup.

Well, Tony can make that decision.

Personally, I think this patch continues to add confusion on top of an
existing mess, and to me provides the proverbial straw that broke the
camel's back.

That being said, the INTC core is an obviously important and sensitive
piece of code so needs to be handled with care.

In case Tony decides to merge this patch and allow a future** cleanup,
I'll provide some comments.

Kevin

** since they rarely happen, we tend to not have too much faith in
   promises of mythical "future" cleanups.  This is not because we don't
   trust you personally, but simply based on based experience.