[PATCH v2 0/4] ARM: OMAP: boards: changes to support dynamic irq alloc

[b-cousson@ti.com (Cousson, Benoit)]

On 3/20/2012 12:39 AM, Tony Lindgren wrote:
> * Cousson, Benoit<b-cousson@ti.com>  [120319 16:00]:
>> Hi Tony,
>>
>> On 3/19/2012 8:17 PM, Tony Lindgren wrote:
>>> * Tarun Kanti DebBarma<tarun.kanti@ti.com>   [120319 05:09]:
>>>> These two patches incorporate changes to OMAP1 and OMAP2 platforms
>>>> board files whereby older references to OMAP_GPIO_IRQ macro are
>>>> now replaced with gpio_to_irq(), thereby getting rid of static
>>>> irq references.
>>>>
>>>> Reference: git://git.kernel.org/pub/scm/linux/kernel/git/arm/arm-soc.git omap/dt
>>>> Commit: 9a0cee711448335ec43eae83272495e9334c0098
>>>
>>> Can you please tell the exact two commits causing this
>>> breakage?
>>
>> Well, this is the GPIO DT + SPARSE_IRQ series I have done. It
>> appears that the boards I have were already using properly
>> gpio_to_irq() and thus were working fine with this series.
>>
>> But this is unfortunately not the case of most OMAP2 and 3 legacy
>> boards that were still using an old OMAP way of converting GPIO to
>> IRQ and were never modified to take advantage of the gpiolib stuff.
>>
>> So if these patches are apply before the GPIO DT + SPARSE_IRQ
>> series, there will be no breakage at all.
>>
>> All the cleanup we have never done before will hurt us at some point
>> when we will start using more extensively newer fmwk (DT,
>> sparse_irq, dmaengine...). It was not done on purpose, but this GPIO
>> series highlighted this remaining static broken mapping inside OMAP
>> boards.
>
> Yes I understand. But still, which patch(s) cause the issue
> so we can put that in the changelog for the fixes?

OK, here they are:
25db711 gpio/omap: Fix IRQ handling for SPARSE_IRQ
384ebe1 gpio/omap: Add DT support to GPIO driver

Tarun,

You should indeed add the references in your cover letter.

>>> I'm baffled how despite all the effort for previnting
>>> issues like this this still happen. These all seem valid
>>> fixes and clean up things, but how come this was not seen
>>> earlier?
>>
>> Maybe because there are much more boards inside mach-omap2 directory
>> than inside my cubicle... :-(
>
> Well somehow we need to make sure that patches get properly
> tested on a reasonable selection of boards. This pretty much
> breaks things for 21 boards out of the 51 board-*.c files :(

What is too bad is that one broken board was enough to figured out the 
issue and fix all the other ones. I just did not have that one :-(

Regards,
Benoit