Re: [PATCH resend 2/2] [RFC] genirq: Set IRQCHIP_SKIP_SET_WAKE for no_irq_chip and dummy_irq_chip

[Geert Uytterhoeven]

Hi Santosh,

On Mon, Jan 12, 2015 at 5:37 PM, santosh shilimkar
<santosh.shilimkar@xxxxxxxxxx> wrote:
> On 1/12/2015 8:04 AM, Geert Uytterhoeven wrote:
>>
>> If no_irq_chip or dummy_irq_chip are used for wake up (e.g. gpio-keys
>> with a simple GPIO controller), the following warning is printed on
>> resume from s2ram:
>>
>>      WANING: CPU: 0 PID: 1046 at kernel/irq/manage.c:537
>> irq_set_irq_wake+0x9c/0xf8()
>>      Unbalanced IRQ 113 wake disable
>>
>> This happens because no_irq_chip and dummy_irq_chip do not implement
>> irq_chip.irq_set_wake(), causing set_irq_wake_real() to return -ENXIO,
>> and irq_set_irq_wake() to reset the wake_depth to zero.
>>
>> Set IRQCHIP_SKIP_SET_WAKE to indicate that irq_chip.irq_set_wake() is
>> not implemented.
>>
>> Signed-off-by: Geert Uytterhoeven <geert+renesas@xxxxxxxxx>
>> ---
>> Alternatively, can't we remove IRQCHIP_SKIP_SET_WAKE, and just check for
>> the presence of irq_chip.irq_set_wake()?
>> I'll be happy to send a patch to do that instead...
>>
>> Is there anything that relies on this -ENXIO error code?
>> All irq_chip implementations that set IRQCHIP_SKIP_SET_WAKE do not
>> implement irq_chip.irq_set_wake(). There are probably more of them that
>> forgot to set IRQCHIP_SKIP_SET_WAKE though.
>> Am I missing something?
>> Commit 60f96b41f71d2a13 ("genirq: Add IRQCHIP_SKIP_SET_WAKE flag")
>> doesn't explain why adding the flag was chosen.
>>
> The flag was added to avoid dummy irq_set_wake() implementation
> as described in the commit.
>
> ------------------
> commit 60f96b41f71d2a13d1c0a457b8b77958f77142d1
> Author: Santosh Shilimkar <santosh.shilimkar@xxxxxx>
> Date:   Fri Sep 9 13:59:35 2011 +0530
>
>     genirq: Add IRQCHIP_SKIP_SET_WAKE flag
>
>     Some irq chips need the irq_set_wake() functionality, but do not
>     require a irq_set_wake() callback. Instead of forcing an empty
>     callback to be implemented add a flag which notes this fact. Check for
>     the flag in set_irq_wake_real() and return success when set.
>
>     Signed-off-by: Santosh Shilimkar <santosh.shilimkar@xxxxxx>
>     Cc: Thomas Gleixner <tglx@xxxxxxxxxxxxx>
> ------------------

I had read that commit description.

> Here is the relevant thread.
> http://lists.infradead.org/pipermail/linux-arm-kernel/2011-September/064590.html
>
> As you can read from thread, the idea is to handle the need at
> genirq level. Either with a flag or a dummy function.

But it's not handled at genirq level. Every driver that doesn't implement the
.irq_set_wake() method has to set the flag. Several of these don't, causing
the warning.

Instead of having to fix them all, can't we remove IRQCHIP_SKIP_SET_WAKE,
and just check for the absence of irq_chip.irq_set_wake() instead?

Is there ever a valid use case for a driver to not provide a .irq_set_wake(),
and not set the flag?

Gr{oetje,eeting}s,

                        Geert

--
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert@xxxxxxxxxxxxxx

In personal conversations with technical people, I call myself a hacker. But
when I'm talking to journalists I just say "programmer" or something like that.
                                -- Linus Torvalds
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/