Re: [RFC] [PATCH 02/62] mpu401:snd_mpu401_uart_new(): split semantic of irq_flags

[Clemens Ladisch <clemens@ladisch.de>]

Takashi Iwai wrote:
> Clemens Ladisch wrote:
> > ALSA: mpu401: clean up interrupt specification
> > 
> > The semantics of snd_mpu401_uart_new()'s interrupt parameters are
> > somewhat counterintuitive:  To prevent the function from allocating its
> > own interrupt, either the irq number must be invalid, or the irq_flags
> > parameter must be zero.  At the same time, the irq parameter being
> > invalid specifies that the mpu401 code has to work without an interrupt
> > allocated by the caller.  This implies that, if there is an interrupt
> > and it is allocated by the caller, the irq parameter must be set to
> > a valid-looking number which then isn't actually used.
> > 
> > With the removal of IRQF_DISABLED, zero becomes a valid irq_flags value,
> > which forces us to handle the parameters differently.
> > 
> > This patch introduces a new flag MPU401_INFO_IRQ_HOOK for when the
> > device interrupt is handled by the caller, and makes the allocation of
> > the interrupt to depend only on the irq parameter.
> > 
> > Signed-off-by: Clemens Ladisch <clemens@ladisch.de>
> 
> This patch looks better indeed.
> 
> However, if we are going to remove IRQF_DISABLED in near future, and
> all the callers use IRQF_DISABLED when requesting an irq, why not
> remove this argument now?  Then the IRQF_DISABLED-removal patch will
> touch less code.

I wanted to change only the minimum amount of code needed to make the
IRQF_DISABLED-removal patch possible, and wait with further cleanups of
snd_mpu401_uart_new() until after that patch to avoid conflicts.

However, moving the IRQF_DISABLED into snd_mpu401_uart_new() avoids
those conflicts, so it's clearly better.


Regards,
Clemens