Re: [PATCH] ASoC: Intel: atom: Remove redundant check to simplify the code

[Andy Shevchenko]

On Mon, Nov 29, 2021 at 09:01:16PM +0200, Andy Shevchenko wrote:
> On Mon, Nov 29, 2021 at 05:11:52PM +0000, Mark Brown wrote:
> > On Mon, Nov 29, 2021 at 10:22:41AM -0600, Pierre-Louis Bossart wrote:
> > > On 11/25/21 1:50 AM, Tang Bin wrote:
> > 
> > > > In the function sst_platform_get_resources(), if platform_get_irq()
> > > > failed, the return should not be zero, as the example in
> > > > platform.c is
> > > >   * int irq = platform_get_irq(pdev, 0)
> > > >   * if (irq < 0)
> > > >   * return irq;
> > > > So remove the redundant check to simplify the code.
> > 
> > > Humm, it's a bit of a gray area.
> > 
> > > the comments for platform_get_irq and platform_get_irq_optional say:
> > 
> > > * Return: non-zero IRQ number on success, negative error number on failure.
> > 
> > > but if you look at platform_get_irq_optional, there are two references
> > > to zero being a possible return value:
> > 
> > Zero is (or was, people were working on changing it partly due to
> > confusion and partly due to moving to newer infrastructure which
> > doesn't use it) a valid IRQ on some architectures.  x86 wasn't one of
> > those though, at least AFAIR.
> 
> I guess it's about x86, but the API returns Linux virtual IRQ and 0 shouldn't
> be among them (hardware IRQ != Linux virtual IRQ). Legacy x86 used 1:1 mapping
> for ISA IRQs (lower 16) among which the Timer IRQ is 0. I believe that timer
> code does not use any of those APIs (it most likely and IIRC has it hardcoded).
> 
> Nevertheless, I have planned to make platform_irq_get_optional() to be optional
> indeed, where we return 0 when there is no IRQ provided and error when it's a
> real error happens. This needs to clean up the current (mis-)use of the API.

Link for previous work: https://lore.kernel.org/lkml/20210331144526.19439-1-andriy.shevchenko@xxxxxxxxxxxxxxx/T/#u

-- 
With Best Regards,
Andy Shevchenko