[PATCH v2 0/2] OMAP3+/4+ PRM: fix of_irq_get() result checks

[Sergei Shtylyov]

On 08/06/2017 09:57 PM, Andy Shevchenko wrote:

>> Here's the set of 2 patches against the 'omap-for-v4.13/fixes' branch of Tony
>> Lingren's 'linux-omap.git' repo.  I'm  fixing of_irq_get() DT API result
>> checks.
>>
>> [1/2] ARM: OMAP3+: PRM: fix of_irq_get() result check
>> [2/2] ARM: OMAP4+: PRM: fix of_irq_get() result checks
> 
> What IRQ domain did you consider? In case of STRICT the mapping will
> be 1:1 and 0 is _valid_ IRQ number for hwirq.

    of_irq_get() returns virtual IRQ #'s, no? And for those 0 is considered 
non-valid by Linus (I personally don't agree with that).

> 0 in Linux kernel IRQ numbering often means NO_IRQ.

    NO_IRQ was "considered harmful" AFAIK. :-)

> But there might be exceptions.

> So, it would be nice to see if there is a problem on certain platform
> if the patch is not applied, or is it problem on OF IRQ code, or
> wrong interpratation of 0?

    of_irq_get() is certainly problematic with its returning both 0 and 
negative error # on failure. I tried to fix that back in 2016 but that patch 
got shot down:

http://marc.info/?t=146447243300002&r=1

    I've now audited all the call sites of of_irq_get() and that resulted in a 
number of fixes.

MBR, Sergei