[PATCH v4 3/5] irqchip: Add DT binding doc for the virtual irq demuxer chip

[Boris Brezillon]

On Wed, 11 Feb 2015 16:32:31 +0000
Mark Rutland <mark.rutland at arm.com> wrote:

> On Wed, Feb 11, 2015 at 04:15:15PM +0000, Boris Brezillon wrote:
> > On Wed, 11 Feb 2015 15:57:20 +0000
> > Mark Rutland <mark.rutland at arm.com> wrote:
> > 
> > > [...]
> > > 
> > > > > > > So for the flag at request time approach to work, all the drivers using
> > > > > > > the interrupt would have to flag they're safe in that context.
> > > > > > 
> > > > > > Something like IRQF_"I can share the line with a timer" I guess?  That wouldn't
> > > > > > hurt and can be checked at request time even.
> > > > > 
> > > > > I guess that would have to imply IRQF_SHARED, so we'd have something
> > > > > like:
> > > > > 
> > > > > IRQF_SHARED_SUSPEND_OK - This handler is safe to call spuriously during
> > > > > 			 suspend in the case the line is shared. The
> > > > > 			 handler will not access unavailable hardware
> > > > > 			 or kernel infrastructure during this period.
> > > > > 
> > > > > #define __IRQF_SUSPEND_SPURIOUS		0x00040000
> > > > > #define	IRQF_SHARED_SUSPEND_OK		(IRQF_SHARED | __IRQF_SUSPEND_SPURIOUS)
> > > > 
> > > > What about
> > > > 
> > > > #define __IRQF_TIMER_SIBLING_OK	0x00040000
> > > > #define	IRQF_SHARED_TIMER_OK	(IRQF_SHARED | __IRQF_TIMER_SIBLING_OK)
> > > > 
> > > > The "suspend" part is kind of a distraction to me here, because that really
> > > > only is about sharing an IRQ with a timer and the "your interrupt handler
> > > > may be called when the device is suspended" part is just a consequence of that.
> > > 
> > > My rationale was that you didn't really care who else was using the IRQ
> > > (e.g. the timer); you're just stating that you can survive being called
> > > during suspend (which is what the driver may need to check for in the
> > > handler if the device happens to be powered down or whatever).
> > > 
> > > So I guess I see it the other way around. This is essentially claiming
> > > we can handle sharing with IRQF_NO_SUSPEND rather than IRQF_TIMER.
> > > 
> > > > So IMO it's better to have "TIMER" in the names to avoid encouraging people to
> > > > abuse this for other purposes not related to timers.
> > > 
> > > In the end a name is a name, and if you think IRQF_SHARED_TIMER_OK is
> > > better I shan't complain.
> > > 
> > > The fundamental issue I'm concerned with is addressed by this approach.
> > 
> > Okay then, is anyone taking care of submitting such a patch (Mark ?) ?
> 
> I'll have the core patch shortly.
> 
> I'll need to ask for your help tagging the relevant drivers and testing.

For the list of impacted drivers, you can have a look at this series [1]
(patches 2 to 5), and I'll take care of the testing part once every one
has agreed on the solution ;-).

[1]https://lkml.org/lkml/2014/12/15/552

-- 
Boris Brezillon, Free Electrons
Embedded Linux and Kernel engineering
http://free-electrons.com