From mboxrd@z Thu Jan  1 00:00:00 1970
From: tglx@linutronix.de (Thomas Gleixner)
Date: Sun, 2 Aug 2015 11:10:21 +0200 (CEST)
Subject: [PATCH 1/3] clocksource: atmel-st: Remove irq handler when clock
 event is unused
In-Reply-To: <alpine.DEB.2.11.1507202245310.18576@nanos>
References: <1437161608-26782-1-git-send-email-alexandre.belloni@free-electrons.com>
 <1437161608-26782-2-git-send-email-alexandre.belloni@free-electrons.com>
 <alpine.DEB.2.11.1507180942360.18576@nanos>
 <20150718222308.GA30489@piout.net>
 <alpine.DEB.2.11.1507201103550.18576@nanos> <20150720193726.GJ15539@piout.net>
 <alpine.DEB.2.11.1507202245310.18576@nanos>
Message-ID: <alpine.DEB.2.11.1508021100390.3825@nanos>
To: linux-arm-kernel@lists.infradead.org
List-Id: linux-arm-kernel.lists.infradead.org

On Mon, 20 Jul 2015, Thomas Gleixner wrote:

> On Mon, 20 Jul 2015, Alexandre Belloni wrote:
> 
> > On 20/07/2015 at 11:04:30 +0200, Thomas Gleixner wrote :
> > > That's why I used setup/remove_irq() in the stale RT changes.
> > > 
> > 
> > Hum, actually, you hit the same thing with setup/remove_irq (and I'm not sure
> > to follow why):
> 
> It's the kmalloc of the proc interface. So back to the drawing
> board. I'll think about it tomorrow with brain awake.

So the only idea I could come up with is the old approach of creating
a dummy irq chip for the interrupts which share that timer/uart/rtc
whatever line.

I think Boris Brezillon had implemented it at some point, but it was
shot down for reasons I can't remember. That would allow us to simply
have disable_irq_nosync/enable_irq in the relevant code pathes and be
done with it. I guess I need to find the references and give it a try
again if nobody beats me to it.

Thanks,

	tglx