Re: [PATCH PREEMPT_RT]: On AT91 ARM: GPIO Interrupt handling can/will stall forever - Russell King

From: Russell King - ARM Linux <linux@arm.linux.org.uk>
To: Daniel Walker <dwalker@mvista.com>
Cc: Steven Rostedt <rostedt@goodmis.org>,
	Remy Bohmer <linux@bohmer.net>,
	RT <linux-rt-users@vger.kernel.org>,
	linux-kernel <linux-kernel@vger.kernel.org>,
	ARM Linux Mailing List  <linux-arm-kernel@lists.arm.linux.org.uk>,
	Thomas Gleixner <tglx@linutronix.de>, Ingo Molnar <mingo@elte.hu>
Subject: Re: [PATCH PREEMPT_RT]: On AT91 ARM: GPIO Interrupt handling can/will stall forever
Date: Thu, 29 Nov 2007 09:04:00 +0000	[thread overview]
Message-ID: <20071129090359.GA14728@flint.arm.linux.org.uk> (raw)
In-Reply-To: <1196291955.27964.47.camel@imap.mvista.com>

On Wed, Nov 28, 2007 at 03:19:14PM -0800, Daniel Walker wrote:
> On Wed, 2007-11-28 at 23:03 +0000, Russell King - ARM Linux wrote:
> > On Wed, Nov 28, 2007 at 04:13:07PM -0500, Steven Rostedt wrote:
> > > 
> > > 
> > > On Wed, 28 Nov 2007, Daniel Walker wrote:
> > > 
> > > >
> > > > Ignoring the ARM side of things for a sec, handle_simple_irq() will
> > > > mask() the interrupt in the special case that an interrupt is already in
> > > > the processes of being handled.. handle_simple_irq() also unmasks when
> > > > it finishes handling an interrupt (something real time adds for some
> > > > reason) ..
> > > >
> > > > In terms of threading the irq everything is the same except there is no
> > > > unmask() call when the thread finishes ..
> > > >
> > > 
> > > OK, to be honest, I never fully understood the concept of this
> > > "simple_irq". I figured it was because of the ARM architecture.
> > 
> > If you read what I said compared with what Daniel said, you'll see that
> > adding the mask/unmask is _pointless_ because for the case where the
> > simple handler should be used, there is _no_ hardware masking available
> > except via the parent interrupt signal.
> > 
> > So actually Daniel's argument misses the basic point - that using
> > handle_simple_irq for non-simple IRQs is just WRONG.
> 
> Well we've got at least two ARM boards which need the additional unmask
> to work correctly with interrupt threading .. So there must be at least
> two ARM boards using handle_simple_irq incorrectly .. It sounds like you
> would prefer we send patches to move those handle_simple_irq users into
> another method ?

If you read my explaination of the purpose of handle_simple_irq() you'll
see that for the _correct_ case where this handler should be used, the
mask and unmask can only be empty.  To repeat for the third time - it's
for the case where the hardware supplies NO way to mask the interrupt.

So adding calls to the mask/unmask methods would be pointless for the
correct use case.

If you do have the ability to mask, then you should be using the level or
edge handlers.