From mboxrd@z Thu Jan  1 00:00:00 1970
From: benh@kernel.crashing.org (Benjamin Herrenschmidt)
Date: Sat, 12 Jun 2010 15:12:53 +1000
Subject: [RFC,PATCH 1/2] Add a common struct clk
In-Reply-To: <19474.172.742782.972629@ipc1.ka-ro>
References: <1275636608.606606.450179637764.0.gpush@pororo>
	<201006111557.12249.jeremy.kerr@canonical.com>
	<19473.61547.684572.647641@ipc1.ka-ro>
	<201006111718.47426.jeremy.kerr@canonical.com>
	<19474.172.742782.972629@ipc1.ka-ro>
Message-ID: <1276319573.1962.180.camel@pasglop>
To: linux-arm-kernel@lists.infradead.org
List-Id: linux-arm-kernel.lists.infradead.org

On Fri, 2010-06-11 at 11:23 +0200, Lothar Wa?mann wrote:
> Hi,
> 
> > > Using a mutex in clk_enable()/clk_disable() is a bad idea, since that
> > > makes it impossible to call those functions in interrupt context.
> > 
> > Do we do this at the moment? I know at least one implementation of clk_enable 
> > uses a mutex for locking.
> > 
> You are probably talking about the Freescale i.MX51 kernel, that won't
> even boot, if you enable CONFIG_DEBUG_KERNEL, CONFIG_DEBUG_SPINLOCK,
> CONFIG_DEBUG_LOCKDEP and CONFIG_DEBUG_SPINLOCK_SLEEP.
> The mutex in the clock implementation is one of the reasons.

Regardless. Clocks generally take time to enable. I don't believe doing
clock enable/disable at hard irq context is a great idea.

If you really want to do something like that, you can always use either
threaded interrupts, or if you know your clock is off, mask & defer your
handling to a work queue yourself.

Unless we have enough case of very fast switching clocks that would
really benefit for that but from my experience, when a device can issue
interrupts, it should have its clocks on, unless it's some kind of
"wakeup" interrupt in which case it can safely be delayed.

Cheers,
Ben.