From mboxrd@z Thu Jan  1 00:00:00 1970
From: alexander.stein@systec-electronic.com (Alexander Stein)
Date: Wed, 29 Jun 2016 10:30:55 +0200
Subject: [PATCH v2 2/2] ARM: at91: pm: switch to the PIE infrastructure
In-Reply-To: <20160629075714.GP29249@piout.net>
References: <1467153886-21144-1-git-send-email-alexandre.belloni@free-electrons.com>
 <2349417.jWuiI2GXql@ws-stein> <20160629075714.GP29249@piout.net>
Message-ID: <1503370.EB259eFnzi@ws-stein>
To: linux-arm-kernel@lists.infradead.org
List-Id: linux-arm-kernel.lists.infradead.org

On Wednesday 29 June 2016 09:57:14, Alexandre Belloni wrote:
> On 29/06/2016 at 08:12:21 +0200, Alexander Stein wrote :
> > > +#if defined(CONFIG_CPU_V7)
> > > +	dsb();
> > > +	wfi();
> > > +#else
> > > +	asm volatile ("mcr	p15, 0, %0, c7, c0, 4" \
> > > +		      : : "r" (0) : "memory");
> > > +#endif
> > 
> > Why not defining wfi() for __LINUX_ARM_ARCH__ < 7 as it is done for dsb()
> > and friends in arch/arm/include/asm/barrier.h? So you can get rid of that
> > #if completly.
> 
> Well, Russell said it was not useful because "there's no architected WFI
> instruction which doesn't have CPU specific issues (hence why we have
> cpu_do_idle() to abstract that)"

Wouldn't you suffer from this here as well? Anyway, the inline assembly above 
is what is actually wfi but using CP15. I checked several proc-*.S in do_idle 
and they all have
> "mcr     p15, 0, r0, c7, c0, 4           @ Wait for interrupt"
at some point. so this inline assembly could actually be implemented on wfi() 
macro.

Best regards,
Alexander