Re: [PATCH] ARM: remove redundant irq disable at halt and restart

[Felipe Balbi]

On Fri, Oct 24, 2014 at 09:50:29PM +0200, Johan Hovold wrote:
> On Fri, Oct 24, 2014 at 02:42:50PM -0500, Felipe Balbi wrote:
> > On Fri, Oct 24, 2014 at 09:28:45PM +0200, Johan Hovold wrote:
> > > On Fri, Oct 24, 2014 at 02:21:11PM -0500, Felipe Balbi wrote:
> > > > On Fri, Oct 24, 2014 at 02:16:27PM -0500, Felipe Balbi wrote:
> > > > > On Fri, Oct 24, 2014 at 09:06:32PM +0200, Johan Hovold wrote:
> > > > > > Remove redundant local_irq_disable() at machine halt and restart.
> > > > > > 
> > > > > > Since commit 44424c34049f ("ARM: 7803/1: Fix deadlock scenario with
> > > > > > smp_send_stop()") interrupts are disabled before stopping secondary
> > > > > > CPUs.
> > > > > 
> > > > > Assuming this is correct, you should have:
> > > > > 
> > > > > Fixes: 44424c3 (ARM: 7803/1: Fix deadlock scenario with smp_send_stop())
> > > > > Cc: <stable@xxxxxxxxxxxxxxx> # v3.12+
> > > 
> > > It's not a bug. Just a redundant disabling of already disabled
> > > interrupts, something which could possibly lead someone to believe that
> > > interrupts could be re-enabled by the power-off handler.
> 
> I meant re-enabled by arm_pm_restart().
> 
> > I didn't dig any of this out but I'll assume you did :-) So I withdraw
> > my comment ;-)
> > 
> > > > > > diff --git a/arch/arm/kernel/process.c b/arch/arm/kernel/process.c
> > > > > > index a35f6ebbd2c2..5663ab57cf07 100644
> > > > > > --- a/arch/arm/kernel/process.c
> > > > > > +++ b/arch/arm/kernel/process.c
> > > > > > @@ -195,7 +195,6 @@ void machine_halt(void)
> > > > > >  	local_irq_disable();
> > > > > >  	smp_send_stop();
> > > > > >  
> > > > > > -	local_irq_disable();
> > > > > >  	while (1);
> > > > > >  }
> > > > > >  
> > > > > > @@ -237,7 +236,6 @@ void machine_restart(char *cmd)
> > > > > >  
> > > > > >  	/* Whoops - the platform was unable to reboot. Tell the user! */
> > > > > >  	printk("Reboot failed -- System halted\n");
> > > > > > -	local_irq_disable();
> > > > > 
> > > > > ... but wouldn't this reintroduce the the buck which that commit fixed ?
> > > > 
> > > > s/buck/bug :-) my fingers have a mind of their own, aparently.
> > > 
> > > :)
> > > 
> > > No, the interrupts would still be disabled.
> > 
> > alright... so far I couldn't find where IRQs are disable before
> > machine_power_off() is called. Starting a do_poweroff(), couldn't find
> > it... Oh well, I'll keep digging.
> 
> It's done a few lines above in the same function. ;)
> 
> void machine_restart(char *cmd)
> {
>         local_irq_disable();
> 	^^^
>         smp_send_stop();
> 
>         arm_pm_restart(reboot_mode, cmd);
> 
>         /* Give a grace period for failure to restart of 1s */
>         mdelay(1000);
> 
>         /* Whoops - the platform was unable to reboot. Tell the user! */
>         printk("Reboot failed -- System halted\n");
>         local_irq_disable();
>         while (1);
> }
> 
> [ and similarly in machine_power_off(). ]

oh, now I get it :-) Nevermind, I completely missed that it was
duplicated. My bad.

Reviewed-by: Felipe Balbi <balbi@xxxxxx>

-- 
balbi

Attachment: signature.asc
Description: Digital signature