Re: [PATCH] ppc64: Fix cpu_up race

[Joel Schopp]

I've tested this patch extensively and it does fix the race.

In more detail the race is that the cpu can be running before it is 
marked in the cpu_online_map.  While running it can go idle and the idle 
loop checks cpu_online_map, sees that it is not set, and assumes that a 
hotplug remove is in progress and that it should stop itself.  By the 
time the cpu is marked online it is stopped.

Srivatsa Vaddagiri wrote:

> On Thu, Jul 29, 2004 at 03:30:11PM +1000, Anton Blanchard wrote:
>
> > > I think Anton pushed this to ameslab. I presume there should be some mechanism
> > > for ameslab stuff to be included mainline. If there is such mechanism, then
> > > I dont want to push the patch myself.
> >
> > It looks good, could you send it directly to akpm, ccing me?
>
>
> Andrew,
> 	Patch below fixes a cpu_up race in PPC64. Please apply.
>
>
> Signed-off-by : Srivatsa Vaddagiri <vatsa@xxxxxxxxxx>
> Signed-off-by : Anton Blanchard <anton@xxxxxxxxx>
>
>
> ---
>
>  linux-2.6.8-rc2-vatsa/arch/ppc64/kernel/smp.c |   10 +++++++++-
>  1 files changed, 9 insertions(+), 1 deletion(-)
>
> diff -puN arch/ppc64/kernel/smp.c~ppc64_cpu_up_fix arch/ppc64/kernel/smp.c
> --- linux-2.6.8-rc2/arch/ppc64/kernel/smp.c~ppc64_cpu_up_fix	2004-07-30 10:33:55.000000000 +0530
> +++ linux-2.6.8-rc2-vatsa/arch/ppc64/kernel/smp.c	2004-07-30 10:34:02.000000000 +0530
> @@ -927,7 +927,11 @@ int __devinit __cpu_up(unsigned int cpu)
>  
>  	if (smp_ops->give_timebase)
>  		smp_ops->give_timebase();
> -	cpu_set(cpu, cpu_online_map);
> +
> +	/* Wait until cpu puts itself in the online map */
> +	while (!cpu_online(cpu))
> +		cpu_relax();
> +
>  	return 0;
>  }
>  
> @@ -963,6 +967,10 @@ int __devinit start_secondary(void *unus
>  #endif
>  #endif
>  
> +	spin_lock(&call_lock);
> +	cpu_set(cpu, cpu_online_map);
> +	spin_unlock(&call_lock);
> +
>  	local_irq_enable();
>  
>  	return cpu_idle(NULL);
>
> _
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/