[PATCH] ARM: imx: do not bring up unavailable cores

[Will Deacon]

On Tue, Apr 02, 2013 at 03:32:26PM +0100, Shawn Guo wrote:
> The i.MX6 Quad can be fused as i.MX6 Dual chip, and similarly i.MX6
> DualLite can be fused as i.MX6 Solo.  The actual number of available
> cores can be found out from SCU.
> 
> Since we do not reflect the fusing thing in device tree, the function
> arm_dt_init_cpu_maps() will always call set_cpu_possible(true) for 4
> cores on i.MX6 Quad/Dual and 2 cores for i.MX6 DualLite/Solo.  This
> causes failures when kernel tries to bring those unavailable cores
> online.  For example, the following failure message will be seen when
> booting an i.MX6 Solo chip.
> 
>   CPU1: failed to come online
> 
> Though kernel will still boot fine, the message is somehow annoying.
> Let's get rid of it by calling set_cpu_possible(false) on those
> unavailable cores.
> 
> While at it, the set_cpu_possible(true) for available cores is removed,
> since it's already been done in arm_dt_init_cpu_maps().
> 
> Signed-off-by: Shawn Guo <shawn.guo at linaro.org>
> ---
>  arch/arm/mach-imx/platsmp.c |    4 ++--
>  1 file changed, 2 insertions(+), 2 deletions(-)
> 
> diff --git a/arch/arm/mach-imx/platsmp.c b/arch/arm/mach-imx/platsmp.c
> index 7c0b03f..7061bde 100644
> --- a/arch/arm/mach-imx/platsmp.c
> +++ b/arch/arm/mach-imx/platsmp.c
> @@ -79,8 +79,8 @@ static void __init imx_smp_init_cpus(void)
>  
>  	ncores = scu_get_core_count(scu_base);
>  
> -	for (i = 0; i < ncores; i++)
> -		set_cpu_possible(i, true);
> +	for (i = ncores; i < NR_CPUS; i++)
> +		set_cpu_possible(i, false);
>  }

So you're passing the wrong DT to the kernel and then getting annoyed about
some warnings in the dmesg? Assuming the bootloader knows how many CPUs
there are, shouldn't that be sorting out the CPU nodes before passing the
dtb to the kernel?

Will