[PATCH] bsps/arm: fix off-by-1 in gicv3 processor count

[Gedare Bloom]

Hi Sebastian,


On Wed, Aug 4, 2021 at 11:15 AM Gedare Bloom <gedare at rtems.org> wrote:
>
> ---
>  bsps/shared/dev/irq/arm-gicv3.c | 4 ++--
>  1 file changed, 2 insertions(+), 2 deletions(-)
>
> diff --git a/bsps/shared/dev/irq/arm-gicv3.c b/bsps/shared/dev/irq/arm-gicv3.c
> index ea123d325e..95021f6ddf 100644
> --- a/bsps/shared/dev/irq/arm-gicv3.c
> +++ b/bsps/shared/dev/irq/arm-gicv3.c
> @@ -551,11 +551,11 @@ uint32_t arm_gic_irq_processor_count(void)
>      for (i = 0; i < CPU_MAXIMUM_PROCESSORS; ++i) {
>        volatile gic_redist *redist = gicv3_get_redist(i);
>
> +      ++cpu_count;
> +
>        if ((redist->icrtyper & GIC_REDIST_ICRTYPER_LAST) != 0) {
>          break;
>        }
> -
> -      ++cpu_count;
>      }
Can you check my thinking here? I'm trying to implement SMP support
for the versal on aarch64. The call to arm_gic_irq_processor_count()
returns 1 through this code path. Since cpu_count is initialized to 0,
I think we need to increment it each time we find an active processor,
until we find the last active processor, which we also need to include
in the count, because the cpu_count should be NUM_PROC + 1?

After I make this change, I encounter other problems with the
secondary processor initialization, but at least it tries to
initialize/wait for core#1 then.

Thanks,
Gedare

>    }
>
> --
> 2.25.1
>