Re: [PATCH 4/8] perf/arm-cmn: Improve build-time assertions

[Mark Rutland <mark.rutland@arm.com>]

On Fri, Aug 09, 2024 at 08:15:43PM +0100, Robin Murphy wrote:
> These days we can use static_assert() in the logical place rather than
> jamming a BUILD_BUG_ON() into the nearest function scope. And since it
> is now so convenient to do so, let's add an extra one to reinforce the
> dtm_idx bitmap as well.
> 
> Signed-off-by: Robin Murphy <robin.murphy@arm.com>
> ---
>  drivers/perf/arm-cmn.c | 4 +++-
>  1 file changed, 3 insertions(+), 1 deletion(-)
> 
> diff --git a/drivers/perf/arm-cmn.c b/drivers/perf/arm-cmn.c
> index c9a2b21a7aec..8f7a1a6f8ab7 100644
> --- a/drivers/perf/arm-cmn.c
> +++ b/drivers/perf/arm-cmn.c
> @@ -577,6 +577,9 @@ struct arm_cmn_hw_event {
>  	bool wide_sel;
>  	enum cmn_filter_select filter_sel;
>  };
> +static_assert(sizeof(struct arm_cmn_hw_event) <= offsetof(struct hw_perf_event, target));

Moving this out seems fine.

> +/* Currently XPs are the node type we can have most of; others top out at 128 */
> +static_assert(sizeof_field(struct arm_cmn_hw_event, dtm_idx) >= CMN_MAX_XPS / 4);

I reckon we can make this true by construction (as on the last patch),
and avoid the need for the assert entirely.

Mark.

>  
>  #define for_each_hw_dn(hw, dn, i) \
>  	for (i = 0, dn = hw->dn; i < hw->num_dns; i++, dn++)
> @@ -587,7 +590,6 @@ struct arm_cmn_hw_event {
>  
>  static struct arm_cmn_hw_event *to_cmn_hw(struct perf_event *event)
>  {
> -	BUILD_BUG_ON(sizeof(struct arm_cmn_hw_event) > offsetof(struct hw_perf_event, target));
>  	return (struct arm_cmn_hw_event *)&event->hw;
>  }
>  
> -- 
> 2.39.2.101.g768bb238c484.dirty
>