Re: [PATCH] sched: Have BUG_ON() check if linker sched classes don't line up correctly

From: Peter Zijlstra
Date: Fri Jun 19 2020 - 18:39:20 EST


On Fri, Jun 19, 2020 at 06:02:19PM -0400, Steven Rostedt wrote:
> From: Steven Rostedt (VMware) <rostedt@xxxxxxxxxxx>
>
> If the sched_class structures do not match how the compiler thinks they
> are in an array, it can cause hard to debug bugs. Change the BUG_ON()
> from making sure each of the sched classes are in the proper order, to
> also making sure they are off by the proper amount.
>
> Signed-off-by: Steven Rostedt (VMware) <rostedt@xxxxxxxxxxx>
> ---
> diff --git a/kernel/sched/core.c b/kernel/sched/core.c
> index 7ad864dc3ac5..876d7ecdab52 100644
> --- a/kernel/sched/core.c
> +++ b/kernel/sched/core.c
> @@ -6635,12 +6635,13 @@ void __init sched_init(void)
> unsigned long ptr = 0;
> int i;
>
> - BUG_ON(&idle_sched_class > &fair_sched_class ||
> - &fair_sched_class > &rt_sched_class ||
> - &rt_sched_class > &dl_sched_class);
> + /* Make sure the linker didn't screw up */
> + BUG_ON(&idle_sched_class + 1 != &fair_sched_class ||
> + &fair_sched_class + 1 != &rt_sched_class ||
> + &rt_sched_class + 1 != &dl_sched_class);
>
> #ifdef CONFIG_SMP
> - BUG_ON(&dl_sched_class > &stop_sched_class);
> + BUG_ON(&dl_sched_class + 1 != &stop_sched_class);
> #endif

Nice, I'll fold that.