Re: [PATCH bpf-next] selftests/bpf: fix off-by-one in bpf_cpumask_populate related selftest

[Paul Chaignon <paul.chaignon@gmail.com>]

On Mon, Apr 20, 2026 at 09:37:34AM +0000, Matt Bobrowski wrote:
> The test_populate test uses >= instead of > when checking if the
> runtime nr_cpus exceeds the bit capacity of a cpumask_t.
> 
> On a system where the physical CPU core count perfectly matches the
> CONFIG_NR_CPUS upper bound (e.g. nr_cpus = 512 and CONFIG_NR_CPUS =
> 512), the condition nr_cpus >= CPUMASK_TEST_MASKLEN * 8 evaluates to
> true (512 >= 512). This incorrectly causes the test to fail with an
> error value of 3.
> 
> A 512-bit cpumask_t provides enough bits (indices 0 through 511) to
> represent 512 CPUs. The subsequent bpf_for(i, 0, nr_cpus) loop
> iterates up to nr_cpus - 1 (511), which perfectly aligns with the
> maximum valid index of the bitmask.
> 
> Change the condition to nr_cpus > CPUMASK_TEST_MASKLEN * 8 to fix the
> false positive failure on these systems.
> 
> Fixes: 918ba2636d4e ("selftests: bpf: add bpf_cpumask_populate selftests")
> Signed-off-by: Matt Bobrowski <mattbobrowski@google.com>

Acked-by: Paul Chaignon <paul.chaignon@gmail.com>

> ---
>  tools/testing/selftests/bpf/progs/cpumask_success.c | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
> 
> diff --git a/tools/testing/selftests/bpf/progs/cpumask_success.c b/tools/testing/selftests/bpf/progs/cpumask_success.c
> index 0e04c31b91c0..774706e7b058 100644
> --- a/tools/testing/selftests/bpf/progs/cpumask_success.c
> +++ b/tools/testing/selftests/bpf/progs/cpumask_success.c
> @@ -866,7 +866,7 @@ int BPF_PROG(test_populate, struct task_struct *task, u64 clone_flags)
>  	 * access NR_CPUS, the upper bound for nr_cpus, so we infer
>  	 * it from the size of cpumask_t.
>  	 */
> -	if (nr_cpus < 0 || nr_cpus >= CPUMASK_TEST_MASKLEN * 8) {
> +	if (nr_cpus < 0 || nr_cpus > CPUMASK_TEST_MASKLEN * 8) {
>  		err = 3;
>  		goto out;
>  	}
> -- 
> 2.54.0.rc1.513.gad8abe7a5a-goog
> 
>