Re: [PATCH v2] selftests/bpf: Make x86 preempt_count access compatible across v6.14+

[Alexei Starovoitov]

On Thu, Jan 29, 2026 at 6:19 PM Changwoo Min <changwoo@xxxxxxxxxx> wrote:
>
> Recent x86 kernels export __preempt_count as a ksym, while some old kernels
> between v6.1 and v6.14 expose the preemption counter via
> pcpu_hot.preempt_count. The existing selftest helper unconditionally
> dereferenced __preempt_count, which breaks BPF program loading on such old
> kernels.
>
> Make the x86 preemption count lookup version-agnostic by:
> - Marking __preempt_count and pcpu_hot as weak ksyms.
> - Introducing a BTF-described pcpu_hot___local layout with
>   preserve_access_index.
> - Selecting the appropriate access path at runtime using ksym availability
>   and bpf_ksym_exists() and bpf_core_field_exists().
>
> This allows a single BPF binary to run correctly across kernel versions
> (e.g., v6.18 vs. v6.13) without relying on compile-time version checks.
>
> Signed-off-by: Changwoo Min <changwoo@xxxxxxxxxx>
> ---
>  .../testing/selftests/bpf/bpf_experimental.h  | 21 +++++++++++++++++--
>  1 file changed, 19 insertions(+), 2 deletions(-)
>
> diff --git a/tools/testing/selftests/bpf/bpf_experimental.h b/tools/testing/selftests/bpf/bpf_experimental.h
> index a39576c8ba04..acabf869d2c4 100644
> --- a/tools/testing/selftests/bpf/bpf_experimental.h
> +++ b/tools/testing/selftests/bpf/bpf_experimental.h
> @@ -614,7 +614,13 @@ extern int bpf_cgroup_read_xattr(struct cgroup *cgroup, const char *name__str,
>
>  extern bool CONFIG_PREEMPT_RT __kconfig __weak;
>  #ifdef bpf_target_x86
> -extern const int __preempt_count __ksym;
> +extern const int __preempt_count __ksym __weak;
> +
> +struct pcpu_hot___local {
> +       int preempt_count;
> +} __attribute__((preserve_access_index));
> +
> +extern struct pcpu_hot___local pcpu_hot __ksym __weak;
>  #endif
>
>  struct task_struct___preempt_rt {
> @@ -624,7 +630,18 @@ struct task_struct___preempt_rt {
>  static inline int get_preempt_count(void)
>  {
>  #if defined(bpf_target_x86)
> -       return *(int *) bpf_this_cpu_ptr(&__preempt_count);
> +       /* By default, read the per-CPU __preempt_count. */
> +       if (bpf_ksym_exists(&__preempt_count))
> +               return *(int *) bpf_this_cpu_ptr(&__preempt_count);
> +
> +       /* If __preempt_count does not exist, try to read preempt_count under

Please use proper kernel comment style in the future.
We don't use old networking style in bpf anymore.
I fixed it up while applying.

Also use [PATCH bpf-next] in the subject to help CI identify the tree.