On Sat, Jan 10, 2026 at 11:38 AM Menglong Dong [off-list ref] wrote:

On 2026/1/10 10:40, Alexei Starovoitov wrote:

quoted

On Wed, Jan 7, 2026 at 6:25 PM Menglong Dong [off-list ref] wrote:

quoted

+       } else if (func_id == special_kfunc_list[KF_bpf_session_is_return]) {
+               if (prog->expected_attach_type == BPF_TRACE_FSESSION)
+                       addr = (unsigned long)bpf_fsession_is_return;

...

quoted

+bool bpf_fsession_is_return(void *ctx)
+{
+       /* This helper call is inlined by verifier. */
+       return !!(((u64 *)ctx)[-1] & (1 << BPF_TRAMP_M_IS_RETURN));
+}
+

Why do this specialization and introduce a global function
that will never be called, since it will be inlined anyway?

Ah, the specialization and the definition of the global function
is not unnecessary. I thought that it's kinda fallback solution

      ^
typo: is not necessary

that we define the function even if it is inlined by the verifier.

quoted

Remove the first hunk and make the 2nd a comment instead of a real function?

Agree. So it will be:

+static bool bpf_fsession_is_return(void *ctx)
+{
+       /* This helper call is implemented and inlined by the verifier, and the logic is:
+         *   return !!(((u64 *)ctx)[-1] & (1 << BPF_TRAMP_M_IS_RETURN));
+         */
+        return false;
+}

quoted