On 18/08/2025 10:36, David Hildenbrand wrote:

quoted

+
+TEST_F(prctl_thp_disable_completely, fork)
+{
+    int ret = 0;
+    pid_t pid;
+
+    /* Make sure prctl changes are carried across fork */
+    pid = fork();
+    ASSERT_GE(pid, 0);
+
+    if (!pid)
+        prctl_thp_disable_completely_test(_metadata, self->pmdsize, variant->thp_policy);
+

Skimming over this once more ... this raises two questions

(a) There is nothing to wait for in the child
(b) Does it work when we return in the child from this function?

I think (b) works by design of the kselftest_harness, as this function is
itself executed from a child process.

Regarding (a), it might be cleaner to just

Makes sense, thanks for pointing this out! Have sent the fixlets.