> Race and functional tests need to drive khugepaged in step: set up a > layout, let one full scan pass over it, check the result. The khugepaged > selftest already waits for full_scans to advance by two, but only makes
"advance by two"? > progress if scan_sleep_millisecs happens to be short. > > Lift it into khugepaged_full_pass() and drive it through sysfs: any store Lift what? > to scan_sleep_millisecs wakes the daemon, so the barrier completes whatever > the scan cadence. A store can be lost when the daemon is between scans, so > it keeps storing until the pass lands; a store to an awake daemon costs > nothing and queues no extra pass. > > One wake completes one pass only if the whole mm list fits in a scan batch, > so callers need a large pages_to_scan. > > Settings pushes must not start passes either. A store to either sleep knob > wakes the daemon, so thp_write_settings() now writes a khugepaged knob only > when its value changes. The other knobs do not wake, but writing them > uniformly costs nothing. thp_update_num() is exported for tests that want > the same restraint. This very much elaborates on "what" and "how" and very little on "why". I'd change the balance. > Assisted-by: Claude-Code:claude-opus-5 > Tested-by: Muhammad Usama Anjum <[email protected]> > Signed-off-by: Kiryl Shutsemau (Meta) <[email protected]> > > diff --git a/tools/testing/selftests/mm/hugepage_settings.c > b/tools/testing/selftests/mm/hugepage_settings.c > index d7917dce3aba..992efee17b71 100644 > --- a/tools/testing/selftests/mm/hugepage_settings.c > +++ b/tools/testing/selftests/mm/hugepage_settings.c > @@ -183,6 +183,19 @@ void thp_read_settings(struct thp_settings *settings) > } > } > > +/* > + * Write only on change: a store to either sleep knob wakes khugepaged -- > + * __sleep_millisecs_store() clears khugepaged_sleep_expire and wakes the > + * queue -- and settings pushes/pops must not start scan passes nobody > + * asked for; khugepaged_full_pass() is the only sanctioned wake. The > + * other knobs do not wake, but writing them the same way costs nothing. > + */ Hmm, the function name already tells that it only writes on change. Do we need this long text along with it? Surely "khugepaged_full_pass() is ithe only sanctioned mode" is out of line for infrastrucure helper description. -- Sincerely yours, Mike.
