On Fri, 2022-03-11 at 18:38 +0100, Eric Farman wrote:
> In the routine test_stop_store_status(), the "running" part of
> the test checks a few of the fields in lowcore (to verify the
> "STORE STATUS" part of the SIGP order), and then ensures that
> the CPU has stopped. But this is backwards, according to the
> Principles of Operation:
> The addressed CPU performs the stop function, fol-
> lowed by the store-status operation (see “Store Sta-
> tus” on page 4-82).
>
> If the CPU were not yet stopped, the contents of the lowcore
> fields would be unpredictable. It works today because the
> library functions wait on the stop function, so the CPU is
> stopped by the time it comes back. Let's first check that the
> CPU is stopped first, just to be clear.
>
> While here, add the same check to the second part of the test,
> even though the CPU is explicitly stopped prior to the SIGP.
>
> Fixes: fc67b07a4 ("s390x: smp: Test stop and store status on a
> running and stopped cpu")
> Signed-off-by: Eric Farman <farman@xxxxxxxxxxxxx>
Reviewed-by: Nico Boehr <nrb@xxxxxxxxxxxxx>
