Re: [PATCH v2 4/5] kscan: Use preemption model accessors

On Thu, 11 Nov 2021 at 10:11, Marco Elver [off-list ref] wrote:

Subject s/kscan/kcsan/

On Wed, Nov 10, 2021 at 08:24PM +0000, Valentin Schneider wrote:

quoted

Per PREEMPT_DYNAMIC, checking CONFIG_PREEMPT doesn't tell you the actual
preemption model of the live kernel. Use the newly-introduced accessors
instead.

Signed-off-by: Valentin Schneider <redacted>

Reviewed-by: Marco Elver <elver@google.com>

Though it currently doesn't compile as a module due to missing
EXPORT_SYMBOL of is_preempt*().

quoted

---
 kernel/kcsan/kcsan_test.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/kernel/kcsan/kcsan_test.c b/kernel/kcsan/kcsan_test.c
index dc55fd5a36fc..14d811eb9a21 100644
--- a/kernel/kcsan/kcsan_test.c
+++ b/kernel/kcsan/kcsan_test.c

@@ -1005,13 +1005,13 @@ static const void *nthreads_gen_params(const void *prev, char *desc)
      else
              nthreads *= 2;

-     if (!IS_ENABLED(CONFIG_PREEMPT) || !IS_ENABLED(CONFIG_KCSAN_INTERRUPT_WATCHER)) {
+     if (!is_preempt_full() || !IS_ENABLED(CONFIG_KCSAN_INTERRUPT_WATCHER)) {

In case you introduce the 5th helper I suggested
(is_preempt_full_or_rt() or whatever you'll call it), this one can be
switched, because this check really does want to know if "at least
full preemption" and not "precisely full preemption".

Thanks,
-- Marco