On Sat, Sep 18, 2021 at 09:27:08AM +0800, Hao Sun [off-list ref] wrote:

Would it be better to add a `data_race` macro to the corresponding
location so that the false report can be disabled?

Something like this

--- a/kernel/cgroup/rstat.c
+++ b/kernel/cgroup/rstat.c

@@ -35,7 +35,7 @@ void cgroup_rstat_updated(struct cgroup *cgrp, int cpu)
         * instead of NULL, we can tell whether @cgrp is on the list by
         * testing the next pointer for NULL.
         */
-       if (cgroup_rstat_cpu(cgrp, cpu)->updated_next)
+       if (data_race(cgroup_rstat_cpu(cgrp, cpu)->updated_next))
                return;

        raw_spin_lock_irqsave(cpu_lock, flags);

?

Makes sense to me. Will you send a patch (if this resolves your KCSAN
noise)?

(IIUC, this becase more visible after commit aa48e47e3906 ("memcg:
infrastructure to flush memcg stats") v5.15-rc1 but it was present since
d8ef4b38cb69 ("Revert "cgroup: Add memory barriers to plug
cgroup_rstat_updated() race window"") v5.7.)

See https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/tools/memory-model/Documentation/access-marking.txt#n58
for more details.

(Interesting, learning...)

Thanks,
Michal