]> git.ipfire.org Git - thirdparty/kernel/linux.git/commit
sched/ext: Avoid null ptr traversal when ->put_prev_task() is called with NULL next
authorJohn Stultz <jstultz@google.com>
Sat, 6 Dec 2025 02:22:03 +0000 (02:22 +0000)
committerTejun Heo <tj@kernel.org>
Mon, 8 Dec 2025 17:18:13 +0000 (07:18 -1000)
commit12b5cd99a05f7cbc2ceb88b3b9601d404ef2236a
tree9dd51125260cb0476a295745d4443487126757a8
parent517a44d18537ef8ab888f71197c80116c14cee0a
sched/ext: Avoid null ptr traversal when ->put_prev_task() is called with NULL next

Early when trying to get sched_ext and proxy-exe working together,
I kept tripping over NULL ptr in put_prev_task_scx() on the line:
  if (sched_class_above(&ext_sched_class, next->sched_class)) {

Which was due to put_prev_task() passes a NULL next, calling:
  prev->sched_class->put_prev_task(rq, prev, NULL);

put_prev_task_scx() already guards for a NULL next in the
switch_class case, but doesn't seem to have a guard for
sched_class_above() check.

I can't say I understand why this doesn't trip usually without
proxy-exec. And in newer kernels there are way fewer
put_prev_task(), and I can't easily reproduce the issue now
even with proxy-exec.

But we still have one put_prev_task() call left in core.c that
seems like it could trip this, so I wanted to send this out for
consideration.

tj: put_prev_task() can be called with NULL @next; however, when @p is
queued, that doesn't happen, so this condition shouldn't currently be
triggerable. The connection isn't straightforward or necessarily reliable,
so add the NULL check even if it can't currently be triggered.

Link: http://lkml.kernel.org/r/20251206022218.1541878-1-jstultz@google.com
Signed-off-by: John Stultz <jstultz@google.com>
Reviewed-by: Andrea Righi <arighi@nvidia.com>
Signed-off-by: Tejun Heo <tj@kernel.org>
kernel/sched/ext.c