]> www.infradead.org Git - users/jedix/linux-maple.git/commit
sched_ext: Use SCX_CALL_OP_TASK in task_tick_scx
authorChuyi Zhou <zhouchuyi@bytedance.com>
Wed, 12 Feb 2025 13:09:35 +0000 (21:09 +0800)
committerTejun Heo <tj@kernel.org>
Thu, 13 Feb 2025 16:57:33 +0000 (06:57 -1000)
commitf5717c93a1b999970f3a64d771a1a9ee68cc37d0
tree7472bc84549f8f52fc69bd8f7e086ce4d188ba32
parent2e2006c91c842c551521434466f9b4324719c9a7
sched_ext: Use SCX_CALL_OP_TASK in task_tick_scx

Now when we use scx_bpf_task_cgroup() in ops.tick() to get the cgroup of
the current task, the following error will occur:

scx_foo[3795244] triggered exit kind 1024:
  runtime error (called on a task not being operated on)

The reason is that we are using SCX_CALL_OP() instead of SCX_CALL_OP_TASK()
when calling ops.tick(), which triggers the error during the subsequent
scx_kf_allowed_on_arg_tasks() check.

SCX_CALL_OP_TASK() was first introduced in commit 36454023f50b ("sched_ext:
Track tasks that are subjects of the in-flight SCX operation") to ensure
task's rq lock is held when accessing task's sched_group. Since ops.tick()
is marked as SCX_KF_TERMINAL and task_tick_scx() is protected by the rq
lock, we can use SCX_CALL_OP_TASK() to avoid the above issue. Similarly,
the same changes should be made for ops.disable() and ops.exit_task(), as
they are also protected by task_rq_lock() and it's safe to access the
task's task_group.

Fixes: 36454023f50b ("sched_ext: Track tasks that are subjects of the in-flight SCX operation")
Signed-off-by: Chuyi Zhou <zhouchuyi@bytedance.com>
Signed-off-by: Tejun Heo <tj@kernel.org>
kernel/sched/ext.c