On 6/1/22 16:43, Tejun Heo wrote:
On Wed, Jun 01, 2022 at 04:37:17PM -0700, Tadeusz Struk wrote:
Yes, but as far as I can see the percpu_ref_kill_and_confirm(&css->refcnt, css_killed_ref_fn)
doesn't change the value of the refcnt, it just causes the css_killed_ref_fn() to be called
Yeah, the base ref is special for percpu_ref.
on it. Only css_get() & css_put() modify the refcnt value.
And for the "free the thing" the css_killed_work_fn() does that.
It calls offline_css(css) and css_put(css) for the whole css hierarchy.
Yeah, the freeing path depends on the css_put(css) invoking css_release()
which schedules the work item which actually frees. Am I misunderstanding
something here?
What I'm trying to say is that it's not really a ref imbalance problem.
I think once the kill_css() has been called on a css, and it is enqueued to be
killed, we shouldn't call css_put(css) on it anymore outside of the "killed call
flow path". It will all be handled by the css_killed_ref_fn() function.
The fact the css_release() is called (via cgroup_kn_unlock()) just after
kill_css() causes the css->destroy_work to be enqueued twice on the same WQ
(cgroup_destroy_wq), just with different function. This results in the
BUG: corrupted list in insert_work issue.
--
Thanks,
Tadeusz
