On Mon, Feb 03, 2020 at 03:32:28PM +0100, Michal Koutný wrote: > On Sun, Feb 02, 2020 at 10:37:02AM +0100, Christian Brauner <christian.brauner@xxxxxxxxxx> wrote: > > cgroup_post_fork() is called past the point of no return for fork and > > cgroup_css_set_put_fork() is explicitly documented as only being > > callable before forks point of no return: > I missed this and somehow incorrectly assumed it's called at the end of > fork too. I find the css_set refcounting correct now. > > BTW any reason why not to utilize cgroup_css_set_put_fork() for the > regular cleanup in cgroup_post_fork() too? Hmyeah, should be doable if we do: kargs->cset = NULL; cgroup_css_set_put_fork(kargs); Christian