On Tue, Mar 04, 2025 at 02:10:18PM +0100, Oleg Nesterov wrote: > I will read this series later, but I see nothing wrong after a quick glance. > > Minor nit below... > > On 03/04, Christian Brauner wrote: > > > > Record the exit code and cgroupid in do_exit() > ^^^^^^^^^^^^ > this is no longer true. In release_task(). Yes, thanks for catching this! > > > @@ -254,6 +255,7 @@ void release_task(struct task_struct *p) > > write_lock_irq(&tasklist_lock); > > ptrace_release_task(p); > > thread_pid = get_pid(p->thread_pid); > > + pidfs_exit(p); > > __exit_signal(p); > > And the next patch rightly moves pidfs_exit() up outside of tasklist. > > Why not call it before write_lock_irq(&tasklist_lock) from the very > beginning? Yes, sorry that was my intention. pidfs_exit() can sleep due to dput(). Thanks!
