On 04/07, Oleg Nesterov wrote:
>
> On 04/06, Jens Axboe wrote:
> >
> > +extern struct callback_head task_work_exited;
> > +
> > static inline void
> > init_task_work(struct callback_head *twork, task_work_func_t func)
> > {
> > @@ -19,7 +21,7 @@ void __task_work_run(void);
> >
> > static inline bool task_work_pending(void)
> > {
> > - return current->task_works;
> > + return current->task_works && current->task_works != &task_work_exited;
> ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
>
> Well, this penalizes all the current users, they can't hit work_exited.
>
> IIUC, this is needed for the next change which adds task_work_run() into
> io_ring_ctx_wait_and_kill(), right?
>
> could you explain how the exiting can call io_ring_ctx_wait_and_kill()
> after it passed exit_task_work() ?
Note also that currently we assume that task_work_run() must not be called
by the exiting process (except exit_task_work). If io_ring adds the precedent
we need change the PF_EXITING logic in task_work_run().
Oleg.
