Re: [PATCH 1/5] sched: Convert thread_group_cputime() to use for_each_thread()

[Oleg Nesterov <oleg@redhat.com>]

On 04/10, Peter Zijlstra wrote:
>
> On Thu, Apr 10, 2014 at 07:29:38PM +0200, Oleg Nesterov wrote:
>
> > > +static inline __deprecated
> > > +struct task_struct *next_thread(const struct task_struct *p)
> > >  {
> >
> > Not sure... But probably fine too.
> >
> > I already killed some users of next_thread(). This reminds me about
> > next_tid(), probably it should be converted too.
> >
> > As for, say, __exit_signal() it really needs next_thread(). We can fix
> > it instead of deprecating, or we can add another one with another name.
>
> Well, your Changelog said that next_thread() was faulty too;

If lockless. Sure.

> if
> __exit_signal() is the only site where it is correct we can open-code it
> there. If there's more we should probably create a new function and
> audit all current sites.

So far I can see only 2 users, __exit_signal() and complete_signal().

May be next_tid(), but in fact it needs next_thread_or_zero(). And this
connects to check_hung_task(). What they actually (OK, perhaps) need is
for_each_thread_continue, like list_for_each_entry_continue_rcu.

We will see. I didn't finish this all because I was distracted (and lazy
of course ;).

My plan (== TODO) was:

	- check all current users, carefully convert those who need
	  a special attention (because the semantics differs a bit)

	- the same for thread_group_empty()

	- reimplement next_thread/while_each_thread via ->thread_head

	- kill task_struct->thread_group

	- all other users can be converted later

Oleg.