Re: [PATCH] ptrace: cleanup ptrace_init_task()->ptrace_link() path

[Oleg Nesterov <oleg@redhat.com>]

On 10/30, Andrew Morton wrote:
>
> On Sat, 31 Oct 2009 01:55:07 +0100 Oleg Nesterov <oleg@redhat.com> wrote:
>
> > On 10/30, Andrew Morton wrote:
> > >
> > > On Fri, 30 Oct 2009 00:56:56 +0100
> > > Oleg Nesterov <oleg@redhat.com> wrote:
> > >
> > > >  ptrace
> > >
> > > Speaking of which, I'm still sitting on
> > > do_wait-optimization-do-not-place-sub-threads-on-task_struct-children-list.patch.
> >
> > (this patch has nothing to do with ptrace)
> >
> > > Should I drop it?
> >
> > Why? I think this is good optimization and imho cleanup.
> >
> > There is no point to have sub-thread in ->children list and this
> > slows down do_wait() if a child has a lot of threads, it has to
> > iterate over all sub-threads just to filter them out.
> >
>
> On 17 Sep you said:
>
> : Yes, risky... God knows who can do list_for_each(->children) and expect to
> : find the sub-threads. But this is obviously good optimization/simplification.
> :
> : It is just ugly to place sub-threads on ->children list, this buys nothing
> : but slown downs do_wait(). (this was needed, afaics, to handle ptraced but
> : not re-parented threads a long ago).
>
> so that's why I didn't merge it into 2.6.32.  Is the patch still
> considered "risky"?

I hope not, it didn't cause any problems during 3 months in -mm.

Oleg.