Re: [patch] ptrace-fix-2.5.33-A1

[Ingo Molnar]

> > tsk->children: tsk's children, which are either untraced or traced by
> > 	tsk.  They have p->parent == p->real_parent == tsk.
> > 	Chained in p->sibling.
> 
> no - the way i wrote it originally was that only untraced children should
> be on the tsk->children list. Traced tasks will be on the debugger's
> ->children list, plus will be on the real parent's ->ptrace_children list.

i cant see how a ->children list that includes all children (traced and
untraced) can be useful. In fact it can be harmful: the debugger needs the
traced children to be on his own child-list, ie. there's a conflict of use
on the ->sibling list field. The ->ptrace_list entry is supposed to help
this situation - obviously use of ->ptrace_list and ->sibling has to be
mutually exclusive, no two parents can have the same task linked over
->sibling.

	Ingo