Re: [PATCH v2] ptrace: fix ptrace vs tasklist_lock race on PREEMPT_RT.

[Oleg Nesterov]

On 04/04, Oleg Nesterov wrote:
>
> Cough. Somehow I can hardly understand v2. For example, if we fix
> wait_task_inactive() correctly, then why ptrace_freeze_traced()
> should take saved_state into account? And how can unfreeze_traced
> hit saved_state == __TASK_TRACED ?

OK, somehow I forgot that ptrace_freeze_traced() is called before
wait_task_inactive(), so it does need to check/change saved_state.

But still, ptrace_unfreeze_traced() can't see task->saved_state ==
__TASK_TRACED, right ?


Oleg.