From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S262765AbTLNWRv (ORCPT ); Sun, 14 Dec 2003 17:17:51 -0500 Received: (majordomo@vger.kernel.org) by vger.kernel.org id S262776AbTLNWRv (ORCPT ); Sun, 14 Dec 2003 17:17:51 -0500 Received: from mx1.elte.hu ([157.181.1.137]:21134 "EHLO mx1.elte.hu") by vger.kernel.org with ESMTP id S262765AbTLNWRu (ORCPT ); Sun, 14 Dec 2003 17:17:50 -0500 Date: Sun, 14 Dec 2003 23:17:02 +0100 (CET) From: Ingo Molnar Reply-To: Ingo Molnar To: Linus Torvalds Cc: Petr Vandrovec , Kernel Mailing List , Andrew Morton , Roland McGrath Subject: Re: [patch] Re: Problem with exiting threads under NPTL In-Reply-To: Message-ID: References: <20031214052516.GA313@vana.vc.cvut.cz> MIME-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII X-ELTE-SpamVersion: SpamAssassin ELTE 1.0 X-ELTE-VirusStatus: clean X-ELTE-SpamCheck: no X-ELTE-SpamCheck-Details: score=-4.9, required 5.9, BAYES_00 -4.90 X-ELTE-SpamLevel: X-ELTE-SpamScore: -4 Sender: linux-kernel-owner@vger.kernel.org X-Mailing-List: linux-kernel@vger.kernel.org On Sun, 14 Dec 2003, Linus Torvalds wrote: > Even though the parent ignores SIGCHLD it _can_ be running on another > CPU in "wait4()". And since we drop the tasklist lock before we do the > "release_task()" on the leader, and since the leader is still marked > TASK_ZOMBIE, we may have _two_ different people trying to release it. > First the parent, and then the last thread that still remembers the > leader. are you sure this can happen? eligible_child() does this: if (p->exit_signal == -1 && !p->ptrace) return 0; so can the parallel situation really happen? Ingo