From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (qmail 19153 invoked by alias); 27 May 2010 22:25:52 -0000 Received: (qmail 19143 invoked by uid 22791); 27 May 2010 22:25:51 -0000 X-SWARE-Spam-Status: No, hits=-2.1 required=5.0 tests=AWL,BAYES_00,T_RP_MATCHES_RCVD X-Spam-Check-By: sourceware.org Received: from mail.codesourcery.com (HELO mail.codesourcery.com) (38.113.113.100) by sourceware.org (qpsmtpd/0.43rc1) with ESMTP; Thu, 27 May 2010 22:25:46 +0000 Received: (qmail 22505 invoked from network); 27 May 2010 22:25:44 -0000 Received: from unknown (HELO orlando.localnet) (pedro@127.0.0.2) by mail.codesourcery.com with ESMTPA; 27 May 2010 22:25:44 -0000 From: Pedro Alves To: Jan Kratochvil Subject: Re: [patch] Fix crash on /proc/PID/stat race Date: Thu, 27 May 2010 22:50:00 -0000 User-Agent: KMail/1.12.2 (Linux/2.6.31-21-generic; KDE/4.3.2; x86_64; ; ) Cc: gdb-patches@sourceware.org References: <20100527175404.GA16087@host0.dyn.jankratochvil.net> <201005272000.26518.pedro@codesourcery.com> <20100527212037.GA24735@host0.dyn.jankratochvil.net> In-Reply-To: <20100527212037.GA24735@host0.dyn.jankratochvil.net> MIME-Version: 1.0 Content-Type: Text/Plain; charset="iso-8859-1" Content-Transfer-Encoding: 7bit Message-Id: <201005272325.40987.pedro@codesourcery.com> X-IsSubscribed: yes Mailing-List: contact gdb-patches-help@sourceware.org; run by ezmlm Precedence: bulk List-Id: List-Subscribe: List-Archive: List-Post: List-Help: , Sender: gdb-patches-owner@sourceware.org X-SW-Source: 2010-05/txt/msg00669.txt.bz2 On Thursday 27 May 2010 22:20:37, Jan Kratochvil wrote: > On Thu, 27 May 2010 21:00:26 +0200, Pedro Alves wrote: > > Why are we trying to get at the core if we know the process > > is gone? Since the process is already waited for, I'm surprised > > the fopen succeeded in the first place. On a couple of quick tests, > > I always see fopen failing. It sounds like a kernel bug. Can't we > > just skip the core_of_thread call for > > TARGET_WAITKING_EXITED|TARGET_WAITKING_SIGNALLED? > > An additional patch like this one? Yes, exactly. Thanks. > It is IMO not correct for GDB to crash on unexpected /proc/** content. Okay, I don't mind your original patch that much. I think that it adds a bit of complexity by being paranoid. As long as a process hasn't been waited for, the /stat entry should exist, even if the process is zombie. Anyway, if you want to put that one in, it's okay, but please don't lose the comment below: > - p = strchr (p, ')') + 2; /* skip ")" and a whitespace. */ > + if (p != NULL) > + p = strchr (p, ')'); and I don't think the fopen race comment makes sense as is anymore. Also, IWBN if gdbserver was fixed similarly, but I won't ask you to do that. :-) -- Pedro Alves