From owner-freebsd-current@FreeBSD.ORG Fri Aug 28 07:31:57 2009 Return-Path: Delivered-To: freebsd-current@freebsd.org Received: from mx1.freebsd.org (mx1.freebsd.org [IPv6:2001:4f8:fff6::34]) by hub.freebsd.org (Postfix) with ESMTP id B319610685C8 for ; Fri, 28 Aug 2009 07:31:57 +0000 (UTC) (envelope-from mandrews@bit0.com) Received: from magnum.bit0.com (magnum.bit0.com [207.246.88.226]) by mx1.freebsd.org (Postfix) with ESMTP id 7B01891C7B for ; Fri, 28 Aug 2009 04:21:53 +0000 (UTC) Received: from localhost (localhost [127.0.0.1]) by magnum.bit0.com (Postfix) with ESMTP id 1674CF7431; Fri, 28 Aug 2009 00:21:37 -0400 (EDT) X-Virus-Scanned: amavisd-new at bit0.com Received: from magnum.bit0.com ([127.0.0.1]) by localhost (magnum.int.bit0.com [127.0.0.1]) (amavisd-new, port 10024) with ESMTP id sLRmi8q0Guls; Fri, 28 Aug 2009 00:21:27 -0400 (EDT) Received: from beast.int.bit0.com (beast.int.bit0.com [172.27.0.2]) by magnum.bit0.com (Postfix) with ESMTP; Fri, 28 Aug 2009 00:21:27 -0400 (EDT) Date: Fri, 28 Aug 2009 00:21:26 -0400 (EDT) From: Mike Andrews X-X-Sender: mandrews@beast.int.bit0.com To: "Li, Qing" In-Reply-To: Message-ID: References: <109727.111234.43119@localhost> <20090827181522.81970.qmail@mailgate.gta.com> User-Agent: Alpine 2.00 (BSF 1167 2008-08-23) MIME-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII; format=flowed Cc: freebsd-current@freebsd.org, Larry Baird Subject: RE: 8.0-BETA3 reproducible panic under load X-BeenThere: freebsd-current@freebsd.org X-Mailman-Version: 2.1.5 Precedence: list List-Id: Discussions about the use of FreeBSD-current List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , X-List-Received-Date: Fri, 28 Aug 2009 07:32:01 -0000 On Thu, 27 Aug 2009, Li, Qing wrote: >> >> Not sure if this is what you hit or not. I once saw something > similar. >> The call to LLE_FREE() in ip_output doesn't expect NULL. Adding >> if (ro->ro_lle != NULL) >> just before the call to LLE_FREE() may fix your crashes. >> > > Actually, LLE_FREE() should not be called at all here because the > "if (!nortfree)" block test applies to route entries that are not > from the flow-table. And if the route entry is not from the > flow-table, then the rt_lle field is invalid. > > I will commit the fix. I commented the LLE_FREE() call out over here, and it seems to have stabilized the machine -- no panics in over 5 hours, where previously i could get one within 5 minutes. Thanks! Any ETA on when the commit will hit RELENG_8?