From owner-freebsd-arm@freebsd.org Wed Sep 16 02:21:42 2015 Return-Path: Delivered-To: freebsd-arm@mailman.ysv.freebsd.org Received: from mx1.freebsd.org (mx1.freebsd.org [IPv6:2001:1900:2254:206a::19:1]) by mailman.ysv.freebsd.org (Postfix) with ESMTP id 2F76F9C2E42 for ; Wed, 16 Sep 2015 02:21:42 +0000 (UTC) (envelope-from fbsd@www.zefox.net) Received: from www.zefox.net (www.zefox.net [69.239.235.194]) (using TLSv1 with cipher DHE-RSA-CAMELLIA256-SHA (256/256 bits)) (Client did not present a certificate) by mx1.freebsd.org (Postfix) with ESMTPS id E7F631210 for ; Wed, 16 Sep 2015 02:21:41 +0000 (UTC) (envelope-from fbsd@www.zefox.net) Received: from www.zefox.net (localhost [127.0.0.1]) by www.zefox.net (8.14.9/8.14.5) with ESMTP id t8G2JvHT017482; Tue, 15 Sep 2015 19:19:57 -0700 (PDT) (envelope-from fbsd@www.zefox.net) Received: (from fbsd@localhost) by www.zefox.net (8.14.9/8.14.5/Submit) id t8G2JvFF017481; Tue, 15 Sep 2015 19:19:57 -0700 (PDT) (envelope-from fbsd) Date: Tue, 15 Sep 2015 19:19:57 -0700 From: bob prohaska To: Warner Losh Cc: bob prohaska , freebsd-arm@freebsd.org Subject: Re: Reproducible crashes on RPI2 under 11-CURRENT using stress2 Message-ID: <20150916021957.GA17471@www.zefox.net> References: <20150903001037.GC33831@www.zefox.net> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: User-Agent: Mutt/1.4.2.3i X-BeenThere: freebsd-arm@freebsd.org X-Mailman-Version: 2.1.20 Precedence: list List-Id: "Porting FreeBSD to ARM processors." List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , X-List-Received-Date: Wed, 16 Sep 2015 02:21:42 -0000 Hi Warner, On Wed, Sep 02, 2015 at 06:26:45PM -0600, Warner Losh wrote: > The interrupt controller code is not SMP safe. It needs to be fixed or > you'll see instability under load. Stuff like these crashes. > > Run a UP kernel if you need to do high loads. > Does the interrupt controller change behavior when in single user? If not the problems may lie elsewhere.. Running an OS build in single user mode on RPI2 (with necessary services such as swap and powerd turned on) seems to result in a uniformly successful compilation. So far no smsc0 warnings and no crashes in about four tries. Times are a bit slow, ~18 hours. Top reports all four cpu cores active, which suggests that multiprocessing is also active. Kernel config is default, with the ULE scheduler. Thanks for reading, and any insights! bob prohaska