From mboxrd@z Thu Jan 1 00:00:00 1970 Received: with ECARTIS (v1.0.0; list linux-mips); Wed, 08 Jul 2009 19:40:57 +0200 (CEST) Received: from h5.dl5rb.org.uk ([81.2.74.5]:49756 "EHLO h5.dl5rb.org.uk" rhost-flags-OK-OK-OK-OK) by ftp.linux-mips.org with ESMTP id S1492482AbZGHRks (ORCPT ); Wed, 8 Jul 2009 19:40:48 +0200 Received: from h5.dl5rb.org.uk (localhost.localdomain [127.0.0.1]) by h5.dl5rb.org.uk (8.14.3/8.14.3) with ESMTP id n68HefIT004424; Wed, 8 Jul 2009 18:40:41 +0100 Received: (from ralf@localhost) by h5.dl5rb.org.uk (8.14.3/8.14.3/Submit) id n68HeeWo004422; Wed, 8 Jul 2009 18:40:40 +0100 Date: Wed, 8 Jul 2009 18:40:40 +0100 From: Ralf Baechle To: David Daney Cc: linux-mips@linux-mips.org Subject: Re: [PATCH] MIPS: Avoid clobbering struct pt_regs in kthreads (v2). Message-ID: <20090708174040.GA4069@linux-mips.org> References: <1247072870-14460-1-git-send-email-ddaney@caviumnetworks.com> MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <1247072870-14460-1-git-send-email-ddaney@caviumnetworks.com> User-Agent: Mutt/1.5.18 (2008-05-17) Return-Path: X-Envelope-To: <"|/home/ecartis/ecartis -s linux-mips"> (uid 0) X-Orcpt: rfc822;linux-mips@linux-mips.org Original-Recipient: rfc822;linux-mips@linux-mips.org X-archive-position: 23681 X-ecartis-version: Ecartis v1.0.0 Sender: linux-mips-bounce@linux-mips.org Errors-to: linux-mips-bounce@linux-mips.org X-original-sender: ralf@linux-mips.org Precedence: bulk X-list: linux-mips On Wed, Jul 08, 2009 at 10:07:50AM -0700, David Daney wrote: > The resume() implementation octeon_switch.S examines the saved > cp0_status register. We were clobbering the entire pt_regs structure > in kernel threads leading to random crashes. > > When switching away from a kernel thread, the saved cp0_status is > examined and if bit 30 is set it is cleared and the CP2 state saved > into the pt_regs structure. Since the kernel thread stack overlaid > the pt_regs structure this resulted in a corrupt stack. When the > kthread with the corrupt stack was resumed, it could crash if it used > any of the data in the stack that was clobbered. > > We fix it by moving the kernel thread stack down so it doesn't overlay > pt_regs. > > Differences from v1: Don't adjust the sp by an additional 32 bytes, it > was not needed. Also fix up __KSTK_TOS and > task_pt_regs. Thanks for fixing and testing the issues I raised on IRC. Next I'm wonding what impact the uninitialized state of the stack frame we allocate may have. I think we're ok - but I need to stare at this for a few more minutes. Ralf