From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1751978AbaBLMvZ (ORCPT ); Wed, 12 Feb 2014 07:51:25 -0500 Received: from cdptpa-outbound-snat.email.rr.com ([107.14.166.226]:33219 "EHLO cdptpa-oedge-vip.email.rr.com" rhost-flags-OK-OK-OK-FAIL) by vger.kernel.org with ESMTP id S1751562AbaBLMvY (ORCPT ); Wed, 12 Feb 2014 07:51:24 -0500 Date: Wed, 12 Feb 2014 07:50:59 -0500 From: Steven Rostedt To: Dave Chinner Cc: Linus Torvalds , Jens Axboe , Tejun Heo , Dave Jones , Al Viro , Eric Sandeen , Linux Kernel , xfs@oss.sgi.com Subject: Re: 3.14-rc2 XFS backtrace because irqs_disabled. Message-ID: <20140212075059.0e3d41ab@gandalf.local.home> In-Reply-To: <20140212083513.GF13997@dastard> References: <20140212004403.GA17129@redhat.com> <20140212010941.GM18016@ZenIV.linux.org.uk> <20140212040358.GA25327@redhat.com> <20140212042215.GN18016@ZenIV.linux.org.uk> <20140212054043.GB13997@dastard> <20140212055027.GA28502@redhat.com> <20140212061038.GC13997@dastard> <20140212063150.GD13997@dastard> <20140212083513.GF13997@dastard> X-Mailer: Claws Mail 3.9.3 (GTK+ 2.24.22; x86_64-pc-linux-gnu) MIME-Version: 1.0 Content-Type: text/plain; charset=US-ASCII Content-Transfer-Encoding: 7bit X-RR-Connecting-IP: 107.14.168.130:25 X-Cloudmark-Score: 0 Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org On Wed, 12 Feb 2014 19:35:13 +1100 Dave Chinner wrote: > And it's leaf functions that the CONFIG_STACK_TRACER doesn't catch > on x86-64 (at least, according to the documentation). > CONFIG_DEBUG_STACK_USAGE output is showing up to 800 bytes more > stack usage than the tracer. As such, I also think that > CONFIG_DEBUG_STACK_USAGE output is a more reliable iindication of > stack usage because it is canary based and so captures the very > worst case usage of the process's stack... Yeah, with the new fentry (adding the mcount call before setting up the stack frame), the function tracer can not catch leaf functions, as it is called before the leaf function's frame is set up. Hmm, I wonder if I should add a config to disable fentry and go back to the old mcount that gets called after setting up the stack frame. This will lead to better stack tracing, but you lose out on all the benefits of fentry. -- Steve From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: from relay.sgi.com (relay3.corp.sgi.com [198.149.34.15]) by oss.sgi.com (Postfix) with ESMTP id D3F457F50 for ; Wed, 12 Feb 2014 06:51:27 -0600 (CST) Received: from cuda.sgi.com (cuda1.sgi.com [192.48.157.11]) by relay3.corp.sgi.com (Postfix) with ESMTP id 7CC86AC007 for ; Wed, 12 Feb 2014 04:51:27 -0800 (PST) Received: from cdptpa-oedge-vip.email.rr.com (cdptpa-outbound-snat.email.rr.com [107.14.166.226]) by cuda.sgi.com with ESMTP id OPrPRWexZDNJXfei for ; Wed, 12 Feb 2014 04:51:24 -0800 (PST) Date: Wed, 12 Feb 2014 07:50:59 -0500 From: Steven Rostedt Subject: Re: 3.14-rc2 XFS backtrace because irqs_disabled. Message-ID: <20140212075059.0e3d41ab@gandalf.local.home> In-Reply-To: <20140212083513.GF13997@dastard> References: <20140212004403.GA17129@redhat.com> <20140212010941.GM18016@ZenIV.linux.org.uk> <20140212040358.GA25327@redhat.com> <20140212042215.GN18016@ZenIV.linux.org.uk> <20140212054043.GB13997@dastard> <20140212055027.GA28502@redhat.com> <20140212061038.GC13997@dastard> <20140212063150.GD13997@dastard> <20140212083513.GF13997@dastard> MIME-Version: 1.0 List-Id: XFS Filesystem from SGI List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit Errors-To: xfs-bounces@oss.sgi.com Sender: xfs-bounces@oss.sgi.com To: Dave Chinner Cc: Jens Axboe , Eric Sandeen , Linux Kernel , xfs@oss.sgi.com, Al Viro , Tejun Heo , Dave Jones , Linus Torvalds On Wed, 12 Feb 2014 19:35:13 +1100 Dave Chinner wrote: > And it's leaf functions that the CONFIG_STACK_TRACER doesn't catch > on x86-64 (at least, according to the documentation). > CONFIG_DEBUG_STACK_USAGE output is showing up to 800 bytes more > stack usage than the tracer. As such, I also think that > CONFIG_DEBUG_STACK_USAGE output is a more reliable iindication of > stack usage because it is canary based and so captures the very > worst case usage of the process's stack... Yeah, with the new fentry (adding the mcount call before setting up the stack frame), the function tracer can not catch leaf functions, as it is called before the leaf function's frame is set up. Hmm, I wonder if I should add a config to disable fentry and go back to the old mcount that gets called after setting up the stack frame. This will lead to better stack tracing, but you lose out on all the benefits of fentry. -- Steve _______________________________________________ xfs mailing list xfs@oss.sgi.com http://oss.sgi.com/mailman/listinfo/xfs