From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1755018AbZHUTPo (ORCPT ); Fri, 21 Aug 2009 15:15:44 -0400 Received: (majordomo@vger.kernel.org) by vger.kernel.org id S1752984AbZHUTPo (ORCPT ); Fri, 21 Aug 2009 15:15:44 -0400 Received: from smtp1.linux-foundation.org ([140.211.169.13]:44179 "EHLO smtp1.linux-foundation.org" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1752461AbZHUTPn (ORCPT ); Fri, 21 Aug 2009 15:15:43 -0400 Date: Fri, 21 Aug 2009 12:13:03 -0700 (PDT) From: Linus Torvalds X-X-Sender: torvalds@localhost.localdomain To: Ingo Molnar cc: Andrew Morton , linux-tip-commits@vger.kernel.org, Arjan van de Ven , Alan Cox , Dave Jones , Kyle McMartin , Greg KH , linux-kernel@vger.kernel.org, hpa@zytor.com, mingo@redhat.com, catalin.marinas@arm.com, a.p.zijlstra@chello.nl, jens.axboe@oracle.com, fweisbec@gmail.com, stable@kernel.org, srostedt@redhat.com, tglx@linutronix.de Subject: Re: [tip:tracing/urgent] tracing: Fix too large stack usage in do_one_initcall() In-Reply-To: <20090821190226.GA28519@elte.hu> Message-ID: References: <20090821111450.GA32037@elte.hu> <20090821104820.60948082.akpm@linux-foundation.org> <20090821190226.GA28519@elte.hu> User-Agent: Alpine 2.01 (LFD 1184 2008-12-16) MIME-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org On Fri, 21 Aug 2009, Ingo Molnar wrote: > > > > That's why I think the async thing could fix this - if we _force_ > > async calls to be asynchronous, you won't have the deep callchains > > for all the device discovery thing. > > Agreed. OTOH we have deep callchains in things like execve() too > which seem to be a lot harder to fix - and those have been around > for the past ~10 years since i've been looking at max-stacktraces. > I think 4K doesnt cut it anymore. I agree that we have stack traces that are too deep in other areas too. At the same time, it does tend to be the case that initcalls are special due to having those long chains of detection (PCI -> driver -> bus -> device), so targeting them specially is probably a good idea. (There are other "special" chains that tend to be long, like "mount": you often end up doing things like reading root inode information etc down a fairly deep chain). And we probably should really complain more actively about big stack frames. We have that CONFIG_FRAME_WARN thing, but it's set very high. Linus