From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1757991AbaITSIM (ORCPT ); Sat, 20 Sep 2014 14:08:12 -0400 Received: from casper.infradead.org ([85.118.1.10]:46864 "EHLO casper.infradead.org" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1756794AbaITSIK (ORCPT ); Sat, 20 Sep 2014 14:08:10 -0400 Date: Sat, 20 Sep 2014 20:08:00 +0200 From: Peter Zijlstra To: Steven Rostedt Cc: Joe Perches , Jan Kara , Markus Trippelsdorf , Geert Uytterhoeven , "linux-kernel@vger.kernel.org" , Andrew Morton Subject: Re: [PATCH] printk: git rid of [sched_delayed] message for printk_deferred Message-ID: <20140920180800.GS2832@worktop.localdomain> References: <20140916173328.6306a5c2@gandalf.local.home> <20140917141816.GO2840@worktop.localdomain> <20140917102255.5cd03071@gandalf.local.home> <20140917223633.GE2848@worktop.localdomain> <20140917203135.6db2ee5e@gandalf.local.home> <20140918173414.GU2840@worktop.localdomain> <20140920051224.GA5573@quack.suse.cz> <20140920154733.GM2832@worktop.localdomain> <1411229447.24444.53.camel@joe-AO725> <20140920123001.4ac60dba@gandalf.local.home> MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <20140920123001.4ac60dba@gandalf.local.home> User-Agent: Mutt/1.5.22.1 (2013-10-16) Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org On Sat, Sep 20, 2014 at 12:30:01PM -0400, Steven Rostedt wrote: > On Sat, 20 Sep 2014 09:10:47 -0700 > Joe Perches wrote: > > > On Sat, 2014-09-20 at 17:47 +0200, Peter Zijlstra wrote: > > > On a whole, printk() is entirely useless for debugging these days, its > > > far too fragile/unreliable to be taken seriously so I really don't care > > > on that point either. > > > > That's unfortunate. > > > > Care to enumerate the issues that you believe make > > printk too fragile/unreliable for debugging? > > I seldom use printk these days. It's far too limited in its uses. For > one, most things worth debugging happen thousands of times a second, > and printk will just slow things down to a crawl if it is used. > Another, is that it can not be used in most critical sections (NMI > handlers and anything that deals with the scheduler). Also, as it no > longer blocks when another CPU is doing a printk, a bug can happen > which crashes the system and the output of that bug will never get > printed due to the delayed output from another CPU having the console > lock. If you're trying to debug a dying machine the early_serial_console which basically only does a stream of OUTBs to the right port is the only reliable thing. For anything else, trace_printk(). Now all I need to figure out is how to keep suspend from killing the early_serial_console setup :-)