Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1752782AbdI1MHh (ORCPT ); Thu, 28 Sep 2017 08:07:37 -0400 Received: from merlin.infradead.org ([205.233.59.134]:53156 "EHLO merlin.infradead.org" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1751873AbdI1MHg (ORCPT ); Thu, 28 Sep 2017 08:07:36 -0400 Date: Thu, 28 Sep 2017 14:07:22 +0200 From: Peter Zijlstra To: "Levin, Alexander (Sasha Levin)" Cc: Sergey Senozhatsky , Sasha Levin , Thomas Gleixner , Ingo Molnar , "linux-kernel@vger.kernel.org List" , "pmladek@suse.com" , "sergey.senozhatsky@gmail.com" , "rostedt@goodmis.org" Subject: Re: sched: serial port lockdep warning when offlining CPUs Message-ID: <20170928120722.ir5tcclsg336ou2n@hirez.programming.kicks-ass.net> References: <20170928093847.zgcgvxb7ajiaywbr@hirez.programming.kicks-ass.net> <20170928103032.GA2380@jagdpanzerIV.localdomain> <20170928113607.GB30818@jagdpanzerIV.localdomain> <20170928114528.wilknwkyjecqrcaa@sasha-lappy> <20170928120532.cjidduvjj5xygaid@hirez.programming.kicks-ass.net> MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <20170928120532.cjidduvjj5xygaid@hirez.programming.kicks-ass.net> User-Agent: NeoMutt/20170609 (1.8.3) Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org Content-Length: 1005 Lines: 24 On Thu, Sep 28, 2017 at 02:05:32PM +0200, Peter Zijlstra wrote: > (please linewrap normal text at 72 chars) > > On Thu, Sep 28, 2017 at 11:45:30AM +0000, Levin, Alexander (Sasha Levin) wrote: > > Okay, that seems to have delt with the callbacks supressed issue. > > > > I also seem to be hitting it as a result of the scheduler triggering a > > WARN, where print_stack_trace(), for example, just uses printk(). > > > > Now I'm not sure what's your approach here, do we keep converting > > printk to printk_deferred one by one until we whacked every mole? > > Nah.. WARN's _should_ not trigger. If they do we fix that. > > Getting a lockdep splat after a WARN is something I can live with. > > And you really don't want that deferred nonsense for WARNs, because that > means you'll risk not over getting the output. Note that I'm not likely to care too much either way, because I'm simply not using any of that printk() nonsense ;-) I'll repost my patches now; otherwise I'll forget again anyway.