From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1756170Ab1LNJjI (ORCPT ); Wed, 14 Dec 2011 04:39:08 -0500 Received: from [205.233.59.134] ([205.233.59.134]:50409 "EHLO merlin.infradead.org" rhost-flags-FAIL-FAIL-OK-OK) by vger.kernel.org with ESMTP id S1755032Ab1LNJjF convert rfc822-to-8bit (ORCPT ); Wed, 14 Dec 2011 04:39:05 -0500 Message-ID: <1323855488.28489.16.camel@twins> Subject: Re: printk() vs tty_io From: Peter Zijlstra To: Linus Torvalds Cc: Greg KH , "Theodore Ts'o" , Alan Cox , linux-kernel , Ingo Molnar Date: Wed, 14 Dec 2011 10:38:08 +0100 In-Reply-To: References: <1323804803.9082.40.camel@twins> Content-Type: text/plain; charset="UTF-8" Content-Transfer-Encoding: 8BIT X-Mailer: Evolution 3.2.1- Mime-Version: 1.0 Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org On Tue, 2011-12-13 at 15:52 -0800, Linus Torvalds wrote: > On Tue, Dec 13, 2011 at 11:33 AM, Peter Zijlstra wrote: > > > > I've been poking at reducing the constraints on printk(), like make it > > work under rq->lock etc.. > > You aren't supposed register a console that wakes things up. But the > only console that honors that afaik is the traditional vt console. > *Maybe* the network console, I didn't check. > > I *assume* you only get this lockdep warning if you have a serial console? I only ever use serial, I'll try and have a go at reproducing any of this on a machine that actually has a screen attached. Anyway, would it make any sense to start enforcing this 'rule'? Can we reasonably make the serial stuff not wake things? Let alone the fbdev/ksm consoles that seem popular these days. Thing is, if everybody and their dog are using ksm, and we cannot make ksm console wake-free, there's a very limited point to my endeavor to make printk() work under rq->lock etc..