From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1755877AbaCZTGV (ORCPT ); Wed, 26 Mar 2014 15:06:21 -0400 Received: from cantor2.suse.de ([195.135.220.15]:51572 "EHLO mx2.suse.de" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1751580AbaCZTGT (ORCPT ); Wed, 26 Mar 2014 15:06:19 -0400 Date: Wed, 26 Mar 2014 20:06:15 +0100 From: Jan Kara To: One Thousand Gnomes Cc: Jan Kara , Andrew Morton , LKML , pmladek@suse.cz, Frederic Weisbecker , Steven Rostedt Subject: Re: [PATCH 6/8] printk: Start printing handover kthreads on demand Message-ID: <20140326190615.GB18118@quack.suse.cz> References: <1395770101-24534-1-git-send-email-jack@suse.cz> <1395770101-24534-7-git-send-email-jack@suse.cz> <20140326171641.0e117018@alan.etchedpixels.co.uk> MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <20140326171641.0e117018@alan.etchedpixels.co.uk> User-Agent: Mutt/1.5.21 (2010-09-15) Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org On Wed 26-03-14 17:16:41, One Thousand Gnomes wrote: > On Tue, 25 Mar 2014 18:54:59 +0100 > Jan Kara wrote: > > > Start kthreads for handing over printing only when printk.offload_chars > > is set to value > 0 (i.e., when print offloading gets enabled). > > The original set it was suggested that having all these threads around > that 99.99999% of users don't need was a waste and why not start them > when needed. Also I asked why you couldn't just schedule_work the jobs > onto CPUs instead. > > I don't recall seeing a reply to that - and it doesn't appear fixed here ? I replied to you (https://lkml.org/lkml/2014/3/14/176) regarding why people refused schedule_work() to be used. Regarding starting threads only when needed: Starting threads from console_unlock() is IHMO a no-go because that can happen in a context in which we cannot just call create_kthread(). So the solution I proposed - and noone really objected - was what this patch does. Also note that we agreed with Andrew that after this feature gets some testing, we will change the default to disabled so that most people don't pay the unnecessary overhead. Honza -- Jan Kara SUSE Labs, CR