From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1751819Ab3J2TJF (ORCPT ); Tue, 29 Oct 2013 15:09:05 -0400 Received: from mx1.redhat.com ([209.132.183.28]:29244 "EHLO mx1.redhat.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1751195Ab3J2TJC (ORCPT ); Tue, 29 Oct 2013 15:09:02 -0400 Date: Tue, 29 Oct 2013 20:10:16 +0100 From: Oleg Nesterov To: Ingo Molnar Cc: Peter Zijlstra , Steven Rostedt , Dave Sullivan , linux-kernel@vger.kernel.org Subject: Re: [PATCH 1/1] hung_task debugging: Add tracepoint to report the hang Message-ID: <20131029191016.GB27834@redhat.com> References: <20131019161807.GA7431@redhat.com> <20131019161828.GA7439@redhat.com> <20131020084837.GB16236@gmail.com> MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <20131020084837.GB16236@gmail.com> User-Agent: Mutt/1.5.18 (2008-05-17) Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org On 10/20, Ingo Molnar wrote: > > * Oleg Nesterov wrote: > > > Currently check_hung_task() prints a warning if it detects the problem, > > but it is not convenient to watch the system logs if user-space wants to > > be notified about the hang. > > > > Add the new trace_sched_process_hang() into check_hung_task(), this way > > a user-space monitor can easily wait for the hang and potentially > > resolve a problem. > > I'm wondering, is the data of trace_console() in kernel/printk/printk.c > not sufficient? Probably yes... I do not think they disable CONFIG_PRINTK. But this is obviously much less convenient, they will need to parse the text. And the user-space watchdog will be woken up much more often than necessary. And they could probably simply read /var/log or interact with syslogd somehow, but they specially asked for something better and more robust. But of course, I understand that every tracepoint should be justified. So if you do not like this change I try to convince them to use trace_console(). > If it's not enough then it might be better to add a higher level printk > tracepoint instead - that can catch hung_task messages and (much) more. Not sure I understand... I mean I do not understand why this is really better for them, except this will simplify the parsing a bit. Anyway I'd prefer to not send another doubtful patch ;) Thanks. Oleg.