From mboxrd@z Thu Jan  1 00:00:00 1970
From: Sebastian Siewior <bigeasy@linutronix.de>
Subject: Re: [PATCH PREEMPT RT] rt-mutex: fix deadlock in device mapper
Date: Thu, 23 Nov 2017 15:42:31 +0100
Message-ID: <20171123144231.GD2384@linutronix.de>
References: <20171117145744.t366d2ztxj2qqnco@linutronix.de>
 <1511030230.12841.42.camel@gmx.de>
 <alpine.LRH.2.02.1711201607230.24006@file01.intranet.prod.int.rdu2.redhat.com>
 <1511234430.7672.26.camel@gmx.de>
 <alpine.DEB.2.20.1711210931460.1782@nanos>
 <1511255884.7672.49.camel@gmx.de>
 <alpine.LRH.2.02.1711211107550.12859@file01.intranet.prod.int.rdu2.redhat.com>
 <1511285619.30074.13.camel@gmx.de>
 <alpine.LRH.2.02.1711211450410.10766@file01.intranet.prod.int.rdu2.redhat.com>
 <1511299251.30074.72.camel@gmx.de>
Mime-Version: 1.0
Content-Type: text/plain; charset=utf-8
Content-Transfer-Encoding: 8BIT
Cc: Mikulas Patocka <mpatocka@redhat.com>,
        Thomas Gleixner <tglx@linutronix.de>,
        linux-kernel@vger.kernel.org, Ingo Molnar <mingo@redhat.com>,
        Steven Rostedt <rostedt@goodmis.org>,
        linux-rt-users@vger.kernel.org
To: Mike Galbraith <efault@gmx.de>
Return-path: <linux-kernel-owner@vger.kernel.org>
Content-Disposition: inline
In-Reply-To: <1511299251.30074.72.camel@gmx.de>
Sender: linux-kernel-owner@vger.kernel.org
List-Id: linux-rt-users.vger.kernel.org

On 2017-11-21 22:20:51 [+0100], Mike Galbraith wrote:
> On Tue, 2017-11-21 at 14:56 -0500, Mikulas Patocka wrote:
> > 
> > If we don't have any reason why it is needed to unplug block requests when 
> > a spinlock is taken - so let's not do this.
> 
> That's perfectly fine.  I guess I shouldn't have even mentioned having
> encountered unplug at mutex being insufficient.

While at it, I intend to drop
fs-jbd2-pull-your-plug-when-waiting-for-space.patch from the -RT queue
for v4.14 which does

--- a/fs/jbd2/checkpoint.c
+++ b/fs/jbd2/checkpoint.c
@@ -116,6 +116,8 @@ void __jbd2_log_wait_for_space(journal_t
 	nblocks = jbd2_space_needed(journal);
 	while (jbd2_log_space_left(journal) < nblocks) {
 		write_unlock(&journal->j_state_lock);
+		if (current->plug)
+			io_schedule();
 		mutex_lock(&journal->j_checkpoint_mutex);
 
 		/*

and is/was probably a workaround for the missing schedule while blocking
on mutex/rwsem.

> 	-Mike

Sebastian