From mboxrd@z Thu Jan  1 00:00:00 1970
Return-Path: <linux-kernel-owner@vger.kernel.org>
Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand
	id S1754855AbcDDJBx (ORCPT <rfc822;w@1wt.eu>);
	Mon, 4 Apr 2016 05:01:53 -0400
Received: from bombadil.infradead.org ([198.137.202.9]:52745 "EHLO
	bombadil.infradead.org" rhost-flags-OK-OK-OK-OK) by vger.kernel.org
	with ESMTP id S1754816AbcDDJBv (ORCPT
	<rfc822;linux-kernel@vger.kernel.org>);
	Mon, 4 Apr 2016 05:01:51 -0400
Date: Mon, 4 Apr 2016 11:01:43 +0200
From: Peter Zijlstra <peterz@infradead.org>
To: Dmitry Vyukov <dvyukov@google.com>
Cc: Thomas Gleixner <tglx@linutronix.de>, Ingo Molnar <mingo@redhat.com>,
        "H. Peter Anvin" <hpa@zytor.com>, "x86@kernel.org" <x86@kernel.org>,
        LKML <linux-kernel@vger.kernel.org>, Ingo Molnar <mingo@kernel.org>,
        Oleg Nesterov <oleg@redhat.com>,
        syzkaller <syzkaller@googlegroups.com>,
        Kostya Serebryany <kcc@google.com>,
        Alexander Potapenko <glider@google.com>,
        Sasha Levin <sasha.levin@oracle.com>
Subject: Re: lockdep WARNING in get_online_cpus
Message-ID: <20160404090143.GW3448@twins.programming.kicks-ass.net>
References: <CACT4Y+a82mMqLh+faQAhvPs28xMkm_ezyAm9O3r5JzynDpKK4w@mail.gmail.com>
 <CACT4Y+YO6DNEPB5BD7G4hw07XbYCZ=4v4QyR9su1jpOEQPNeEg@mail.gmail.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <CACT4Y+YO6DNEPB5BD7G4hw07XbYCZ=4v4QyR9su1jpOEQPNeEg@mail.gmail.com>
User-Agent: Mutt/1.5.21 (2012-12-30)
Sender: linux-kernel-owner@vger.kernel.org
List-ID: <linux-kernel.vger.kernel.org>
X-Mailing-List: linux-kernel@vger.kernel.org

On Mon, Apr 04, 2016 at 10:19:05AM +0200, Dmitry Vyukov wrote:
> 
> This happens in CONFIG_DEBUG_LOCKDEP code. Is it a bug in lockdep?

We hope not; but it is a new test.

So lockdep needs to check each current lock stack against the recorded
lock dependencies to see if we've gotten ourselves a cycle. Doing this
check is _expensive_.

So what lockdep does is it computes a hash for each lock stack and only
if we've not seen this hash before (actually truncated since we don't
have a full 64bit hashtable) do we go look for cycles.

The new check tries to detect hash-collisions in this cache. A collision
would result in not checking for cycles, even if we've not seen the
stack before.

You've managed to tickle this.

Now, last week I found some bugs in there, and Alfredo added a pretty
printer, so maybe try and add these patches to your testing?

  git://git.kernel.org/pub/scm/linux/kernel/git/peterz/queue.git locking/urgent

> But I always see at the same stack involving perf and jump_label...

So you have a simple reproducer? So that I can have a go at this.