[Top] [All Lists]

Re: xfs lockdep trace after unlink

To: "Paul E. McKenney" <paulmck@xxxxxxxxxxxxxxxxxx>
Subject: Re: xfs lockdep trace after unlink
From: Dave Jones <davej@xxxxxxxxxx>
Date: Tue, 8 Oct 2013 17:59:47 -0400
Cc: Linux Kernel <linux-kernel@xxxxxxxxxxxxxxx>, xfs@xxxxxxxxxxx
Delivered-to: xfs@xxxxxxxxxxx
In-reply-to: <20131008213910.GX5790@xxxxxxxxxxxxxxxxxx>
Mail-followup-to: Dave Jones <davej@xxxxxxxxxx>, "Paul E. McKenney" <paulmck@xxxxxxxxxxxxxxxxxx>, Linux Kernel <linux-kernel@xxxxxxxxxxxxxxx>, xfs@xxxxxxxxxxx
References: <20131008212056.GA7467@xxxxxxxxxx> <20131008213910.GX5790@xxxxxxxxxxxxxxxxxx>
User-agent: Mutt/1.5.21 (2010-09-15)
On Tue, Oct 08, 2013 at 02:39:10PM -0700, Paul E. McKenney wrote:
 > On Tue, Oct 08, 2013 at 05:20:56PM -0400, Dave Jones wrote:
 > > I was deleting a kernel tree, when this happened..
 > > RCU, or xfs ?
 > > 
 > I have to ask...  What happens when you bump up MAX_LOCKDEP_CHAINS?
We eat up a even more memory I guess.

kernel/lockdep_internals.h:#define MAX_LOCKDEP_CHAINS_BITS      15
kernel/lockdep_internals.h:#define MAX_LOCKDEP_CHAINS   (1UL << 

I had to increase MAX_LOCKDEP_ENTRIES on my local tree already because
I was hitting that limit. 

 > There is a patch to avoid an RCU/scheduler/perf deadlock, which may
 > be found below.  But this stack doesn't look to me to be matching that
 > deadlock.

I'll add that to my tree until it gets mainlined.



<Prev in Thread] Current Thread [Next in Thread>