[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <20200526212707.GE2483@worktop.programming.kicks-ass.net>
Date: Tue, 26 May 2020 23:27:07 +0200
From: Peter Zijlstra <peterz@...radead.org>
To: Waiman Long <longman@...hat.com>
Cc: Qian Cai <cai@....pw>, Ingo Molnar <mingo@...hat.com>,
Will Deacon <will.deacon@....com>, linux-kernel@...r.kernel.org
Subject: Re: [PATCH] locking/lockdep: Increase MAX_LOCKDEP_ENTRIES by half
On Tue, May 26, 2020 at 04:30:58PM -0400, Waiman Long wrote:
> On 5/26/20 3:56 PM, Peter Zijlstra wrote:
> > On Tue, May 26, 2020 at 02:58:50PM -0400, Qian Cai wrote:
> >
> > > I still don't understand why reading all sysfs files on this system
> > > could increase that much, but here is the lockdep file after
> > > running sysfs read to see if you could spot anything obviously,
> > >
> > > https://cailca.github.io/files/lockdep.txt
> > 00000000f011a2a5 OPS: 20 FD: 45 BD: 1 .+.+: kn->active#834
> >
> > is that somewhere near the number of CPUs you have?
> >
> > Anyway, there's very long "kn->active#..." chains in there, which seems
> > to suggest some annotation is all sorts of buggered.
> >
> It is actually one active lock per instance of the kerfs_node structures.
> That means more than 800 sysfs files are accessed in some way. As we could
> have much more than 800 sysfs files in the system, we could easily overwhelm
> the lockdep tables if we really try to access all of them.
A lock per instance is crazy, that's not what lockdep is made for.
Fixing this seems like a far better idea than increasing the numbers.
Powered by blists - more mailing lists