Re: WARNING: ODEBUG bug in netdev_freemem (2)
From: Thomas Gleixner
Date: Mon Jun 24 2019 - 09:19:03 EST
On Mon, 24 Jun 2019, Dmitry Vyukov wrote:
> On Mon, Jun 24, 2019 at 2:08 PM Eric Dumazet <eric.dumazet@xxxxxxxxx> wrote:
> > >>> ------------[ cut here ]------------
> > >>> ODEBUG: free active (active state 0) object type: timer_list hint:
> > >>> delayed_work_timer_fn+0x0/0x90 arch/x86/include/asm/paravirt.h:767
> > >>
> > >> One of the cleaned up devices has left an active timer which belongs to a
> > >> delayed work. That's all I can decode out of that splat. :(
> > >
> > > Hi Thomas,
> > >
> > > If ODEBUG would memorize full stack traces for object allocation
> > > (using lib/stackdepot.c), it would make this splat actionable, right?
> > > I've fixed https://bugzilla.kernel.org/show_bug.cgi?id=203969 for this.
> > >
> > Not sure this would help in this case as some netdev are allocated through a generic helper.
> > The driver specific portion might not show up in the stack trace.
> > It would be nice here to get the work queue function pointer,
> > so that it gives us a clue which driver needs a fix.
Hrm. Let me think about a way to achieve that after I handled that
regression which is on my desk.
> I see. But isn't the workqueue callback is cleanup_net in this case
> and is in the stack?
> cleanup_net+0x3fb/0x960 net/core/net_namespace.c:553
> process_one_work+0x989/0x1790 kernel/workqueue.c:2269
That's the work which does the cleanup, but I doubt that this is part of
the offending net_device.