lkml.org 
[lkml]   [2014]   [Feb]   [7]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
Date
From
SubjectRe: mmotm 2014-02-05 list_lru_add lockdep splat
On Thu, Feb 06, 2014 at 02:18:24PM -0800, Hugh Dickins wrote:
> On Thu, 6 Feb 2014, Johannes Weiner wrote:
> > On Wed, Feb 05, 2014 at 07:50:10PM -0800, Hugh Dickins wrote:
> > > ======================================================
> > > [ INFO: SOFTIRQ-safe -> SOFTIRQ-unsafe lock order detected ]
> > > 3.14.0-rc1-mm1 #1 Not tainted
> > > ------------------------------------------------------
> > > kswapd0/48 [HC0[0]:SC0[0]:HE0:SE1] is trying to acquire:
> > > (&(&lru->node[i].lock)->rlock){+.+.-.}, at: [<ffffffff81117064>] list_lru_add+0x80/0xf4
> > >
> > > s already holding:
> > > (&(&mapping->tree_lock)->rlock){..-.-.}, at: [<ffffffff81108c63>] __remove_mapping+0x3b/0x12d
> > > which would create a new lock dependency:
> > > (&(&mapping->tree_lock)->rlock){..-.-.} -> (&(&lru->node[i].lock)->rlock){+.+.-.}
> >
> > Thanks for the report. The first time I saw this on my own machine, I
> > misinterpreted it as a false positive (could have sworn the "possible
> > unsafe scenario" section looked different, too).
> >
> > Looking at it again, there really is a deadlock scenario when the
> > shadow shrinker races with a page cache insertion or deletion and is
> > interrupted by the IO completion handler while holding the list_lru
> > lock:
> >
> > > Possible interrupt unsafe locking scenario:
> > >
> > > CPU0 CPU1
> > > ---- ----
> > > lock(&(&lru->node[i].lock)->rlock);
> > > local_irq_disable();
> > > lock(&(&mapping->tree_lock)->rlock);
> > > lock(&(&lru->node[i].lock)->rlock);
> > > <Interrupt>
> > > lock(&(&mapping->tree_lock)->rlock);
> >
> > Could you please try with the following patch?
>
> Sure, that fixes it for me (with one trivial correction appended), thanks.
> But don't imagine I've given it anything as demanding as thought!
>
> --- hannes/mm/list_lru.c 2014-02-06 08:50:25.104032277 -0800
> +++ hughd/mm/list_lru.c 2014-02-06 08:58:36.884043965 -0800
> @@ -143,7 +143,7 @@ int list_lru_init_key(struct list_lru *l
> }
> return 0;
> }
> -EXPORT_SYMBOL_GPL(list_lru_init);
> +EXPORT_SYMBOL_GPL(list_lru_init_key);
>
> void list_lru_destroy(struct list_lru *lru)
> {

Oops, yes, I usually do non-modular builds. Thanks, will merge this
into the above patch unless Andrew beats me to it.


\
 
 \ /
  Last update: 2014-02-08 00:41    [W:0.047 / U:0.060 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site