Re: next-20190708: kernel BUG at lib/lockref.c:189, softlockups in shrink_dcache...?

[Al Viro]

On Wed, Jul 10, 2019 at 05:09:12PM -0700, Andrew Morton wrote:
> On Wed, 10 Jul 2019 22:13:11 +0200 Pavel Machek <pavel@xxxxxx> wrote:
> 
> > Hi!
> > 
> > I'm getting some nastyness from lockref / memory management.
> > 
> > Any ideas? Any ideas who to talk to?
> > 
> 
> I'd be suspecting Al's a99d7580f66e737 ("Teach shrink_dcache_parent()
> to cope with mixed-filesystem shrink lists").

It is, and it's already fixed.  See 9bdebc2bd1c4 with fixes folded in;
the incremental is

diff --git a/fs/dcache.c b/fs/dcache.c
index d8732cf2e302..01b8cae41a71 100644
--- a/fs/dcache.c
+++ b/fs/dcache.c
@@ -1555,7 +1555,9 @@ void shrink_dcache_parent(struct dentry *parent)
 		d_walk(parent, &data, select_collect2);
 		if (data.victim) {
 			struct dentry *parent;
+			spin_lock(&data.victim->d_lock);
 			if (!shrink_lock_dentry(data.victim)) {
+				spin_unlock(&data.victim->d_lock);
 				rcu_read_unlock();
 			} else {
 				rcu_read_unlock();
 
> This appears to have been added to -next around the -rc6 timeframe,
> which seems awfully late?

I'm not sure if I'll be asking to pull that series this window,
actually ;-/

Sigh...  This cycle had been a mess - spent a month or so very unpleasantly
sick ;-/