Re: [4.4.4-rt11] Possiblie recursive locking detected in kswapd / mb_cache_shrink_scan()

[Josh Cartwright <joshc@ni.com>]

[-- Attachment #1: Type: text/plain, Size: 1514 bytes --]

On Sat, Mar 19, 2016 at 09:33:34AM +0100, Thomas Gleixner wrote:
> On Mon, 14 Mar 2016, Luis Claudio R. Goncalves wrote:
> 
> > Hello,
> > 
> > I just started using 4.4.4-rt11 this afternoon and saw the lockdep splat below.
> > Is this something new or did I miss a patch floating on the list?
> 
> That's new. Looking into it.

It looks like the way that INIT_HLIST_BL_HEAD() is authored, all
instances of the contained raw spinlock will share the same lockdep
class.  That, coupled with the following in mb_cache_shrink(), is
triggering the lockdep violation:

	void
	mb_cache_shrink(struct block_device *bdev)
	{
		[..]
				/*
				 * Prevent any find or get operation on the entry.
				 */
				hlist_bl_lock(ce->e_block_hash_p);
				hlist_bl_lock(ce->e_index_hash_p);

  Josh

diff --git a/include/linux/list_bl.h b/include/linux/list_bl.h
index 44f0b55..d13428a 100644
--- a/include/linux/list_bl.h
+++ b/include/linux/list_bl.h
@@ -42,13 +42,18 @@ struct hlist_bl_node {
 	struct hlist_bl_node *next, **pprev;
 };
 
-static inline void INIT_HLIST_BL_HEAD(struct hlist_bl_head *h)
-{
-	h->first = NULL;
 #ifdef CONFIG_PREEMPT_RT_BASE
-	raw_spin_lock_init(&h->lock);
+#define INIT_HLIST_BL_HEAD(h)		\
+do {					\
+	(h)->first = NULL;		\
+	raw_spin_lock_init(&(h)->lock);	\
+} while (0)
+#else
+#define INIT_HLIST_BL_HEAD(h)		\
+do {					\
+	(h)->first = NULL;		\
+} while (0)
 #endif
-}
 
 static inline void INIT_HLIST_BL_NODE(struct hlist_bl_node *h)
 {

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 473 bytes --]