[PATCH] rcu: Fix wrong check in list_splice_init_rcu()

[PATCH] rcu: Fix wrong check in list_splice_init_rcu()

[Jan H. Schönherr]

Hi.

If I don't miss something obvious, the following patch would
be a good idea.

There are currently two users of list_splice_init_rcu(): IPMI and btrfs.
For these, list_splice_init_rcu() is currently a NOP.

If I interpret the code correctly, this fixes a memory leak for
IPMI and something probably worse for btrfs.

The call from within btrfs was introduced only after v2.6.39 with commit
1f78160 ("Btrfs: using rcu lock in the reader side of devices list") by
converting a list_splice_init() call.

Therefore, this patch might be a candidate for v3.0.

Regards
Jan

----8<----
From: Jan H. Schönherr <schnhrr@cs.tu-berlin.de>

Splicing something into an empty list should work, while there is
nothing to do if the source list is empty.

Signed-off-by: Jan H. Schönherr <schnhrr@cs.tu-berlin.de>
---
 include/linux/rculist.h |    2 +-
 1 files changed, 1 insertions(+), 1 deletions(-)

diff --git a/include/linux/rculist.h b/include/linux/rculist.h
index e3beb31..d079290 100644
--- a/include/linux/rculist.h
+++ b/include/linux/rculist.h
@@ -183,7 +183,7 @@ static inline void list_splice_init_rcu(struct list_head *list,
 	struct list_head *last = list->prev;
 	struct list_head *at = head->next;
 
-	if (list_empty(head))
+	if (list_empty(list))
 		return;
 
 	/* "first" and "last" tracking list, so initialize it. */
-- 
1.7.6

Re: [PATCH] rcu: Fix wrong check in list_splice_init_rcu()

[Paul E. McKenney]

On Tue, Jul 19, 2011 at 09:10:26PM +0200, Jan H. Schönherr wrote:
> Hi.
> 
> If I don't miss something obvious, the following patch would
> be a good idea.
> 
> There are currently two users of list_splice_init_rcu(): IPMI and btrfs.
> For these, list_splice_init_rcu() is currently a NOP.
> 
> If I interpret the code correctly, this fixes a memory leak for
> IPMI and something probably worse for btrfs.
> 
> The call from within btrfs was introduced only after v2.6.39 with commit
> 1f78160 ("Btrfs: using rcu lock in the reader side of devices list") by
> converting a list_splice_init() call.
> 
> Therefore, this patch might be a candidate for v3.0.

Excellent catch!  This one has been around for some time, so my guess
is that it goes into v3.1.  Either way, I have queued it, thank you!

							Thanx, Paul

> Regards
> Jan
> 
> ----8<----
> From: Jan H. Schönherr <schnhrr@cs.tu-berlin.de>
> 
> Splicing something into an empty list should work, while there is
> nothing to do if the source list is empty.
> 
> Signed-off-by: Jan H. Schönherr <schnhrr@cs.tu-berlin.de>
> ---
>  include/linux/rculist.h |    2 +-
>  1 files changed, 1 insertions(+), 1 deletions(-)
> 
> diff --git a/include/linux/rculist.h b/include/linux/rculist.h
> index e3beb31..d079290 100644
> --- a/include/linux/rculist.h
> +++ b/include/linux/rculist.h
> @@ -183,7 +183,7 @@ static inline void list_splice_init_rcu(struct list_head *list,
>  	struct list_head *last = list->prev;
>  	struct list_head *at = head->next;
> 
> -	if (list_empty(head))
> +	if (list_empty(list))
>  		return;
> 
>  	/* "first" and "last" tracking list, so initialize it. */
> -- 
> 1.7.6
> 
> --
> To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
> Please read the FAQ at  http://www.tux.org/lkml/

Re: [PATCH] rcu: Fix wrong check in list_splice_init_rcu()

[Chris Mason]

Excerpts from Paul E. McKenney's message of 2011-07-19 16:27:52 -0400:
> On Tue, Jul 19, 2011 at 09:10:26PM +0200, Jan H. Schönherr wrote:
> > Hi.
> > 
> > If I don't miss something obvious, the following patch would
> > be a good idea.
> > 
> > There are currently two users of list_splice_init_rcu(): IPMI and btrfs.
> > For these, list_splice_init_rcu() is currently a NOP.
> > 
> > If I interpret the code correctly, this fixes a memory leak for
> > IPMI and something probably worse for btrfs.
> > 
> > The call from within btrfs was introduced only after v2.6.39 with commit
> > 1f78160 ("Btrfs: using rcu lock in the reader side of devices list") by
> > converting a list_splice_init() call.
> > 
> > Therefore, this patch might be a candidate for v3.0.
> 
> Excellent catch!  This one has been around for some time, so my guess
> is that it goes into v3.1.  Either way, I have queued it, thank you!

A good catch and thanks for the cc.  The btrfs code in question here has
some other problems, and it's fine if this patch doesn't make 3.0.

-chris