Add each lock class to the all_lock_classes list when it is
first registered.
Previously, lock classes were added to all_lock_classes when
the lock class was first used. Since one of the uses of the
list is to find unused locks, this didn't work well.
Signed-off-by: Dale Farnsworth <[email protected]>
---
kernel/lockdep.c | 8 ++++----
1 files changed, 4 insertions(+), 4 deletions(-)
diff --git a/kernel/lockdep.c b/kernel/lockdep.c
index 723bd9f..e630127 100644
--- a/kernel/lockdep.c
+++ b/kernel/lockdep.c
@@ -779,6 +779,10 @@ register_lock_class(struct lockdep_map *lock, unsigned int subclass, int force)
* parallel walking of the hash-list safe:
*/
list_add_tail_rcu(&class->hash_entry, hash_head);
+ /*
+ * Add it to the global list of classes:
+ */
+ list_add_tail_rcu(&class->lock_entry, &all_lock_classes);
if (verbose(class)) {
graph_unlock();
@@ -2282,10 +2286,6 @@ static int mark_lock(struct task_struct *curr, struct held_lock *this,
return 0;
break;
case LOCK_USED:
- /*
- * Add it to the global list of classes:
- */
- list_add_tail_rcu(&this->class->lock_entry, &all_lock_classes);
debug_atomic_dec(&nr_unused_locks);
break;
default:
--
1.5.3.4
On Fri, 2008-02-01 at 08:21 -0700, Dale Farnsworth wrote:
> Add each lock class to the all_lock_classes list when it is
> first registered.
>
> Previously, lock classes were added to all_lock_classes when
> the lock class was first used.
> Since one of the uses of the list is to find unused locks, this didn't
> work well.
You mean in lockdep_proc.c to generate the statistics?
Another potential issue might be count_matching_names() generating a
wrong class version.
On Sun, Feb 03, 2008 at 04:21:02PM +0100, Peter Zijlstra wrote:
> On Fri, 2008-02-01 at 08:21 -0700, Dale Farnsworth wrote:
> > Add each lock class to the all_lock_classes list when it is
> > first registered.
> >
> > Previously, lock classes were added to all_lock_classes when
> > the lock class was first used.
>
> > Since one of the uses of the list is to find unused locks, this didn't
> > work well.
>
> You mean in lockdep_proc.c to generate the statistics?
Yes, that's where I noticed it. On my platform, there was one
unused lock which was missed.
>
> Another potential issue might be count_matching_names() generating a
> wrong class version.
I haven't looked into that issue yet.
-Dale
On Sun, 2008-02-03 at 12:47 -0700, Dale Farnsworth wrote:
> On Sun, Feb 03, 2008 at 04:21:02PM +0100, Peter Zijlstra wrote:
> > On Fri, 2008-02-01 at 08:21 -0700, Dale Farnsworth wrote:
> > > Add each lock class to the all_lock_classes list when it is
> > > first registered.
> > >
> > > Previously, lock classes were added to all_lock_classes when
> > > the lock class was first used.
> >
> > > Since one of the uses of the list is to find unused locks, this didn't
> > > work well.
> >
> > You mean in lockdep_proc.c to generate the statistics?
>
> Yes, that's where I noticed it. On my platform, there was one
> unused lock which was missed.
OK, I'll clarify the description and take the patch. Thanks!