lockdep: register_lock_class() fix
PATCH 1/1 lockdep: register_lock_class() fix
Subject: [patch] lockdep: register_lock_class() fix
From: Ingo Molnar <[email protected]>
the hash_lock must only ever be taken with irqs disabled. This happens in all the important places, except one codepath: register_lock_class(). The race should trigger rarely because register_lock_class() is quite rare and single-threaded (happens during init most of the time).
The fix is to disable irqs.
( bug found live in -rt: there preemption is alot more agressive and
preempting with the hash-lock held caused a lockup.)
Signed-off-by Ingo Molnar <[email protected]>
--- kernel/lockdep.c | 6 ++++++ 1 file changed, 6 insertions(+)
kernel/lockdep.c +6 -0
@@ -1183,6 +1183,7 @@ register_lock_class(struct lockdep_map *struct lockdep_subclass_key *key;struct list_head *hash_head;struct lock_class *class; +unsigned long flags; class =look_up_lock_class (lock, subclass);if (likely (class))
@@ -1204,6 +1205,7 @@ register_lock_class(struct lockdep_map * key = lock->key->subkeys + subclass; hash_head =classhashentry (key); +raw_local_irq_save (flags);__raw_spin_lock (&hash_lock);/* * We have to do the hash-walk again, to avoid races
@@ -1218,6 +1220,7 @@ register_lock_class(struct lockdep_map **/ if (nr_lock_classes >=MAX_LOCKDEP_KEYS ) {__raw_spin_unlock (&hash_lock); +raw_local_irq_restore (flags);debug_locks_off ();printk ("BUG: MAX_LOCKDEP_KEYS too low!\n" );printk ("turning off the locking correctness validator.\n" );
@@ -1240,15 +1243,18 @@ register_lock_class(struct lockdep_map *if (verbose (class)) {__raw_spin_unlock (&hash_lock); +raw_local_irq_restore (flags);printk ("\nnew class %p: %s" , class->key, class->name);if (class->name_version >1 )printk ("#%d" , class->name_version);printk ("\n" );dump_stack (); +raw_local_irq_save (flags);__raw_spin_lock (&hash_lock); } out_unlock_set:__raw_spin_unlock (&hash_lock); +raw_local_irq_restore (flags);if (!subclass || force) lock->class_cache = class;