[PATCH] locking/lockdep: Fix UBSAN warnings

[PATCH] locking/lockdep: Fix UBSAN warnings

[Kobe Wu]

Fix complaints from UBSAN about signed integer overflow.

========================================================================
UBSAN: Undefined behaviour in kernel/locking/lockdep.c:2998:3
signed integer overflow:
2147483647 + 1 cannot be represented in type 'int'
Call trace:
 dump_backtrace+0x0/0x470
 show_stack+0x14/0x20
 dump_stack+0xc8/0x11c
 ubsan_epilogue+0x14/0xa8
 handle_overflow+0x138/0x1a8
 __ubsan_handle_add_overflow+0x10/0x18
 trace_hardirqs_off_caller+0x1a0/0x268
 trace_hardirqs_off+0x1c/0x28
 rcu_irq_enter_irqson+0x18/0x50
 handle_IPI+0x4cc/0x898
 gic_handle_irq+0x1f4/0x240

When system has been running for a long time, signed integer counters
are not enough for some lockdep statistics. Such as the warning above,
it occurs at debug_atomic_inc(hardirqs_off_events). Using unsigned long
counters can satisfy the requirement. Besides, most of other lockdep
statistics are unsigned. It is better to use unsigned int instead of int.

Remove unused variables.
- max_recursion_depth
- nr_cyclic_check_recursions
- nr_find_usage_forwards_recursions
- nr_find_usage_backwards_recursions

Signed-off-by: Kobe Wu <kobe-cp.wu-NuS5LvNUpcJWk0Htik3J/w@public.gmane.org>
---
 kernel/locking/lockdep_internals.h |   36 ++++++++++++++++--------------------
 1 file changed, 16 insertions(+), 20 deletions(-)

diff --git a/kernel/locking/lockdep_internals.h b/kernel/locking/lockdep_internals.h
index 150ec3f..cc83568 100644
--- a/kernel/locking/lockdep_internals.h
+++ b/kernel/locking/lockdep_internals.h
@@ -131,7 +131,6 @@ extern void get_usage_chars(struct lock_class *class,
 extern unsigned int nr_softirq_chains;
 extern unsigned int nr_process_chains;
 extern unsigned int max_lockdep_depth;
-extern unsigned int max_recursion_depth;
 
 extern unsigned int max_bfs_queue_depth;
 
@@ -160,25 +159,22 @@ extern void get_usage_chars(struct lock_class *class,
  * and we want to avoid too much cache bouncing.
  */
 struct lockdep_stats {
-	int	chain_lookup_hits;
-	int	chain_lookup_misses;
-	int	hardirqs_on_events;
-	int	hardirqs_off_events;
-	int	redundant_hardirqs_on;
-	int	redundant_hardirqs_off;
-	int	softirqs_on_events;
-	int	softirqs_off_events;
-	int	redundant_softirqs_on;
-	int	redundant_softirqs_off;
-	int	nr_unused_locks;
-	int	nr_redundant_checks;
-	int	nr_redundant;
-	int	nr_cyclic_checks;
-	int	nr_cyclic_check_recursions;
-	int	nr_find_usage_forwards_checks;
-	int	nr_find_usage_forwards_recursions;
-	int	nr_find_usage_backwards_checks;
-	int	nr_find_usage_backwards_recursions;
+	unsigned long  chain_lookup_hits;
+	unsigned int   chain_lookup_misses;
+	unsigned long  hardirqs_on_events;
+	unsigned long  hardirqs_off_events;
+	unsigned long  redundant_hardirqs_on;
+	unsigned long  redundant_hardirqs_off;
+	unsigned long  softirqs_on_events;
+	unsigned long  softirqs_off_events;
+	unsigned long  redundant_softirqs_on;
+	unsigned long  redundant_softirqs_off;
+	int            nr_unused_locks;
+	unsigned int   nr_redundant_checks;
+	unsigned int   nr_redundant;
+	unsigned int   nr_cyclic_checks;
+	unsigned int   nr_find_usage_forwards_checks;
+	unsigned int   nr_find_usage_backwards_checks;
 
 	/*
 	 * Per lock class locking operation stat counts
-- 
1.7.9.5

Re: [PATCH] locking/lockdep: Fix UBSAN warnings

[Peter Zijlstra]

On Wed, Jun 12, 2019 at 01:49:04PM +0800, Kobe Wu wrote:
> Fix complaints from UBSAN about signed integer overflow.
> 
> ========================================================================
> UBSAN: Undefined behaviour in kernel/locking/lockdep.c:2998:3
> signed integer overflow:

I don't mind the patch, but this is an UBSAN bug and cannot therefore be
a reason.

Re: [PATCH] locking/lockdep: Fix UBSAN warnings

[Kobe-CP Wu]

On Wed, 2019-06-12 at 09:34 +0200, Peter Zijlstra wrote:
> On Wed, Jun 12, 2019 at 01:49:04PM +0800, Kobe Wu wrote:
> > Fix complaints from UBSAN about signed integer overflow.
> > 
> > ========================================================================
> > UBSAN: Undefined behaviour in kernel/locking/lockdep.c:2998:3
> > signed integer overflow:
> 
> I don't mind the patch, but this is an UBSAN bug and cannot therefore be
> a reason.

I think the description could be amended as follows.
There is unnecessary to mention the UBSAN bug.

locking/lockdep: add size of counters for lockdep statistics

When system has been running for a long time, signed integer counters 
are not enough for some lockdep statistics. Using unsigned long counters
can satisfy the requirement. Besides, most of lockdep statistics are 
unsigned. It is better to use unsigned int instead of int.