On 09/03/2014 10:40 AM, Peter Zijlstra wrote:

On Wed, Sep 03, 2014 at 07:20:04AM -0400, Peter Hurley wrote:

quoted

Hi Peter,

On 09/03/2014 05:27 AM, Peter Zijlstra wrote:

quoted

On Tue, Sep 02, 2014 at 05:39:22PM -0400, Peter Hurley wrote:

quoted

Provide method for non-essential or non-critical code to warn of
invariant errors.

CC: Ingo Molnar <mingo@kernel.org>
CC: Peter Zijlstra <peterz@infradead.org>
CC: Thomas Gleixner <redacted>
Signed-off-by: Peter Hurley <redacted>
---
 include/linux/spinlock.h         | 1 +
 include/linux/spinlock_api_smp.h | 1 +
 include/linux/spinlock_api_up.h  | 1 +
 3 files changed, 3 insertions(+)

diff --git a/include/linux/spinlock.h b/include/linux/spinlock.h
index 3f2867f..8a9aaf1 100644
--- a/include/linux/spinlock.h
+++ b/include/linux/spinlock.h

@@ -394,6 +394,7 @@ static inline int spin_can_lock(spinlock_t *lock)
 }
 
 #define assert_spin_locked(lock)	assert_raw_spin_locked(&(lock)->rlock)
+#define warn_not_spin_locked(lock)	warn_not_raw_spin_locked(&(lock)->rlock)
 
 /*
  * Pull the atomic_t declaration:

diff --git a/include/linux/spinlock_api_smp.h b/include/linux/spinlock_api_smp.h
index 42dfab8..0ddd499 100644
--- a/include/linux/spinlock_api_smp.h
+++ b/include/linux/spinlock_api_smp.h

@@ -18,6 +18,7 @@
 int in_lock_functions(unsigned long addr);
 
 #define assert_raw_spin_locked(x)	BUG_ON(!raw_spin_is_locked(x))
+#define warn_not_raw_spin_locked(x)	WARN_ON_ONCE(!raw_spin_is_locked(x))

No we should remove assert_spin_locked() not add to it. Use
lockdep_assert_held() instead.

I probably should have been more descriptive in the changelog: this
is not for a test configuration, but rather, an assertion in an
exported api.

So ?

So a lockdep-only assert is unlikely to draw attention to existing bugs,
especially in established drivers.