locking/mutex: Initialize mutex::magic even when LOCKDEP=n

[PATCH] locking/mutex: Initialize mutex::magic even when LOCKDEP=n

Posted by Boqun Feng 6 days, 8 hours ago

When DEBUG_MUTEXES=y and LOCKDEP=n, mutex_lock() still checks on
->magic, hence debug_mutex_init() should be called in
mutex_init_generic() as well. While we are at it, decouple LOCKDEP
logic from debug_mutex_init(), because in this way debug_mutex_init()
only needs one parameter, and we now have mutex_init_lockep() for
LOCKDEP=y scenarios.

Reported-by: Stephen Rothwell <sfr@canb.auug.org.au>
Closes: https://lore.kernel.org/lkml/20251117202214.4f710f02@canb.auug.org.au/
Reported-by: Nathan Chancellor <nathan@kernel.org>
Closes: https://lore.kernel.org/lkml/20251121215819.GA1374726@ax162/
Fixes: 3572e2edc7b6 ("locking/mutex: Redo __mutex_init()")
Reviewed-by: Waiman Long <longman@redhat.com>
Signed-off-by: Boqun Feng <boqun.feng@gmail.com>
---
Peter,

Feel free to fold it into commit 3572e2edc7b6 ("locking/mutex: Redo
__mutex_init()"), just resend it properly so it won't fall off your
radar ;-)

 kernel/locking/mutex-debug.c | 10 +---------
 kernel/locking/mutex.c       |  8 +++++++-
 kernel/locking/mutex.h       |  5 ++---
 3 files changed, 10 insertions(+), 13 deletions(-)

diff --git a/kernel/locking/mutex-debug.c b/kernel/locking/mutex-debug.c
index 949103fd8e9b..2c6b02d4699b 100644
--- a/kernel/locking/mutex-debug.c
+++ b/kernel/locking/mutex-debug.c
@@ -78,16 +78,8 @@ void debug_mutex_unlock(struct mutex *lock)
 	}
 }
 
-void debug_mutex_init(struct mutex *lock, const char *name,
-		      struct lock_class_key *key)
+void debug_mutex_init(struct mutex *lock)
 {
-#ifdef CONFIG_DEBUG_LOCK_ALLOC
-	/*
-	 * Make sure we are not reinitializing a held lock:
-	 */
-	debug_check_no_locks_freed((void *)lock, sizeof(*lock));
-	lockdep_init_map_wait(&lock->dep_map, name, key, 0, LD_WAIT_SLEEP);
-#endif
 	lock->magic = lock;
 }
 
diff --git a/kernel/locking/mutex.c b/kernel/locking/mutex.c
index f3bb352a368d..2a1d165b3167 100644
--- a/kernel/locking/mutex.c
+++ b/kernel/locking/mutex.c
@@ -51,6 +51,7 @@ static void __mutex_init_generic(struct mutex *lock)
 #ifdef CONFIG_MUTEX_SPIN_ON_OWNER
 	osq_lock_init(&lock->osq);
 #endif
+	debug_mutex_init(lock);
 }
 
 static inline struct task_struct *__owner_task(unsigned long owner)
@@ -173,7 +174,12 @@ static __always_inline bool __mutex_unlock_fast(struct mutex *lock)
 void mutex_init_lockep(struct mutex *lock, const char *name, struct lock_class_key *key)
 {
 	__mutex_init_generic(lock);
-	debug_mutex_init(lock, name, key);
+
+	/*
+	 * Make sure we are not reinitializing a held lock:
+	 */
+	debug_check_no_locks_freed((void *)lock, sizeof(*lock));
+	lockdep_init_map_wait(&lock->dep_map, name, key, 0, LD_WAIT_SLEEP);
 }
 EXPORT_SYMBOL(mutex_init_lockep);
 #endif /* !CONFIG_DEBUG_LOCK_ALLOC */
diff --git a/kernel/locking/mutex.h b/kernel/locking/mutex.h
index 2e8080a9bee3..9ad4da8cea00 100644
--- a/kernel/locking/mutex.h
+++ b/kernel/locking/mutex.h
@@ -59,8 +59,7 @@ extern void debug_mutex_add_waiter(struct mutex *lock,
 extern void debug_mutex_remove_waiter(struct mutex *lock, struct mutex_waiter *waiter,
 				      struct task_struct *task);
 extern void debug_mutex_unlock(struct mutex *lock);
-extern void debug_mutex_init(struct mutex *lock, const char *name,
-			     struct lock_class_key *key);
+extern void debug_mutex_init(struct mutex *lock);
 #else /* CONFIG_DEBUG_MUTEXES */
 # define debug_mutex_lock_common(lock, waiter)		do { } while (0)
 # define debug_mutex_wake_waiter(lock, waiter)		do { } while (0)
@@ -68,6 +67,6 @@ extern void debug_mutex_init(struct mutex *lock, const char *name,
 # define debug_mutex_add_waiter(lock, waiter, ti)	do { } while (0)
 # define debug_mutex_remove_waiter(lock, waiter, ti)	do { } while (0)
 # define debug_mutex_unlock(lock)			do { } while (0)
-# define debug_mutex_init(lock, name, key)		do { } while (0)
+# define debug_mutex_init(lock)				do { } while (0)
 #endif /* !CONFIG_DEBUG_MUTEXES */
 #endif /* CONFIG_PREEMPT_RT */
-- 
2.50.1 (Apple Git-155)

Re: [PATCH] locking/mutex: Initialize mutex::magic even when LOCKDEP=n

Posted by Peter Zijlstra 3 days, 14 hours ago

On Tue, Nov 25, 2025 at 06:54:21AM -0800, Boqun Feng wrote:
> When DEBUG_MUTEXES=y and LOCKDEP=n, mutex_lock() still checks on
> ->magic, hence debug_mutex_init() should be called in
> mutex_init_generic() as well. While we are at it, decouple LOCKDEP
> logic from debug_mutex_init(), because in this way debug_mutex_init()
> only needs one parameter, and we now have mutex_init_lockep() for
> LOCKDEP=y scenarios.
> 
> Reported-by: Stephen Rothwell <sfr@canb.auug.org.au>
> Closes: https://lore.kernel.org/lkml/20251117202214.4f710f02@canb.auug.org.au/
> Reported-by: Nathan Chancellor <nathan@kernel.org>
> Closes: https://lore.kernel.org/lkml/20251121215819.GA1374726@ax162/
> Fixes: 3572e2edc7b6 ("locking/mutex: Redo __mutex_init()")
> Reviewed-by: Waiman Long <longman@redhat.com>
> Signed-off-by: Boqun Feng <boqun.feng@gmail.com>
> ---
> Peter,
> 
> Feel free to fold it into commit 3572e2edc7b6 ("locking/mutex: Redo
> __mutex_init()"), just resend it properly so it won't fall off your
> radar ;-)

Clearly I've been suffering too much mail again :/ Let me go frob stuff.

Re: [PATCH] locking/mutex: Initialize mutex::magic even when LOCKDEP=n

Posted by Stephen Rothwell 3 days, 18 hours ago

Hi all,

On Tue, 25 Nov 2025 06:54:21 -0800 Boqun Feng <boqun.feng@gmail.com> wrote:
>
> When DEBUG_MUTEXES=y and LOCKDEP=n, mutex_lock() still checks on
> ->magic, hence debug_mutex_init() should be called in  
> mutex_init_generic() as well. While we are at it, decouple LOCKDEP
> logic from debug_mutex_init(), because in this way debug_mutex_init()
> only needs one parameter, and we now have mutex_init_lockep() for
> LOCKDEP=y scenarios.
> 
> Reported-by: Stephen Rothwell <sfr@canb.auug.org.au>
> Closes: https://lore.kernel.org/lkml/20251117202214.4f710f02@canb.auug.org.au/
> Reported-by: Nathan Chancellor <nathan@kernel.org>
> Closes: https://lore.kernel.org/lkml/20251121215819.GA1374726@ax162/
> Fixes: 3572e2edc7b6 ("locking/mutex: Redo __mutex_init()")
> Reviewed-by: Waiman Long <longman@redhat.com>
> Signed-off-by: Boqun Feng <boqun.feng@gmail.com>
> ---
> Peter,
> 
> Feel free to fold it into commit 3572e2edc7b6 ("locking/mutex: Redo
> __mutex_init()"), just resend it properly so it won't fall off your
> radar ;-)
> 
>  kernel/locking/mutex-debug.c | 10 +---------
>  kernel/locking/mutex.c       |  8 +++++++-
>  kernel/locking/mutex.h       |  5 ++---
>  3 files changed, 10 insertions(+), 13 deletions(-)

Any chance if this being applied?  I am still seeing the boot warning.

-- 
Cheers,
Stephen Rothwell

Re: [PATCH] locking/mutex: Initialize mutex::magic even when LOCKDEP=n

Posted by Peter Zijlstra 3 days, 13 hours ago

On Fri, Nov 28, 2025 at 04:08:15PM +1100, Stephen Rothwell wrote:
> Hi all,
> 
> On Tue, 25 Nov 2025 06:54:21 -0800 Boqun Feng <boqun.feng@gmail.com> wrote:
> >
> > When DEBUG_MUTEXES=y and LOCKDEP=n, mutex_lock() still checks on
> > ->magic, hence debug_mutex_init() should be called in  
> > mutex_init_generic() as well. While we are at it, decouple LOCKDEP
> > logic from debug_mutex_init(), because in this way debug_mutex_init()
> > only needs one parameter, and we now have mutex_init_lockep() for
> > LOCKDEP=y scenarios.
> > 
> > Reported-by: Stephen Rothwell <sfr@canb.auug.org.au>
> > Closes: https://lore.kernel.org/lkml/20251117202214.4f710f02@canb.auug.org.au/
> > Reported-by: Nathan Chancellor <nathan@kernel.org>
> > Closes: https://lore.kernel.org/lkml/20251121215819.GA1374726@ax162/
> > Fixes: 3572e2edc7b6 ("locking/mutex: Redo __mutex_init()")
> > Reviewed-by: Waiman Long <longman@redhat.com>
> > Signed-off-by: Boqun Feng <boqun.feng@gmail.com>
> > ---
> > Peter,
> > 
> > Feel free to fold it into commit 3572e2edc7b6 ("locking/mutex: Redo
> > __mutex_init()"), just resend it properly so it won't fall off your
> > radar ;-)
> > 
> >  kernel/locking/mutex-debug.c | 10 +---------
> >  kernel/locking/mutex.c       |  8 +++++++-
> >  kernel/locking/mutex.h       |  5 ++---
> >  3 files changed, 10 insertions(+), 13 deletions(-)
> 
> Any chance if this being applied?  I am still seeing the boot warning.

Should be sorted now, sorry for the delay!

Re: [PATCH] locking/mutex: Initialize mutex::magic even when LOCKDEP=n

Posted by Stephen Rothwell 3 days, 13 hours ago

Hi Peter,

On Fri, 28 Nov 2025 10:55:52 +0100 Peter Zijlstra <peterz@infradead.org> wrote:
>
> Should be sorted now, sorry for the delay!

Excellent, thanks.

-- 
Cheers,
Stephen Rothwell