[PATCH v2 0/4] locking: Some locking code cleanups

[PATCH v2 0/4] locking: Some locking code cleanups

[Waiman Long]

 v2:
  - Add Boqun's "Reviewed-by" tags & twist the mutex patch as suggested
    by him.

This series contains a number of cleanup locking patches.

Waiman Long (4):
  locking/qspinlock: Fix 'wait_early' set but not used warning
  locking/mutex: Clean up mutex.h
  locking/rwsem: Clarify that RWSEM_READER_OWNED is just a hint
  locking/rwsem: Make DEBUG_RWSEMS and PREEMPT_RT mutually exclusive

 include/linux/mutex.h               | 8 ++------
 kernel/locking/qspinlock_paravirt.h | 2 +-
 kernel/locking/rwsem.c              | 6 +++---
 lib/Kconfig.debug                   | 4 ++--
 4 files changed, 8 insertions(+), 12 deletions(-)

-- 
2.39.3

[PATCH v2 1/4] locking/qspinlock: Fix 'wait_early' set but not used warning

[Waiman Long]

When CONFIG_LOCK_EVENT_COUNTS is off, the wait_early variable will be
set but not used. This is expected. Recent compilers will not generate
wait_early code in this case.

Add the __maybe_unused attribute to wait_early for suppressing this
W=1 warning.

Reported-by: kernel test robot <lkp@intel.com>
Closes: https://lore.kernel.org/oe-kbuild-all/202312260422.f4pK3f9m-lkp@intel.com/
Signed-off-by: Waiman Long <longman@redhat.com>
Reviewed-by: Boqun Feng <boqun.feng@gmail.com>
---
 kernel/locking/qspinlock_paravirt.h | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/kernel/locking/qspinlock_paravirt.h b/kernel/locking/qspinlock_paravirt.h
index 6a0184e9c234..ae2b12f68b90 100644
--- a/kernel/locking/qspinlock_paravirt.h
+++ b/kernel/locking/qspinlock_paravirt.h
@@ -294,8 +294,8 @@ static void pv_wait_node(struct mcs_spinlock *node, struct mcs_spinlock *prev)
 {
 	struct pv_node *pn = (struct pv_node *)node;
 	struct pv_node *pp = (struct pv_node *)prev;
+	bool __maybe_unused wait_early;
 	int loop;
-	bool wait_early;
 
 	for (;;) {
 		for (wait_early = false, loop = SPIN_THRESHOLD; loop; loop--) {
-- 
2.39.3

[PATCH v2 2/4] locking/mutex: Clean up mutex.h

[Waiman Long]

CONFIG_DEBUG_MUTEXES and CONFIG_PREEMPT_RT are mutually exclusive. They
can't be both set at the same time.  Move up the mutex_destroy() function
declaration and the __DEBUG_MUTEX_INITIALIZER() macro above the "#ifndef
CONFIG_PREEMPT_RT" section to eliminate duplicated mutex_destroy()
declaration.

Also remove the duplicated mutex_trylock() function declaration in the
CONFIG_PREEMPT_RT section.

Signed-off-by: Waiman Long <longman@redhat.com>
Reviewed-by: Boqun Feng <boqun.feng@gmail.com>
---
 include/linux/mutex.h | 8 ++------
 1 file changed, 2 insertions(+), 6 deletions(-)

diff --git a/include/linux/mutex.h b/include/linux/mutex.h
index 7e208d46ba5b..67edc4ca2bee 100644
--- a/include/linux/mutex.h
+++ b/include/linux/mutex.h
@@ -32,11 +32,9 @@
 # define __DEP_MAP_MUTEX_INITIALIZER(lockname)
 #endif
 
-#ifndef CONFIG_PREEMPT_RT
-
 #ifdef CONFIG_DEBUG_MUTEXES
 
-#define __DEBUG_MUTEX_INITIALIZER(lockname)				\
+# define __DEBUG_MUTEX_INITIALIZER(lockname)				\
 	, .magic = &lockname
 
 extern void mutex_destroy(struct mutex *lock);
@@ -49,6 +47,7 @@ static inline void mutex_destroy(struct mutex *lock) {}
 
 #endif
 
+#ifndef CONFIG_PREEMPT_RT
 /**
  * mutex_init - initialize the mutex
  * @mutex: the mutex to be initialized
@@ -101,9 +100,6 @@ extern bool mutex_is_locked(struct mutex *lock);
 
 extern void __mutex_rt_init(struct mutex *lock, const char *name,
 			    struct lock_class_key *key);
-extern int mutex_trylock(struct mutex *lock);
-
-static inline void mutex_destroy(struct mutex *lock) { }
 
 #define mutex_is_locked(l)	rt_mutex_base_is_locked(&(l)->rtmutex)
 
-- 
2.39.3

[PATCH v2 3/4] locking/rwsem: Clarify that RWSEM_READER_OWNED is just a hint

[Waiman Long]

Clarify in the comments that the RWSEM_READER_OWNED bit in the owner
field is just a hint, not an authoritative state of the rwsem.

Signed-off-by: Waiman Long <longman@redhat.com>
Reviewed-by: Boqun Feng <boqun.feng@gmail.com>
---
 kernel/locking/rwsem.c | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/kernel/locking/rwsem.c b/kernel/locking/rwsem.c
index 2340b6d90ec6..c6d17aee4209 100644
--- a/kernel/locking/rwsem.c
+++ b/kernel/locking/rwsem.c
@@ -35,7 +35,7 @@
 /*
  * The least significant 2 bits of the owner value has the following
  * meanings when set.
- *  - Bit 0: RWSEM_READER_OWNED - The rwsem is owned by readers
+ *  - Bit 0: RWSEM_READER_OWNED - rwsem may be owned by readers (just a hint)
  *  - Bit 1: RWSEM_NONSPINNABLE - Cannot spin on a reader-owned lock
  *
  * When the rwsem is reader-owned and a spinning writer has timed out,
@@ -1002,8 +1002,8 @@ rwsem_down_read_slowpath(struct rw_semaphore *sem, long count, unsigned int stat
 
 	/*
 	 * To prevent a constant stream of readers from starving a sleeping
-	 * waiter, don't attempt optimistic lock stealing if the lock is
-	 * currently owned by readers.
+	 * writer, don't attempt optimistic lock stealing if the lock is
+	 * very likely owned by readers.
 	 */
 	if ((atomic_long_read(&sem->owner) & RWSEM_READER_OWNED) &&
 	    (rcnt > 1) && !(count & RWSEM_WRITER_LOCKED))
-- 
2.39.3

[PATCH v2 4/4] locking/rwsem: Make DEBUG_RWSEMS and PREEMPT_RT mutually exclusive

[Waiman Long]

The debugging code enabled by CONFIG_DEBUG_RWSEMS will only be
compiled in when CONFIG_PREEMPT_RT isn't set. There is no point to
allow CONFIG_DEBUG_RWSEMS to be set in a kernel configuration where
CONFIG_PREEMPT_RT is also set. Make them mutually exclusive.

Signed-off-by: Waiman Long <longman@redhat.com>
Reviewed-by: Boqun Feng <boqun.feng@gmail.com>
---
 lib/Kconfig.debug | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/lib/Kconfig.debug b/lib/Kconfig.debug
index 975a07f9f1cc..cb695bc76d30 100644
--- a/lib/Kconfig.debug
+++ b/lib/Kconfig.debug
@@ -1303,7 +1303,7 @@ config PROVE_LOCKING
 	select DEBUG_SPINLOCK
 	select DEBUG_MUTEXES if !PREEMPT_RT
 	select DEBUG_RT_MUTEXES if RT_MUTEXES
-	select DEBUG_RWSEMS
+	select DEBUG_RWSEMS if !PREEMPT_RT
 	select DEBUG_WW_MUTEX_SLOWPATH
 	select DEBUG_LOCK_ALLOC
 	select PREEMPT_COUNT if !ARCH_NO_PREEMPT
@@ -1426,7 +1426,7 @@ config DEBUG_WW_MUTEX_SLOWPATH
 
 config DEBUG_RWSEMS
 	bool "RW Semaphore debugging: basic checks"
-	depends on DEBUG_KERNEL
+	depends on DEBUG_KERNEL && !PREEMPT_RT
 	help
 	  This debugging feature allows mismatched rw semaphore locks
 	  and unlocks to be detected and reported.
-- 
2.39.3

[tip: locking/core] locking/rwsem: Make DEBUG_RWSEMS and PREEMPT_RT mutually exclusive

[tip-bot2 for Waiman Long]

The following commit has been merged into the locking/core branch of tip:

Commit-ID:     f22f71322a18e90e127f2fc2c11cc2d5191bc4c6
Gitweb:        https://git.kernel.org/tip/f22f71322a18e90e127f2fc2c11cc2d5191bc4c6
Author:        Waiman Long <longman@redhat.com>
AuthorDate:    Thu, 22 Feb 2024 10:05:40 -05:00
Committer:     Ingo Molnar <mingo@kernel.org>
CommitterDate: Wed, 28 Feb 2024 13:08:38 +01:00

locking/rwsem: Make DEBUG_RWSEMS and PREEMPT_RT mutually exclusive

The debugging code enabled by CONFIG_DEBUG_RWSEMS=y will only be
compiled in when CONFIG_PREEMPT_RT isn't set. There is no point to
allow CONFIG_DEBUG_RWSEMS to be set in a kernel configuration where
CONFIG_PREEMPT_RT is also set. Make them mutually exclusive.

Signed-off-by: Waiman Long <longman@redhat.com>
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Reviewed-by: Boqun Feng <boqun.feng@gmail.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Link: https://lore.kernel.org/r/20240222150540.79981-5-longman@redhat.com
---
 lib/Kconfig.debug | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/lib/Kconfig.debug b/lib/Kconfig.debug
index 975a07f..cb695bc 100644
--- a/lib/Kconfig.debug
+++ b/lib/Kconfig.debug
@@ -1303,7 +1303,7 @@ config PROVE_LOCKING
 	select DEBUG_SPINLOCK
 	select DEBUG_MUTEXES if !PREEMPT_RT
 	select DEBUG_RT_MUTEXES if RT_MUTEXES
-	select DEBUG_RWSEMS
+	select DEBUG_RWSEMS if !PREEMPT_RT
 	select DEBUG_WW_MUTEX_SLOWPATH
 	select DEBUG_LOCK_ALLOC
 	select PREEMPT_COUNT if !ARCH_NO_PREEMPT
@@ -1426,7 +1426,7 @@ config DEBUG_WW_MUTEX_SLOWPATH
 
 config DEBUG_RWSEMS
 	bool "RW Semaphore debugging: basic checks"
-	depends on DEBUG_KERNEL
+	depends on DEBUG_KERNEL && !PREEMPT_RT
 	help
 	  This debugging feature allows mismatched rw semaphore locks
 	  and unlocks to be detected and reported.

[tip: locking/core] locking/rwsem: Clarify that RWSEM_READER_OWNED is just a hint

[tip-bot2 for Waiman Long]

The following commit has been merged into the locking/core branch of tip:

Commit-ID:     d566c78659eccf085f905fd266fc461de92eaa8f
Gitweb:        https://git.kernel.org/tip/d566c78659eccf085f905fd266fc461de92eaa8f
Author:        Waiman Long <longman@redhat.com>
AuthorDate:    Thu, 22 Feb 2024 10:05:39 -05:00
Committer:     Ingo Molnar <mingo@kernel.org>
CommitterDate: Wed, 28 Feb 2024 13:08:38 +01:00

locking/rwsem: Clarify that RWSEM_READER_OWNED is just a hint

Clarify in the comments that the RWSEM_READER_OWNED bit in the owner
field is just a hint, not an authoritative state of the rwsem.

Signed-off-by: Waiman Long <longman@redhat.com>
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Reviewed-by: Boqun Feng <boqun.feng@gmail.com>
Link: https://lore.kernel.org/r/20240222150540.79981-4-longman@redhat.com
---
 kernel/locking/rwsem.c | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/kernel/locking/rwsem.c b/kernel/locking/rwsem.c
index 2340b6d..c6d17ae 100644
--- a/kernel/locking/rwsem.c
+++ b/kernel/locking/rwsem.c
@@ -35,7 +35,7 @@
 /*
  * The least significant 2 bits of the owner value has the following
  * meanings when set.
- *  - Bit 0: RWSEM_READER_OWNED - The rwsem is owned by readers
+ *  - Bit 0: RWSEM_READER_OWNED - rwsem may be owned by readers (just a hint)
  *  - Bit 1: RWSEM_NONSPINNABLE - Cannot spin on a reader-owned lock
  *
  * When the rwsem is reader-owned and a spinning writer has timed out,
@@ -1002,8 +1002,8 @@ rwsem_down_read_slowpath(struct rw_semaphore *sem, long count, unsigned int stat
 
 	/*
 	 * To prevent a constant stream of readers from starving a sleeping
-	 * waiter, don't attempt optimistic lock stealing if the lock is
-	 * currently owned by readers.
+	 * writer, don't attempt optimistic lock stealing if the lock is
+	 * very likely owned by readers.
 	 */
 	if ((atomic_long_read(&sem->owner) & RWSEM_READER_OWNED) &&
 	    (rcnt > 1) && !(count & RWSEM_WRITER_LOCKED))

[tip: locking/core] locking/mutex: Simplify <linux/mutex.h>

[tip-bot2 for Waiman Long]

The following commit has been merged into the locking/core branch of tip:

Commit-ID:     d6cac0b6b0115fd0a5f51a49401473626e4e4fe7
Gitweb:        https://git.kernel.org/tip/d6cac0b6b0115fd0a5f51a49401473626e4e4fe7
Author:        Waiman Long <longman@redhat.com>
AuthorDate:    Thu, 22 Feb 2024 10:05:38 -05:00
Committer:     Ingo Molnar <mingo@kernel.org>
CommitterDate: Wed, 28 Feb 2024 13:08:37 +01:00

locking/mutex: Simplify <linux/mutex.h>

CONFIG_DEBUG_MUTEXES and CONFIG_PREEMPT_RT are mutually exclusive. They
can't be both set at the same time.  Move up the mutex_destroy() function
declaration and the __DEBUG_MUTEX_INITIALIZER() macro above the "#ifndef
CONFIG_PREEMPT_RT" section to eliminate duplicated mutex_destroy()
declaration.

Also remove the duplicated mutex_trylock() function declaration in the
CONFIG_PREEMPT_RT section.

Signed-off-by: Waiman Long <longman@redhat.com>
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Reviewed-by: Boqun Feng <boqun.feng@gmail.com>
Link: https://lore.kernel.org/r/20240222150540.79981-3-longman@redhat.com
---
 include/linux/mutex.h | 8 ++------
 1 file changed, 2 insertions(+), 6 deletions(-)

diff --git a/include/linux/mutex.h b/include/linux/mutex.h
index 7e208d4..67edc4c 100644
--- a/include/linux/mutex.h
+++ b/include/linux/mutex.h
@@ -32,11 +32,9 @@
 # define __DEP_MAP_MUTEX_INITIALIZER(lockname)
 #endif
 
-#ifndef CONFIG_PREEMPT_RT
-
 #ifdef CONFIG_DEBUG_MUTEXES
 
-#define __DEBUG_MUTEX_INITIALIZER(lockname)				\
+# define __DEBUG_MUTEX_INITIALIZER(lockname)				\
 	, .magic = &lockname
 
 extern void mutex_destroy(struct mutex *lock);
@@ -49,6 +47,7 @@ static inline void mutex_destroy(struct mutex *lock) {}
 
 #endif
 
+#ifndef CONFIG_PREEMPT_RT
 /**
  * mutex_init - initialize the mutex
  * @mutex: the mutex to be initialized
@@ -101,9 +100,6 @@ extern bool mutex_is_locked(struct mutex *lock);
 
 extern void __mutex_rt_init(struct mutex *lock, const char *name,
 			    struct lock_class_key *key);
-extern int mutex_trylock(struct mutex *lock);
-
-static inline void mutex_destroy(struct mutex *lock) { }
 
 #define mutex_is_locked(l)	rt_mutex_base_is_locked(&(l)->rtmutex)
 

[tip: locking/core] locking/qspinlock: Fix 'wait_early' set but not used warning

[tip-bot2 for Waiman Long]

The following commit has been merged into the locking/core branch of tip:

Commit-ID:     ca4bc2e07b716509fd279d2b449bb42f4263a9c8
Gitweb:        https://git.kernel.org/tip/ca4bc2e07b716509fd279d2b449bb42f4263a9c8
Author:        Waiman Long <longman@redhat.com>
AuthorDate:    Thu, 22 Feb 2024 10:05:37 -05:00
Committer:     Ingo Molnar <mingo@kernel.org>
CommitterDate: Wed, 28 Feb 2024 13:08:37 +01:00

locking/qspinlock: Fix 'wait_early' set but not used warning

When CONFIG_LOCK_EVENT_COUNTS is off, the wait_early variable will be
set but not used. This is expected. Recent compilers will not generate
wait_early code in this case.

Add the __maybe_unused attribute to wait_early for suppressing this
W=1 warning.

Reported-by: kernel test robot <lkp@intel.com>
Signed-off-by: Waiman Long <longman@redhat.com>
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Reviewed-by: Boqun Feng <boqun.feng@gmail.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Link: https://lore.kernel.org/r/20240222150540.79981-2-longman@redhat.com

Closes: https://lore.kernel.org/oe-kbuild-all/202312260422.f4pK3f9m-lkp@intel.com/
---
 kernel/locking/qspinlock_paravirt.h | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/kernel/locking/qspinlock_paravirt.h b/kernel/locking/qspinlock_paravirt.h
index 6a0184e..ae2b12f 100644
--- a/kernel/locking/qspinlock_paravirt.h
+++ b/kernel/locking/qspinlock_paravirt.h
@@ -294,8 +294,8 @@ static void pv_wait_node(struct mcs_spinlock *node, struct mcs_spinlock *prev)
 {
 	struct pv_node *pn = (struct pv_node *)node;
 	struct pv_node *pp = (struct pv_node *)prev;
+	bool __maybe_unused wait_early;
 	int loop;
-	bool wait_early;
 
 	for (;;) {
 		for (wait_early = false, loop = SPIN_THRESHOLD; loop; loop--) {

Re: [tip: locking/core] locking/qspinlock: Fix 'wait_early' set but not used warning

[Ingo Molnar]

* tip-bot2 for Waiman Long <tip-bot2@linutronix.de> wrote:

> The following commit has been merged into the locking/core branch of tip:
> 
> Commit-ID:     ca4bc2e07b716509fd279d2b449bb42f4263a9c8
> Gitweb:        https://git.kernel.org/tip/ca4bc2e07b716509fd279d2b449bb42f4263a9c8
> Author:        Waiman Long <longman@redhat.com>
> AuthorDate:    Thu, 22 Feb 2024 10:05:37 -05:00
> Committer:     Ingo Molnar <mingo@kernel.org>
> CommitterDate: Wed, 28 Feb 2024 13:08:37 +01:00
> 
> locking/qspinlock: Fix 'wait_early' set but not used warning
> 
> When CONFIG_LOCK_EVENT_COUNTS is off, the wait_early variable will be
> set but not used. This is expected. Recent compilers will not generate
> wait_early code in this case.
> 
> Add the __maybe_unused attribute to wait_early for suppressing this
> W=1 warning.
> 
> Reported-by: kernel test robot <lkp@intel.com>
> Signed-off-by: Waiman Long <longman@redhat.com>
> Signed-off-by: Ingo Molnar <mingo@kernel.org>
> Reviewed-by: Boqun Feng <boqun.feng@gmail.com>
> Cc: Linus Torvalds <torvalds@linux-foundation.org>
> Link: https://lore.kernel.org/r/20240222150540.79981-2-longman@redhat.com
> 
> Closes: https://lore.kernel.org/oe-kbuild-all/202312260422.f4pK3f9m-lkp@intel.com/
> ---
>  kernel/locking/qspinlock_paravirt.h | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
> 
> diff --git a/kernel/locking/qspinlock_paravirt.h b/kernel/locking/qspinlock_paravirt.h
> index 6a0184e..ae2b12f 100644
> --- a/kernel/locking/qspinlock_paravirt.h
> +++ b/kernel/locking/qspinlock_paravirt.h
> @@ -294,8 +294,8 @@ static void pv_wait_node(struct mcs_spinlock *node, struct mcs_spinlock *prev)
>  {
>  	struct pv_node *pn = (struct pv_node *)node;
>  	struct pv_node *pp = (struct pv_node *)prev;
> +	bool __maybe_unused wait_early;
>  	int loop;
> -	bool wait_early;

On a second thought, shouldn't this be solved via lockevent_cond_inc()'s
!CONFIG_LOCK_EVENT_COUNTS stub explicitly marking the variable as used, via 
!something like:

   #define lockevent_cond_inc(ev, c)		do { (void)(c); } while (0)

or so, instead of uglifying the usage site?

Thanks,

	Ingo

Re: [tip: locking/core] locking/qspinlock: Fix 'wait_early' set but not used warning

[Waiman Long]

On 3/11/24 06:57, Ingo Molnar wrote:
> * tip-bot2 for Waiman Long <tip-bot2@linutronix.de> wrote:
>
>> The following commit has been merged into the locking/core branch of tip:
>>
>> Commit-ID:     ca4bc2e07b716509fd279d2b449bb42f4263a9c8
>> Gitweb:        https://git.kernel.org/tip/ca4bc2e07b716509fd279d2b449bb42f4263a9c8
>> Author:        Waiman Long <longman@redhat.com>
>> AuthorDate:    Thu, 22 Feb 2024 10:05:37 -05:00
>> Committer:     Ingo Molnar <mingo@kernel.org>
>> CommitterDate: Wed, 28 Feb 2024 13:08:37 +01:00
>>
>> locking/qspinlock: Fix 'wait_early' set but not used warning
>>
>> When CONFIG_LOCK_EVENT_COUNTS is off, the wait_early variable will be
>> set but not used. This is expected. Recent compilers will not generate
>> wait_early code in this case.
>>
>> Add the __maybe_unused attribute to wait_early for suppressing this
>> W=1 warning.
>>
>> Reported-by: kernel test robot <lkp@intel.com>
>> Signed-off-by: Waiman Long <longman@redhat.com>
>> Signed-off-by: Ingo Molnar <mingo@kernel.org>
>> Reviewed-by: Boqun Feng <boqun.feng@gmail.com>
>> Cc: Linus Torvalds <torvalds@linux-foundation.org>
>> Link: https://lore.kernel.org/r/20240222150540.79981-2-longman@redhat.com
>>
>> Closes: https://lore.kernel.org/oe-kbuild-all/202312260422.f4pK3f9m-lkp@intel.com/
>> ---
>>   kernel/locking/qspinlock_paravirt.h | 2 +-
>>   1 file changed, 1 insertion(+), 1 deletion(-)
>>
>> diff --git a/kernel/locking/qspinlock_paravirt.h b/kernel/locking/qspinlock_paravirt.h
>> index 6a0184e..ae2b12f 100644
>> --- a/kernel/locking/qspinlock_paravirt.h
>> +++ b/kernel/locking/qspinlock_paravirt.h
>> @@ -294,8 +294,8 @@ static void pv_wait_node(struct mcs_spinlock *node, struct mcs_spinlock *prev)
>>   {
>>   	struct pv_node *pn = (struct pv_node *)node;
>>   	struct pv_node *pp = (struct pv_node *)prev;
>> +	bool __maybe_unused wait_early;
>>   	int loop;
>> -	bool wait_early;
> On a second thought, shouldn't this be solved via lockevent_cond_inc()'s
> !CONFIG_LOCK_EVENT_COUNTS stub explicitly marking the variable as used, via
> !something like:
>
>     #define lockevent_cond_inc(ev, c)		do { (void)(c); } while (0)
>
> or so, instead of uglifying the usage site?

Right, that should work too. Thanks for the suggestion. I will post 
another to do that.

Cheers,
Longman