Skip to contentSkip to conversation

lockstat: fix a typo

PATCH In review Yuanhan Liu 5 participants 17 messages Jan 24, 2013 – Feb 22, 2013 3 messages not in the archive Documentation

lore
Share & tools
lore
#0 lore raw

PATCH 1/2 lockstat: fix a typo

It should be CONFIG_LOCK_STAT

---
 Documentation/lockstat.txt |    2 +-
 1 files changed, 1 insertions(+), 1 deletions(-)
Documentation/lockstat.txt +1 -1
index cef00d4..dd2f7b2 100644
@@ -65,7 +65,7 @@ that had to wait on lock acquisition.
 
  - CONFIGURATION
 
-Lock statistics are enabled via CONFIG_LOCK_STATS.
+Lock statistics are enabled via CONFIG_LOCK_STAT.
 
  - USAGE
 
Signature · 1 line
1.7.7.6
#1 lore raw

PATCH 2/2 mutex: use spin_[un]lock instead of arch_spin_[un]lock

Use spin_[un]lock instead of arch_spin_[un]lock in mutex-debug.h so that we can collect the lock statistics of spin_lock_mutex from /proc/lock_stat.

---
 kernel/mutex-debug.h |    4 ++--
 1 files changed, 2 insertions(+), 2 deletions(-)
kernel/mutex-debug.h +2 -2
index 0799fd3..556c0bc 100644
@@ -43,13 +43,13 @@ static inline void mutex_clear_owner(struct mutex *lock)
 							\
 		DEBUG_LOCKS_WARN_ON(in_interrupt());	\
 		local_irq_save(flags);			\
-		arch_spin_lock(&(lock)->rlock.raw_lock);\
+		spin_lock(lock);			\
 		DEBUG_LOCKS_WARN_ON(l->magic != l);	\
 	} while (0)
 
 #define spin_unlock_mutex(lock, flags)				\
 	do {							\
-		arch_spin_unlock(&(lock)->rlock.raw_lock);	\
+		spin_unlock(lock);				\
 		local_irq_restore(flags);			\
 		preempt_check_resched();			\
 	} while (0)
Signature · 1 line
1.7.7.6
#2 lore raw

Re: [PATCH 2/2] mutex: use spin_[un]lock instead of arch_spin_[un]lock

Quoted 15 lines · Yuanhan Liu

Use spin_[un]lock instead of arch_spin_[un]lock in mutex-debug.h so that we can collect the lock statistics of spin_lock_mutex from /proc/lock_stat.

---
 kernel/mutex-debug.h |    4 ++--
 1 files changed, 2 insertions(+), 2 deletions(-)
kernel/mutex-debug.h +0 -0
index 0799fd3..556c0bc 100644

kernel/mutex-debug.h static inline void mutex_clear_owner(struct mutex *lock) · line 46

Quoted 5 lines
kernel/mutex-debug.h +1 -1
@@ -43,13 +43,13 @@ static inline void mutex_clear_owner(struct mutex *lock)
 							\
 		DEBUG_LOCKS_WARN_ON(in_interrupt());	\
 		local_irq_save(flags);			\
-		arch_spin_lock(&(lock)->rlock.raw_lock);\
+		spin_lock(lock);			\

But in that case it could probably use the spin_lock_irqsave() primitive, right?

kernel/mutex-debug.h static inline void mutex_clear_owner(struct mutex *lock) · line 54

Quoted 9 lines
kernel/mutex-debug.h +1 -1
 		DEBUG_LOCKS_WARN_ON(l->magic != l);	\
 	} while (0)
 
 #define spin_unlock_mutex(lock, flags)				\
 	do {							\
-		arch_spin_unlock(&(lock)->rlock.raw_lock);	\
+		spin_unlock(lock);				\
 		local_irq_restore(flags);			\
 		preempt_check_resched();			\

And here spin_unlock_irqrestore().

Thanks,

Ingo

#3 lore raw

Re: [PATCH 2/2] mutex: use spin_[un]lock instead of arch_spin_[un]lock

Quoted 13 lines · Ingo Molnar
Quoted 10 lines · Yuanhan Liu

Use spin_[un]lock instead of arch_spin_[un]lock in mutex-debug.h so that we can collect the lock statistics of spin_lock_mutex from /proc/lock_stat.

---
 kernel/mutex-debug.h |    4 ++--
 1 files changed, 2 insertions(+), 2 deletions(-)
Quoted 13 lines

kernel/mutex-debug.h static inline void mutex_clear_owner(struct mutex *lock) · line 46

Quoted 10 lines
kernel/mutex-debug.h +1 -1
index 0799fd3..556c0bc 100644
@@ -43,13 +43,13 @@ static inline void mutex_clear_owner(struct mutex *lock)
 							\
 		DEBUG_LOCKS_WARN_ON(in_interrupt());	\
 		local_irq_save(flags);			\
-		arch_spin_lock(&(lock)->rlock.raw_lock);\
+		spin_lock(lock);			\

But in that case it could probably use the spin_lock_irqsave() primitive, right?

Right, in that case I should use spin_lock_irqsave.

But one question, why we use spin_lock at kernel/mutex.h, while use 'local_irq_save(); arch_spin_lock' at kernel/mutex-debug.h?

Shouldn't we keep it consistent? Say use spin_lock_irqsave?

Thanks.

	--yliu
Quoted 16 lines

kernel/mutex-debug.h static inline void mutex_clear_owner(struct mutex *lock) · line 54

Quoted 10 lines
kernel/mutex-debug.h +1 -1
 		DEBUG_LOCKS_WARN_ON(l->magic != l);	\
 	} while (0)
 
 #define spin_unlock_mutex(lock, flags)				\
 	do {							\
-		arch_spin_unlock(&(lock)->rlock.raw_lock);	\
+		spin_unlock(lock);				\
 		local_irq_restore(flags);			\
 		preempt_check_resched();			\

And here spin_unlock_irqrestore().

Thanks,

Ingo

#4 lore raw

Re: [PATCH 2/2] mutex: use spin_[un]lock instead of arch_spin_[un]lock

Quoted 28 lines · Yuanhan Liu
Quoted 13 lines · Ingo Molnar
Quoted 10 lines · Yuanhan Liu

Use spin_[un]lock instead of arch_spin_[un]lock in mutex-debug.h so that we can collect the lock statistics of spin_lock_mutex from /proc/lock_stat.

---
 kernel/mutex-debug.h |    4 ++--
 1 files changed, 2 insertions(+), 2 deletions(-)
Quoted 13 lines

kernel/mutex-debug.h static inline void mutex_clear_owner(struct mutex *lock) · line 46

Quoted 10 lines
kernel/mutex-debug.h +1 -1
index 0799fd3..556c0bc 100644
@@ -43,13 +43,13 @@ static inline void mutex_clear_owner(struct mutex *lock)
 							\
 		DEBUG_LOCKS_WARN_ON(in_interrupt());	\
 		local_irq_save(flags);			\
-		arch_spin_lock(&(lock)->rlock.raw_lock);\
+		spin_lock(lock);			\

But in that case it could probably use the spin_lock_irqsave() primitive, right?

Quoted 7 lines

Right, in that case I should use spin_lock_irqsave.

But one question, why we use spin_lock at kernel/mutex.h, while use 'local_irq_save(); arch_spin_lock' at kernel/mutex-debug.h?

Shouldn't we keep it consistent? Say use spin_lock_irqsave?

I think we did it to increase performance with lockdep enabled - this particular lockdep annotation, given the short codepaths, is not that hard to verify - and if it breaks it will break a thousand mutex locking places in the kernel.

So maybe it's better to leave it alone - maybe add a comment that explains the reason.

Thanks,

Ingo

#5 lore raw

Re: [PATCH 2/2] mutex: use spin_[un]lock instead of arch_spin_[un]lock

Quoted 37 lines · Ingo Molnar
Quoted 34 lines · Yuanhan Liu
Quoted 13 lines · Ingo Molnar
Quoted 10 lines · Yuanhan Liu

Use spin_[un]lock instead of arch_spin_[un]lock in mutex-debug.h so that we can collect the lock statistics of spin_lock_mutex from /proc/lock_stat.

---
 kernel/mutex-debug.h |    4 ++--
 1 files changed, 2 insertions(+), 2 deletions(-)
Quoted 13 lines

kernel/mutex-debug.h static inline void mutex_clear_owner(struct mutex *lock) · line 46

Quoted 10 lines
kernel/mutex-debug.h +1 -1
index 0799fd3..556c0bc 100644
@@ -43,13 +43,13 @@ static inline void mutex_clear_owner(struct mutex *lock)
 							\
 		DEBUG_LOCKS_WARN_ON(in_interrupt());	\
 		local_irq_save(flags);			\
-		arch_spin_lock(&(lock)->rlock.raw_lock);\
+		spin_lock(lock);			\

But in that case it could probably use the spin_lock_irqsave() primitive, right?

Right, in that case I should use spin_lock_irqsave.

But one question, why we use spin_lock at kernel/mutex.h, while use 'local_irq_save(); arch_spin_lock' at kernel/mutex-debug.h?

Quoted 6 lines
Quoted 1 line

Shouldn't we keep it consistent? Say use spin_lock_irqsave?

I think we did it to increase performance with lockdep enabled - this particular lockdep annotation, given the short codepaths, is not that hard to verify - and if it breaks it will break a thousand mutex locking places in the kernel.

Thanks for the explanation.

Quoted 3 lines

So maybe it's better to leave it alone - maybe add a comment that explains the reason.

Sorry, I may not get your point clearly. Should I make another patch to convert 'local_irq_save(..); arch_spin_lock(..);' at kernel/mutex-debug.h to spin_lock_irqsave() then?

Thanks.

	--yliu
#6 lore raw

Re: [PATCH 2/2] mutex: use spin_[un]lock instead of arch_spin_[un]lock

Quoted 45 lines · Yuanhan Liu
Quoted 37 lines · Ingo Molnar
Quoted 34 lines · Yuanhan Liu
Quoted 13 lines · Ingo Molnar
Quoted 10 lines · Yuanhan Liu

Use spin_[un]lock instead of arch_spin_[un]lock in mutex-debug.h so that we can collect the lock statistics of spin_lock_mutex from /proc/lock_stat.

---
 kernel/mutex-debug.h |    4 ++--
 1 files changed, 2 insertions(+), 2 deletions(-)
Quoted 13 lines

kernel/mutex-debug.h static inline void mutex_clear_owner(struct mutex *lock) · line 46

Quoted 10 lines
kernel/mutex-debug.h +1 -1
index 0799fd3..556c0bc 100644
@@ -43,13 +43,13 @@ static inline void mutex_clear_owner(struct mutex *lock)
 							\
 		DEBUG_LOCKS_WARN_ON(in_interrupt());	\
 		local_irq_save(flags);			\
-		arch_spin_lock(&(lock)->rlock.raw_lock);\
+		spin_lock(lock);			\

But in that case it could probably use the spin_lock_irqsave() primitive, right?

Right, in that case I should use spin_lock_irqsave.

But one question, why we use spin_lock at kernel/mutex.h, while use 'local_irq_save(); arch_spin_lock' at kernel/mutex-debug.h?

Quoted 6 lines
Quoted 1 line

Shouldn't we keep it consistent? Say use spin_lock_irqsave?

I think we did it to increase performance with lockdep enabled - this particular lockdep annotation, given the short codepaths, is not that hard to verify - and if it breaks it will break a thousand mutex locking places in the kernel.

Quoted 8 lines

Thanks for the explanation.

Quoted 3 lines

So maybe it's better to leave it alone - maybe add a comment that explains the reason.

Sorry, I may not get your point clearly. Should I make another patch to convert 'local_irq_save(..); arch_spin_lock(..);' at kernel/mutex-debug.h to spin_lock_irqsave() then?

No, I'd suggest to add a comment that explains why there's no lockdep annotation in that place.

Thanks,

Ingo

#7 lore raw

Re: [PATCH 2/2] mutex: use spin_[un]lock instead of arch_spin_[un]lock

In reply to Yuanhan Liu
Quoted 25 lines · On Thu, 24 Jan 2013 17:22:45 +0800 Yuanhan Liu <[email protected]>…

Use spin_[un]lock instead of arch_spin_[un]lock in mutex-debug.h so that we can collect the lock statistics of spin_lock_mutex from /proc/lock_stat.

...

kernel/mutex-debug.h +2 -2
@@ -43,13 +43,13 @@ static inline void mutex_clear_owner(struct mutex *lock)
 							\
 		DEBUG_LOCKS_WARN_ON(in_interrupt());	\
 		local_irq_save(flags);			\
-		arch_spin_lock(&(lock)->rlock.raw_lock);\
+		spin_lock(lock);			\
 		DEBUG_LOCKS_WARN_ON(l->magic != l);	\
 	} while (0)
 
 #define spin_unlock_mutex(lock, flags)				\
 	do {							\
-		arch_spin_unlock(&(lock)->rlock.raw_lock);	\
+		spin_unlock(lock);				\
 		local_irq_restore(flags);			\
 		preempt_check_resched();			\
 	} while (0)
Quoted 1 line

From my reading of the c2f21ce2e3128 changelog, this patch might screw

up the -rt kernel:

locking: Implement new raw_spinlock

Now that the raw_spin name space is freed up, we can implement raw_spinlock and the related functions which are used to annotate the locks which are not converted to sleeping spinlocks in preempt-rt.

A side effect is that only such locks can be used with the low level lock fsunctions which circumvent lockdep.

For !rt spin_* functions are mapped to the raw_spin* implementations.

Also, I believe your patch permits this cleanup:

kernel/mutex-debug.h renamed from kernel/mutex-debug.h~mutex-use-spin_lock-instead-of-arch_spin_lock-fix +2 -4
@@ -42,14 +42,12 @@ static inline void mutex_clear_owner(str
 		struct mutex *l = container_of(lock, struct mutex, wait_lock); \
 							\
 		DEBUG_LOCKS_WARN_ON(in_interrupt());	\
-		local_irq_save(flags);			\
-		spin_lock(lock);			\
+		spin_lock_irqsave(lock, flags);		\
 		DEBUG_LOCKS_WARN_ON(l->magic != l);	\
 	} while (0)
 
 #define spin_unlock_mutex(lock, flags)				\
 	do {							\
-		spin_unlock(lock);				\
-		local_irq_restore(flags);			\
+		spin_unlock_irqrestore(lock, flags);		\
 		preempt_check_resched();			\
 	} while (0)

But we should hear from the -rt guys (at least!) before proceeding with this, please.

#8 lore raw

Re: [PATCH 2/2] mutex: use spin_[un]lock instead of arch_spin_[un]lock

Quoted 33 lines · Andrew Morton
Quoted 25 lines · On Thu, 24 Jan 2013 17:22:45 +0800 Yuanhan Liu <[email protected]>…

Use spin_[un]lock instead of arch_spin_[un]lock in mutex-debug.h so that we can collect the lock statistics of spin_lock_mutex from /proc/lock_stat.

...

kernel/mutex-debug.h +2 -2
@@ -43,13 +43,13 @@ static inline void mutex_clear_owner(struct mutex *lock)
 							\
 		DEBUG_LOCKS_WARN_ON(in_interrupt());	\
 		local_irq_save(flags);			\
-		arch_spin_lock(&(lock)->rlock.raw_lock);\
+		spin_lock(lock);			\
 		DEBUG_LOCKS_WARN_ON(l->magic != l);	\
 	} while (0)
 
 #define spin_unlock_mutex(lock, flags)				\
 	do {							\
-		arch_spin_unlock(&(lock)->rlock.raw_lock);	\
+		spin_unlock(lock);				\
 		local_irq_restore(flags);			\
 		preempt_check_resched();			\
 	} while (0)
Quoted 1 line

From my reading of the c2f21ce2e3128 changelog, this patch might screw

up the -rt kernel:

locking: Implement new raw_spinlock

Quoted 8 lines

Now that the raw_spin name space is freed up, we can implement raw_spinlock and the related functions which are used to annotate the locks which are not converted to sleeping spinlocks in preempt-rt.

A side effect is that only such locks can be used with the low level lock fsunctions which circumvent lockdep.

For !rt spin_* functions are mapped to the raw_spin* implementations.

Actually this code is never called when we enable -rt.

At the top of -rt's mutex.h header:

#ifdef CONFIG_PREEMPT_RT_FULL

# include <linux/mutex_rt.h>
#else

[...]

Which basically includes the rest of the mutex.h file.

The kernel/Makefile has:

ifneq ($(CONFIG_PREEMPT_RT_FULL),y)
obj-y += mutex.o
obj-$(CONFIG_DEBUG_MUTEXES) += mutex-debug.o
obj-y += rwsem.o
endif

That's because the -rt kernel uses the priority inheritance rtmutex. Pretty much the same one that userspace futexes use.

These changes are fine and wont hurt -rt. But thanks for think about us :-)

Quoted 13 lines

Also, I believe your patch permits this cleanup:

kernel/mutex-debug.h renamed from kernel/mutex-debug.h~mutex-use-spin_lock-instead-of-arch_spin_lock-fix +1 -2
@@ -42,14 +42,12 @@ static inline void mutex_clear_owner(str
 		struct mutex *l = container_of(lock, struct mutex, wait_lock); \
 							\
 		DEBUG_LOCKS_WARN_ON(in_interrupt());	\
-		local_irq_save(flags);			\
-		spin_lock(lock);			\
+		spin_lock_irqsave(lock, flags);		\

kernel/mutex-debug.h static inline void mutex_clear_owner(str · line 53

Quoted 10 lines
kernel/mutex-debug.h renamed from kernel/mutex-debug.h~mutex-use-spin_lock-instead-of-arch_spin_lock-fix +1 -2
 		DEBUG_LOCKS_WARN_ON(l->magic != l);	\
 	} while (0)
 
 #define spin_unlock_mutex(lock, flags)				\
 	do {							\
-		spin_unlock(lock);				\
-		local_irq_restore(flags);			\
+		spin_unlock_irqrestore(lock, flags);		\
 		preempt_check_resched();			\
 	} while (0)

Actually this perhaps hurts lockdep. We want to keep the arch_spin_(un)lock() versions because each spin_lock() and spin_unlock() needs to be verified by lockdep. Lockdep also verifies mutex locks. But with this change, for every mutex, it's going to also analyze a spin_lock and spin_unlock twice each (one for mutex lock and one for unlock). As this is just locking the mutex internals, it may not be necessary to debug it via lockdep. Hence we probably want to keep the arch_* version.

Quoted 3 lines

But we should hear from the -rt guys (at least!) before proceeding with this, please.

-rt fine, lockdep not so fine.

-- Steve

#9 lore raw

Re: [PATCH 2/2] mutex: use spin_[un]lock instead of arch_spin_[un]lock

Quoted 2 lines · On Thu, 24 Jan 2013 19:50:57 -0500 Steven Rostedt <[email protected]> wrote:

These changes are fine and wont hurt -rt. But thanks for think about us :-)

Thanks (tglx) for writing a useful changelog ;)

Quoted 15 lines
Quoted 15 lines

Also, I believe your patch permits this cleanup:

kernel/mutex-debug.h renamed from kernel/mutex-debug.h~mutex-use-spin_lock-instead-of-arch_spin_lock-fix +1 -2
@@ -42,14 +42,12 @@ static inline void mutex_clear_owner(str
 		struct mutex *l = container_of(lock, struct mutex, wait_lock); \
 							\
 		DEBUG_LOCKS_WARN_ON(in_interrupt());	\
-		local_irq_save(flags);			\
-		spin_lock(lock);			\
+		spin_lock_irqsave(lock, flags);		\
 		DEBUG_LOCKS_WARN_ON(l->magic != l);	\
 	} while (0)
 
Quoted 16 lines

kernel/mutex-debug.h static inline void mutex_clear_owner(str · line 53

Quoted 7 lines
kernel/mutex-debug.h renamed from kernel/mutex-debug.h~mutex-use-spin_lock-instead-of-arch_spin_lock-fix +1 -2
 #define spin_unlock_mutex(lock, flags)				\
 	do {							\
-		spin_unlock(lock);				\
-		local_irq_restore(flags);			\
+		spin_unlock_irqrestore(lock, flags);		\
 		preempt_check_resched();			\
 	} while (0)

Actually this perhaps hurts lockdep. We want to keep the arch_spin_(un)lock() versions because each spin_lock() and spin_unlock() needs to be verified by lockdep. Lockdep also verifies mutex locks. But with this change, for every mutex, it's going to also analyze a spin_lock and spin_unlock twice each (one for mutex lock and one for unlock). As this is just locking the mutex internals, it may not be necessary to debug it via lockdep. Hence we probably want to keep the arch_* version.

In what way is this actually a problem? lockdep will have more work to do (and given the frequency of mutex_lock/unlock, that overhead may be significant). Anything else?

#10 lore raw

Re: [PATCH 2/2] mutex: use spin_[un]lock instead of arch_spin_[un]lock

Quoted 10 lines · Andrew Morton
Quoted 9 lines

Actually this perhaps hurts lockdep. We want to keep the arch_spin_(un)lock() versions because each spin_lock() and spin_unlock() needs to be verified by lockdep. Lockdep also verifies mutex locks. But with this change, for every mutex, it's going to also analyze a spin_lock and spin_unlock twice each (one for mutex lock and one for unlock). As this is just locking the mutex internals, it may not be necessary to debug it via lockdep. Hence we probably want to keep the arch_* version.

Quoted 3 lines

In what way is this actually a problem? lockdep will have more work to do (and given the frequency of mutex_lock/unlock, that overhead may be significant). Anything else?

Ingo probably can answer this better than I can. I'll let him do it when he wakes up tomorrow.

-- Steve

#11 lore raw

Re: [PATCH 2/2] mutex: use spin_[un]lock instead of arch_spin_[un]lock

In reply to Andrew Morton
Quoted 6 lines · Andrew Morton
Quoted 3 lines · On Thu, 24 Jan 2013 17:22:45 +0800 Yuanhan Liu <[email protected]>…

Use spin_[un]lock instead of arch_spin_[un]lock in mutex-debug.h so that we can collect the lock statistics of spin_lock_mutex from /proc/lock_stat.

So, as per the discussion we don't want this patch, because we are using raw locks there to keep mutex lockdep overhead low. The value of lockdep-checking such a basic locking primitive is minimal - it's rarely tweaked and if it breaks we won't have a bootable kernel to begin with.

So instead I suggested a different patch: adding a comment to explain why we don't lockdep-cover the mutex code spinlocks.

kernel/mutex-debug.h static inline void mutex_clear_owner(str · line 45

Quoted 11 lines

Also, I believe your patch permits this cleanup:

kernel/mutex-debug.h renamed from kernel/mutex-debug.h~mutex-use-spin_lock-instead-of-arch_spin_lock-fix +1 -2
@@ -42,14 +42,12 @@ static inline void mutex_clear_owner(str
 		struct mutex *l = container_of(lock, struct mutex, wait_lock); \
 							\
 		DEBUG_LOCKS_WARN_ON(in_interrupt());	\
-		local_irq_save(flags);			\
-		spin_lock(lock);			\
+		spin_lock_irqsave(lock, flags);		\

Yes, I mentioned that yesterday, but we really don't want the change to begin with.

Thanks,

Ingo

A message that is not in the archive (sent off-list or not yet received): CABBm8944FSNDZ2u=PJy0CsZmz=Vm7e+Y5jX+3UFLN5NOSGRgWQ@mail.gmail.com

#13 lore raw

Re: [PATCH 2/2] mutex: use spin_[un]lock instead of arch_spin_[un]lock

Quoted 40 lines · nan chen

2013/1/25 Ingo Molnar <[email protected]>

Quoted 38 lines · Andrew Morton
Quoted 7 lines
Quoted 3 lines · On Thu, 24 Jan 2013 17:22:45 +0800 Yuanhan Liu <[email protected]>…

Use spin_[un]lock instead of arch_spin_[un]lock in mutex-debug.h so that we can collect the lock statistics of spin_lock_mutex from /proc/lock_stat.

So, as per the discussion we don't want this patch, because we are using raw locks there to keep mutex lockdep overhead low. The value of lockdep-checking such a basic locking primitive is minimal - it's rarely tweaked and if it breaks we won't have a bootable kernel to begin with.

So instead I suggested a different patch: adding a comment to explain why we don't lockdep-cover the mutex code spinlocks.

Quoted 3 lines

Also, I believe your patch permits this cleanup:

---

a/kernel/mutex-debug.h~mutex-use-spin_lock-instead-of-arch_spin_lock-fix

Review static inline void mutex_clear_owner(str · line 42

Quoted 3 lines

+++ a/kernel/mutex-debug.h

(quoted diff) +0 -0
@@ -42,14 +42,12 @@ static inline void mutex_clear_owner(str
              struct mutex *l = container_of(lock, struct mutex,

wait_lock); \

kernel/mutex-debug.h static inline void mutex_clear_owner(str · line 45

Quoted 5 lines
kernel/mutex-debug.h renamed from kernel/mutex-debug.h~mutex-use-spin_lock-instead-of-arch_spin_lock-fix +1 -2
                                                      \
              DEBUG_LOCKS_WARN_ON(in_interrupt());    \
-             local_irq_save(flags);                  \
-             spin_lock(lock);                        \
+             spin_lock_irqsave(lock, flags);         \

Yes, I mentioned that yesterday, but we really don't want the change to begin with.

Thanks,

Quoted 7 lines
Quoted 2 lines

Ingo

Hi,

Looks like in mutex.h, it does not disable local interrupt. But why the code disable local interrupt in mutex-debug.h?

To protect against preemption I suspect. preempt_disable() could be used in the mutex-debug.h variant I suppose.

Thanks,

Ingo

A message that is not in the archive (sent off-list or not yet received): CABBm896XGVkubML9TyW_MTXMovd6nHoK0v24jbEDJOipzNqpDA@mail.gmail.com

#15 lore raw

Re: [PATCH 2/2] mutex: use spin_[un]lock instead of arch_spin_[un]lock

Quoted 57 lines · nan chen

2013/1/25 Ingo Molnar <[email protected]>

Quoted 55 lines · nan chen
Quoted 26 lines

2013/1/25 Ingo Molnar <[email protected]>

Quoted 23 lines · Andrew Morton
Quoted 7 lines
Quoted 3 lines · On Thu, 24 Jan 2013 17:22:45 +0800 Yuanhan Liu <[email protected]>…

Use spin_[un]lock instead of arch_spin_[un]lock in mutex-debug.h so that we can collect the lock statistics of spin_lock_mutex from /proc/lock_stat.

So, as per the discussion we don't want this patch, because we are using raw locks there to keep mutex lockdep overhead low. The value of lockdep-checking such a basic locking primitive is minimal - it's rarely tweaked and if it breaks we won't have a bootable kernel to begin with.

So instead I suggested a different patch: adding a comment to explain why we don't lockdep-cover the mutex code spinlocks.

Quoted 3 lines

Also, I believe your patch permits this cleanup:

---

a/kernel/mutex-debug.h~mutex-use-spin_lock-instead-of-arch_spin_lock-fix

Quoted 15 lines
Quoted 15 lines

Review static inline void mutex_clear_owner(str · line 42

Quoted 3 lines

+++ a/kernel/mutex-debug.h

(quoted diff) +0 -0
@@ -42,14 +42,12 @@ static inline void mutex_clear_owner(str
              struct mutex *l = container_of(lock, struct mutex,

wait_lock); \

kernel/mutex-debug.h static inline void mutex_clear_owner(str · line 45

Quoted 5 lines
kernel/mutex-debug.h renamed from kernel/mutex-debug.h~mutex-use-spin_lock-instead-of-arch_spin_lock-fix +1 -2
                                                      \
              DEBUG_LOCKS_WARN_ON(in_interrupt());    \
-             local_irq_save(flags);                  \
-             spin_lock(lock);                        \
+             spin_lock_irqsave(lock, flags);         \

Yes, I mentioned that yesterday, but we really don't want the change to begin with.

Thanks,

Quoted 7 lines
Quoted 2 lines

Ingo

Hi,

Looks like in mutex.h, it does not disable local interrupt. But why the code disable local interrupt in mutex-debug.h?

To protect against preemption I suspect. preempt_disable() could be used in the mutex-debug.h variant I suppose.

Quoted 7 lines
Quoted 4 lines

Thanks,

Ingo

spin_lock() itself already protects against preemption.

Yes, but mutex-debug.h does not use spin_lock(), it uses arch_spin_lock():

#define spin_lock_mutex(lock, flags)                    \
        do {                                            \
                struct mutex *l = container_of(lock, struct mutex, wait_lock); \
                                                        \
                DEBUG_LOCKS_WARN_ON(in_interrupt());    \
                local_irq_save(flags);                  \
                arch_spin_lock(&(lock)->rlock.raw_lock);\
                DEBUG_LOCKS_WARN_ON(l->magic != l);     \
        } while (0)

The original question was why mutex-debug.h (unmodified) uses irq disabling.

Thanks,

Ingo

A message that is not in the archive (sent off-list or not yet received): CABBm8979+XVRA-MQRdczMbStJR7DyPwNUvQL+RTuQL0sECOo8g@mail.gmail.com

#17 lore raw

Re: [PATCH 2/2] mutex: use spin_[un]lock instead of arch_spin_[un]lock

Quoted 14 lines · nan chen
Quoted 14 lines

#define spin_lock_mutex(lock, flags) \
do { \
struct mutex *l = container_of(lock, struct mutex,

wait_lock); \
                                                        \
                DEBUG_LOCKS_WARN_ON(in_interrupt());    \
                local_irq_save(flags);                  \
                arch_spin_lock(&(lock)->rlock.raw_lock);\
                DEBUG_LOCKS_WARN_ON(l->magic != l);     \
        } while (0)

The original question was why mutex-debug.h (unmodified) uses irq disabling.

Quoted 7 lines
Quoted 4 lines

Thanks,

Ingo

So, the reason why you use WARN_ON here is try to keep things going?

What would be the alternative?

Thanks,

Ingo

#18 lore raw

[tip:core/locking] locking/stat: Fix a typo

In reply to Yuanhan Liu
Commit-ID:  5b43715b9d27e4e6620264113169bb8e4e607205
Gitweb:     http://git.kernel.org/tip/5b43715b9d27e4e6620264113169bb8e4e607205
Author:     Yuanhan Liu <[email protected]>
AuthorDate: Thu, 24 Jan 2013 17:22:44 +0800
Committer:  Ingo Molnar <[email protected]>
CommitDate: Thu, 24 Jan 2013 10:59:43 +0100

locking/stat: Fix a typo

s/STATS/STAT

---
 Documentation/lockstat.txt | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
Documentation/lockstat.txt +1 -1
index cef00d4..dd2f7b2 100644
@@ -65,7 +65,7 @@ that had to wait on lock acquisition.
 
  - CONFIGURATION
 
-Lock statistics are enabled via CONFIG_LOCK_STATS.
+Lock statistics are enabled via CONFIG_LOCK_STAT.
 
  - USAGE
 
#19 lore raw

[tip:core/locking] locking/stat: Fix a typo

In reply to Yuanhan Liu
locking/stat: Fix a typo
Commit-ID:  0be5c8ff58cf7a66019af2f1236daff731ed318c
Gitweb:     http://git.kernel.org/tip/0be5c8ff58cf7a66019af2f1236daff731ed318c
Author:     Yuanhan Liu <[email protected]>
AuthorDate: Thu, 24 Jan 2013 17:22:44 +0800
Committer:  Ingo Molnar <[email protected]>
CommitDate: Tue, 19 Feb 2013 08:42:37 +0100

locking/stat: Fix a typo

s/STATS/STAT

---
 Documentation/lockstat.txt | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
Documentation/lockstat.txt +1 -1
index cef00d4..dd2f7b2 100644
@@ -65,7 +65,7 @@ that had to wait on lock acquisition.
 
  - CONFIGURATION
 
-Lock statistics are enabled via CONFIG_LOCK_STATS.
+Lock statistics are enabled via CONFIG_LOCK_STAT.
 
  - USAGE