LKML Archive on lore.kernel.org
help / color / mirror / Atom feed
* [PATCH 0/2] enable irqs when waiting for rwlocks
@ 2008-10-22  8:33 Petr Tesarik
  2008-10-22  8:34 ` [PATCH 1/2] Allow rwlocks to re-enable interrupts Petr Tesarik
  2008-10-22 17:31 ` [PATCH 0/2] enable irqs when waiting for rwlocks Rick Jones
  0 siblings, 2 replies; 9+ messages in thread
From: Petr Tesarik @ 2008-10-22  8:33 UTC (permalink / raw)
  To: linux-kernel, linux-ia64; +Cc: tee

Hello,

SGI has observed that on large systems, interrupts are not serviced for
a long period of time when waiting for a rwlock. The following patch
series re-enables irqs while waiting for the lock, resembling the code
which is already there for spinlocks.

I only made the ia64 version, because the patch adds some overhead to
the fast path. In particular, I wasn't able to make a reasonably
lightweight implementation for x86, so I assume there are not such large
systems based on x86(_64). Of course, the possibility to implement
raw_{read|write}_lock_flags for any architecture is still there.

Petr Tesarik



^ permalink raw reply	[flat|nested] 9+ messages in thread

* [PATCH 1/2] Allow rwlocks to re-enable interrupts
  2008-10-22  8:33 [PATCH 0/2] enable irqs when waiting for rwlocks Petr Tesarik
@ 2008-10-22  8:34 ` Petr Tesarik
  2008-10-22  8:45   ` Peter Zijlstra
  2008-10-22 17:31 ` [PATCH 0/2] enable irqs when waiting for rwlocks Rick Jones
  1 sibling, 1 reply; 9+ messages in thread
From: Petr Tesarik @ 2008-10-22  8:34 UTC (permalink / raw)
  To: linux-kernel, linux-ia64; +Cc: tee

Pass the original flags to rwlock arch-code, so that it can re-enable
interrupts if implemented for that architecture.
    
Initially, make __raw_read_lock_flags and __raw_write_lock_flags
stubs which just do the same thing as their non-flags variants.

Signed-off-by: Petr Tesarik <ptesarik@suse.cz>
---
 arch/alpha/include/asm/spinlock.h    |    3 +++
 arch/arm/include/asm/spinlock.h      |    3 +++
 arch/ia64/include/asm/spinlock.h     |    3 +++
 arch/mips/include/asm/spinlock.h     |    2 ++
 arch/powerpc/include/asm/spinlock.h  |    3 +++
 arch/s390/include/asm/spinlock.h     |    3 +++
 arch/sh/include/asm/spinlock.h       |    3 +++
 arch/sparc/include/asm/spinlock_32.h |    2 ++
 arch/sparc/include/asm/spinlock_64.h |    2 ++
 include/asm-cris/arch-v32/spinlock.h |    2 ++
 include/asm-m32r/spinlock.h          |    3 +++
 include/asm-parisc/spinlock.h        |    3 +++
 include/asm-x86/spinlock.h           |    3 +++
 include/linux/spinlock.h             |    6 ++++++
 kernel/spinlock.c                    |    8 ++++++++
 15 files changed, 49 insertions(+), 0 deletions(-)

diff --git a/arch/alpha/include/asm/spinlock.h b/arch/alpha/include/asm/spinlock.h
index aeeb125..e38fb95 100644
--- a/arch/alpha/include/asm/spinlock.h
+++ b/arch/alpha/include/asm/spinlock.h
@@ -166,6 +166,9 @@ static inline void __raw_write_unlock(raw_rwlock_t * lock)
 	lock->lock = 0;
 }
 
+#define __raw_read_lock_flags(lock, flags) __raw_read_lock(lock)
+#define __raw_write_lock_flags(lock, flags) __raw_write_lock(lock)
+
 #define _raw_spin_relax(lock)	cpu_relax()
 #define _raw_read_relax(lock)	cpu_relax()
 #define _raw_write_relax(lock)	cpu_relax()
diff --git a/arch/arm/include/asm/spinlock.h b/arch/arm/include/asm/spinlock.h
index 2b41ebb..c13681a 100644
--- a/arch/arm/include/asm/spinlock.h
+++ b/arch/arm/include/asm/spinlock.h
@@ -217,6 +217,9 @@ static inline int __raw_read_trylock(raw_rwlock_t *rw)
 /* read_can_lock - would read_trylock() succeed? */
 #define __raw_read_can_lock(x)		((x)->lock < 0x80000000)
 
+#define __raw_read_lock_flags(lock, flags) __raw_read_lock(lock)
+#define __raw_write_lock_flags(lock, flags) __raw_write_lock(lock)
+
 #define _raw_spin_relax(lock)	cpu_relax()
 #define _raw_read_relax(lock)	cpu_relax()
 #define _raw_write_relax(lock)	cpu_relax()
diff --git a/arch/ia64/include/asm/spinlock.h b/arch/ia64/include/asm/spinlock.h
index 0229fb9..0a61961 100644
--- a/arch/ia64/include/asm/spinlock.h
+++ b/arch/ia64/include/asm/spinlock.h
@@ -213,6 +213,9 @@ static inline int __raw_read_trylock(raw_rwlock_t *x)
 	return (u32)ia64_cmpxchg4_acq((__u32 *)(x), new.word, old.word) == old.word;
 }
 
+#define __raw_read_lock_flags(lock, flags) __raw_read_lock(lock)
+#define __raw_write_lock_flags(lock, flags) __raw_write_lock(lock)
+
 #define _raw_spin_relax(lock)	cpu_relax()
 #define _raw_read_relax(lock)	cpu_relax()
 #define _raw_write_relax(lock)	cpu_relax()
diff --git a/arch/mips/include/asm/spinlock.h b/arch/mips/include/asm/spinlock.h
index 1a1f320..de03e47 100644
--- a/arch/mips/include/asm/spinlock.h
+++ b/arch/mips/include/asm/spinlock.h
@@ -479,6 +479,8 @@ static inline int __raw_write_trylock(raw_rwlock_t *rw)
 	return ret;
 }
 
+#define __raw_read_lock_flags(lock, flags) __raw_read_lock(lock)
+#define __raw_write_lock_flags(lock, flags) __raw_write_lock(lock)
 
 #define _raw_spin_relax(lock)	cpu_relax()
 #define _raw_read_relax(lock)	cpu_relax()
diff --git a/arch/powerpc/include/asm/spinlock.h b/arch/powerpc/include/asm/spinlock.h
index f56a843..f6491ae 100644
--- a/arch/powerpc/include/asm/spinlock.h
+++ b/arch/powerpc/include/asm/spinlock.h
@@ -287,6 +287,9 @@ static inline void __raw_write_unlock(raw_rwlock_t *rw)
 	rw->lock = 0;
 }
 
+#define __raw_read_lock_flags(lock, flags) __raw_read_lock(lock)
+#define __raw_write_lock_flags(lock, flags) __raw_write_lock(lock)
+
 #define _raw_spin_relax(lock)	__spin_yield(lock)
 #define _raw_read_relax(lock)	__rw_yield(lock)
 #define _raw_write_relax(lock)	__rw_yield(lock)
diff --git a/arch/s390/include/asm/spinlock.h b/arch/s390/include/asm/spinlock.h
index df84ae9..f3861b0 100644
--- a/arch/s390/include/asm/spinlock.h
+++ b/arch/s390/include/asm/spinlock.h
@@ -172,6 +172,9 @@ static inline int __raw_write_trylock(raw_rwlock_t *rw)
 	return _raw_write_trylock_retry(rw);
 }
 
+#define __raw_read_lock_flags(lock, flags) __raw_read_lock(lock)
+#define __raw_write_lock_flags(lock, flags) __raw_write_lock(lock)
+
 #define _raw_read_relax(lock)	cpu_relax()
 #define _raw_write_relax(lock)	cpu_relax()
 
diff --git a/arch/sh/include/asm/spinlock.h b/arch/sh/include/asm/spinlock.h
index e793181..6028356 100644
--- a/arch/sh/include/asm/spinlock.h
+++ b/arch/sh/include/asm/spinlock.h
@@ -216,6 +216,9 @@ static inline int __raw_write_trylock(raw_rwlock_t *rw)
 	return (oldval > (RW_LOCK_BIAS - 1));
 }
 
+#define __raw_read_lock_flags(lock, flags) __raw_read_lock(lock)
+#define __raw_write_lock_flags(lock, flags) __raw_write_lock(lock)
+
 #define _raw_spin_relax(lock)	cpu_relax()
 #define _raw_read_relax(lock)	cpu_relax()
 #define _raw_write_relax(lock)	cpu_relax()
diff --git a/arch/sparc/include/asm/spinlock_32.h b/arch/sparc/include/asm/spinlock_32.h
index bf2d532..46f91ab 100644
--- a/arch/sparc/include/asm/spinlock_32.h
+++ b/arch/sparc/include/asm/spinlock_32.h
@@ -177,6 +177,8 @@ static inline int __read_trylock(raw_rwlock_t *rw)
 #define __raw_write_unlock(rw)	do { (rw)->lock = 0; } while(0)
 
 #define __raw_spin_lock_flags(lock, flags) __raw_spin_lock(lock)
+#define __raw_read_lock_flags(rw, flags)   __raw_read_lock(rw)
+#define __raw_write_lock_flags(rw, flags)  __raw_write_lock(rw)
 
 #define _raw_spin_relax(lock)	cpu_relax()
 #define _raw_read_relax(lock)	cpu_relax()
diff --git a/arch/sparc/include/asm/spinlock_64.h b/arch/sparc/include/asm/spinlock_64.h
index 120cfe4..e856cd0 100644
--- a/arch/sparc/include/asm/spinlock_64.h
+++ b/arch/sparc/include/asm/spinlock_64.h
@@ -230,9 +230,11 @@ static int inline __write_trylock(raw_rwlock_t *lock)
 }
 
 #define __raw_read_lock(p)	__read_lock(p)
+#define __raw_read_lock_flags(p, f) __read_lock(p)
 #define __raw_read_trylock(p)	__read_trylock(p)
 #define __raw_read_unlock(p)	__read_unlock(p)
 #define __raw_write_lock(p)	__write_lock(p)
+#define __raw_write_lock_flags(p, f) __write_lock(p)
 #define __raw_write_unlock(p)	__write_unlock(p)
 #define __raw_write_trylock(p)	__write_trylock(p)
 
diff --git a/include/asm-cris/arch-v32/spinlock.h b/include/asm-cris/arch-v32/spinlock.h
index 0d5709b..129756b 100644
--- a/include/asm-cris/arch-v32/spinlock.h
+++ b/include/asm-cris/arch-v32/spinlock.h
@@ -121,6 +121,8 @@ static  inline int __raw_write_trylock(raw_rwlock_t *rw)
 	return 1;
 }
 
+#define _raw_read_lock_flags(lock, flags) _raw_read_lock(lock)
+#define _raw_write_lock_flags(lock, flags) _raw_write_lock(lock)
 
 #define _raw_spin_relax(lock)	cpu_relax()
 #define _raw_read_relax(lock)	cpu_relax()
diff --git a/include/asm-m32r/spinlock.h b/include/asm-m32r/spinlock.h
index f5cfba8..dded923 100644
--- a/include/asm-m32r/spinlock.h
+++ b/include/asm-m32r/spinlock.h
@@ -316,6 +316,9 @@ static inline int __raw_write_trylock(raw_rwlock_t *lock)
 	return 0;
 }
 
+#define __raw_read_lock_flags(lock, flags) __raw_read_lock(lock)
+#define __raw_write_lock_flags(lock, flags) __raw_write_lock(lock)
+
 #define _raw_spin_relax(lock)	cpu_relax()
 #define _raw_read_relax(lock)	cpu_relax()
 #define _raw_write_relax(lock)	cpu_relax()
diff --git a/include/asm-parisc/spinlock.h b/include/asm-parisc/spinlock.h
index f3d2090..fae03e1 100644
--- a/include/asm-parisc/spinlock.h
+++ b/include/asm-parisc/spinlock.h
@@ -187,6 +187,9 @@ static __inline__ int __raw_write_can_lock(raw_rwlock_t *rw)
 	return !rw->counter;
 }
 
+#define __raw_read_lock_flags(lock, flags) __raw_read_lock(lock)
+#define __raw_write_lock_flags(lock, flags) __raw_write_lock(lock)
+
 #define _raw_spin_relax(lock)	cpu_relax()
 #define _raw_read_relax(lock)	cpu_relax()
 #define _raw_write_relax(lock)	cpu_relax()
diff --git a/include/asm-x86/spinlock.h b/include/asm-x86/spinlock.h
index 157ff7f..132f5d3 100644
--- a/include/asm-x86/spinlock.h
+++ b/include/asm-x86/spinlock.h
@@ -357,6 +357,9 @@ static inline void __raw_write_unlock(raw_rwlock_t *rw)
 		     : "+m" (rw->lock) : "i" (RW_LOCK_BIAS) : "memory");
 }
 
+#define __raw_read_lock_flags(lock, flags) __raw_read_lock(lock)
+#define __raw_write_lock_flags(lock, flags) __raw_write_lock(lock)
+
 #define _raw_spin_relax(lock)	cpu_relax()
 #define _raw_read_relax(lock)	cpu_relax()
 #define _raw_write_relax(lock)	cpu_relax()
diff --git a/include/linux/spinlock.h b/include/linux/spinlock.h
index e0c0fcc..9e3fe36 100644
--- a/include/linux/spinlock.h
+++ b/include/linux/spinlock.h
@@ -148,9 +148,11 @@ do {								\
  extern int _raw_spin_trylock(spinlock_t *lock);
  extern void _raw_spin_unlock(spinlock_t *lock);
  extern void _raw_read_lock(rwlock_t *lock);
+#define _raw_read_lock_flags(lock, flags) _raw_read_lock(lock)
  extern int _raw_read_trylock(rwlock_t *lock);
  extern void _raw_read_unlock(rwlock_t *lock);
  extern void _raw_write_lock(rwlock_t *lock);
+#define _raw_write_lock_flags(lock, flags) _raw_write_lock(lock)
  extern int _raw_write_trylock(rwlock_t *lock);
  extern void _raw_write_unlock(rwlock_t *lock);
 #else
@@ -160,9 +162,13 @@ do {								\
 # define _raw_spin_trylock(lock)	__raw_spin_trylock(&(lock)->raw_lock)
 # define _raw_spin_unlock(lock)		__raw_spin_unlock(&(lock)->raw_lock)
 # define _raw_read_lock(rwlock)		__raw_read_lock(&(rwlock)->raw_lock)
+# define _raw_read_lock_flags(lock, flags) \
+		__raw_read_lock_flags(&(lock)->raw_lock, *(flags))
 # define _raw_read_trylock(rwlock)	__raw_read_trylock(&(rwlock)->raw_lock)
 # define _raw_read_unlock(rwlock)	__raw_read_unlock(&(rwlock)->raw_lock)
 # define _raw_write_lock(rwlock)	__raw_write_lock(&(rwlock)->raw_lock)
+# define _raw_write_lock_flags(lock, flags) \
+		__raw_write_lock_flags(&(lock)->raw_lock, *(flags))
 # define _raw_write_trylock(rwlock)	__raw_write_trylock(&(rwlock)->raw_lock)
 # define _raw_write_unlock(rwlock)	__raw_write_unlock(&(rwlock)->raw_lock)
 #endif
diff --git a/kernel/spinlock.c b/kernel/spinlock.c
index 29ab207..f769d8a 100644
--- a/kernel/spinlock.c
+++ b/kernel/spinlock.c
@@ -121,7 +121,11 @@ unsigned long __lockfunc _read_lock_irqsave(rwlock_t *lock)
 	local_irq_save(flags);
 	preempt_disable();
 	rwlock_acquire_read(&lock->dep_map, 0, 0, _RET_IP_);
+#ifdef CONFIG_LOCKDEP
 	LOCK_CONTENDED(lock, _raw_read_trylock, _raw_read_lock);
+#else
+	_raw_read_lock_flags(lock, &flags);
+#endif
 	return flags;
 }
 EXPORT_SYMBOL(_read_lock_irqsave);
@@ -151,7 +155,11 @@ unsigned long __lockfunc _write_lock_irqsave(rwlock_t *lock)
 	local_irq_save(flags);
 	preempt_disable();
 	rwlock_acquire(&lock->dep_map, 0, 0, _RET_IP_);
+#ifdef CONFIG_LOCKDEP
 	LOCK_CONTENDED(lock, _raw_write_trylock, _raw_write_lock);
+#else
+	_raw_write_lock_flags(lock, &flags);
+#endif
 	return flags;
 }
 EXPORT_SYMBOL(_write_lock_irqsave);



^ permalink raw reply	[flat|nested] 9+ messages in thread

* Re: [PATCH 1/2] Allow rwlocks to re-enable interrupts
  2008-10-22  8:34 ` [PATCH 1/2] Allow rwlocks to re-enable interrupts Petr Tesarik
@ 2008-10-22  8:45   ` Peter Zijlstra
  2008-10-22  8:58     ` Petr Tesarik
  0 siblings, 1 reply; 9+ messages in thread
From: Peter Zijlstra @ 2008-10-22  8:45 UTC (permalink / raw)
  To: Petr Tesarik; +Cc: linux-kernel, linux-ia64, tee, Ingo Molnar

On Wed, 2008-10-22 at 10:34 +0200, Petr Tesarik wrote:
> c b/kernel/spinlock.c
> index 29ab207..f769d8a 100644
> --- a/kernel/spinlock.c
> +++ b/kernel/spinlock.c
> @@ -121,7 +121,11 @@ unsigned long __lockfunc _read_lock_irqsave(rwlock_t *lock)
>  	local_irq_save(flags);
>  	preempt_disable();
>  	rwlock_acquire_read(&lock->dep_map, 0, 0, _RET_IP_);
> +#ifdef CONFIG_LOCKDEP
>  	LOCK_CONTENDED(lock, _raw_read_trylock, _raw_read_lock);
> +#else
> +	_raw_read_lock_flags(lock, &flags);
> +#endif
>  	return flags;
>  }

That should be CONFIG_LOCK_STAT.

But aside from that, I really don't like this change, I'd rather you'd
create a LOCK_CONTENDED_FLAGS() that can deal with this.



^ permalink raw reply	[flat|nested] 9+ messages in thread

* Re: [PATCH 1/2] Allow rwlocks to re-enable interrupts
  2008-10-22  8:45   ` Peter Zijlstra
@ 2008-10-22  8:58     ` Petr Tesarik
  2008-10-22 17:24       ` Peter Zijlstra
  0 siblings, 1 reply; 9+ messages in thread
From: Petr Tesarik @ 2008-10-22  8:58 UTC (permalink / raw)
  To: Peter Zijlstra; +Cc: linux-kernel, linux-ia64, tee, Ingo Molnar

On Wed, 2008-10-22 at 10:45 +0200, Peter Zijlstra wrote:
> On Wed, 2008-10-22 at 10:34 +0200, Petr Tesarik wrote:
> > c b/kernel/spinlock.c
> > index 29ab207..f769d8a 100644
> > --- a/kernel/spinlock.c
> > +++ b/kernel/spinlock.c
> > @@ -121,7 +121,11 @@ unsigned long __lockfunc _read_lock_irqsave(rwlock_t *lock)
> >  	local_irq_save(flags);
> >  	preempt_disable();
> >  	rwlock_acquire_read(&lock->dep_map, 0, 0, _RET_IP_);
> > +#ifdef CONFIG_LOCKDEP
> >  	LOCK_CONTENDED(lock, _raw_read_trylock, _raw_read_lock);
> > +#else
> > +	_raw_read_lock_flags(lock, &flags);
> > +#endif
> >  	return flags;
> >  }
> 
> That should be CONFIG_LOCK_STAT.

Ugh. Fine with me, but is CONFIG_LOCKDEP correct in _spin_lock_irqsave,
or should it also read CONFIG_LOCK_STAT?

> But aside from that, I really don't like this change, I'd rather you'd
> create a LOCK_CONTENDED_FLAGS() that can deal with this.

No problem. I could then also use it for _spin_lock_irqsave, if the
answer to the above question is use CONFIG_LOCK_STAT there as well.

Petr Tesarik



^ permalink raw reply	[flat|nested] 9+ messages in thread

* Re: [PATCH 1/2] Allow rwlocks to re-enable interrupts
  2008-10-22  8:58     ` Petr Tesarik
@ 2008-10-22 17:24       ` Peter Zijlstra
  2008-10-22 19:05         ` Matthew Wilcox
  0 siblings, 1 reply; 9+ messages in thread
From: Peter Zijlstra @ 2008-10-22 17:24 UTC (permalink / raw)
  To: Petr Tesarik; +Cc: linux-kernel, linux-ia64, tee, Ingo Molnar

On Wed, 2008-10-22 at 10:58 +0200, Petr Tesarik wrote:
> On Wed, 2008-10-22 at 10:45 +0200, Peter Zijlstra wrote:
> > On Wed, 2008-10-22 at 10:34 +0200, Petr Tesarik wrote:
> > > c b/kernel/spinlock.c
> > > index 29ab207..f769d8a 100644
> > > --- a/kernel/spinlock.c
> > > +++ b/kernel/spinlock.c
> > > @@ -121,7 +121,11 @@ unsigned long __lockfunc _read_lock_irqsave(rwlock_t *lock)
> > >  	local_irq_save(flags);
> > >  	preempt_disable();
> > >  	rwlock_acquire_read(&lock->dep_map, 0, 0, _RET_IP_);
> > > +#ifdef CONFIG_LOCKDEP
> > >  	LOCK_CONTENDED(lock, _raw_read_trylock, _raw_read_lock);
> > > +#else
> > > +	_raw_read_lock_flags(lock, &flags);
> > > +#endif
> > >  	return flags;
> > >  }
> > 
> > That should be CONFIG_LOCK_STAT.
> 
> Ugh. Fine with me, but is CONFIG_LOCKDEP correct in _spin_lock_irqsave,
> or should it also read CONFIG_LOCK_STAT?

Yep.

> > But aside from that, I really don't like this change, I'd rather you'd
> > create a LOCK_CONTENDED_FLAGS() that can deal with this.
> 
> No problem. I could then also use it for _spin_lock_irqsave, if the
> answer to the above question is use CONFIG_LOCK_STAT there as well.

If you create LOCK_CONTEDED_FLAGS() the whole issue goes away nicely.

^ permalink raw reply	[flat|nested] 9+ messages in thread

* Re: [PATCH 0/2] enable irqs when waiting for rwlocks
  2008-10-22  8:33 [PATCH 0/2] enable irqs when waiting for rwlocks Petr Tesarik
  2008-10-22  8:34 ` [PATCH 1/2] Allow rwlocks to re-enable interrupts Petr Tesarik
@ 2008-10-22 17:31 ` Rick Jones
  2008-10-23 14:12   ` Robin Holt
  1 sibling, 1 reply; 9+ messages in thread
From: Rick Jones @ 2008-10-22 17:31 UTC (permalink / raw)
  To: Petr Tesarik; +Cc: linux-kernel, linux-ia64, tee

Petr Tesarik wrote:
> Hello,
> 
> SGI has observed that on large systems, interrupts are not serviced for
> a long period of time when waiting for a rwlock. The following patch
> series re-enables irqs while waiting for the lock, resembling the code
> which is already there for spinlocks.

Perhaps I'm just out in left field, but that (and the similar behaviour 
for obtaining a spinlock?) feels like treating a symptom rather than a 
root cause where the root cause would appear to be long lock hold 
times/contention?

rick jones

^ permalink raw reply	[flat|nested] 9+ messages in thread

* Re: [PATCH 1/2] Allow rwlocks to re-enable interrupts
  2008-10-22 17:24       ` Peter Zijlstra
@ 2008-10-22 19:05         ` Matthew Wilcox
  2008-10-22 19:19           ` Peter Zijlstra
  0 siblings, 1 reply; 9+ messages in thread
From: Matthew Wilcox @ 2008-10-22 19:05 UTC (permalink / raw)
  To: Peter Zijlstra; +Cc: Petr Tesarik, linux-kernel, linux-ia64, tee, Ingo Molnar

On Wed, Oct 22, 2008 at 07:24:31PM +0200, Peter Zijlstra wrote:
> > No problem. I could then also use it for _spin_lock_irqsave, if the
> > answer to the above question is use CONFIG_LOCK_STAT there as well.
> 
> If you create LOCK_CONTEDED_FLAGS() the whole issue goes away nicely.

Should it also be used for _spin_lock_irq()?  I'm puzzled why it's only
used for _irqsave().

(should _spin_lock_bh() re-enable BHs while waiting?  Is it just not big
enough of a deal?)

-- 
Matthew Wilcox				Intel Open Source Technology Centre
"Bill, look, we understand that you're interested in selling us this
operating system, but compare it to ours.  We can't possibly take such
a retrograde step."

^ permalink raw reply	[flat|nested] 9+ messages in thread

* Re: [PATCH 1/2] Allow rwlocks to re-enable interrupts
  2008-10-22 19:05         ` Matthew Wilcox
@ 2008-10-22 19:19           ` Peter Zijlstra
  0 siblings, 0 replies; 9+ messages in thread
From: Peter Zijlstra @ 2008-10-22 19:19 UTC (permalink / raw)
  To: Matthew Wilcox; +Cc: Petr Tesarik, linux-kernel, linux-ia64, tee, Ingo Molnar

On Wed, 2008-10-22 at 13:05 -0600, Matthew Wilcox wrote:
> On Wed, Oct 22, 2008 at 07:24:31PM +0200, Peter Zijlstra wrote:
> > > No problem. I could then also use it for _spin_lock_irqsave, if the
> > > answer to the above question is use CONFIG_LOCK_STAT there as well.
> > 
> > If you create LOCK_CONTEDED_FLAGS() the whole issue goes away nicely.

Gah, I looked at it again, and that #ifdef isn't only to select between
LOCK_CONTENDED and not, but we can't actually have the re-enable for
anything lockdep.

So I was wrong.

> Should it also be used for _spin_lock_irq()?  I'm puzzled why it's only
> used for _irqsave().

Right, not sure how this maze is done.

The thing is, with spin_lock_irq() you know the irq state and can do the
enable unconditionally - then again, with ticket locks we cannot do it
at all.

The _flags() version needs the flags to see if irqs was enabled before
we entered the op, if it wasn't we cannot go around enabling them.

> (should _spin_lock_bh() re-enable BHs while waiting?  Is it just not big
> enough of a deal?)

Doubt it.. dunno.. personally I'd rather see softirqs die sooner rather
than later.

^ permalink raw reply	[flat|nested] 9+ messages in thread

* Re: [PATCH 0/2] enable irqs when waiting for rwlocks
  2008-10-22 17:31 ` [PATCH 0/2] enable irqs when waiting for rwlocks Rick Jones
@ 2008-10-23 14:12   ` Robin Holt
  0 siblings, 0 replies; 9+ messages in thread
From: Robin Holt @ 2008-10-23 14:12 UTC (permalink / raw)
  To: Rick Jones; +Cc: Petr Tesarik, linux-kernel, linux-ia64, tee

On Wed, Oct 22, 2008 at 10:31:53AM -0700, Rick Jones wrote:
> Petr Tesarik wrote:
>> Hello,
>>
>> SGI has observed that on large systems, interrupts are not serviced for
>> a long period of time when waiting for a rwlock. The following patch
>> series re-enables irqs while waiting for the lock, resembling the code
>> which is already there for spinlocks.
>
> Perhaps I'm just out in left field, but that (and the similar behaviour  
> for obtaining a spinlock?) feels like treating a symptom rather than a  
> root cause where the root cause would appear to be long lock hold  
> times/contention?

Sometimes lock contention on large systems will take a few seconds to
pass.  This is normal behavior which simply can not be eliminated.  For
those cases, we need to rely upon being able to re-enable interrupts and
allowing other operations to continue normally.  Simply put, in some
cases, nothing more can be done.

Thanks,
Robin

^ permalink raw reply	[flat|nested] 9+ messages in thread

end of thread, other threads:[~2008-10-23 14:12 UTC | newest]

Thread overview: 9+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2008-10-22  8:33 [PATCH 0/2] enable irqs when waiting for rwlocks Petr Tesarik
2008-10-22  8:34 ` [PATCH 1/2] Allow rwlocks to re-enable interrupts Petr Tesarik
2008-10-22  8:45   ` Peter Zijlstra
2008-10-22  8:58     ` Petr Tesarik
2008-10-22 17:24       ` Peter Zijlstra
2008-10-22 19:05         ` Matthew Wilcox
2008-10-22 19:19           ` Peter Zijlstra
2008-10-22 17:31 ` [PATCH 0/2] enable irqs when waiting for rwlocks Rick Jones
2008-10-23 14:12   ` Robin Holt

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).