[PATCH] Directed yield: cpu_relax variants for spinlocks and rw-locks

On systems running with virtual cpus there is optimization potential in
regard to spinlocks and rw-locks.  If the virtual cpu that has taken a lock
is known to a cpu that wants to acquire the same lock it is beneficial to
yield the timeslice of the virtual cpu in favour of the cpu that has the
lock (directed yield).

With CONFIG_PREEMPT="n" this can be implemented by the architecture without
common code changes.  Powerpc already does this.

With CONFIG_PREEMPT="y" the lock loops are coded with _raw_spin_trylock,
_raw_read_trylock and _raw_write_trylock in kernel/spinlock.c.  If the lock
could not be taken cpu_relax is called.  A directed yield is not possible
because cpu_relax doesn't know anything about the lock.  To be able to
yield the lock in favour of the current lock holder variants of cpu_relax
for spinlocks and rw-locks are needed.  The new _raw_spin_relax,
_raw_read_relax and _raw_write_relax primitives differ from cpu_relax
insofar that they have an argument: a pointer to the lock structure.

Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Haavard Skinnemoen <hskinnemoen@atmel.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

15 files changed, 60 insertions, 0 deletions

diff --git a/include/asm-alpha/spinlock.h b/include/asm-alpha/spinlock.h
index 0c294c9b0c55..aeeb125f6851 100644
--- a/include/asm-alpha/spinlock.h
+++ b/include/asm-alpha/spinlock.h
@@ -166,4 +166,8 @@ static inline void __raw_write_unlock(raw_rwlock_t * lock)
 	lock->lock = 0;
 }
 
+#define _raw_spin_relax(lock)	cpu_relax()
+#define _raw_read_relax(lock)	cpu_relax()
+#define _raw_write_relax(lock)	cpu_relax()
+
 #endif /* _ALPHA_SPINLOCK_H */
diff --git a/include/asm-arm/spinlock.h b/include/asm-arm/spinlock.h
index 01b7c26a3038..861092fbaa53 100644
--- a/include/asm-arm/spinlock.h
+++ b/include/asm-arm/spinlock.h
@@ -218,4 +218,8 @@ 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_spin_relax(lock)	cpu_relax()
+#define _raw_read_relax(lock)	cpu_relax()
+#define _raw_write_relax(lock)	cpu_relax()
+
 #endif /* __ASM_SPINLOCK_H */
diff --git a/include/asm-cris/arch-v32/spinlock.h b/include/asm-cris/arch-v32/spinlock.h
index 52df72a62232..5f43df0a5fb4 100644
--- a/include/asm-cris/arch-v32/spinlock.h
+++ b/include/asm-cris/arch-v32/spinlock.h
@@ -160,4 +160,8 @@ static __inline__ int is_write_locked(rwlock_t *rw)
 	return rw->counter < 0;
 }
 
+#define _raw_spin_relax(lock)	cpu_relax()
+#define _raw_read_relax(lock)	cpu_relax()
+#define _raw_write_relax(lock)	cpu_relax()
+
 #endif /* __ASM_ARCH_SPINLOCK_H */
diff --git a/include/asm-i386/spinlock.h b/include/asm-i386/spinlock.h
index b0b3043f05e1..c18b71fae6b3 100644
--- a/include/asm-i386/spinlock.h
+++ b/include/asm-i386/spinlock.h
@@ -205,4 +205,8 @@ static inline void __raw_write_unlock(raw_rwlock_t *rw)
 				 : "+m" (rw->lock) : : "memory");
 }
 
+#define _raw_spin_relax(lock)	cpu_relax()
+#define _raw_read_relax(lock)	cpu_relax()
+#define _raw_write_relax(lock)	cpu_relax()
+
 #endif /* __ASM_SPINLOCK_H */
diff --git a/include/asm-ia64/spinlock.h b/include/asm-ia64/spinlock.h
index 9e83210dc312..ff857e31738a 100644
--- a/include/asm-ia64/spinlock.h
+++ b/include/asm-ia64/spinlock.h
@@ -213,4 +213,8 @@ 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_spin_relax(lock)	cpu_relax()
+#define _raw_read_relax(lock)	cpu_relax()
+#define _raw_write_relax(lock)	cpu_relax()
+
 #endif /*  _ASM_IA64_SPINLOCK_H */
diff --git a/include/asm-m32r/spinlock.h b/include/asm-m32r/spinlock.h
index f9f90727a4a1..f5cfba81ee10 100644
--- a/include/asm-m32r/spinlock.h
+++ b/include/asm-m32r/spinlock.h
@@ -316,4 +316,8 @@ static inline int __raw_write_trylock(raw_rwlock_t *lock)
 	return 0;
 }
 
+#define _raw_spin_relax(lock)	cpu_relax()
+#define _raw_read_relax(lock)	cpu_relax()
+#define _raw_write_relax(lock)	cpu_relax()
+
 #endif	/* _ASM_M32R_SPINLOCK_H */
diff --git a/include/asm-mips/spinlock.h b/include/asm-mips/spinlock.h
index 4c1a1b53aeaf..c8d5587467bb 100644
--- a/include/asm-mips/spinlock.h
+++ b/include/asm-mips/spinlock.h
@@ -328,4 +328,8 @@ static inline int __raw_write_trylock(raw_rwlock_t *rw)
 }
 
 
+#define _raw_spin_relax(lock)	cpu_relax()
+#define _raw_read_relax(lock)	cpu_relax()
+#define _raw_write_relax(lock)	cpu_relax()
+
 #endif /* _ASM_SPINLOCK_H */
diff --git a/include/asm-parisc/spinlock.h b/include/asm-parisc/spinlock.h
index a93960e232cf..e1825530365d 100644
--- a/include/asm-parisc/spinlock.h
+++ b/include/asm-parisc/spinlock.h
@@ -152,4 +152,8 @@ static __inline__ int __raw_write_can_lock(raw_rwlock_t *rw)
 	return !rw->counter;
 }
 
+#define _raw_spin_relax(lock)	cpu_relax()
+#define _raw_read_relax(lock)	cpu_relax()
+#define _raw_write_relax(lock)	cpu_relax()
+
 #endif /* __ASM_SPINLOCK_H */
diff --git a/include/asm-powerpc/spinlock.h b/include/asm-powerpc/spinlock.h
index c31e4382a775..eaccac8327f2 100644
--- a/include/asm-powerpc/spinlock.h
+++ b/include/asm-powerpc/spinlock.h
@@ -285,5 +285,9 @@ static __inline__ void __raw_write_unlock(raw_rwlock_t *rw)
 	rw->lock = 0;
 }
 
+#define _raw_spin_relax(lock)	cpu_relax()
+#define _raw_read_relax(lock)	cpu_relax()
+#define _raw_write_relax(lock)	cpu_relax()
+
 #endif /* __KERNEL__ */
 #endif /* __ASM_SPINLOCK_H */
diff --git a/include/asm-ppc/spinlock.h b/include/asm-ppc/spinlock.h
index 5c64b75f0295..fccaf5531e57 100644
--- a/include/asm-ppc/spinlock.h
+++ b/include/asm-ppc/spinlock.h
@@ -161,4 +161,8 @@ static __inline__ void __raw_write_unlock(raw_rwlock_t *rw)
 	rw->lock = 0;
 }
 
+#define _raw_spin_relax(lock)	cpu_relax()
+#define _raw_read_relax(lock)	cpu_relax()
+#define _raw_write_relax(lock)	cpu_relax()
+
 #endif /* __ASM_SPINLOCK_H */
diff --git a/include/asm-s390/spinlock.h b/include/asm-s390/spinlock.h
index ce3edf6d63b3..5f00feaf1be6 100644
--- a/include/asm-s390/spinlock.h
+++ b/include/asm-s390/spinlock.h
@@ -154,4 +154,8 @@ static inline int __raw_write_trylock(raw_rwlock_t *rw)
 	return _raw_write_trylock_retry(rw);
 }
 
+#define _raw_spin_relax(lock)	cpu_relax()
+#define _raw_read_relax(lock)	cpu_relax()
+#define _raw_write_relax(lock)	cpu_relax()
+
 #endif /* __ASM_SPINLOCK_H */
diff --git a/include/asm-sh/spinlock.h b/include/asm-sh/spinlock.h
index 846322d4c35d..54458fd24981 100644
--- a/include/asm-sh/spinlock.h
+++ b/include/asm-sh/spinlock.h
@@ -100,4 +100,8 @@ static inline int __raw_write_trylock(raw_rwlock_t *rw)
 	return 0;
 }
 
+#define _raw_spin_relax(lock)	cpu_relax()
+#define _raw_read_relax(lock)	cpu_relax()
+#define _raw_write_relax(lock)	cpu_relax()
+
 #endif /* __ASM_SH_SPINLOCK_H */
diff --git a/include/asm-sparc/spinlock.h b/include/asm-sparc/spinlock.h
index 1c75474ba1df..557d08959d2f 100644
--- a/include/asm-sparc/spinlock.h
+++ b/include/asm-sparc/spinlock.h
@@ -154,6 +154,10 @@ static inline int __raw_write_trylock(raw_rwlock_t *rw)
 #define __raw_spin_lock_flags(lock, flags) __raw_spin_lock(lock)
 #define __raw_read_trylock(lock) generic__raw_read_trylock(lock)
 
+#define _raw_spin_relax(lock)	cpu_relax()
+#define _raw_read_relax(lock)	cpu_relax()
+#define _raw_write_relax(lock)	cpu_relax()
+
 #define __raw_read_can_lock(rw) (!((rw)->lock & 0xff))
 #define __raw_write_can_lock(rw) (!(rw)->lock)
 
diff --git a/include/asm-sparc64/spinlock.h b/include/asm-sparc64/spinlock.h
index bd5ffc76bc7e..0006fe9f8c7a 100644
--- a/include/asm-sparc64/spinlock.h
+++ b/include/asm-sparc64/spinlock.h
@@ -241,6 +241,10 @@ static int inline __write_trylock(raw_rwlock_t *lock)
 #define __raw_read_can_lock(rw)		(!((rw)->lock & 0x80000000UL))
 #define __raw_write_can_lock(rw)	(!(rw)->lock)
 
+#define _raw_spin_relax(lock)	cpu_relax()
+#define _raw_read_relax(lock)	cpu_relax()
+#define _raw_write_relax(lock)	cpu_relax()
+
 #endif /* !(__ASSEMBLY__) */
 
 #endif /* !(__SPARC64_SPINLOCK_H) */
diff --git a/include/asm-x86_64/spinlock.h b/include/asm-x86_64/spinlock.h
index 3daf5b005905..05ef097ba55b 100644
--- a/include/asm-x86_64/spinlock.h
+++ b/include/asm-x86_64/spinlock.h
@@ -133,4 +133,8 @@ static inline void __raw_write_unlock(raw_rwlock_t *rw)
 				: "=m" (rw->lock) : : "memory");
 }
 
+#define _raw_spin_relax(lock)	cpu_relax()
+#define _raw_read_relax(lock)	cpu_relax()
+#define _raw_write_relax(lock)	cpu_relax()
+
 #endif /* __ASM_SPINLOCK_H */