[Xen-devel] [PATCH 15/20] x86/ticketlock: prevent compiler reordering into locked region

[Jeremy Fitzhardinge <jeremy@xxxxxxxx>]

From: Jeremy Fitzhardinge <jeremy.fitzhardinge@xxxxxxxxxx>

Add a barrier() at the end of __raw_spin_unlock() to prevent instructions
from after the locked region from being reordered into it.  In theory doing
so shouldn't cause any problems, but in practice, the system locks up
under load...

Signed-off-by: Jeremy Fitzhardinge <jeremy.fitzhardinge@xxxxxxxxxx>
---
 arch/x86/include/asm/spinlock.h |    2 ++
 1 files changed, 2 insertions(+), 0 deletions(-)

diff --git a/arch/x86/include/asm/spinlock.h b/arch/x86/include/asm/spinlock.h
index d6de5e7..158b330 100644
--- a/arch/x86/include/asm/spinlock.h
+++ b/arch/x86/include/asm/spinlock.h
@@ -189,6 +189,8 @@ static __always_inline void 
arch_spin_unlock(arch_spinlock_t *lock)
        next = lock->tickets.head + 1;
        __ticket_unlock_release(lock);
        __ticket_unlock_kick(lock, next);
+
+       barrier();              /* prevent reordering into locked region */
 }
 
 static inline int arch_spin_is_locked(arch_spinlock_t *lock)
-- 
1.7.2.3


_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxxxxxxxx
http://lists.xensource.com/xen-devel