On x86, atomic_cond_read_relaxed will busy-wait with a cpu_relax() loop, so it is desirable to increase the number of times we spin on the qspinlock lockword when it is found to be transitioning from pending to locked.
According to Waiman Long: | Ideally, the spinning times should be at least a few times the typical | cacheline load time from memory which I think can be down to 100ns or | so for each cacheline load with the newest systems or up to several | hundreds ns for older systems. which in his benchmarking corresponded to 512 iterations. Cc: Peter Zijlstra <[email protected]> Cc: Ingo Molnar <[email protected]> Suggested-by: Waiman Long <[email protected]> Signed-off-by: Will Deacon <[email protected]> --- arch/x86/include/asm/qspinlock.h | 2 ++ 1 file changed, 2 insertions(+) diff --git a/arch/x86/include/asm/qspinlock.h b/arch/x86/include/asm/qspinlock.h index 5e16b5d40d32..2f09915f4aa4 100644 --- a/arch/x86/include/asm/qspinlock.h +++ b/arch/x86/include/asm/qspinlock.h @@ -7,6 +7,8 @@ #include <asm-generic/qspinlock_types.h> #include <asm/paravirt.h> +#define _Q_PENDING_LOOPS (1 << 9) + #define queued_spin_unlock queued_spin_unlock /** * queued_spin_unlock - release a queued spinlock -- 2.1.4
