[PATCH] powerpc: During context switch, check before setting mm_cpumask

From: Anton Blanchard <redacted>

During context switch, switch_mm() sets our current CPU in mm_cpumask.
We can avoid this atomic sequence in most cases by checking before
setting the bit.

Testing on a POWER8 using our context switch microbenchmark:

tools/testing/selftests/powerpc/benchmarks/context_switch \
	--process --no-fp --no-altivec --no-vector

Performance improves 2%.

Signed-off-by: Anton Blanchard <redacted>
---
 arch/powerpc/include/asm/mmu_context.h | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/arch/powerpc/include/asm/mmu_context.h b/arch/powerpc/include/asm/mmu_context.h
index 475d1be..5c45114 100644
--- a/arch/powerpc/include/asm/mmu_context.h
+++ b/arch/powerpc/include/asm/mmu_context.h

@@ -72,7 +72,8 @@ static inline void switch_mm(struct mm_struct *prev, struct mm_struct *next,
 			     struct task_struct *tsk)
 {
 	/* Mark this context has been used on the new CPU */
-	cpumask_set_cpu(smp_processor_id(), mm_cpumask(next));
+	if (!cpumask_test_cpu(smp_processor_id(), mm_cpumask(next)))
+		cpumask_set_cpu(smp_processor_id(), mm_cpumask(next));
 
 	/* 32-bit keeps track of the current PGDIR in the thread struct */
 #ifdef CONFIG_PPC32

-- 
2.7.4

On 03/10/16 17:40, Anton Blanchard wrote:

I think this makes sense, in fact I think in the longer term we can
even use __set_bit() reorder-able version since we have a sync
coming out of schedule(). The read side for TLB flush can use a RMB

Acked-by: Balbir Singh <bsingharora@gmail.com>

Balbir Singh.

On Tue, 2016-10-04 at 10:25 +1100, Balbir Singh wrote:

I think this makes sense, in fact I think in the longer term we can
even use __set_bit() reorder-able version since we have a sync
coming out of schedule(). The read side for TLB flush can use a RMB

No, that wouldn't be atomic vs. other threads accessing the same
bitmap.

Ben.

On 04/10/16 10:58, Benjamin Herrenschmidt wrote:

On Tue, 2016-10-04 at 10:25 +1100, Balbir Singh wrote:

quoted

I think this makes sense, in fact I think in the longer term we can
even use __set_bit() reorder-able version since we have a sync
coming out of schedule(). The read side for TLB flush can use a RMB

No, that wouldn't be atomic vs. other threads accessing the same
bitmap.

Somethings distorted my thought process. Thanks!

Balbir Singh.

On Mon, 2016-03-10 at 06:40:29 UTC, Anton Blanchard wrote:

From: Anton Blanchard <redacted>

During context switch, switch_mm() sets our current CPU in mm_cpumask.
We can avoid this atomic sequence in most cases by checking before
setting the bit.

Testing on a POWER8 using our context switch microbenchmark:

tools/testing/selftests/powerpc/benchmarks/context_switch \
	--process --no-fp --no-altivec --no-vector

Performance improves 2%.

Signed-off-by: Anton Blanchard <redacted>
Acked-by: Balbir Singh <bsingharora@gmail.com>

Applied to powerpc next, thanks.

https://git.kernel.org/powerpc/c/bb85fb5803270c52863b983596c2a0

cheers