[PATCH v6 11/11] powerpc/mm/book3s64/pgtable: Uses counting method to skip serializing

[Leonardo Bras]

For each cpu in cpumask, checks if it's running a lockless pagetable
walk. Then, run serialize_against_pte_lookup() only on these cpus.

serialize_agains_pte_lookup() can take a long while when there are a
lot of cpus in cpumask.

This method is intended to reduce this waiting, while not impacting
too much on the lockless pagetable walk.

Signed-off-by: Leonardo Bras <leonardo@xxxxxxxxxxxxx>
---
 arch/powerpc/mm/book3s64/pgtable.c | 16 +++++++++++++++-
 1 file changed, 15 insertions(+), 1 deletion(-)

diff --git a/arch/powerpc/mm/book3s64/pgtable.c b/arch/powerpc/mm/book3s64/pgtable.c
index bb138b628f86..4822ff1aac4b 100644
--- a/arch/powerpc/mm/book3s64/pgtable.c
+++ b/arch/powerpc/mm/book3s64/pgtable.c
@@ -96,8 +96,22 @@ static DEFINE_PER_CPU(int, lockless_pgtbl_walk_counter);
  */
 void serialize_against_pte_lookup(struct mm_struct *mm)
 {
+	int cpu;
+	struct cpumask cm;
+
 	smp_mb();
-	smp_call_function_many(mm_cpumask(mm), do_nothing, NULL, 1);
+
+	/*
+	 * Fills a new cpumask only with cpus that are currently doing a
+	 * lockless pagetable walk. This reduces time spent in this function.
+	 */
+	cpumask_clear(&cm);
+	for_each_cpu(cpu, mm_cpumask((mm))) {
+		if (per_cpu(lockless_pgtbl_walk_counter, cpu) > 0)
+			cpumask_set_cpu(cpu, &cm);
+	}
+
+	smp_call_function_many(&cm, do_nothing, NULL, 1);
 }
 
 /* begin_lockless_pgtbl_walk: Must be inserted before a function call that does
-- 
2.24.1