mm: fs: invalidate bh_lrus for only cold path

The kernel test robot reported the regression of fio.write_iops[1] with
commit 8cc621d2f45d ("mm: fs: invalidate BH LRU during page migration").

Since lru_add_drain is called frequently, invalidate bh_lrus there could
increase bh_lrus cache miss ratio, which needs more IO in the end.

This patch moves the bh_lrus invalidation from the hot path( e.g.,
zap_page_range, pagevec_release) to cold path(i.e., lru_add_drain_all,
lru_cache_disable).

Zhengjun Xing confirmed
 "I test the patch, the regression reduced to -2.9%"

[1] https://lore.kernel.org/lkml/20210520083144.GD14190@xsang-OptiPlex-9020/
[2] 8cc621d2f45d, mm: fs: invalidate BH LRU during page migration

Link: https://lkml.kernel.org/r/20210907212347.1977686-1-minchan@kernel.org
Signed-off-by: Minchan Kim <minchan@kernel.org>
Reported-by: kernel test robot <oliver.sang@intel.com>
Reviewed-by: Chris Goldsworthy <cgoldswo@codeaurora.org>
Tested-by: "Xing, Zhengjun" <zhengjun.xing@intel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
diff --git a/mm/swap.c b/mm/swap.c
index 897200d..af3cad4 100644
--- a/mm/swap.c
+++ b/mm/swap.c
@@ -620,7 +620,6 @@ void lru_add_drain_cpu(int cpu)
 		pagevec_lru_move_fn(pvec, lru_lazyfree_fn);
 
 	activate_page_drain(cpu);
-	invalidate_bh_lrus_cpu(cpu);
 }
 
 /**
@@ -703,6 +702,20 @@ void lru_add_drain(void)
 	local_unlock(&lru_pvecs.lock);
 }
 
+/*
+ * It's called from per-cpu workqueue context in SMP case so
+ * lru_add_drain_cpu and invalidate_bh_lrus_cpu should run on
+ * the same cpu. It shouldn't be a problem in !SMP case since
+ * the core is only one and the locks will disable preemption.
+ */
+static void lru_add_and_bh_lrus_drain(void)
+{
+	local_lock(&lru_pvecs.lock);
+	lru_add_drain_cpu(smp_processor_id());
+	local_unlock(&lru_pvecs.lock);
+	invalidate_bh_lrus_cpu();
+}
+
 void lru_add_drain_cpu_zone(struct zone *zone)
 {
 	local_lock(&lru_pvecs.lock);
@@ -717,7 +730,7 @@ static DEFINE_PER_CPU(struct work_struct, lru_add_drain_work);
 
 static void lru_add_drain_per_cpu(struct work_struct *dummy)
 {
-	lru_add_drain();
+	lru_add_and_bh_lrus_drain();
 }
 
 /*
@@ -858,7 +871,7 @@ void lru_cache_disable(void)
 	 */
 	__lru_add_drain_all(true);
 #else
-	lru_add_drain();
+	lru_add_and_bh_lrus_drain();
 #endif
 }