From 3b92f2ca6f9fed90d050bb7ebb192bf39394a542 Mon Sep 17 00:00:00 2001
From: Peter Zijlstra <a.p.zijlstra@chello.nl>
Date: Fri, 3 Jul 2009 08:44:37 -0500
Subject: [PATCH] mm: page_alloc reduce lock sections further

commit 46167aec68f48cbbeff23cae9173bc4d19a7bcda in tip.

Split out the pages which are to be freed into a separate list and
call free_pages_bulk() outside of the percpu page allocator locks.

Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com>
---
 include/linux/list.h |    3 +++
 mm/page_alloc.c      |   44 ++++++++++++++++++++++++++++++++++++--------
 2 files changed, 39 insertions(+), 8 deletions(-)

diff --git a/include/linux/list.h b/include/linux/list.h
index 8392884..703cd07 100644
--- a/include/linux/list.h
+++ b/include/linux/list.h
@@ -359,6 +359,9 @@ static inline void list_splice_tail_init(struct list_head *list,
 #define list_first_entry(ptr, type, member) \
 	list_entry((ptr)->next, type, member)
 
+#define list_last_entry(ptr, type, member) \
+	list_entry((ptr)->prev, type, member)
+
 /**
  * list_for_each	-	iterate over a list
  * @pos:	the &struct list_head to use as a loop cursor.
diff --git a/mm/page_alloc.c b/mm/page_alloc.c
index 2eaf763..f54d527 100644
--- a/mm/page_alloc.c
+++ b/mm/page_alloc.c
@@ -600,8 +600,9 @@ static void free_pcppages_bulk(struct zone *zone, int count,
 	int migratetype = 0;
 	int batch_free = 0;
 	int to_free = count;
+	unsigned long flags;
 
-	spin_lock(&zone->lock);
+	spin_lock_irqsave(&zone->lock, flags);
 	zone->all_unreclaimable = 0;
 	zone->pages_scanned = 0;
 
@@ -630,22 +631,27 @@ static void free_pcppages_bulk(struct zone *zone, int count,
 			/* MIGRATE_MOVABLE list may include MIGRATE_RESERVEs */
 			__free_one_page(page, zone, 0, page_private(page));
 			trace_mm_page_pcpu_drain(page, 0, page_private(page));
+#ifdef CONFIG_PREEMPT_RT
+			cond_resched_lock(&zone->lock);
+#endif
 		} while (--to_free && --batch_free && !list_empty(list));
 	}
 	__mod_zone_page_state(zone, NR_FREE_PAGES, count);
-	spin_unlock(&zone->lock);
+	spin_unlock_irqrestore(&zone->lock, flags);
 }
 
 static void free_one_page(struct zone *zone, struct page *page, int order,
 				int migratetype)
 {
-	spin_lock(&zone->lock);
+	unsigned long flags;
+
+	spin_lock_irqsave(&zone->lock, flags);
 	zone->all_unreclaimable = 0;
 	zone->pages_scanned = 0;
 
 	__free_one_page(page, zone, order, migratetype);
 	__mod_zone_page_state(zone, NR_FREE_PAGES, 1 << order);
-	spin_unlock(&zone->lock);
+	spin_unlock_irqrestore(&zone->lock, flags);
 }
 
 static void __free_pages_ok(struct page *page, unsigned int order)
@@ -1040,6 +1046,16 @@ static int rmqueue_bulk(struct zone *zone, unsigned int order,
 	return i;
 }
 
+static void
+isolate_pcp_pages(int count, struct list_head *src, struct list_head *dst)
+{
+	while (count--) {
+		struct page *page = list_last_entry(src, struct page, lru);
+		list_move(&page->lru, dst);
+	}
+}
+
+
 #ifdef CONFIG_NUMA
 /*
  * Called from the vmstat counter updater to drain pagesets of this
@@ -1051,6 +1067,7 @@ static int rmqueue_bulk(struct zone *zone, unsigned int order,
  */
 void drain_zone_pages(struct zone *zone, struct per_cpu_pages *pcp)
 {
+	LIST_HEAD(free_list);
 	unsigned long flags;
 	int to_drain;
 	int this_cpu;
@@ -1060,9 +1077,10 @@ void drain_zone_pages(struct zone *zone, struct per_cpu_pages *pcp)
 		to_drain = pcp->batch;
 	else
 		to_drain = pcp->count;
-	free_pcppages_bulk(zone, to_drain, pcp);
+	isolate_pcp_pages(to_drain, &pcp->list, &free_list);
 	pcp->count -= to_drain;
 	unlock_cpu_pcp(flags, this_cpu);
+	free_pages_bulk(zone, to_drain, &free_list, 0);
 }
 #endif
 
@@ -1081,6 +1099,8 @@ static void drain_pages(unsigned int cpu)
 	for_each_populated_zone(zone) {
 		struct per_cpu_pageset *pset;
 		struct per_cpu_pages *pcp;
+		LIST_HEAD(free_list);
+		int count;
 
 		__lock_cpu_pcp(&flags, cpu);
 		pset = per_cpu_ptr(zone->pageset, cpu);
@@ -1091,9 +1111,11 @@ static void drain_pages(unsigned int cpu)
 			continue;
 		}
 		pcp = &pset->pcp;
-		free_pcppages_bulk(zone, pcp->count, pcp);
+		isolate_pcp_pages(pcp->count, &pcp->list, &free_list);
+		count = pcp->count;
 		pcp->count = 0;
 		unlock_cpu_pcp(flags, cpu);
+		free_pages_bulk(zone, count, &free_list, 0);
 	}
 }
 
@@ -1200,7 +1222,7 @@ void free_hot_cold_page(struct page *page, int cold)
 	struct per_cpu_pages *pcp;
 	unsigned long flags;
 	int migratetype;
-	int this_cpu, wasMlocked = __TestClearPageMlocked(page);
+	int count, this_cpu, wasMlocked = __TestClearPageMlocked(page);
 
 	trace_mm_page_free_direct(page, 0);
 	kmemcheck_free_shadow(page, 0);
@@ -1246,8 +1268,14 @@ void free_hot_cold_page(struct page *page, int cold)
 		list_add(&page->lru, &pcp->lists[migratetype]);
 	pcp->count++;
 	if (pcp->count >= pcp->high) {
-		free_pcppages_bulk(zone, pcp->batch, pcp);
+		LIST_HEAD(free_list);
+
+		isolate_pcp_pages(pcp->batch, &pcp->list, &free_list);
 		pcp->count -= pcp->batch;
+		count = pcp->batch;
+		put_zone_pcp(zone, flags, this_cpu);
+		free_pages_bulk(zone, count, &free_list, 0);
+		return;
 	}
 
 out:
-- 
1.7.0.4