[PATCH v3 09/24] mm/balloon_compaction: remove dependency on page lock

["David Hildenbrand (Red Hat)" <david@kernel.org>]

Let's stop using the page lock in balloon code and instead use only the
balloon_device_lock.

As soon as we set the PG_movable_ops flag, we might now get isolation
callbacks for that page as we are no longer holding the page lock. In
there, we'll simply synchronize using the balloon_device_lock.

So in balloon_page_isolate() lookup the balloon_dev_info through
page->private under balloon_device_lock.

It's crucial that we update page->private under the balloon_device_lock,
so the isolation callback can properly deal with concurrent deflation.

Consequently, make sure that balloon_page_finalize() is called under
balloon_device_lock as we remove a page from the list and clear
page->private. balloon_page_insert() is already called with the
balloon_device_lock held.

Note that the core will still lock the pages, for example in
isolate_movable_ops_page(). The lock is there still relevant for
handling the PageMovableOpsIsolated flag, but that can be later changed
to use an atomic test-and-set instead, or moved into the movable_ops
backends.

Acked-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: David Hildenbrand (Red Hat) <david@kernel.org>
---
 include/linux/balloon_compaction.h | 25 ++++++++++----------
 mm/balloon_compaction.c            | 38 ++++++++++--------------------
 2 files changed, 25 insertions(+), 38 deletions(-)

diff --git a/include/linux/balloon_compaction.h b/include/linux/balloon_compaction.h
index 9a8568fcd477d..ad594af6ed100 100644
--- a/include/linux/balloon_compaction.h
+++ b/include/linux/balloon_compaction.h
@@ -12,25 +12,27 @@
  * is derived from the page type (PageOffline()) combined with the
  * PG_movable_ops flag (PageMovableOps()).
  *
+ * Once the page type and the PG_movable_ops are set, migration code
+ * can initiate page isolation by invoking the
+ * movable_operations()->isolate_page() callback
+ *
+ * As long as page->private is set, the page is either on the balloon list
+ * or isolated for migration. If page->private is not set, the page is
+ * either still getting inflated, or was deflated to be freed by the balloon
+ * driver soon. Isolation is impossible in both cases.
+ *
  * As the page isolation scanning step a compaction thread does is a lockless
  * procedure (from a page standpoint), it might bring some racy situations while
  * performing balloon page compaction. In order to sort out these racy scenarios
  * and safely perform balloon's page compaction and migration we must, always,
  * ensure following these simple rules:
  *
- *   i. Setting the PG_movable_ops flag and page->private with the following
- *	lock order
- *	    +-page_lock(page);
- *	      +--spin_lock_irq(&balloon_pages_lock);
+ *   i. Inflation/deflation must set/clear page->private under the
+ *      balloon_pages_lock
  *
  *  ii. isolation or dequeueing procedure must remove the page from balloon
  *      device page list under balloon_pages_lock
  *
- * The functions provided by this interface are placed to help on coping with
- * the aforementioned balloon page corner case, as well as to ensure the simple
- * set of exposed rules are satisfied while we are dealing with balloon pages
- * compaction / migration.
- *
  * Copyright (C) 2012, Red Hat, Inc.  Rafael Aquini <aquini@redhat.com>
  */
 #ifndef _LINUX_BALLOON_COMPACTION_H
@@ -93,8 +95,7 @@ static inline struct balloon_dev_info *balloon_page_device(struct page *page)
  * @balloon : pointer to balloon device
  * @page    : page to be assigned as a 'balloon page'
  *
- * Caller must ensure the page is locked and the spin_lock protecting balloon
- * pages list is held before inserting a page into the balloon device.
+ * Caller must ensure the balloon_pages_lock is held.
  */
 static inline void balloon_page_insert(struct balloon_dev_info *balloon,
 				       struct page *page)
@@ -119,7 +120,7 @@ static inline gfp_t balloon_mapping_gfp_mask(void)
  *			   balloon list for release to the page allocator
  * @page: page to be released to the page allocator
  *
- * Caller must ensure that the page is locked.
+ * Caller must ensure the balloon_pages_lock is held.
  */
 static inline void balloon_page_finalize(struct page *page)
 {
diff --git a/mm/balloon_compaction.c b/mm/balloon_compaction.c
index a0fd779bbd012..75763c73dbd52 100644
--- a/mm/balloon_compaction.c
+++ b/mm/balloon_compaction.c
@@ -20,15 +20,7 @@ static DEFINE_SPINLOCK(balloon_pages_lock);
 static void balloon_page_enqueue_one(struct balloon_dev_info *b_dev_info,
 				     struct page *page)
 {
-	/*
-	 * Block others from accessing the 'page' when we get around to
-	 * establishing additional references. We should be the only one
-	 * holding a reference to the 'page' at this point. If we are not, then
-	 * memory corruption is possible and we should stop execution.
-	 */
-	BUG_ON(!trylock_page(page));
 	balloon_page_insert(b_dev_info, page);
-	unlock_page(page);
 	if (b_dev_info->adjust_managed_page_count)
 		adjust_managed_page_count(page, -1);
 	__count_vm_event(BALLOON_INFLATE);
@@ -93,22 +85,12 @@ size_t balloon_page_list_dequeue(struct balloon_dev_info *b_dev_info,
 	list_for_each_entry_safe(page, tmp, &b_dev_info->pages, lru) {
 		if (n_pages == n_req_pages)
 			break;
-
-		/*
-		 * Block others from accessing the 'page' while we get around to
-		 * establishing additional references and preparing the 'page'
-		 * to be released by the balloon driver.
-		 */
-		if (!trylock_page(page))
-			continue;
-
 		list_del(&page->lru);
 		if (b_dev_info->adjust_managed_page_count)
 			adjust_managed_page_count(page, 1);
 		balloon_page_finalize(page);
 		__count_vm_event(BALLOON_DEFLATE);
 		list_add(&page->lru, pages);
-		unlock_page(page);
 		dec_node_page_state(page, NR_BALLOON_PAGES);
 		n_pages++;
 	}
@@ -213,13 +195,19 @@ EXPORT_SYMBOL_GPL(balloon_page_dequeue);
 static bool balloon_page_isolate(struct page *page, isolate_mode_t mode)
 
 {
-	struct balloon_dev_info *b_dev_info = balloon_page_device(page);
+	struct balloon_dev_info *b_dev_info;
 	unsigned long flags;
 
-	if (!b_dev_info)
-		return false;
-
 	spin_lock_irqsave(&balloon_pages_lock, flags);
+	b_dev_info = balloon_page_device(page);
+	if (!b_dev_info) {
+		/*
+		 * The page already got deflated and removed from the
+		 * balloon list.
+		 */
+		spin_unlock_irqrestore(&balloon_pages_lock, flags);
+		return false;
+	}
 	list_del(&page->lru);
 	b_dev_info->isolated_pages++;
 	spin_unlock_irqrestore(&balloon_pages_lock, flags);
@@ -253,9 +241,6 @@ static int balloon_page_migrate(struct page *newpage, struct page *page,
 	unsigned long flags;
 	int rc;
 
-	VM_BUG_ON_PAGE(!PageLocked(page), page);
-	VM_BUG_ON_PAGE(!PageLocked(newpage), newpage);
-
 	/*
 	 * When we isolated the page, the page was still inflated in a balloon
 	 * device. As isolated balloon pages cannot get deflated, we still have
@@ -293,10 +278,11 @@ static int balloon_page_migrate(struct page *newpage, struct page *page,
 	}
 
 	b_dev_info->isolated_pages--;
-	spin_unlock_irqrestore(&balloon_pages_lock, flags);
 
 	/* Free the now-deflated page we isolated in balloon_page_isolate(). */
 	balloon_page_finalize(page);
+	spin_unlock_irqrestore(&balloon_pages_lock, flags);
+
 	put_page(page);
 
 	return 0;
-- 
2.52.0