mm: strictly require elevated page refcount in isolate_lru_page()

isolate_lru_page() must be called only with stable reference to the page,
this is what is written in the comment above it, this is reasonable.

current isolate_lru_page() users and its page extra reference sources:

 mm/huge_memory.c:
  __collapse_huge_page_isolate()	- reference from pte

 mm/memcontrol.c:
  mem_cgroup_move_parent()		- get_page_unless_zero()
  mem_cgroup_move_charge_pte_range()	- reference from pte

 mm/memory-failure.c:
  soft_offline_page()			- fixed, reference from get_any_page()
  delete_from_lru_cache() - reference from caller or get_page_unless_zero()
	[ seems like there bug, because __memory_failure() can call
	  page_action() for hpages tail, but it is ok for
	  isolate_lru_page(), tail getted and not in lru]

 mm/memory_hotplug.c:
  do_migrate_range()			- fixed, get_page_unless_zero()

 mm/mempolicy.c:
  migrate_page_add()			- reference from pte

 mm/migrate.c:
  do_move_page_to_node_array()		- reference from follow_page()

 mlock.c:				- various external references

 mm/vmscan.c:
  putback_lru_page()			- reference from isolate_lru_page()

It seems that all isolate_lru_page() users are ready now for this
restriction.  So, let's replace redundant get_page_unless_zero() with
get_page() and add page initial reference count check with VM_BUG_ON()

Signed-off-by: Konstantin Khlebnikov <khlebnikov@openvz.org>
Cc: Andi Kleen <andi@firstfloor.org>
Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Cc: Mel Gorman <mel@csn.ul.ie>
Cc: Lee Schermerhorn <lee.schermerhorn@hp.com>
Cc: Rik van Riel <riel@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

diff --git a/mm/vmscan.c b/mm/vmscan.c
index d303b60f4c2abb309a85392e94b8a33a25ed73d0..890f54184d9a7e5647289dfafb5a858054a94567 100644 (file)
--- a/mm/vmscan.c
+++ b/mm/vmscan.c
@@ -1207,13 +1207,16 @@ int isolate_lru_page(struct page *page)
 {
        int ret = -EBUSY;
 
+       VM_BUG_ON(!page_count(page));
+
        if (PageLRU(page)) {
                struct zone *zone = page_zone(page);
 
                spin_lock_irq(&zone->lru_lock);
-               if (PageLRU(page) && get_page_unless_zero(page)) {
+               if (PageLRU(page)) {
                        int lru = page_lru(page);
                        ret = 0;
+                       get_page(page);
                        ClearPageLRU(page);
 
                        del_page_from_lru_list(zone, page, lru);