mm, page_alloc: prevent infinite loop in buffered_rmqueue()

In DEBUG_VM kernel, we can hit infinite loop for order == 0 in
buffered_rmqueue() when check_new_pcp() returns 1, because the bad page
is never removed from the pcp list.  Fix this by removing the page
before retrying.  Also we don't need to check if page is non-NULL,
because we simply grab it from the list which was just tested for being
non-empty.

Fixes: 479f854 ("mm, page_alloc: defer debugging checks of pages allocated from the PCP")
Link: http://lkml.kernel.org/r/20160530090154.GM2527@techsingularity.net
Signed-off-by: Vlastimil Babka <vbabka@suse.cz>
Signed-off-by: Mel Gorman <mgorman@techsingularity.net>
Reported-by: Naoya Horiguchi <n-horiguchi@ah.jp.nec.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm/page_alloc.c

@@ -2615,11 +2615,12 @@ struct page *buffered_rmqueue(struct zone *preferred_zone,
 				page = list_last_entry(list, struct page, lru);
 			else
 				page = list_first_entry(list, struct page, lru);
-		} while (page && check_new_pcp(page));
 
-		__dec_zone_state(zone, NR_ALLOC_BATCH);
-		list_del(&page->lru);
-		pcp->count--;
+			__dec_zone_state(zone, NR_ALLOC_BATCH);
+			list_del(&page->lru);
+			pcp->count--;
+
+		} while (check_new_pcp(page));
 	} else {
 		/*
 		 * We most definitely don't want callers attempting to