free_area[] bitmap elimination [3/3]

diff -puN mm/page_alloc.c~eliminate-bitmap-free mm/page_alloc.c
@@ -157,6 +157,37 @@ static void destroy_compound_page(struct
 #endif		/* CONFIG_HUGETLB_PAGE */
 
 /*
+ * This function checks whether a page is free && is the buddy
+ * we can do coalesce if
+ * (a) the buddy is free and
+ * (b) the buddy is on the buddy system
+ * (c) the buddy has the same order.
+ * for recording page's order, we use private field and PG_private.
+ */
+static inline int page_is_buddy(struct page *page, int order)
+{
+	if (page_count(page) == 0 &&
+	    PagePrivate(page) &&
+	    !PageReserved(page) &&
+            page_order(page) == order) {
+		/* check, check... see free_pages_check() */
+		if (page_mapped(page) ||
+		    page->mapping != NULL ||
+		    (page->flags & (
+			    1 << PG_lru	|
+			    1 << PG_locked	|
+			    1 << PG_active	|
+			    1 << PG_reclaim	|
+			    1 << PG_slab	|
+			    1 << PG_swapcache |
+			    1 << PG_writeback )))
+			bad_page(__FUNCTION__, page);
+		return 1;
+	}
+	return 0;
+}
+
+/*
  * Freeing function for a buddy system allocator.
  *
  * The concept of a buddy system is to maintain direct-mapped table

@@ -168,9 +199,12 @@ static void destroy_compound_page(struct
  * at the bottom level available, and propagating the changes upward
  * as necessary, plus some accounting needed to play nicely with other
  * parts of the VM system.
- * At each level, we keep one bit for each pair of blocks, which
- * is set to 1 iff only one of the pair is allocated.  So when we
- * are allocating or freeing one, we can derive the state of the
+ *
+ * At each level, we keep a list of pages, which are head of chunk of
+ * pages at the level. A page, which is a head of chunks, has its order
+ * in page structure itself and PG_private flag is set. we can get an 
+ * order of a page by calling  page_order().
+ * So we are allocating or freeing one, we can derive the state of the
  * other.  That is, if we allocate a small block, and both were   
  * free, the remainder of the region must be split into blocks.   
  * If a block is freed, and its buddy is also free, then this

@@ -178,43 +212,53 @@ static void destroy_compound_page(struct
  *
  * -- wli
  */
-
 static inline void __free_pages_bulk (struct page *page, struct page *base,
 		struct zone *zone, struct free_area *area, unsigned int order)
 {
-	unsigned long page_idx, index, mask;
-
+	unsigned long page_idx, mask;
 	if (order)
 		destroy_compound_page(page, order);
 	mask = (~0UL) << order;
 	page_idx = page - base;
 	if (page_idx & ~mask)
 		BUG();
-	index = page_idx >> (1 + order);
-
 	zone->free_pages += 1 << order;
-	while (order < MAX_ORDER-1) {
-		struct page *buddy1, *buddy2;
+	BUG_ON(bad_range(zone,page));
 
+	while (order < MAX_ORDER-1) {
+		struct page *buddy1;
 		BUG_ON(area >= zone->free_area + MAX_ORDER);
-		if (!__test_and_change_bit(index, area->map))
+		buddy1 = base + (page_idx ^ (1 << order));
+		if (order >= zone->aligned_order) {
+			/* we need range check */
+#ifdef CONFIG_VIRTUAL_MEM_MAP  
+			/* This check is necessary when
+			   1. there may be holes in zone.
+			   2. a hole is not aligned in this order.
+			   currently, VIRTUAL_MEM_MAP case, is only case.
+			   Is there better call than pfn_valid ?
+			*/
+			if (!pfn_valid(zone->zone_start_pfn + (page_idx ^ (1 << order))))
+				break;
+#endif		
+			/* this order in this zone is not aligned. */
+			if (bad_range(zone, buddy1))
+				break;
+		}
+		if (!page_is_buddy(buddy1, order))
 			/*
-			 * the buddy page is still allocated.
+			 *  the buddy page is still allocated.
 			 */
 			break;
-
-		/* Move the buddy up one level. */
-		buddy1 = base + (page_idx ^ (1 << order));
-		buddy2 = base + page_idx;
-		BUG_ON(bad_range(zone, buddy1));
-		BUG_ON(bad_range(zone, buddy2));
 		list_del(&buddy1->lru);
+		invalidate_page_order(buddy1);
 		mask <<= 1;
 		order++;
 		area++;
-		index >>= 1;
 		page_idx &= mask;
 	}
+	/* record the final order of the page */
+	set_page_order((base + page_idx), order);
 	list_add(&(base + page_idx)->lru, &area->free_list);
 }