--- /dev/null
+diff -uprN linux-2.6.28/block/genhd.c linux-2.6.28.new/block/genhd.c
+--- linux-2.6.28/block/genhd.c 2011-03-13 15:15:43.815647000 +0100
++++ linux-2.6.28.new/block/genhd.c 2011-03-15 10:13:35.145764805 +0100
+@@ -1129,6 +1129,8 @@ struct gendisk *alloc_disk_node(int mino
+ disk->part_tbl->part[0] = &disk->part0;
+
+ disk->minors = minors;
++ disk->flags |=
++ (GENHD_FL_REMAP_SWAPPED_PAGES | GENHD_FL_NOTIFY_REMAPPED_ONLY);
+ rand_initialize_disk(disk);
+ disk_to_dev(disk)->class = &block_class;
+ disk_to_dev(disk)->type = &disk_type;
+diff -uprN linux-2.6.28/include/linux/blkdev.h linux-2.6.28.new/include/linux/blkdev.h
+--- linux-2.6.28/include/linux/blkdev.h 2011-02-01 09:54:54.519982520 +0100
++++ linux-2.6.28.new/include/linux/blkdev.h 2011-02-01 10:15:39.369903561 +0100
+@@ -1068,6 +1068,8 @@ struct block_device_operations {
+ int (*media_changed) (struct gendisk *);
+ int (*revalidate_disk) (struct gendisk *);
+ int (*getgeo)(struct block_device *, struct hd_geometry *);
++ /* this callback is with swap_lock and sometimes page table lock held */
++ void (*swap_slot_free_notify) (struct block_device *, unsigned long);
+ struct module *owner;
+ };
+
+diff -uprN linux-2.6.28/include/linux/genhd.h linux-2.6.28.new/include/linux/genhd.h
+--- linux-2.6.28/include/linux/genhd.h 2011-03-13 15:23:58.275368057 +0100
++++ linux-2.6.28.new/include/linux/genhd.h 2011-03-15 10:14:01.575121499 +0100
+@@ -113,6 +113,8 @@ struct hd_struct {
+ #define GENHD_FL_UP 16
+ #define GENHD_FL_SUPPRESS_PARTITION_INFO 32
+ #define GENHD_FL_EXT_DEVT 64 /* allow extended devt */
++#define GENHD_FL_REMAP_SWAPPED_PAGES 128
++#define GENHD_FL_NOTIFY_REMAPPED_ONLY 256
+
+ #define BLK_SCSI_MAX_CMDS (256)
+ #define BLK_SCSI_CMD_PER_LONG (BLK_SCSI_MAX_CMDS / (sizeof(long) * 8))
+diff -uprN linux-2.6.28/mm/swapfile.c linux-2.6.28.new/mm/swapfile.c
+--- linux-2.6.28/mm/swapfile.c 2011-02-01 09:54:31.434289623 +0100
++++ linux-2.6.28.new/mm/swapfile.c 2011-03-15 10:14:50.998178343 +0100
+@@ -270,10 +270,23 @@ out:
+ return NULL;
+ }
+
++static void swap_entry_update(struct swap_info_struct *p, unsigned long offset)
++{
++ if (offset < p->lowest_bit)
++ p->lowest_bit = offset;
++ if (offset > p->highest_bit)
++ p->highest_bit = offset;
++ if (p->prio > swap_info[swap_list.next].prio)
++ swap_list.next = p - swap_info;
++ nr_swap_pages++;
++ p->inuse_pages--;
++}
++
+ static int swap_entry_free(struct swap_info_struct *p, unsigned long offset)
+ {
+ int count = p->swap_map[offset];
+ unsigned old;
++ struct gendisk *disk;
+
+ if (count >= SWAP_MAP_MAX)
+ return count;
+@@ -283,28 +296,40 @@ static int swap_entry_free(struct swap_i
+ if (count)
+ return count;
+
+- spin_lock(&p->remap_lock);
++ disk = p->bdev->bd_disk;
+
+- if (offset < p->lowest_bit)
+- p->lowest_bit = offset;
+- if (offset > p->highest_bit)
+- p->highest_bit = offset;
+- if (p->prio > swap_info[swap_list.next].prio)
+- swap_list.next = p - swap_info;
+- nr_swap_pages++;
+- p->inuse_pages--;
++ if (p->swap_remap) {
++ spin_lock(&p->remap_lock);
++ swap_entry_update(p, offset);
++ }
++ else {
++ swap_entry_update(p, offset);
++ if (disk->fops->swap_slot_free_notify)
++ disk->fops->swap_slot_free_notify(p->bdev, offset);
++ return 0;
++ }
+
+ /* Re-map the page number */
+ old = p->swap_remap[offset] & 0x7FFFFFFF;
+ /* Zero means it was not re-mapped */
+- if (!old)
+- goto out;
++ if (!old) {
++ /* Skip notify if flag is set or the page is used */
++ if ((disk->flags & GENHD_FL_NOTIFY_REMAPPED_ONLY) ||
++ (p->swap_remap[offset] & 0x80000000))
++ goto out;
++
++ old = offset;
++ goto notify;
++ }
+ /* Clear the re-mapping */
+ p->swap_remap[offset] &= 0x80000000;
+ /* Mark the re-mapped page as unused */
+ p->swap_remap[old] &= 0x7FFFFFFF;
+ /* Record how many free pages there are */
+ p->gaps_exist += 1;
++notify:
++ if (disk->fops->swap_slot_free_notify)
++ disk->fops->swap_slot_free_notify(p->bdev, old);
+ out:
+ spin_unlock(&p->remap_lock);
+ return 0;
+@@ -1110,6 +1135,8 @@ sector_t map_swap_page(struct swap_info_
+ struct swap_extent *start_se = se;
+ unsigned old;
+
++ if (!sis->swap_remap)
++ goto out;
+ /*
+ * Instead of using the offset we are given, re-map it to the next
+ * sequential position.
+@@ -1159,7 +1186,7 @@ sector_t map_swap_page(struct swap_info_
+ offset = old;
+ }
+ spin_unlock(&sis->remap_lock);
+-
++out:
+ for ( ; ; ) {
+ struct list_head *lh;
+
+@@ -1517,8 +1544,10 @@ SYSCALL_DEFINE1(swapoff, const char __us
+ p->flags = 0;
+ spin_unlock(&swap_lock);
+ mutex_unlock(&swapon_mutex);
+- kfree(p->gap_pool_arr);
+- vfree(p->swap_remap);
++ if (p->swap_remap) {
++ kfree(p->gap_pool_arr);
++ vfree(p->swap_remap);
++ }
+ vfree(swap_map);
+ inode = mapping->host;
+ if (S_ISBLK(inode->i_mode)) {
+@@ -1832,15 +1861,17 @@ SYSCALL_DEFINE2(swapon, const char __use
+ error = -ENOMEM;
+ goto bad_swap;
+ }
+- swap_remap = vmalloc(maxpages * sizeof(unsigned));
+- if (!swap_remap) {
+- error = -ENOMEM;
+- goto bad_swap;
++ if (p->bdev->bd_disk->flags & GENHD_FL_REMAP_SWAPPED_PAGES) {
++ swap_remap = vmalloc(maxpages * sizeof(unsigned));
++ if (!swap_remap) {
++ error = -ENOMEM;
++ goto bad_swap;
++ }
++ memset(swap_remap, 0, maxpages * sizeof(unsigned));
+ }
+
+ error = 0;
+ memset(swap_map, 0, maxpages * sizeof(short));
+- memset(swap_remap, 0, maxpages * sizeof(unsigned));
+ for (i = 0; i < swap_header->info.nr_badpages; i++) {
+ int page_nr = swap_header->info.badpages[i];
+ if (page_nr <= 0 || page_nr >= swap_header->info.last_page)
+@@ -1872,13 +1903,15 @@ SYSCALL_DEFINE2(swapon, const char __use
+ goto bad_swap;
+ }
+
+- p->gap_pool_arr = kmalloc(sizeof(struct swap_gap_node)*
+- SWAP_GAP_TREE_SIZE, GFP_KERNEL);
+- if (!p->gap_pool_arr) {
+- error = -ENOMEM;
+- goto bad_swap;
++ if (swap_remap) {
++ p->gap_pool_arr = kmalloc(sizeof(struct swap_gap_node)*
++ SWAP_GAP_TREE_SIZE, GFP_KERNEL);
++ if (!p->gap_pool_arr) {
++ error = -ENOMEM;
++ goto bad_swap;
++ }
++ p->gaps_tree = RB_ROOT;
+ }
+- p->gaps_tree = RB_ROOT;
+
+ mutex_lock(&swapon_mutex);
+ spin_lock(&swap_lock);
+@@ -1889,11 +1922,13 @@ SYSCALL_DEFINE2(swapon, const char __use
+ p->prio = --least_priority;
+ p->swap_map = swap_map;
+ p->swap_remap = swap_remap;
+- p->gap_next = 1;
+- p->gap_end = p->max - 1;
+- p->gaps_exist = p->max - 1;
+- spin_lock_init(&p->remap_lock);
+- mutex_init(&p->remap_mutex);
++ if (swap_remap) {
++ p->gap_next = 1;
++ p->gap_end = p->max - 1;
++ p->gaps_exist = p->max - 1;
++ spin_lock_init(&p->remap_lock);
++ mutex_init(&p->remap_mutex);
++ }
+ p->flags = SWP_ACTIVE;
+ nr_swap_pages += nr_good_pages;
+ total_swap_pages += nr_good_pages;
+@@ -1932,7 +1967,8 @@ bad_swap_2:
+ p->swap_file = NULL;
+ p->flags = 0;
+ spin_unlock(&swap_lock);
+- vfree(swap_remap);
++ if (swap_remap)
++ vfree(swap_remap);
+ vfree(swap_map);
+ if (swap_file)
+ filp_close(swap_file, NULL);