mm: swap: SWP_SYNCHRONOUS_IO: skip swapcache only if swapped page has…

… no other reference

When SWP_SYNCHRONOUS_IO swapped-in pages are shared by several
processes, it can cause unnecessary memory wastage by skipping swap
cache.  Because, with swapin fault by read, they could share a page if
the page were in swap cache.  Thus, it avoids allocating same content
new pages.

This patch makes the swapcache skipping work only if the swap pte is
non-sharable.

[[email protected]: coding-style fixes]
Link: http://lkml.kernel.org/r/[email protected]
Signed-off-by: Minchan Kim <[email protected]>
Cc: Dan Williams <[email protected]>
Cc: Ross Zwisler <[email protected]>
Cc: Hugh Dickins <[email protected]>
Cc: Christoph Hellwig <[email protected]>
Cc: Ilya Dryomov <[email protected]>
Cc: Jens Axboe <[email protected]>
Cc: Sergey Senozhatsky <[email protected]>
Cc: Huang Ying <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

include/linux/swap.h

@@ -463,6 +463,7 @@ extern unsigned int count_swap_pages(int, int);
 extern sector_t map_swap_page(struct page *, struct block_device **);
 extern sector_t swapdev_block(int, pgoff_t);
 extern int page_swapcount(struct page *);
+extern int __swap_count(struct swap_info_struct *si, swp_entry_t entry);
 extern int __swp_swapcount(swp_entry_t entry);
 extern int swp_swapcount(swp_entry_t entry);
 extern struct swap_info_struct *page_swap_info(struct page *);
@@ -589,6 +590,11 @@ static inline int page_swapcount(struct page *page)
 	return 0;
 }
 
+static inline int __swap_count(struct swap_info_struct *si, swp_entry_t entry)
+{
+	return 0;
+}
+
 static inline int __swp_swapcount(swp_entry_t entry)
 {
 	return 0;

mm/memory.c

@@ -2890,15 +2890,8 @@ int do_swap_page(struct vm_fault *vmf)
 	if (!page) {
 		struct swap_info_struct *si = swp_swap_info(entry);
 
-		if (!(si->flags & SWP_SYNCHRONOUS_IO)) {
-			if (vma_readahead)
-				page = do_swap_page_readahead(entry,
-					GFP_HIGHUSER_MOVABLE, vmf, &swap_ra);
-			else
-				page = swapin_readahead(entry,
-					GFP_HIGHUSER_MOVABLE, vma, vmf->address);
-			swapcache = page;
-		} else {
+		if (si->flags & SWP_SYNCHRONOUS_IO &&
+				__swap_count(si, entry) == 1) {
 			/* skip swapcache */
 			page = alloc_page_vma(GFP_HIGHUSER_MOVABLE, vma, vmf->address);
 			if (page) {
@@ -2908,6 +2901,14 @@ int do_swap_page(struct vm_fault *vmf)
 				lru_cache_add_anon(page);
 				swap_readpage(page, true);
 			}
+		} else {
+			if (vma_readahead)
+				page = do_swap_page_readahead(entry,
+					GFP_HIGHUSER_MOVABLE, vmf, &swap_ra);
+			else
+				page = swapin_readahead(entry,
+				       GFP_HIGHUSER_MOVABLE, vma, vmf->address);
+			swapcache = page;
 		}
 
 		if (!page) {

mm/swapfile.c

@@ -1328,6 +1328,13 @@ int page_swapcount(struct page *page)
 	return count;
 }
 
+int __swap_count(struct swap_info_struct *si, swp_entry_t entry)
+{
+	pgoff_t offset = swp_offset(entry);
+
+	return swap_count(si->swap_map[offset]);
+}
+
 static int swap_swapcount(struct swap_info_struct *si, swp_entry_t entry)
 {
 	int count = 0;