[PATCH 3/4] reuse unused swap entry if necessary

[KAMEZAWA Hiroyuki]

From: KAMEZAWA Hiroyuki <kamezawa.hiroyu@xxxxxxxxxxxxxx>

Now, we can know a swap entry is just used as SwapCache via swap_map,
without looking up swap cache.

Then, we have a chance to reuse swap-cache-only swap entries in
get_swap_pages().

This patch tries to free swap-cache-only swap entries if swap is
not enough.
Note: We hit following path when swap_cluster code cannot find
a free cluster. Then, vm_swap_full() is not only condition to allow
the kernel to reclaim unused swap.

Signed-off-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@xxxxxxxxxxxxxx>
---
 mm/swapfile.c |   39 +++++++++++++++++++++++++++++++++++++++
 1 file changed, 39 insertions(+)

Index: new-trial-swapcount2/mm/swapfile.c
===================================================================
--- new-trial-swapcount2.orig/mm/swapfile.c
+++ new-trial-swapcount2/mm/swapfile.c
@@ -73,6 +73,25 @@ static inline unsigned short make_swap_c
 	return ret;
 }
 
+static int
+try_to_reuse_swap(struct swap_info_struct *si, unsigned long offset)
+{
+	int type = si - swap_info;
+	swp_entry_t entry = swp_entry(type, offset);
+	struct page *page;
+	int ret = 0;
+
+	page = find_get_page(&swapper_space, entry.val);
+	if (!page)
+		return 0;
+	if (trylock_page(page)) {
+		ret = try_to_free_swap(page);
+		unlock_page(page);
+	}
+	page_cache_release(page);
+	return ret;
+}
+
 /*
  * We need this because the bdev->unplug_fn can sleep and we cannot
  * hold swap_lock while calling the unplug_fn. And swap_lock
@@ -294,6 +313,18 @@ checks:
 		goto no_page;
 	if (offset > si->highest_bit)
 		scan_base = offset = si->lowest_bit;
+
+	/* reuse swap entry of cache-only swap if not busy. */
+	if (vm_swap_full() && si->swap_map[offset] == SWAP_HAS_CACHE) {
+		int ret;
+		spin_unlock(&swap_lock);
+		ret = try_to_reuse_swap(si, offset);
+		spin_lock(&swap_lock);
+		if (ret)
+			goto checks; /* we released swap_lock. retry. */
+		goto scan; /* In some racy case */
+	}
+
 	if (si->swap_map[offset])
 		goto scan;
 
@@ -375,6 +406,10 @@ scan:
 			spin_lock(&swap_lock);
 			goto checks;
 		}
+		if (vm_swap_full() && si->swap_map[offset] == SWAP_HAS_CACHE) {
+			spin_lock(&swap_lock);
+			goto checks;
+		}
 		if (unlikely(--latency_ration < 0)) {
 			cond_resched();
 			latency_ration = LATENCY_LIMIT;
@@ -386,6 +421,10 @@ scan:
 			spin_lock(&swap_lock);
 			goto checks;
 		}
+		if (vm_swap_full() && si->swap_map[offset] == SWAP_HAS_CACHE) {
+			spin_lock(&swap_lock);
+			goto checks;
+		}
 		if (unlikely(--latency_ration < 0)) {
 			cond_resched();
 			latency_ration = LATENCY_LIMIT;

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/