mm: zswap: support exclusive loads

Commit 71024cb4a0 ("frontswap: remove frontswap_tmem_exclusive_gets")
removed support for exclusive loads from frontswap as it was not used. 
Bring back exclusive loads support to frontswap by adding an "exclusive"
output parameter to frontswap_ops->load.

On the zswap side, add a module parameter to enable/disable exclusive
loads, and a config option to control the boot default value.  Refactor
zswap entry invalidation in zswap_frontswap_invalidate_page() into
zswap_invalidate_entry() to reuse it in zswap_frontswap_load() if
exclusive loads are enabled.

With exclusive loads, we avoid having two copies of the same page in
memory (compressed & uncompressed) after faulting it in from zswap.  On
the other hand, if the page is to be reclaimed again without being
dirtied, it will be re-compressed.  Compression is not usually slow, and a
page that was just faulted in is less likely to be reclaimed again soon.

Link: https://lkml.kernel.org/r/20230607195143.1473802-1-yosryahmed@google.com
Signed-off-by: Yosry Ahmed <yosryahmed@google.com>
Suggested-by: Yu Zhao <yuzhao@google.com>
Acked-by: Johannes Weiner <hannes@cmpxchg.org>
Cc: Dan Streetman <ddstreet@ieee.org>
Cc: Domenico Cerasuolo <cerasuolodomenico@gmail.com>
Cc: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
Cc: Nhat Pham <nphamcs@gmail.com>
Cc: Seth Jennings <sjenning@redhat.com>
Cc: Vitaly Wool <vitaly.wool@konsulko.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

include/linux/frontswap.h

@@ -10,7 +10,7 @@
 struct frontswap_ops {
 	void (*init)(unsigned); /* this swap type was just swapon'ed */
 	int (*store)(unsigned, pgoff_t, struct page *); /* store a page */
-	int (*load)(unsigned, pgoff_t, struct page *); /* load a page */
+	int (*load)(unsigned, pgoff_t, struct page *, bool *); /* load a page */
 	void (*invalidate_page)(unsigned, pgoff_t); /* page no longer needed */
 	void (*invalidate_area)(unsigned); /* swap type just swapoff'ed */
 };

mm/Kconfig

@@ -46,6 +46,22 @@ config ZSWAP_DEFAULT_ON
 	  The selection made here can be overridden by using the kernel
 	  command line 'zswap.enabled=' option.
 
+config ZSWAP_EXCLUSIVE_LOADS_DEFAULT_ON
+	bool "Invalidate zswap entries when pages are loaded"
+	depends on ZSWAP
+	help
+	  If selected, exclusive loads for zswap will be enabled at boot,
+	  otherwise it will be disabled.
+
+	  If exclusive loads are enabled, when a page is loaded from zswap,
+	  the zswap entry is invalidated at once, as opposed to leaving it
+	  in zswap until the swap entry is freed.
+
+	  This avoids having two copies of the same page in memory
+	  (compressed and uncompressed) after faulting in a page from zswap.
+	  The cost is that if the page was never dirtied and needs to be
+	  swapped out again, it will be re-compressed.
+
 choice
 	prompt "Default compressor"
 	depends on ZSWAP

mm/frontswap.c

@@ -206,6 +206,7 @@ int __frontswap_load(struct page *page)
 	int type = swp_type(entry);
 	struct swap_info_struct *sis = swap_info[type];
 	pgoff_t offset = swp_offset(entry);
+	bool exclusive = false;
 
 	VM_BUG_ON(!frontswap_ops);
 	VM_BUG_ON(!PageLocked(page));
@@ -215,9 +216,14 @@ int __frontswap_load(struct page *page)
 		return -1;
 
 	/* Try loading from each implementation, until one succeeds. */
-	ret = frontswap_ops->load(type, offset, page);
+	ret = frontswap_ops->load(type, offset, page, &exclusive);
-	if (ret == 0)
+	if (ret == 0) {
 		inc_frontswap_loads();
+		if (exclusive) {
+			SetPageDirty(page);
+			__frontswap_clear(sis, offset);
+		}
+	}
 	return ret;
 }
 

mm/zswap.c

@@ -138,6 +138,10 @@ static bool zswap_non_same_filled_pages_enabled = true;
 module_param_named(non_same_filled_pages_enabled, zswap_non_same_filled_pages_enabled,
 		   bool, 0644);
 
+static bool zswap_exclusive_loads_enabled = IS_ENABLED(
+		CONFIG_ZSWAP_EXCLUSIVE_LOADS_DEFAULT_ON);
+module_param_named(exclusive_loads, zswap_exclusive_loads_enabled, bool, 0644);
+
 /*********************************
 * data structures
 **********************************/
@@ -1340,12 +1344,22 @@ shrink:
 	goto reject;
 }
 
+static void zswap_invalidate_entry(struct zswap_tree *tree,
+				   struct zswap_entry *entry)
+{
+	/* remove from rbtree */
+	zswap_rb_erase(&tree->rbroot, entry);
+
+	/* drop the initial reference from entry creation */
+	zswap_entry_put(tree, entry);
+}
+
 /*
  * returns 0 if the page was successfully decompressed
  * return -1 on entry not found or error
 */
 static int zswap_frontswap_load(unsigned type, pgoff_t offset,
-				struct page *page)
+				struct page *page, bool *exclusive)
 {
 	struct zswap_tree *tree = zswap_trees[type];
 	struct zswap_entry *entry;
@@ -1415,6 +1429,10 @@ stats:
 freeentry:
 	spin_lock(&tree->lock);
 	zswap_entry_put(tree, entry);
+	if (!ret && zswap_exclusive_loads_enabled) {
+		zswap_invalidate_entry(tree, entry);
+		*exclusive = true;
+	}
 	spin_unlock(&tree->lock);
 
 	return ret;
@@ -1434,13 +1452,7 @@ static void zswap_frontswap_invalidate_page(unsigned type, pgoff_t offset)
 		spin_unlock(&tree->lock);
 		return;
 	}
-
+	zswap_invalidate_entry(tree, entry);
-	/* remove from rbtree */
-	zswap_rb_erase(&tree->rbroot, entry);
-
-	/* drop the initial reference from entry creation */
-	zswap_entry_put(tree, entry);
-
 	spin_unlock(&tree->lock);
 }