mm/zpool: update zswap to use zpool
authorDan Streetman <ddstreet@ieee.org>
Wed, 6 Aug 2014 23:08:40 +0000 (16:08 -0700)
committerLinus Torvalds <torvalds@linux-foundation.org>
Thu, 7 Aug 2014 01:01:23 +0000 (18:01 -0700)
Change zswap to use the zpool api instead of directly using zbud.  Add a
boot-time param to allow selecting which zpool implementation to use,
with zbud as the default.

Signed-off-by: Dan Streetman <ddstreet@ieee.org>
Tested-by: Seth Jennings <sjennings@variantweb.net>
Cc: Weijie Yang <weijie.yang@samsung.com>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Nitin Gupta <ngupta@vflare.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
mm/Kconfig
mm/zswap.c

index 12179b8c3b8980c020fc7e7a6c6da8f44abff8c9..886db2158538572ee52790b31d93b4baeb8b187f 100644 (file)
@@ -535,7 +535,7 @@ config ZSWAP
        bool "Compressed cache for swap pages (EXPERIMENTAL)"
        depends on FRONTSWAP && CRYPTO=y
        select CRYPTO_LZO
-       select ZBUD
+       select ZPOOL
        default n
        help
          A lightweight compressed cache for swap pages.  It takes
index 008388fe7b0ff923b2300504b3a7717869fe12f8..032c21eeab2b0f05ebe73a9a7d1fd820651f3cef 100644 (file)
@@ -34,7 +34,7 @@
 #include <linux/swap.h>
 #include <linux/crypto.h>
 #include <linux/mempool.h>
-#include <linux/zbud.h>
+#include <linux/zpool.h>
 
 #include <linux/mm_types.h>
 #include <linux/page-flags.h>
@@ -45,8 +45,8 @@
 /*********************************
 * statistics
 **********************************/
-/* Number of memory pages used by the compressed pool */
-static u64 zswap_pool_pages;
+/* Total bytes used by the compressed storage */
+static u64 zswap_pool_total_size;
 /* The number of compressed pages currently stored in zswap */
 static atomic_t zswap_stored_pages = ATOMIC_INIT(0);
 
@@ -89,8 +89,13 @@ static unsigned int zswap_max_pool_percent = 20;
 module_param_named(max_pool_percent,
                        zswap_max_pool_percent, uint, 0644);
 
-/* zbud_pool is shared by all of zswap backend  */
-static struct zbud_pool *zswap_pool;
+/* Compressed storage to use */
+#define ZSWAP_ZPOOL_DEFAULT "zbud"
+static char *zswap_zpool_type = ZSWAP_ZPOOL_DEFAULT;
+module_param_named(zpool, zswap_zpool_type, charp, 0444);
+
+/* zpool is shared by all of zswap backend  */
+static struct zpool *zswap_pool;
 
 /*********************************
 * compression functions
@@ -168,7 +173,7 @@ static void zswap_comp_exit(void)
  *            be held while changing the refcount.  Since the lock must
  *            be held, there is no reason to also make refcount atomic.
  * offset - the swap offset for the entry.  Index into the red-black tree.
- * handle - zbud allocation handle that stores the compressed page data
+ * handle - zpool allocation handle that stores the compressed page data
  * length - the length in bytes of the compressed page data.  Needed during
  *          decompression
  */
@@ -284,15 +289,15 @@ static void zswap_rb_erase(struct rb_root *root, struct zswap_entry *entry)
 }
 
 /*
- * Carries out the common pattern of freeing and entry's zbud allocation,
+ * Carries out the common pattern of freeing and entry's zpool allocation,
  * freeing the entry itself, and decrementing the number of stored pages.
  */
 static void zswap_free_entry(struct zswap_entry *entry)
 {
-       zbud_free(zswap_pool, entry->handle);
+       zpool_free(zswap_pool, entry->handle);
        zswap_entry_cache_free(entry);
        atomic_dec(&zswap_stored_pages);
-       zswap_pool_pages = zbud_get_pool_size(zswap_pool);
+       zswap_pool_total_size = zpool_get_total_size(zswap_pool);
 }
 
 /* caller must hold the tree lock */
@@ -409,7 +414,7 @@ cleanup:
 static bool zswap_is_full(void)
 {
        return totalram_pages * zswap_max_pool_percent / 100 <
-               zswap_pool_pages;
+               DIV_ROUND_UP(zswap_pool_total_size, PAGE_SIZE);
 }
 
 /*********************************
@@ -525,7 +530,7 @@ static int zswap_get_swap_cache_page(swp_entry_t entry,
  * the swap cache, the compressed version stored by zswap can be
  * freed.
  */
-static int zswap_writeback_entry(struct zbud_pool *pool, unsigned long handle)
+static int zswap_writeback_entry(struct zpool *pool, unsigned long handle)
 {
        struct zswap_header *zhdr;
        swp_entry_t swpentry;
@@ -541,9 +546,9 @@ static int zswap_writeback_entry(struct zbud_pool *pool, unsigned long handle)
        };
 
        /* extract swpentry from data */
-       zhdr = zbud_map(pool, handle);
+       zhdr = zpool_map_handle(pool, handle, ZPOOL_MM_RO);
        swpentry = zhdr->swpentry; /* here */
-       zbud_unmap(pool, handle);
+       zpool_unmap_handle(pool, handle);
        tree = zswap_trees[swp_type(swpentry)];
        offset = swp_offset(swpentry);
 
@@ -573,13 +578,13 @@ static int zswap_writeback_entry(struct zbud_pool *pool, unsigned long handle)
        case ZSWAP_SWAPCACHE_NEW: /* page is locked */
                /* decompress */
                dlen = PAGE_SIZE;
-               src = (u8 *)zbud_map(zswap_pool, entry->handle) +
-                       sizeof(struct zswap_header);
+               src = (u8 *)zpool_map_handle(zswap_pool, entry->handle,
+                               ZPOOL_MM_RO) + sizeof(struct zswap_header);
                dst = kmap_atomic(page);
                ret = zswap_comp_op(ZSWAP_COMPOP_DECOMPRESS, src,
                                entry->length, dst, &dlen);
                kunmap_atomic(dst);
-               zbud_unmap(zswap_pool, entry->handle);
+               zpool_unmap_handle(zswap_pool, entry->handle);
                BUG_ON(ret);
                BUG_ON(dlen != PAGE_SIZE);
 
@@ -652,7 +657,7 @@ static int zswap_frontswap_store(unsigned type, pgoff_t offset,
        /* reclaim space if needed */
        if (zswap_is_full()) {
                zswap_pool_limit_hit++;
-               if (zbud_reclaim_page(zswap_pool, 8)) {
+               if (zpool_shrink(zswap_pool, 1, NULL)) {
                        zswap_reject_reclaim_fail++;
                        ret = -ENOMEM;
                        goto reject;
@@ -679,7 +684,7 @@ static int zswap_frontswap_store(unsigned type, pgoff_t offset,
 
        /* store */
        len = dlen + sizeof(struct zswap_header);
-       ret = zbud_alloc(zswap_pool, len, __GFP_NORETRY | __GFP_NOWARN,
+       ret = zpool_malloc(zswap_pool, len, __GFP_NORETRY | __GFP_NOWARN,
                &handle);
        if (ret == -ENOSPC) {
                zswap_reject_compress_poor++;
@@ -689,11 +694,11 @@ static int zswap_frontswap_store(unsigned type, pgoff_t offset,
                zswap_reject_alloc_fail++;
                goto freepage;
        }
-       zhdr = zbud_map(zswap_pool, handle);
+       zhdr = zpool_map_handle(zswap_pool, handle, ZPOOL_MM_RW);
        zhdr->swpentry = swp_entry(type, offset);
        buf = (u8 *)(zhdr + 1);
        memcpy(buf, dst, dlen);
-       zbud_unmap(zswap_pool, handle);
+       zpool_unmap_handle(zswap_pool, handle);
        put_cpu_var(zswap_dstmem);
 
        /* populate entry */
@@ -716,7 +721,7 @@ static int zswap_frontswap_store(unsigned type, pgoff_t offset,
 
        /* update stats */
        atomic_inc(&zswap_stored_pages);
-       zswap_pool_pages = zbud_get_pool_size(zswap_pool);
+       zswap_pool_total_size = zpool_get_total_size(zswap_pool);
 
        return 0;
 
@@ -752,13 +757,13 @@ static int zswap_frontswap_load(unsigned type, pgoff_t offset,
 
        /* decompress */
        dlen = PAGE_SIZE;
-       src = (u8 *)zbud_map(zswap_pool, entry->handle) +
-                       sizeof(struct zswap_header);
+       src = (u8 *)zpool_map_handle(zswap_pool, entry->handle,
+                       ZPOOL_MM_RO) + sizeof(struct zswap_header);
        dst = kmap_atomic(page);
        ret = zswap_comp_op(ZSWAP_COMPOP_DECOMPRESS, src, entry->length,
                dst, &dlen);
        kunmap_atomic(dst);
-       zbud_unmap(zswap_pool, entry->handle);
+       zpool_unmap_handle(zswap_pool, entry->handle);
        BUG_ON(ret);
 
        spin_lock(&tree->lock);
@@ -811,7 +816,7 @@ static void zswap_frontswap_invalidate_area(unsigned type)
        zswap_trees[type] = NULL;
 }
 
-static struct zbud_ops zswap_zbud_ops = {
+static struct zpool_ops zswap_zpool_ops = {
        .evict = zswap_writeback_entry
 };
 
@@ -869,8 +874,8 @@ static int __init zswap_debugfs_init(void)
                        zswap_debugfs_root, &zswap_written_back_pages);
        debugfs_create_u64("duplicate_entry", S_IRUGO,
                        zswap_debugfs_root, &zswap_duplicate_entry);
-       debugfs_create_u64("pool_pages", S_IRUGO,
-                       zswap_debugfs_root, &zswap_pool_pages);
+       debugfs_create_u64("pool_total_size", S_IRUGO,
+                       zswap_debugfs_root, &zswap_pool_total_size);
        debugfs_create_atomic_t("stored_pages", S_IRUGO,
                        zswap_debugfs_root, &zswap_stored_pages);
 
@@ -895,16 +900,26 @@ static void __exit zswap_debugfs_exit(void) { }
 **********************************/
 static int __init init_zswap(void)
 {
+       gfp_t gfp = __GFP_NORETRY | __GFP_NOWARN;
+
        if (!zswap_enabled)
                return 0;
 
        pr_info("loading zswap\n");
 
-       zswap_pool = zbud_create_pool(GFP_KERNEL, &zswap_zbud_ops);
+       zswap_pool = zpool_create_pool(zswap_zpool_type, gfp, &zswap_zpool_ops);
+       if (!zswap_pool && strcmp(zswap_zpool_type, ZSWAP_ZPOOL_DEFAULT)) {
+               pr_info("%s zpool not available\n", zswap_zpool_type);
+               zswap_zpool_type = ZSWAP_ZPOOL_DEFAULT;
+               zswap_pool = zpool_create_pool(zswap_zpool_type, gfp,
+                                       &zswap_zpool_ops);
+       }
        if (!zswap_pool) {
-               pr_err("zbud pool creation failed\n");
+               pr_err("%s zpool not available\n", zswap_zpool_type);
+               pr_err("zpool creation failed\n");
                goto error;
        }
+       pr_info("using %s pool\n", zswap_zpool_type);
 
        if (zswap_entry_cache_create()) {
                pr_err("entry cache creation failed\n");
@@ -928,7 +943,7 @@ pcpufail:
 compfail:
        zswap_entry_cache_destory();
 cachefail:
-       zbud_destroy_pool(zswap_pool);
+       zpool_destroy_pool(zswap_pool);
 error:
        return -ENOMEM;
 }