[Qemu-devel] [RFC V3 16/24] qcow2: Allow creation of images using deduplication.

["Benoît Canet" <benoit@irqsave.net>]

todo: Change qemu-img output so it reflect the dedup cluster size.

Signed-off-by: Benoit Canet <benoit@irqsave.net>
---
 block/qcow2.c |   94 +++++++++++++++++++++++++++++++++++++++++++++++++--------
 block/qcow2.h |    2 ++
 2 files changed, 84 insertions(+), 12 deletions(-)

diff --git a/block/qcow2.c b/block/qcow2.c
index 76d2340..e641049 100644
--- a/block/qcow2.c
+++ b/block/qcow2.c
@@ -277,6 +277,11 @@ static int qcow2_mark_dirty(BlockDriverState *bs)
     return qcow2_add_feature(bs, QCOW2_INCOMPAT_DIRTY);
 }
 
+static int qcow2_activate_dedup(BlockDriverState *bs)
+{
+    return qcow2_add_feature(bs, QCOW2_INCOMPAT_DEDUP);
+}
+
 /*
  * Clears an incompatible feature bit and flushes before if necessary.
  * Only call this function when there are no pending requests, it does not
@@ -913,6 +918,11 @@ static void qcow2_close(BlockDriverState *bs)
     BDRVQcowState *s = bs->opaque;
     g_free(s->l1_table);
 
+    if (s->has_dedup) {
+        qcow2_cache_flush(bs, s->dedup_cluster_cache);
+        qcow2_cache_destroy(bs, s->dedup_cluster_cache);
+    }
+
     qcow2_cache_flush(bs, s->l2_table_cache);
     qcow2_cache_flush(bs, s->refcount_block_cache);
 
@@ -1231,7 +1241,8 @@ static int preallocate(BlockDriverState *bs)
 static int qcow2_create2(const char *filename, int64_t total_size,
                          const char *backing_file, const char *backing_format,
                          int flags, size_t cluster_size, int prealloc,
-                         QEMUOptionParameter *options, int version)
+                         QEMUOptionParameter *options, int version,
+                         bool dedup)
 {
     /* Calculate cluster_bits */
     int cluster_bits;
@@ -1258,8 +1269,10 @@ static int qcow2_create2(const char *filename, int64_t total_size,
      * size for any qcow2 image.
      */
     BlockDriverState* bs;
+    BDRVQcowState *s;
     QCowHeader header;
-    uint8_t* refcount_table;
+    uint8_t *tables;
+    int size;
     int ret;
 
     ret = bdrv_create_file(filename, options);
@@ -1301,10 +1314,11 @@ static int qcow2_create2(const char *filename, int64_t total_size,
         goto out;
     }
 
-    /* Write an empty refcount table */
-    refcount_table = g_malloc0(cluster_size);
-    ret = bdrv_pwrite(bs, cluster_size, refcount_table, cluster_size);
-    g_free(refcount_table);
+    /* Write an empty refcount table + extra space for dedup table if needed */
+    size = cluster_size * (dedup ? 2 : 1);
+    tables = g_malloc0(size);
+    ret = bdrv_pwrite(bs, cluster_size, tables, size);
+    g_free(tables);
 
     if (ret < 0) {
         goto out;
@@ -1325,7 +1339,8 @@ static int qcow2_create2(const char *filename, int64_t total_size,
         goto out;
     }
 
-    ret = qcow2_alloc_clusters(bs, 2 * cluster_size);
+    size += cluster_size;
+    ret = qcow2_alloc_clusters(bs, size);
     if (ret < 0) {
         goto out;
 
@@ -1335,11 +1350,32 @@ static int qcow2_create2(const char *filename, int64_t total_size,
     }
 
     /* Okay, now that we have a valid image, let's give it the right size */
+    s = bs->opaque;
+    size = (total_size + (dedup ? s->cluster_sectors : 0)) * BDRV_SECTOR_SIZE;
     ret = bdrv_truncate(bs, total_size * BDRV_SECTOR_SIZE);
     if (ret < 0) {
         goto out;
     }
 
+    if (dedup) {
+        s->has_dedup = true;
+        s->dedup_table_offset = cluster_size * 2;
+        s->dedup_table_size = cluster_size / sizeof(uint64_t);
+
+        ret = qcow2_activate_dedup(bs);
+        if (ret < 0) {
+            goto out;
+        }
+
+        ret = qcow2_update_header(bs);
+        if (ret < 0) {
+            goto out;
+        }
+
+        /* minimal init */
+        s->dedup_cluster_cache = qcow2_cache_create(bs, DEDUP_CACHE_SIZE);
+    }
+
     /* Want a backing file? There you go.*/
     if (backing_file) {
         ret = bdrv_change_backing_file(bs, backing_file, backing_format);
@@ -1365,15 +1401,30 @@ out:
     return ret;
 }
 
+static int qcow2_warn_if_version_3_is_needed(int version,
+                                             bool has_feature,
+                                             const char *feature)
+{
+    if (version < 3 && has_feature) {
+        fprintf(stderr, "%s only supported with compatibility "
+                "level 1.1 and above (use compat=1.1 or greater)\n",
+                feature);
+        return -EINVAL;
+    }
+    return 0;
+}
+
 static int qcow2_create(const char *filename, QEMUOptionParameter *options)
 {
     const char *backing_file = NULL;
     const char *backing_fmt = NULL;
     uint64_t sectors = 0;
     int flags = 0;
+    int ret;
     size_t cluster_size = DEFAULT_CLUSTER_SIZE;
     int prealloc = 0;
     int version = 2;
+    bool dedup = false;
 
     /* Read out options */
     while (options && options->name) {
@@ -1411,24 +1462,38 @@ static int qcow2_create(const char *filename, QEMUOptionParameter *options)
             }
         } else if (!strcmp(options->name, BLOCK_OPT_LAZY_REFCOUNTS)) {
             flags |= options->value.n ? BLOCK_FLAG_LAZY_REFCOUNTS : 0;
+        } else if (!strcmp(options->name, BLOCK_OPT_DEDUP)) {
+            dedup = options->value.n ? true : false;
         }
         options++;
     }
 
+    if (dedup) {
+        cluster_size = 4096;
+    }
+
     if (backing_file && prealloc) {
         fprintf(stderr, "Backing file and preallocation cannot be used at "
             "the same time\n");
         return -EINVAL;
     }
 
-    if (version < 3 && (flags & BLOCK_FLAG_LAZY_REFCOUNTS)) {
-        fprintf(stderr, "Lazy refcounts only supported with compatibility "
-                "level 1.1 and above (use compat=1.1 or greater)\n");
-        return -EINVAL;
+    ret = qcow2_warn_if_version_3_is_needed(version,
+                                            flags & BLOCK_FLAG_LAZY_REFCOUNTS,
+                                            "Lazy refcounts");
+    if (ret < 0) {
+        return ret;
+    }
+    ret = qcow2_warn_if_version_3_is_needed(version,
+                                            dedup,
+                                            "Deduplication");
+    if (ret < 0) {
+        return ret;
     }
 
     return qcow2_create2(filename, sectors, backing_file, backing_fmt, flags,
-                         cluster_size, prealloc, options, version);
+                         cluster_size, prealloc, options, version,
+                         dedup);
 }
 
 static int qcow2_make_empty(BlockDriverState *bs)
@@ -1731,6 +1796,11 @@ static QEMUOptionParameter qcow2_create_options[] = {
         .type = OPT_FLAG,
         .help = "Postpone refcount updates",
     },
+    {
+        .name = BLOCK_OPT_DEDUP,
+        .type = OPT_FLAG,
+        .help = "Live deduplication",
+    },
     { NULL }
 };
 
diff --git a/block/qcow2.h b/block/qcow2.h
index 9d08bf9..90dcdd9 100644
--- a/block/qcow2.h
+++ b/block/qcow2.h
@@ -56,6 +56,8 @@
 /* Must be at least 4 to cover all cases of refcount table growth */
 #define REFCOUNT_CACHE_SIZE 4
 
+#define DEDUP_CACHE_SIZE 4
+
 #define DEFAULT_CLUSTER_SIZE 65536
 
 /* indicate that the hash structure is empty and miss offset */
-- 
1.7.10.4