[Qemu-devel] [RFC V5 18/36] block: Add qemu-img dedup create option.

["Benoît Canet" <benoit@irqsave.net>]

Signed-off-by: Benoit Canet <benoit@irqsave.net>
---
 block/qcow2.c             |  113 +++++++++++++++++++++++++++++++++++++++------
 block/qcow2.h             |    2 +
 include/block/block_int.h |    1 +
 3 files changed, 103 insertions(+), 13 deletions(-)

diff --git a/block/qcow2.c b/block/qcow2.c
index f046a77..835554d 100644
--- a/block/qcow2.c
+++ b/block/qcow2.c
@@ -276,6 +276,11 @@ 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
@@ -907,6 +912,11 @@ 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);
 
@@ -1266,7 +1276,8 @@ 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, uint8_t hash_algo)
 {
     /* Calculate cluster_bits */
     int cluster_bits;
@@ -1293,8 +1304,10 @@ 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);
@@ -1336,10 +1349,11 @@ 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 = dedup ? 2 : 1;
+    tables = g_malloc0(size * cluster_size);
+    ret = bdrv_pwrite(bs, cluster_size, tables, size * cluster_size);
+    g_free(tables);
 
     if (ret < 0) {
         goto out;
@@ -1350,7 +1364,7 @@ static int qcow2_create2(const char *filename, int64_t total_size,
     /*
      * And now open the image and make it consistent first (i.e. increase the
      * refcount of the cluster that is occupied by the header and the refcount
-     * table)
+     * table and the eventual dedup table)
      */
     BlockDriver* drv = bdrv_find_format("qcow2");
     assert(drv != NULL);
@@ -1360,7 +1374,8 @@ static int qcow2_create2(const char *filename, int64_t total_size,
         goto out;
     }
 
-    ret = qcow2_alloc_clusters(bs, 2 * cluster_size);
+    size++; /* Add a cluster for the header */
+    ret = qcow2_alloc_clusters(bs, size * cluster_size);
     if (ret < 0) {
         goto out;
 
@@ -1370,11 +1385,33 @@ 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;
     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);
+        s->dedup_hash_algo = hash_algo;
+
+        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,
+                                                    s->hash_block_size);
+    }
+
     /* Want a backing file? There you go.*/
     if (backing_file) {
         ret = bdrv_change_backing_file(bs, backing_file, backing_format);
@@ -1400,15 +1437,41 @@ 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 int8_t qcow2_get_dedup_hash_algo(char *value)
+{
+    if (!strcmp(value, "sha256")) {
+        return QCOW_HASH_SHA256;
+    }
+
+    error_printf("Unsupported deduplication hash algorithm.\n");
+    return -EINVAL;
+}
+
 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;
+    int8_t hash_algo = 0;
 
     /* Read out options */
     while (options && options->name) {
@@ -1446,24 +1509,43 @@ 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) &&
+                   options->value.s) {
+            hash_algo = qcow2_get_dedup_hash_algo(options->value.s);
+            if (hash_algo < 0) {
+                return hash_algo;
+            }
+            dedup = true;
         }
         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, hash_algo);
 }
 
 static int qcow2_make_empty(BlockDriverState *bs)
@@ -1766,6 +1848,11 @@ static QEMUOptionParameter qcow2_create_options[] = {
         .type = OPT_FLAG,
         .help = "Postpone refcount updates",
     },
+    {
+        .name = BLOCK_OPT_DEDUP,
+        .type = OPT_STRING,
+        .help = "Deduplication",
+    },
     { NULL }
 };
 
diff --git a/block/qcow2.h b/block/qcow2.h
index 59432fd..f987328 100644
--- a/block/qcow2.h
+++ b/block/qcow2.h
@@ -60,6 +60,8 @@
 /* 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
 
 #define HASH_LENGTH 32
diff --git a/include/block/block_int.h b/include/block/block_int.h
index f83ffb8..b7ed3e6 100644
--- a/include/block/block_int.h
+++ b/include/block/block_int.h
@@ -55,6 +55,7 @@
 #define BLOCK_OPT_SUBFMT            "subformat"
 #define BLOCK_OPT_COMPAT_LEVEL      "compat"
 #define BLOCK_OPT_LAZY_REFCOUNTS    "lazy_refcounts"
+#define BLOCK_OPT_DEDUP             "dedup"
 
 typedef struct BdrvTrackedRequest BdrvTrackedRequest;
 
-- 
1.7.10.4