[Qemu-devel] [RFC V5 31/62] qcow2: Use large L2 table for deduplication.

["Benoît Canet" <benoit@irqsave.net>]

Signed-off-by: Benoit Canet <benoit@irqsave.net>
---
 block/qcow2-cluster.c  |    2 +-
 block/qcow2-refcount.c |   22 +++++++++++++++-------
 block/qcow2.c          |    8 ++++++--
 3 files changed, 22 insertions(+), 10 deletions(-)

diff --git a/block/qcow2-cluster.c b/block/qcow2-cluster.c
index c016e85..8ad4740 100644
--- a/block/qcow2-cluster.c
+++ b/block/qcow2-cluster.c
@@ -236,7 +236,7 @@ static int l2_allocate(BlockDriverState *bs, int l1_index, uint64_t **table)
             goto fail;
         }
 
-        memcpy(l2_table, old_table, s->cluster_size);
+        memcpy(l2_table, old_table, s->l2_size << 3);
 
         ret = qcow2_cache_put(bs, s->l2_table_cache, (void**) &old_table);
         if (ret < 0) {
diff --git a/block/qcow2-refcount.c b/block/qcow2-refcount.c
index 3077a9f..f305510 100644
--- a/block/qcow2-refcount.c
+++ b/block/qcow2-refcount.c
@@ -536,12 +536,15 @@ fail:
  */
 static int update_cluster_refcount(BlockDriverState *bs,
                                    int64_t cluster_index,
-                                   int addend)
+                                   int addend,
+                                   bool is_l2)
 {
     BDRVQcowState *s = bs->opaque;
     int ret;
 
-    ret = update_refcount(bs, cluster_index << s->cluster_bits, 1, addend,
+    int size = is_l2 ? s->l2_size << 3 : 1;
+
+    ret = update_refcount(bs, cluster_index << s->cluster_bits, size, addend,
                           false);
     if (ret < 0) {
         return ret;
@@ -666,7 +669,7 @@ int64_t qcow2_alloc_bytes(BlockDriverState *bs, int size)
         if (free_in_cluster == 0)
             s->free_byte_offset = 0;
         if ((offset & (s->cluster_size - 1)) != 0)
-            update_cluster_refcount(bs, offset >> s->cluster_bits, 1);
+            update_cluster_refcount(bs, offset >> s->cluster_bits, 1, false);
     } else {
         offset = qcow2_alloc_clusters(bs, s->cluster_size);
         if (offset < 0) {
@@ -676,7 +679,7 @@ int64_t qcow2_alloc_bytes(BlockDriverState *bs, int size)
         if ((cluster_offset + s->cluster_size) == offset) {
             /* we are lucky: contiguous data */
             offset = s->free_byte_offset;
-            update_cluster_refcount(bs, offset >> s->cluster_bits, 1);
+            update_cluster_refcount(bs, offset >> s->cluster_bits, 1, false);
             s->free_byte_offset += size;
         } else {
             s->free_byte_offset = offset;
@@ -817,7 +820,10 @@ int qcow2_update_snapshot_refcount(BlockDriverState *bs,
                     } else {
                         uint64_t cluster_index = (offset & L2E_OFFSET_MASK) >> s->cluster_bits;
                         if (addend != 0) {
-                            refcount = update_cluster_refcount(bs, cluster_index, addend);
+                            refcount = update_cluster_refcount(bs,
+                                                               cluster_index,
+                                                               addend,
+                                                               false);
                         } else {
                             refcount = get_refcount(bs, cluster_index);
                         }
@@ -849,7 +855,9 @@ int qcow2_update_snapshot_refcount(BlockDriverState *bs,
 
 
             if (addend != 0) {
-                refcount = update_cluster_refcount(bs, l2_offset >> s->cluster_bits, addend);
+                refcount = update_cluster_refcount(bs,
+                                                   l2_offset >> s->cluster_bits,
+                                                   addend, true);
             } else {
                 refcount = get_refcount(bs, l2_offset >> s->cluster_bits);
             }
@@ -1145,7 +1153,7 @@ static int check_refcounts_l1(BlockDriverState *bs,
             /* Mark L2 table as used */
             l2_offset &= L1E_OFFSET_MASK;
             inc_refcounts(bs, res, refcount_table, refcount_table_size,
-                l2_offset, s->cluster_size);
+                l2_offset, s->l2_size << 3);
 
             /* L2 tables are cluster aligned */
             if (l2_offset & (s->cluster_size - 1)) {
diff --git a/block/qcow2.c b/block/qcow2.c
index 7ef9170..f70c24b 100644
--- a/block/qcow2.c
+++ b/block/qcow2.c
@@ -432,7 +432,11 @@ static int qcow2_open(BlockDriverState *bs, int flags)
     s->cluster_bits = header.cluster_bits;
     s->cluster_size = 1 << s->cluster_bits;
     s->cluster_sectors = 1 << (s->cluster_bits - 9);
-    s->l2_bits = s->cluster_bits - 3; /* L2 is always one cluster */
+    if (s->incompatible_features & QCOW2_INCOMPAT_DEDUP) {
+        s->l2_bits = 17; /* 64 * 16 KB L2 to compensate smaller cluster size */
+    } else {
+        s->l2_bits = s->cluster_bits - 3; /* L2 is always one cluster */
+    }
     s->l2_size = 1 << s->l2_bits;
     bs->total_sectors = header.size / 512;
     s->csize_shift = (62 - (s->cluster_bits - 8));
@@ -469,7 +473,7 @@ static int qcow2_open(BlockDriverState *bs, int flags)
     }
 
     /* alloc L2 table/refcount block cache */
-    s->l2_table_cache = qcow2_cache_create(bs, L2_CACHE_SIZE, s->cluster_size);
+    s->l2_table_cache = qcow2_cache_create(bs, L2_CACHE_SIZE, s->l2_size << 3);
     s->refcount_block_cache = qcow2_cache_create(bs, REFCOUNT_CACHE_SIZE,
                                                  s->cluster_size);
 
-- 
1.7.10.4