Patchwork [RFC] qcow2: group refcount updates during cow

login
register
mail settings
Submitter Frediano Ziglio
Date July 28, 2011, 1:50 p.m.
Message ID <1311861017-13425-1-git-send-email-freddy77@gmail.com>
Download mbox | patch
Permalink /patch/107259/
State New
Headers show

Comments

Frediano Ziglio - July 28, 2011, 1:50 p.m.
Well, I think this is the first real improve patch.
Is more a RFC than a patch. Yes, some lines are terrible!
It collapses refcount decrement during cow.
From a first check time executing 015 test passed from about 600 seconds
to 70.
This at least prove that refcount updates counts!
Some doubt:
1- place the code in qcow2-refcount.c as it update only refcount and not
  cluster?
2- allow some sort of "begin transaction" / "commit" / "rollback" like 
  databases instead?
3- allow changing tables from different coroutines?

1) If you have a sequence like (1, 2, 4) probably these clusters are all in
the same l2 table but with this code you get two write instead of one.
I'm thinking about a function in qcow2-refcount.c that accept an array of cluster
instead of a start + len.

Signed-off-by: Frediano Ziglio <freddy77@gmail.com>
---
 block/qcow2-cluster.c |   36 ++++++++++++++++++++++++++++++++++--
 1 files changed, 34 insertions(+), 2 deletions(-)

Patch

diff --git a/block/qcow2-cluster.c b/block/qcow2-cluster.c
index 81cf77d..da17365 100644
--- a/block/qcow2-cluster.c
+++ b/block/qcow2-cluster.c
@@ -675,10 +675,42 @@  int qcow2_alloc_cluster_link_l2(BlockDriverState *bs, QCowL2Meta *m)
      * Also flush bs->file to get the right order for L2 and refcount update.
      */
     if (j != 0) {
+        int64_t old_start = 0, old_end = -2;
+        int count = 0;
         for (i = 0; i < j; i++) {
-            qcow2_free_any_clusters(bs,
-                be64_to_cpu(old_cluster[i]) & ~QCOW_OFLAG_COPIED, 1);
+            old_cluster[i] = be64_to_cpu(old_cluster[i]) & ~QCOW_OFLAG_COPIED;
         }
+        // XXX sort old_cluster
+        for (i = 0; i < j; i++) {
+            int64_t cluster = old_cluster[i];
+
+            /* group if contiguos */
+            if (old_end + 1 == (cluster >> s->cluster_bits)) {
+                ++old_end;
+                continue;
+            }
+
+            /* handle */
+            if (old_end > 0) {
+                qcow2_free_any_clusters(bs, old_start << s->cluster_bits, old_end - old_start + 1);
+                count += old_end - old_start + 1;
+            }
+            old_end = -2;
+
+            /* handle compressed separately */
+            if ((cluster & QCOW_OFLAG_COMPRESSED)) {
+                qcow2_free_any_clusters(bs, cluster, 1);
+                continue;
+            }
+
+            /* start a new group */
+            old_start = old_end = cluster >> s->cluster_bits;
+        }
+        if (old_end > 0) {
+            qcow2_free_any_clusters(bs, old_start << s->cluster_bits, old_end - old_start + 1);
+            count += old_end - old_start + 1;
+        }
+        assert(count == j);
     }
 
     ret = 0;