[Qemu-devel] [PATCH 1/2] qcow2: Clear L2 table cache after write error

[Qemu-devel] [PATCH 1/2] qcow2: Clear L2 table cache after write error

[Kevin Wolf]

If the L2 table was already updated in cache, but writing it to disk has
failed, we must not continue using the changed version in the cache to stay
consistent with what's on the disk.

Signed-off-by: Kevin Wolf <kwolf@redhat.com>
---
 block/qcow2-cluster.c |    1 +
 1 files changed, 1 insertions(+), 0 deletions(-)

diff --git a/block/qcow2-cluster.c b/block/qcow2-cluster.c
index c11680d..ed5c4b2 100644
--- a/block/qcow2-cluster.c
+++ b/block/qcow2-cluster.c
@@ -696,6 +696,7 @@ int qcow2_alloc_cluster_link_l2(BlockDriverState *bs, QCowL2Meta *m)
 
     ret = write_l2_entries(bs, l2_table, l2_offset, l2_index, m->nb_clusters);
     if (ret < 0) {
+        qcow2_l2_cache_reset(bs);
         goto err;
     }
 
-- 
1.6.6.1

[Qemu-devel] [PATCH 2/2] qcow2: Fix error handling in l2_allocate

[Kevin Wolf]

l2_allocate has some intermediate states in which the image is inconsistent.
Change the order to write to the L1 table only after the new L2 table has
successfully been initialized.

Also reset the L2 cache in failure case, it's very likely wrong.

Signed-off-by: Kevin Wolf <kwolf@redhat.com>
---
 block/qcow2-cluster.c |   23 +++++++++++++----------
 1 files changed, 13 insertions(+), 10 deletions(-)

diff --git a/block/qcow2-cluster.c b/block/qcow2-cluster.c
index ed5c4b2..244b4a7 100644
--- a/block/qcow2-cluster.c
+++ b/block/qcow2-cluster.c
@@ -239,14 +239,6 @@ static int l2_allocate(BlockDriverState *bs, int l1_index, uint64_t **table)
         return l2_offset;
     }
 
-    /* update the L1 entry */
-
-    s->l1_table[l1_index] = l2_offset | QCOW_OFLAG_COPIED;
-    ret = write_l1_entry(bs, l1_index);
-    if (ret < 0) {
-        return ret;
-    }
-
     /* allocate a new entry in the l2 cache */
 
     min_index = l2_cache_new_entry(bs);
@@ -261,7 +253,7 @@ static int l2_allocate(BlockDriverState *bs, int l1_index, uint64_t **table)
         ret = bdrv_pread(bs->file, old_l2_offset, l2_table,
             s->l2_size * sizeof(uint64_t));
         if (ret < 0) {
-            return ret;
+            goto fail;
         }
     }
     /* write the l2 table to the file */
@@ -269,7 +261,14 @@ static int l2_allocate(BlockDriverState *bs, int l1_index, uint64_t **table)
     ret = bdrv_pwrite(bs->file, l2_offset, l2_table,
         s->l2_size * sizeof(uint64_t));
     if (ret < 0) {
-        return ret;
+        goto fail;
+    }
+
+    /* update the L1 entry */
+    s->l1_table[l1_index] = l2_offset | QCOW_OFLAG_COPIED;
+    ret = write_l1_entry(bs, l1_index);
+    if (ret < 0) {
+        goto fail;
     }
 
     /* update the l2 cache entry */
@@ -279,6 +278,10 @@ static int l2_allocate(BlockDriverState *bs, int l1_index, uint64_t **table)
 
     *table = l2_table;
     return 0;
+
+fail:
+    qcow2_l2_cache_reset(bs);
+    return ret;
 }
 
 static int count_contiguous_clusters(uint64_t nb_clusters, int cluster_size,
-- 
1.6.6.1

Re: [Qemu-devel] [PATCH 2/2] qcow2: Fix error handling in l2_allocate

[Stefan Hajnoczi]

On Thu, May 20, 2010 at 3:48 PM, Kevin Wolf <kwolf@redhat.com> wrote:
> l2_allocate has some intermediate states in which the image is inconsistent.
> Change the order to write to the L1 table only after the new L2 table has
> successfully been initialized.

Looks good.

Stefan