[Qemu-devel] [PATCH] vmdk: Fix next_cluster_sector for compressed write

[Qemu-devel] [PATCH] vmdk: Fix next_cluster_sector for compressed write

[Fam Zheng]

This fixes the bug introduced by commit c6ac36e (vmdk: Optimize cluster
allocation).

Sometimes, write_len could be larger than cluster size, because it
contains both data and marker.  We must advance next_cluster_sector in
this case, otherwise the image gets corrupted.

Reported-by: Antoni Villalonga <qemu-list@friki.cat>
Signed-off-by: Fam Zheng <famz@redhat.com>
---
 block/vmdk.c | 14 ++++++++++----
 1 file changed, 10 insertions(+), 4 deletions(-)

diff --git a/block/vmdk.c b/block/vmdk.c
index 1c5e2ef..4b4a862 100644
--- a/block/vmdk.c
+++ b/block/vmdk.c
@@ -1302,6 +1302,8 @@ static int vmdk_write_extent(VmdkExtent *extent, int64_t cluster_offset,
     uLongf buf_len;
     const uint8_t *write_buf = buf;
     int write_len = nb_sectors * 512;
+    int64_t write_offset;
+    int64_t write_end_sector;
 
     if (extent->compressed) {
         if (!extent->has_marker) {
@@ -1320,10 +1322,14 @@ static int vmdk_write_extent(VmdkExtent *extent, int64_t cluster_offset,
         write_buf = (uint8_t *)data;
         write_len = buf_len + sizeof(VmdkGrainMarker);
     }
-    ret = bdrv_pwrite(extent->file,
-                        cluster_offset + offset_in_cluster,
-                        write_buf,
-                        write_len);
+    write_offset = cluster_offset + offset_in_cluster,
+    ret = bdrv_pwrite(extent->file, write_offset, write_buf, write_len);
+
+    write_end_sector = DIV_ROUND_UP(write_offset + write_len, BDRV_SECTOR_SIZE);
+
+    extent->next_cluster_sector = MAX(extent->next_cluster_sector,
+                                      write_end_sector);
+
     if (ret != write_len) {
         ret = ret < 0 ? ret : -EIO;
         goto out;
-- 
1.9.3

Re: [Qemu-devel] [PATCH] vmdk: Fix next_cluster_sector for compressed write

[Kevin Wolf]

Am 10.09.2015 um 10:58 hat Fam Zheng geschrieben:
> On Thu, 09/10 11:48, Radoslav Gerganov wrote:
> > On 10.09.2015 11:15, Fam Zheng wrote:
> > > On Thu, 09/10 10:53, Radoslav Gerganov wrote:
> > >> When the VMDK is streamOptimized (or compressed), the
> > >> next_cluster_sector must not be incremented by a fixed number of
> > >> sectors. Instead of this, it must be rounded up to the next consecutive
> > >> sector. Fixing this results in much smaller compressed images.
> > >>
> > >> Signed-off-by: Radoslav Gerganov <rgerganov@vmware.com>
> > > 
> > > Is this patch enough to produce a valid streamOptimized image? Because I
> > > remember there were reports on some other issues with footer or grain markers.
> > > 
> > > Fam
> > 
> > The VMDK spec says that each marker and its associated block must begin on
> > sector boundary which is true before and after my patch.  However, with my
> > patch we don't put empty sectors when we achieve good compression for the
> > current cluster but start the next cluster on the next consecutive sector.
> > 
> > I have also verified that images produced with my patch can be booted in
> > VMware Workstation.
> 
> Thanks, the change looks good to me.
> 
> Reviewed-by: Fam Zheng <famz@redhat.com>

Thanks, applied to the block branch.

Radoslav, for your next patch please remember to always keep qemu-devel
CCed, even if you're sending the patch to a more specific mailing list.
I'm adding it to this mail so that people who aren't subscribed to
qemu-block have a chance to notice the patch.

Kevin