Previouly, convert from ISO to VMDK with subformat=streamOptimized fails:
$ ./qemu-img convert -O vmdk -o subformat=streamOptimized foo.iso bar.vmdk
VMDK: can't write to allocated cluster for streamOptimized
qemu-img: error while writing sector 64: Input/output error
Because current code in qemu-img.c uses the normal convert loop, rather than
the compressed == true loop. In VMDK streamOptimized, writes must be in cluster
granularity, because overlapped write to an allocated cluster is -EIO.
This series adds is_compressed into BlockDriverInfo, and uses compressed
convertion loop if the target block driver sets this field to true.
It also implements .bdrv_get_info and .bdrv_write_compressed in VMDK driver to
fit into this framework.
Adds a test case to cover the code path in question: source image cluster size
is smaller.
v4: Rename .is_compressed -> .needs_compressed_writes. (Stefan)
has -> have in commit log in patch 3.
v3: Finally revisit and address Stefan's review comments on v2 from 2 months
ago. Sorry for being so slow on this one.
The general approach is the same, changes include:
- Split originial "[PATCH v2 5/5] mirror: Check for bdrv_get_info result"
and
send to list separately, because it's irrelevant with this series.
- Don't report cluster_size in bdrv_get_info if it's a flat extent, no need
to change type of BlockDriverInfo.cluster_size to 64 bits. So drop
"[PATCH v2 3/5] block: Change BlockDriverInfo.cluster_size to 64 bits".
- Add qemu-iotests case as [PATCH 4/4].
Fam Zheng (4):
qemu-img: Convert by cluster size if target is compressed
vmdk: Implement .bdrv_write_compressed
vmdk: Implement .bdrv_get_info()
qemu-iotests: Test converting to streamOptimized from small cluster
size
block/vmdk.c | 35 +++++++++++++++++++++++++++++++++++
include/block/block.h | 4 ++++
qemu-img.c | 1 +
tests/qemu-iotests/059 | 7 +++++++
tests/qemu-iotests/059.out | 8 ++++++++
5 files changed, 55 insertions(+)
--
1.9.2
