Re: [Qemu-block] [PATCH] vmdk: Set vmdk parent backing_format to vmdk

[Max Reitz]

On 11.04.19 12:43, Sam Eiderman wrote:
> Are you going to fix this when you apply the patch or should I?

I can’t fix it because Kevin has applied this patch to his branch.  I’d
prefer it if there weren’t commits that break patches (even if it’s just
temporary), so I think right now it’s mostly up to Kevin whether to
unstage the patch or fix it.

Maybe it makes sense for me to send a patch to fix it, then Kevin can
decide whether to squash it in.  (And if he doesn’t, we’ll at least have
the fix.)

Max

>> On 10 Apr 2019, at 23:07, Max Reitz <address@hidden> wrote:
>>
>> On 10.04.19 21:15, Max Reitz wrote:
>>> On 26.03.19 20:58, Sam Eiderman wrote:
>>>> Commit fb2105b ("vmdk: Support version=3 in VMDK descriptor files") fixed
>>>> the probe function to correctly guess vmdk descriptors with version=3.
>>>>
>>>> This solves the issue where vmdk snapshot with parent vmdk descriptor
>>>> containing "version=3" would be treated as raw instead vmdk.
>>>>
>>>> In the future case where a new vmdk version is introduced, we will again
>>>> experience this issue, even if the user will provide "-f vmdk" it will
>>>> only apply to the tip image and not to the underlying "misprobed" parent
>>>> image.
>>>>
>>>> The code in vmdk.c already assumes that the backing file of vmdk must be
>>>> vmdk (see vmdk_is_cid_valid which returns 0 if backing file is not
>>>> vmdk).
>>>>
>>>> So let's make it official by supplying the backing_format as vmdk.
>>>>
>>>> Reviewed-by: Mark Kanda <address@hidden>
>>>> Reviewed-By: Liran Alon <address@hidden>
>>>> Reviewed-by: Arbel Moshe <address@hidden>
>>>> Signed-off-by: Shmuel Eiderman <address@hidden>
>>>> ---
>>>> block/vmdk.c | 2 ++
>>>> 1 file changed, 2 insertions(+)
>>>
>>> Because this is not in master yet, maybe it’s still time for me to say
>>> that this breaks iotests 110 and 126 (requires a trivial fix to the
>>> reference output).
>>
>> Correction: It’s a bit more complicated because both tests should work
>> for other formats, too.  So I guess the backing file format should be
>> filtered from the output.
>>
>> Max
>>
>

signature.asc
Description: OpenPGP digital signature