Re: [Qemu-devel] [PATCH v3 2/2] qapi: Change BlockDirtyInfo to list

[Fam Zheng <famz@redhat.com>]

On 2013年11月14日 10:13, Wenchao Xia wrote:
> 于 2013/11/13 22:40, Fam Zheng 写道:
>> On Wed, Nov 13, 2013 at 10:19 PM, Kevin Wolf <kwolf@redhat.com> wrote:
>>> Am 13.11.2013 um 11:29 hat Fam Zheng geschrieben:
>>>> We have multiple dirty bitmaps in BDS now, switch QAPI to allow query
>>>> it (BlockInfo.dirty_bitmaps), and also drop old BlockInfo.dirty.
>>>>
>>>> Signed-off-by: Fam Zheng <famz@redhat.com>
>>>
>>>> diff --git a/qapi-schema.json b/qapi-schema.json
>>>> index 81a375b..931d710 100644
>>>> --- a/qapi-schema.json
>>>> +++ b/qapi-schema.json
>>>> @@ -948,8 +948,8 @@
>>>>   # @tray_open: #optional True if the device has a tray and it is open
>>>>   #             (only present if removable is true)
>>>>   #
>>>> -# @dirty: #optional dirty bitmap information (only present if the
>>>> dirty
>>>> -#         bitmap is enabled)
>>>> +# @dirty-bitmaps: #optional dirty bitmaps information (only present
>>>> if the
>>>> +#                 driver has one or more dirty bitmaps)
>>>>   #
>>>>   # @io-status: #optional @BlockDeviceIoStatus. Only present if the
>>>> device
>>>>   #             supports it and the VM is configured to stop on errors
>>>> @@ -963,7 +963,7 @@
>>>>     'data': {'device': 'str', 'type': 'str', 'removable': 'bool',
>>>>              'locked': 'bool', '*inserted': 'BlockDeviceInfo',
>>>>              '*tray_open': 'bool', '*io-status': 'BlockDeviceIoStatus',
>>>> -           '*dirty': 'BlockDirtyInfo' } }
>>>> +           '*dirty-bitmaps': ['BlockDirtyInfo'] } }
>>>>
>>>>   ##
>>>>   # @query-block:
>>>
>>> I believe this is of limited use; if you ever have more than one dirty
>>> bitmap, we're lacking information to associate it with the job it
>>> belongs to. One option would be to extend BlockDirtyInfo to indicate
>>> this, but another might be to actually extend other commands like
>>> query-block-jobs to return information on the dirty bitmap associated
>>> with a specific job.
>>>
>>> I've applied it to block-next anyway, we still have some time to
>>> reconsider for 1.8.
>>>
>>
>> Another case for this may be user enabled external dirty bitmap, which
>> could be standalone from any block job. E.g. when we introduce a QMP
>> command like:
>>
>> block-new-dirty-bitmap device=foo file=bar.bitmap
>>
>> This could be some code in block.c, could be a block job (really
>> necessary?), or a block filter. I'm not sure...
>>
>> Fam
>>
>    This command is for sure useful, but not quite a core block fuction,
> so hope it would not be in block.c. I think there is another problem
> need to solve: how to let user read "bar.bitmap", three options here:
> qemu-img dump, a new qmp command, a library. It seems a library is
> better(probably qmp interface is also needed).
>

Yes, block.c is certainly too ad hoc and I don't like it either.

I didn't go deep into this yet, I think we need to discuss these in a 
separate thread. Indeed there's much to decide: where the code lies, 
what format we would use and how to ensure consistency, etc.

Fam