Re: [Qemu-devel] RFC v2: blockdev_add & friends, brief rationale, QMP docs

[Markus Armbruster]

Avi Kivity <address@hidden> writes:

> On 06/15/2010 05:54 PM, Markus Armbruster wrote:
>> Avi Kivity<address@hidden>  writes:
>>
>>    
>>> On 06/15/2010 04:27 PM, Markus Armbruster wrote:
>>>      
>>>>        
>>>>> I'm only talking about the interface, not the implementation.
>>>>> Internal design details shouldn't be exposed.
>>>>>
>>>>> For the implementation, I imagine you can create an empty blockdev
>>>>> during guest device creation and treat blockdev_add/blockdev_del as
>>>>> media change/eject.
>>>>>
>>>>>          
>>>> If blockdev_del only ejects media, then we need another command to get
>>>> rid of a blockdev.
>>>>
>>>>        
>>> No.  If you have a blockdev just to satisfy the guest device's need
>>> for a blockdev pointer, you can delete it automatically when the last
>>> reference goes.
>>>      
>> That's not how netdev and chardev behave.
>>    
>
> Ok.  To me duplicate argument lists suggest a lack of orthogonality, though.
>
> How about
>
>   blockdev_add id=...
>   media_attach blockdev=..., media-parameters
>   media_detach blockdev=...
>   blockdev_del id=...
>
> So blockdev_add/del define/remove a slot for the media,
> media_attach/detach connect it to media.

Yes, those are orthogonal operations.  Two more are side effects of
device_add and device_del: attach device model, detach device model.

My proposal provides two inessential operations purely for convenience:
define slot + connect media, and disconnect media + remove slot.  These
patterns are very common.

>>>> Unless you propose that blockdev_del merely ejects media if the blockdev
>>>> is being used by a device, but destroys the blockdev outright if not.
>>>> But that would be sick, wouldn't it?
>>>>
>>>>        
>>> Create a blockdev implicitly with guest device creation, and use
>>> blockdev_add (or media_attach) to attach the media.
>>>
>>> (but that creates a window where the guest device is visible but media
>>> is not yet inserted?)
>>>      
>> Think so, for hot plug.
>>
>> Actually, the device model driver would reject an empty blockdev, unless
>> it's a device with removable media, such as a CD-ROM.
>>
>> Having a device with fixed media go through a "no media yet" state
>> during initialization just complicates things.  Defining the media
>> *before* creating the device is much simpler.
>>    
>
> That is true.  Maybe we should just ignore the duplication.

It's fine to create empty blockdev, insert media, attach device model,
even if the device model wants fixed media.

Nevertheless, I think my convenience shortcuts make sense.