Re: [PATCH v3 4/9] qapi/misc: Restrict balloon-related commands to machine code

[Markus Armbruster <armbru@redhat.com>]

Philippe Mathieu-Daudé <philmd@redhat.com> writes:

> On 5/26/20 11:31 AM, Philippe Mathieu-Daudé wrote:
>> +Laurent
>> 
>> On 5/26/20 11:04 AM, Markus Armbruster wrote:
>>> Philippe Mathieu-Daudé <philmd@redhat.com> writes:
>>>
>>>> On 5/26/20 9:38 AM, Markus Armbruster wrote:
>>>>> Philippe Mathieu-Daudé <philmd@redhat.com> writes:
>>>>>
>>>>>> Signed-off-by: Philippe Mathieu-Daudé <philmd@redhat.com>
>>>>>
>>>>> A brief note on why restricting "to machine code" is useful would be
>>>>> nice.  Same for the other patches.
>>>>>
>>>>> Acked-by: Markus Armbruster <armbru@redhat.com>
>>>>>
>>>>
>>>> What about this?
>>>>
>>>> "QEMU can do system-mode and user-mode emulation.
>>>> Only system mode emulate a machine.
>>>> Remove this feature from the user-mode emulation."
>>>
>>> Is is a feature of user-mode emulation before the patch?  Or is it just
>>> dead code?
>>>
>>> Hint: QMP commands tend to be dead code when the program doesn't expose
>>> a QMP monitor :)
>> 
>> Maybe a 'corollary' question, "How user-mode users use QMP?"
>> 
>
> I can't find a way to start a user-mode process with a QMP socket, is
> there one?

As far as I can tell, only qemu-system-FOO and qemu-storage-daemon
provide QMP monitors:

* Monitors need to be initialized with monitor_init().  Two callers:
  monitor_init_hmp(), monitor_init_qmp().

* monitor_init() calls both, and is the common wat to create a monitor.
  Called by vl.c via monitor_init_opts(), and by qemu-storage-daemon.

* monitor_init_hmp() has additional callers, but HMP doesn't matter
  here.