Re: [Qemu-devel] [PATCH 3/3] add cpu_set qmp command

[Igor Mammedov <imammedo@redhat.com>]

On 01/19/2012 11:24 AM, Jan Kiszka wrote:
> On 2012-01-19 10:38, Igor Mammedov wrote:
>> On 01/17/2012 03:18 PM, Jan Kiszka wrote:
>>> On 2012-01-17 14:17, Igor Mammedov wrote:
>>>> Signed-off-by: Igor Mammedov<imammedo@redhat.com>
>>>> ---
>>>>    qapi-schema.json |    9 +++++++++
>>>>    qmp-commands.hx  |   26 ++++++++++++++++++++++++++
>>>>    qmp.c            |   15 +++++++++++++++
>>>>    3 files changed, 50 insertions(+), 0 deletions(-)
>>>>
>>>> diff --git a/qapi-schema.json b/qapi-schema.json
>>>> index 44cf764..05cc582 100644
>>>> --- a/qapi-schema.json
>>>> +++ b/qapi-schema.json
>>>> @@ -903,6 +903,15 @@
>>>>    { 'command': 'cpu', 'data': {'index': 'int'} }
>>>>
>>>>    ##
>>>> +# @cpu_set
>>>> +#
>>>> +# Sets specified cpu to online/ofline mode
>>>> +#
>>>> +# Notes: semantics is : cpu_set x online|offline
>>>> +##
>>>> +{ 'command': 'cpu_set', 'data': {'cpu_index': 'int', 'status': 'str'} }
>>>> +
>>>> +##
>>>>    # @memsave:
>>>>    #
>>>>    # Save a portion of guest memory to a file.
>>>> diff --git a/qmp-commands.hx b/qmp-commands.hx
>>>> index 7e3f4b9..ef1ac1e 100644
>>>> --- a/qmp-commands.hx
>>>> +++ b/qmp-commands.hx
>>>> @@ -348,6 +348,32 @@ Note: CPUs' indexes are obtained with the 'query-cpus' command.
>>>>    EQMP
>>>>
>>>>        {
>>>> +        .name       = "cpu_set",
>>>> +        .args_type  = "cpu_index:i,status:s",
>>>> +        .mhandler.cmd_new = qmp_marshal_input_cpu_set,
>>>> +    },
>>>> +
>>>> +SQMP
>>>> +cpu_set
>>>> +-------
>>>> +
>>>> +Sets virtual cpu to online/ofline state
>>>> +
>>>> +Arguments:
>>>> +
>>>> +- "cpu_index": virtual cpu index (json-int)
>>>> +- "status": desired state of cpu, online/offline (json-string)
>>>> +
>>>> +Example:
>>>> +
>>>> +->   { "execute": "cpu_set",
>>>> +             "arguments": { "cpu_index": 2,
>>>> +                            "status": "online" } }
>>>> +<- { "return": {} }
>>>> +
>>>> +EQMP
>>>> +
>>>> +    {
>>>>            .name       = "memsave",
>>>>            .args_type  = "val:l,size:i,filename:s,cpu:i?",
>>>>            .mhandler.cmd_new = qmp_marshal_input_memsave,
>>>> diff --git a/qmp.c b/qmp.c
>>>> index c74dde6..e2b268d 100644
>>>> --- a/qmp.c
>>>> +++ b/qmp.c
>>>> @@ -101,6 +101,21 @@ void qmp_cpu(int64_t index, Error **errp)
>>>>        /* Just do nothing */
>>>>    }
>>>>
>>>> +void qmp_cpu_set(int64_t cpu_index, const char *status, Error **errp)
>>>> +{
>>>> +    int state;
>>>> +
>>>> +    if (!strcmp(status, "online")) {
>>>> +        state = 1;
>>>> +    } else if (!strcmp(status, "offline")) {
>>>> +        state = 0;
>>>> +    } else {
>>>> +        error_set(errp, QERR_INVALID_PARAMETER, status);
>>>> +        return;
>>>> +    }
>>>> +    qemu_system_cpu_hot_add(cpu_index, state);
>>>> +}
>>>> +
>>>>    #ifndef CONFIG_VNC
>>>>    /* If VNC support is enabled, the "true" query-vnc command is
>>>>       defined in the VNC subsystem */
>>>
>>> This shouldn't go upstream. We rather need qdev'ified CPUs that can be
>>> added and removed as any other device.
>>>
>>
>> Jan,
>>
>> Thanks for review!
>> Then I'll drop this patch and re-post the other ones after fixing them.
>
> Well, you will still need the control logic of this patch, thus need to
> think about converting x86 CPUs to qdev. Otherwise, testing will only be
> possible over qemu-kvm.
>
> Jan
>

I've agree that converting x86 CPUs to qdev is needed. I've seen RFC
patches for qev-ifying ppc CPU but they haven't made to upstream yet. I'll
look at them as model for x86 CPU conversion but I can't promise fast
results here (I'm complete newbie in this). I'm open to suggestions if
you have a better idea how to do it or a better example.

I think there is sense in having in the tree the first 2 patches separately
from qdev-ifying cpu patches, since they provide a basic infrastructure
for vcpu hot-plug and we will reduce difference between qemu and qemu-kvm
in this parts of code. In addition, it will open a road for hot-unplug
patches that people send now against qemu-kvm.

It still will be possible to play with hot-plug using 3rd patch off the tree
while qdev-ifed cpu patch(es) in works.

-- 
Thanks,
  Igor