[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [Qemu-devel] [PATCH] cpu: don't allow negative core id
From: |
Laurent Vivier |
Subject: |
Re: [Qemu-devel] [PATCH] cpu: don't allow negative core id |
Date: |
Wed, 2 Aug 2017 15:50:36 +0200 |
User-agent: |
Mozilla/5.0 (X11; Linux x86_64; rv:52.0) Gecko/20100101 Thunderbird/52.2.1 |
On 02/08/2017 15:42, Philippe Mathieu-Daudé wrote:
> Hi Laurent,
>
> On Wed, Aug 2, 2017 at 7:32 AM, Laurent Vivier <address@hidden> wrote:
>> With pseries machine type a negative core-id is not managed properly:
>> -1 gives an inaccurate error message ("core -1 already populated"),
>> -2 crashes QEMU (core dump)
>>
>> As it seems a negative value is invalid for any architecture,
>> instead of checking this in spapr_core_pre_plug() I think it's better
>> to check this in the generic part, core_prop_set_core_id()
>
> Why is this property signed? If there is not reason to use it negative,
> is it possible to use object_property_add(.."uint"..)?
You should be right:
{ 'struct': 'NumaNodeOptions',
'data': {
'*nodeid': 'uint16',
'*cpus': ['uint16'],
'*mem': 'size',
'*memdev': 'str' }}
but
{ 'struct': 'CpuInstanceProperties',
'data': { '*node-id': 'int',
'*socket-id': 'int',
'*core-id': 'int',
'*thread-id': 'int'
}
}
But I'm not sure it's a good idea to change the API now.
> Also what about core_prop_set_nr_threads()? It might coredump the
> same way.
In pseries case, there is another test in the spapr part that prevents that.
Thanks,
Laurent