Re: [Qemu-devel] [PATCH v4 08/11] numa: Extend the command-line to provide memory latency and bandwidth information

[Igor Mammedov <imammedo@redhat.com>]

On Thu, 6 Jun 2019 08:23:47 -0500
Eric Blake <eblake@redhat.com> wrote:

> On 6/6/19 2:47 AM, Tao Xu wrote:
> > On 6/5/2019 10:40 PM, Igor Mammedov wrote:  
> >> On Wed,  8 May 2019 14:17:23 +0800
> >> Tao Xu <tao3.xu@intel.com> wrote:
> >>  
> >>> From: Liu Jingqi <jingqi.liu@intel.com>
> >>>
> >>> Add -numa hmat-lb option to provide System Locality Latency and
> >>> Bandwidth Information. These memory attributes help to build
> >>> System Locality Latency and Bandwidth Information Structure(s)
> >>> in ACPI Heterogeneous Memory Attribute Table (HMAT).
> >>>
> >>> Signed-off-by: Liu Jingqi <jingqi.liu@intel.com>
> >>> Signed-off-by: Tao Xu <tao3.xu@intel.com>
> >>> ---  
> > ...  
> >>> +##
> >>> +{ 'struct': 'NumaHmatLBOptions',
> >>> +  'data': {
> >>> +   'initiator': 'uint16',
> >>> +   'target': 'uint16',
> >>> +   'hierarchy': 'HmatLBMemoryHierarchy',
> >>> +   'data-type': 'HmatLBDataType',  
> >> I think union will be better here with data-type used as discriminator,
> >> on top of that you'll be able to drop a bit of error checking above since
> >> QAPI's union will not allow user to mix latency and bandwidth.
> >>  
> > Hi Igor,
> > 
> > I have quesion here, the 'hmat-lb' is a member of a union 'NumaOptions',
> > it seems can' use a union as a member of union.  
> 
> It should be technically possible to expand the QAPI generators to allow
> one union as a branch within another union, so long as there are no
> collisions in identifiers, if that makes for the smartest on-the-wire
> representation.

It would save quite a bit of boiler plate error checking in numa code,
but since I don't know much about QAPI to make meaningful suggestion
how to implement it, I won't insist on using union.