[libvirt] [PATCH 07/23] cpu: Return model from virCPUModelIsAllowed

[Jiri Denemark]

On Thu, Oct 12, 2017 at 15:50:29 -0400, John Ferlan wrote:
> 
> 
> On 10/04/2017 10:58 AM, Jiri Denemark wrote:
> > If a given CPU model is supported by the hypervisor, we want to know
> > more about it, e.g., what features may block its usage on the current
> > host and such details are stored in the virDomainCapsCPUModelsPtr list
> > which virCPUModelIsAllowed uses to check whether the CPU model is
> > supported. Thus if the CPU model is found in the list we can directly
> > return a pointer to the corresponding virDomainCapsCPUModel if the
> > caller needs to look at the details.
> > 
> > Signed-off-by: Jiri Denemark <jdenemar at redhat.com>
> > ---
> >  src/cpu/cpu.c       | 18 ++++++++++++++----
> >  src/cpu/cpu.h       |  3 ++-
> >  src/cpu/cpu_ppc64.c |  2 +-
> >  src/cpu/cpu_x86.c   |  2 +-
> >  4 files changed, 18 insertions(+), 7 deletions(-)
> > 
> > diff --git a/src/cpu/cpu.c b/src/cpu/cpu.c
> > index b815ed383a..48290a471b 100644
> > --- a/src/cpu/cpu.c
> > +++ b/src/cpu/cpu.c
> > @@ -819,24 +819,34 @@ virCPUDataParse(const char *xmlStr)
> >   *
> >   * @model: CPU model to be checked
> >   * @models: list of supported CPU models
> > + * @hvModel: pointer to matching model from @models will be returned here
> 
> As later pointed out - 'hv' == hypervisor version.
> 
> In any case, the 'ModelIsAllowed" now feels overloaded returning more
> than just true/false that a typical "Is" type function would return.
> Sorry, I don't have suggestions, so unless someone else is looking and
> has a better name, then just go with it.

Hmm, I have a suggestion. I should just create a separate function
virDomainCapsCPUModelsGet or *Find which would just return a pointer to
the CPU model structure from the models array and change
virCPUModelIsAllowed into a simple wrapper around the new function. The
callers which would pass non-NULL hvModel pointer would just call the
underlying virDomainCapsCPUModelsGet function directly.

> The only other concern is that since you're returning a pointer in the
> middle of some array, any concerns over some other thread changing
> things and freeing what you're looking at?

No. If there were any concerns like this even going through the array
would be dangerous. Anyway, virDomainCapsCPUModelsPtr is a reference
counted object which never changes once it's created. It can only go
away and be replaced with another one.

Jirka