Hi Greg,
On Thu, Jun 17, 2021 at 10:20 AM Jing Zhang <jingzhangos@xxxxxxxxxx> wrote:
>
> Hi Greg,
>
> On Thu, Jun 17, 2021 at 12:56 AM Greg KH <gregkh@xxxxxxxxxxxxxxxxxxx> wrote:
> >
> > On Thu, Jun 17, 2021 at 04:41:44AM +0000, Jing Zhang wrote:
> > > + struct kvm_stats_desc {
> > > + __u32 flags;
> > > + __s16 exponent;
> > > + __u16 size;
> > > + __u32 offset;
> > > + __u32 unused;
> > > + char name[0];
> > > + };
> >
> > <snip>
> >
> > > +The ``unused`` fields are reserved for future support for other types of
> > > +statistics data, like log/linear histogram.
> >
> > you HAVE to set unused to 0 for now, otherwise userspace does not know
> > it is unused, right? And then, really it is "used", so why not just say
> > that now? It's tricky, but you have to get this right now otherwise you
> > can never use it in the future.
> >
> Sure, will do that.
> > > +The ``name`` field points to the name string of the statistics data. The name
> >
> > It is not a pointer, it is the data itself.
> >
> Will fix it.
> > > +string starts at the end of ``struct kvm_stats_desc``.
> > > +The maximum length (including trailing '\0') is indicated by ``name_size``
> > > +in ``struct kvm_stats_header``.
> >
> > I thought we were replacing [0] arrays with [], are you sure you should
> > be declaring this as [0]? Same for all structures in this document (and
> > code).
> >
> The reason to declare it as [0] is to have the flexibility to change the maximum
> length of KVM stats name. For now, the max len is defined as 48, which can
> be read from the header. Then the userspace can get the length of descriptor by
> adding sizeof(struct_kvm_stats_desc) + 48. Whenever the max len is changed
> in KVM, the userspace would not have to update code to reflect that.
> However, if we are OK to restrict the maximum KVM stats' length to 48
> (or any other
> number), we can just declear it with [] instead of [0].
Please ignore my above comments.
You are right. We can just replace all zero-length arrays [0] with a
flexible array
member []. Thanks.
> > thanks,
> >
> > greg k-h
>
> Thanks,
> Jing
