Re: [PATCH v2 17/18] ppc/pnv: Introduce user creatable pnv-phb4 devices

[Daniel Henrique Barboza]

On 1/8/22 08:11, Cédric Le Goater wrote:
> > > > +        object_property_set_int(OBJECT(phb), "index",
> > > > +                                phb->phb_id, &error_abort);
> > > > +
> > > > +        pnv_phb4_set_stack_phb_props(stack, phb);
> > > > +
> > > > +        /* Assign the phb to the stack */
> > > > +        stack->phb = phb;
> > >
> > > The stack object should check the validity of the stack->phb pointer always.
> >
> >
> > What do you mean by "check the validity"?
>
> I am thinking of the usage of 'stack->phb', for instance in the routine
> pnv_phb4_update_regions(). We should add an assert there.

Oh ok. I'll add an assert in this point (and in the code above that where you
also pointed it out.

And about pnv_phb4_update_regions(), v3 will need to handle it but not with
an assert but with a phb != NULL verification to become a NO-OP if phb is NULL.
It's an easier and less intrusive fix than trying to change the order of the
XSCOM initialization in stk_realize().

I will (hopefully) explain it better in the v3.


Thanks,


Daniel

> Your changes seem to cleanup the stack <-> phb relation quite a lot. Which
> is good.
>
> Thanks,
>
> C.