On Mon, 14 Oct 2024 at 19:50, Bernhard Beschow <shen...@gmail.com> wrote:
>
>
>
> Am 14. Oktober 2024 12:47:52 UTC schrieb Peter Maydell
> <peter.mayd...@linaro.org>:
> >> +typedef struct Lan9118PhyState {
> >> + uint32_t status;
> >> + uint32_t control;
> >> + uint32_t advertise;
> >> + uint32_t ints;
> >> + uint32_t int_mask;
> >> + IRQState irq;
> >> + bool link_down;
> >> +} Lan9118PhyState;
> >
> >This takes state that was in a QOM object, and moves it
> >into something that's kind of a device but not a QOM
> >object. I think we should avoid that, because at some
> >point somebody's going to have to QOMify this.
> >
> >Making this a QOM device is a bit awkward for migration
> >compatibility, unfortunately.
>
> Do we care about migration compatibility here? Or is it
> sufficient to check the version? In the latter case I could
> QOMify it.
Doing a quick grep it looks like the lan9118 is only
used in a set of Arm boards and none of them are ones where
we care about migration across versions. So I think we're
ok to break compat with a version-bump. We should mention
the affected boards in the commit message.
-- PMM
