Re: [PATCH] hw/char/mchp_pfsoc_mmuart: QOM'ify PolarFire MMUART

[Philippe Mathieu-Daudé]

On 9/23/21 12:52, Philippe Mathieu-Daudé wrote:
> On 9/23/21 12:41, Peter Maydell wrote:
> > On Thu, 23 Sept 2021 at 11:29, Philippe Mathieu-Daudé 
> > <f4bug@amsat.org> wrote:
> > > On 9/23/21 07:16, Bin Meng wrote:> On Sun, Sep 19, 2021 at 2:07 AM
> > > Philippe Mathieu-Daudé <f4bug@amsat.org> wrote:
> > > > > +static void mchp_pfsoc_mmuart_realize(DeviceState *dev, Error **errp)
> > > > > +{
> > > > > +    MchpPfSoCMMUartState *s = MCHP_PFSOC_UART(dev);
> > > > > +
> > > > > +    qdev_prop_set_uint8(DEVICE(&s->serial_mm), "regshift", 2);
> > > > > +    qdev_prop_set_uint32(DEVICE(&s->serial_mm), "baudbase", 399193);
> > > > > +    qdev_prop_set_uint8(DEVICE(&s->serial_mm), "endianness",
> > > > > +                        DEVICE_LITTLE_ENDIAN);
> > > >
> > > > It looks like serial_mm_init() does one more thing:
> > > >
> > > >       qdev_set_legacy_instance_id(DEVICE(smm), base, 2);
> > > >
> > > > I am not sure what that is.
> > >
> > > I'll defer on Paolo / Marc-André for that part, I think this is for
> > > migrating legacy (x86?) machines, which is not the case.
> >
> > Yes, this is only needed for backwards-compatibility of incoming
> > migration data with old versions of QEMU which implemented migration
> > of devices with hand-rolled code. If a device didn't previously
> > handle migration at all then it should not now be setting the
> > legacy instance ID.
>
> Thanks. I'll try to add that in the documentation somewhere.
>
> > Speaking of migration, I notice that this QOM conversion does
> > not add a vmstate, which usually we do as part of the QOM conversion.
> > The device is also missing a reset method.

Sigh, you reminded me of this series:
"hw: Mark devices with no migratable fields"
https://lore.kernel.org/qemu-devel/20210117192446.23753-19-f4bug@amsat.org/

> OK, I'll add that in a previous patch.
>
> Thanks,
>
> Phil.