On Fri, Jan 24, 2025 at 03:29:43PM +0200, Tomi Valkeinen wrote:
> Hi,
>
> On 15/01/2025 16:29, Sakari Ailus wrote:
> > Moi,
> >
> > On Fri, Jan 10, 2025 at 11:14:18AM +0200, Tomi Valkeinen wrote:
> > > @@ -2956,6 +3033,36 @@ static int ub960_init_rx_ports_ub9702(struct ub960_data *priv)
> > > if (ret)
> > > return ret;
> > > + for_each_active_rxport(priv) {
> > > + if (it.rxport->ser.addr >= 0) {
> > > + /*
> > > + * Set serializer's I2C address if set in the dts file,
> > > + * and freeze it to prevent updates from the FC.
> > > + */
> > > + ub960_rxport_write(priv, it.nport, UB960_RR_SER_ID,
> > > + it.rxport->ser.addr << 1 |
> > > + UB960_RR_SER_ID_FREEZE_DEVICE_ID,
> > > + &ret);
> > > + }
> > > +
> > > + /* Set serializer I2C alias with auto-ack */
> > > + ub960_rxport_write(priv, it.nport, UB960_RR_SER_ALIAS_ID,
> > > + it.rxport->ser.alias << 1 |
> > > + UB960_RR_SER_ALIAS_ID_AUTO_ACK, &ret);
> > > +
> > > + if (ret)
> > > + return ret;
> > > + }
> > > +
> > > + for_each_active_rxport(priv) {
> > > + if (fwnode_device_is_compatible(it.rxport->ser.fwnode,
> > > + "ti,ds90ub971-q1")) {
> >
> > I guess one instance is fine but consider using match data instead.
>
> This is checking the model of the (remote) serializer, not the deserializer
> (which this driver deals with).
>
> Unfortunately we need to do some early configuration before the serializer
> driver probes, and that configuration depends on the serializer model.
>
> We could, perhaps, get the serializer models once and store it in the rxport
> data for later use.
Ack.
--
Sakari Ailus
