Re: [PATCH net-next v2 05/10] net: dsa: realtek: rtl8365mb: describe the ports of each chip
From: Oleksij Rempel
Date: Mon Nov 16 2026 - 08:36:33 EST
On Thu, Nov 12, 2026 at 04:08:47PM +0100, Alvin Šipraga wrote:
> On Thu, Nov 12, 2026 at 09:37:48AM +0100, Oleksij Rempel wrote:
> > Record which ports each chip has, report one more than its highest
> > port to DSA as the port count, and reject a devicetree port the chip
> > does not have. Lower the family maximum to 8 to match.
> >
> > Commit ca5ecd4246d4 ("net: dsa: realtek: rtl8365mb: correct the max
> > number of ports") raised the port count of all chips to 11, but no
> > supported chip has a port above 7.
> >
> > Since commit 51a4a22301b0 ("net: dsa: realtek: rtl8365mb: use dsa
> > helpers for port iteration"), setup also programs unused ports, so
> > every probe writes per-port registers of ports 8..10. Their learn-limit
> > writes go to 0x0A28..0x0A2A, above the learn-limit registers of ports
> > 0..7. We have not seen a visible effect from these writes.
>
> I'm not quite sure what the point of the patch is (sorry, maybe it's
> clear later in the series, but I didn't get that far yet). Was the setup
> of nonexistent ports clobbering some other registers?
Not that I could see. I saw no effect from the writes to ports 8..10.
The reason is the rest of the series. The QoS code in the later
patches handles ports 0..7 only. With 11 ports, DSA accepts a
devicetree port 8..10, and the QoS code would then write a register
that does not hold that port's field. With the per-chip port mask, the
driver no longer accepts a port the chip does not have. I reworded the
commit message to say this.
> Also in the blamed commit there was:
>
> -/* RTL8370MB and RTL8310SR, possibly suportable by this driver, have 10 ports */
> -#define RTL8365MB_MAX_NUM_PORTS 10
> +#define RTL8365MB_MAX_NUM_PORTS 11
>
> So maybe it needs bumping in the future if support is added for those
> chips. That's why I'm wondering if you saw something off?
Good point. v3 drops that.
> > dsa_switch_for_each_port(dp, ds) {
> > + if (dsa_port_is_unused(dp))
> > + continue;
>
> You can use dsa_switch_for_each_available_port() here.
Done in v3.
> > +
> > + if (!(mb->chip_info->port_mask & BIT(dp->index))) {
> > + dev_err(priv->dev, "port %d does not exist on %s\n",
> > + dp->index, mb->chip_info->name);
> > + ret = -EINVAL;
> > + goto out_teardown_irq;
> > + }
> > +
Thanks,
Oleksij