Re: [PATCH net-next 1/4 v4] net: dsa: rtl8366rb: Support disabling learning

On 10/5/21 4:07 PM, Linus Walleij wrote:

On Tue, Oct 5, 2021 at 9:59 AM Alvin Šipraga [off-list ref] wrote:

quoted

On 10/4/21 10:57 PM, Linus Walleij wrote:

quoted

quoted

BTW: all the patches i have left are extensions to RTL8366RB
specifically so I think it should be fine for you to submit patches
for your switch on top of net-next, maybe we can test this
on you chip too, I suspect it works the same on all Realtek
switches?

Generally speaking I don't think that the patches you have sent for 66RB
are particularly relevant for the 65MB because the register layout and
some chip semantics are totally different.

I was mainly thinking about the crazy VLAN set-up that didn't use
port isolation and which is now deleted. But  maybe you were not
using the rtl8366.c file either? Just realtek-smi.c?

Ah, I was just not using those particularly freaky VLAN functions from 
the rtl8366 library. I still use vlan_{add,del} and the MIB counter 
helpers though, as these seem to be OK. I have been keeping up-to-date 
with your changes to rtl8366.c and tested them locally with my subdriver 
and they are working like a charm. So we should still benefit from some 
level of code reuse.

	Alvin

quoted

Regarding CPU port learning
for the RTL8365MB chip: right now I am playing around with the "third
way" Vladimir suggested, by enabling learning selectively only for
bridge-layer packets (skb->offload_fwd_mark == true). To begin with I'm
not even sure if you have this capability with the RTL8366RB.

This will be interesting to see!

Yours,
Linus Walleij