netdev - DSA driver support for RTL8367R

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [day] [month] [year] [list]

Message-ID: <CAJq09z6u+=zHPQT05t5pdXCZdTuy=zLiT9N8Yayc+MoFb1L5iA@mail.gmail.com>
Date: Thu, 18 Jan 2024 02:55:43 -0300
From: Luiz Angelo Daros de Luca <luizluca@...il.com>
To: "open list:NETWORKING DRIVERS" <netdev@...r.kernel.org>, Alvin Šipraga <ALSI@...g-olufsen.dk>, 
	Linus Walleij <linus.walleij@...aro.org>, Arınç ÜNAL <arinc.unal@...nc9.com>, 
	Christian Lamparter <chunkeey@...il.com>, Vladimir Oltean <olteanv@...il.com>
Subject: DSA driver support for RTL8367R

Hello,

While net-next is closed, I played with a device that uses RTL8367R.
It might not be a popular version, as most next product iterations
switched to RTL8367RB or something else.

It looks pretty close to the registers rtl8365mb uses, and I could
bring it up with some tweaks. It also reproduced the bug Christian
reported last year
(https://lore.kernel.org/netdev/802305c6-10b6-27e6-d154-83ee0abe3aeb@gmail.com/T/),
which we still need to fix someday (no supported devices affected). My
device also has LEDs that might help expand the LED support to
rtl8365mb. However, I could not enable the CPU tagger. I played with
two registers: RTL8365MB_CPU_PORT_MASK_REG 0x1219 (by default 0x0000)
and RTL8365MB_CPU_CTRL_REG 0x121A (by default 0x0002), but nothing
seems to be changing what I get in eth0: external packets never have
the CPU tag, and the switch does not react to packets with a CPU tag.

GPL packages from products that use the RTL8367R indicate that they
use the rtl8365b "vendor API." That API does have the functions to
enable the CPU tag touching the same registers I mentioned before. I
wonder if that switch does support CPU tagging. Does anyone have a
clue or access to docs/vendor support?

Even without the CPU tag, we could try the 802.1Q tagger. I didn't
find what limitations that kind of tag would introduce, but it might
be enough for some cases (or else ocelot-8021q wouldn't exist). An
802.1Q tagger would also be interesting for those devices that cannot
handle the checksum offload with an alien CPU tag.

Anyway, an 802.1Q tagger might require VLAN support, which rtl8365mb
does not have. Before digging into that, we need to fix that
limitation. Alvin, you were working on forward offloading/vlan
support. Could you share how it is?

Regards,

Luiz