Re: [PATCH/RFC net] net: dsa: lantiq_gswip: honor dsa_db passed to port_fdb_{add,del}

[Daniel Golle <daniel@makrotopia.org>]

On Mon, Aug 11, 2025 at 06:32:42PM +0300, Vladimir Oltean wrote:
> Hi Daniel,
> 
> On Mon, Aug 11, 2025 at 12:02:00AM +0300, Vladimir Oltean wrote:
> > I suggest tools/testing/selftests/net/forwarding/local_termination.sh
> > once dsa_switch_supports_uc_filtering() returns true.
> 
> Since you're working with the lantiq_gswip driver which receives
> relatively few patches...
> 
> I would like to submit this patch to remove the legacy behavior:
> 
> diff --git a/drivers/net/dsa/lantiq_gswip.c b/drivers/net/dsa/lantiq_gswip.c
> index 6b8a5101b0e7..7e11f198ff2b 100644
> --- a/drivers/net/dsa/lantiq_gswip.c
> +++ b/drivers/net/dsa/lantiq_gswip.c
> @@ -886,8 +886,6 @@ static int gswip_setup(struct dsa_switch *ds)
> 
>  	ds->mtu_enforcement_ingress = true;
> 
> -	ds->configure_vlan_while_not_filtering = false;
> -
>  	return 0;
>  }
> 
> however I'm sure that the driver will break, so I have more, in an
> attempt to avoid that :)

This honourable endeavour deserves my full support :)

> 
> Would you please look at the patches I've prepared on this branch and
> reviewing with extra info you might have / giving them a test, one by one?
> I was only able to compile-test them. I also lack proper documentation
> (which I'm sure you lack too), I only saw the "developer resources" from
> Martin Blumenstingl's Github (which lack actual PCE register descriptions)
> https://github.com/xdarklight/ltq-upstream-status
> and the Maxlinear PRPLOS code at
> https://github.com/maxlinear/linux/tree/UPDK_9.1.90/drivers/net/datapath/gswip/switchcore/src
> (which I think is what you were also referencing)

There is also a DSA driver for the newer standalone GSW1xx switch ICs
which are connected to a CPU using either MDIO or SPI for management and
2500Base-X/SGMII or RGMII/RMII for the datapath. I'm working on merging
that with the existing lantiq_gswip driver as those newer switch ICs have
a lot in common with the older Lantiq/Intel in-SoC switches.

You can see my work-in-progress first cleaning up the reference mxl-gsw1xx
driver and then preparing the lantiq_gswip driver for a potential merge
with that driver here:

https://github.com/dangowrt/linux/commits/mxl-gsw1xx-cleanup/

The old Lantiq VRV20x turned out to be difficult to even undergo proper testing
using tools/test/selftest/drivers/net/dsa scripts due to most boards coming with
only 64 MiB or DDR2 RAM -- even with nothing else running and the root
filesystem on-flash tcpdump quickly causes oom when trying to run
local_termination.sh.

In the next days I'm going to receive two boards with 256 MiB of RAM which
I found used for little money on ebay which will allow me to at least test
GSWIP 2.1 (xrx200) and GSWIP 2.2 (xrx330).
I don't think any of those boards comes with an old enough xrx200 SoC to
still contain the GSWIP 2.0 IP...

> 
> The branch over net-next is here:
> https://github.com/vladimiroltean/linux/commits/lantiq-gswip/

Nice, I'll happily take a look tomorrow and can do basic testing on my
TL-W8970 (VRV200, but only 64 MiB of DDR2 RAM), and hopefully the full
DSA selftest suite (or rather the part of it which covers features
actually supported by the driver/hardware) starting with
local_termination.sh once I receive the more resourceful boards with
VRV200 and VRV330 in the next couple of days.