[PATCH net-next v2 07/15] net: dsa: mt7530: move MT753X_MTRAP operations for MT7530

[Arınç ÜNAL]

On 27.04.2024 05:24, Daniel Golle wrote:
> Hi Arınç,
> 
> On Mon, Apr 22, 2024 at 10:15:14AM +0300, Arınç ÜNAL via B4 Relay wrote:
>> From: Arınç ÜNAL <arinc.unal at arinc9.com>
>>
>> On MT7530, the media-independent interfaces of port 5 and 6 are controlled
>> by the MT7530_P5_DIS and MT7530_P6_DIS bits of the hardware trap. Deal with
>> these bits only when the relevant port is being enabled or disabled. This
>> ensures that these ports will be disabled when they are not in use.
>>
>> Do not set MT7530_CHG_TRAP on mt7530_setup_port5() as that's already being
>> done on mt7530_setup().
> 
> Multiple users reported ([1], [2]) that after I've imported the series
> to OpenWrt they noticed that WAN connection on MT7621 boards using
> PHY-muxing to hook up either port 0 or port 4 to GMAC1 no longer works.
> 
> The link still seems to come up, but no data flows. I went ahead and
> confirmed the bug, then started bisecting the patches of this series,
> and ended up identifying this very patch being the culprit.
> 
> I can't exclude that what ever the issue may be is caused by other
> downstream patches we have, but can confirm that removing this patch of
> your series [3] in OpenWrt fixes the issue. Please take a look and as
> the cover letter states you have tested this on some MT7621 board,
> please make sure traffic actually flows on the PHY-muxed port on that
> board after this patch is applied, and if not, please figure out why and
> repost a fixed version of this patch.
> 
> 
> Cheers
> 
> 
> Daniel
> 
> [1]: https://github.com/openwrt/openwrt/issues/15273
> [2]: https://github.com/openwrt/openwrt/issues/15279
> [3]: https://git.openwrt.org/?p=openwrt/openwrt.git;a=commit;h=a8dde7e5bd6d289db6485cf57d3512ea62eaa827

Thanks for reporting this Daniel. I am not happy that I've caused all this
fuss. My testing as described on the cover letter did not include the
hardware design with PHY muxing. Lesson learned; next time, I'll make sure
to test the specific hardware design when I work on the part of the code
that would affect that hardware design.

That said, I've submitted a patch that fixes this issue [1]. I have tested
the hardware design with PHY muxing with this fix applied and I don't
experience this issue anymore.

[1] https://lore.kernel.org/netdev/20240427-for-netnext-mt7530-do-not-disable-port5-when-phy-muxing-v1-1-793cdf9d7707@arinc9.com/

Arınç