On 24/01/2025 23:43, Daniel Golle wrote:
On Fri, Jan 24, 2025 at 11:03:28PM +0000, Chester A. Unal wrote:
On 24/01/2025 22:15, Hauke Mehrtens wrote:
@arinc9:
Currently we have the following known bugs:
[...]
* Ethernet link unstable on some mt7530 switches. Deactivate EEE
(Energy-Efficient Ethernet) as a workaround, see:
https://github.com/openwrt/openwrt/issues/17351
Any idea what could be going on there?
Here are the facts to consider:
We have only started receiving reports of the unstable link issue with the
24.10 branch.
I believe the MT753X DSA subdriver did not receive changes in years with
regards to EEE or any other parts of the code that would cause link
instability. I am giving this opinion as a maintainer of the said
subdriver.
There's a report by Florian [1], stating that the same issue appears with a
PHY that is not part of the MT7530 switch hardware. So the issue is
apparent on a system that does not have the MT7530 switch involved, and the
only common part of the reports is the MediaTek Ethernet driver.
Thanks for the quick summary.
Note that regarding Florian's report, on the WAX206 the RTL8221B-BV-VG
PHY is connected to port 5 of the MT7531AE switch [1] and hence NOT to the
MAC of MT7622 SoC. So it can of course still be an issue with that PHY
(which has been giving us some trouble in general), but I'd be very
confused if you insist that it would be caused by mtk_eth_soc, and would
ask you to explain why you think that would be the case.
Good point. I meant to say that the issue is apparent without the use of
the MT7530 switch PHYs. But as the switch MACs are still in use, we can't
completely factor out MT7530 and, therefore, the MT753X DSA subdriver.
My suspicion had been the EEE operation on the link between the switch
(operating as a PHY) and the MediaTek SoC MAC, controlled by the MediaTek
Ethernet driver. But the switch PHY in question is a simple digital logic
to deliver only the basic functions of a link and is not conformant to the
IEEE Std 802.3-2022 so I don't expect EEE to be available at all on that
link.
As all the reports involve the MT7530 switch MACs, the MediaTek Ethernet
driver must be irrelevant to this specific unstable link issue.
It could be a race condition issue that existed since the EEE support was
brought to the MT753X DSA subdriver and became more noticeable with newer
kernels as things slow down. Remember the PHY initialisation issue that was
experienced heavily on the TP-Link devices with the MT7621 SoC? We solved
it by ensuring the switch as a PHY is initialised first, then the switch
PHYs. It could be a similar case.
Chester A.
_______________________________________________
openwrt-devel mailing list
openwrt-devel@lists.openwrt.org
https://lists.openwrt.org/mailman/listinfo/openwrt-devel
