OpenWRT 24.10 and problems with EEE on ramips/mt7621

[Mateusz Jończyk]

Chester A. Unal on Mon Apr 14 03:11:14 PDT 2025 wrote:*
*
> Hello Felix.
> >/sorry to bother you with this again. />/The WAX206 issue with it's Realtek WAN PHY has long been resolved but 
> the mt7621-ramips plattform is still causing issues for users on the 
> new release due to unstable/sometimes not working ethernet link. />//>/matjon just put some effort into debugging this (to some degree) />/<https://github.com/openwrt/openwrt/issues/17351#issuecomment-2798953937> 
> />//>/And HanabishiRecca noticed that EEE somehow gets enabled even when 
> the receiving side has it disabled. />/<https://github.com/openwrt/openwrt/issues/17351#issuecomment-2799921359> 
> />//>/Do their posts somehow shine a different light on the matter or do 
> you have an idea what could be at fault? /
> No, it doesn't shine a different light.
>
> >/Does EEE itself work fine but gets enabled when it shouldn't be? /
> I can't say for certain. I don't think it matters to do so anyway, for the
> issue at hand.
>
> >//>/I know you are maintaining only the mt7530 part and not all related 
> code so I don't expect you to change anything upstream. Though your 
> opinion on the matter would be very valuable to us here. :) /
> My opinion hasn't changed. I still suspect the same thing that I stated on
> the GitHub issue [1]:
>
> >/I'm going to refrain from disabling EEE for now as I see reports here /that suggest that it is not EEE that is the cause of the problem, but
> rather it reveals an issue of race condition of sorts when EEE is enabled.
> Perhaps the cause is in the MediaTek ethernet driver...

Hello,

I have been working on this issue and trying to debug it.

I have been following this bug report closely and I think that every 
credible
report there (that cannot be explained by another cause) has
been on ramips/mt7621.

That report by Florian Maurer [1] on Netgear WAX206 which you quoted
previously turned out to be a different issue and has already been fixed 
[2].

There was also a report from @raenye, which turned out to be
caused by a bad cable. @lorand-horvath described a problem on RT-AX59U,
(ARM / mediatek target) with the mt7531 external switch, however the
symptoms were very atypical (several flaps during a minute) and
happened only once.

There has also been a report by barcell1 of apparent hardware damage
just after flashing OpenWRT 24.10, which persisted after reverting to 23.05:

|[ 2.774046] mt7530-mdio mdio-bus:1f: reset timeout [ 2.778946] 
mt7530-mdio: probe of mdio-bus:1f failed with error -145|

> Get someone from MediaTek involved. I've got nothing else to add.

Back in 2021, Landen Chao from MediaTek described why disabling EEE is
recommended [4]:

W dniu 20.05.2021 o 09:13, Landen Chao pisze:
> On Thu, 2021-05-20 at 10:38 +0800, DENG Qingfang wrote:
>> On Thu, May 20, 2021 at 02:59:21AM +0200, Andrew Lunn wrote:
>>> > +static void mtk_gephy_config_init(struct phy_device *phydev) > +{ > + /* Disable EEE */ > + phy_write_mmd(phydev, MDIO_MMD_AN, 
>>> MDIO_AN_EEE_ADV, 0);
>>> Is EEE broken on this PHY? Or is this just to get it into a defined
>>> state?
>> As I said in commit message, the initialization (including EEE) is
>> from the vendor driver.
>> I have also tested it with EEE enabled by default on one of my APs,
>> and got occasional link drops.
>>
> EEE of the 10-year-old MT7530 internal gephy has many IOT problems, so
> it is recommended to disable its EEE.
> EEE of the MT7531 internal gephy has been updated, but it is not yet
> widely used.
> Therefore, EEE is disabled in vendor driver.
>
> Landen

Apparently the way EEE has been disabled for mt7530 stopped working
somewhere between Linux 5.15 and 6.6. The present way to disable EEE in
a PHY driver appears appears to be in drivers/net/phy/micrel.c:

 > phydev 
<https://elixir.bootlin.com/linux/v6.6.83/C/ident/phydev>->eee_broken_modes 
<https://elixir.bootlin.com/linux/v6.6.83/C/ident/eee_broken_modes>=-1;

I'm going to reconfirm at Mediatek, but in the end I will probably just
post a patch to mark EEE as broken in the devicetree.

Greetings,

Mateusz

[1] https://github.com/openwrt/openwrt/issues/17351#issuecomment-2608342255

[2] https://github.com/openwrt/openwrt/pull/17701

[3] https://github.com/openwrt/openwrt/issues/17351#issuecomment-2773185300

[4] 
https://lore.kernel.org/all/0adde34f936a2dafca40b06b408d82afe0852327.camel@mediatek.com/