RE: Broken networking on iMX8QXP after suspend after upgrading from 5.10 to 6.1

From: Wei Fang <wei.fang@nxp.com>
To: John Ernberg <john.ernberg@actia.se>,
	"netdev@vger.kernel.org" <netdev@vger.kernel.org>
Cc: Jonas Blixt <jonas.blixt@actia.se>,
	Shenwei Wang <shenwei.wang@nxp.com>,
	Clark Wang <xiaoning.wang@nxp.com>, Andrew Lunn <andrew@lunn.ch>,
	Heiner Kallweit <hkallweit1@gmail.com>,
	Russell King <linux@armlinux.org.uk>
Subject: RE: Broken networking on iMX8QXP after suspend after upgrading from 5.10 to 6.1
Date: Wed, 28 Feb 2024 08:34:45 +0000	[thread overview]
Message-ID: <AM5PR04MB31391FBF10F18312F0E87C6F88582@AM5PR04MB3139.eurprd04.prod.outlook.com> (raw)
In-Reply-To: <521d30d8-91b5-414f-93bd-19f86bba4aa0@actia.se>

> -----Original Message-----
> From: John Ernberg <john.ernberg@actia.se>
> Sent: 2024年2月28日 15:59
> To: Wei Fang <wei.fang@nxp.com>; netdev@vger.kernel.org
> Cc: Jonas Blixt <jonas.blixt@actia.se>; Shenwei Wang
> <shenwei.wang@nxp.com>; Clark Wang <xiaoning.wang@nxp.com>; Andrew
> Lunn <andrew@lunn.ch>; Heiner Kallweit <hkallweit1@gmail.com>; Russell
> King <linux@armlinux.org.uk>
> Subject: Re: Broken networking on iMX8QXP after suspend after upgrading
> from 5.10 to 6.1
>
> >
> > Sorry for the delayed response.
>
> I must equally apologize for the delayed response.
>
> Your patch helped greatly finding the actual root cause of the problem
> (which pre-dates 5.10):
>
> f166f890c8f0 ("net: ethernet: fec: Replace interrupt driven MDIO with polled
> IO")
>
> How 5.10 worked for us is a mystery, because a suspend-resume cycle before
> link up writes to MII_DATA register before fec_restart() is called, which
> restores the MII_SPEED register, triggering the MII_EVENT quirk.
>
> > Have you tried setting mac_management_pm to true after mdiobus
> registration?
> > Just like below:
>
> I have tested your patch and it does fix my issue, with your patch I also
> realized a side-effect of mac_managed_pm in the FEC driver. The PHY will
> never suspend due to the current implementation of fec_suspend() and
> fec_resume().
>
> phy_suspend() and phy_resume() are never called from FEC code.
>
> May I pick up your patch with a signed-off from you? I would like to make it a
> small series adding also suspend/resume of the PHY.
>

Yes, you can pick up my patch as long as you want. :)

> If you want to send it yourself instead, please pick up these tags:
> Fixes: 557d5dc83f68 ("net: fec: use mac-managed PHY PM")
> Closes:
> https://lore.k/
> ernel.org%2Fnetdev%2F1f45bdbe-eab1-4e59-8f24-add177590d27%40actia.s
> e%2F&data=05%7C02%7Cwei.fang%40nxp.com%7Cb0ae4b334bae44f50d34
> 08dc3833314f%7C686ea1d3bc2b4c6fa92cd99c5c301635%7C0%7C0%7C63
> 8447039719541704%7CUnknown%7CTWFpbGZsb3d8eyJWIjoiMC4wLjAwM
> DAiLCJQIjoiV2luMzIiLCJBTiI6Ik1haWwiLCJXVCI6Mn0%3D%7C0%7C%7C%7C&
> sdata=b2Q4eYhXuNHwRfG7kIIteIgUVtVxxo%2BAwqyyt8OdPiU%3D&reserved
> =0
> Reported-by: John Ernberg <john.ernberg@actia.se>
> Tested-by: John Ernberg <john.ernberg@actia.se>
>
> And then I send a separate patch with yours as a dependency.
>
> Thanks! // John Ernberg