RE: [PATCH net-next] net: fec: add pm runtime force suspend and resume support

[Wei Fang]

> -----Original Message-----
> From: Andrew Lunn <andrew@xxxxxxx>
> Sent: 2022年9月6日 22:42
> To: Wei Fang <wei.fang@xxxxxxx>
> Cc: davem@xxxxxxxxxxxxx; edumazet@xxxxxxxxxx; kuba@xxxxxxxxxx;
> pabeni@xxxxxxxxxx; netdev@xxxxxxxxxxxxxxx; linux-kernel@xxxxxxxxxxxxxxx
> Subject: Re: [PATCH net-next] net: fec: add pm runtime force suspend and
> resume support
> 
> On Tue, Sep 06, 2022 at 04:39:23PM +0800, wei.fang@xxxxxxx wrote:
> > From: Wei Fang <wei.fang@xxxxxxx>
> >
> > Force mii bus into runtime pm suspend state during device suspends,
> > since phydev state is already PHY_HALTED, and there is no need to
> > access mii bus during device suspend state. Then force mii bus into
> > runtime pm resume state when device resumes.
> 
> Have you tested this with an Ethernet switch hanging off the MDIO bus?
> It has a life cycle of its own, and i'm not sure it is guaranteed that the switch is
> suspended before the FEC. That is why the MDIO read/write functions have
> there own runtime PM calls, they can be used when the interface itself is
> down.
> 
Sorry, we don't have the product that an Ethernet switch hanging off the MIDO
bus of FEC. But I have tested system suspend/resume on i.MX6UL platform which
has two FEC MAC and share one MDIO bus. I have confirmed that the two PHYs
are suspended before the FEC. So it's safe to force the MDIO bus into runtime
suspend state. In addition, this patch has been already submitted to our local
repository for 3 years and our test team has test it for several times.