RE: [PATCH net v2 2/2] net: fec: Suspend and resume the PHY

[Wei Fang]

> -----Original Message-----
> From: John Ernberg <john.ernberg@xxxxxxxx>
> Sent: 2024年2月29日 18:53
> To: Wei Fang <wei.fang@xxxxxxx>
> Cc: Shenwei Wang <shenwei.wang@xxxxxxx>; Clark Wang
> <xiaoning.wang@xxxxxxx>; dl-linux-imx <linux-imx@xxxxxxx>; David S.
> Miller <davem@xxxxxxxxxxxxx>; Eric Dumazet <edumazet@xxxxxxxxxx>;
> Jakub Kicinski <kuba@xxxxxxxxxx>; Paolo Abeni <pabeni@xxxxxxxxxx>;
> Heiner Kallweit <hkallweit1@xxxxxxxxx>; netdev@xxxxxxxxxxxxxxx;
> linux-kernel@xxxxxxxxxxxxxxx; John Ernberg <john.ernberg@xxxxxxxx>
> Subject: [PATCH net v2 2/2] net: fec: Suspend and resume the PHY
> 
> PHYs that are always-on will not enter their low power modes otherwise as
> they have no regulator to be powered off with.
> 
> Since the PHY is picked up via {of_,}phy_connect() and dropped with
> phy_disconnect() when the link is brought up and down respectively the only
> cases were pm is needed is when the netif is running or or when the link
nit: where

> has never been up.
> 
> To deal with the latter case the PHY is suspended on discovery in probe,
> since it won't be needed until link up.
> 
> Fixes: 557d5dc83f68 ("net: fec: use mac-managed PHY PM")
I'm not sure whether this commit should be blamed. After checking my local
code (not the recent upstream code), fec_suspend() will make the PHY enter 
suspend mode when calling phy_stop(), the specific logic is fec_suspend() -->
phy_stop() --> phy_state_machine() --> phy_suspend (). But the latest upstream
code may have changed this logic. I'm sorry that I don't have time to sit down
and look at the latest code.

> Signed-off-by: John Ernberg <john.ernberg@xxxxxxxx>
> ---
> 
> v2: New patch
> ---
>  drivers/net/ethernet/freescale/fec_main.c | 6 +++++-
>  1 file changed, 5 insertions(+), 1 deletion(-)
> 
> diff --git a/drivers/net/ethernet/freescale/fec_main.c
> b/drivers/net/ethernet/freescale/fec_main.c
> index 8decb1b072c5..c5394a4d8491 100644
> --- a/drivers/net/ethernet/freescale/fec_main.c
> +++ b/drivers/net/ethernet/freescale/fec_main.c
> @@ -2539,8 +2539,10 @@ static int fec_enet_mii_init(struct
> platform_device *pdev)
>  	/* find all the PHY devices on the bus and set mac_managed_pm to true
> */
>  	for (addr = 0; addr < PHY_MAX_ADDR; addr++) {
>  		phydev = mdiobus_get_phy(fep->mii_bus, addr);
> -		if (phydev)
> +		if (phydev) {
>  			phydev->mac_managed_pm = true;
> +			phy_suspend(phydev);
> +		}
>  	}
> 
>  	mii_cnt++;
> @@ -4631,6 +4633,7 @@ static int __maybe_unused fec_suspend(struct
> device *dev)
>  		if (fep->wol_flag & FEC_WOL_FLAG_ENABLE)
>  			fep->wol_flag |= FEC_WOL_FLAG_SLEEP_ON;
>  		phy_stop(ndev->phydev);
> +		phy_suspend(ndev->phydev);
As I aforementioned, if phy_stop() does not suspend the PHY in the latest
code, is it more general to restore the suspend operation in phy_stop()?

>  		napi_disable(&fep->napi);
>  		netif_tx_lock_bh(ndev);
>  		netif_device_detach(ndev);
> @@ -4716,6 +4719,7 @@ static int __maybe_unused fec_resume(struct
> device *dev)
>  		netif_tx_unlock_bh(ndev);
>  		napi_enable(&fep->napi);
>  		phy_init_hw(ndev->phydev);
> +		phy_resume(ndev->phydev);
>  		phy_start(ndev->phydev);
>  	}
>  	rtnl_unlock();
> --
> 2.43.0