Hello Russell, On 14/09/22 21:07, Russell King (Oracle) wrote: > On Wed, Sep 14, 2022 at 03:20:47PM +0530, Siddharth Vadapalli wrote: >> @@ -1427,6 +1471,9 @@ static void am65_cpsw_nuss_mac_link_down(struct phylink_config *config, unsigned >> struct net_device *ndev = port->ndev; >> int tmo; >> >> + /* disable phy */ >> + am65_cpsw_disable_phy(port->slave.ifphy); >> + > > This seems really strange. If you have a serdes interface which > presumably supports SGMII, 1000base-X etc, then link status is sent > across the serdes interface. If you power down the serdes, then you > can't receive the link status, and so mac_link_up() won't be called. > > Are you really sure you want to be enabling and disabling the PHY > in mac_link_down()/mac_link_up() ? Thank you for reviewing the patch. The PHY passed to the "am65_cpsw_disable_phy()" and "am65_cpsw_disable_phy()" functions within the "am65_cpsw_nuss_mac_link_down()" and "am65_cpsw_nuss_mac_link_up()" functions respectively, is the CPSW ethernet MAC's PHY and not the SERDES PHY. The SERDES PHY is powered on through the function call to the "am65_cpsw_init_phy()" function. The calls to the functions "am65_cpsw_enable_phy()" and "am65_cpsw_disable_phy()" within the "am65_cpsw_nuss_mac_link_up()" and "am65_cpsw_nuss_mac_link_down()" functions respectively, try to power on and power off the CPSW ethernet MAC's phy. Looking at it again,they do nothing, since the driver corresponding to the ethernet MAC's PHY which happens to be drivers/phy/ti/phy-gmii-sel.c, does not provide any methods to power on and power off the ethernet MAC's PHY. I have just realized that this is stale code and will remove it in the v2 series. Also, I realize now that I did not invoke "am65_cpsw_disable_phy()" on the SERDES PHY in the driver's remove function. I will fix this in the v2 series. Regards, Siddharth.
