On Tue, Apr 09, 2019 at 06:03:24PM +0300, Sergei Shtylyov wrote: > Hello! > > On 04/09/2019 01:45 PM, Simon Horman wrote: > > >>>> @@ -1979,8 +1985,9 @@ static void ravb_set_delay_mode(struct net_device *ndev) > >>>> priv->phy_interface == PHY_INTERFACE_MODE_RGMII_RXID) > >>>> set |= APSR_DM_RDM; > >>>> > >>>> - if (priv->phy_interface == PHY_INTERFACE_MODE_RGMII_ID || > >>>> - priv->phy_interface == PHY_INTERFACE_MODE_RGMII_TXID) > >>>> + if ((priv->phy_interface == PHY_INTERFACE_MODE_RGMII_ID || > >>>> + priv->phy_interface == PHY_INTERFACE_MODE_RGMII_TXID) && > >>>> + !soc_device_match(ravb_delay_mode_quirk_match)) > >>> > >>> But don't we need to error out of the probing as we can't set the delay mode > >>> requested? > >> > >> Yes, if we can, we should error out. It just depends on if there are > >> broken DT blobs out there. We recently had a lot of pain from broken > >> DT blobs using the at803x PHY and getting RGMII modes wrong. In the > >> long run, it is best to if DT, but i've no idea how many boards are > >> affected. > > > > Hi Andrew, > > > > I suspect there are such blobs out there and I'm not sure > > what the fall-out may or may not be. > > You mean the out-of-tree blobs? Because I'm not seeing any in-kernel > breakage... I am referring to older upstream blobs built from older upstream or out-of-tree BSP kernel source where the phy mode is rgmii-txid or rgmii-id. F.e. Ebisu and upstream v5.0 or out-of-tree R-Car Gen3 BSP v3.8.0.
