It doesn't really make sense for phydev->master_slave_set (the user requested AN master/slave configuration) to take the value MASTER_SLAVE_CFG_UNSUPPORTED. That only exists, AFAICS, to detect implementations which don't report phydev->master_slave_get and phydev->master_slave_state. Permit the most trivial of drivers to exist, that where .config_aneg() accepts any user request, and is implemented as: /* We support anything */ phydev->master_slave_get = phydev->master_slave_set; This is currently rejected by ethnl_update_linkmodes() with the message "master/slave configuration not supported by device", precisely because lsettings->master_slave_cfg (which came from phydev->master_slave_get through phy_ethtool_ksettings_get()) is MASTER_SLAVE_CFG_UNSUPPORTED (coming from phydev->master_slave_get, see implementation above). By making phydev->master_slave_set never hold UNSUPPORTED, we avoid the above confusion (driver does implement master/slave but still gets detected as unsupported by core) without special casing in the driver. Signed-off-by: Vladimir Oltean <vladimir.oltean@xxxxxxx> --- drivers/net/phy/phy_device.c | 1 + 1 file changed, 1 insertion(+) diff --git a/drivers/net/phy/phy_device.c b/drivers/net/phy/phy_device.c index 9cb5aa04b2b5..e374d1a57030 100644 --- a/drivers/net/phy/phy_device.c +++ b/drivers/net/phy/phy_device.c @@ -3276,6 +3276,7 @@ static int phy_probe(struct device *dev) phydev->is_internal = true; phydev->suspended = true; + phydev->master_slave_set = MASTER_SLAVE_CFG_UNKNOWN; /* Deassert the reset signal */ phy_device_reset(phydev, 0); -- 2.34.1
