On Tue, 2021-09-28 at 23:08 +0300, Imre Deak wrote: > On Tue, Sep 28, 2021 at 11:02:37PM +0300, Souza, Jose wrote: > > On Tue, 2021-09-28 at 22:55 +0300, Imre Deak wrote: > > > On Tue, Sep 28, 2021 at 10:45:50PM +0300, Souza, Jose wrote: > > > > > > [...] > > > > > > Would not be possible to use TC_PORT_DISCONNECTED when really > > > > > > disconnected and dropping the use of TC_PORT_TBT_ALT for it? > > > > > > > > > > TC_PORT_DISCONNECTED is the state when the PHY ownership is not held and > > > > > we don't hold any power domains. > > > > > > > > > > TC_PORT_TBT_ALT is the state when the PHY ownership is not held (like > > > > > above), and we hold the power domain needed to block TC-cold. > > > > > > > > Swapping it would make modes names do what their names intend to. > > > > > > > > Up to the point that we only had TBT, TC alt and legacy it was fine to > > > > keep into TBT mode when disconnected but now with a disconnected state > > > > it do not make sense to keep it in TBT mode when disconnected. > > > > > > > > Or you rename it to TC_PORT_UNKNOWN, as it sets to > > > > TC_PORT_DISCONNECTED mode during tc_init() and when going to suspend. > > > > > > Not sure what you mean, because what you describe is what actually > > > happens. From all states icl_tc_phy_disconnect() will change to > > > the disconnected state, which is the state at init time and during > > > suspend or after unloading the driver. > > > > I'm talking about the state when system is up without anything > > connected to the port, after icl_tc_phy_disconnect() sets > > dig_port->tc_mode = TC_PORT_DISCONNECTED, icl_tc_phy_connect() goes > > and set it back to TC_PORT_TBT_ALT. > > Yes, when the port is locked we are in one of TBT-alt, DP-alt or legacy > mode. These are the only modes that make sense for an AUX transfer or a > modeset, for which the lock was taken. Okay at the end of the series mode goes to TC_PORT_DISCONNECTED if no reference is held. > > --Imre
