On Tue, Oct 26, 2021 at 11:48:42PM -0700, Jack Pham wrote:
> The intent of ucsi_get_src_pdos() is to obtain the source's PDOs
> in order to provide the power_supply object the required data to
> report the mininum, maximum and currently operating voltage &
> current should a PD contract be in place.
>
> If however, the port is operating as a PD source, this call would
> invoke GET_PDOS on the partner causing the PPM to send a
> Get_Source_Caps PD message to the port partner which may not make
> sense especially if the partner is not a dual-power role capable
> device. Further, it has been observed that certain DisplayPort
> adapter cables (which are power sink-only devices) even fail to
> bring up the display link after receiving a Get_Source_Caps
> message, suggesting they can't cope well with an unsupported PD
> message to the point that it renders them functionally inoperable.
>
> Fix this by checking the connector status flags for the power
> direction and use this to decide whether to send the GET_PDOs
> query to the partner or the port. This also helps to make the
> power_supply VOLTAGE_{MIN,MAX,NOW} and CURRENT_{MAX,NOW}
> properties more consistent when the port is in source mode.
>
> Signed-off-by: Jack Pham <quic_jackp@xxxxxxxxxxx>
> ---
> Hi Heikki,
>
> Was wrestling with how exactly to do this. The other approach I was
> thinking was to not even do GET_PDOs at all if operating as a source,
> but that would also mean we'd need to add similar checking to the
> VOLTAGE/CURRENT property getters in psy.c so that they would not
> return incorrect/stale data. Since the ONLINE property will already
> be 0 anyway it may make more sense to invalidate the rest of the props?
>
> The patch below is concise though...so that's what I went with ;)
Would it still make sense / help if we had separate power supplies
registered for the port-source, port-sink, partner-source and
partner-sink?
I also think we need to unify these power supplies so that tcpm and
ucsi (and others) always register the same power supplies.
thanks,
--
heikki
