Re: [PATCH] usb: typec: ucsi: Handle connectors already attached at init

["Christian A. Ehrhardt" <lk@c--e.de>]

Hi William,

On Thu, Aug 13, 2026 at 10:38:40AM +0100, William Bright wrote:
> After a cold boot the PPM may already have a partner attached without a
> pending connector change to report. On a CYPD6125 the CCI register reads
> back 0 in ucsi_init(), so ucsi_connector_change() is never called and the
> Type-C mux is left unconfigured for the already-attached port.
> 
> Fix this by additionally checking the connector status of every
> connector and call ucsi_connector_change() for connectors that are
> already attached. Whilst also continuing to call ucsi_connector_change()
> for connectors that report a change via the CCI register.

I'm not sure if this is the right thing to do.

First, there is code in ucsi_register_port() that is supposed to
handle the case where a connection is already present. I wonder
why that does not work for you?

> Fixes: ce4c8d21054a ("usb: typec: ucsi: Fix connector check on init")
> Signed-off-by: William Bright <william.bright@imd-tec.com>
> ---
>  drivers/usb/typec/ucsi/ucsi.c | 7 +++++--
>  1 file changed, 5 insertions(+), 2 deletions(-)
> 
> diff --git a/drivers/usb/typec/ucsi/ucsi.c b/drivers/usb/typec/ucsi/ucsi.c
> index ecbda2a3783c..740ae4a14f7f 100644
> --- a/drivers/usb/typec/ucsi/ucsi.c
> +++ b/drivers/usb/typec/ucsi/ucsi.c
> @@ -2136,8 +2136,11 @@ static int ucsi_init(struct ucsi *ucsi)
>  	mutex_unlock(&ucsi->ppm_lock);
>  	if (ret)
>  		return ret;
> -	if (UCSI_CCI_CONNECTOR(cci))
> -		ucsi_connector_change(ucsi, UCSI_CCI_CONNECTOR(cci));
> +
> +	for (i = 0; i < ucsi->cap.num_connectors; i++)
> +		if (UCSI_CONSTAT(&connector[i], CONNECTED) ||
> +		    UCSI_CCI_CONNECTOR(cci) == i + 1)
> +			ucsi_connector_change(ucsi, i + 1);

ucsi_connector_change() really is supposed to handle an actual event.
The second call to the function will be a NOP because EVENT_PENDING
is already set. So your fix only works for one established connection.

Additionally, calling ucsi_connector_change() will set
EVENT_PENDING which in turn will cause the next ACK to include
UCSI_ACK_CONNECTOR_CHANGE which acks a connector change event
that never happend. That is, unless such an event did happen in the
mean time. In this case you will ACK the event without handling
it.

Also, I seem to recall that some backends don't like it when you
ACK something that did not happen.

Best regards,
Christian

>  	return 0;
>  
> 
> ---
> base-commit: 55431f6adac42f818a0c76594ce2c6c02fe1034d
> change-id: 20260813-ucsi-cold-b35e11991b7c
> 
> Best regards,
> -- 
> William Bright <william.bright@imd-tec.com>
> 
>