SM8250 USB-C DP alt mode: DP AUX times out in reversed (CC2) cable orientation

[Neil Armstrong]

On 9/7/26 18:26, BurningHoryd wrote:
> On 9/7/26 08:40, Neil Armstrong wrote:
>> This means the nb7vpq904m is _not_ responsible for the AUX swapping, other
>> boards have the wcd9395 or the fsa4480 in the middle to use the AUX line
>> for the USB-C Alternate Audio support, and does the swap.
>>
>> Could you check in the vendor tree if there's any reference to those ?
> 
> Checked Retroid's downstream Android DT for this board. There is an
> FSA4480, plus a GPIO AUX switch, and my mainline port models neither.
> 
> 1. fsa4480 at 43, on the same I2C controller as the redriver:
> 
>       i2c at 884000 {
>           fsa4480 at 43 {
>               compatible = "qcom,fsa4480-i2c";
>               reg = <0x43>;
>           };
>           redriver at 1c {                 /* nb7vpq904m equivalent */
>               compatible = "onnn,redriver";
>               status = "disabled";
>           };
>       };
> 
>     The only node that references it is the ASoC machine driver
>     (qcom,kona-asoc-snd, via "fsa4480-i2c-handle") - so downstream it is
>     used for the USB-C analog-audio / MBHC SBU routing, not by the DP
>     driver. The codec is a wcd9385, no USBSS block, so no wcd9395 here.
> 
> 2. The downstream DP driver does its own AUX orientation over GPIO:
> 
>       &dp_display {
>           qcom,dp-gpio-aux-switch;
>           qcom,aux-sel-gpio    = <&tlmm 8  0>;
>           qcom,aux-en-gpio     = <&tlmm 9  0>;
>           qcom,usbplug-cc-gpio = <&tlmm 65 0>;
>       };
> 
>     TLMM 9 there is the same pin my mainline DT currently hands to the
>     nb7vpq904m as "enable-gpios", so that assignment is likely wrong, and
>     TLMM 8 (the orientation select) is not wired up in my port at all.
> 
> 3. The redriver (onnn,redriver @ 0x1c, same I2C address as the
>     nb7vpq904m on my board) is left status = "disabled" downstream.
> 
> So the SBU / DP-AUX orientation switch on this board looks like a
> discrete GPIO-controlled analog mux (TLMM 8 select, TLMM 9 enable), and
> my port is missing it - I modeled the I2C redriver as the
> orientation-switch instead, which is why poking its AUX_CC_REG did
> nothing and only the DP_CC2 (pair-swapping, 2-lane) op-mode ever got an
> AUX transaction through in reverse.
> 
> Does this look like a job for gpio-sbu-mux
> (drivers/usb/typec/mux/gpio-sbu-mux.c) - a node with
> select-gpios = <&tlmm 8>, enable-gpios = <&tlmm 9>, orientation-switch +
> mode-switch, wired to the connector - with the nb7vpq904m kept as
> retimer-switch only (or dropped, since downstream disables it)? Or is
> the FSA4480 in the DP AUX path as well, making fsa,fsa4480 the right
> orientation-switch to add?

Yes if it's a discrete SBU mux, gpio-sbu-mux was designed for that, and
keep the nb7vpq904m and just drop the SBU link to the redriver and replace
it with a link to the SBU mux instead.

Neil

> 
> I'll wire up gpio-sbu-mux and come back with reverse/normal traces
> unless that's the wrong direction.
> 
> Thanks,
> BurningHoryd