Re: [PATCH 02/19] usb: typec: Add thunderbolt switch

[Sven Peter <sven@kernel.org>]

Hi Heikki,

On 9/1/26 13:17, Heikki Krogerus wrote:
> Hi Sven,
>
> On Sun, Aug 30, 2026 at 10:19:20PM +0200, Sven Peter wrote:
>> On Apple Silicon the USB4/Thunderbolt host router can only be brought
>> up after the Type-C PHY has been configured and requires cable details
>> known only to the Type-C port controller.
>>
>> The existing orientation switch only forwards the cable orientation.
>> The mode switch also forwards most of the required details but
>> represents devices which configure or route the Type-C signal path. The
>> ACIO host router is a consumer of that path and has additional ordering
>> requirements: it has to be started after the PHY has been configured
>> and stopped again before the PHY is disabled.
>>
>> The mode switch code calls all handlers in the same order both when
>> entering and leaving a mode and cannot express this reverse teardown
>> order. Treating ACIO as another mode switch could therefore leave the
>> NHI active after its PHY has been disabled, which can result in an
>> asynchronous SError or a SoC watchdog reset.
>>
>> Add a separate Thunderbolt switch which Type-C port drivers can use to
>> forward the negotiated Thunderbolt or USB4 cable details and explicitly
>> order host router setup and teardown around the mode switch. This is
>> similar to a USB role switch: each connection has exactly one host
>> router and the callback controls the lifetime of that functional
>> controller instead of configuring a signal mux.
> You are bypassing and duplicating things that are already supported in
> the typec bus.
>
> Your typec port driver probable does not register the partner altmode
> that you need, so you just need to fix that. After that you should be
> able to get the details you need simply by registering a typec bus
> notifier in the thunderbolt driver. We can add a helper for that if
> needed - the DRM subsystem will use the same notifier with the DP
> altmode soon.

Thanks for the pointer, I think that can indeed work out just fine. The 
bus notifier isn't enough though, I need a notification specifically 
when typec_altmode_update_active() happens (and not when the partner is 
registered as a possible altmode) and that only triggers a sysfs_notify 
and a kobject_uevent unless I'm missing something. Adding a simple 
notifier call chain should be possible there though.


>
> USB4 is meant to be handled as an altmode too, but it looks like it is
> not supported yet - I need to check the status with that. This may
> still need to be fixed, but that should not be a problem.

It isn't, but that indeed should not be hard to add.

Best,


Sven