Re: [GIT PULL] USB/Thunderbolt driver updates for 7.4-rc1

[Jacob Riff <jacob@riff.dk>]

On Fri, 30 Oct 2026 15:28:29 -1000, Linus Torvalds wrote:
> and honestly, this all seems rather excessive. Retry ten times a
> second, for ten seconds? As far as I know, this has never succeeded in
> the history of this laptop, and it has no bad effects apart from the
> silly messages.

Hi Linus,

Sorry about the noise. As written the retry catches every -ENODEV,
including ucsi_register_port() failing on the role switch lookup, which
no retry can change. Fix follows in this thread: it retries only the
PPM-not-ready case, for one second instead of ten, and has been through
six boots on the machine the retry was written for (three hit the race,
all recovered).

As far as I can see, the -ENODEV in your log can only come from the
!CONFIG_USB_ROLE_SWITCH stub of fwnode_usb_role_switch_get(); the real
one returns a switch, NULL or -EPROBE_DEFER. If that is your config,
UCSI has never been able to init there: TYPEC_UCSI builds without
USB_ROLE_SWITCH but then fails on it at probe. That is a separate,
older problem.

Jacob