feat(install): udev reprobe rule recovers driverless devices at boot (#355) [HARDWARE-GATED]

[BANANASJIM]

DRAFT — defense-in-depth safety net, NOT a confirmed fix. Do not merge yet.

Status after multi-agent review (2026-06-05)

Three independent reviewers + a real-hardware repro on the maintainer's Vader 5 concluded:

• The "no device at boot" mechanism (Generic X-Box pad is not hidden for Vader 5 (Bazzite) #355 symptom 4) was already fixed by PR fix(vader5): block only xpad so the pad stays discoverable (#355) #362 (b76edee): the driverless-at-boot state was caused by padctl's own 61-padctl-driver-block.rules unbinding usbhid when block_kernel_drivers briefly included it. Current main blocks only xpad, so nothing recreates the driverless precondition this rule assumes.
• On the maintainer's hardware (current main), baseline is healthy: usbhid bound to the 3 HID interfaces, 3 hidraw nodes present. The driverless state only reproduces by manually unbinding usbhid.
• The likely reason any reporter still fails is stale pre-fix(vader5): block only xpad so the pad stays discoverable (#355) #362 rules from an old install (fixed by reinstalling a post-fix(vader5): block only xpad so the pad stays discoverable (#355) #362 build, which rewrites the 61 rule) — diagnosable now via padctl doctor.
• This rule carries a small claim-fight risk: it could rebind usbhid under the daemon's in-flight libusb claim on an add/bind replay (mitigated but not eliminated by the live test -e guard).

Kept open as a documented safety net for any future regression that genuinely leaves an interface driverless. Merge gate: a real-boot padctl doctor capture from an affected machine showing a supported interface driver -> (none) (truly driverless). If instead it shows usbhid bound + no hidraw, or hidraw present + daemon down, this is the wrong fix and should be closed.

Change

• New generateReprobeRulesFromEntries → 62-padctl-reprobe.rules (sorts after the 61 unbind). One rule per supported device:

  ACTION=="add|bind", SUBSYSTEM=="usb", ENV{DEVTYPE}=="usb_interface", ATTRS{idVendor}=="37d7", ATTRS{idProduct}=="2401", RUN+="/bin/sh -c 'test -e /sys%p/driver || echo %k > /sys/bus/usb/drivers_probe'"
  

  Uses udev %p/%k substitution (filled by udevd, matching the proven 61 rule) — no reliance on shell env export. Live-checks the driver symlink at execution time, so only a still-driverless interface is reprobed.
• Wired into install after the 61 rule; added to /lib + /etc cleanup lists. Keeps the install-time probeAndReprobeDrivers (hybrid).

Test plan

• Layer-0 rule-gen tests (vid/pid, add|bind, usb_interface, drivers_probe target, live %p guard on every line); falsifiability verified (removing the guard fails the tests). CI green.
• Phase-B hardware validation (required before any merge): with the rule installed, udevadm monitor --udev + udevadm trigger --action=add on a driverless Vader interface must rebind usbhid and produce a hidraw node (udevadm test does NOT execute RUN, so it is not sufficient proof).

Refs

• Generic X-Box pad is not hidden for Vader 5 (Bazzite) #355 symptom 4. Investigation found the active cause already fixed (fix(vader5): block only xpad so the pad stays discoverable (#355) #362); this is belt-and-suspenders.

[coderabbitai]

Important

Review skipped

Draft detected.

Please check the settings in the CodeRabbit UI or the .coderabbit.yaml file in this repository. To trigger a single review, invoke the @coderabbitai review command.

⚙️ Run configuration

Configuration used: defaults

Review profile: CHILL

Plan: Pro

Run ID: 87d86082-f5ad-4ed4-8697-e783d9f8b12e

You can disable this status message by setting the reviews.review_status to false in the CodeRabbit configuration file.

Use the checkbox below for a quick retry:

• 🔍 Trigger review

✨ Finishing Touches

🧪 Generate unit tests (beta)

• Create PR with unit tests
• Commit unit tests in branch feat/boot-reprobe

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands and usage tips.}

[BANANASJIM]

Closing. Live padctl doctor output from an affected machine (#355) shows the daemon up with the Vader claimed and state=active, output_fd_alive=true — there is no driverless interface, so the boot-reprobe rule this PR adds addresses a precondition current main no longer produces (the boot-time usbhid unbind was fixed in #362). The real residual symptom is a connect/disconnect flap, tracked separately. Keeping the codebase free of a rule that could rebind usbhid under the daemon's libusb claim.