[eclecticc]

Mouse mode currently has problems in UEFI. First, because UEFI is sampling USB at a very low rate, the dongle report queue gets filled quickly with touchpad motion and drained slowly, which means the cursor ends up with a lot of lag. This is noticeable on Dogwood setup menu, but could be even worse on other BIOS screens.

Second, because PixArt's current mouse mode is missing scroll, there is no mouse wheel functionality. I (and claude) submitted PRs that resolve both.

For the lag, the dongle in mouse mode now accumulates movement the same way a mouse does. For the scroll, mouse mode now uses PTP and emulates a mouse, the same as we did earlier for Mac Mode. This only does scroll, tap to click, and two finger right click, not the other gestures that Mac Mode enables.

In addition to UEFI, this should improve the compatibility issues with TVs that Ray saw. Like with Mac Mode, we may still want to tune the cursor speed and acceleration. I have it a little fast because the touchpad is so small, but others may find it to be too fast.

---

Claude-generated details

Jira: HW-8318. Needed by FrameworkComputer/zmk-private#183.

Change

Two additions to hid_passthrough_hosts:

• tp_hosts_selected_value(id) returns the selected host's own copy of a host-state report, whatever tp_hosts_override() puts on the pad. It returns -ENODEV with no host selected and -EBUSY while a test tool holds the pad.
• tp_hosts_set_change_cb(cb) registers one callback, run on the system work queue whenever that value may have changed:
  • a host's SET that changes its copy,
  • the end of apply_work_cb (host switch, bond change, USB reset, settings load),
  • tp_hosts_hold().

The board uses these to tell whether the selected host has a PTP driver: one that has set Input Mode to PTP. If it hasn't (UEFI setup, smart TVs, boot loaders), the keyboard reads the pad's PTP frames itself, as it already does for Macs. A host's SETs still go to its own copy while the keyboard overrides the pad, so the moment its PTP driver sets PTP, the keyboard hands the pad back.

Testing

• Bench: Framework Desktop, 2026-10-31. Keyboard image = zmk-private integration branch (including #159) + the keyboard change, with this module change on c9d2703 (touchpad: Drop the reports the pad queued while suspended (HW-8311) #3 rebased onto ble: Pace, split and resample mouse reports; let the board read and override the pad (HW-8252) #2).
  • UEFI setup over wired and over the dongle: two-finger scroll and two-finger right-click work.
  • Booting from UEFI into Windows: PTP works normally.
  • Fedora over wired, Bluetooth and the dongle: PTP works normally.
• This branch: the same diff on ble: Pace, split and resample mouse reports; let the board read and override the pad (HW-8252) #2, compile-checked in the keyboard build of zmk-private#183.

Base: stacked on #2 (hw8252-mac-trackpad), which zmk-private#159 pins. Independent of #3 and #4.

🤖 Generated with Claude Code

[Copilot]

🟢 Approval recommended

No active review issues remain for the selected-host touchpad mode lookup and asynchronous host-state notifications.

No open issues 🎉

What changed in this PR

Exposes the selected host’s touchpad mode independently of board overrides, enabling the board to detect PTP support and choose mouse emulation for hosts such as UEFI and smart TVs.

Changes:

• Adds tp_hosts_selected_value() to read the selected host’s stored report value.
• Adds a system-work-queue callback for host-state changes, including feature updates, host switches, resets, and holds.
• Documents the new APIs and provides disabled-configuration stubs.

File	Description
drivers/​input/​hid_passthrough_hosts.h	Declares and documents the new APIs and adds fallback stubs.
drivers/​input/​hid_passthrough_hosts.c	Implements selected-host value lookup and asynchronous change notifications.

_{Review effort: Balanced}