Rebind hid-multitouch on resume to recover Razer Blade AMD trackpad click

[arun9486]

Summary

On Razer Blade laptops with AMD chipsets, the trackpad click stops responding after resume from suspend or hibernate. Motion still works; only the click button events drop. The kernel logs nothing on resume — the driver doesn't notice anything is wrong, so users typically don't realize it's recoverable without a reboot.

The affected hardware is the Cypress 06CB:CDA3 I2C HID touchpad on AMD's AMDI0010 I2C controller. After S3/S4 the device returns with stale firmware state and the I2C HID driver doesn't re-initialize it. Reloading hid_multitouch forces a clean detach + re-bind, which re-runs the protocol negotiation and recovers the click. There's no clean fix in 6.x kernels yet; this is a userspace workaround until that lands.

Changes

• default/systemd/system-sleep/fix-razer-amd-trackpad — post-resume hook that reloads hid_multitouch. Guarded by $1 == "post" so it only fires on resume.
• install/config/hardware/fix-razer-amd-trackpad.sh — DMI-gated installer. Triple check (vendor Razer + AMDI0010 platform device + 06CB:CDA3 HID device) so it's a no-op on Intel Razers and on Razer models with a different touchpad.
• install/config/all.sh — wired into the hardware-fix block next to fix-tuxedo-backlight.sh, matching the existing vendor-fix pattern.

Existing installs won't pick this up — there's no migration. Following the convention of the other fix-*.sh hardware scripts, which also only run on fresh install. Happy to add a migration if preferred.

Test plan

• On affected hardware (Razer Blade 14 RZ09-0370, AMD): sleep cycle without the hook → click broken on resume; with the hook → click works on every resume.
• Detection guard verified: all three gates (DMI vendor, AMDI0010 device, 06CB:CDA3 HID) present on the affected machine.
• Detection guard on non-affected hardware — script exits cleanly with no install (logically correct from the conditions, not separately reproduced).

[Copilot]

Pull request overview

Adds a hardware-specific systemd sleep hook to restore Razer Blade (AMD) trackpad click functionality after suspend/hibernate by re-initializing the HID multitouch stack.

Changes:

• Add a systemd/system-sleep post-resume hook that reloads hid_multitouch.
• Add a DMI + device-ID gated installer script to deploy the hook only on matching Razer/AMD/Cypress hardware.
• Wire the new hardware fix into the main install flow.

Reviewed changes

Copilot reviewed 3 out of 3 changed files in this pull request and generated 1 comment.

File	Description
install/config/hardware/fix-razer-amd-trackpad.sh	Detects the target Razer AMD platform + HID device and installs the system-sleep hook.
default/systemd/system-sleep/fix-razer-amd-trackpad	Runs on resume (“post”) to recover click events by reinitializing hid_multitouch.
install/config/all.sh	Invokes the new hardware fix during installation alongside other hardware-specific scripts.

Tip

If you aren't ready for review, convert to a draft PR.
Click "Convert to draft" or run gh pr ready --undo.
Click "Ready for review" or run gh pr ready to reengage.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[Copilot]

Reloading the whole hid_multitouch module is a global operation: it can disrupt any other devices using that module, and modprobe -r hid_multitouch will fail if the module is in use (in which case the intended detach/rebind won’t happen and the click issue may remain). Consider re-binding only the affected HID device(s) (e.g., unbind/bind the *06CB:CDA3* entries under /sys/bus/hid/drivers/hid-multitouch/) or otherwise targeting the specific device instead of unloading the module.

Suggested change

	# nothing fails from its perspective. Reloading hid_multitouch forces
	# the driver to detach and re-bind, which re-initializes the I2C HID
	# protocol and recovers the click.
	if [[ $1 == "post" ]]; then
	modprobe -r hid_multitouch 2>/dev/null
	modprobe hid_multitouch 2>/dev/null
	# nothing fails from its perspective. Re-binding only the affected
	# hid_multitouch device forces the driver to detach and re-bind, which
	# re-initializes the I2C HID protocol and recovers the click without
	# disrupting other hid_multitouch devices.
	if [[ $1 == "post" ]]; then
	for device_path in /sys/bus/hid/drivers/hid-multitouch/*06CB:CDA3*; do
	[[ -d "$device_path" ]] || continue
	device_name=${device_path##*/}
	printf '%s' "$device_name" > /sys/bus/hid/drivers/hid-multitouch/unbind 2>/dev/null
	printf '%s' "$device_name" > /sys/bus/hid/drivers/hid-multitouch/bind 2>/dev/null
	done

[arun9486]

Good catch — switched to per-device unbind/bind under /sys/bus/hid/drivers/hid-multitouch/ so it targets only the Cypress 06CB:CDA3 entry and is independent of module refcount. Verified on the affected hardware: rebind re-runs the I2C HID protocol negotiation and recovers the click. Updated in the latest force-push.

[Copilot]

Pull request overview

Copilot reviewed 3 out of 3 changed files in this pull request and generated 2 comments.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[Copilot]

echo ... > "$driver/unbind" 2>/dev/null won’t reliably suppress errors from the > redirection itself (bash reports redirection failures before the 2>/dev/null takes effect). If the write fails during resume, this can still emit “Permission denied”/ENOENT messages to the journal. Consider redirecting stderr for the whole operation (e.g., grouping the command and redirecting) or using a tee-based write to ensure errors are intentionally handled/suppressed.

Suggested change

	echo "$name" > "$driver/unbind" 2>/dev/null
	echo "$name" > "$driver/bind" 2>/dev/null
	{ echo "$name" > "$driver/unbind"; } 2>/dev/null
	{ echo "$name" > "$driver/bind"; } 2>/dev/null

[Copilot]

The header comment says the hook “reloads hid_multitouch”, but the installed system-sleep script actually unbinds/rebinds the specific HID device under the hid-multitouch driver. Please update this comment (and ideally the PR title/description) to match the implementation so future readers don’t go looking for a module reload that never happens.