GPD Win Mini 2025 (AI 370 / G1617-02): Gyro not working — two fixes needed

[shivito1]

GPD Win Mini 2025 (AI 370 / G1617-02): Gyro not working — two fixes needed

Device Information

Field	Value
Device	GPD Win Mini 2025
CPU	AMD AI 370
DMI String	G1617-02
OS	CachyOS Deckify
Kernel	7.0.1-1-cachyos-deckify
HHD Version	4.1.8-1
IMU Chip	BMI260 (ACPI name: BMI0160, IIO name: bmi260)

---

Symptom

Gyro does nothing in Steam despite HHD running correctly and DualSense emulation active.
The virtual DualSense device is created and visible in Steam, but tilting the device produces
no gyro output.

HHD log shows the following error approximately 2 seconds after launch:

GPDW  ERROR    IMU Did not send information for 2s. Disabling Gyro Sync.

The IIO device exists and is readable at /sys/bus/iio/devices/iio:device0 with name bmi260,
and raw axis reads work correctly:

cat /sys/bus/iio/devices/iio:device0/name
# bmi260

cat /sys/bus/iio/devices/iio:device0/in_anglvel_x_raw
# 4  (non-zero, sensor is alive)

---

Root Cause 1 — imu.py: bmi260 missing from IMU_NAMES

File: hhd/controller/physical/imu.py

The IMU_NAMES list is used by find_sensor() to locate the IIO device by reading
/sys/bus/iio/devices/iio:deviceX/name and checking if any entry in IMU_NAMES is
a substring of the name string. The match is case-sensitive.

The AI 370's BMI260 chip reports its IIO name as lowercase bmi260. The list only
contains BMI0260 (uppercase ACPI style), so the match silently fails and
find_sensor() returns None. HHD then waits 2 seconds for IMU data that never
arrives and disables gyro sync.

# Current (broken)
IMU_NAMES = ["bmi323-imu", "BMI0160", "BMI0260", "i2c-10EC5280:00", "i2c-BOSC0260:00"]

# Fixed
IMU_NAMES = ["bmi323-imu", "BMI0160", "BMI0260", "bmi260", "bmi160", "i2c-10EC5280:00", "i2c-BOSC0260:00"]

---

Root Cause 2 — __init__.py: hrtimer missing from G1617-02 config

File: hhd/device/gpd/win/__init__.py

The G1617-02 device config does not set "hrtimer": True. In base.py, the
controller loop only sets up the IIO buffer trigger when dconf.get("hrtimer", False)
is True:

if dconf.get("hrtimer", False):
    start_imu = d_timer.open()
if start_imu:
    prepare(d_imu)

Without a trigger assigned, the Linux IIO subsystem refuses to enable the buffer.
Attempting to enable scan elements while the buffer has no trigger returns
Invalid argument. This means even if Fix 1 is applied and the sensor is found,
no data can be read from the buffer because the trigger was never configured.

# Current (broken)
"G1617-02": {
    "name": "GPD Win Mini (2025)",
    "touchpad": True,
    "wincontrols": False,
},

# Fixed
"G1617-02": {
    "name": "GPD Win Mini (2025)",
    "touchpad": True,
    "wincontrols": False,
    "hrtimer": True,
},

---

Confirmation

After applying both fixes, clearing the Python bytecode cache, and restarting
hhd.service, the log now shows:

imu  INFO    Found device 'bmi260' at /sys/bus/iio/devices/iio:device0
imu  INFO    Found device 'bmi260' at /sys/bus/iio/devices/iio:device0

No IMU timeout error appears. Gyro functions correctly in Steam with DualSense
emulation enabled.

---

Additional Notes

• The ACPI firmware on this device reports the chip as BMI0160 in the sysfs path
  (.../i2c-BMI0160:00/iio:device0) but the bmi260 kernel driver correctly
  identifies it as a BMI260. This ACPI naming quirk is what led to the wrong sensor
  name bmi160 being written into state.json by earlier troubleshooting attempts —
  worth noting in case other users report the same confusion.

• The G1617-01 (original GPD Win Mini) config also does not set hrtimer,
  but that device may use a different IMU or driver path. This report is specific
  to the 2025 model with the AI 370 and BMI260.

• Both fixes are minimal, targeted, and do not affect any other device configs.

---

Steps to Reproduce

1. Install HHD on CachyOS Deckify (or any Arch-based distro) on a GPD Win Mini 2025
2. Enable DualSense emulation in HHD
3. Open Steam and attempt to use gyro in any game
4. Observe gyro does nothing
5. Check HHD log: IMU Did not send information for 2s. Disabling Gyro Sync.

---

Reported by a GPD Win Mini 2025 AI 370 user on CachyOS Deckify.
Both fixes verified working on kernel 7.0.1-1-cachyos-deckify, HHD 4.1.8-1.

[antheas]

Is this the same as yours?

#311

[shivito1]

Related but not the same issue — the symptom is similar but the root cause is different.

Issue #311 was solved by blacklisting competing BMI drivers (bmi160/bmi270 loading alongside bmi260), which is a driver conflict problem.

My issue on the GPD Win Mini 2025 (AI 370, G1617-02) is different — I only have bmi260_i2c and bmi260_core loaded, no competing drivers. The two separate problems I found are:

1. IMU_NAMES in imu.py uses case-sensitive matching — my sensor reports as lowercase bmi260 but only BMI0260 (ACPI uppercase) is in the list, so find_sensor() returns None silently with no error logged.
2. G1617-02 config is missing "hrtimer": True — without the hrtimer trigger the IIO buffer cannot be enabled at all, regardless of whether the sensor is found. Attempting to enable scan elements returns Invalid argument until a trigger is assigned.
   Both fixes are required on my device. The driver conflict fix from GPD Win Mini 2024 - CachyOS bmi160 gyro on HHD DualSense not working #311 would not have helped here since I never had competing drivers loaded.

[antheas]

Hhd does not support bmi260_i2c yet. It only supports the BMI270 module.

Last year bmi260 was updated to support the bmi270 unit but it was partial so I did not whitelist it. Did you try the dkms module?

[shivito1]

Yes, gyro is working for me right now with bmi260_i2c loaded — I have not installed the dkms module.
My loaded modules relevant to the IMU:
bmi260_i2c
bmi260_core
industrialio
industrialio_triggered_buffer
The IIO device appears at /sys/bus/iio/devices/iio:device0 with name = bmi260, and raw axis reads work (in_anglvel_x_raw returns live data).
The two changes I made to get HHD to actually read from it:

Added "bmi260" to IMU_NAMES in imu.py so find_sensor() matches the lowercase name the bmi260_i2c driver reports
Added "hrtimer": True to the G1617-02 config in init.py so the IIO buffer trigger gets set up before HHD tries to read from the buffer

Without the hrtimer trigger, enabling scan elements returns Invalid argument — the buffer simply cannot be opened without a trigger assigned first.
I have not tried the dkms module. Would you like me to test with it, or is the bmi260_i2c path something worth supporting directly?

[antheas]

It is worth supporting and you can leave the issue open. It's on the Todo list. Particularly, fixing it so that the built in interrupt works, but I need to go through the kernel and upstream some patchss

hrtimer doubles CPU use and it's not required with the bmi270 module