The real root cause

Short version

The Valve Index bug was not really a blank-EDID bug that needed the headset rebooted. It was ddcutil / libddcutil probing the headset and wedging its EDID EEPROM. That probe is fixed upstream in ddcutil 3.0.2 (and the config file workaround below covers older versions). The fixvr udev rule is now a legacy fallback.

What was actually happening

ddcutil reads DDC/CI data over the display's I2C bus to find out whether a display supports things like brightness control. When it probes the Valve Index (a DisplayPort sink flagged non-desktop), the headset firmware stops answering EDID reads after the access to I2C slave address 0x37.

Once that happens the kernel logs EDID err: 2 / EDID_NO_RESPONSE, amdgpu falls back to a safe 640×480 mode with non-desktop=0, and the headset is unusable until it is power-cycled. Rebooting the HID layer (what fixvr did) got the EDID back, but it treated the symptom, not the cause.

Why only some setups were affected

The root cause is only triggered when something actually runs a DDC/CI probe against the headset. The most common trigger is KDE Plasma's powerdevil (the process behind System Settings → Power Management / power consumption settings), which uses libddcutil to work out whether connected displays allow brightness control. powerdevil probes at boot and login, which is exactly when people saw the headset come up wrong or the system lock up.

So the bug was never universal: people who did not run KDE Plasma, or who did not have powerdevil/ddcutil talking to the headset, often never saw it at all. That is also why fixvr "helped" some people and not others — it was racing another piece of software.

The upstream fix

The issue was tracked in ddcutil as rockowitz/ddcutil#632 and fixed in PR #637, released in ddcutil 3.0.2.

ddcutil now ships a built-in ignore list that includes the Valve Index HMD (VLV / product 0x91a8, model Index HMD). It is equivalent to the manual option:

--ignore-mmid VLV-Index_HMD-37288

Huge thanks to @omus, who diagnosed the root cause, drove the upstream ddcutil fix, and pointed fixvr users at it. 🎉

What you should do

1. Update ddcutil / libddcutil to 3.0.2 or newer

This is the real fix and needs no config file. Make sure your distro package (or your libddcutil shared library, which is what powerdevil actually loads) is at 3.0.2 or later:

bash

ddcutil --version

Then restart powerdevil (or log out/in) so it picks up the new library. The Valve Index should now be ignored by libddcutil.

2. Manually configure ddcutil ignore rules (if you cannot update ddcutil)

On ddcutil versions older than 3.0.2, tell the library to ignore the headset through the ddcutilrc config file. libddcutil reads this, so powerdevil honours it after a restart:

bash

mkdir -p ~/.config/ddcutil
cat >~/.config/ddcutil/ddcutilrc <<EOF
[global]
options = --ignore-mmid VLV-Index_HMD-37288
EOF

3. Verify ddcutil is ignoring your headset

Run ddcutil detect with the headset plugged in to verify the ignore rule is working. Find the entry for the Valve Index which should look similar to:

DDC_disabled
   I2C bus:  /dev/i2c-6
   DRM_connector:           card0-DP-1
   EDID synopsis:
      Mfg id:               VLV - Valve Corporation
      Model:                Index HMD
      Product code:         37288  (0x91a8)
      Serial number:
      Binary serial number: 4294967295 (0xffffffff)
      Model year:           2018
   DDC communication disabled

If the output for ddcutil detect says "Invalid display" instead of "DDC_disabled" as shown in the example then the ignore rule didn't work for your exact model. To fix the ignore rule follow the steps in "Manually configure ddcutil ignore rules" and update the --ignore-mmid to use the "Mfg id", "Model", and "Product" shown in the ddcutil dectect output with your headsets details.

Additionally, comment on this issue with your headset's --ignore-mmid value so we can update ddcutil to ignore your model by default.

4. (Optional) Stop the Index waking the system on suspend

Separate from the EDID wedge, the headset can wake the machine from suspend. Tell the kernel to ignore its remote-wakeup capability by adding this kernel parameter:

usbcore.quirks=28de:2613:j

Add it to your bootloader's kernel command line (grub/systemd-boot/ limine/etc.) and reboot. @omus is working on a kernel patch so this will eventually no longer be needed either.

Do I still need fixvr?

Probably not. If your libddcutil is 3.0.2+ and you added the kernel quirk, the headset should stop wedging and fixvr has nothing left to do.

The udev rule is kept as a legacy fallback for systems that cannot update ddcutil, or that hit a genuinely separate EDID problem. Be aware that it reboots the headset at boot, which can interact badly with other early-boot software (for example it can hang boot under Plymouth), and it does not run on resume from suspend (issue #2). That is why the root fix is preferred.

See Installation if you still want the legacy rule.

The real root cause ​

What was actually happening ​

Why only some setups were affected ​

The upstream fix ​

What you should do ​

1. Update ddcutil / libddcutil to 3.0.2 or newer ​

2. Manually configure ddcutil ignore rules (if you cannot update ddcutil) ​

3. Verify ddcutil is ignoring your headset ​

4. (Optional) Stop the Index waking the system on suspend ​

Do I still need fixvr? ​

References ​