fixing_stuff

A personal knowledge base of fixes, workarounds, and notes for problems that I ran into. Most of them are with old, unsupported, or niche hardware on Linux (Nobara / Fedora). When I solve something, I dump the notes here. Then the next person who searches the same error message gets a result.

Each subdirectory covers one problem: what it was, what did not work, and what did.

---

Index

8bitdo-ultimate-2-wireless — 8BitDo Ultimate 2 Wireless: 2.4GHz dongle ignored by games

The controller works correctly over Bluetooth, and Steam Big Picture detects it. But games do not detect it. Root cause: the dongle's default firmware mode exposes the controller as a composite HID device (gamepad + keyboard + mouse). This breaks the SDL2 GameController classification. Fix: hold B while you power on the controller. This boots it into D-input mode.

Stack: 8BitDo Ultimate 2 Wireless + 2.4GHz dongle, Nobara 43, kernel 6.17.x, SDL2-based games, Unity-on-Proton.

Likely search hits: 8BitDo Ultimate 2 Wireless not working in games, dongle USB ID 2dc8:310b vs 2dc8:6012, composite HID gamepad, SDL2 GameController classification, Steam Input dongle Linux.

hp560sfp-firmware-update — Flashing HP 560SFP+ firmware on non-ProLiant hardware

HPE's firmware update tool refuses to run on consumer motherboards. But the firmware files inside the package support the 560SFP+. Ancient EEPROM firmware on used cards (for example 0x800007C7) can hang POST/GRUB when SFP+ modules link at boot.

Fix:

1. Force-install the RPM.
2. Symlink libpci.so.3.
3. Unbind the ixgbe driver.
4. Run .setup interactively, with no flags.

This bypasses the HPSUM platform check.

Stack: HP 560SFP+ (Intel 82599ES, PCI 8086:10fb / 103c:17d3), ixgbe, Linux x86_64.

Likely search hits: HP 560SFP+ firmware update Linux, HPE firmware-nic-intel-1.27.30, Possible execution of smart component from an older unsupported HPSUM release, NIC firmware flash without ProLiant, Adapter initialization failed during NVM update.

ixgbe-third-party-sfp — Third-party SFP+ modules on Intel ixgbe NICs

Three loosely related issues with non-Intel SFP+ transceivers on ixgbe-driven NICs (Intel X520-DA1, HP 560SFP+, and others):

1. ixgbe rejects unrecognized SFP+ modules by default. To fix this, use allow_unsupported_sfp=1.
2. Some copper RJ45 transceivers (for example RealHD) spoof optical DDM data. As a result, ethtool -m reports nonsense laser and power values. Honest modules like the MikroTik S+RJ10 report -inf dBm zeros. Only temperature and voltage are reliable across modules.
3. RJ45 SFP+ modules run hot in dead-air bracket pockets and need airflow.

To monitor temperature, see the CC-SFP-module-sensor daemon. It is module-agnostic and writes hwmon-format files for CoolerControl.

Stack: Intel X520-DA1 / HP 560SFP+ (Intel 82599ES) or any ixgbe-driven NIC, MikroTik S+RJ10 / RealHD SFP+ → RJ45 transceivers, Nobara / Fedora.

Likely search hits: ixgbe allow_unsupported_sfp, MikroTik S+RJ10, RealHD SFP+ RJ45 10GBase-T, SFP+ DDM spoofed laser power, copper SFP+ thermal throttle, CoolerControl SFP+ custom file sensor, ethtool -m bogus optical readings.

---

Why this repo exists

The solution to a hardware-specific Linux problem is usually buried in a forum post, a kernel mailing list thread, or a five-year-old GitHub issue. This is my own breadcrumb trail back to the working answer. I made it public so others can find it too.

I write these notes on a Nobara Linux 43 (Fedora-based) workstation. But most fixes apply to any modern Fedora or RHEL-family system. Many apply to Linux in general.

Contributing

Issues and PRs are welcome: corrections, better workarounds, or related fixes to add. You can have one of these problems on a different distro, kernel, or hardware revision. If the fix needed changes there, a note about that is useful.

License

MIT. Use, modify, and redistribute freely. No warranty.