[MT7902] mt7921e driver fails to initialize with "Message 00020001 timeout" on Acer Extensa 15 #1062

Activity

1. martinrobot commented on May 12, 2026

   

   martinrobot

   on May 12, 2026

   More actions

   Adding a working-config data point on a different OEM than the OP. May help distinguish whether the MCU timeout is OEM/firmware-specific or chip-universal.

   Hardware

   • Laptop: ASUS Vivobook K6602VV (BIOS K6602VV.311, 2024-04-23)
   • Wi-Fi (PCIe): 14c3:7902, subsystem 1a3b:5520 (AzureWave)
   • Bluetooth (USB): 13d3:3579 (IMC Networks shell; USB descriptor's iManufacturer reads MediaTek Inc., and bluetoothctl show reports Manufacturer: 0x0046 (70) = MediaTek per Bluetooth SIG — cross-confirming the silicon under the IMC USB ID)
   • Combo card: Wi-Fi half PCIe, BT half USB

   Software

   • Ubuntu 24.04.3 LTS, kernel 6.17.0-1020-oem
   • Stock mt7921e / mt7925e don't claim 14c3:7902, so the card sits unbound by default.

   Working config

   Installed abdullaabdullazade/mt7902_driver with the --fallback flag. The flag skips the gen4 Xiaomi-BSP variant (which carries an explicit kernel-panic warning for ASUS hardware with AW-series modules). The fallback path uses:

   • Wi-Fi: hmtheboy154/mt7902 — a strip-down of mainline mt76 that, per the upstream repo's claims, incorporates Sean Wang's Feb 2026 patch series. I haven't audited the patch contents against lore line-by-line.
   • Bluetooth: out-of-tree btusb + btmtk built from mt7902_temp/linux-6.17/drivers/bluetooth/ (the directory matching my running kernel).

   Init traces — no MCU timeout

   Driver bind:

   $ lspci -nnk -s 2c:00.0
   0000:2c:00.0 Network controller [0280]: MEDIATEK Corp. Device [14c3:7902]
       Subsystem: AzureWave Device [1a3b:5520]
       Kernel driver in use: mt7902e
   

   Wi-Fi probe:

   mt7902e 0000:2c:00.0: ASIC revision: 79020000
   mt7902e 0000:2c:00.0: HW/SW Version: 0x8a108a10, Build Time: 20251212032046a
   mt7902e 0000:2c:00.0: WM Firmware Version: ____000000, Build Time: 20251212032127
   mt7902e 0000:2c:00.0 wlp44s0: renamed from wlan0
   

   Bluetooth probe (with patched btmtk):

   Bluetooth: hci0: HW/SW Version: 0x008a008a, Build Time: 20220107141051
   Bluetooth: hci0: Device setup in 2485973 usecs
   Bluetooth: hci0: AOSP extensions version v0.96
   

   Controller identity:

   $ bluetoothctl show
   Controller XX:XX:XX:XX:XX:XX (public)
       Manufacturer: 0x0046 (70)
       Powered: yes
   

   Wi-Fi associates and passes traffic on 2.4 GHz and 5 GHz. Bluetooth pairs and connects. No Message 00020001 timeout. Two cosmetic warnings worth flagging only for completeness: HCI Enhanced Setup Synchronous Connection command is advertised, but not supported (legacy SCO edge case) and AOSP quality report is not supported (Android-only diagnostic). Suspend/resume not yet stress-tested.

   Possible diff vs the OP's failure

   I'm running an out-of-tree driver based on the upstream patches; the OP is on stock mt7921e, which doesn't claim 14c3:7902 at all, so this isn't quite an apples-to-apples comparison. Still, the working config above suggests the chip itself is amenable to the upstream patch series on at least one OEM variant. Things worth checking on the Acer Extensa side:

   • Try the out-of-tree driver before chasing OEM differences. If the OP still hits an MCU timeout with abdullaabdullazade/mt7902_driver --fallback (or the underlying hmtheboy154/mt7902), that's a meaningful new signal — the failure is likely chip-universal or OEM-specific to Acer, not a stock-driver gap.
   • Module-vendor subsystem ID — mine is AzureWave 1a3b:5520; OP reports 1a3b:5524. Different AzureWave SKU; potentially different EEPROM/calibration blob.
   • ASPM — the mt7925e driver exposes a disable_aspm module parameter that several users in adjacent threads have found necessary. The OP already tried pcie_aspm=off as a kernel cmdline param, but the two mechanisms aren't equivalent. Worth checking whether the out-of-tree mt7902e build accepts a similar parameter.
   • BIOS/UEFI Wi-Fi-init differences — pre-driver state of the combo card can vary between vendors.
2. angrykirc commented on Jul 6, 2026

   

   angrykirc

   on Jul 6, 2026

   More actions

   For fellow people, I managed to get MT7902 working on my OpenWrt 25.12.4 by simply installing kmod-mt7921e and putting firmware blobs from https://github.com/hmtheboy154/mt7902/tree/backport/firmware into /lib/firmware/mediatek folder.
   Without these firmware blobs I was getting an 'direct firmware load failed' error.