[Bug 2164785] Re: mt7921e: drv_own timeout → chip reset failed → kernel page fault (hard lock); regression since 22.04→24.04, still on 26.04/7.0.0-29

** Description changed: # Bug report: mt7921e oops after chip reset failed - Collected and prepared in **Cursor** by **Cursor Grok 4.6 (extra high)**. - جمع‌آوری و آماده‌سازی این گزارش در Cursor توسط مدل Cursor Grok 4.6 (extra high) انجام شده است. + **Filed:** https://bugs.launchpad.net/bugs/2164785 (Ubuntu `linux`) + + Collected and prepared in Cursor by Cursor Grok 4.6 (extra high). File against Ubuntu package **linux**. Also relevant: LP#2154385 (s2idle only), Discourse [mt7921 disappears on 26.04 / 7.0.0-27](https://discourse.ubuntu.com/t/mediatek-mt7921-wi-fi- intermittently-disappears-on-ubuntu-26-04-works-on-kernel-7-0-0-22-but- not-reliably-on-7-0-0-27/85118), kernel bz [#220353](https://bugzilla.kernel.org/show_bug.cgi?id=220353). --- ## Title mt7921e: drv_own timeout → chip reset failed → kernel page fault (hard lock); regression since 22.04→24.04, still on 26.04/7.0.0-29 ## Summary Internal MediaTek **MT7921** (`mt7921e`, PCI `14c3:7961`) hard-locks the machine. Sequence: 1. `driver own failed` / `Timeout for driver own` (MCU ownership handshake) 2. `chip reset failed` 3. `BUG: unable to handle page fault` — supervisor write, not-present page (`error_code 0x0002`) A firmware timeout must not panic the box. The oops is on the **reset- failure path**. **Regression timeline (this host):** - Ubuntu **22.04**: same NIC, years without this lockup - After upgrade to **24.04** (still **not** kernel 7.x): the hang appeared - Ubuntu **26.04** kernel **7.0.0-29-generic**: same hang, now captured as a kernel #PF So this is not a 7.0-only regression. 24.04 already used 6.8 HWE `mt7921e`. ## Hardware - ASUS TUF Gaming F17 `FX706HE_FX706HE`, BIOS `FX706HE.316` (2025-06-17; latest ASUS 316) - Tiger Lake-H + i915 + NVIDIA GA107M (RTX 3050 Ti Mobile) - Wi-Fi: `0000:2e:00.0` **14c3:7961** Filogic 330, subsystem AzureWave **1a3b:4680** - Parent: `0000:00:1c.0` Intel `8086:43bf` - Combo BT USB: IMC **13d3:3563** `Wireless_Device` (`btusb`) ## Software ``` Ubuntu 26.04 LTS (resolute), upgraded path 22.04 → 24.04 → 26.04 Linux 7.0.0-29-generic #29-Ubuntu SMP PREEMPT_DYNAMIC Fri Jul 17 20:52:35 UTC 2026 x86_64 linux-firmware 20260319.git217ca6e4.1ubuntu ``` Firmware loaded: ``` ASIC revision: 79610010 HW/SW Version: 0x8a108a10, Build Time: 20260224110909a WM Firmware Version: ____010000, Build Time: 20260224110949 ``` On-disk: ``` WIFI_MT7961_patch_mcu_1_2_hdr.bin.zst sha256 1cad8c509ded3d60567ec8b8665db85a0982591341f46175d1fc94a82b988b58 WIFI_RAM_CODE_MT7961_1.bin.zst sha256 b99a525d7921bf430fe4648ccadb9ce8fd282118cf11a834651016ee4256371d ``` ACPI FADT: ASPM unsupported; `mt7921e: disabling ASPM L1` then `can't disable ASPM; OS doesn't have ASPM control`. Default suspend was s2idle; `mem_sleep_default=deep` was applied later. **Awake oopses still happened** on 7.0.0-29 (not only after resume). LP#2154385 is related but not sufficient. ## Reproduction On 7.0.0-29 with `mt7921e` bound: 1. Associate to an AP 2. Idle ~15s–2 min, or toggle Wi-Fi radio, or warm-reboot after a firmware wedge 3. `driver own failed` storm (~1 Hz, ~23s) → `chip reset failed` → #PF → hard lock (power button) Warm reboot after wedge: ``` pci 0000:00:1c.0: broken device, retraining non-functional downstream link at 2.5GT/s pci 0000:00:1c.0: retraining failed ``` Device missing from lspci until full **Shut Down** or PCI rescan. Rescan can restore the device and oops again within minutes. ## Kernel log (2026-08-21 16:09:24 +0330) Healthy probe ~16:09:25. Then: ``` mt7921e 0000:2e:00.0: driver own failed mt7921e 0000:2e:00.0: Timeout for driver own … repeats … mt7921e 0000:2e:00.0: chip reset failed BUG: unable to handle page fault for address: ffffd3a042d64ff0 #PF: supervisor write access in kernel mode #PF: error_code(0x0002) - not-present page ``` RIP/Call Trace often **not flushed**. Another oops the same day: fault address `ffffd37941ac9ff0`. ## Why not “just install a newer Ubuntu kernel” Official **resolute** archives (2026-08-22): newest **generic** is **7.0.0-30.30**. Changelog for linux-signed 7.0.0-30.30 is packaging resync only; **no mt76/mt7921 entries**. There is **no 7.1/7.2** in Ubuntu apt. Mainline 7.1/7.2 exist only as **unsigned** kernels on kernel.ubuntu.com (Secure Boot must be off). One Discourse user said 7.2-rc2 helped **MT7922**; that is not a verified fix for this MT7921 #PF, and it is not an Ubuntu-supported kernel. `disable_aspm=1`, `pcie_aspm=force`, unbind-on-error, and PCI rescan were tried; they do not stop the oops (unbind can D-state; rescan re- triggers). ## Request 1. Treat reset-failure #PF as a kernel bug: `chip reset failed` must not take down the machine. 2. Bisect 22.04-stable vs 24.04 6.8 HWE mt76/mt7921 (regression started at 24.04 on this laptop). 3. Consider SRU / note 7.0.0-27+ reports vs 7.0.0-22 on 26.04 (other users). ## Current workaround `modprobe.blacklist=mt7921e`, bind `14c3:7961` to `pci-stub`, USB `13d3:3563` unauthorized. USB Ethernet. Internal Wi-Fi not loaded on 7.0.0-29. -- You received this bug notification because you are subscribed to linux in Ubuntu. Matching subscriptions: Bgg, Bmail, Nb https://bugs.launchpad.net/bugs/2164785 Title: mt7921e: drv_own timeout → chip reset failed → kernel page fault (hard lock); regression since 22.04→24.04, still on 26.04/7.0.0-29 Status in linux package in Ubuntu: New Bug description: # Bug report: mt7921e oops after chip reset failed **Filed:** https://bugs.launchpad.net/bugs/2164785 (Ubuntu `linux`) Collected and prepared in Cursor by Cursor Grok 4.6 (extra high). File against Ubuntu package **linux**. Also relevant: LP#2154385 (s2idle only), Discourse [mt7921 disappears on 26.04 / 7.0.0-27](https://discourse.ubuntu.com/t/mediatek-mt7921-wi-fi- intermittently-disappears-on-ubuntu-26-04-works-on- kernel-7-0-0-22-but-not-reliably-on-7-0-0-27/85118), kernel bz [#220353](https://bugzilla.kernel.org/show_bug.cgi?id=220353). --- ## Title mt7921e: drv_own timeout → chip reset failed → kernel page fault (hard lock); regression since 22.04→24.04, still on 26.04/7.0.0-29 ## Summary Internal MediaTek **MT7921** (`mt7921e`, PCI `14c3:7961`) hard-locks the machine. Sequence: 1. `driver own failed` / `Timeout for driver own` (MCU ownership handshake) 2. `chip reset failed` 3. `BUG: unable to handle page fault` — supervisor write, not-present page (`error_code 0x0002`) A firmware timeout must not panic the box. The oops is on the **reset- failure path**. **Regression timeline (this host):** - Ubuntu **22.04**: same NIC, years without this lockup - After upgrade to **24.04** (still **not** kernel 7.x): the hang appeared - Ubuntu **26.04** kernel **7.0.0-29-generic**: same hang, now captured as a kernel #PF So this is not a 7.0-only regression. 24.04 already used 6.8 HWE `mt7921e`. ## Hardware - ASUS TUF Gaming F17 `FX706HE_FX706HE`, BIOS `FX706HE.316` (2025-06-17; latest ASUS 316) - Tiger Lake-H + i915 + NVIDIA GA107M (RTX 3050 Ti Mobile) - Wi-Fi: `0000:2e:00.0` **14c3:7961** Filogic 330, subsystem AzureWave **1a3b:4680** - Parent: `0000:00:1c.0` Intel `8086:43bf` - Combo BT USB: IMC **13d3:3563** `Wireless_Device` (`btusb`) ## Software ``` Ubuntu 26.04 LTS (resolute), upgraded path 22.04 → 24.04 → 26.04 Linux 7.0.0-29-generic #29-Ubuntu SMP PREEMPT_DYNAMIC Fri Jul 17 20:52:35 UTC 2026 x86_64 linux-firmware 20260319.git217ca6e4.1ubuntu ``` Firmware loaded: ``` ASIC revision: 79610010 HW/SW Version: 0x8a108a10, Build Time: 20260224110909a WM Firmware Version: ____010000, Build Time: 20260224110949 ``` On-disk: ``` WIFI_MT7961_patch_mcu_1_2_hdr.bin.zst sha256 1cad8c509ded3d60567ec8b8665db85a0982591341f46175d1fc94a82b988b58 WIFI_RAM_CODE_MT7961_1.bin.zst sha256 b99a525d7921bf430fe4648ccadb9ce8fd282118cf11a834651016ee4256371d ``` ACPI FADT: ASPM unsupported; `mt7921e: disabling ASPM L1` then `can't disable ASPM; OS doesn't have ASPM control`. Default suspend was s2idle; `mem_sleep_default=deep` was applied later. **Awake oopses still happened** on 7.0.0-29 (not only after resume). LP#2154385 is related but not sufficient. ## Reproduction On 7.0.0-29 with `mt7921e` bound: 1. Associate to an AP 2. Idle ~15s–2 min, or toggle Wi-Fi radio, or warm-reboot after a firmware wedge 3. `driver own failed` storm (~1 Hz, ~23s) → `chip reset failed` → #PF → hard lock (power button) Warm reboot after wedge: ``` pci 0000:00:1c.0: broken device, retraining non-functional downstream link at 2.5GT/s pci 0000:00:1c.0: retraining failed ``` Device missing from lspci until full **Shut Down** or PCI rescan. Rescan can restore the device and oops again within minutes. ## Kernel log (2026-08-21 16:09:24 +0330) Healthy probe ~16:09:25. Then: ``` mt7921e 0000:2e:00.0: driver own failed mt7921e 0000:2e:00.0: Timeout for driver own … repeats … mt7921e 0000:2e:00.0: chip reset failed BUG: unable to handle page fault for address: ffffd3a042d64ff0 #PF: supervisor write access in kernel mode #PF: error_code(0x0002) - not-present page ``` RIP/Call Trace often **not flushed**. Another oops the same day: fault address `ffffd37941ac9ff0`. ## Why not “just install a newer Ubuntu kernel” Official **resolute** archives (2026-08-22): newest **generic** is **7.0.0-30.30**. Changelog for linux-signed 7.0.0-30.30 is packaging resync only; **no mt76/mt7921 entries**. There is **no 7.1/7.2** in Ubuntu apt. Mainline 7.1/7.2 exist only as **unsigned** kernels on kernel.ubuntu.com (Secure Boot must be off). One Discourse user said 7.2-rc2 helped **MT7922**; that is not a verified fix for this MT7921 #PF, and it is not an Ubuntu-supported kernel. `disable_aspm=1`, `pcie_aspm=force`, unbind-on-error, and PCI rescan were tried; they do not stop the oops (unbind can D-state; rescan re- triggers). ## Request 1. Treat reset-failure #PF as a kernel bug: `chip reset failed` must not take down the machine. 2. Bisect 22.04-stable vs 24.04 6.8 HWE mt76/mt7921 (regression started at 24.04 on this laptop). 3. Consider SRU / note 7.0.0-27+ reports vs 7.0.0-22 on 26.04 (other users). ## Current workaround `modprobe.blacklist=mt7921e`, bind `14c3:7961` to `pci-stub`, USB `13d3:3563` unauthorized. USB Ethernet. Internal Wi-Fi not loaded on 7.0.0-29. To manage notifications about this bug go to: https://bugs.launchpad.net/ubuntu/+source/linux/+bug/2164785/+subscriptions