Adding cross-vendor confirmation: this bug is reproducible and
DKMS-fixable on ONEXPLAYER SUPER X hardware.

Hardware
  ONEXPLAYER SUPER X (handheld PC)
  AMD Ryzen AI MAX+ 395, 128GB RAM
  MT7925 WiFi card [14c3:0717] on PCI 0000:c2:00.0

Software
  Ubuntu 24.04 LTS
  Kernel: tested on 6.17.0-22-generic and 6.17.0-23-generic (HWE)
  Driver: stock mt7925e from in-tree mt76 stack

Symptom pattern (pre-fix)
  Full system freezes 2-3 times per week. REISUB unresponsive,
  recovery only via hard power-off. Trigger pattern was multi-radio
  AP roaming via phone hotspot (3 BSSIDs across 2.4/5/6 GHz under
  one SSID). Captured call traces showed classic AB-BA deadlock
  between cfg80211_wiphy_work (holding wiphy mutex, calling
  cancel_work_sync on ROC) and mt7925_roc_work (waiting on that
  mutex). page_pool_release_retry() warnings preceded user-visible
  failure by hours.

Frequently overlapped with xHCI host controller death on s2idle
resume — the suspend/resume bug family. The xHCI death triggered
WiFi roam attempts in a degraded state, which then hit the ROC
deadlock and panicked the system.

Fix verified
  mt76-mt7925-dkms v1.5.0 from Zac Bowling's Cloudsmith repo.
  https://github.com/zbowling/mt7925

Verification protocol
  14-day window: April 29 (DKMS install) to May 13.
  Days 1-7: BSSID pinned to a single 5GHz radio while monitoring.
  Days 8-14: pin removed, auto-roaming re-enabled to exercise the
  previously-broken code path.
  Results: 44 roam attempts logged across days 8-14, 0 mt7925
  errors at error level, 0 hung-task signatures, 0 panics.
  xHCI deaths still occurred occasionally (separate AMD platform
  bug, not in mt7925's domain) but no longer cascaded into WiFi
  ROC deadlock. System uptime measured in days vs hours pre-fix.

Conclusion
  Bowling's patches resolve both bug families on this hardware:
  the ROC deadlock directly, and the xHCI-cascade-to-WiFi-panic
  by breaking the failure chain. DKMS auto-rebuilt cleanly across
  the kernel update from 6.17.0-22 to 6.17.0-23.

Adding to the existing Framework hardware reports — this confirms
the bug and fix on a different AMD-based vendor's system. Same
chip family, same trigger pattern, same fix.

-- 
You received this bug notification because you are a member of Ubuntu
Bugs, which is subscribed to Ubuntu.
https://bugs.launchpad.net/bugs/2137291

Title:
  mt7925e: Kernel panic NULL pointer dereference in
  mt76_connac_mcu_uni_add_dev during WiFi reset - Framework Desktop

To manage notifications about this bug go to:
https://bugs.launchpad.net/ubuntu/+source/linux/+bug/2137291/+subscriptions


-- 
ubuntu-bugs mailing list
[email protected]
https://lists.ubuntu.com/mailman/listinfo/ubuntu-bugs

Reply via email to