Tested v2 on Ubuntu kernel 7.0.0-34-generic (rebased build, patch applied with no offsets). Results:
Test procedure: 3x rapid unload/load cycles of the patched modules (deliberately targeting the window that produced the v1 WARNING), adapter unplug during active transfer, and a full clean shutdown. Results: all three v1 symptoms are gone - zero WARNING at kthread_park, zero "rx urb mismatch" (mt76u_complete_rx), zero "chip reset failed" across the whole test boot. Unplug under load: no call traces, no hung tasks, interface teardown immediate, device re-probes cleanly on replug. Shutdown completed without hanging. Full kernel log of the test boot available on request. Three observations from code review / testing: 1. "Message ... timeout" lines still appear on module unload with an active connection (e.g. 00020002 = MCU_UNI_CMD(BSS_INFO_UPDATE), 000046ed = MCU_EXT_CMD(MAC_INIT_CTRL), sent from the mac80211 teardown path). These are cosmetic: with MT76_MCU_RESET set, mt76_mcu_get_response returns immediately (the messages appear ~33 ms after disconnect starts, not after the 3 s mcu timeout), and the mt792x_reset they trigger is now correctly suppressed by the new MT76_REMOVED check. Suggestion: check MT76_MCU_RESET in mt7921_mcu_parse_response before the dev_err + mt792x_reset to silence these misleading messages (same pattern exists in mt7925/mcu.c). 2. The commit message says this aligns mt7921 with mt7925, but in the 7.0.x tree mt7925_mac_reset_work does not have the MT76_REMOVED check - so mt7925 USB devices would still be exposed to the same race. You may want to include the equivalent fix for mt7925 in the series. 3. Minor: with the flags set on disconnect entry, the teardown MCU commands (MAC disable, BSS removal) and wfsys_reset in mt792xu_cleanup no longer reach the chip (-EIO). Harmless in practice - probe resets the chip on next load and the device recovers cleanly every time in our tests - but worth being aware of. Planning to run v2 as daily driver; will report back after 1-2 weeks of regular use. -- You received this bug notification because you are a member of Ubuntu Bugs, which is subscribed to Ubuntu. https://bugs.launchpad.net/bugs/2167595 Title: mt7921u: USB reset after -110 timeouts deadlocks in mt7921_abort_roc, blocks rtnl_lock and shutdown To manage notifications about this bug go to: https://bugs.launchpad.net/ubuntu/+source/linux/+bug/2167595/+subscriptions -- ubuntu-bugs mailing list [email protected] https://lists.ubuntu.com/mailman/listinfo/ubuntu-bugs
