v3 reviewed, built and deployed on the Precision 3650
(7.0.0-34-generic). Results:

- Diff review v2->v3: changes are limited to the three declared items; sha256 
verified.
- 2x rmmod/insmod cycles: MCU teardown noise went from 15 "Message ... timeout" 
lines per unload (v2, measured with the same test) to 0. No WARNING, no 
kthread_park, no blocked tasks.
- Unplug under active TX load (~750 pkt/s): clean teardown, the only kernel 
output was "USB disconnect" plus a single benign "tx urb failed: -71". 
Interface re-probed and reconnected after replug.
- Permanently installed (modules in /lib/modules/.../updates/) and running 
since 2026-09-27; boot-time check confirms v3 loads from disk, dmesg clean.

Two small notes from the review, mainly for the upstream submission:

1. In mt7925_mac_reset_work() the new MT76_REMOVED checks are correct
and useful (they prevent wake_queues/vif_connect_iter/regd_change on a
removed device), but the description's "do not clear MT76_RESET" wording
doesn't apply there - that function never touches MT76_RESET on the
mt7925 side.

2. The MCU timeout silencing in
mt7921_mcu_parse_response()/mt7925_mcu_parse_response() affects all
buses (mt7921e/mt7921s/mt7925e as well, since parse_response is shared),
not just USB. The logic holds on every bus - a real runtime MCU timeout
with no flags set still logs and triggers mt792x_reset() as before, we
verified the return path is unchanged - but it's worth mentioning in the
changelog. Skipping mt792x_reset() during teardown is also a small
correctness win: v2 was re-queueing reset_work onto a workqueue right
after disconnect had cancelled it and shortly before mt76_free_device()
destroys it.

Thanks for the quick turnaround on all three findings. From our side v3
looks ready for upstream submission; happy to be listed as Tested-by if
useful (Piotr, this Launchpad account).

-- 
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

Reply via email to