On 2024-09-25 08:19, Johannes Berg wrote:

On Wed, 2024-09-25 at 03:30 +0000, Ping-Ke Shih wrote:

...

I think the cause is picking commit 268f84a82753 ("wifi: cfg80211: check wiphy mutex is held for wdev mutex"), and cfg80211_is_all_idle() called by disconnect_work() uses wdev_lock() but not wiphy_lock().

Yeah seems like a stable only problem (CC them), this is with kernel version 6.6.51-00141-ga1649b6f8ed6 according to the warning.

...

I'm not sure if we should simply revert the picked commit 268f84a82753 or should pick more commits.

I don't see why it was picked up in the first place, so I guess I'd say remove it. We won't want to redo the locking on a stable kernel, I'd think.

Thanks! I tried reverting 268f84a82753, then as you guys expected the issue is gone. So will continue to run with this commit reverted. But would be good if we can decide how to handle this for stable releases (such as 6.6) and see if we can revert it their also in upcoming 6.6.53.

...

By the way, I think the latest kernel will not throw these messages.

Agree, that seems unlikely.

johannes

Sounds good. I will see if I can give it a try on later kernel also, to verify. It will require some work to get my device up and running on a later version so might take some time until I can give feedback. I was a bit fooled that I got the issue when moving from 6.6.30 to 6.6.51 + cherry-picking all latest rtw88 related changes for my LM842 (8822cu) usb dongle, but great that you guys figured it out so quickly.