Carlo Caione

Yeah, hitting the same problem here also with on 4.14.12.

@lwfinger output of bisection: ``` f2aa694b7459f20d37199651bc2a4495a5559502 is the first bad commit commit f2aa694b7459f20d37199651bc2a4495a5559502 Author: Johannes Berg Date: Tue Oct 17 21:56:20 2017 +0200 cfg80211: fix connect/disconnect edge cases commit 51e13359cd5ea34acc62c90627603352956380af...

Some info about this (just a braindump). The problem seems to be originated in https://github.com/torvalds/linux/blob/51e13359cd5ea34acc62c90627603352956380af/net/wireless/sme.c#L1077 since `prev_bssid` is not being set by https://github.com/torvalds/linux/blob/51e13359cd5ea34acc62c90627603352956380af/net/wireless/nl80211.c#L8942

@karora no, that's set by the userspace when using `nl80211` on reconnection. I checked `wpa_supplicant` and it appears that it should set it correctly (see https://w1.fi/cgit/hostap/plain/wpa_supplicant/ChangeLog). I'm still looking into...

Just to update this on resume `NL80211_ATTR_PREV_BSSID` is not set by `wpa_supplicant`. This is only triggering the problem but probably it's correct and not a problem in itself.

Just opened a PR for this issue

Some nits but nothing really blocking

> Hi @alexanderwachter @carlocaione , are there any results from the architecture meeting? This was not put in the agenda AFAIK. > In our internal discussions, it's better to have...

> @carlocaione Can this now be merged? We are in code freeze now. This must wait for the release to be out before merging.

Is still any definitive solution to this? I'm still getting this when using the jack (compiled from master) on a `type multi` alsa PCM.