Re: LinuxKPI (iwlwifi, rtw8x) Please test! Important! (in stable/15 in three days)

From: Bjoern A. Zeeb <bz_at_FreeBSD.org>
Date: Sat, 07 Mar 2026 09:39:09 UTC
On Thu, 5 Mar 2026, ltning-freebsd-wireless@anduin.net wrote:

> On 3/5/26 14:58, Bjoern A. Zeeb wrote:
>> Thanks to the three of you who have tested the last round and reported 
>> back.
>> That was very much appreciated and helpful!
>
> Count me in, though a bit late! I immediately pulled and built the sources 
> when I saw your first message on the thread, and while things seem to work 
> mostly like before, I have had two very-rarely-seen crashes in relatively 
> short order:
> - one complete wedge when doing 'service netif restart wlan0' (obviously no 
> debug info available)
> - one failure to wake from sleep (black screen, hard poweroff required)

In theory on a debug kernel and crash dumps turned on there are ways to
blindly acquire a crash dump (if not really hard locked up) even if drm-kmod
has a blank or stale screen.


> I'm not saying I've never seen this before, but it's been ... months? So I 
> can't help but shake the feeling it's at least plausibly related to the 
> changes.
>
> On my actual problem with iwlwifi - the firmware still needs to crash at 
> least once after resuming from sleep before I can successfully connect to a 
> network. It will (upon 'service netif restart') associate, but never get an 
> IP, then the firmware crashes, and then it works.
>
> Since I've provided logs for this before I will not right now, however I am 
> rebuilding again now with your most recent changes, and will collect logs if 
> it happens again (and you think it is useful).

Yes please do as with underlying things changing details may change too and
give me more hints.  I know S/R is still an open TODO here and it seems most
of you are more blessed than me to actually have a laptop which will resume.

Just to fill my cache, you are also using the work around from
https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=263632#c27


>> This time you'll likely not need long testing time.  Maybe a change of
>> association or if you want a netif restart will be good tests.
>
> I will torture it properly. Swear!

Thanks!

-- 
Bjoern A. Zeeb                                                     r15:7