> Date: Thu, 5 Feb 2026 23:52:21 +0100
> From: Stefan Sperling <stsp@stsp.name>
> 
> On Thu, Feb 05, 2026 at 11:24:39PM +0100, Mark Kettenis wrote:
> > > Tests of this patch while using roaming on any iwx(4) device would be
> > > much appreciated.
> > > 
> > > Bz device firmware throws a fatal firmware error if we attempt to remove
> > > the firmware AP station before removing associated crypto keys. We currently
> > > rely on firmware to implicitly remove keys along with the station and this
> > > will no longer work on Bz.
> > > The same might be true for MA devices, not sure. I cannot test them.
> > 
> > Tested this diff on:
> > 
> > iwx0 at pci0 dev 20 function 3 "Intel Wi-Fi 6 AX210" rev 0x20, msix
> > iwx0: hw rev 0x440, fw 83.d24e06ed.0, pnvm 285b3568, address dc:45:46:e7:d5:33
> > 
> > This is FW version 83, so it must be MA ;).
> > 
> > Doesn't seem to make a difference here.  But I'm not roaming between
> > access points.
> > 
> > On the bright side, it seems this device works a lot better now.  But
> > I'm also using a different access point now.  Will do a bit more
> > testing before declaring that MA now work.
> 
> There is something going very wrong with Rx somehow. I see lots of
> packet loss (~85%) towards the Bz device. The worst symptoms are on
> 5GHz channels where I cannot even run tcpbench properly and ping can
> take well over 10 seconds between ping and response packets.
> 
> On 2GHz channels it tends to work a lot better but not perfectly either.
> tcpbench usually gives me about 10 Mbit/s but sometimes it's just as
> bad as it is on 5GHz.
> 
> I suspect this is the same issue some people have reported on MA.

Hard to tell, but MA certainly has problems as well.  I think I now
get a somewhat reliable connection because the laptop is sitting less
than a meter away from the "new" access point.  If I switch to using
the "old" access point, I don't even get a DHCP lease :(.

I think your diff helps though when I switch back to the "new" access
point.  So I'd suggest you should just go for it.

> Unfortunately, I haven't yet managed to track it down what is happening.
> It feels like the PHY wasn't properly tuned to the channel or something.

I'll see if I can explore that direction for my MA device.