Re: [PATCH wireless-next] wifi: cfg80211: don't apply HT flags to S1G channels

[Johannes Berg <johannes@sipsolutions.net>]

On Tue, 2026-01-20 at 16:22 +1100, Lachlan Hodges wrote:
> > > Playing around with some 5 GHz configurations in OpenWRT it appears that only
> > > so much information can be conveyed and indeed there are configurations that
> > > will be rejected by cfg80211 (or maybe hostapd in some cases, not sure).
> > 
> > Really? Got an example, perhaps even with hostapd log to see how it's
> > rejected? That clearly contradicts what I wrote above which I really did
> > believe to be true until this moment ;-)
> 
> Sorry, it is indeed cfg80211 that is not accepting the chandef configuration.

Maybe I still know something ;-)

> channel=165
> 
> vht_oper_centr_freq_seg0_idx=171

> This is the result of selecting mode AX, channel 165 with a width of 80MHz.

Well, I guess that could be done in some countries, U-NII-4 goes up to
channel 177 inclusive? But not for Australia which only has up to 173
(according to Wikipedia, maybe it's changing, I didn't check better
sources now.)

> The GL-MT3000 using the mt76 driver with mt7915e driver (?)
> 
> cfg80211              299008  4 mt7915e,mt76_connac_lib,mt76,mac80211
> compat                 12288  3 mt76,mac80211,cfg80211
> mac80211              581632  3 mt7915e,mt76_connac_lib,mt76
> mt76                   73728  2 mt7915e,mt76_connac_lib
> mt76_connac_lib        45056  1 mt7915e
> 
> And for hostapd output (Sorry I don't have anything more verbose)
> 
> Tue Jan 20 04:10:43 2026 daemon.err hostapd: nl80211: kernel reports: (extension) channel is disabled

Right, makes sense, it's at least saying channel 177 doesn't exist or
cannot be used, perhaps already on 169/173 depending on what the device
knows about U-NII-4.

> Tue Jan 20 04:10:43 2026 daemon.err hostapd: Could not set channel for kernel driver
> Tue Jan 20 04:10:43 2026 daemon.err hostapd: Interface initialization failed
> 
> where _cfg80211_chandef_usable() is (rightfully) rejecting the configuration.
> This simply appears to be the logic to generate the config doesn't cover this
> case properly. 

Right.

> Anyways, I don't think this is all that useful too be honest,
> it just comes down to ensuring the static configurations are known especially
> since this range is specified as AUTO-BW anyway as per the regdom.

Yeah, or checking against the regulatory, or something ... but basically
channel 165 + 80 MHz isn't _really_ possible then, since the
channelization plan says then you should use 171 as it does now.

> > > I
> > > suppose what I'm asking is - how much do we need to protect usermode from
> > > this? Do you feel it's worth including a flag that somewhat emulates
> > > NOHT40+/-? I know much of that logic is quite old and S1G is "modern" ... ?
> > 
> > It's always nice to have userspace be aware of things, but I guess if
> > it's a configuration that's statically known to be invalid, it wouldn't
> > matter so much? 

But I think in the case you describe above, userspace should well have
been able to, at least in theory, predict that this would happen, based
on the channel flags.

> > But I don't think I've understood the S1G angle - are
> > you talking about primary 1 MHz? Primary 2 MHz?
> 
> Putting primaries aside and just thinking of operating - eventually we would
> need a NO_2MHZ flag as India for example only has 1MHz channels.. however
> this is well into the future .. we first need to add support for other
> countries :').

:)

> But yes as you said since these configurations are static it's not really a
> big deal and more up to userspace to be aware of it. I think what we have in
> cfg80211 now is good enough and seems to be working well when integrating
> into an actual userspace stack.

OK, sounds good.

> > We just posted a lot of hwsim changes for NAN - better send yours sooner
> > rather than later ;-)
> 
> Yes I have seen, lots of Wi-Fi 8 documents aswell so I assume that will get busy
> soon too. Well see .. Lots to do!!

I don't think hwsim will get too many wifi8 changes, but yeah there are
a lot of things brewing now. Seems you (mostly?) got just ahead of it
though, so it shouldn't be much of an issue, I'd think.

johannes