Extended Capabilities bit values overwritten by the kernel

[Johannes Berg]

On Thu, 2025-10-23 at 18:23 +0200, Pablo MARTIN-GOMEZ wrote:
> 
> > You're probably the only/first person to ever want to _remove_
> > capabilities that the driver has :)

> It is more a question of testing a capability and its performance's 
> impact than to remove it completely :)

Well, OK, fair :)

> > Somehow it appears that we never considered the case of a capability
> > that's _completely_ implemented in mac80211/driver/device but someone
> > explicitly does _not_ want to advertise it.

> It's bit more complicated than that, at least in the case of TWT 
> Responder in ath12k.  When we don't set NL80211_ATTR_TWT_RESPONDER flag 
> in hostapd, the ath12k driver sends a command to the device to disable 
> TWT, so we end up in a situation where the device does not support the 
> feature anymore but we are still advertising its support in the 
> capabilities IE.

Huh, right. Then really this should've been handled like FTM responder
where we have a separate nl80211 feature flag, and userspace is
responsible for setting up both NL80211_ATTR_TWT_RESPONDER and the
extended capability flag together...

Not sure it's too late for that now? Jeff? Could/should we do
https://p.sipsolutions.net/d8d7eb632e66bfeb.txt with a corresponding
hostapd patch? Seems like the cleaner/better way to do it given that now
we end up in a situation where TWT responder is advertised but not
supported on older hostapd/hostapd with it not configured? And
apparently you already have this internally with
QCA_WLAN_VENDOR_FEATURE_HT_VHT_TWT_RESPONDER ...

> > I'm not sure I know what the right solution is ... Obviously not
> > advertising some features could easily be done in hostapd/wpa_s by just
> > clearing the bits you don't want, but that wouldn't actually disable the
> > feature in the lower level. So it'd probably work for TWT responder, but
> > not for TWT requester maybe, if the request would be triggered by
> > something in lower levels?
> Ok, so for my case, I just need to clear the bit, the feature is already 
> disabled in the lower levels. If I go for that solution in 
> hostapd/wpa_s, the issue is going to be can it break another driver?
> 

I don't think there's really any other driver using it.

johannes