Re: [PATCH] wifi: brcmfmac: cfg80211: Report port_authorized for 4-way HS offload

[Janne Grunau]

On Wed, Sep 16, 2026 at 10:32:38AM +0200, Johannes Berg wrote:
> On Tue, 2026-09-15 at 21:05 +0200, Janne Grunau wrote:
> > On Tue, Sep 15, 2026 at 03:22:50PM +0200, Johannes Berg wrote:
> > > On Sat, 2026-09-12 at 11:40 +0200, Janne Grunau wrote:
> > > > 
> > > > This will conflict with the "brcmfmac: add FT/OKC roaming offload
> > > > support" [3] series but is written in a way to easily resolve the
> > > > conflicts.
> > > 
> > > I ... no, I'm not going to take that conflict, and I applied the FT/OKC
> > > series to wireless-next. Please create this patch on top of wireless-
> > > next, if needed at all, and then you can backport something later to
> > > stable if needed.
> > 
> > That is very unfortunate. I hoped this change could merged as a fix for
> > 7.3.
> > 
> > The change is required. wpa_supplicant 2.12 (and unpatched 2.11) will
> > run into its authentication timeout and disconnect WPA2/3 PSK networks
> > on brcmfmac devices with firmware offload support.
> > 
> > I'll rebase onto wireless-next and send a v2 but I think it's serious
> > issue in the process that it will take until after v7.4-rc1 (7+ weeks
> > from now) before this fix will land in stable kernel releases.
> 
> That process issue is yours - wpa_s 2.11 was released *well over two
> years ago*, and nobody cared to fix the issue since. AFAICT this could
> have been fixed in kernel 6.11 or so, but now it's suddenly urgent
> enough to make life more complicated for everyone? ...

yes, the kernel change fell of the table 2 years ago. Nothing can be
done about that anymore.

The issue became urgent due to the wpa_supplicant 2.12 release. That was
more than one month ago but I became aware of it and the reintroduction
of the authentication issue last week after Gentoo shipped
wpa_supplicant 2.12.

The patch to avoid this issue in wpa_supplicant 2.11 does not apply to
wpa_supplicant 2.12. So distributions either drop the patch and are
broken or adept it for 2.12. One distro maintainer was reluctant to
carry a patch for a second release. Based on this I disagree that fixing
the issue quickly in stable kernels makes life more complicated for
everyone. It would make life easier for distro maintainers and users of
distributions shipping unpatched wpa_supplicant 2.12.

Janne