Received-SPF: pass (google.com: domain of linux-wireless+bounces-2066-linux.lists.archive=gmail.com@vger.kernel.org designates 2604:1380:40f1:3f00::1 as permitted sender) client-ip=2604:1380:40f1:3f00::1;
Message-ID: <f7b6c875dd41e792795ee3a7cea542b98c949a7d.camel@sipsolutions.net>
Subject: Re: [REGRESSION] 6.7 broke wifi "AP is in CSA process, reject auth"
From: Johannes Berg <johannes@sipsolutions.net>
To: coldolt <andypalmadi@gmail.com>
Cc: ayala.beker@intel.com, linux-wireless@vger.kernel.org, 
	regressions@lists.linux.dev
Date: Tue, 16 Jan 2024 22:08:45 +0100
In-Reply-To: <CAJvGw+BF+LWAzOKa8aJvGcdORWWjRDZObXupkCF5VAgSEm21nA@mail.gmail.com>
References: 
	<CAJvGw+DQhBk_mHXeu6RTOds5iramMW2FbMB01VbKRA4YbHHDTA@mail.gmail.com>
	 <dae4bb032ad8b0c9fa6547de5e869e51f9f6e766.camel@sipsolutions.net>
	 <CAJvGw+ATM3B09KbuLM0VsmU9GZRSO-ZP6ffwvvOa75xVHR3-kw@mail.gmail.com>
	 <d14d6b4c02282d6ef48a936b1ac8be71c80a8835.camel@sipsolutions.net>
	 <CAJvGw+BF+LWAzOKa8aJvGcdORWWjRDZObXupkCF5VAgSEm21nA@mail.gmail.com>
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable
User-Agent: Evolution 3.50.3 (3.50.3-1.fc39) 
Precedence: bulk
MIME-Version: 1.0

On Tue, 2024-01-16 at 22:37 +0200, coldolt wrote:
> On Tue, Jan 16, 2024 at 1:43=E2=80=AFPM Johannes Berg <johannes@sipsoluti=
ons.net> wrote:
> > so mode=3D=3D1 indicates quiet, new_operating_class/new_ch_num are actu=
ally
> > the channel it's currently on, and count is 9.
> >=20
> > Can you say if it actually changes the count? Maybe capture on channel
> > 36 using the NIC as a sniffer what it does over time:
> > https://wireless.wiki.kernel.org/en/users/drivers/iwlwifi/debugging#air=
_sniffing
>=20
> If I keep checking it the line "Unknown IE (60): 01 16 24 09" seems to
> always stay the same, the 09 doesn't change, it's the same today as it
> was yesterday.

Interesting.

> I captured the channel 36 for 15 mins, here is the 34MB file in
> gdrive: https://drive.google.com/file/d/1yqDb3g3Cfttm4W-Jb5AA51nLQ7OglWhl=
/view?usp=3Dsharing

Thanks! I got it, feel free to delete.

It's interesting, the AP is not including the element in the beacon, but
is unconditionally and always including it in the probe responses.

> > Initially I'd say though that if this situation persists, then your AP
> > is having some problems and we'd not have stayed connected without the
> > patch in question either. If you want, maybe revert and see what the
> > symptom is then?
>=20
> I now compiled and installed commit c09c4f3 and its parent 2bf57b0.
> The 2bf57b0 works great, connects immediately and I used it for 30+
> mins, also tried to connect/disconnect 5+ times smoothly, no symptoms.
> The c09c4f3 has the problem described originally, never connects even
> if trying 10+ times in 5+ minutes, keeps outputting the same dmesg
> message "AP is in CSA process, reject auth".

Yes, I didn't think of it being broken enough to include it in probe
responses but not beacons ... So it makes sense that w/o that we could
connect and not even disconnect, because while connected we ignore
(E)CSA in probe responses...

I'll think about it for a bit and ask some folks, but I guess we'll just
remove the check for probe responses from this ... there's little
guarantee anyway that this works well. Or maybe if we find it in probe
response check if we have a recent beacon? But that's tricky, we may
not, if we got a probe response on the channel but no beacon.

johannes