MIME-Version: 1.0
From:   Tim K <tpkuester@gmail.com>
Date:   Tue, 7 Mar 2023 18:29:20 -0500
Message-ID: <CA+shoWTCExx7UQsC07mXe0FnuQDC6FGs_Vfd0qgVV+9V6wGrQw@mail.gmail.com>
Subject: rtw88: Monitor Mode Stability Issues
To:     linux-wireless@vger.kernel.org
Cc:     s.hauer@pengutronix.de, Larry.Finger@lwfinger.net
Content-Type: text/plain; charset="UTF-8"
Precedence: bulk

All,

Please excuse me if I am breaking mailing list etiquette, this is my
first email here. I was advised to CC some developers, so please
forgive me if this is too forward, or in the wrong place.

I am experimenting with an 8822bu dongle, and doing a packet capture
on monitor mode similar to kismet. (I have a simple bash loop doing
channel hopping in one window including HT40 and 80 MHz channels, and
tcpdump in another.) After a varying window of time (between 1 to 10
hours) the monitor interface simply stops receiving packets. There are
no output messages in the kernel otherwise indicating there is a
problem.

Cycling the interface doesn't resolve the situation, I have to rmmod
the driver, and then re-modprobe it. This allows me to get packets
again, but the packets stop in another few hours.

I am running this on a BeagleBone (Debian 10, Linux 4.19) and have
just setup a second copy on an STM32MP1 (Debian 11, Linux 5.15). I am
running the driver from lwfinger's GitHub repository at 4bede29. I
have cross compiled the drivers with cpp-arm-linux-gnueabihf 4:8.3.0-1
(Debian).

Interestingly enough, there seems to be a parallel issue. Using iw to
probe the phy info, it seems that all the 5 GHz channels are marked
"no IR", even after I set the regulatory domain to "US". At one point,
changing the regulatory domain caused some kernel messages to output
about "rtw_get_tx_power_params [rtw_core]" from phy.c line 2069 --
this was on, I believe, 3af004d.

Any ideas on what could be causing this, or steps to help troubleshoot
/ provide more information?

- Tim