On Tue, Nov 10, 2020 at 09:59:50PM +0100, Linus Lüssing wrote:
> On Tue, Nov 10, 2020 at 07:20:39AM +0100, Linus Lüssing wrote:
> > [...]
> >
> > The issue was introduced with the following commit:
> >
> > f4bfdc5e571e ("iwlwifi: mvm: stop supporting swcrypto and bt_coex_active module parameters")
> > * first affected tag: v5.8-rc1
> >
>
> PS: As this commit mentioned bt_coex_active, I retried with a
> vanilla 5.9.6 kernel while leaving bt_coex_active at its
> default value. That is leaving it enabled while all previous tests
> I did had it disabled.
>
> However I still get the Bluetooth A2DP freeze and subsequent
> kernel panics.
> [...]
I did a few more tests and found out that it was the old iwlwifi
firmware causing the kernel panics for me when Bluetooth co-existence
is enabled.
With firmware-iwlwifi_20170823-1_all.deb on Debian I can reproduce
the issue, with firmware-iwlwifi_20180518-1~bpo9+1_all.deb or
firmware-iwlwifi_20210818-1_all.deb I can't.
Also, I can still reproduce the kernel panic with firmware-iwlwifi at
version 20170823-1 and with a recent Linux kernel on Debian Sid
(linux-image-5.17.0-1-amd64, 5.17.3-1). So nothing which has fixed
it in the upstream kernel since v5.8-rc1.
I'm a bit surprised that a non-free firmware can create kernel
panics in "random" code paths. But maybe that's expected as
whatever is running the iwlwifi firmware has access to more memory
areas than I would like it to have? Let me know if I should dig
deeper, if there is something that should/could be fixed in the
upstream, opensource iwlwifi driver to prevent such kernel panics.
Regards, Linus
PS: firmware-iwlwifi_20170823-1_all.deb seems unavailable on Debian
at the moment, even the archives. But I found a copy in the Kali
Linux archives:
http://old.kali.org/kali/pool/non-free/f/firmware-nonfree/firmware-iwlwifi_20170823-1_all.deb
