netdev - NETIF_F_LLTX breaks iwlwifi

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [day] [month] [year] [list]

Message-ID: <20190730141827.25fc4136@endymion>
Date:   Tue, 30 Jul 2019 14:18:27 +0200
From:   Jean Delvare <jdelvare@...e.de>
To:     Felix Fietkau <nbd@....name>,
        Toke Høiland-Jørgensen 
        <toke@...hat.com>, Johannes Berg <johannes.berg@...el.com>
Cc:     "David S. Miller" <davem@...emloft.net>,
        linux-wireless@...r.kernel.org, netdev@...r.kernel.org,
        linux-kernel@...r.kernel.org
Subject: NETIF_F_LLTX breaks iwlwifi

Hi Felix, Toke, Johannes,

After updating to kernel 5.2, I started losing wireless network on my
workstation a few minutes after boot. I could restart the network
service to get it back, but it would go away again a few minutes later.
No error message logged, but somehow the network traffic was no long
being processed.

My hardware is:

05:00.0 Network controller [0280]: Intel Corporation Wireless 8265 / 8275 [8086:24fd] (rev 78)

This is an Intel 8265 PCIe WiFI adapter by Gigabyte, model GC-WB867D-I,
which worked flawlessly for me until then.

I bisected it down to:

commit 8dbb000ee73be2c05e34756739ce308885312a29 (refs/bisect/bad)
Author: Felix Fietkau
Date:   Sat Mar 16 18:06:34 2019 +0100

    mac80211: set NETIF_F_LLTX when using intermediate tx queues

So whatever the commit message says, it is apparently not safe to run
TX handlers on multiple CPUs in parallel for this specific driver /
device.

Unless someone has an immediate explanation as to why it broke the
iwlwifi driver and the actual bug is in iwlwifi and it can be fixed
quickly and easily there, I would suggest that the above commit is
reverted for the time being, as apparently it wasn't fixing anything
but was just a performance optimization.

I am available to do any amount of tests or debugging, given the
guidance.

Thanks,
-- 
Jean Delvare
SUSE L3 Support