Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]


Groups > linux.debian.kernel > #66423 > unrolled thread

Bug#952660: src:linux: macvlan multicast/broadcast regression in stretch

Started byBen Hutchings <ben.hutchings@codethink.co.uk>
First post2020-02-27 00:50 +0100
Last post2020-04-22 20:40 +0200
Articles 3 — 3 participants

Back to article view | Back to linux.debian.kernel


Contents

  Bug#952660: src:linux: macvlan multicast/broadcast regression in stretch Ben Hutchings <ben.hutchings@codethink.co.uk> - 2020-02-27 00:50 +0100
    Bug#952660: src:linux: macvlan multicast/broadcast regression in stretch Dave Page <dave.page@codethink.co.uk> - 2020-04-07 14:30 +0200
    Bug#952660: src:linux: macvlan multicast/broadcast regression in stretch Tero Marttila <terom@fixme.fi> - 2020-04-22 20:40 +0200

#66423 — Bug#952660: src:linux: macvlan multicast/broadcast regression in stretch

FromBen Hutchings <ben.hutchings@codethink.co.uk>
Date2020-02-27 00:50 +0100
SubjectBug#952660: src:linux: macvlan multicast/broadcast regression in stretch
Message-ID<zEeBb-52z-3@gated-at.bofh.it>
Package: src:linux
Version: 4.9.210-1
Severity: important
Tags: upstream fixed-upstream patch

Linux 4.9.209 included:

commit 9b266c6c12b055d51f5004e9b7285a4c97627311
Author: Eric Dumazet <edumazet@google.com>
Date:   Mon Jan 6 12:30:48 2020 -0800

    macvlan: do not assume mac_header is set in macvlan_broadcast()

which fixed some TX cases but broke the RX case.  When handling a
received multicast or broadcast packet, macvlan_broadcast() now reads
the destination address from the wrong place.  The packets may then
fail to match the multicast filters that they should.  This is fixed
in 4.9.211 by:

commit bde7568224a8d1fca99d00ec3f35d9f8fdc50cc6
Author: Eric Dumazet <edumazet@google.com>
Date:   Tue Jan 14 13:00:35 2020 -0800

    macvlan: use skb_reset_mac_header() in macvlan_queue_xmit()

This is a major regression for VM hosts using macvlan/macvtap as
ARP and IPv6 neighbour discovery became quite unreliable.

Ben.

-- System Information:
Debian Release: 10.3
  APT prefers stable
  APT policy: (990, 'stable'), (500, 'stable-updates'), (500, 'unstable')
Architecture: amd64 (x86_64)
Foreign Architectures: i386, armhf, arm64

Kernel: Linux 4.19.0-8-amd64 (SMP w/2 CPU cores)
Locale: LANG=en_GB.UTF-8, LC_CTYPE=en_GB.UTF-8 (charmap=UTF-8), LANGUAGE=en_GB.UTF-8 (charmap=UTF-8)
Shell: /bin/sh linked to /bin/dash
Init: systemd (via /run/systemd/system)
LSM: AppArmor: enabled

-- 
Ben Hutchings, Software Developer                         Codethink Ltd
https://www.codethink.co.uk/                 Dale House, 35 Dale Street
                                     Manchester, M1 2HF, United Kingdom

[toc] | [next] | [standalone]


#66700

FromDave Page <dave.page@codethink.co.uk>
Date2020-04-07 14:30 +0200
Message-ID<zSVwC-804-1@gated-at.bofh.it>
In reply to#66423
Hi there,

We've been able to work around this thanks to Ben's help, but are Debian 
planning on releasing an update to fix this regression caused by the 
9.12 point release?

Cheers,
Dave

-- 
Dave Page, Operations Team Manager                        Codethink Ltd
Telephone: +44 7762 840 414       3rd Floor, Dale House, 35 Dale Street
https://www.codethink.co.uk/         Manchester, M1 2HF, United Kingdom
We respect your privacy.   See https://www.codethink.co.uk/privacy.html

[toc] | [prev] | [next] | [standalone]


#66859

FromTero Marttila <terom@fixme.fi>
Date2020-04-22 20:40 +0200
Message-ID<zYsrT-wb-3@gated-at.bofh.it>
In reply to#66423
This seems to also affect keepalived `use_vmac` configurations, where 
the keepalived VIPs are configured on macvlan interfaces. After a 
stretch kernel update from 4.9.0-11-amd64 (4.9.189) to 4.9.0-12-amd64 
(4.9.209) the keepalived VIPs stop responding to ARP queries, causing 
the VIP to remain configured but unreachable on the VRRP master.

[toc] | [prev] | [standalone]


Back to top | Article view | linux.debian.kernel


csiph-web