Re: Regression: net/ipv6/mld running system out of memory (not a leak)

On Tue, Feb 18, 2020 at 07:55:36AM +0100, Rafał Miłecki wrote:

I'm sorry for a late reply, I spent that time for switching my devices
to some newer kernel. I wanted to make sure we are not chasing a bug
that's long time fixed now.

This problem still exists in the 5.4.18.

OK, bad news...

quoted

Hmm, I'm surprised that IGMP works for you, as it requires enable IPv6
forwarding. Do you have a lot IPv6 multicast groups on your device?

The thing is I don't really use IPv6. There are some single IPv6 packets
in my network (e.g. MDNS packets) but nothing significant.

For my testing purposes I access my access points using ssh and it's the
only real traffic. There are no wireless devices connected to my testing
devices. They are just running monitor mode interfaces without any real
traffic.

quoted

What dose `ip maddr list` show?

# ip maddr list
1:      lo
        inet  224.0.0.1
        inet6 ff02::1
        inet6 ff01::1
7:      br-lan
        link  33:33:00:00:00:01
        link  33:33:00:00:00:02
        link  01:00:5e:00:00:01
        link  33:33:ff:7a:fc:80
        link  33:33:ff:00:00:00
        inet  224.0.0.1
        inet6 ff02::1:ff00:0
        inet6 ff02::1:ff7a:fc80
        inet6 ff02::2
        inet6 ff02::1
        inet6 ff01::1

No much ipv6 traffic, no much IPv6 multicast groups. Only occurred with
IPv6 forwarding enabled...Any possibility(although unlikely) that there
is a loop for ipv6 multicast traffic under br-lan?

Maybe we can use perf kmem to trace kernel memory statistics.

Thanks
Hangbin