Subject: Re: [vserver] End of life of 4.9
From: Herbert Poetzl <herbert@13thfloor.at>
Date: Tue, 17 Jan 2023 05:44:15 +0100

On Sun, Jan 15, 2023 at 04:18:50PM +1030, Adrian Butterworth wrote:
> Thanks Ghislain.

Thanks for the updates and testing! Appreciated!

> The existing 4.9.312-vs2.3.9.13 patch worked for me against
> 4.9.337 with a simple bump of "SUBLEVEL = 337" to the Makefile.

> I got a clean compile & am about to test on a non critical
> host. Ghislain, I know you run some regression tests but not
> sure which & where you got them.

> Given this is the last 4.9 it would be good if Herbert can
> generate an official patch & put it on the site.

Will happen shortly.

> Ideally Herbert is finding time to do the major work, but I
> have been considering options, if a serious security hole needs
> urgently patching before then.

> One option is the Civil Infrastructure "Super Long Term
> Support" (SLTS) kernels. These are maintained for 32-bit ARM
> and x86-64 systems in very long-lived infrastructure. ref:
> https://www.cip-project.org/

> They are already supporting 4.4.x to 2026 with a likely
> extension to 2036, so it should be possible to apply the
> patch-4.4.302-vs2.3.9.9.diff against their ongoing releases
> with minimal effort. I'll look into this further next week.

This is good news and I will check the 4.4 cip repo
and create an official patch for that as well.

> For me its going backwards a little, but I already backport the
> e1000e drivers to 4.9 for current generation hardware & expect
> it will be fine on 4.4.

> Ideally a small jump forward would be better. The latter
> announced SLTS kernels are the 4.19.x & 5.10.x series.

> A bump of the 4.9.x vserver patches to 4.19.x may be small
> enough to be done relatively quickly & would get us to 2024
> with LTS & at least another decade of SLTS. 

I remeber running into weird issues with 4.14 back then,
but that might not be a real problem with 4.19, so I'll
give that a try as well, no promises there though.

> Perhaps Herbert already has an idea on this. 

My main goal is to get 6.x at least partially working
so that it can be tested and slowly extended.

There are a number of ideas which were thrown around how
to handle things like network isolation via bpf or
utilize ip namespaces instead of the current xid setup
(which might even give us systemd support?)

My main concern at the moment is the userspace side,
because any changes we do on the kernel side, certainly
needs some kind of adjustment in userspace.

But for sure, some kind of LT 4.x will buy us some time
we need to sort those things out.

Best,
Herbert

> If not I'll run the 4.9 patches against it when I get a chance
> to get a handle on the scale of conflicts.

> regards
> Adrian


> On Mon, January 9, 2023 17:08, Ghislain Adnet wrote:
>> hi,
>> https://www.phoronix.com/news/Linux-4.9.337-LTS-Over
>> LINUX KERNEL --
>> The Linux 4.9 kernel was released back in 2016 and Greg Kroah-Hartman today
> issued the final point release for that kernel series with the Long Term
> Support (LTS) period now expired.
>> After Linux 4.9 was in an upgraded six-year LTS period, Linux 4.9.337 was
> released this morning as the final update to that kernel series. hope bertl
> has good news for us !
>> --
>> cordialement,
>> Ghislain ADNET.
>> AQUEOS.