Re: y2038 backport to v5.4

From: gregkh@xxxxxxxxxxxxxxxxxxx
Date: Mon Aug 17 2020 - 11:14:45 EST


On Mon, Aug 17, 2020 at 03:00:24PM +0000, Roosen Henri wrote:
> On Mon, 2020-08-17 at 16:35 +0200, gregkh@xxxxxxxxxxxxxxxxxxx wrote:
> > On Mon, Aug 17, 2020 at 02:15:16PM +0000, Roosen Henri wrote:
> > > On Tue, 2020-06-09 at 16:18 +0200, Arnd Bergmann wrote:
> > > > On Tue, Jun 9, 2020 at 2:36 PM Roosen Henri <
> > > > Henri.Roosen@xxxxxxxxxxxxx> wrote:
> > > > > Hi Arnd,
> > > > >
> > > > > I hope you are well and could answer me a quick question.
> > > > >
> > > > > I've read on the kernel mailing-list that initially there was
> > > > > an
> > > > > intention to backport the final y2038 patches to v5.4. We're
> > > > > currently targeting to use the v5.4 LTS kernel for a project
> > > > > which
> > > > > should be y2038 compliant.
> > > > >
> > > > > I couldn't find all of the y2038-endgame patches in the current
> > > > > v5.4-stable branch. Are there any patches still required to be
> > > > > backported in order for v5.4 to be y2038 compliant, or can the
> > > > > remaining patches be ignored (because of only cleanup?)? Else,
> > > > > is
> > > > > there still an intention to get the v5.4 LTS kernel y2038
> > > > > compliant?
> > > >
> > > > I don't think there are currently any plans to merge my y2038-
> > > > endgame
> > > > branch
> > > > into the official linux-5.4 lts kernel, but you should be able to
> > > > just pull from
> > > >
> > > > https://git.kernel.org/pub/scm/linux/kernel/git/arnd/playground.git/log/?h=y2038-endgame
> > > >
> > > > and get the same results. If you see any problems with that,
> > > > please
> > > > report
> > > > that to me with Cc to the mailing list and perhaps gregkh, so I
> > > > can
> > > > see if
> > > > I can resolve it by rebasing my patches, or if he would like to
> > > > merge
> > > > the
> > > > patches.
> > >
> > > Pulling the y2038-endgame branch does lead to some conflicts, which
> > > are
> > > currently still kinda staightforward to solve.
> > >
> > > However I'd be very interested in getting this branch merged to
> > > v5.4,
> > > so we don't run into more difficult merge conflicts the coming
> > > years
> > > where the v5.4-LTS still gets stable updates (Dec, 2025) and
> > > possibly
> > > to get any related fixes from upstream.
> > >
> > > @Greg: any chance to get the y2038-endgame merged into v5.4.y?
> >
> > I have no idea what this really means, and what it entails, but odds
> > are, no :)
>
> I fully understand, thanks for your statement on this.
>
> >
> > Why not just use a newer kernel? Why are you stuck using a 5.4
> > kernel
> > for a device that has to live in 2038? That feels very foolish to
> > me...
>
> Oh I agree on that :) It's just that these are currently customer
> requirements.

Are you sure that customers really understand what they want?

Usually they want a well-supported, stable, system. Why do they care
about a specific kernel version? That feels odd.

Good luck!

greg k-h