systemd: 249 -> 250 by jonringer · Pull Request #156096 · NixOS/nixpkgs

jonringer · 2022-01-21T20:51:52Z

Motivation for this change

Continuation of #150491

Opening PR so I can start large build job before getting on a plane.

Things done

flokli · 2022-01-21T21:04:42Z

This needs to target staging, and needs to be bumped to systemd/systemd-stable v250.3.

Furthermore, we should at least skim over the changes from v250..v250.3 and verify that they don't break any NixOS things.

jonringer · 2022-01-21T22:45:11Z

Yea, traveling right now. I'll mark as draft until it's actually ready

jonringer · 2022-01-21T22:54:05Z

I'll do a proper rebase and merge conflict analysis once I get home.

jonringer · 2022-01-22T23:00:14Z

@ofborg eval

nixos/doc/manual/release-notes/rl-2205.section.md

pkgs/os-specific/linux/systemd/default.nix

Izorkin · 2022-01-23T20:12:25Z

@jonringer commit systemd: 249.5 -> 250.3 needed rename to systemd: 249.7 -> 250.3

jonringer · 2022-01-23T21:00:28Z

rebased on latest master, will rebase onto staging once it's ready to merge. I would like to wait a staging cycle, as there was the python changes merged to staging, which I would prefer to stabilize in a different staging-next cycle from systemd changes

Mic92 · 2022-01-24T07:16:56Z

I saw your PR too late and opened up #156438

pkgs/os-specific/linux/systemd/default.nix

Mic92 · 2022-01-24T11:45:36Z

I am closing this just now to avoid pings on rebase...

jonringer · 2022-02-03T01:46:42Z

I'm running a hydra job on this: https://hydra.jonringer.us/jobset/nixpkgs/systemd250

flokli · 2022-02-28T09:33:18Z

I think we could probably merge into staging after next staging-next cycle begins

@jonringer there have been some more staging cycles (#158080, #158080).

This also needs a rebase, as there have been some conflicts. Is there anything else preventing this from being merged? (I see the PR title still contains a "WIP" too).

Mic92 · 2022-02-28T12:59:42Z

I am using this for some time on my laptop now without noticeable impact so far.

jonringer · 2022-02-28T15:11:37Z

@flokli I was going to use this staging cycle for glibc #133431 (comment). Once it begins, I think we can merge this.

jonringer · 2022-02-28T15:11:47Z

I can still rebase though

ius

Fails to build on EFI enabled RISC-V because the requested EFI linker (efi-ld=gold) is unsupported. According to Wikipedia gold only supports x86, x86-64, ARM, PowerPC, TileGX.

~~If there is no particular reason to pick gold over bfd, I suggest passing -Defi-ld=bfd instead.~~

Edit: I suggest just removing the option altogether, so it will figure out a suitable default by itself. See comment about changed semantics below as to why this is okay for v250.

ius · 2022-03-05T15:55:12Z

pkgs/os-specific/linux/systemd/default.nix

Did this build before this PR?

Yes. In systemd v249.7 the variable was used to invoke the linker directly (see meson.build), so that's why ld was valid and defaulted to bfd.

In systemd v250.3 it's passed to -fuse-ld instead. meson.build

Actually, because of these new semantics, we can just drop the option. It will figure out the default linker by itself, or use bfd if it's anything other than bfd or gold.

#150491 (comment)

We don't have to do that as we already set all the feature flags to null. Setting individual libraries to null instead of disabling their feature flag will lead with bad example that will cause each of the features to be disabled with multiple flags in the systemdMinimal variant. If a dependency is pulled in via another feature we should disable that rather than setting it to null. Overriding a given package should be the last resort.

This allows us to make test-only dependencies optional in builds that aren't running tests (sadly all of our builds).

When initializing a system (e.g. first boot / livecd) we have no good reference source for time. systemd-timesyncd however would revert back to its configured fallback time (in our case 01.01.1980). Since we probably don't want to hardcode a specific date as fallback we are now using the current system time (wherever that might have come from) to initialize the reference clock file. The only systems that might be remotely affected by this change are machines that have highly unreliable RTCs or those where the battery that backs the RTC is running empty. Historically these systems always had a tough time with anything time related and likely required manual intervention. For stateless systems (those that wipe / between reboots or our installer CDs) this has the consequence that time will always be reset to whatever the system comes up with on boot. This is likely the correct time coming from an RTC. No harm done here the situation is likely unchanged for them. For stateful systems (those that retain the / partition across reboots) there shouldn't be a change at all. They'll provide an initial clock value once on their lifetime (during first boot / after installation). From then onwards systemd-timesyncd will update the file with the newer fallback time (that will be picked up on the next boot).

This helps systemd during runtime to make decisions about the sanity of the system clock. See the references news article for more details on the matter.

We do not run them, so it is unnecessary work.

As reported in NixOS#156096 (review), this fails to build on EFI enabled RISC-V because the requested EFI linker (efi-ld=gold) is unsupported. According to Wikipedia gold only supports x86, x86-64, ARM, PowerPC, TileGX. Removing this option alltogether will cause meson to figure out the default linker by itself.

flokli · 2022-03-05T20:35:05Z

I rebased this onto latest staging, fixed the conflicts, and dropped -Defi-ld=gold as suggested in #156096 (comment).

Considering this has been open for quite some time, and tested by some people already, I'll run the nixosTests.systemd{-boot} tests once more, and then we should merge this.

flokli · 2022-03-05T21:29:05Z

nixosTests.systemd-boot succeeded.

nixosTests.systemd couldn't run, due to /nix/store/i3axx4kkdbfl01r0dfsps9qnj8biryxh-python3.9-scikit-build-0.12.0.drv being broken right now. However, this has been sufficiently tested previously.

Let's get this into staging.

vcunat · 2022-03-17T08:51:03Z

This merge broke nixosTests.containers-ip. I checked the commit right before this merge and the merge commit (after some bisection steps, as I started with a range of many hundred commits).

Maybe also some other small-channel blockers from this list regressed together with it: https://hydra.nixos.org/eval/1748773#tabs-still-fail (you can see logs there or reproduce cheaply on that particular commit)

flokli · 2022-03-17T09:27:53Z

Thanks for the heads-up, I'll take a look.

flokli · 2022-03-17T21:47:08Z

I built this locally. The container startup fails while trying to bind-mount /nix/var/nix/daemon-socket into the container:

machine # [    5.364468] systemd[1]: Starting Container 'webserver'...
machine # [    5.448067] container webserver[879]: Spawning container webserver on /var/lib/containers/webserver.
machine # [    5.448712] container webserver[879]: Press ^] three times within 1s to kill container.
machine # [    5.450500] container webserver[885]: Failed to stat /nix/var/nix/daemon-socket: No such file or directory
machine # [    5.454873] systemd[1]: container@webserver.service: Main process exited, code=exited, status=1/FAILURE
machine # [    5.455510] systemd[1]: container@webserver.service: Failed with result 'exit-code'.
machine # [    5.456098] systemd[1]: Failed to start Container 'webserver'.
machine # [    5.763319] systemd[1]: container@webserver.service: Scheduled restart job, restart counter is at 2.
machine # [    5.769093] systemd[1]: Stopped Container 'webserver'.
machine # [    5.770123] systemd[1]: Starting Container 'webserver'...

I ran the interactive test driver, and looked at the VM. Indeed /nix/var/nix/daemon-socket doesn't exist, because nix-daemon.socket failed to activate:

[  237.187747] systemd[1]: Nix Daemon Socket was skipped because of a failed condition check (ConditionPathIsReadWrite=/nix/var/nix/daemon-socket).

flokli · 2022-03-17T21:59:08Z

Okay, I figured this out - this is actually a bug in the upstream nix-daemon.socket unit.

I have a workaround in nixpkgs, and will send a fix to the Nix repo.

flokli · 2022-03-17T22:13:18Z

@vcunat the workaround is in #164644, and the proper fix in Nix is in NixOS/nix#6282.

ofborg bot added the 2.status: merge conflict This PR has merge conflicts with the target branch label Jan 21, 2022

ofborg bot removed the 2.status: merge conflict This PR has merge conflicts with the target branch label Jan 22, 2022

ofborg bot requested review from andir, edolstra, flokli and kloenk January 23, 2022 00:04

flokli reviewed Jan 23, 2022

View reviewed changes

nixos/doc/manual/release-notes/rl-2205.section.md Outdated Show resolved Hide resolved

flokli reviewed Jan 23, 2022

View reviewed changes

pkgs/os-specific/linux/systemd/default.nix Outdated Show resolved Hide resolved

ofborg bot requested a review from flokli January 23, 2022 19:46

Mic92 reviewed Jan 24, 2022

View reviewed changes

pkgs/os-specific/linux/systemd/default.nix Outdated Show resolved Hide resolved

Mic92 reviewed Jan 24, 2022

View reviewed changes

pkgs/os-specific/linux/systemd/default.nix Outdated Show resolved Hide resolved

Mic92 mentioned this pull request Jan 24, 2022

systemd: v249.7 -> v250.3 #156438

Closed

13 tasks

Mic92 closed this Jan 24, 2022

Mic92 reopened this Jan 24, 2022

aszlig mentioned this pull request Feb 2, 2022

systemd-confinement: Fix with services using restrictive UMasks #147639

Draft

ofborg bot requested review from WilliButz, globin, kalbasit and mmahut February 7, 2022 00:06

ius suggested changes Mar 5, 2022

View reviewed changes

andir and others added 9 commits March 5, 2022 21:26

systemd: reformat code with nixpkgs-fmt

0423158

systemd: 249.7 -> 250.3

3869ce7

systemd: remove unused lvm2 input

0c852e1

systemd: introduce withTests flag

e6280a6

This allows us to make test-only dependencies optional in builds that aren't running tests (sadly all of our builds).

systemd: add the release timestamp into the build

49267a9

This helps systemd during runtime to make decisions about the sanity of the system clock. See the references news article for more details on the matter.

systemd: do not patch test files

f592c5a

We do not run them, so it is unnecessary work.

flokli force-pushed the systemd-250 branch from 1a054c4 to f5c243d Compare March 5, 2022 20:32

flokli merged commit 180c7c7 into NixOS:staging Mar 5, 2022

Uh oh!

Conversation

jonringer commented Jan 21, 2022

Motivation for this change

Things done

Uh oh!

flokli commented Jan 21, 2022

Uh oh!

jonringer commented Jan 21, 2022

Uh oh!

jonringer commented Jan 21, 2022

Uh oh!

jonringer commented Jan 22, 2022

Uh oh!

Uh oh!

Uh oh!

Izorkin commented Jan 23, 2022

Uh oh!

jonringer commented Jan 23, 2022 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Mic92 commented Jan 24, 2022

Uh oh!

Uh oh!

Uh oh!

Mic92 commented Jan 24, 2022

Uh oh!

jonringer commented Feb 3, 2022

Uh oh!

flokli commented Feb 28, 2022

Uh oh!

Mic92 commented Feb 28, 2022

Uh oh!

jonringer commented Feb 28, 2022

Uh oh!

jonringer commented Feb 28, 2022

Uh oh!

ius left a comment • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

ius Mar 5, 2022 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

flokli Mar 5, 2022

Choose a reason for hiding this comment

Uh oh!

ius Mar 5, 2022 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

mweinelt Mar 5, 2022

Choose a reason for hiding this comment

Uh oh!

flokli commented Mar 5, 2022

Uh oh!

flokli commented Mar 5, 2022

Uh oh!

vcunat commented Mar 17, 2022

Uh oh!

flokli commented Mar 17, 2022 via email

Uh oh!

flokli commented Mar 17, 2022

Uh oh!

flokli commented Mar 17, 2022

Uh oh!

flokli commented Mar 17, 2022

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

8 participants

jonringer commented Jan 23, 2022 •

edited

Loading

ius left a comment •

edited

Loading

ius Mar 5, 2022 •

edited

Loading

ius Mar 5, 2022 •

edited

Loading