-
Notifications
You must be signed in to change notification settings - Fork 59
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
rawhide: upgrades fail with new openssh-9.0p1-10.fc38.1 #1394
Comments
Instead of a %post that code should be run as a systemd unit. |
Pushed a draft PR in https://src.fedoraproject.org/rpms/openssh/pull-request/39 |
The issue described in coreos/fedora-coreos-tracker#1394 has been fixed upstream in fedora-selinux/selinux-policy#1574 and landed in `selinux-policy-38.6-1.fc38`.
I've updated my PR from the discussion at the meeting (same link). |
We discussed this during the community meeting today. We were unable to convince the maintainer to continue support for relaxed file permissions on host keys, but we did find out how to discover the in use hostkeys (as mentioned in #1394 (comment) @travier updated https://src.fedoraproject.org/rpms/openssh/pull-request/39 to use the new method). We also discussed ways we could roll this out to FCOS users. @travier @jlebon and I will get together tomorrow to flesh out the strategy there. |
opened https://bugzilla.redhat.com/show_bug.cgi?id=2172956 so we can submit this as a freeze exception for Fedora 38 beta. |
@travier @jlebon and I discussed this last week. The two options we discussed were:
I believe we are leaning towards |
We discussed this in today's community meeting:
|
New updates landed in bodhi with the fix:
These were fast-tracked in coreos/fedora-coreos-config@d3b0047 This issue never made it into a released production stream of Fedora CoreOS so we can close it now. |
@dustymabe Did you test this script on a system with a non |
The chmod/chonw calls in https://src.fedoraproject.org/rpms/openssh/pull-request/40#_3__35 are missing quotes. But agree it's unlikely to happen. |
sigh.. of course I didn't. I liked the |
How do we tell if https://github.com/openssh/openssh-portable/blob/36c6c3eff5e4a669ff414b9daf85f919666e8e03/authfile.c#L108 ever gets translated? |
I think just calling it as |
I'll PR that change if anyone knows the answer to #1394 (comment) |
@dustymabe Not sure I understand the question. The idea behind |
Looking at the openssh-portable codebase, I'm not actually sure OpenSSH is internationalized at all. |
Indeed. I see a few |
OK, I hope we get away with this as it makes me very uncomfortable to parse error output as root and do chmod/chown changes like that. Maybe we should remove those files completely once we've made a barrier release. |
We're kind of bound by what the |
Presumably we'll need to continue supporting custom configs with mode 640 host keys indefinitely. I've added a regression test for that part in coreos/fedora-coreos-config#2285. The test also implicitly exercises the upgrade path, but that isn't an explicit goal, since I wanted to make the test independent of the particular mitigation technique (or of whether any mitigation is needed at all). |
I'm not sure we can 100%. We can support it if the user wrote it out via Ignition (like your test does), but if they configure it some other way (config management of some sort) after the |
Right, I meant via Ignition. Config management is out of scope for FCOS. |
Fedora will have to keep it until Fedora 39 (2 releases) and can drop it in Fedora 40 as they support skipping a release. We have barrier releases so we could remove all of this earlier. |
Because of #1394 (comment), don't we have the opposite problem? Once the maintainers decide to drop it, we'll need to decide whether to keep carrying the mitigation or not. Ideally we can just give a heads up to users and let it drop out. |
OK, if users created host keys with But I'd say that's reasonable as it would be new systems, not existing systems that would break, so announcing the change and waiting should be good enough. |
Exclude the embedded SSH private key from Red Hat secret scanning. For coreos/fedora-coreos-tracker#1394.
coreos-status email: https://hackmd.io/aw7c1xNLRNSxcomW63Carw |
Looks like we've had one report of translations messing us up on this in the wild:
|
Nevermind. I think it was a typo from the user. |
Updates in https://src.fedoraproject.org/rpms/fedora-release/pull-request/252. The change only works for FCOS right now. |
https://bodhi.fedoraproject.org/updates/FEDORA-2023-bb12d0efe6 > This would be good to test via a fast-track in FCOS CI. |
Exclude the embedded SSH private key from Red Hat secret scanning. For coreos/fedora-coreos-tracker#1394.
Exclude the embedded SSH private key from Red Hat secret scanning. For coreos/fedora-coreos-tracker#1394.
The upgrade tests started failing in
38.20230126.91.0
. One of the transitions there is:If you recreate the upgrade test yourself and have serial console to a machine:
then (to rebase to a local build):
then after reboot from the serial console I can see the sshd service is failed and why:
Which means that existing systems that are upgraded will need their host keys migrated to adhere to the new stricter permissions requirements.
In the distgit commit for this change there is an RPM scriptlet to migrate the existing host keys to conforming permissions.
We need to figure out how to handle this on OSTree systems.
The text was updated successfully, but these errors were encountered: