-
Notifications
You must be signed in to change notification settings - Fork 29
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
Auth failure on Centos9 Stream #39
Comments
I don't know how to track down the problem, but I suspect it is related to el9 distros using OpenSSL 3.0 and something having changed there. I'm experiencing the same problem with Rocky Linux 9. |
To track down the problem, could you please enable LEGACY crypto policy? |
I enabled the policy and things started to work again. Let me know if you need more information. Happy to help. |
Try to use the last build of OpenSSH (-19 in C9S) |
I don't have C9S, but I found the SRPM for the -19 OpenSSH and built it on an updated Rocky 9. After installing the newer versions of packages for what I already have it works with the DEFAULT instead of LEGACY crypto profiles. |
For what it's worth this looks like it could be a C9 specific issue
somehow, there's a bug open at
https://bugzilla.redhat.com/show_bug.cgi?id=2070113
Not sure this is strictly a OpenSSL3/LEGACY issue, as I'm using the same
on F36 (which also has OpenSSL3) with the same crypto settings, and that
works fine.
…On Mon, Jul 18, 2022 at 03:05:34PM -0700, shafer wrote:
I don't have C9S, but I found the SRPM for the -19 OpenSSH and built it on an updated Rocky 9. After installing the newer versions of packages for what I already have it works with the DEFAULT instead of LEGACY crypto profiles.
--
Reply to this email directly or view it on GitHub:
#39 (comment)
You are receiving this because you authored the thread.
Message ID: ***@***.***>
|
Yes, and this bug was fixed recently. I don't remember exactly to what extent SHA1-disablement patches are brought to F36. |
Run command update-crypto-policies --set DEFAULT:SHA1 in RHEL 9.0, then reboot. The problem disappeared. But my pubkey format in .ssh/authorized_keys is SHA256:xxx. It has nothing to do with SHA1. |
Yes. The problem is that to proof the key ownership, sha1 signature is used internally. |
Would you happen to know which particular bit in the crypto policies
governs this specific use case? I'm running the same setup on C9S and
Fedora, and the Fedora boxes mention a whole lot less SHA1 in
/etc/crypto-policies/, yet things still work. Does RHEL9 carry more
patches to SSH that extends the crypto-policy coverage?
|
Another problem also disappeared after running update-crypto-policies --set DEFAULT:SHA1 in RHEL 9.0. I'm trying to setup a SSH hostbased authentication cluster with RHEL 9.0. Passwordless login between nodes always failed:
After running update-crypto-policies --set DEFAULT:SHA1, Passwordless login between nodes working too:
|
OpenSSH in RHEL/Fedora uses OpenSSL 3.0 features to make FIPS certification easier. So the changes in CP are not located in the the OpenSSH file but in the OpenSSL one. |
For those interested, the important setting is
[evp_properties]
rh-allow-sha1-signatures = yes
in /etc/crypto-policies/back-ends/opensslcnf.config. This defaults to
"yes" on fedora in the DEFAULT policy, but to "no" on RHEL9 with the
DEFAULT policy.
|
I'm running the same pam_ssh_agent_auth configuration across a number of RH derived systems: Centos7. Centos8 Stream, Centos9 Stream, Fedora.
This configuration works on all of them, except for C9S, and I can't really figure out why.
Installation of the pam_ssh_agent_auth lib, configuration of
/etc/pam.d/sudo
and/etc/sudoers
, creation of users and installation of.../authorized_keys
is managed through ansible, and is the same on all of the above.On C9S, attempting to sudo (as the user
sun
) results in the below logs:Setting a password for the user, and using that to
sudo
works, so the user is allowed tosudo
in general.In comparison, on all other systems (here a C8S) the same sequence results in these logs:
As far as I can tell there's no difference in how pam_ssh_agent_auth is built between the different OSs. Any hint as to where in the code this might go wrong would be appreciated.
The text was updated successfully, but these errors were encountered: