We now treat env REMOVE_USER as username vai GSSAPI auth, and it's from principal's service part. for example: - djoe of djoe@EXAMPLE.COM - bdavid of bdavid@EXAMPLE.COM - service/hostname of service/hostname@EXAMPLE.COM
REMOVE_USER
But we could also create a user with a specified principal whose service part is not exactly it's username for old kerberos way. for example: - bdavid of djoe@EXAMPLE.COM - shortname of service/hostname@EXAMPLE.COM
Then via GSSAPI. bdavid@EXAMPLE.COM will be authenticated as djoe, and authentication as service/hostname@EXAMPLE.COM won't get the real username - shortname.
IMO, that would be a potential security issue.
I'm not sure where the security issue is. Creating users requires admin privileges, and creating arbitrary krb credentials is presumably a privileged action within whatever kerberos infra the Koji instance is using.
I can see how an admin mistake in user creation could cause a problem, but that is the nature of admin access. The ability of root to run rm -rf / doesn't constitute a security bug.,
rm -rf /
Maybe I'm missing something?
I assume that, if the admins of kerberos infra don't take the change of administration of koji, they wouldn't know if the principal is used by more than one koji user. Because for GSSAPI, auth is based on username, and for krbV it's based on krb_principal. The benefit of changing GSSAPI auth to check krb_principal is reducing such kind of administration mistake.
username
krb_principal
But, it's not strictly a security bug.
Metadata Update from @julian8628: - Issue private status set to: False (was: True)
PR #1419
Metadata Update from @yulwang: - Issue tagged with: groomed
Metadata Update from @dgregor: - Issue untagged with: groomed - Issue priority set to: High (was: Normal) - Issue set to the milestone: 1.18
Metadata Update from @dgregor: - Issue tagged with: groomed
Please see https://pagure.io/koji/pull-request/1419#comment-86430
Metadata Update from @mikem: - Custom field Size adjusted to None - Issue set to the milestone: 1.19 (was: 1.18)
We'll fix this shortly after 1.18
Commit 8a0c39f4 fixes this issue
Metadata Update from @dgregor: - Issue assigned to julian8628
This issue has been migrated to Fedora Forge: https://forge.fedoraproject.org/koji/koji/issues/1400
Please continue any further discussion there.