-
Notifications
You must be signed in to change notification settings - Fork 694
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
[xenial] Investigate if PAM common-auth customizations are still needed #3963
Comments
It seems like ecryptfs pam module [0] was and still is used by Ubuntu's encrypted home directory [1] functionality. The install documentation instructs users to not enable this feature [2]. Based on the above, I believe it's safe for us to remove this, perhaps making a note in the release notes that this was done, in case an organization has accidentally enabled this feature. Given the critical nature of this configuration file, we want to ensure its state is consistent and predictable across installed: To be mindful of existing installs, using the suggested [0] https://help.ubuntu.com/lts/serverguide/ecryptfs.html
|
If we decide to remove this config option, I see two options:
Option 2 seems like the safest to me, if we can rely on admin intervention for the upgrade. Is it realistic to expect admins to run a |
Agreed that the logic is no longer necessary. Since 0.3.1 (2015-03-23) we've explicitly recommended against encrypting the home directory, and for several releases prior to that, we stated that there was "no need" to encrypt the home directories. Given that breadth of time, we can rather safely assume that instances in the wild are not using the feature (cc @redshiftzero for visibility), in case we wanted to clean up on active systems—a solution to which option 1 above refers. So, no concerns with removing the Ansible logic, as stated in option 2. As for cleaning up via postinst, we'd do well to have the most predictable system state possible going into the Xenial migration, so I'm inclined toward cleaning up. The question becomes: is anything broken in regard to leaving this reference in place? #3519 notes that syslog reports problems with the line as currently structured even on Trusty. Over in #3491 (comment), I myself noted that I commented it out—presumably based on log messages, shame on me for not providing more detail. Given that #3491 (comment) and #3519 were filed on the same day, I suspect they're related to the same report, and not indicative of a wider problem. If we cannot point to any specific breakage under Trusty or Xenial, then option 2 does seem quite reasonable, and our best course of action give the timelines we're working with for the migration. |
The error described in the tickets above ( I think it would be more prudent to revert to the default upstream configuration in all cases.
That should take care of new installs. For existing installs I see 2 options:
If we are planning on having an upgrade playbook, the second option strikes me as the much safer one. |
Given the lack of functional problems reported during the Xenial testing, it seems fine to leave those files as-is for existing instances. Action to close out this ticket should be:
We still have a bit of config drift there, but resetting to the known good values is the safest action going forward, and doesn't involve hacky postinst scripts. |
Examples of failure output in log lines, to inform writing config regression tests in the PR to resolve:
In particular it's the |
The PAM
common-auth
customizations include declarations forpam_ecryptfs.so
which prove problematic on Xenial; commenting out appears to resolve. More research required. We may be able to remove these additions entirely, now that we no longer support 2FA via challenge-response.Part of #3204.
The text was updated successfully, but these errors were encountered: