Hi,

Three questions below about login classes, on the themes of what's a sensible way to work with them and how they relate to user privilege escalation (not at all I hope).


WHAT I'M TRYING TO DO:
I am going to run some processes that have higher memory and file descriptor quota requirements than the other processes that OpenBSD runs otherwise.

I will have some designated user accounts for those processes to run in (user accounts a, b and c, all non-root), so I can as well have the quotas associated with those user accounts -

My idea is that OpenBSD's pre-installed services have optimal settings already, so I shouldn't alter any of their settings, but instead, I better associate my separate memory and descriptor quotas which are especially designated for those particular activities that I am up to, to those user accounts I designated for my activities (a, b, c).

(Also I like those users to have a particular default umask for all files they create.)


QUESTION 1:
Looking for a "best practice" way of doing this, it is quite clear from the documentation that I should attribute those special quotas to a login class that I have and that is designated for my activities, and then assign those users (a, b, c) to that login class.

I got it right, right?


QUESTION 2:
To understand the ramifications of the login class concept:

(As long as noone flips on the "wheel" option for any login class,) a user's login class belonging *DOES NOT* imply any particular user privileges of any kind, right?

And in particular, the "staff" login group does not imply any particular significant privileges, right?

(I see that it has an ":ignorenologin" setting, that would be all. So no particular OS call abilities, and no particular admin-like privileges.)


QUESTION 3:
Looking in /etc/master.passwd at what login class belonging present users have,

 * "root" belongs to the "daemon" login class,

* "_pbuild" belongs to the "pbuild" login class, "unbound" belongs to the "unbound" login class,

* and all other users that are pre-setup at OS install time either belong to the "daemon" login class or to no login class at all, which is interpreted as belonging to the "default" login class.

This means we have one login class predefined by OpenBSD at installation time namely the "staff" login class, which has no member users at OS install time.

This means that I can use the "staff" login class for anything I want - it's even proper to say that the "staff" login class is really well suited for my use case as it already exists so why create a new one, right?


So, all in all, do you think it makes sense for me to add the extra memory and file descriptor quotas and umask rules to the "staff" login class and attribute my users to it, or should I create a new login class e.g. a class "high_resource_users"?


Thanks,
Tinker

Reply via email to