Re: Clarifications of LSM API

From: Valdis.Kletnieks@private
Date: Tue Jun 29 2004 - 09:30:08 PDT

  • Next message: Stephen Smalley: "Re: Clarifications of LSM API"

    On Tue, 29 Jun 2004 11:49:36 EDT, James Morris said:
    
    > The optimization for a single LSM is good, but won't the common case be
    > two LSMs (e.g. capabilities + something else) ?  If so, I'd suggest having
    > two statically allocated entries as primary & secondary, then any further
    > LSMs can be stacked dynamically.  Even then, unless a distro was planning
    > on shipping three or more stacked LSMs by default, I'd wonder if it was
    > really going to be useful in a wider sense.
    
    I could *easily* see "3 or more" LSMs being shipped... Consider:
    
    1) SELinux (I'll even grant you not counting this as a "stacked")..
    2) A harden-/tmp LSM like what I *started* working on..
    3) a harden-chroot LSM like Jail..
    4) a Capabilities LSM...
    
    Over 65% of the LSM that I did isn't even stuff I *wanted* to do, it was stuff
    that got dragged in because I ended up needing to do all of 2,3 and 4....
    
    And sites that have some need to implement a "Users/Processes of Type X can't
    do Operation Y" to satisfy some local requirement (possibly not easily
    expressed in SELinux terms - consider a rule like "Processes run by Graduate
    Students may not access the Foo Shared Resource between 9AM and 5PM M-F"...)
    may be looking at a *fifth* LSM to implement one or two rules....
    
    
    
    



    This archive was generated by hypermail 2b30 : Tue Jun 29 2004 - 09:31:33 PDT