Re: Willing to change LSM so secondary defaults correct

From: dwheelerat_private
Date: Sat Dec 28 2002 - 22:20:40 PST

  • Next message: Stephen D. Smalley: "Re: [patch] Hook for d_instantiate"

    Crispen:
    > David: assuming Greg comes up with concrete complaints,
    > what is your rebuttal?
    
    I'm not sure what you're asking, but I'll try to reply anyway.
    The two main issues that Greg has raised are:
    (1) modules won't promise to work with arbitrary other modules, and
    (2) any stacker will (probably) be slow.
    
    For the first point, I agree with Greg; I don't expect modules to
    promise to work with all other modules.  I instead expect that either
    (a) module writers will list specific other modules that they will work with
    (creating a "closed club"), or (b) administrators will have to make
    the determination of whether or not modules can be combined, at their risk.
    For (b), I think that there are many useful special cases where
    modules _CAN_ be shown to work together that it's worth doing.
    E.G., if you can have special-case modules that only prohibit certain
    dangerous actions, and don't use the security blobs, there's a
    a better chance of them working together.  There's no guarantee.
    E.G., the owlsm /tmp file stuff should combine well with many other
    security policies.
    
    The second point is a valid concern.
    This was especially a problem in my first version of the stacker,
    which invoked a lock on every hook invocation (ouch!!).
    However, things have gotten better; by default the latest version only
    locks when a new LSM module is inserted; otherwise it just walks a linked list
    and calls functions in the list. In theory, it SHOULD be relatively cheap
    (yes, there are instruction cache issues, but still...).
    
    However, as Greg says:
    >The last time I looked at the "stacking module" it looked like it had
    >the potential to greatly slow down things, but running real benchmarks
    >would be the only way to tell this.
    
    And I completely agree with Greg; benchmarks are the ONLY valid way to
    really answer the question.
    
    
    _______________________________________________
    linux-security-module mailing list
    linux-security-moduleat_private
    http://mail.wirex.com/mailman/listinfo/linux-security-module
    



    This archive was generated by hypermail 2b30 : Sun Dec 29 2002 - 10:52:09 PST