Re: A Comment from User Space

From: David Wagner (dawat_private)
Date: Tue Apr 24 2001 - 12:42:02 PDT

  • Next message: Kent Crispin: "Re: A Comment from User Space"

    Chris Wright  wrote:
    >is it possible to collapse before_foo into check_foo?
    
    Yes.  The main reason to separate them is to deal with multiple
    registrants who want to hook on the same event.  You call check_foo()
    on all registrants first, then if all of them allow the event, you
    call before_foo() on all of them.  If you have multiple registrants,
    this behaviour can't be emulated with a single "int pre_foo();" hook,
    and if you go with a single "int pre_foo();" hook now and later want
    to support multiple registrants, you may well have to change the
    interface in a non-backwards-compatible way.
    
    The tradeoff is that, in return for ability for the interface to
    upgrade gracefully to multiple registrants per hook without major
    changes to most modules, you end up with double the number of hook
    locations, which is unfortunate.  I'll let you decide which is preferred.
    
    Personally, my guess is that in the long run, if we want to deal with
    multiple registrants, we might want the interface the kernel sees
    (security_ops) might be different from the interface module-writers
    see.  However, in the short run I know everyone would like to get
    things working just in the simple case of a single module per process
    (and I think this view makes a *lot* of sense), so arguably the best
    strategy in the short run is to ignore multiple registrants for the
    current prototype and pay the costs of changing the interface later.
    My main point is just to keep in mind the possibility that there may
    be costs to be born later, if we later want to support multiple policies
    per process.
    
    P.S. I'm not sure whether Janus uses any other post-hooks: I'll have
    to take a closer look.  Off the top of my head, I think the post-hooks
    we used were for fork()/vfork()/... and the process-death event.
    
    _______________________________________________
    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 : Tue Apr 24 2001 - 12:45:00 PDT