MAC before DAC vs DAC before MAC

From: richard offer (offerat_private)
Date: Wed Jul 25 2001 - 11:56:53 PDT

  • Next message: Greg KH: "Re: Patch: Socket hooks"

    We've been over this before, but to summaraize and concentrate the
    discussions into hopefully a mutually dissapointing conclusion, here goes.
    
    
    There are two projects (SubDomain, our B1 code) that have mutually
    conflicting requirements.
    
    
    SubDomain has a policy that allows normal DAC checks to procede and then
    enforces MAC checks if it gets that far. For its policy (I think of it as
    containing), this makes perfect sense.
    
    B1 want to perform the MAC check first, if that is acceptable, then the
    normal DAC checks can procede. Partly this is audit driven (if a check
    would return a MAC fail, it should, not a DAC fail) partly customer
    expectation. 
    
    
    Possible Solutions
    ==================
    
    Implement DAC before MAC
        Pro:        No code changes, SubDomain is happy
        Con:        SGI unhappy, fails litmus test of supporting all reasonable
    
                    policies
    
    Implement MAC before DAC        
        Pro:        SGI happy,
        Con:        SubDomain unhappy, extra work, fails litmus test of 
                    supporting all reasonable policies
    
    Hooks pre and post DAC logic
        Pro:        Works for both camps
        Con:        People belive it wont pass muster by the kernel list.
                    
    
    Single Hook to replace all DAC logic
        Pro:        Works for both camps
        Con:        Means moving DAC logic out of the kernel into the module,
                    People belive it wont pass muster by the kernel list,
                    extra work.
    
    Build time selection of hook placement
    
    #ifdef MAC_BEFORE_DAC
            security_ops->hook()
    #endif
            /* in-line DAC logic */
    #ifndef MAC_BEFORE_DAC
            security_ops->hook()
    #endif
    
        Pro:        works for both camps
        Con:        REALLY ugly, #ifdefs, possibly limits policies that could
    be 
                    loaded.
    
    
    
    Those are the only options I can see, anyone any others ?
    
    
    richard.
    
    _______________________________________________
    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 : Wed Jul 25 2001 - 11:57:35 PDT