Re: backward compat / access (was Re: Benchmarks)

From: Flavien Lebarbe (flavienat_private)
Date: Tue Apr 17 2001 - 16:09:55 PDT

  • Next message: Crispin Cowan: "BoF in Oakland?"

    Hello,
    
    
    aleph1at_private wrote:
    > 
    > You are proposing that the default behavior be to fail open. This type
    > of behavior is not normally considered secure. If a new system call
    > is added to the kernel it may provide a way to bypass the policy
    > enforced the security module.
    
    A newly introduced syscall changes kernel version. Module will not 
    load. Module recompilation against another kernel version is the 
    responsability of :
    1) The module writer :
    	#if (LINUX_VERSION_CODE > 0x02051a)
    	#error SECURITY HOLE : GET A NEW VERSION OF THIS MODULE
    2) the person who recompiles it and blindly removes the above error.
    
    
    Anyway, let's try to find something to avoid above situation :
    
    Default values set at kernel startup have to be "allow", otherwise 
    nothing will be able to start (no fork, no open, no read...
    pretty stuck there ! :) ). 
    What to do with NULL pointer in kernel_set_check_hooks (or whatever 
    it will be called) ? Several solutions :
    
    1) Refuse to initialize. Period. (kinda hard, but hey, kernel hackers
    have to know what they're doing ! )
    2) Default policy "deny" (could be problematic in the future if
    one wants to load several modules one after another, each dealing
    with fs|network|processes parts of the checks)
    3) More likely : Mandatory flag : NULL_KEEPS_WHAT_EXISTS,
       NULL_IS_ALLOW, NULL_IS_DENY. Then the module writer decides 
       what it wants to do, knowing what he's doing.
    
    
    Thanks for pointing this out, encouraging more thinking ! :-))
    
    Flavien.
    
    _______________________________________________
    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 17 2001 - 16:11:08 PDT