Re: intercepting system calls

From: Casey Schaufler (caseyat_private)
Date: Fri Apr 13 2001 - 16:11:49 PDT

  • Next message: David Wagner: "Re: The bootstrap process"

    David Wagner wrote:
    
    > In Unix, file descriptors are capabilities.
    
    Whoa boy, now we're getting to the meat of the subject!
    
    In the model that we've used to evaluate U2X systems
    a file descriptor is a name for a file system object.
    The name is private to the process, and the access control
    policy to an object is different depending on what
    kind of name (FD vs. path) you use. The FD policy
    can be looser than the path name policy because the
    name is only recognized after access checks have been
    made on a path name.
    
    > In the end, I guess modules that want good performance can pick
    > a policy with Unix-like semantics where there is no need to mediate
    > read()/write(), and modules that want radically-different policies
    > will need to mediate read()/write() and pay the performance cost.
    > I don't see any way to avoid this tradeoff, and as long as both
    > options are open to both parties, it seems like everyone can go
    > home happy.  Am I right?
    
    I think that's the nut of it.
    
    -- 
    
    Casey Schaufler				Manager, Trust Technology, SGI
    caseyat_private				voice: 650.933.1634
    casey_pat_private			Pager: 888.220.0607
    
    _______________________________________________
    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 : Fri Apr 13 2001 - 16:13:44 PDT