Re: Crashing NT with Native Calls

From: David LeBlanc (dleblancat_private)
Date: Mon Sep 28 1998 - 05:19:40 PDT

  • Next message: *unknown*: "Re: 1+2=3, +++ATH0=Old school DoS"

    At 06:19 PM 9/25/98 +0200, Gigi Mori wrote:
    
    >The bug resides in TCPIP.SYS InternalIoControl dispatch routine where
    lacks a check on current IRQ level before processing the query information
    request.
    
    I think you mean IRQL, which is a little different.  At any rate, this one
    isn't too surprising - any time you start dealing directly with drivers,
    you're dealing with performance-oriented, trusted code.  It is likely that
    all of the other code that calls into this sets the IRQL before going in.
    
    The trade-off to fix all of this is that, of course, to add checks all over
    the place for IRQL level, etc will incur a performance penalty.
    
    There are some interesting security implications in terms of who can do
    what to device drivers - see http://www.sysinternals.com/devsec.htm for a
    good article on this.  An excerpt:
    
    "The vast majority of device objects created have the following
    permissions: Everyone:Read/Write (R/W) (the Everyone built-in Security
    Identifier -SID - is also known as the World SID), System:Full Access, and
    Administrator:Full Access."
    
    IMHO, it would be really nice to have documented APIs implemented in user
    mode for all of the command line apps - for example, nbtstat would be nice
    to be able to call from inside an app.  Obviously what you're doing here is
    trying to do the same thing as netstat from inside an app.
    
    
    David LeBlanc
    dleblancat_private
    



    This archive was generated by hypermail 2b30 : Fri Apr 13 2001 - 14:17:57 PDT