DoS on HP ProCurve 4000M switch (possibly others)

From: Jon Snyder (jonat_private)
Date: Thu Feb 28 2002 - 17:45:52 PST

  • Next message: Justin Piszcz: "[Fwd: BUG: [Kernel 2.4.18 - IP Tables 1.2.4] ?]"

    Advisory Vitals:
    
    Name:				HP ProCurve 4000M nmap DoS
    Affected Products:	HP ProCurve 4000M (J4121A), possibly others
    Firmware Versions:	C.08.22 and C.09.09 both tested vulnerable
    Relevant Vendor URL:	http://www.hp.com/rnd/
    Vendor Contacted:		9/10/2001; 1/16/2002
    
    
    Summary:
    
    nmap portscans cause a DoS on the HP ProCurve 4000M Ethernet switch.
    Depending on the version of firmware, after portscanning the management IP
    address of the switch it is no longer possible to use telnet to manage the
    device.  However, the switch continues to process ICMP messages and SNMP
    PDUs normally, and frames switched by the device also appear unaffected.
    
    
    Details:
    
    Only the HP ProCurve 4000M was tested; a number of other products run the
    same firmware image and may or may not be vulnerable.
    Firmware C.07.01 does not appear to be vulnerable to this issue; numerous
    successive and varied nmap scans against the switch did not affect its
    ability to accept new telnet sessions.
    
    C.08.22 and C.09.09 are vulnerable.  One nmap portscan against the switch's
    management IP address renders the switch unable to accept new telnet
    sessions.  Port 23 remains open, but no text is displayed once connected.
    Eventually (after a number of minutes) this state changes and the switch is
    again able to accept incoming telnet sessions, but a single nmap portscan or
    OS detection attempt immediately renders the switch inaccessible via telnet
    once again.
    
    Existing telnet sessions to the switch appear unaffected during and after
    the portscan.  Also, SNMP continues to function normally, and the switch is
    ping-able even in its 'dead telnet' state.
    
    Console access to the switch does not appear affected.  Rebooting the switch
    is the only way to regain the ability to telnet to it, once it is stuck in
    the described state.
    
    Exacerbating this issue is that the source of the nmap portscan does not
    have to be on the 'Authorized IP Managers' list in the switch for this DoS
    to occur.
    
    
    Vendor Notification:
    
    HP initially confirmed this issue on 9/10/2001 and assigned trouble ticket
    #3200180647.  After some initially positive discussions, I didn't hear from
    them for some time, and called back on 1/16/2002 when I was given another
    case number, #1430333405.  Haven't heard anything since.  Everyone I have
    dealt with at HP has been very friendly, and in all other respects I am very
    happy with the ProCurve switches I have used, but this issue remains
    unresolved.
    
    
    Workaround:
    None known.  A number of bugs have been fixed since C.07.01 and that version
    is no longer available via HP's web site, so running it may not be a viable
    option.  Isolating the management address of the switch from networks that
    may intentionally or unintentionally portscan the switch is the best
    solution in lieu of new firmware from HP.
    
    
    ----------
    Jon Snyder
    



    This archive was generated by hypermail 2b30 : Fri Mar 01 2002 - 11:11:21 PST