[ 
https://issues.apache.org/jira/browse/ZOOKEEPER-2693?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15876406#comment-15876406
 ] 

ASF GitHub Bot commented on ZOOKEEPER-2693:
-------------------------------------------

Github user hanm commented on a diff in the pull request:

    https://github.com/apache/zookeeper/pull/179#discussion_r102275734
  
    --- Diff: src/java/main/org/apache/zookeeper/server/NettyServerCnxn.java ---
    @@ -267,10 +267,17 @@ private boolean checkFourLetterWord(final Channel 
channel,
         {
             // We take advantage of the limited size of the length to look
             // for cmds. They are all 4-bytes which fits inside of an int
    -        String cmd = FourLetterCommands.getCmdMapView().get(len);
    -        if (cmd == null) {
    +        if (!FourLetterCommands.isKnown(len)) {
                 return false;
             }
    +
    +        // ZOOKEEPER-2693: don't execute 4lw if it's not enabled.
    +        String cmd = FourLetterCommands.getCommandString(len);
    +        if (!FourLetterCommands.isEnabled(cmd)) {
    +            LOG.debug("Command {} is not executed because it is not white 
listed.", cmd);
    +            return true;
    --- End diff --
    
    @rakeshadr hmmmm I think we did not even close server sockets today for 
four letter words in general, see 
    
https://github.com/apache/zookeeper/blob/master/src/java/main/org/apache/zookeeper/server/NIOServerCnxn.java#L343
    and
    
https://github.com/apache/zookeeper/blob/master/src/java/main/org/apache/zookeeper/server/NettyServerCnxn.java#L388
    
    This is not a problem in general for the command line interface because in 
that case client socket will close first and then server socket will close as a 
result of client socket close... however if someone writes a client that opens 
and holds a socket then server will not close the socket even after 4lw finish 
execution. This probably is a by design as it allows clients to pipe 4lw 
commands w/o re-opening sockets but I see this is another potential point of 
vulnerability where a server could run out of sockets..
    
    I think we should probably close the sockets in the two links I posted in 
the beginning. Let me know what you think @rakeshadr.


> DOS attack on wchp/wchc four letter words (4lw)
> -----------------------------------------------
>
>                 Key: ZOOKEEPER-2693
>                 URL: https://issues.apache.org/jira/browse/ZOOKEEPER-2693
>             Project: ZooKeeper
>          Issue Type: Bug
>          Components: security, server
>    Affects Versions: 3.4.0, 3.5.1, 3.5.2
>            Reporter: Patrick Hunt
>            Assignee: Michael Han
>            Priority: Blocker
>             Fix For: 3.4.10, 3.5.3
>
>         Attachments: ZOOKEEPER-2693-01.patch
>
>
> The wchp/wchc four letter words can be exploited in a DOS attack on the ZK 
> client port - typically 2181. The following POC attack was recently published 
> on the web:
> https://webcache.googleusercontent.com/search?q=cache:_CNGIz10PRYJ:https://www.exploit-db.com/exploits/41277/+&cd=14&hl=en&ct=clnk&gl=us
> The most straightforward way to block this attack is to not allow access to 
> the client port to non-trusted clients - i.e. firewall the ZooKeeper service 
> and only allow access to trusted applications using it for coordination.



--
This message was sent by Atlassian JIRA
(v6.3.15#6346)

Reply via email to