Re: Unacceptable ACK in SYN_RCVD state handling

[Murali Bashyam <[e-mail]>]

Bob Braden wrote:

>   *> From [e-mail]  Fri Apr  5 11:22:21 2002
>   *> Date: Fri, 05 Apr 2002 11:14:22 -0800
>   *> From: Murali Bashyam <[e-mail]>
>   *> X-Accept-Language: en
>   *> MIME-Version: 1.0
>   *> To: [e-mail]
>   *> Subject: Unacceptable ACK in SYN_RCVD state handling
>   *> Content-Transfer-Encoding: 7bit
>   *> X-AntiVirus: scanned by AMaViS 0.2.1
>   *>
>   *> Hi
>   *>
>   *> In SYN_RECEIVED state after sending a SYN-ACK to the peer, when TCP
>   *> receives a unacceptable ACK
>   *> i.e the ACK lies outside the receive window, it is supposed to drop the
>   *> incoming ACK and respond with a pure ACK as it should for any packet
>   *> that is completely outside the receive window. This is reqd as per RFC
>   *> 793.
>   *>
>   *> In BSD TCP code (FREEBSD/NETBSD) i notice that in SYN-RECEIVED state any
>   *>
>   *> outgoing segment generated would carry the SYN bit as well as the ACK
>   *> bit set. So BSD code, when it responds with a ACK in the above scenario
>   *> also has the SYN bit set. Is this acceptable behaviour? Linux responds
>   *> with a pure ACK in this scenario.
>   *>
>   *> Murali
>   *>
>   *>
>   *>
>
> Hmmm.  I was about to say that:
>
>         The TCP has sent a SYN,ACK and its SYN has not been acknowledged.  It
>         is still in SYN RECEIVED state, so retransmitting the SYN seems to me
>         to be acceptable behavior in this situation.

>
>
> But glancing at RFC 793, it appears that it says to send a RST segment
> at that point (ACK bit on and unacceptable ACK in SYN RECEIVED state,
> page 72 of RFC 793.)

No, the scenario that applies is described on page 68 (sequence number check),
in this
case the ACK carries a sequence number that lies outside the receive window and
hence
the action recommended on page 68 is the following

        If an incoming segment is not acceptable, an acknowledgment
        should be sent in reply (unless the RST bit is set, if so drop
        the segment and return):

          <SEQ=SND.NXT><ACK=RCV.NXT><CTL=ACK>

        After sending the acknowledgment, drop the unacceptable segment
        and return.

The point u mentioned above is the conflict i am having as well, it seems to me
that retransmitting
a SYN-ACK seems the right thing to do. But the BSD stack doesn't seem to be
doing that.

Murali




>
>
> Bob Braden