RE: 2.2.14 TCP close problem - race?

From: Tony Gale (gale@syntax.dera.gov.uk)
Date: Wed May 24 2000 - 06:03:56 EDT

  • Next message: Sebastian Ip: "Re: MASQ problem wiht 2.3.45 and Netfilter"

    Sounds like the server doesn't read all the data from the socket
    before close'ing it - this results in a RST, and is correct behaviour.

    Whats the app?

    -tony

    On 24-May-2000 Charlie Brady wrote:
    >
    > A colleague reports the following. Is this a known problem?
    > Comments? More
    > info required? Fixes?
    >
    > ====================
    >
    > We have a problem with some server code and some apparently
    > spurious
    > RST packets.
    >
    > The client establishes a session, sends a request and receives
    > ECONNRESET
    > instead of the data expected. The server side reports no errors. It
    > accepts the connection, reads the request and writes the response
    > (2
    > write calls) and calls close. tcpdump shows everything normal until
    > after the ack by the receiver of the server data. This produces the
    > RST response. There are no FIN packets.
    >
    > Workarounds:
    > sleep(1) in the server code before the close.
    > using 4 byte aligned and sized data packets
    > use a solaris 2.6 system for the server end.
    >
    > My uneducated guess is a race of some sort in the TCP close routine
    > where the closed state is being marked prematurely before all data
    > is
    > sent and acked.
    >
    > Kernel version is 2.2.14-5.0 (RedHat)
    >
    > --
    > Chris Maltby "One man is as good as another. In fact, probably
    > better."
    > chris@sw.oz.au
    >
    >
    > -
    > To unsubscribe from this list: send the line "unsubscribe
    > linux-kernel" in
    > the body of a message to majordomo@vger.rutgers.edu
    > Please read the FAQ at http://www.tux.org/lkml/ 

    ---
E-Mail: Tony Gale <gale@syntax.dera.gov.uk>
This here's the wattle,
The emblem of our land.
You can stick it in a bottle;
You can hold it in your hand.
Amen!
		-- Monty Python

    The views expressed above are entirely those of the writer
and do not represent the views, policy or understanding of
any other person or official body.

    -
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/

    This archive was generated by hypermail 2b29 : Wed May 24 2000 - 06:11:53 EDT