W3C home > Mailing lists > Public > ietf-http-wg-old@w3.org > May to August 2000

apache source included....Re: CRLF on POST requests, where/how specified (repost from prior bad subject line)

From: Brad Taylor <btaylor@neonsys.com>
Date: Tue, 20 Jun 2000 14:08:30 -0500
To: "Life is hard, and then you die" <ronald@innovation.ch>
cc: "Fred Bohle" <fbohle@neonsys.com>, http-wg@cuckoo.hpl.hp.com
Message-ID: <86256904.006931C1.00@internet1.neonsys.com>

06/20/2000 02:08 PM
Brad Taylor@NEON
Brad Taylor@NEON
Brad Taylor@NEON
06/20/2000 02:08 PM
06/20/2000 02:08 PM

 * More machine-dependent networking gooo... on some systems,
 * you've got to be *really* sure that all the packets are acknowledged
 * before closing the connection, since the client will not be able
 * to see the last response if their TCP buffer is flushed by a RST
 * packet from us, which is what the server's TCP stack will send
 * if it receives any request data after closing the connection.
 * In an ideal world, this function would be accomplished by simply
 * setting the socket option SO_LINGER and handling it within the
 * server's TCP stack while the process continues on to the next request.
 * Unfortunately, it seems that most (if not all) operating systems
 * block the server process on close() when SO_LINGER is used.
 * For those that don't, see USE_SO_LINGER below.  For the rest,
 * we have created a home-brew lingering_close.
 * Many operating systems tend to block, puke, or otherwise mishandle
 * calls to shutdown only half of the connection.  You should define
 * NO_LINGCLOSE in ap_config.h if such is the case for your system.

/* Special version of timeout for lingering_close */

static void lingerout(int sig)
#ifdef NETWARE
    if (alarms_blocked) {
     alarm_pending = 1;

    if (!current_conn) {
     ap_longjmp(jmpbuffer, 1);
    ap_bsetflag(current_conn->client, B_EOUT, 1);
    current_conn->aborted = 1;

static void linger_timeout(void)
#ifdef NETWARE
    timeout_name = "lingering close";
    ap_set_callback_and_alarm(lingerout, MAX_SECS_TO_LINGER);

/* Since many clients will abort a connection instead of closing it,
 * attempting to log an error message from this routine will only
 * confuse the webmaster.  There doesn't seem to be any portable way to
 * distinguish between a dropped connection and something that might be
 * worth logging.
static void lingering_close(request_rec *r)
    char dummybuf[512];
    struct timeval tv;
    fd_set lfds;
    int select_rv;
    int lsd;

    /* Prevent a slow-drip client from holding us here indefinitely */


    /* Send any leftover data to the client, but never try to again */

    if (ap_bflush(r->connection->client) == -1) {
    ap_bsetflag(r->connection->client, B_EOUT, 1);

    /* Close our half of the connection --- send the client a FIN */

    lsd = r->connection->client->fd;

    if ((shutdown(lsd, 1) != 0) || r->connection->aborted) {

    /* Set up to wait for readable data on socket... */


    /* Wait for readable data or error condition on socket;
     * slurp up any data that arrives...  We exit when we go for an
     * interval of tv length without getting any more data, get an error
     * from select(), get an error or EOF on a read, or the timer expires.

    do {
     /* We use a 2 second timeout because current (Feb 97) browsers
      * fail to close a connection after the server closes it.  Thus,
      * to avoid keeping the child busy, we are only lingering long enough
      * for a client that is actively sending data on a connection.
      * This should be sufficient unless the connection is massively
      * losing packets, in which case we might have missed the RST anyway.
      * These parameters are reset on each pass, since they might be
      * changed by select.
#ifdef NETWARE

     FD_SET(lsd, &lfds);
     tv.tv_sec = 2;
     tv.tv_usec = 0;

     select_rv = ap_select(lsd + 1, &lfds, NULL, NULL, &tv);

    } while ((select_rv > 0) &&
             (read(lsd, dummybuf, sizeof dummybuf) > 0));

    /* Should now have seen final ack.  Safe to finally kill socket */


#endif /* ndef NO_LINGCLOSE */

"Life is hard, and then you die" <ronald@innovation.ch> on 06/20/2000
01:21:07 PM

To:   Fred Bohle/Dev/Neon
cc:   http-wg@cuckoo.hpl.hp.com (bcc: Brad Taylor/Neon)

Subject:  Re: CRLF on POST requests, where/how specified (repost from prior
      bad subject line)

On Tue, Jun 20, 2000 at 12:51:11PM -0500, Fred Bohle wrote:
>      No, not the CRLF between the headers and the body.  We are asking
>      about a CRLF that follows the body.  It does not seem to be in
>      the 1.0 spec, and the 1.1 spec seems to specifically prohibit it.
>      And yet, IE5 and Netscape both will send a CRLF after the body
>      for Content-type: application/x-www-url-encoded.

The spec clearly disallows this extra CRLF, i.e. it's a bug in those
browsers. However, there is an easy workaround which many people use:
when reading the request line, ignore all whitespace (or empty lines)
until you hit the actual request line.

>      What do other web servers do to handle this CRLF?  We find that
>      if we decide to close the connection with the end of the response
>      we generate, (Connection: close) and THEN the CRLF arrives from
>      the client, the TCP layer will generate a Reset packet.  This
>      causes (IE5 at least) the client to fail processing the response
>      we just sent.  So the application stops, dead in the water.

This is a little tricky, but not much. Basically you need to do a
shutdown on the socket, i.e. only close for sends; then you keep
reading from the socket (and just discard the data) until either you get
an EOF (in which case the client did the close), or until some timeout
(a few seconds) in case the client doesn't close the connection. Note
that this problem is a general problem (i.e. not limited to the extra
CRLF), because you can get the same effect when a client pipelines



P.S. take a look at the Apache source: read_request_line() in
     for the first problem, and lingering_close() in http_main.c for
     the second - the comments are enlightening.
Received on Tuesday, 20 June 2000 20:09:40 UTC

This archive was generated by hypermail 2.3.1 : Wednesday, 7 January 2015 14:40:24 UTC