I found the problem.  It was a typo on my part in the configuration file (added as I was trying different things to work out why it was seg-faulting earlier).

To find it I just loaded on an earlier version of Chrome with "better" error messages and recognised it as something I'd seen before.

Regards,

Peter

On 30/05/13 19:08, Peter Dunkley wrote:
The reserved bits are part of the WebSocket header that gets put before the actual data.

Looking in Wireshark those bits are (correctly) all 0. But it is not unusual for Chrome to present an error message that is completely inaccurate and it is probably something else in the header that is the problem.

Tomorrow morning I plan to roll-back the build and I should be able to produce two traces that are almost identical (as far as the payload goes) - one for each version.  At that point I will be able to do a proper bit-level comparison between the data in the IP packet.

That might at least give a clue as to which set of commits the problem might have appeared in :-)

Regards,

Peter

On 30 May 2013, at 18:03, Daniel-Constantin Mierla <miconda@gmail.com> wrote:

Hello,

you can have some fun and then revert :-)

Are the reserved bits before or after the websocket payload?

Cheers,
Daniel

On 5/30/13 6:58 PM, Peter Dunkley wrote:
Hi Daniel,

I don't think it is an MSRP issue.  It's just I have first spotted it with MSRP and now can't risk touching my SIP WebSocket server.

My suspicion is that (if it is related to the changes made in the last 10 days) it is something that happens during or after a call to tcp_send().

Regards,

Peter


On 30/05/13 17:56, Daniel-Constantin Mierla wrote:

On 5/30/13 6:44 PM, Peter Dunkley wrote:
I don't have a commit ID for it working as the system that is working was installed from RPMs.  However, It was compiled at 18:39:25 (BST) on May 20 2013.

It will have been a clean pull from Git at the time it was built.
After that date, from core perspective, sctp was moved as a module, syn_branch parameter removed and pv cache used for confg vars, but it doesn't look like should be a direct effect on msrp. A bit before was removal of IPv6 define.

Maybe is a side effect with some memory overflow. Can you compile with MEMDBG=1 and see if you get some errors from memory manager?

Cheers,
Daniel


Peter

On 30/05/13 17:36, Daniel-Constantin Mierla wrote:

On 5/30/13 6:17 PM, Peter Dunkley wrote:
When latest Kamailio git master sends a message over WebSocket (specifically an MSRP reply on my test system) I get the error,
One or more reserved bits are on: reserved1 = 1, reserved2 = 0, reserved3 = 0
in Google Chrome.  This happens for WebSockets over TCP and WebSockets over TLS.  It doesn't happen with a build of Kamailio git master from around two weeks ago.

Have there been any changes in the network code over the last couple of weeks that might have had an effect on what Kamailio puts out on the wire for TCP and TLS?
I don't recall any, but if you give the commit id of the version you are running and it is ok, we can check the rest of the commits till today.

Cheers,
Daniel

The code to send an MSRP reply over WebSocket is in modules/msrp/msrp_netio.c:msrp_reply()
        if (unlikely((env->srcinfo.proto == PROTO_WS
                        || env->srcinfo.proto == PROTO_WSS)
                        && sr_event_enabled(SREV_TCP_WS_FRAME_OUT))) {
                struct tcp_connection *con = tcpconn_get(env->srcinfo.id, 0, 0,
                                                                0, 0);
                ws_event_info_t wsev;

                if (con == NULL)
                {
                        LM_WARN("TCP/TLS connection for WebSocket could not be"
                                "found\n");
                        return -1;
                }

                memset(&wsev, 0, sizeof(ws_event_info_t));
                wsev.type = SREV_TCP_WS_FRAME_OUT;
                wsev.buf = rplbuf;
                wsev.len = p - rplbuf;
                wsev.id = con->id;
                return sr_event_exec(SREV_TCP_WS_FRAME_OUT, (void *) &wsev);
        }

The code that handles the SREV_TCP_WS_FRAME_OUT event is in modules/websocket/ws_frame.c and basically involves:
  • filling in the WebSocket message header
  • identifying the correct TCP/TLS connection
  • setting some flags (for example, SND_F_FORCE_CON_REUSE)
  • calling tcp_send()

These areas of the code haven't been changed for months.

Regards,

Peter




_______________________________________________
sr-dev mailing list
sr-dev@lists.sip-router.org
http://lists.sip-router.org/cgi-bin/mailman/listinfo/sr-dev

-- 
Daniel-Constantin Mierla - http://www.asipto.com
http://twitter.com/#!/miconda - http://www.linkedin.com/in/miconda
Kamailio Advanced Training, San Francisco, USA - June 24-27, 2013
  * http://asipto.com/u/katu *


_______________________________________________
sr-dev mailing list
sr-dev@lists.sip-router.org
http://lists.sip-router.org/cgi-bin/mailman/listinfo/sr-dev



_______________________________________________
sr-dev mailing list
sr-dev@lists.sip-router.org
http://lists.sip-router.org/cgi-bin/mailman/listinfo/sr-dev

-- 
Daniel-Constantin Mierla - http://www.asipto.com
http://twitter.com/#!/miconda - http://www.linkedin.com/in/miconda
Kamailio Advanced Training, San Francisco, USA - June 24-27, 2013
  * http://asipto.com/u/katu *


_______________________________________________
sr-dev mailing list
sr-dev@lists.sip-router.org
http://lists.sip-router.org/cgi-bin/mailman/listinfo/sr-dev



_______________________________________________
sr-dev mailing list
sr-dev@lists.sip-router.org
http://lists.sip-router.org/cgi-bin/mailman/listinfo/sr-dev

-- 
Daniel-Constantin Mierla - http://www.asipto.com
http://twitter.com/#!/miconda - http://www.linkedin.com/in/miconda
Kamailio Advanced Training, San Francisco, USA - June 24-27, 2013
  * http://asipto.com/u/katu *
_______________________________________________
sr-dev mailing list
sr-dev@lists.sip-router.org
http://lists.sip-router.org/cgi-bin/mailman/listinfo/sr-dev


_______________________________________________
sr-dev mailing list
sr-dev@lists.sip-router.org
http://lists.sip-router.org/cgi-bin/mailman/listinfo/sr-dev