[Iaxclient-devel] Closing In On A Killer... News in the search for the IAX2 Bug

[Steven S. <ss...@so...>]

I have narrowed the section of the code where the inbound call messages
are being lost to a particular sub-section of the 
Ok.  I have narrowed the iax_header_to_event function.  My simple
logging seems to have trapped where the error is occurring.  Now, I need
somebody with more experience to help me figure out WHY we are getting
these errors.

[FYI - I DO have the latest version of the code with all of Steve
Underwood's changes]

Here is my trace:

7897468 Log Message-> Processing Message...
7897484 Log Message-> Packet arrived out of order (expecting 2, got 0)
(frametype = 6, subclass = 1)
7897500 Log Message-> Sending Ack Anyway!
7897500 Log Message-> Returning Null! Exiting In ReSend Section!

7899468 Log Message-> Processing Message...
7899484 Log Message-> Packet arrived out of order (expecting 2, got 0)
(frametype = 6, subclass = 1)
7899484 Log Message-> Sending Ack Anyway!
7899500 Log Message-> Returning Null! Exiting In ReSend Section!

7904203 Log Message-> Processing Message...
7904218 Log Message-> Packet arrived out of order (expecting 2, got 1)
(frametype = 6, subclass = 5)
7904234 Log Message-> Sending Ack Anyway!
7904234 Log Message-> Returning Null! Exiting In ReSend Section!

[FYI - The numbers to the left of the trace are approximate ms since
bootup from the GetTickCount() Win32 function.]

This is coming in the iax_header_to_event function in a section with a
comment that says "Check where we are".  It appears to be validating the
packet arrival sequence.  The above trace shows an out-of-order arrival
of two IAX_COMMAND_NEW messages, and an IAX_COMMAND_HANGUP message when
the non-working call is abandoned at the far end. [See below for a copy
of the code as I am using it...]

Can anybody tell me what would cause the parser to view the incoming
COMMAND_NEW messages as out-of-order.  The "expecting 2, got 0" makes it
look like it was expecting some kind of confirmation for a previous
action or message?  Can we safely bypass this?  Steve, Steve, Michael,
Dan - do you guys know what the preceding message and why we failed to
receive it?

Thanks!

Steve Sokol
Sokol & Associates, LLC




[My Code Including The Debugging.  Attached is the iax.c file from my
system.]

	/* Check where we are */
		if (ntohs(fh->dcallno) & IAX_FLAG_RETRANS)
			updatehistory = 0;
			
		if ((session->iseqno != fh->oseqno) &&
			(session->iseqno || ((subclass !=
IAX_COMMAND_TXCNT) && (subclass != IAX_COMMAND_TXACC)) || (subclass !=
AST_FRAME_IAX))) {
			if (
			  ((subclass != IAX_COMMAND_ACK) &&
			  (subclass != IAX_COMMAND_INVAL) &&
			  (subclass != IAX_COMMAND_TXCNT) &&
			  (subclass != IAX_COMMAND_TXACC) &&
			  (subclass != IAX_COMMAND_VNAK)) ||
			  (fh->type != AST_FRAME_IAX)) {
			 	/* If it's not an ACK packet, it's out
of order. */
				DEBU(G "Packet arrived out of order
(expecting %d, got %d) (frametype = %d, subclass = %d)\n", 
					session->iseqno, fh->oseqno,
fh->type, subclass);
				char Error[200]; 
				sprintf(Error, "Packet arrived out of
order (expecting %d, got %d) (frametype = %d, subclass = %d)\n", 
					session->iseqno, fh->oseqno,
fh->type, subclass);
				iax_debug_log((LPCSTR) Error);
				if (session->iseqno > fh->oseqno) {
					/* If we've already seen it, ack
it XXX There's a border condition here XXX */
					if ((fh->type != AST_FRAME_IAX)
|| ((subclass != IAX_COMMAND_ACK) && (subclass != IAX_COMMAND_INVAL))) {
						DEBU(G "Acking
anyway\n");
						iax_debug_log("Sending
Ack Anyway!");
						/* XXX Maybe we should
handle its ack to us, but then again, it's probably outdated anyway, and
if
						   we have anything to
send, we'll retransmit and get an ACK back anyway XXX */
	
send_command_immediate(session, AST_FRAME_IAX, IAX_COMMAND_ACK, ts,
NULL, 0,fh->iseqno);
					} else {
						iax_debug_log("Sending
Absolutely Nothing...");	
					}
				} else {
					/* Send a VNAK requesting
retransmission */
					iax_debug_log("Sending Vnak!");
					iax2_vnak(session);
				}
				iax_debug_log("Returning Null! Exiting
In ReSend Section!");
				return NULL; //****** THIS IS WHERE WE
EXIT WITH NO EVENT BEING GENERATED!!!
			}
		} else {
			/* Increment unless it's an ACK or VNAK */
			if (((subclass != IAX_COMMAND_ACK) &&
			    (subclass != IAX_COMMAND_INVAL) &&
			    (subclass != IAX_COMMAND_TXCNT) &&
			    (subclass != IAX_COMMAND_TXACC) &&
				(subclass != IAX_COMMAND_VNAK)) ||
			    (fh->type != AST_FRAME_IAX))
				session->iseqno++;
		}