RTP Tools (Version 1.20)

The authors cannot provide support for compiling or running the rtptools. We will gladly accept bug fixes, but all other email regarding the rtptools will be ignored.

Description

The rtptools distribution consists of a number of small applications that can be used for processing RTP data.

rtpplay
Play back RTP sessions recorded by rtpdump
rtpsend
Generate RTP packets from textual description, generated by hand or rtpdump
rtpdump
Parse and print RTP packets, generating output files suitable for rtpplay and rtpsend
rtptrans
RTP translator between unicast and multicast networks; also translates between VAT and RTP formats.

Installation

Sources for a variety of platforms and binaries for Windows are available from http://www.cs.columbia.edu/IRT/rtptools/download.

The RTP tools should compile on any Posix-compliant platform supporting sockets, as well as Windows/NT/95/98/2000 (Win32). They have been tested on SunOS 4.1, SunOS 5.x (Solaris), Linux, NT 4.0, SGI Irix, and HP-UX. Edit the directories and libraries at the top of Makefile and type make. The compiler must support ANSI C: gcc does, Sun's old /usr/ucb/cc does not.

Note: You must use the sun4 architecture for SunOS 4.1.x and sun5 for SunOS 5.x (Solaris). You will get system call errors if you do not.

For Unix systems, type
./configure; make
To install RTP tools on WIN32 machine, please follow the following steps:
*.dsp files are project files. *.dsw file and workspace file.
User can open the workspace file and use 'batch compile' to compile all the projects.
  1. In Visual C++ 6.0, open workspace file rtptools.dsw.
  2. In VC menu Build, use Batch Build to build all the tools.
  3. All the rtptools will be created under "debug\" directory.

For quite a character, who desire to compile on Borland C++ Builder, please open dump_bcb.bpr, play_bcb.bpr, send_bcb.bpr and trans_bcb.bpr under bcb directory. Only pressing ctr-F9 needed for compilation, and the tool will be generated on the same directory.

General Usage Hints

Network addresses can be either multicast or unicast addresses, unless stated otherwise. They may be specified in dotted-decimal notation (e.g., 224.2.0.1) or as a host name (e.g., lupus.fokus.gmd.de). Port numbers must be given as decimal numbers in the range of 1 to 65535. Network addresses are specified as destination/port/ttl. The time-to-live (ttl) value is optional and only applies to multicast.

For all commands, the flag -h or -? will print a short usage summary.

Unless otherwise noted, input is taken from stdin, and output sent to stdout. The extension .rtp is suggested for files generated in rtpdump -F dump format.

rtpplay

rtpplay [-T] [-v] [-f file] [-p profile] [-s sourceport] [-b begin] [-e end] destination/port[/ttl]

rtpplay reads RTP session data, recorded by rtpdump -F dump from either the file or stdin, if file is not specified, sending it to network address destination and port port with a time-to-live value of ttl.

If the flag -T is given, the timing between packets corresponds to the arrival timing rather than the RTP timestamps. Otherwise, for RTP data packets, the timing given by the RTP timestamps is used, smoothing interarrival jitter and restoring packet sequence. RTCP packets are still sent with their original timing. This may cause the relative order of RTP and RTCP packets to be changed.

The source port(localport) for outgoing packets can be set with the -s flag. A random port is chosen if this flag is not specified.

The whole file is played unless the begin or end times are specified. Times are measured in seconds and fractions from the beginning of the recording.

The RTP clock frequency is read from the profile file if given; the default profile (RFC 1890) is used if not. The profile file contains lines with two fields each: the first is the numeric payload type, the second the clock frequency. The values read from the profile file are silently ignored if the -T flag is used.

If you want to loop a particular file, it is easiest to put the rtpplay command in a shell script.

The -v flag has rtpplay display the packets generated on stdout.

rtpplay uses the hsearch (3C) library, which may not be available on all operating systems.

rtpdump

rtpdump [-F format] [-t duration] [-x bytes] [-f file] [-o outputfile] address/port

rtpdump listens on the address and port pair for RTP and RTCP packets and dumps a processed version to outputfile if specified or stdout otherwise.

If file is specified, the file is used instead of the network address. If no network address is given, file input is expected from stdin. The file must have been recorded using the rtpdump dump format.

The recording duration is measured in minutes.

From each packet, only the first bytes of the payload are dumped (only applicable for "dump" and "hex" formats).

Supported formats are:

format text/binary description
dump binary dump in binary format, suitable for rtpplay. The format is as follows: The file starts with
#!rtpplay1.0 address/port\n

The version number indicates the file format version, not the version of RTP tools used to generate the file. The current file format version is 1.0.

This is followed by one binary header (RD_hdr_t) and one RD_packet_t structure for each received packet. All fields are in network byte order. The RTP and RTCP packets are recorded as-is.

typedef struct {
  struct timeval start;  /* start of recording (GMT) */
  u_int32 source;        /* network source (multicast address) */
  u_int16 port;          /* UDP port */
} RD_hdr_t;

typedef struct {
  u_int16 length;    /* length of packet, including this header (may 
                        be smaller than plen if not whole packet recorded) */
  u_int16 plen;      /* actual header+payload length for RTP, 0 for RTCP */
  u_int32 offset;    /* milliseconds since the start of recording */
} RD_packet_t;
header like "dump", but don't save audio/video payload
payload only audio/video payload
ascii text parsed packets (default), suitable for rtpsend:
844525628.240592 RTP len=176 from=131.136.234.103:46196 v=2 p=0 x=0
   cc=0 m=0 pt=5 (IDVI,1,8000) seq=28178 ts=954052737 ssrc=0x124e2b58
844525628.243123 RTCP len=128 from=139.88.27.43:53154 
 (RR ssrc=0x125bd36f p=0 count=1 len=7
(ssrc=bc64b658 fraction=0.503906 lost=4291428375 last_seq=308007791
  jit=17987961 lsr=2003335488 dlsr=825440558)
 )
 (SDES p=0 count=1 len=23
  (src=0x125bd36f CNAME="yywhy@139.88.27.43" NAME="Michael Baldizzi
  (NASA LeRC)" TOOL="vat-4.0a8" EMAIL="mbaldizzi@lerc.nasa.gov" )
 )
hex like ascii, but with hex dump of payload
rtcp like ascii, but only RTCP packets
short RTP or vat data in tabular form: [-]time ts [seq], where a - indicates a set marker bit. The sequence number seq is only used for RTP packets.
844525727.800600 954849217 30667
844525727.837188 954849537 30668
844525727.877249 954849857 30669
844525727.922518 954850177 30670

rtpsend

rtpsend sends an RTP packet stream with configurable parameters. This is intended to test RTP features. The RTP or RTCP headers are read from a file, generated by hand, a test program or rtpdump (format "ascii").

rtpsend [-a] [-l] [-s sourceport] [-f file] destination/port[/ttl]

Packets are sent with a time-to-live value ttl.

If data is read from a file instead of stdin, the -l(loop) flag resends the same sequence of packets again and again.

The source port(localport) for outgoing packets can be set with the -s flag. A random port is chosen if this flag is not specified.

If the -a flag is specified, rtpsend includes a router alert IP option in RTCP packets. This is used by the YESSIR resource reservation protoccol.

The file file contains the description of the packets to be sent. Within the file, each entry starts with a time value, in seconds, relative to the beginning of the trace. The time value must appear at the beginning of a line, without white space. Within an RTP or RTCP packet description, parameters may appear in any order, without white space around the equal sign. Lines are continued with initial white space on the next line. Comment lines start with #. Strings are enclosed in quotation marks.

<time> RTP 
   v=<version>
   p=<padding>
   x=<extension>
   m=<marker>
   pt=<payload type>
   ts=<time stamp>
   seq=<sequence number>
   ssrc=<SSRC>
   cc=<CSRC count>
   csrc=<CSRC>
   data=<hex payload>
   ext_type=<type of extension>
   ext_len=<length of extension header>
   ext_data=<hex extension data>
   len=<packet size in bytes(including header)>
<time> RTCP (SDES v=<version> 
              (src=<source> cname="..." name="...")
              (src=<source> ...)
            )
            (SR v=<version>
              ssrc=<SSRC of data source>
              p=<padding>
              count=<number of sources>
              len=<length>
              ntp=<NTP timestamp>
              psent=<packet sent>
              osent=<octets sent>
                (ssrc=<SSRC of source>
                 fraction=<loss fraction>
                 lost=<number lost>
                 last_seq=<last sequence number>
                 jit=<jitter>
                 lsr=<last SR received>
                 dlsr=<delay since last SR>
                ) 
            )

rtptrans

rtptrans [host]/port[/ttl] [host]/port[/ttl] [...]

rtptrans RTP/RTCP packets arriving from one of the addresses to all other addresses. Addresses can be a multicast or unicast. TTL values for unicast addresses are ignored. (Actually, doesn't check whether packets are RTP or not.)

Additionally, the translator can translate VAT packets into RTP packets. VAT control packets are translated into RTCP SDES packets with a CNAME and a NAME entry. However, this is only intended to be used in the following configuration: VAT packets arriving on a multicast connection are translated into RTP and sent over a unicast link. RTP packets are not (yet) translated into VAT packets and and all packets arriving on unicast links are not changed at all. Therefore, currently mainly the following topology is supported: multicast VAT -> translator -> unicast RTP; and on the way back it should lokk like this multicast VAT <- translator <- unicast VAT. This means that the audio agent on the unicast link should be able use both VAT and RTP.

Authors

The rtptools were written by Henning Schulzrinne, with enhancements by Ping Pan and Akira Tsukamoto. rtptrans was written by Dorgham Sisalem and enhanced by Steve Casner.


Program history
Last updated by Henning Schulzrinne