[FFmpeg-devel] [PATCH 1/3] avformat/rtsp: add support for satip://

[Aman Karmani]

On Tue, Dec 22, 2020 at 12:17 PM Andriy Gelman <andriy.gelman at gmail.com>
wrote:

> On Mon, 21. Dec 16:35, Aman Karmani wrote:
> > From: Aman Karmani <aman at tmm1.net>
> >
> > The SAT>IP protocol[1] is similar to RTSP. However SAT>IP servers
> > are assumed to speak only MP2T, so DESCRIBE is not used in the same
> > way. When no streams are active, DESCRIBE will return 404 according
> > to the spec (see section 3.5.7). When streams are active, DESCRIBE
> > will return a list of all current streams along with information
> > about their signal strengths.
> >
> > Previously, attemping to use ffmpeg with a rtsp:// url that points
> > to a SAT>IP server would work with some devices and fail due to 404
> > response on others. Further, if the SAT>IP server was already
> > streaming, ffmpeg would incorrectly consume the DESCRIBE SDP response
> > and join an existing tuner instead of requesting a new session with
> > the URL provided by the user. These issues have been noted by many
> > users across the internet[2][3].
> >
> > This commit adds proper spec-compliant support for SAT>IP, including:
> >
> > - support for the satip:// psuedo-protocol
> > - avoiding the use of DESCRIBE
> > - parsing and consuming the com.ses.streamID response header
> > - using "Transport: RTP/AVP;unicast" because the optional "/UDP"
> >   suffix confuses some servers
> >
> > [1]
> https://www.satip.info/sites/satip/files/resource/satip_specification_version_1_2_2.pdf
> > [2]
> https://stackoverflow.com/questions/61194344/does-ffmpeg-violate-the-satip-specification-describe-syntax
> > [3] https://github.com/kodi-pvr/pvr.iptvsimple/issues/196
> > ---
> >  libavformat/rtsp.c    | 52 ++++++++++++++++++++++++++++++++++++++-----
> >  libavformat/rtsp.h    |  6 +++++
> >  libavformat/rtspdec.c |  1 +
> >  3 files changed, 53 insertions(+), 6 deletions(-)
> >
> > diff --git a/libavformat/rtsp.c b/libavformat/rtsp.c
> > index c7ffa07d9e..4a863dbac3 100644
> > --- a/libavformat/rtsp.c
> > +++ b/libavformat/rtsp.c
> > @@ -252,6 +252,24 @@ static void
> finalize_rtp_handler_init(AVFormatContext *s, RTSPStream *rtsp_st,
> >      }
> >  }
> >
> > +static int init_satip_stream(AVFormatContext *s)
> > +{
> > +    RTSPState *rt = s->priv_data;
> > +    RTSPStream *rtsp_st;
> > +    rtsp_st = av_mallocz(sizeof(RTSPStream));
> > +    if (!rtsp_st)
> > +        return AVERROR(ENOMEM);
> > +    rtsp_st->stream_index = -1;
> > +    dynarray_add(&rt->rtsp_streams,
> > +                 &rt->nb_rtsp_streams, rtsp_st);
> > +    av_strlcpy(rtsp_st->control_url,
> > +               rt->control_uri, sizeof(rtsp_st->control_url));
> > +    rtsp_st->sdp_payload_type = 33; // MP2T
> > +    init_rtp_handler(&ff_mpegts_dynamic_handler, rtsp_st, NULL);
> > +    finalize_rtp_handler_init(s, rtsp_st, NULL);
> > +    return 0;
> > +}
> > +
> >  /* parse the rtpmap description: <codec_name>/<clock_rate>[/<other
> params>] */
> >  static int sdp_parse_rtpmap(AVFormatContext *s,
> >                              AVStream *st, RTSPStream *rtsp_st,
> > @@ -1116,6 +1134,9 @@ void ff_rtsp_parse_line(AVFormatContext *s,
> >      } else if (av_stristart(p, "Content-Type:", &p)) {
> >          p += strspn(p, SPACE_CHARS);
> >          av_strlcpy(reply->content_type, p, sizeof(reply->content_type));
> > +    } else if (av_stristart(p, "com.ses.streamID:", &p)) {
> > +        p += strspn(p, SPACE_CHARS);
> > +        av_strlcpy(reply->stream_id, p, sizeof(reply->stream_id));
> >      }
> >  }
> >
> > @@ -1495,8 +1516,10 @@ int ff_rtsp_make_setup_request(AVFormatContext
> *s, const char *host, int port,
> >          rtp_opened:
> >              port = ff_rtp_get_local_rtp_port(rtsp_st->rtp_handle);
> >          have_port:
>
> > -            snprintf(transport, sizeof(transport) - 1,
> > -                     "%s/UDP;", trans_pref);
> > +            av_strlcat(transport, trans_pref, sizeof(transport));
>
> transport needs to be initialized to an empty string. Otherwise av_strlcat
> will
> keep all the random bytes before the first NULL.


Thanks for catching this. Updated the v2 patch to use av_strlcpy, which
makes more sense anyway given this code is inside a loop.



>
> --
> Andriy
>