This adds basic support for new 2.0 features, though the protocol is
subposdely backward incompatible, most semantics are the sames.
This commit adds:
- features:
* version negotiation
* pipelined requests support
* Media-Properties support
* Accept-Ranges support
- APIs:
* gst_rtsp_media_seekable
The RTSP methods that have been removed when using 2.0 now return
BAD_REQUEST.
https://bugzilla.gnome.org/show_bug.cgi?id=781446
The RTSP server will not timeout an idle RTSP connection
(note this is different from doing timeout on a RTSP
session).
At least for Apache this is a problem when running RTSP over
HTTPS since it uses one of the threads (there is a rather
limited number) that are available for handling requests.
https://bugzilla.gnome.org/show_bug.cgi?id=771830
These signals let the application validate the requests, configure the
media/stream in a certain way and also generate error status code in
case of error or bad request.
https://bugzilla.gnome.org/show_bug.cgi?id=758062
Call session filter with filter_session_media as paramer in
client_unwatch_session if using drop_backlog = FALSE.
In client_unwatch_session its allowed to grow the watchs backlog.
If using drop_backlog = FALSE and the backlog is full it will cause
a deadlock when setting session media state to NULL
if the backlog is not allowed to grow.
https://bugzilla.gnome.org/show_bug.cgi?id=771983
This is basically reverting changes introduced in commit f62a9a7,
because it was introducing various regressions:
- It introduces a leak of udpsrc elements that got wrongly fixed by adding
an hash table in commit cba045e. We should have at most 4 udpsrc for unicast:
ipv4/ipv6, rtp/rtcp. They can be reused for all unicast clients.
- If a mcast client connects, it creates a new socket in SETUP to try to respect
the destination/port given by the client in the transport, and overrides the
socket already set on the udpsink element. That means that if we already had a
client connected, the source address on the udp packets it receives suddenly
changes.
- If a 2nd mcast client connects, the destination/port in its transport is
ignored but its transport wasn't updated.
What this patch does:
- Revert back to create udpsrc/udpsink for unicast clients on DESCRIBE.
- Always have a tee+queue when udp is enabled. This could be optimized
again in a later patch, but is more complicated. If no unicast clients
connects then those elements are useless, this could be also optimized
in a later patch.
- When mcast transport is added, it creates a new set of udpsrc/udpsink,
seperated from those for unicast clients. Since we already support only
one mcast address, we also create only one set of elements.
https://bugzilla.gnome.org/show_bug.cgi?id=766612
For NTP and PTP clocks we signal the actual clock that is used and signal
the direct media clock offset.
For all other clocks we at least signal that it's the local sender clock.
This allows receivers to know which clock was used to generate the media and
its RTP timestamps. Receivers can then implement network synchronization,
either absolute or at least relative by getting the sender clock rate directly
via NTP/PTP instead of estimating it from RTP timestamps and packet receive
times.
https://bugzilla.gnome.org/show_bug.cgi?id=760005
Postpone the allocation of the UDP sockets until we know
what transport has been chosen by the client.
Both unicast and multicast UDP sources are created in one
function.
https://bugzilla.gnome.org/show_bug.cgi?id=757488
Without TEARDOWN it might be desireable to keep the media running and continue
sending data to the client, even if the RTSP connection itself is
disconnected.
Only do this for session medias that have only UDP transports. If there's at
least on TCP transport, it will stop working and cause problems when the
connection is disconnected.
https://bugzilla.gnome.org/show_bug.cgi?id=758999
SETUP request from clients needs to suspend the media to clear the
prerolled buffers. Otherwise it will not affect the prerolled buffer
and the prerolled buffers will be incorrect (for example block-size
from setup request will not affect the prerolled buffer unless the
media is suspended).
https://bugzilla.gnome.org/show_bug.cgi?id=758268
Add "check-requirements" signal and vfunc to allow application
(and subclasses) to check the requirements.
Based on patch from Hyunjun Ko <zzoon.ko@samsung.com>
https://bugzilla.gnome.org/show_bug.cgi?id=749417
When calling gst_rtsp_watch_write_data in gstrtspconnection.c and
backlog is empty it can happen that just a part of a message will be
sent and rest is in backlog queue. If then flush during teardown
just a part of message will be sent.This can lead to client miss
teardown response since it expect to get the last part of message.
The flushing during teardown was introduced to fix a deadlock that now
is fixed more generally in handle_request by temporary setting backlog
size to unlimited.
Fixes https://bugzilla.gnome.org/show_bug.cgi?id=749845
If the media was just not seekable, we continue from whatever position we are
and let the client decide if that is what is wanted or not.
Only if the actual seek failed, we can't really recover and should error out.
RFC4566 Section 5.2 says that it should make the username, session id,
nettype, addrtype and unicast address tuple globally unique. Always using
1188340656180883 is not going to guarantee that: https://xkcd.com/221/
Instead let's create a 64 bit random number, which at least brings us
closer to the goal of global uniqueness.
https://tools.ietf.org/html/rfc4566#section-5.2
We add a trailing \0 in GstRTSPConnection to make parsing of
string message bodies easier (e.g. the SDP from DESCRIBE) but
for actual data this means we have to drop it or otherwise
create invalid data.
The default implementation of configure_client_transport() in
rtsp-client uses the session media when it chooses channels for
interleaved traffic.
https://bugzilla.gnome.org/show_bug.cgi?id=739112
If the media has been managed by a session media, it should not be
cached in the client any longer. The GstRTSPSessionMedia object is now
responsible for unpreparing the GstRTSPMedia object using
gst_rtsp_media_unprepare(). Unprepare the media when finalizing the
session media.
https://bugzilla.gnome.org/show_bug.cgi?id=739112
We need to set session medias to NULL without the client lock otherwise
we can end up in a deadlock if another thread is waiting for the lock
and media unprepare is also waiting for that thread to end.
https://bugzilla.gnome.org/show_bug.cgi?id=737690
If the backlog limit is kept two cases of deadlocks may be
encountered when streaming over TCP. Without the backlog
limit this deadlocks can not happen, at the expence of
memory usage.
Fixes https://bugzilla.gnome.org/show_bug.cgi?id=737631
As long as gst-rtsp-server can successfully send RTP/RTCP data to
clients then the client must be reading. This change makes the server
timeout the connection if the client stops reading.
Fixes https://bugzilla.gnome.org/show_bug.cgi?id=736647
Allow the send backlog in the RTSP watch to grow to unlimited size while
attempting to bring the media pipeline to NULL due to a session
expiring. Without this change the appsink element cannot change state
because it is blocked while rendering data in the new_sample callback.
This callback will block until it has successfully put the data into the
send backlog. There is a chance that the send backlog is full at this
point which means that the callback may block for a long time, possibly
forever. Therefore the media pipeline may also be prevented from
changing state for a long time.
Fixes https://bugzilla.gnome.org/show_bug.cgi?id=736647
rtsp-client.c:2553:50: error: cast to pointer from integer of different size [-Werror=int-to-pointer-cast]
Just in case that guint8 doesn't fit in a pointer. Just in case ...
We need to raise the backlog limits before pausing the pipeline or else
the appsink might be blocking in the render method in wait_backlog() and
we would deadlock waiting for paused.
Fixes https://bugzilla.gnome.org/show_bug.cgi?id=736322
link/unlink of the transport in a session was done to keep track of all
TCP transports and to send RTP/RTCP data to the streams. We can simplify
that by putting all the TCP transports in a hashtable indexed with the
channel number.
We also don't need to link/unlink the transports when we pause/resume
the streams. The same effect is already achieved when we pause/play the
media. Indeed, when we pause the media, the transport is removed from
the media and the callbacks will not be called anymore.
See https://bugzilla.gnome.org/show_bug.cgi?id=736041
Fixes a crash when close() is called while merging clients
in handle_tunnel(). In that case close() would destroy the
watch while it is still being used in handle_tunnel().
https://bugzilla.gnome.org/show_bug.cgi?id=735570
Release the object lock before calling the filter functions. We need to
keep a cookie to detect when the list changed during the filter
callback. We also keep a hashtable to make sure we only call the filter
function once for each object in case of concurrent modification.
Fixes https://bugzilla.gnome.org/show_bug.cgi?id=732950
This extra ref will be dropped when all client sessions have been
removed. A session is removed when a client sends teardown, closes its
endpoint of the TCP connection or the sessions expires.
Fixes https://bugzilla.gnome.org/show_bug.cgi?id=732226
Once we manage a media in a session, we can't unmanage it anymore
without destroying it. Therefore, first check everything before we
manage the media, otherwise if something is wrong we have no way to
unmanage the media.
If we created a new session and something went wrong, remove the session
again. Fixes a leak in the unit test.
Don't just keep a weak ref to the session objects but use a hard ref. We
will be notified when a session is removed from the pool (expired) with
the new session-removed signal.
Don't automatically close the RTSP connection when all the sessions of
a client are removed, a client can continue to operate and it can create
a new session if it wants. If you want to remove the client from the
server, you have to use gst_rtsp_server_client_filter() now.
Based on patch from Ognyan Tonchev <ognyan.tonchev at axis.com>
See https://bugzilla.gnome.org/show_bug.cgi?id=732226
Add encryption and authentication key length parameters to MIKEY. For
the encoders, the key lengths are obtained from the cipher and auth
algorithms set in the caps. For the decoders, they are obtained while
parsing the key management from the client.
Fixes https://bugzilla.gnome.org/show_bug.cgi?id=730472
Coverity was moaning about unreachable code, and I think it was just
confused by { being before the label. We'll see if it pops up again.
Coverity 1197705
When we have too many messages queued for a client (currently hardcoded
to 100) we overflow and drop the messages. Add a drop-backlog property
to control this behaviour. Setting this property to FALSE will retry
to send the messages to the client by waiting for more room in the
backlog.
Fixes https://bugzilla.gnome.org/show_bug.cgi?id=725898
Add method to create the RTPInfo for a stream-transport.
Add method to create the RTPInfo for all stream-transports in a
session-media.
Use the session-media RTPInfo code in client. This allows us to refactor
another method to link the TCP callbacks.
Implement a vmethod that can be used to configure the media and the
streams based on the current context. Handle the blocksize handling in
the default handler.
See https://bugzilla.gnome.org/show_bug.cgi?id=720667
If the SETUP url contains a query it must be appended to the control
path so that it matches any already created stream in the media. The
query will also be appended to the session media path.
