Not entirely wait-free: allows waiting on a fd, and in general fd
notifications for poll-loop integration between threads.
Change-Id: I4f17042baf76d086ce6b20eb99402dc64c22c657
There is no obvious reason why we would want to complicate the code by
caching pointers, since pointer traversal is probably not a performance
bottleneck, and if it is we should rather take a look at our dozens of
linked lists first..
Change-Id: I2456ba63598f76200d53e00223abf60bb36a49c0
Usually only one MGCP client per application is present. Then the log
lines from mgcp_client.c will be distinguishable without additional
information. When the application is using a pool of MGWs, then the
various MGCP Client instances become hard to distinguish.
- Add a possibility to set a description (name) for each MGW pool
member. When no description is set, use the domain name.
- Output the pool member name on each log line in mgcp_client.c
and mgcp_client_pool.c
Change-Id: I53ff5445c8e5faffa4ef908ffb1fdb1f47ea2904
Related: SYS#5091
They are mostly not even as large as the talloc header used to
dynamically allocate them, and they are also not "shared" by anything.
Change-Id: I7b46d531c5d3b53984f2ce44538116973f6a074d
No need to complicate audio codes with pointers, our "usual" string is
barely larger than a poointer, five times smaller than a talloc header,
and most importantly not really optional anyway...
Change-Id: Icc41643050a5e1ca3c66f307d60b6911ba1b8032
The change I19f67db1c56473f47338b56114f6bbae8981d067 removes the
policy_cb and change_cb callback funtions, but it does not remove
the related ratecounters.
Change-Id: I53aa3c890555055466e86b09a359375a10d3be7b
At the moment the MGCP Client only supports one MGW per application.
Depending on the requirements of the application one MGW might not offer
the performance needed. Lets add support for an MGCP Client pool that is
backward compatible to existing applications.
Change-Id: Icaaba0e470e916eefddfee750b83f5f65291a6b0
Related: SYS#5091
Depending on the usecase of osmo_mpcg_client it may be helpful to send a
DLCX to certain endpoints. Usually this would be a wildcarded endpoint
that resets the entire trunk to drop lingering RTP flows which may still
present after a restart/crash, but it might be also a group of specific
endpoints. The user may specify an arbitrary amount of endpoints where
the mgcp client will send a DLCX to. It does not matter if the endpoints
are wildcarded or not.
Change-Id: I47e7ff858d5067b46d52329be5f362ff61c0dff8
Related: SYS#5535
The rate counter and stats item groups, which are allocated in
mgcp_ratectr.c are freed using a talloc destructor that is set in the
context of the item we just allocated using the stats / rate counter API
functions. When we do that, we risk overwriting an already existing
talloc destructor. Lets instead implement own free functions and set
those as talloc_destructor from above (trunk and MGCP config)
Change-Id: Ifc5091e9f95cc721e58d1eb2e55b97102c497706
Related: OS#5201
The two callback functions policy_cb and change_cb are essentially dead
code. They also make the code more difficult to read and understand.
Lets remove them.
Change-Id: I19f67db1c56473f47338b56114f6bbae8981d067
We are currently counting events in rate counters, but there is
currently no way to get a sample of the current situation of the trunk
usage. In particular how many endpoints are currently in use.
This is a corrected version of:
Ib7b654168dc3512f55e45cc4755dc1f6f423d023
Change-Id: I6d3a74f6087512130d85002348787bffc672de81
Related: SYS#5201
The request handler handle_delete_con currently rejects wildcarded DLCX
requests even though a wildcarded DLCX would be a valuable tool to
remove lingering connections from the trunk in case osmo-bsc has to be
restarted.
Change-Id: I5c2de6b2b61ee64ba9c0618fd20e8fc2fe6a5ed3
Related: SYS#5535
This reverts commit 6bad138c96.
Reason for revert: heap-use-after-free during 'make check'
in mgcp_test.c test_retransmission()
Change-Id: I96792a719c9c7273676ab9ffe0b9e2aae4c23166
Related: OS#5201
There is not always an endp pointer present when mgcp_check_param() is
called but we always have a trunk pointer. Lets add a trunk parameter so
that the function can pick LOGPTRUNK when endp is not available.
Change-Id: I7327c5a105e7f0e20cabf64623ff9f36fd83bbb8
Related: SYS#5535
We are currently counting events in rate counters, but there is
currently no way to get a sample of the current situation of the trunk
usage. In particular how many endpoints are currently in use.
Change-Id: Ib7b654168dc3512f55e45cc4755dc1f6f423d023
Related: SYS#5535
At the moment the MGCP request handling and message parsing is not
clearly separated. The function mgcp_parse_header() in mgcp_msg.c is
also responsible for resolving an endpoint. This leads to unclear layer
separation. We eventually end up in a situation where we can not execute
any request handler without beeing able to resolve an endpoint, however
this is necessary if we want to implement wildcarded DLCX resquests.
In the current situation a wildcarded DLCX is not possible to implement
as we always have to resolve a an to get to the trunk which we need to
iterate. However, we just can't resolve a free endpoint in a situation
where all endpoints on te trunk are in use.
We have to refactor the request handler so that the parsing in mgcp_msg
only extracts us the endpoint name. The resolving is then done in
mgcp_handle_message() in mgcp_protocol.c. Then we are able to decide
what to do if we are unable to resolve an endpoint but still be able to
resolve the trunk.
This patch does not change the behaviour of osmo-mgw yet, but it lays
the foundation for request handler implementations that can still
perform useful actions if no endpoint but a trunk has been resolved. A
wilcarded DLCX is such a case. It does not need an endpoint, just the
trunk.
Change-Id: I9f519d8a0ee8a513fa1e74acf3ee7dbc0991cdde
Related: SYS#5535
the trunk_nr is in struct mgcp_trunk. The trunk number can not be
negative and there is no magic value that makes use of the fact that it
could be negative. Lets use unsigned int to make this less irretating.
Change-Id: I5d0e1d76adb8c92d84331a0aca2496908e41d621
Related: SYS#5535
osmo-mgw currently does only a very simple detection method for
wildcarded requests, but it makes sense to split this detection
off into a separate function so that it can be used from different code
locations and we still have it at one place only.
Change-Id: I27018c01afb8acabfcf5d435c996cc9806e52d6b
Related: SYS#5535
The function mgcp_rtp_end_config() takes an mgcp_endpoint as
parameter. The header file does not declare the mgcp_endp struct
and it also does not include mgcp_endp.h because that also would
cause problems. Since the endp parameter in mgcp_rtp_end_config()
is only a pointer we can simply forward declare it.
This patch will currently not change anythig, but it will prevent
compiler warnings when we remove the endp pointer from
struct mgcp_parse_data in a follow up patch
Related: SYS#5535
Change-Id: I07a4d6f9d5334b1f4cf4b262482b8a67b1384398
The rate counter group is currently only referenced by an index. In a
system with multiple trunks this makes it difficult to say which rate
counter group belongs to which trunk sinde the index that is used does
not necessarly corespond to a specific trunk.
Since rate counter groups can now get a human readable name assigned, we
should do that.
Also E1 specific rate counters only make sense for E1-trunks, so they
should not be present on the virtual trunk.
Change-Id: I5e7f0e9081a06af48e284afa5c36a095b2847704
On a deployed osmo-mgw with RTP traffic coming from a thirdparty
RTP source, it was usual to see log messages like following one from
time to time:
"The input timestamp has an alignment error of 159 on SSRC"
Doing a quick traffic analysis showed that the above mentioned RTP
source was generating traffic from time to time containing RTP packets
with the Marker (M) bit.
Those messages were logged because the verification & patching funcions
in osmo-mgw were not Marker-bit aware. Hence, this patch implements
support for Marker bit when handling RTP packets.
The Marker bit is usually used as a start of a talkspurt, and has to be
considered a syncrhonization point, where timestamp and relation to real
time don't need to match with last received RTP packet in the stream.
Related: SYS#5498
Change-Id: I1fb449eda49e82607649122b9b9d983a9e5983fa
The following sequence of events was seen frequently in a osmo-mgw
instance running on the field with heavy traffic:
"""
endpoint:rtpbridge/1@mgw CRCX: creating new connection ...
mgcp_network.c:236 endpoint:rtpbridge/1@mgw CI:1C8CCFA9 Failed to send dummy RTP packet.
"""
Allegedly, that happens because at CRCX time the remote address may
still not be known, hence we end up trying to send a dummy rtp packet
to, for instance, host 0.0.0.0 port 0, which will of course fail.
Let's avoid sending it if the address is not yet known.
Similary, same issue could be seen during MDCX, since at MDCX we don't
necessarily need to have a valid addr+port (there could be several MDCX
and only last one set it).
Finally, the keepalive timer also needs the check, since it iterates
over all connections, and it could be that some is still not fully
configured.
Related: SYS#5498
Change-Id: I8ceafda691146823b12232b4a804b4ce74acbdc8
The constant MGCP_DUMMY_LOAD is used ambigously. Sometimes it is used as
initalizer for an array, sometimes it is used as a single byte. Also the
name is not very expressive. Lets refactor this.
Change-Id: I21d96cefeeb647958bfa1e22a0ea030884746fad
Related: OS#4005
When mgcp_udp_send() is called, the memory where *buf is pointing to is
never modified. It should be marked as const.
Change-Id: Iac90de5beb19bf52586ac6ffeab9eb5152edf339
So far an mgcp_client user can get the RTP address+port information that
the MGW has returned upon a CRCX. Add this function to return the other
RTP end, i.e. address+port that the MGW was told to send RTP to.
This will be used to fix the MGCP in osmo-bsc, which so far mixes up the
two RTP ends and compares the MSC's RTP address+port with the MGW's one
and hence fails to skip unnecessary MDCX.
Change-Id: Ibb488925827d9dc0ccb1f8d6d84728745d086793
This is useful for affecting the 802.1Q PCP value without any separate
external packet filter rules for classification.
Change-Id: I69136c6dd114c24b1dace034e75dba5157bac37e
Depends: libosmocore.git I89abffcd125e6d073338a5c6437b9433220e1823
libosmocore If22988735fe05e51226c6b091a5348dcf1208cdf introduces
an even more convenient mechanism for specifying the DSCP of
an IP socket via OSMO_SOCK_F_DSCP()
Change-Id: If0b11dea08716ed3952a25b546b2a9bd013857bf
We have VTY options that allow to set the DSCP value. However, we
then call a function to set the TOS bits in the kernel. This is
very wrong. The DSCP is only the upper 6 bits of the 8-bit TOS
value, and hence we are mussing that translation.
As libosmocore now has a helper function osmo_sock_set_dscp(),
let's make use of it and don't care about the low-level details.
However, this means we need to finally remove the deprecated
alias for "rtp ip-tos <0-255>".
Closes: OS#5137
Change-Id: I9c18c90273be97aedd2ad212b82f650e35c32851
Depends: libosmocore.git Ia4ba389a5b7e3e9d5f17a742a900d6fd68c08e40
To make clear which configuration changes (configure terminal)
apply when, add appropriate user attributes to VTY commands.
Change-Id: I2d9487801b3b78f94577264b56d217c926ef76a9
Related: SYS#4937, OS#1601
The initialization of the E1 line in mgcp_e1.c is controlled internally,
there is no need to expose the function mgcp_e1_init(), lets make it
static and remove the prefix mgcp_
Change-Id: I6aba1c55c9b1d729709ee1fba2994c77bd848a9b
There exist trunk_loop commands, which sets an trunk->audio_loop
variable, however all it does is to turn on a log message. There is no
actual implementation present. Lets set the VTY commands to
DEFUN_DEPRECATED and remove the variable.
Change-Id: I72b0f8b908e32643e6e3db6ac024371b13c074a1
Otherwise some state is kept, like the previous CallId, which may then
provoke issues next time the endpoint is to be used.
Change-Id: I3ac4f4542c1c8c877127c64acce6c82b458f697f
It doesn't make sense to call the function several times since anyway we
are only binding during
allocate_port()->mgcp_bind_net_rtp_port()->bind_rtp()->mgcp_create_bind()->osmo_sock_init2().
Let's better calculate the local IP addr once and use that stored value.
THis is a previous step towards next commit updating the local IP addr
and re-bindng if encessary.
Change-Id: I803b99c5e5fe0f92a5bf6796d8c25df88d1608e6
This commit allows for fully IPv6 systems to work fine. However, if a
remote endpoint still wants to use IPv4, it will fail since at this
point osmo-mgw still doesn't re-bind the local end of the connection to
an IPv4 after having initially bound it to an IPv6 one. This kind of
scenarios get fixed in next commits.
TODO: really bind the socket if a different IP address is requested.
Change-Id: I8ed94bd3f674f498e6ba315f44a351fff9c1be15
This commit contains the bulky work of moving all address parsing to
support IPv6 together with IPv4.
Some specific patches required for full IPv6+IPv4 support requiring
behavioral changes come after this one.
Full Osmux IPv6 support is left out of the scope of this patch.
Depends: libosmocore.git Ie07a38b05b7888885dba4ae795e9f3d9a561543d (> 1.4.0)
Depends: libosmocore.git I59bf4b4b3ed14766a5a5285923d1ffa9fc8b2294 (> 1.4.0)
Change-Id: I504ca776d88fd852bbaef07060c125980db3fdd7
The API and related implementation fields are not used internally nor
externally, and only support IPv4. Let's simply deprecate the API and
drop all the uneeded implementation.
Change-Id: I905d4c4efabb6b4a4bc5c02e956808777243cadc
If "0.0.0.0", the default, is passed together with an IPv6 configured
address (ex: "mgw remote-ip ::1") in VTY, socket creation will fail due
to address version mismatch (because getaddrinfo() returns only an IPv4
address in the local result set).
If instead NULL is passed, then 2 entries are returned, one in IPv4 and
one in IPv6, and osmo_sock_init2 is smart enough to take one or another
when passed AF_UNSPEC.
Change-Id: I1be6f3b71486ce1782ba6b8c62f25145b42ec894
Eric Wild
Philipp Maier
Neels Hofmeyr
Pau Espin
Neels Hofmeyr
Harald Welte