libosmocore > 1.4.0 is required (master, not yet released) since some
fixes done in osmo-bsc code where not cherry-picked to libosmocore APIs.
Depends: libosmocore.git I2bf5635b8536b11d69774d17ac1908019633e3af
Change-Id: I7d5e5ddd174463c2a3d957c8245d2911ce013681
Steve Langasek <steve.langasek@ubuntu.com> submitted some patches
against downstream osmo-bsc 1.3.0 because some possible null derefences
were detected by the compiler on Ubuntu s390x. Code has eveolved since
then and patch doesn't apply directly anymore, since related code
changed (we now use osmo_count in bsc_subscr_get).
The compiled allegedly claimed some null dereference in gsm_lchan_name.
In general code using that function seems to be doing checks for
existing lchan before calling it, or assuming the lchan pointer is not
null, so I couldn't find any major issue.
However, let's add a OSMO_ASSERT to make sure we can easily identify the
issue if an issue ever happens there, since the gsm_lchan_name should
clearly only be called on non null pointers.
Change-Id: If4d12cb1d95ee2a89244bb8f27df839871667387
This is needed in order to to proper feature support verification for
IPv6 when configuring the NSVC.
Before this patch, there could be a race condition where NSVC FSM
checked for BTS feature BTS_FEAT_IPV6_NSVC before it was negotiated
through BTS Get Attributes (Ack).
Fixes: OS#4870
Change-Id: I7c207eee0e331995ae04acec014fbd13d4d16280
Before this patch, Get Attributes was sent quicklyafter the OML link
became up, even if the BTS/BB_TRANSC objects were still powered off,
which is wrong since attributes should only be available after the
objects transition out of the Power off state.
Furthermore, information about get attr response already received will
be required in future patches to delay NSVC setting.
Related: OS#4870
Change-Id: I8ec39c7e1f956ffce9aecd58a5590c43200ba086
There's no real need to retrieve the trx before passing it to the
function, we can do that in the function itself and hence also simplify
the function itself.
Related: OS#4870
Change-Id: I7181510c5021ff2712c09ebc6ec8b13fdd8e8dc2
The condition was set in st_op_disabled_notinstalled_on_enter():
"""
site_mgr->peer_has_no_avstate_offline = (bts->type == GSM_BTS_TYPE_NANOBTS);
"""
However, at startup of the FSM the oneneter func of the default initial
state is not called. In any case, if called it would be too early since
the type is not known yet (because its parsed later on at the VTY with
the "type" command, that's after the "bts X" node is called and the
bts_sm is allocated.
So we need to make sure to always enable it, also for nanobts.
Change-Id: Ic6049a44ae3fca1b8e968fe800c268f579e7cad4
The only real 1-1 relationship between BTS NM objects is the one between
GPRS Cell and BTS (which is actually a BTS cell).
In our current osmo-bts implementation we don't care much since we only
handle 1-cell BTSses, but let's make the data structure organization
more generic.
Implementation notes:
The gsm_bts_sm is moved to its own file, APIs to allocate are added and
the new public object is hooked correctly in the allocation process of
osmo-bsc.
Change-Id: I06461b7784fa2a78de37383406e35beae85fbad8
MS Classmark 3 is optional, and thus can be NULL.
Change-Id: I4f1455a3db4972ea9843564b590e405c51083b47
Fixes: I39ae439d05562b35b2e47774dc92f8789fea1a57
Fixes: CID#215593 "Explicit null dereferenced"
The CTRL connect already includes 30 retries with .1 seconds delay each, so
there is no point in adding a fixed two second delay that unnecessarily drags
out test duration for each test.
Change-Id: I830fabbd037d6e945956a99aa3f5e4d4015ea165
In order to activate FACCH/SACCH repetition on the BTS, the classmark 3
IE in the CLASSMARK CHANGE message must be parsed and depending on the
Repeated ACCH Capability bit the RSL_IE_OSMO_REP_ACCH_CAP is added to
the RSL CHAN ACT und RSL CHAN MODE MODIFY. Since
RSL_IE_OSMO_REP_ACCH_CAP is a propritary IE, it may only be added for
BTS type osmo-bts.
Change-Id: I39ae439d05562b35b2e47774dc92f8789fea1a57
Related: OS#4796 SYS#5114
To be able to control the FACCH/SACCH repetition behavior inside the
BTS a one byte flag is sent to the BTS together with the
RSL_IE_OSMO_REP_ACCH_CAP IE. This patch adds the necessary VTY commands.
The sending of the flag is implemented in a follow-up patch.
See also: I39ae439d05562b35b2e47774dc92f8789fea1a57
Related: SYS#5114, OS#4796, OS#4794, OS#4795
Depends: libosmocore I6dda239e9cd7033297bed1deb5eb1d9f87b8433f
Change-Id: I083eaa2c30478912426e9c24a506f0b88836e190
When ICMI becomes zero for 'start-mode auto', the smod bits will remain
whatever start-mode was set in the previous osmo-bsc config. Instead, osmo-bsc
should clear the smod bits for 'start-mode auto' so that its MultiRate Config
does not vary depending on what was previously configured.
Change-Id: I1ec5bad0bce01cc425ee05ecf70c83ec662a226a
So far, the smod bits reflecting the start mode never made it to the
MultiRate Config IE: gsm48_mr_cfg_from_gsm0808_sc_cfg() always sets them
to zero, and the mode filtering fails to carry the BTS set bits.
Set smod bits according to BTS config after the filtering of available
modes.
Change-Id: I49691df01745a7c485bf165e897872c35fc4b147
Send the proper ICMI value, instead of always sending ICMI=1 regardless
of the AMR start-mode setting.
To avoid test fallout from the fix, we should first merge
osmo-ttcn3-hacks I4cff01c37d5c7e301e9a01f773b7e009a789519b
Change-Id: I577ff590d7588fd7e3ee4846c7955ab8f84cf2b1
Previous code relied on abort() switching sigaction to SIG_FDL +
retriggering SIGABRT in case the signal handler returns, which would
then generate the coredump + terminate the process.
However, if a SIGABRT is received from somewhere else (kill -SIGABRT),
then the process would print the talloc report and continue running,
which is not desired.
Change-Id: Iff920ff3dbeb48bd871b7578470f27fe9d0f9516
Fixes: OS#4865
The IPV6_NSVC feature is required to use IPv6 NSVC to connect
to the SGSN. The features in general should be dynamic discovered for
the BTS which support this.
Workaround for OS#4870
Related: SYS#4915, OS#4870
Change-Id: I711efca931012b8e66516f2721390e9dbdbb72a8
Previous code relied on abort() switching sigaction to SIG_FDL +
retriggering SIGABRT in case the signal handler returns, which would
then generate the coredump + terminate the process.
However, if a SIGABRT is received from somewhere else (killall -SIGABRT
osmo-bsc), then the process would print the talloc report and continue
running, which is not desired.
Change-Id: I125288283af630efa20d64505e319636964a0982
Fixes: OS#4865
This was started in 2017 before the NITB split. Let's align it with
reality, mostly in terms of naming and making explicit what was
implemented.
Change-Id: Ib4e66b565a1edfc5ea8c1446f911011bfd4b1642
Shows undesired behavior of moving a TCH/H from a fully used dyn TS
leading to two half-used dyn TS, rather than moving the half-used dyn TS
that would lead to completely freeing a dyn TS.
Change-Id: I3beaa95d12ca0a19d4d1a732f3e22558e68ee29c
When a used timeslot gets moved to another timeslot for load balancing, prefer
moving a dynamic timeslot, as illustrated by handover_test.c test 30.
Rationale: freeing up a dynamic timeslot is better for PDCH availability, as
well as for flexibility in timeslots. Test 30 shows that when freeing a static
TCH/F even though a dynamic one with identical ratings is in use, later
handovers to a TCH/H may become impossible, because no more dynamic timeslots
are available to switch to TCH/H. A freed dynamic timeslot allows congestion
resolution to continue and reduce more TCH/F to TCH/H.
The scope of this preference is per-TRX, where the RXLEV ratings used for
picking a target lchan are the same by definition. In other words, this never
overrules picking another lchan that has better RXLEV.
Among lchans on dynamic timeslots that could be moved, this code favors moving
later lchans; mainly because it makes for a simpler condition in the code.
Change-Id: Ic221b8d2687cdec0bf94410c84a4da43853f0900
There are four places deciding which of 2 lchans to move, depending on average
db ratings. Upcoming patches will enrich that decision for better handling of
dynamic timeslots, so have one common function for these to avoid code dup.
Change-Id: I745dc95cf564dd330295cecb4d64dccebf55163f
Show undesired behavior of opening up another TCH/H timeslot even though
another TCH/H timeslot still has room for a second lchan. This is particularly
bad for dynamic timeslots, reducing PDCH as well as TCH/F resources.
Change-Id: If222835af92d832b848824e5466bdcaf9af8a614
In the test, show the undesired behavior of moving non-dynamic timeslots first,
because they are the first to be considered in the congestion resolution loop.
The behavior will be fixed in Ic221b8d2687cdec0bf94410c84a4da43853f0900.
Change-Id: I09ab9f2f79fa434c7279cb4d09899f69b047aa55
So far the 'meas-rep' expects the index of the MS in order of appearance, to
send a measurement report for that MS. That may seem like a simplification for
human readability, but in fact, it is confusing for a human to keep track which
MS has which index, and to which timeslot it has been handovered. When reading
the tests, it is much easier to read exactly which timeslot the measurement
report is being sent to; which MS came first is inherently not as interesting
as which lchans are currently occupied.
Change-Id: I7ab2ef604196cfe96112dfefb9aa930368bf4812
When a test wants to use N lchans, so far it has to issue N 'create-ms'
commands. Many lchan distributions cannot be easily obtained by this, because
'crate-ms' always takes the first free lchan. To shorten test setup and easily
allow arbitrary lchan distributions, this command sets specific lchans to
specific usage states in one line.
Change-Id: I6e90679bf22609fb72e53e3d11748a4070a38159
In handover tests, it can be hard to follow which lchans are currently
occupied. The expect-ts-use command is like an assertion for a specific
distribution of used lchans, and helps understanding test intentions.
Change-Id: I6f4b573de7faace2d6bfe4f58b568e69c74dd07d
Before this, handover_test channel activation and release on dynamic timeslots
failed to reflect the correct pchan type that osmo-bsc uses.
Change-Id: I3284b4565ad2c3b124442f4373243da1518206ce
Adjust 'create-bts' command so that it allows multiple TRX and arbitrary
timeslot configurations. It is now possible to write tests for dynamic
timeslots and multiple TRX.
Change-Id: Ic645cea671aa4798804666b8886f11bab5351e11
This "global" array shadows the bsc_gsmnet state and is not needed. Look up the
BTS in bsc_gsmnet like all of osmo-bsc does.
Change-Id: Ieb27403b97124771e4d28b9c69bf7c36288f396d
I will add a command to create a single BTS with arbitrary timeslot
configurations.
To limit amount of editing of the current tests, keep the current 'create-bts'
command, which creates N identically configured BTS, but rename it to
'create-n-bts'.
Thus allow the upcoming single-BTS command to take the name 'create-bts'.
Change-Id: I4853771cf23b509b7f278d04c57883332ede786d
Preparation for Ic645cea671aa4798804666b8886f11bab5351e11 to allow arbitrary
timeslot configurations in the tests.
Change-Id: Ie0d0c85552ba24802392423b60cfa3919e87563f
Allow passing timeslot config to create_bts().
Preparation for Ic645cea671aa4798804666b8886f11bab5351e11 to allow arbitrary
timeslot configurations in the tests.
Change-Id: I3ae898be4a503060c6bf30ae89115a75461ffc27
Preparation for Ic645cea671aa4798804666b8886f11bab5351e11 to allow arbitrary
timeslot configurations in the tests.
Change-Id: I8ed468da53f7b0e5ba8125d1bebafebe8c5becbb
Preparation for Ic645cea671aa4798804666b8886f11bab5351e11 to allow arbitrary
timeslot configurations in the tests.
Change-Id: I6eb11e9f98c827cfcd819990db38b6848bf18b27
The function bssmap_handle_cipher_mode() suggests to check if an lchan
is actually present when it gets called, but it only checks for conn.
This might lead to a segfault later in the execution path.
Change-Id: I3103ec89cd6dce1a11ea8e9f8187373e4114e852
Pau Espin Pedrol
Vadim Yanitskiy
Pau Espin
Philipp Maier
Harald Welte