The BTS_CTR_RACH_DROP counter is being increased in case if a RACH
request is ignored due to exceeding ToA value, but remains untouched
in case of exceeding BER (Bit Error Rate). Let's fix this.
Change-Id: Ia02e781d6c47d9d8012a4c8846fe4b731aab74d7
Before this patch we had:
* osmo-bts-trx internally using 1/256th bit/symbol period
* osmo-bts-sysmo internally using 1/4 bit/smbol period
* PCU interface using 1/4
* L1SAP interface using 1/4
* measurement processing code on top of L1SAP using 1/256
So for sysmo/lc15/octphy we are not loosing resolution, but for
osmo-bts-trx we're arbitrarily reducing the resolution via L1SAP
only then to compute with higher resolution again.
Let's change L1SAP to use 1/256 bits and hence not loose any resolution.
This requires a corresponding change in libosmocore for l1sap.h, which
is found in Change-Id Ibb58113c2819fe2d6d23ecbcfb8b3fce4055025d
Change-Id: If9b0f617845ba6c4aa47969f521734388197c9a7
When decoding RACH bursts, we should use a BER threshold in order to
help distinguish 'ghost' RACH bursts from real RACH bursts.
The theoretical ideal threshold according to some papers is 7 out of 41
bits qhich aquals to Eb/N0 of 0 dB = 0.1707 (17.07%)
We add a new 'ber10k' parameter to the RACH indication l1sap primitive
(needs separate change for libosmocore), and then fill this value from
osmo-bts-{sysmo,lc15,trx,octphy}. The common part above L1SAP then
applies the threshold, which can be changed from vty using the
"max-ber10k-rach <0-10000>"
command available at the BTS node. The unit is BER in 1/10000, i.e. a
value of 100 equals 1% bit error rate.
Change-Id: Ic41c11f6312a36baa2738547e8dcec80829457f8
In the past, rach_busy counting was performed below L1SAP, while
reporting was handled above. This lead to subtle differences between
the BTS models, such as osmo-bts-trx missing to increment rach_busy.
Let's move the rach_busy counting above L1SAP to share more code.
This means we need libosmocore Change-Id
I9439810c3a3ad89ea0302753617b850749af887c for the additional required
parameters in ph_rach_ind_param, as well as libosmocore Change-id
I2b1926a37bde860dcfeb0d613eb55a71271928c5 for osmo-bts-trx to determine
the RACH bit error rate.
Change-Id: I3b989580cb38082e3fd8fc50a11fedda13991092
Closes: OS#3003
info_meas_ind on the L1SAP always allowed the lower layers to pass
in whether a given measurement is part of the "SUB", or not.
However, the existing l1sap code before this patch simply drops this
information, despite the measurement.c code also having "is_sub" state.
Let's make sure this state is passed from L1SAP into measurement
processing as intended.
Fact is, none of our current lower-layers actually set this is_sub flag
for their primitives passed up in L1SAP, but at least now *if* they
would set that flag, the measurement code would process it as intended.
Related: OS#2978
Change-Id: Ibed2e8d7563b471c6b5dd2214ac4765caf31ed2a
This is currently only used for logging, but will be needed for proper
RX{LEV,QUAL}-SUB reporting in upcoming patches.
Related: OS#2978
Change-Id: I07fd06e8a379cab7c0c2eb111c3f5600037d3c9e
Let's make sure whenever we do have a frame number, we print it as
context in the related log line
Change-Id: I751d5ddb3322fce489bc241459738cbcc55c890b
A recurrent kernel crash in sysmobts (several kernel versions)
corrupting kernel memory in random places has been investigated and
reproduced by placing a call against an MSC sending RTP
with bandwidth-efficient AMR payload to osmo-bts-sysmo.
The osmo-bts-sysmo in turn sends the payload to the femtobts related
kernel modules via a msgq, which most probably fail to handle correctly
this bw-efficient AMR payload and corrupt the kernel memory.
First approach was to drop the bw-efficient AMR payloads lower in the
stack in sysmo specific code (l1if_tch_encode), but as there's no bts
model in osmo-bts actually supporting bw-efficient AMR, let's drop it
early in the incoming path for all models to avoid further problems.
Related: SYS#4063
Change-Id: If0c9233c628c724de4ab74e58e3e2affac79e6d0
Previously we always printed error on startup:
l1sap.c:461 Invalid condition detected: Frame difference is 627202-0=627202 > 1!
This is expected and of little practical use. Let's suppress it
by checking that FN was actually initialized.
Change-Id: I4f7cf285e437db0e980e1a2a3ab48a7255945448
Problem:
lchan->tch.dtx.dl_amr_fsm struct failed to allocate in l1sap_chan_act routine
in l1sap.c due to illegal characters contained in lchan->name which are passed to
osmo_fsm_inst_alloc routine. As a result, lchan->tch.dtx.dl_amr_fsm is NULL
causing BTS crashed (SEG FAULT) when trying to access this struct.
Below is snapshot of crash log obtained by GDB:
...
Fri Nov 24 18:13:55 2017 <0000> rsl.c:1653 payload type: 98
Fri Nov 24 18:13:55 2017 <0000> rsl.c:1463 (bts=0,trx=0,ts=2,ss=0)
RSL Tx IPAC_MDCX_ACK (local 127.0.0.1:11538, remote 127.0.0.1:30012)
Program received signal SIGSEGV, Segmentation fault.
0x00031930 in dtx_dl_amr_fsm_step (lchan=lchan@entry=0xb69592a8,
rtp_pl=rtp_pl@entry=0x87ae8 " \024\351Y\363_\337\345\351f\177\373\300\210\201\200\210",
rtp_pl_len=17, fn=1728481, l1_payload=0x10dd25 "", marker=marker@entry=true,
len=len@entry=0x10ddc4 "\024", ft_out=0xbefff7d7 "\002",
ft_out@entry=0xbefff7cf "\276\341_\032") at msg_utils.c:233
233 msg_utils.c: No such file or directory.
...
Fix:
* Use different formatting for lchan name passed to osmo_fsm_inst_alloc routine
* Refuse channel activation if FSM could not be generated (as opposed to crash)
Related: OS#2606
Reported-by: Minh-Quang Nguyen <minh-quang.nguyen@nutaq.com>
Change-Id: I929ce3703dc57acf8db569ae0e346265644d0b3c
Recently while testing new osmo-mgw, big RTP packets (around 4K bytes,
see OS#2625 for more info), were being received on the BTS, which was
aborting with the following message:
"msgb(0xff208): Not enough tailroom msgb_put (348 < 1488)"
The crash can be reproduced in a sysmobts as well as on my PC locally
with osmo-bts-trx. I used osmo-bts-trx to test that the patch solved the
abort.
Fixes: OS#2624
Change-Id: Idfde1dacc3dc3d3d5e239cf1f7e39ade7fc25975
When determining if a frame is a fill frame or not, the case
statement only conditionally handled AGCH and PCH cases. In
case a non-fill-frame was observed, the return value was uninitialized,
resulting in some non-fill-frames to be missed from GMSTAP
Change-Id: I7b46c720e34cb8ef9a91ae5da28a050439a1937d
Closes: Coverity CID#174175
* explicitly set SAPI when sending data_ind to PCU
* drop unused receiving code for BCCH SAPI
* send SI13 when PCU is connected
* send SI13 when new SI is received
Change-Id: I9e83ef792585aa962f99897d9973cef12f186bcf
Related: OS#2400
There's very little point in sending fill frames (such as empty PAGING)
or dummy UI frames via GSMTAP all the time. They serve no purpose other
than to bloat the log files and make it more difficult for users to find
the interesting bits among all this noise.
Change-Id: Icd18dafb235933c9e6aa9d98ddd8fac1522cc9ac
So far, L1SAP code is hiding RSL_CHAN_OSMO_PDCH from the bts specific
code below L1SAP. This is some kind of a hack/workaround, making code
and debug output / logs more difficult to understand.
So let's teach the lower layer how to treat RSL_CHAN_OSMO_PDCH and
remove the "hiding" code from the common l1sap.c code.
Change-Id: Iaaa833febe45b82166d3901f10cc5466a7591c19
The RSL Channel ID is best read / interpreted as hex value, not as
decimal, primarily due to the fact that it is a bit-mask of various
fields.
Change-Id: I9b72a67407870b485e7f7e8a72fa1ad30fc8ed4d
Internal l1sap dump_gsmtime has been moved to libosmocore as osmo_dump_gsmtime.
Remove use of internal function and replace with the libosmocore version.
Depends on libosmocore Ib5452e2c20f53006c0f6d197fb055728947125d8
Change-Id: Ia2f89965d970ed5bbb8c0d4f591a043e58c4bd66
After latest changes, l2 expects to receive an event for every TCH
frame, that is, no TCH frame event should be lost on that layer. We
should now then be safe returning always GSM_RTP_DURATION.
The code which used to calculate the variable duration is left there to
assert that indeed we are not longer having this kind of issues.
Change-Id: I9d112c6db142be138e71393e77129e6d069d9973
We want to always call l1if_tch_rx and l1sap_up in order to avoid losing triggering
events on the upper layer.
With this change, the upper layer will increase correctly seq + ts for
RTP. It will then send an RTP packet with only the header and no payload, which is
not correct but at least we avoid drifting the RTP clock. Upcoming patch
in the series solves this issue.
This patch assumes that we are not lossing data events from the physical
layer and that we receive an event every 20ms, even if the MS is not
transmitting due to DTX.
Depends on libosmocore If4ae20c22b881e94585dad710f17b9e37f77bf82
Change-Id: If5df8940fab833eb4e3ed851880b66987d356031
When we're in loopback testing mode for PDTCH, we must make sure to
avoid adding PTCCH blocks to the queue. Only PDTCH blocks must be
enqueued. For the transmit (downlink) side, we already had the PTCCH
check in place.
Change-Id: I7ef40d9bdf74a99375bc6568ed9483499664bf6f
This add support for BTS-side lookback of PDTCH channels.
If lchan-loopback is enabled, We take the uplink frames as received in
PH-DATA.ind and put them into the dl_tch_queue for that lchan. When we
receive PH-RTS.ind, we dequeue frames from that queue and transmit them
in downlink. If no frame is found in queue, we transmit an empty
(all-zero) frame of 23 bytes (CS-1).
Change-Id: Idd07a3f4a88c38398d3e844333c0104e2de23864
When the lchan is in loopback mode, we loop back all uplink blocks into
downlink blocks. We do not processs any RTP frames for that lchan
anymore. Rather, we discard those RTP frames to avoid mixing
looped-back samples with those received from remote.
Change-Id: I29ef4963e9c491c94c413cbc10436a2388c04d9b
In some situations (e.g. when trying to do measurements/testing on the
BTS receiver / uplink) it is useful to have a way to disable the radio
link timeout and keep any channel open until deactivated, irrespective
of whether (valid) data is received or not.
This adds a related feature that can be activated by using an
osmocom-specific value of 0xff for the TS 12.21 Connection Failure
Criterion (9.4.14).
Change-Id: I736f21f6528db5c16fa80cdb905af20673797be5
Computing the measurement results on in l1sap_info_time_ind() all
at once may peak the host CPU. On smaller systems (arm based
sysmobts) this might cause a noticable delay of other important
tasks (e.g. passing l2 messages back and forth) It makes more
sense to compute the measurement results continously when
l1sap_info_meas_ind() is executed.
Change-Id: Iecb9a30c0d716bfc88221cd752b1ffdc74269e30
RTP jitter increases continuously because duration is not
updated when speech resumes in DTX mode.
Change-Id: Ib51ed95a449369222c957b3acebd9ce1f66c5435
Add optional MS timing offset (3GPP TS 45.010 § 1.2) to RSL MEASUREMENT
RESULT (3GPP TS 48.058 § 8.4.8). The value is calculated either directly
from corresponding BTS measurement or from 3GPP TS 48.058 § 9.3.17
Access Delay (for known TA) and is invalidated after RSL report is sent
until new measurement indication or RACH is received.
Change-Id: I4dfe5c48834a083e757d5de3236a02e15a238b28
Related: OS#1574
In some cases, when successive mobile originated calls are made, the LAPDm UA
message gets lost because the channel is relased to early. Too overcome the
problem we do not send relase indications immediately. Instead a flag will be
set and the message stored and sent on the next TCH-RTS-IND.
This commit adds the functionality to store the release indication msg, to
rsl.c. It also addes the mechanism to forward the release indication to l1sap.c
See also coresponding change in openbsc.git:
Change-Id I15fc1ef8e9e83f009bde96de9a8e95702cffbce6
This patch is is a slightly improved/reformatted version of:
95d1f15ad1
Change-Id: Ie4f70c75f0137b4bd72d579b3a32575bac2fca38
* make oml_tx_failure_event_rep() static and use osmo_signal_dispatch()
wrapped into oml_fail_rep() to trigger event reports outside of oml.c
instead of directly calling into OML layer
* remove unnecessary formatting from text messages
Related: OS#1615
Change-Id: I738555c547926e97b325ab53763c0076c42309bc
This bug was introduced in the recently merged
commit 1e399f888e
aka change-id I87f40f5f160a4f6750c4f3d06997fc4f24049303
Fixes: coverity-scan CID#160156 and CID#160155
Change-Id: I88ed1b3e59213acdf97f88eda097b8172b952a5e
The counting of the expired rach slots in l1sap.c is not correctly
implemented. This commit fixes the implementation. The expired
rach slots are now conted correctly according to the configured
channel combination. If a CCCH and SDCCH are combined, only the
frames related to rach slots are counted.
Change-Id: I87f40f5f160a4f6750c4f3d06997fc4f24049303
* Unlike in AMR FR, in AMR HR incoming ONSET have to be treated
differently depending on whether we've recently sent SID UPDATE or
EMPTY frame. Split ST_SID_U FSM state into 2 states to accommodate for
that and make sure that additional states specific to AMR HR are not
used for AMR FR.
* Avoid sending E_VOICE and E_SID_U in corresponding states
as those do not initiate FSM state transitions anyway. This decrease
extra load from FSM signalling which otherwise would be triggered on
per-frame basis.
* Introduce separate signal for SID First P1 -> P2 transition to avoid
confusion with E_COMPL and E_SID_U initiated transitions from P1
state.
* Don't init DTX FSM for SDCCH channels.
Change-Id: I229ba39a38a223fada4881fc9aca35d3639371f8
Related: OS#1801
Previously timestamp was always adjusted according to FN difference. In
case of ONSET event this causes unnecessary TS gap with subsequent
speech packet. Fix this by checking Marker bit before performing
adjustment.
Change-Id: I9bf4b45aa990dd4014334dd846f43f793366056c
Related: OS#1801
Having RTP metadata is useful for debugging - save Sequence Number and
Timestamp next to Marker bit from RTP header.
Change-Id: I359b3bcb74fbfc071547fe2f9d837829374fe997
Do not assume that 1 == BS_AG_BLKS_RES but take that information from
SI3. Note: due to current implementation quirks we activate channels
before SI3 obtained, than we deactivate channels upon receiving SI3 and
activate them again. This might not be necessary once we migrate to
proper OML state machines.
This affects lc15 and sysmo hw.
Change-Id: I11377b12680ac3b2f77f80e742b6f0af63fc9c1e
Related: OS#1575
Vadim Yanitskiy
Harald Welte
Pau Espin
Philipp Maier