this config option overwrites the EARFCN config that is typically
used to tell the UE the bands to scan for cells. When custom
frequencies are used, this option allows to set them explicitly.
Change-Id: Ice070ea6755e273d916db2dc941068d33bbe206a
this allows to use the lteue to generate DL traffic. the
server listens on the default spgw address.
Change-Id: Iad081e57e2fd9d3cbc3e59e4f729a4721ad778c9
the UEs count param, if specified, is mapped into the config
values of the EPC so it can be used for template rendering.
Since we only have one count parameter, its added to the first
subscriber. 1 is the default value and its used when no count
param is specified.
this allows to run the Amarisoft UE simulator with the Amarisoft EPC
without having to create subscriber entries for each simulated UE.
Change-Id: I29885791f716c204d0b6f18ba134885bae853b6f
Despite open5gs allowing to change the GTP ports in the config file, in
reality changing those values to something else than the standard prot
will fail. Hence, we must use the standard port. As a result, we must
use different IP addresses in each process to avoid ip+port collisions.
Let's use some loopback addresses which shouldn't require extra
configuration on the host, and still only requiring 1 run_node as per
existing EPCs, with the limitation that only 1 open5gs EPC instance can
be run at one in a given run_node.
Related: https://github.com/open5gs/open5gs/issues/897
Change-Id: Id3062c6ad9d6de4c6066547e1e46edad5da285c1
when they are greater than 0 they are written as config paramter.
if they are -1 they are disbaled and automatic gain calibration is
used.
Change-Id: I473ff3ae679784178574d2f76b612dbf77180490
throughput tests require the UE object to have the num_carriers
member. For Android UEs, set it to 1 by default.
Change-Id: I297d24bb578a7db0eb7629be4f2cc0d54624467d
the ltesim_server is part of the UE package but needs to be executed
on the EPC run node. Therefore its handy to have the binary in the epc
trial tgz.
Change-Id: I21ac57d1c1547c1cf4d2394c8593f58ff2c78697
we've provided only all_log_level so far but sometimes it's needed
to select the level per layer. This patch adds the ability
to do so for the NAS layer in the UE.
Change-Id: Iab2bce65e8af81f6d344849c97952e6441cb2846
srsENB now supports PDCP discard so we set it to a sane value.
Also enable status reporting for UL.
Change-Id: Ib6ad0ff57b95a1f220f87c1ede0f6eacc5fb21ff
with many parallel jobs running on one machine we should
relax the timeout a bit to avoid false positives.
Change-Id: Ib32a07559bd5a4b0165f6d3431c5ac57cc050f44
we previously mixed component specific and component agnostic APIs
(stdout vs. log file for example) for setting and retrieving KPI.
This patch propose to use a single abstract get_kpis() method for
all components that can be enriched with component-specific
stuff as desired.
In the case of srsLTE blocks, the main implementation will
remain in srslte_common() and is shared among srsENB/srsUE/srsEPC.
The KPI analyzer in srslte_common() extract and also manages
all three KPI sources (log, csv and stdout) independently.
In addition to the get_kpis() method that always returns a flat
dictionary, it also exposes get_kpi_tree() that return
a dict of KPI dicts that will be used for the Junit.xml generation.
Change-Id: I4bacc6b8a0cb92a581edfb947100b57022265265
osmo-bts.git ae09c8acb4aa93284cdb44f8bbdc14533dc4fa52 modified code to
avoid calling pcu_tx_info_ind() if pcu socket was not connected. Since
osmo-gsm-tester relied on log message sent from within that funtion to
find out whether BTS was able to handle PCUIF connections, that log line
is not printed anymore and hence the function ready_for_pcu() fails to
ever return true.
As a result, gprs tests for sysmo and oc2g bts types always fail with a
timeout.
Other BTS types are not affected (such as osmo-bts-trx) because in there
we simply check whether the osmo-bts process is running.
Change-Id: Ibde4b8b27b8300b4b9197c0f8d82a6ccbf3b0522
we've detected a possible race condition during the Msg3
transmission that caused the thread that sets the Msg3 grant
to be delayed. The PHY worker that executed TTI+2 finished
by that time already and didn't even see the pending UL grant.
This is issue is more likely to happen on loaded system,
for example when running parallel ZMQ jobs. We therefore decided
to reduce the number of parallel PHY workers to 2 so the issue
is circumvented.
Change-Id: Ibdb42a1705d87b6d343201458c8fe397398802bc
we've already incrased the t3413 value in 9c7fd618d2
but realized it may still be too low in some cases.
we set it to 12s now.
Change-Id: I9c3c59c0ff8f539a44b420f8ba31eea016c211ce
the specific cell scenarios files already contain cell-related
information so having a extra ncells scenario seems too much overhead.
furthermore, in more complex tests its helps to reduce the filename/path
lenght to avoid hitting the 255 character limit
Change-Id: If022d9e8d239805ee1280fb52660e68dda7484a0
This patch enables setting cipher and integrity algorithms
in Amarisoft eNB and srsENB via scenario files. If no
settings are defined following defaults are applied:
- Cipher algorithm: EEA0, EEA2, EEA1
- Integrity algorithm: EIA2, EIA1, EIA0
Example of setting cipher algorithms:
- 4g:srsue-rftype@uhd+srsenb-rftype@uhd+mod-enb-cipher@eea1+mod-enb-cipher@eea0+mod-enb-nprb@6
Change-Id: I595206b7d49016fb6d0aec175c828d9537c53886
implement as noop for Amarisoft eNB, srsENB will send q+Enter to stdin,
which is implemented in class srslte_common()
Change-Id: Ide606e1a6b523997215aa2fa39d4d56ae1f49181
the tests already check for sent PRACHs, this will
also check for received PRACHs on the eNB to make
sure we don't detect any fake PRACHs during the tests.
Change-Id: Ib23056b1ad4b18a7dde6cd0c997f5114ac2acb61
this patch adds the stdout counter to count events happening
on the stdout (known from the UE already) to the common
process class so they can also be used from the eNB (and other objects)
In addition, we add a PRACH counter to be used for tests.
Change-Id: I434f072b8aa6f4dce9f90889c6b40832f6798ff8
To expand the test capacities we would like to introduce
Android UEs as new modems. Currently the following tests
are supported:
- Ping
- iPerf3 DL/UL
- RRC Mobile MT Ping
In the following is a small description.
Prerequisites:
- Android UE
- Rooted (Ping, iPerf, RRC Idle MT Ping)
- Qualcomm baseband with working diag_mdlog (RRC Idle MT Ping)
- iPerf3
- Dropbear
- OGT Slave Unit
- Android SDK Platform-Tools
(https://developer.android.com/studio/releases/platform-tools#downloads)
- Pycrate (https://github.com/P1sec/pycrate)
- SCAT
clone https://github.com/bedrankara/scat/ & install dependencies
checkout branch ogt
symlink scat (ln -s ~/scat/scat.py /usr/local/bin/scat)
Infrastructure explaination:
The Android UEs are connected to the OGT Units via USB. We
activate tethering and set up a SSH server (with Dropbear).
We chose tethering over WiFi to have a more stable route
for the ssh connection. We forward incoming connections to
the OGT unit hosting the Android UE(s) on specific ports
to the UEs via iptables. This enables OGT to issue commands
directly to the UEs. In case of local execution we use ADB
to issue commands to the AndroidUE. The set up was tested
with 5 Android UEs connected in parallel but it should be
scalable to the number of available IPs in the respective
subnet. Furthermore, we need to cross compile Dropbear
and iPerf3 to use them on the UEs. These tools have to be
added to the $PATH variable of the UEs.
Examplary set up:
In this example we have two separate OGT units (master
and slave) and two Android UEs that are connected to the
slave unit. An illustration may be found here: https://ibb.co/6BXSP2C
On UE 1:
ip address add 192.168.42.130/24 dev rndis0
ip route add 192.168.42.0/24 dev rndis0 table local_network
dropbearmulti dropbear -F -E -p 130 -R -T /data/local/tmp/authorized_keys -U 0 -G 0 -N root -A
On UE 2:
ip address add 192.168.42.131/24 dev rndis0
ip route add 192.168.42.0/24 dev rndis0 table local_network
dropbearmulti dropbear -F -E -p 131 -R -T /data/local/tmp/authorized_keys -U 0 -G 0 -N root -A
On OGT slave unit:
sudo ip link add name ogt type bridge
sudo ip l set eth0 master ogt
sudo ip l set enp0s20f0u1 master ogt
sudo ip l set enp0s20f0u2 master ogt
sudo ip a a 192.168.42.1/24 dev ogt
sudo ip link set ogt up
Now we have to manually connect to every UE from OGT Master
to set up SSH keys and verify that the setup works.
Therefore, use:
ssh -p [UE-PORT] root@[OGT SLAVE UNIT's IP]
Finally, to finish the setup procedure create the
remote_run_dir for Android UEs on the slave unit like
following:
mkdir /osmo-gsm-tester-androidue
chown jenkins /osmo-gsm-tester-androidue
Example for modem in resource.conf:
- label: mi5g
type: androidue
imsi: '901700000034757'
ki: '85E9E9A947B9ACBB966ED7113C7E1B8A'
opc: '3E1C73A29B9C293DC5A763E42C061F15'
apn:
apn: 'srsapn'
mcc: '901'
mnc: '70'
select: 'True'
auth_algo: 'milenage'
features: ['4g', 'dl_qam256', 'qc_diag']
run_node:
run_type: ssh
run_addr: 100.113.1.170
ssh_user: jenkins
ssh_addr: 100.113.1.170
ue_ssh_port: 130
adb_serial_id: '8d3c79a7'
scat_parser:
run_type: local
run_addr: 127.0.0.1
adb_serial_id: '8d3c79a7'
Example for default-suites.conf:
- 4g:ms-label@mi5g+srsenb-rftype@uhd+mod-enb-nprb@25+mod-enb-txmode@1
Change-Id: I79a5d803e869a868d4dac5e0d4c2feb38038dc5c
this executed the PRACH search in the main thread so
PHY workers and PRACH processing don't diverge in loaded
scenarios
Change-Id: I97eddb5f9ca682251ded1f73093c2d0a49462bd6
we've seen that the default value of 2s might be a bit tight.
the value is network dependent so no real default value exists.
however, 6s seems to be a good compromise.
Change-Id: Icc75ee048c2cc605ab26c4c87a039519998f7e57
it turned out that we have to reduce the MCS when using QAM256,
especially for 6 PRB as subframe 0 and 5 contains PBCH and PSS
signals, so the available REs are reduced.
The eNB scheduler now has this limitation in mind and lowers the MCS.
Change-Id: I0e38fe28002fd68c768cc8dcffcf74f4f190df02
this was causing failed tests because to achieve QAM64 rates both
eNB and UE need to support it and have it activated.
Change-Id: I599df92d69eeb56a5d44327de08f004222cff073
since ZMQ runs are not using wall clock anyway, measuring
TTI execution isn't useful, disable it therefore to avoid
misleading warnings.
Change-Id: I5c2cb0abcfce0ee67806f6611356f4d5d180541d
make sure to have the right eNB that matches the
cells neighbor ECIDs, otherwise the eNB may trigger
a S1 HO if the eNB ID is set elsewhere
Change-Id: Id494ac1984846e78a6456bb6a17e72d63c9dacf5
the order of checks needs to go from high to low, i.e. the higher
release feature (e.g. qam256) needs to be checked and set first.
in theory it should also be possible to have a CA-capable UE
that does not support QAM256, but for srsUE we announce both anyway.
Change-Id: I2fa49f0cb5d80db412a811ceeb380359c8ad67a7
AlaiaL
Pau Espin