OpenAirInterface LTE

Notes on a working OAI LTE project using a Samsung Galaxy S4 Mini, mainly pointers to build guides used and summary of obstacles encountered.

Result: Data transmission with an off the shelf cell phone up to 8Mbps (iperf) with very limited range using two antenna for tx and rx (proper duplexer project pending). That is, the phone must be experimentally positioned a less than a half meter from the LimeSDR and moved around until a good constellation display is found for best results.

Guides for installation:
[1]Main OAI page for setup with usrp device
[2]open-cells LimeSDR setup
[3]Recent (8/22/2017) all-in-one box build of OAI eNodeB and EPC components.

Hardware used:
Dell OptiPlex 9010 - quad core i7-3770 CPU @ 3.40GHz with USB3 support and hyperthreading turned off per OAI [4]recommendations
Bash script to turn off hyperthreading (/usr/local/bin/set-hyper-threading) from discussion [5].
Ubuntu 16.04 LTS Xenial - with low latency kernel

root@DellOptiPlex9010:~# uname -a
Linux DellOptiPlex9010 4.13.0-36-lowlatency #40~16.04.1-Ubuntu SMP PREEMPT Sat Feb 17 00:18:34 UTC 2018 x86_64 x86_64 x86_64 GNU/Linux

The latest available lowlatency kernel can be found with a search for the kernel module gpt.ko that will be needed to run the gateway module

# apt-file search gtp.ko  | grep lowlatency
linux-image-4.13.0-36-lowlatency: /lib/modules/4.13.0-36-lowlatency/kernel/drivers/net/gtp.ko
# apt install linux-image-4.13.0-36-lowlatency

Beware if that particular kernel is no longer the primary boot image in grub.conf after an apt upgrate - or manually pick the kernel on startup, or change grub.conf to boot that particular image [6] - example /etc/default/grub:

GRUB_DEFAULT="Advanced options for Ubuntu>Ubuntu, with Linux 4.13.0-36-lowlatency"

So now we have a platform with i7-3770, no hyperthreading and lowlatency ready for OAI LTE.
And off we go:

~/src/oai$ git clone https://gitlab.eurecom.fr/oai/openairinterface5g.git     Near 200MB 
cd  openairinterface5g
~/src/oai/openairinterface5g$ source oaienv
~/src/oai/openairinterface5g$ ./build_oai -I --install-optional-packages          <-- left this running in a screen, installs a bunch of pkgs -- hope it does not break my 2G stuff!

---

later ----

came home to find a question in the screen, about allowing non-root users to run wireshark packet capture - choose the not-recommended 'yes'
then fail on python ssl - had to fix with

python -m easy_install --upgrade pyOpenSSL

from [7] re-run ./build_oai above and completed successfully
Next run

~/src/oai/openairinterface5g$ source oaienv     Since I exited the screen with env set
~/src/oai/openairinterface5g$ ./cmake_targets/build_oai --eNB -w LMSSDR -c -C -x
< ... >
-- Build files have been written to: /home/chuck/src/oai/openairinterface5g/cmake_targets/lte_build_oai/build
Compiling lte-softmodem
Log file for compilation has been written to: /home/chuck/src/oai/openairinterface5g/cmake_targets/log/lte-softmodem.Rel14.txt
lte-softmodem compiled
Log file for compilation has been written to: /home/chuck/src/oai/openairinterface5g/cmake_targets/log/oai_lmssdrdevif.Rel14.txt
oai_lmssdrdevif compiled
liboai_device.so is linked to LMSSDR device library
10. Bypassing the Tests ...

More on build options:
A fellow in the discourse how-to-install-limesdr-on-openinterface-enodeb uses: ./build_oai -I --eNB -x -w LMSSDR Add -x to enable xforms (soft scope), -w hardware EXMIMO, USRP, BLADERF, ETHERNET, LMSSDR, None (Default) --eNB Makes the LTE softmodem -I Installs required packages such as LibXML, asn1.1 compiler, freediameter, ... <-- I did this seperately above

Above command had -c clean Erase all files to make a rebuild from start
-C clean-all Erase all files made by previous compilations, installations

The open-cells limesdr-installation used

./cmake_targets/build_oai -c -w LMSSDR --eNB --UE

for -c clean, -w hardware, --eNB and also --UE Makes the UE specific parts (ue_ip, usim, nvram) from the given configuration file -- default given config file is /home/chuck/src/oai/openairinterface5g/openair3/NAS/TOOLS/ue_eurecom_test_sfr.conf
What you end up with after that build is:

 targets/bin/liboai_device.so ->  targets/bin/liboai_lmssdrdevif.so.Rel14
 targets/bin/liboai_lmssdrdevif.so.Rel14
 targets/bin/lte-softmodem.Rel14

and any custom LimeSDR tweaks like setting external clock reference or printing confirmation of antenna uses go in

targets/ARCH/LMSSDR/USERSPACE/LIB/lms_lib.cpp

plus it's just fun to read in itself with /usr/local/include/lime/LimeSuite.h open in another term. Of course rebuild lte-softmodem after any tweaks or experiments.

Do Download and Patch EPC from [8]
Install 3rd party software for EPC

source oaienv
cd scripts
./build_hss -i
Installing mysql -  root user password:  Pa$$word            <intentionally including example password> 
installs freeDiameter - auth like RADIUS  https://en.wikipedia.org/wiki/Diameter_(protocol)
installs apache, php    -- for phpMyAdmin   pick apache
Configure database for phpmyadmin with dbconfig-common? Yes
phpmyadmin password:   Pa$$word

That was to install all the prereqs for hss - it will be built later, meanwhile prepare for

 $ ./build_mme -i
freeDiameter - no
asn1c rev - no
libgtpnl - yes      hope this does not conflict with osmosgsn, ggsn  GTP
wireshark - done previously (non-root CAN run it)


 $ ./build_spgw -i
libftpnl - no this time

Next, actually compile them - just up-arrow and delete the -i

./build_hss
'/home/chuck/src/oai/openair-cn/build/hss/build/oai_hss' -> '/usr/local/bin/oai_hss'
oai_hss installed
$ ls /usr/local/bin/oai_hss
/usr/local/bin/oai_hss

./build_mme
mme compiled
'/home/chuck/src/oai/openair-cn/build/mme/build/mme' -> '/usr/local/bin/mme'
mme installed
auth_request compiled
'/home/chuck/src/oai/openair-cn/build/mme/build/auth_request' -> '/usr/local/bin/auth_request'
auth_request installed

./build_spgw
spgw compiled
'/home/chuck/src/oai/openair-cn/build/spgw/build/spgw' -> '/usr/local/bin/spgw'
spgw installed

We are going to use this configuration:

HSS is on localhost: 127.0.0.1
eNB is on 127.0.0.10
MME is on 127.0.0.20
SPGW is on 127.0.0.30

I learned something about loopback lo interface here - you already have all 255^3 addresses available ready to use!

# ping 127.90.90.90
PING 127.90.90.90 (127.90.90.90) 56(84) bytes of data.
64 bytes from 127.90.90.90: icmp_seq=1 ttl=64 time=0.025 ms
64 bytes from 127.90.90.90: icmp_seq=2 ttl=64 time=0.032 ms

So backup and edit the config file openairinterface5g/targets/PROJECTS/GENERIC-LTE-EPC/CONF/enb.band7.tm1.25PRB.lmssdr.conf to use this:

mme_ip_address = ( { ipv4 = "127.0.0.20";          -- rest the same

NETWORK_INTERFACES : 
     ENB_INTERFACE_NAME_FOR_S1_MME = "lo";
     ENB_IPV4_ADDRESS_FOR_S1_MME = "127.0.0.10/8";

     ENB_INTERFACE_NAME_FOR_S1U = "lo";
     ENB_IPV4_ADDRESS_FOR_S1U = "127.0.0.10/8";

Here is a good discussion of the nodes and interfaces (S1-MME, S1-U, etc) to help visualize the interconnects [9]
Install This Configuration for EPC in home ~ This uses the package d/l from [10]

sudo mkdir -p /usr/local/etc/oai
sudo cp -rp ~/opencells-mods/config_epc/* /usr/local/etc/oai

cd src/oai/openair-cn; source oaienv; cd scripts
./check_hss_s6a_certificate /usr/local/etc/oai/freeDiameter hss.OpenAir5G.Alliance
HSS S6A: Did not find valid certificate in /usr/local/etc/oai/freeDiameter
HSS S6A: generating new certificate in /usr/local/etc/oai/freeDiameter...
Creating HSS certificate for user 'hss.OpenAir5G.Alliance'
...
Certificate is to be certified until Feb 24 18:51:15 2019 GMT (365 days)

Write out database with 1 new entries
Data Base Updated
/home/chuck/src/oai/openair-cn/scripts
HSS S6A: Found valid certificate in /usr/local/etc/oai/freeDiameter

So that cert is good for ONE YEAR Warning if still using it then

# less /usr/local/etc/oai/freeDiameter/hss.cert.pem
         Validity
            Not Before: Feb 24 18:51:15 2018 GMT
            '''Not After : Feb 24 18:51:15 2019''' GMT

./check_mme_s6a_certificate /usr/local/etc/oai/freeDiameter mme.OpenAir5G.Alliance
File /usr/local/etc/oai/freeDiameter/mme.cert.pem not found
MME S6A: Did not find valid certificate in /usr/local/etc/oai/freeDiameter
MME S6A: generating new certificate in /usr/local/etc/oai/freeDiameter...
Creating MME certificate for user 'mme.OpenAir5G.Alliance'
...
Certificate is to be certified until Feb 24 18:53:02 2019 GMT (365 days)

Write out database with 1 new entries
Data Base Updated
/home/chuck/src/oai/openair-cn/scripts
MME S6A: Found valid certificate in /usr/local/etc/oai/freeDiameter

Then in /usr/local/etc/oai/spgw.conf change SGI to YOUR Internet facing interface:

PGW_INTERFACE_NAME_FOR_SGI = "enp0s31f6";  
to 
PGW_INTERFACE_NAME_FOR_SGI = "eno1";

This is already set:
PGW_MASQUERADE_SGI = "yes";

Warning: the config file /usr/local/etc/oai/hss.conf contains:

## HSS options
#OPERATOR_key = "1006020f0a478bf6b699f15c062e42b3"; # OP key matching your database
#OPERATOR_key = "11111111111111111111111111111111"; # OP key matching your database

The OPERATOR_key or 'OP' is an important part of the UE authentication, along with the key Ki and OPc. In fact, when the hss starts up is recalculates all the subscribers OPc from their key and this OPERATOR_key so they have to match what you use to write the OPc to the SIM card later. I picked a random OP (openssl rand -hex 16) and calculated the OPc from that with auchss.py [11].

Next discussion about MCC,MNC and headache getting an ebay special Galaxy S4Mini GT-I9195 [12] working on LTE with Sysmocom SIMS
While GSM is fairly easy to authenticate, you can use about any sim, LTE requires a SIM for which you know the Key Ki.
That means going ahead and buying the Sysmocom sysmoUSIM-SJS1 10 pack with the ADM keys.
They are very good and email the info for each card before you get them - the IMSI ICCID ACC PIN1 PUK1 PIN2 PUK2 Ki OPC ADM1 KIC1 KID1 and KIK1.
Another option might be piswords cards from Alibaba or ebay.
Will also need an affordable Scm Microsystems USB Smart Card Reader SCR3310V2.

The target UE, the SGS4Mini, that worked fine with osmocom gsm/gprs and had an Option for LTE bands

Band 3 - dl 1805 to 1880
Band 7 - dl 2620 to 2690
Band 8 - dl 925 to 960
Band 20 - dl 791 to 821

expecting to use band 7 which has a known working lte-softmodem configuration file for the LimeSDR (enb.band7.tm1.25PRB.lmssdr.conf). However on plugging in the SIM, the LTE
option disappeared, leaving only WCDMA UMTS and good ol' gsm. Much head scratching, searching and finding the Service Mode for the phone [13], finally tried changing the SIM MCC/MNC - mobile country code, mobile network code [14]. The Sysmocom SIMS come with 901/70 which looks like Satellite Network, unused network code. Tried the default OAI MCC/MNC 208/93 France, Unused network code, also UK, no luck. Tried a US code and got the LTE option but would not register. Finally discovered the IMEI of the phone contains a code for the country or origin [15], which in this case turns out to be Finland - tried a carrier there 244/91 and during tests with the easy to setup OpenLTE found the network "FI SONERA" finally. So the target UE likes 244/91 OK. Another helpful tool is the phone service mode mentioned above, *#0011# and then hit (menu) Back and then (menu) Key Input and enter Q0000 and wait. Then Select UMTS -> Debug Screen -> Phone Control -> Network Control -> Band Selection -> LTE Band and can pick 3,7,8,20 or * to narrow and speed up the network search.

For SIM programming this is a very useful page [16] that uses the strategy of copying a user that already exists in the example mysql oai_db 'users' table. They will also need to be put in the 'pdn' packet-data-network as detailed in step 2 of section 2.3 of [17]

The next obstacle was authentication, the OP issue mentioned above. A fail to connect issue was due to a confusion about OP and OPc programmed in the card.
/usr/local/etc/oai/hss.conf was setup with this as a random guess

## HSS options
#OPERATOR_key = "1006020f0a478bf6b699f15c062e42b3"; # OP key matching your database
OPERATOR_key = "11111111111111111111111111111111"; # OP key matching your database

That should give an OPc of:

OP: 11111111111111111111111111111111
Ki:  39860FDF8D531CDF383582C4AEEFA607

~/src/sysmo-usim-tool$ ./auchss.py -o 11111111111111111111111111111111 -k 39860FDF8D531CDF383582C4AEEFA607
OP: 11111111111111111111111111111111
KI: 39860FDF8D531CDF383582C4AEEFA607
OPc: 327ed2b3a3437b08d5ad35875d222f29      <-- this 

Matches db:

mysql> select hex(OPc) from users where imsi='244910000022771';
+----------------------------------+
| hex(OPc)                         |
+----------------------------------+
| 327ED2B3A3437B08D5AD35875D222F29 |
+----------------------------------+

but not the sim. I don't know how to get an OP to put in hss.conf from the existing OPc and Ki so reprogram the card with that OPc - use:

python pySim-prog.py --pcsc-device=0 --type="sysmoUSIM-SJS1" --mcc=244 --mnc=91 --imsi=244910000022771 --opc=327ED2B3A3437B08D5AD35875D222F29 --ki=39860FDF8D531CDF383582C4AEEFA607 --iccid=8988211000000227713 --pin-adm=57942614 --acc=0002

--OR-- create a new OP

~$ openssl rand -hex 16
2f22315911b5ff00591be8b3898b4c09

~/src/sysmo-usim-tool$ ./auchss.py -o 2f22315911b5ff00591be8b3898b4c09 -k 39860FDF8D531CDF383582C4AEEFA607
OP: 2f22315911b5ff00591be8b3898b4c09
KI: 39860FDF8D531CDF383582C4AEEFA607
OPc: 95dbcca435e94223cbf8990860bdae22

and program the card with:
python pySim-prog.py --pcsc-device=0 --type="sysmoUSIM-SJS1" --mcc=244 --mnc=91 --imsi=244910000022771 --opc=95dbcca435e94223cbf8990860bdae22 --ki=39860FDF8D531CDF383582C4AEEFA607 --iccid=8988211000000227713 --pin-adm=57942614 --acc=0002

To test that - update /usr/local/etc/oai/hss.conf with:

OPERATOR_key = "2f22315911b5ff00591be8b3898b4c09"; # random from openssl rand -hex 16

and startup

./run_hss
...
IMSI: 244910000022771Key: 39.86.0f.df.8d.53.1c.df.38.35.82.c4.ae.ef.a6.07.
OPc: 32.7e.d2.b3.a3.43.7b.08.d5.ad.35.87.5d.22.2f.29.
RijndaelKeySchedule: K 39860FDF8D531CDF383582C4AEEFA607
Compute opc:
K: 39860FDF8D531CDF383582C4AEEFA607
In: 2F22315911B5FF00591BE8B3898B4C09
Rinj: BAF9FDFD245CBD2392E371BBE936E22B
Out: 95DBCCA435E94223CBF8990860BDAE22
Query: UPDATE `users` SET `OPc`=UNHEX('95dbcca435e94223cbf8990860bdae22') WHERE `users`.`imsi`='244910000022771'
IMSI 244910000022771 Updated OPc 327ed2b3a3437b08d5ad35875d222f29 -> 95dbcca435e94223cbf8990860bdae22
...

correct! Just needs to program the SIM and it should connect OK this time :\

The options in pysim are:

-k KI, --ki=KI Ki (default is to randomize)
-o OPC, --opc=OPC OPC (default is to randomize)
--op=OP Set OP to derive OPC from OP and KI

~/src/pysim$ python pySim-prog.py --pcsc-device=0 --type="sysmoUSIM-SJS1" --mcc=244 --mnc=91 --imsi=244910000022771 --opc=95dbcca435e94223cbf8990860bdae22 --ki=39860FDF8D531CDF383582C4AEEFA607 --iccid=8988211000000227713 --pin-adm=57942614 --acc=0002 
Insert card now (or CTRL-C to cancel)
Generated card parameters :
> Name : Magic
> SMSP : e1ffffffffffffffffffffffff0581005155f5ffffffffffff000000
> ICCID : 8988211000000227713
> MCC/MNC : 244/91
> IMSI : 244910000022771
> Ki : 39860FDF8D531CDF383582C4AEEFA607
> OPC : 95dbcca435e94223cbf8990860bdae22
> ACC : 0002

Programming ...
Done !

That made progress indeed - at first could still not find the network - then fired up OpenLTE and let it fail on band20 a while - switched back to OAI band7 and soon as I changed the settings in service mode for LTE band selection it attached and got logs immediately!

[MAC][I][UL_failure_indication] [eNB 0][UE 0/80e6] Frame 620 subframeP 6 Signaling UL Failure for UE 0 on CC_id 0 (timer 0)
[PHY][E]ERROR: Format 1A: rb_alloc (1ff) > RIV_max (144)
[PHY][I][eNB 0][RAPROC] Frame 627 Terminating ra_proc for harq 4, UE 1
[MAC][I][rx_sdu] [eNB 627] Frame 6, Subframe 0 CC_id 0 MAC CE_LCID 27 (ce 0/3): CRNTI 80e6 (UE_id 0) in Msg3
[MAC][I][rx_sdu] [eNB 0] CC_id 0 MAC CE_LCID 29 : Received short BSR LCGID = 1 bsr = 33
[PHY][W][eNB 0, CC 0] frame 733, subframe 4, UE 0: ULSCH consecutive error count reached 20, triggering UL Failure
[MAC][I][UL_failure_indication] [eNB 0][UE 0/80e6] Frame 733 subframeP 4 Signaling UL Failure for UE 0 on CC_id 0 (timer 0)
[PHY][E]ERROR: Format 1A: rb_alloc (1ff) > RIV_max (144)
[PHY][I][eNB 0][RAPROC] Frame 739 Terminating ra_proc for harq 4, UE 1
[MAC][I][rx_sdu] [eNB 739] Frame 6, Subframe 0 CC_id 0 MAC CE_LCID 27 (ce 0/3): CRNTI 80e6 (UE_id 0) in Msg3
[MAC][I][rx_sdu] [eNB 0] CC_id 0 MAC CE_LCID 30: Received long BSR LCGID0 = 0 LCGID1 = 0 LCGID2 = 0 LCGID3 = 0
[PHY][I]UE 0 : rnti 80e6
[MAC][I][eNB_dlsch_ulsch_scheduler] UE rnti 80e6 : in synch, PHR 40 dB CQI 10
[RRC][I]UE rnti 80e6 failure timer 0/20000
[PHY][W][eNB 0, CC 0] frame 733, subframe 8, UE 0: ULSCH consecutive error count reached 20, triggering UL Failure

tcpdump -i gtp0 gets this

17:41:56.461835 IP 172.16.0.10.47668 > ord30s25-in-f206.1e100.net.https: Flags [P.], seq 1:217, ack 1, win 229, options [nop,nop,TS val 43729 ecr 7671639], length 216
17:41:56.491983 IP ord30s25-in-f206.1e100.net.https > 172.16.0.10.47668: Flags [.], ack 217, win 170, options [nop,nop,TS val 7671748 ecr 43729], length 0

1e100 is a googol, the UE phoning home. it is pingable from the OAI host

# ping 172.16.0.10
PING 172.16.0.10 (172.16.0.10) 56(84) bytes of data.
64 bytes from 172.16.0.10: icmp_seq=1 ttl=64 time=73.9 ms
64 bytes from 172.16.0.10: icmp_seq=4 ttl=64 time=61.2 ms

The last puzzle piece with the lousy bursty throughput and those log errors:

[PHY][W][eNB 0, CC 0] frame 733, subframe 4, UE 0: ULSCH consecutive error count reached 20, triggering UL Failure
[MAC][I][UL_failure_indication] [eNB 0][UE 0/80e6] Frame 733 subframeP 4 Signaling UL Failure for UE 0 on CC_id 0 (timer 0)
[PHY][E]ERROR: Format 1A: rb_alloc (1ff) > RIV_max (144)

was partly solved by using the lte-softmodem -d switch Enable soft scope and L1 and L2 stats (Xforms), since it was built with the -x --xforms option, and partly by randomly moving the phone around and noticing there was a sweet spot where Firefox would download and install very fast.

Screenshot of a good connection