Hi, osmocom team members.
In order to have an internet connection in UE, we have run OsmoGGSN project
to set up GGSN. Also, SGSN is developed in our system.
At this step, UE connects to our network core and we can query to an
arbitrary site in the UE browser and we see the answer of the query without
any problem and everything is OK. But after about 30 or 40 seconds, UE
sends "Deactivate PDP Context Request" with SM Cause "SM Cause: Protocol
error, unspecified (111)" and the connection is reset and a new "Activate
PDP Context Request" is required. What is the reason and how can we solve
it?
--
*When there is much light, The shadow is deep...*
Hi all,
in recent months - after many years of no patches at all - the kernel
GTP driver has received significant interest in terms of contributions.
Given the presumed few users of the GTP tunnel driver, as well as the
tendency to use super ancient versions of software in the telecom world,
I think that the usual "let's have the users test -rcX kernels" approach
is unlikely going to work.
Within Osmocom (a group of FOSS projects implementing various cellular
standards, protocol stacks and network elements) we are used to rather
comprehensive functional test suites that we execute automatically
at least once per 24 hours. That is for the 99% of our software that is
userspace code.
For the kernel GTP code, it's of course not that simple, and we never
had any related testing so far.
In 2021, 5 years after the GTP kernel driver was merged mainline, we now
finally have set up some test jobs for kernel GTP.
Those jobs execute the same GGSN test suite that we run against the userspace
dataplane (handling GTP-U and GTP-C in a userspace program "osmo-ggsn"). The
only difference is that they configure osmo-ggsn to use the kernel GTP driver
for GTP-U.
Those tests are executed by jenkins, in the following jobs:
* ttcn3-ggsn-test-kernel-latest-torvalds
https://jenkins.osmocom.org/jenkins/view/TTCN3/job/ttcn3-ggsn-test-kernel-l…
runs GGSN test suite against latest released osmo-ggsn version with
kernel-gtp of torvalds/linux.git test results at
https://jenkins.osmocom.org/jenkins/view/TTCN3/job/ttcn3-ggsn-test-kernel-l…
(make sure to click "Expand All")
* ttcn3-ggsn-test-kernel-latest-torvalds
https://jenkins.osmocom.org/jenkins/view/TTCN3/job/ttcn3-ggsn-test-kernel-l…
runs GGSN test suite against latest released osmo-ggsn version with
kernel-gtp of net-next.git
test results at https://jenkins.osmocom.org/jenkins/view/TTCN3/job/ttcn3-ggsn-test-kernel-l… (make sure to click "Expand All")
* ttcn3-ggsn-test-kernel-latest
https://jenkins.osmocom.org/jenkins/view/TTCN3/job/ttcn3-ggsn-test-kernel-l…
runs GGSN test suite against latest released osmo-ggsn version with
kernel-gtp of Debian 10 kernel package
test results at https://jenkins.osmocom.org/jenkins/view/TTCN3/job/ttcn3-ggsn-test-kernel-l… (make sure to click "Expand All")
* ttcn3-ggsn-test-latest
https://jenkins.osmocom.org/jenkins/view/TTCN3/job/ttcn3-ggsn-test-latest/
runs GGSN test suite against latest released osmo-ggsn version with
userspace GTP-U, not using any kernel GTP driver
test results at https://jenkins.osmocom.org/jenkins/view/TTCN3/job/ttcn3-ggsn-test-latest/t…
this serves as a base line that shows the test suite can fully pass,
and any failures in the aove are either bugs or lack of features (like
IPv6)
Contrary to the above jenkins jobs which all run automatically once
every 24 hours, there is also one jenkins job for manual execution:
* ttcn3-ggsn-test-kernel-git
https://jenkins.osmocom.org/jenkins/view/TTCN3/job/ttcn3-ggsn-test-kernel-g…
this is the same test suite and osmo-ggsn as above, but a developer
can manually trigger the job and specify
* URL of the kernel git repo to build
* branch of the kernel git repo to build
* whether to use osmo-ggsn latest tag or master
If any one is working on the kernel GTP driver and wants to get access
to triggering that jenkins job in order to test your patches/branches
before submission, please register an account at
https://jenkins.osmocom.org/ and contact me privately.
We are always happy about any contributions in terms of extending test
coverage. This could be done by e.g. adding jobs using other P-GW/GGSN
software than osmo-ggsn, or by extending the coverage of the test cases
of the test suite.
For anyone interested in more details, please see our redmine issue
https://osmocom.org/issues/3208 tracks the development of the tests.
The test suite source code (in TTCN-3) is at
https://git.osmocom.org/osmo-ttcn3-hacks/tree/ggsn_tests
with containerization + configs at
https://git.osmocom.org/docker-playground/tree/ttcn3-ggsn-test
I'd like to thank Oliver Smith <osmith(a)sysmocom.de> for creating the
above automatic test CI integration. Development of this was funded
by sysmocom (https://sysmocom.de/)
If you have any questions, please feel free to reach out to Oliver
and/or me.
Regards,
Harald
--
- Harald Welte <laforge(a)gnumonks.org> http://laforge.gnumonks.org/
============================================================================
"Privacy in residential applications is a desirable marketing option."
(ETSI EN 300 175-7 Ch. A6)
Dear Osmocom community,
This topic has been past due for way too many years by now:
A re-organization of our major mailing lists.
I would like to propose the following changes. Pleas let me know if you
have any comments or feedback. I'm aware that renaming will mean people
have to update their mail filter rules, but I think we're long past the
point where the names of some of our lists started to confuse users.
== openbsc(a)lists.osmocom.org ==
* openbsc doesn't exist anymore since OsmoNITB, which is also obsolete
* does already cover anything "Osmocom CNI" related
* Proposed new name: osmocom-cni(a)lists.osmocom.org
== osmocom-net-gprs(a)lists.osmocom.org ==
This date back to when GPRS was a highly experimental add-on to our GSM
code base. This list should simply be merged with openbsc@ as osmocom-cni(a)lists.osmocom.org
== simtrace(a)lists.osmocom.org ==
Historically was created to cover only the simtrace project.
We should rename this to osmocom-simcard(a)lists.osmocom.org or something
along those lines.
I would like to suggest it covers
* SIMtrace / SIMtrace2 hardware + firmware
* pySim and related tools for working with SIM/USIM/UICC cards
* any other information / discussion related to SIM/USIM/UICC cards,
like OTA, ARA-M, ...
== osmodevcon(a)lists.osmocom.org ==
This has been a private list for people attending OsmoDevCon
I would like to open up list membership to the general public, and ensure
it also covers the new OsmoDevCall. We could then have discussions regarding
feedback, topics, scheduling, etc. on that list.
Maybe rename it to osmocom-events(a)lists.osmocom.org instead?
To differentiate: osmocom-event-orga(a)lists.osmocom.org should remain a
private list related to organizational / administrative topics of those
involved with organizing future events.
== nextepc(a)lists.osmocom.org ==
Should have been renamed to open5gs(a)lists.osmocom.org quite some time
ago, I simply forgot about it. My apologies.
--
- Harald Welte <laforge(a)osmocom.org> http://laforge.gnumonks.org/
============================================================================
"Privacy in residential applications is a desirable marketing option."
(ETSI EN 300 175-7 Ch. A6)