Fronthaul xran library update to F release (2. and final step towards M-plane integration)
- xran library support -> both E and F releases
- RU delay profile removed -> not used in neither of the releases
- DU delay profile updated
- one-way delay measurements support is improved/fixed in the F release.
However, not implemented here as it depends on the RU capabilities
FAPI P7 Refactoring and unitary test creation
This MR is a follow-up to the work done in !2714 this time separating
the P7 FAPI messages into its own library (nr_fapi_p7).
In the same way, utility functions to copy, free and compare have been
added for each message, as well as an addition function to get the
allocated size for a message, which is used for the messages that can be
too big to fit in a buffer that's allocated in the stack.
Unitary tests have been added for each of the messages, testing
pack/unpack and the utility functions added.
Some users of the OAI CI complain that since we build with -Werror, it
is difficult to see all warnings/errors and fix them, necessitating
multiple attempts at times to see all warnings. To allow them to see
more warnings, use ninja option -k10 to keep going after errors, and
show more errors at once.
There are command line options that one might pass to ninja/make, but
for which cmake does not provide an interface. For instance, make/ninja
support -k/-kN to "keep going" after errors. Cmake does not provide an
option, but it is still possible to pass this. Make a build_oai option
to allow this, used in the next commit.
Imscope updates
- Using imgui docking branch - allows window docking.
- Moved scopes to different windows to allow docking.
- Disabled scatterplot for time domain samples - issues with amount of
vertices caused segfaults.
- Disabled IQ heatmaps for RX IQ samples on gNB and UE - this was
incorrect and currently I don't know how to fix it. This will be
reenabled once it is.
- added a simple menu and disabled demo windows by default.
Update NR UE threading model
This updates the NR UE threading model by preventing calling threadpool
from threadpool and fixes deadlock issues related to processSlotTX. By
making processSlotTX run to completion instead of starting and waiting
for other threads to finish it prevents locking thread pool cores,
therefore prevening deadlocks.
There is a slight difference here compared to the previous version where
I do not use the UL actor as I believe it is not necessary,
processSlotTX satisfies the run-to-completion requirement now. If in the
future any UL procedure would be parallelized using the thread pool it
should either be done in a run-to-completion model or the processSlotTX
function should be taken off the thread pool.
The documentation has been updated.
4G/5G FAPI code is located in the same file. Since 5G code calls 5G
specific functions, these functions need to be defined in 4G.
The solution would be to separate the FAPI changes into 4G and 5G.
Remove dummy_subframe variable, which is initialized once in
start_request() and then copied to memory on the heap. It is not needed.
In nr_start_request(), it should not be necessary, but we keep it for
the moment to not make more changes than necessary.
This avoids a segfault for me, but I am not exactly sure why.
The previous code fills random number in the TDD config, which leads to
unrealistic TDD patterns. Filling wtih a specific, realistic TDD config
allowed to fix the bug in the next commit.
Testing showed that some messages could be at least slightly bigger than
the previous maximum size of 8192. Given that the maximum transport
block could be larger, increase this value to 32768.
unpack_dl_tti_ssb_pdu_rel15_value() reads two additional bytes,
seemingly for RX power information. However, table 3-79 does not
contain such info, so I remove to match the corresponding pack function
pack_dl_tti_ssb_pdu_rel15_value().
TODO: tx power info relevant?
Since tag 2024.w45/commit f54ca5714e,
standalone mode/--sa is the default and does not exist. Correspondingly,
it has been removed from the documentation.
For users reading recent documentation, but running an older version of
OAI, there might be a mismatch in that --sa is necessary to run OAI, but
not given in the documentation. Provide explanatory text to make users
aware of this.
Run the last job of gNB pusch channel estimation inline
In case of 1 antenna this prevents the usage of threadpool which
eliminates the thread pool overhead. In case of >1 antenna the
threadpool contention should be reduced providing smaller benefits.
NR UE RRC T304 in phy-test mode
RRC timer T304 is stopped upon completion of the associated RA handover
procedure. There is no RA procedure in phy-test mode so T304 should not
be started to avoid unwanted expiration.
NR MSG3 scheduling improvements
- fix wrong condition on MSG3 allocation assertion message
- remove assertion altogether and handle properly the case where MSG3
can't be allocated
T2: minor fix for 4x4 processing
- correct max number of LDPC segments for T2 processing
- use pointer arithmetics in retrieve_ldpc_enc_op() in T2 lib to
optimize loop processing
If we are pulling FlexRIC, overwrite the tag to use with what has been
provided on the command line. This is because we hardcode the tag schema
"branchname-commitId[0:8]" everywhere else, but this does not apply for
FlexRIC, and we have to overwrite this here.
- pass parameter directly instead of indirectly via class
- harmonize to svr_id to read through common code in the next commits
- pass images to express which images will be deleted
- refactor function to use "with" for cls_cmd, and return success (or
not)
- reuse cls_containerize.Containerize.PullImage()
- pass parameter directly instead of indirectly via class
- always read node name and use this instead of "sometimes" looking it
up from a class member. With "sometimes" I mean that if no node name
is given, it will take it from self.eNBIPAddress, which is confusing
at best if we can give the node name directly
- consistently put the node name in the xml file
- pass parameter directly instead of indirectly via class
- call it simply "images", to harmonize with clean-function later
- harmonize to svr_id as in the previous commit
- rewrite function to use with-statement with cls_cmd
- optionally pass a tag to use (which will be used in a test, and can be
used to pull flexric)
- Introduce Pull_Image() which does the actual pulling, can pull from
any registry, and will be reused in the case of pulling from cluster.
The asserts might be used to intentionally stop the running process.
Further, it might be an annoyance if it happen unintentionally. Just
print an error message instead, and return to FlexRIC that nothing is to
be sent.
The CI pipeline RAN-NSA-B200-Module-LTEBOX-Container uses the same core
network (ltebox) as the pipelines RAN-LTE-{TDD,FDD}-LTEBOX-Container,
triggered through the same XML files. However, the core networks are on
different hosts, configured through (Jenkins) pipeline configuration. A
future MR will move the entire core network configuration into the XML
files [1]. To facilitate this, create separate XML files for the NSA
pipeline.
[1] The advantage is that it will be easier to reason with only the XML
files what is going to be run where and how.
- pass parameter directly instead of indirectly via class
- harmonize to svr_id to harmonize parameter with parameters in the next
commit (and towards a future harmonization of directly taking node
names instead of "IDs")
The queue for finished jobs of PUSCH channel estimation is only used inside
nr_pusch_channel_estimation(). Since initialization is fast, it can be
declared on the stack in the function where it is used.
There seem to be randomly occurring deadlocks when running the UE in the
CI. MR !3091 should fix this. While waiting for this, apply a workaround
by increasing the number of worker pools in the UE to 12.
Since tag 2024.w45/commit f54ca5714e,
standalone mode/--sa is the default and does not exist. Correspondingly,
it has been removed from the documentation.
For users reading recent documentation, but running an older version of
OAI, there might be a mismatch in that --sa is necessary to run OAI, but
not given in the documentation. Provide explanatory text to make users
aware of this.
In some runs of a multi-UE deployment, we see segfaults because
cell_group->spCellConfig is NULL. We should not segfault in that case,
so check for spCellConfig.
When the UE list is full, the iterator will run through all
MAX_MOBILES_PER_GNB UEs. It expects that eventually the pointer is NULL,
which is not the case if we have MAX_MOBILES_PER_GNB (or it simply
depends on the following memory). So give one array entry more, to
ensure that the last one is NULL.
Upon press of ^C, the UE triggers a NAS deregistration. This does not
work in do-ra or phy-test modes, because there is no NAS. Therefore, in
non-SA modes, just stop the UE directly.
If the RLC cannot retransmit a package, 3GPP specs say that we should
declare radio link failure (RLF). Up to this commit, the code did
nothing though. In this commit, add a per-UE RLF callback; the RLC will
call into this function to signal that RLF has been detected through
maximum retransmissions on a particular RLC bearer.
CI: Cleanup of cls_physim.py
- Remove unused code and file
- Rework functions for CUDA and T2 log analysis
- Check if test was successful for nr_ulsim and nr_dlsim in
RAN-gNB-N300-Timing-Phytest-LDPC pipeline
In the last commits, SSHconnection class has been removed from
cls_containerize.py. We therefore don't need the username and password,
as cls_cmd uses certificates only. Remove the usage of the username and
the password, and harmonize the use of GetCredentials().
Recent refactoring replaced the CreateWorkspace function call with a
dedicated XML step (for clarity, and because it avoids needless
recreation of workspaces). In the proxy function, the same
CreateWorkspace xml step has been added, cloning the OAI RAN repo, when
we actually need the repository for the proxy. Hence, fix this by
calling the CreateWorkspace() function manually, but only when
necessary, to build the proxy. (It would only be a problem if the
pipeline would rebuild the proxy, which did not happen because not much
work is happening on the proxy repo).
Further, refactoring and simplication of the proxy build function were
done:
- never tag proxy with develop, that makes no sense
- overwrite CreateWorkspace() info after having created and used the
workspace (the "if" to build the proxy)
- simplify copying of logs
- add simple unit test
- remove check for ubuntu; the dependcy is "docker". Simplify the code
to not obscure with a class variable.
The first thing Build_Proxy does is to remove the repository. Further,
Build_Proxy does not use the OAI RAN repo, but the proxy repo. The
Create_Workspace() is therefore completely useless.
Remove CheckAndAddRoute(), as it is complicated to read, uses a
multitude of hostname/IP/password combinations. Replace with scripts on
the machines, which, as of this commit, are:
- porcepix: ~/after-reboot.sh
- ofqot: ~/after-reboot.sh
- nepes: ~/after-reboot.sh
- obelix: ~/how-to-resetup-obelix-out-of-reboot.sh
It is clearer in CI logs to use absolute paths, and that is also a
"preference" of the new SSH class (cls_cmd.py) in the sense that some
functions don't handle "cd" gracefully (because of underlying Python
library paramiko).
For zip, though, we should change to have nice relative paths in the
archives, so use cd <path> && zip ... to achieve that, instead of
changing to a directory. This way, the path on which we operate is still
visible in the logs in each line.
The clang build fails with
/oai-ran/openair1/PHY/MODULATION/gen_75KHz.cpp:16:23: error: variable length arrays in C++ are a Clang extension [-Werror,-Wvla-cxx-extension]
16 | complex<double> t[len];
| ^~~
/oai-ran/openair1/PHY/MODULATION/gen_75KHz.cpp:16:23: note: function parameter 'len' with unknown value cannot be used in a constant expression
So ignore this warning (and make the build pass), since OAI uses VLAs in
various places.
The previous version would put the user-defined options first. However,
the default options include e.g., -Wall. If a user tries to override
(see for instance the next commit) -Wall or other options might override
this and not honor the user choice.
To always have the user have the last work, put user-defined compile
options last.
THere are four possibilities, for the three options (phy_test, do_ra,
nsa):
- None is set -> SA
- Only one is set of each -> corresponding mode
Introduce a function softmodem_verify_mode() to check for these.
fix blocking bugs in command nr_dlsim -n100 -e27 -s35 -b 273 -R273 -x4 -y4 -z4
errors in main OAI code:
- one stack overflow
- one variable array of size 0
- one direct usage of unaligned SIMD
and small fixes inside the tests tool nr_dlsim to handle this large configuration (4layers, 100MHz)
use a better complex vector multiplication function
Since commit 041b2b2d7a, the CI main loop
iterates over all steps, even if there has been a problem during an
earlier XML step. In all cases, UndeployObject() is marked to always
run, as we always need to make sure that all deployed services are
terminated. UndeployObject() always recovers logs. Hence, it's not
necessary to do it here as well; it leads to duplicated logs.
This is visible in all pipeline with a failed deployment: the logs
appear twice.
create_workspace.sh used to check out develop by default. However, this
is unnecessary because the script later checks out a specific reference
(which can be a commit ID, which is not accepted by git clone, the
reason we don't check out during clone); the repository will simply be
empty in the beginning, which should not be a problem.
Further, it is problematic because we use this script for the build of
the L2sim proxy, and the repository does not have develop; thus, without
this modification, L2sim proxy build will always fail.
telnetsrv uses the Intel/AMD-specific "cpuid" instruction, which is not
available on ARM machines. Remove the call.
The previous code was checking for the number of cores, which is
retained by a call to get_nprocs(). I did not find an equivalent for the
number of threads in C. In C++, there would be
https://en.cppreference.com/w/cpp/thread/thread/hardware_concurrency, in
case it's relevant for anybody.
chore(ci): creating a develop tag on top of develop-commit-id tag for the local private registry
Create a 2nd tag of develop for all images on the local registry. We
are doing this in order to simplify cross-repository testing between
FlexRIC and RAN.
Fix two issues with softmodem yaml files
- Fix a typo that prevents casting max_chan to integer in nrue.uicc.yaml
- Remove unused field ssb_PositionsInBurst_PR from gNB example config
yaml file
Fix Regression 4G
- Fix various bugs and/or regressions in 4G code
- Also, the CI does not test RFsim 4G with --sanitize, enables this to
detect problems.
Add Fedora 41, build_oai cleanup
- Add Fedora 41 to list of supported distributions
- Remove some dependencies in RHEL that are not necessary
- Cleanup some build_oai documentation of non-supported options
Simplify logs in gNB&UE
- reduce the amount of printf() that outputs useless info
- in MAC, PHY, NGAP, RRC, mark some "important" UE events with LOG_A for
better visibility -> can put log_config.global_log_level analysis to
only see important stuff
- in RRC, use some macro to consistently print logs
- track the cellID of a UE in RRC
small fix for 3 digit mnc usage in sib1
When 3 digit MNC is used in PLMN, during preparation of SIB1, first
digit of MNC was not taking into account. UE derives Serving Network
Name (SNN) using PLMN info from sib1. Because of 5GCore and UE has
different SNNs, this leads to XRES*, RES* mismatch during Authentication
Procedure between 5GCore and UE. (SNN is used for derivation of these
keys)
Add functional DFT test
- clean up code on (i)dft calls
- add DFT test for some DFT sizes against trivial implementation: it
tests a simple square signal in only real values
- correctly scale DFT-16
Robert Schmidt