CI: Rename RAN-SA-2x2-Module-CN5G pipeline
RAN-SA-2x2-Module-CN5G performs tests not only with 2x2 but also with a
4x4 MIMO configuration. The pipeline name has been generalized to
RAN-SA-Multi-Antenna-CN5G to cover multiple antenna configurations.
RRC: Multi-cell handling architecture
This MR implements a foundational multi-cell handling architecture for
RRC by separating DU management from cell management and establishing a
hierarchical cell organization within each DU container. The changes
enable proper multi-cell support where each DU can serve multiple cells,
addressing the previous limitation where DU and cell management were
tightly coupled with the assumption of one cell per DU.
Key Changes:
- Data Structure Separation: Separated for DU containers (DU-specific
info) from cell containers (cell-specific data)
- Global DU and Cell Trees: Cells are now stored in global RB trees in
RRC, similarly to DUs
- Du-specific cell seq_arr: within each DU container (du->cells) is
stored a seq_arr of cells belonging to the DU
- UE Cell Association Framework: Introduced serving_cells array to track
PCell and SCells for each UE with servCellIndex indexing
- Cell Management Library: Extracted cell management functions into
rrc_cell_management.c/h for better testability and maintainability
Other relevant changes:
- F1AP Refactoring: cell list definitions, heap allocation to prevent
stack overflow
- 5G-S-TMSI Helpers: Common helper functions for 3GPP TS 23.003
compliant construction
- Integration Tests: Comprehensive test suite for DU lookup, cell
lookup, and UE cell association
- Cell Duplication Checks: Validation in F1 DU configuration update to
prevent duplicate cells
- Documentation: Updated RRC developer documentation with multi-cell
architecture details
Validated with RFsim in re-establishment and handover scenarios.
Remove VCD signal dumper from NR code
We don't test VCD in 5G NR code (and haven't tested it in 4G for years).
Remove the VCD traces in 5G code to avoid any ambiguity. 4G code remains
in place, as well as the tools to analyze existing traces.
Fix for split 7.2 DU configurations including DU & RU ethernet addresses
Fix parsing of DU configurations that include both RU and DU ethernet
addresses in the fhi_72 section. The extraneous DU addresses are used in
MPlane to configure the RU.
RRC QoS refactor: PDU Session Modify support, SDAP fixes, and code improvements
The goal of this MR is to prepare the RRC/NGAP layer for the multi-QoS
support. The MR:
- enables PDU Session Resource Modify support in RRC/NGAP/E1AP
- fixes SDAP header mapping
- enables SDAP by default
- refactors QoS flow handling in RRC/NGAP
- improves error handling, and code maintainability in RRC/NGAP
Changes:
- NGAP PDU Session Modify:
* add NGAP PDU Session Resource Modify Request/Response handling with
proper error reporting for unrecognized sessions and AMF UE NGAP ID
validation
- E1AP Bearer Context Modification:
* add DRB To Remove List IE to support DRB release during modify
* rename numQosFlow2Setup to numQosFlowsMod in DRB To Modify List
- RRC QoS flow handling:
* enable multi-QoS flow support per DRB
* add utilities to map NGAP QoS parameters to E1AP
* filter QoS flows by DRB ID
- SDAP configuration:
* Fix SDAP header UL/DL mapping between internal representation and
ASN.1 enum (invert mapping)
* change SDAP header fields from long to bool for clarity
- F1 DRB setup refactoring:
* Extract shared helpers for F1 DRB setup list filling
* reduce duplication between E1 response and direct DRB setup paths
- AMBR management:
* Store AMBR in UE context
* validate AMBR is set before bearer setup
- Configuration:
* remove unused drbs configuration parameter
* enable SDAP by default (change default from false to true)
RAN-SA-2x2-Module-CN5G performs tests not only with 2x2 but also with a 4x4 MIMO
configuration. The pipeline name has been generalized to RAN-SA-Multi-Antenna-CN5G
to cover multiple antenna configurations.
Integrate GPU-Accelerated Channel Simulation + CI Test Pipeline
This MR builds on top of MR !3588 (feat: GPU Accelerated Channel
Simulation Pipeline).
The main portion of the changes is directly taken from MR !3588, and
focuses on integrating the GPU-accelerated channel simulation. On top of
that, this MR adds CI-related configurations to introduce a new channel
simulation testing pipeline into the CI --> RAN-Channel-Simulation
This MR introduces a CUDA-based pipeline to offload the
multipath_channel and add_noise channel simulation functions to the GPU.
This enables large-scale, real-time physical layer simulations not
possible on the CPU.
For full documentation, please see the file at openair1/SIMULATION/TOOLS/DOC/gpu_acceleration.md.
System & Compiler Requirements: An NVIDIA GPU with a compatible driver
and the CUDA Toolkit installed.
This feature was developed and tested on the GH200 server using the GCC
12 toolchain. To ensure reproducibility, building with GCC 12 is highly
recommended. Build Instructions
The primary flag to enable this feature is -DCUDA_ENABLE=ON. The
following command shows a sample configuration used during development
on the GH200 server:
./build_oai --phy_simulators --cmake-opt "\
-DCUDA_ENABLE=ON \
-DUSE_UNIFIED_MEMORY=ON \
-DUSE_ATS_MEMORY=OFF \
-DCMAKE_C_COMPILER=/usr/bin/gcc-12 \
-DCMAKE_CXX_COMPILER=/usr/bin/g++-12 \
-DENABLE_TESTS=ON"
Memory Model Configuration
The pipeline supports multiple CUDA memory management modes, which can
be configured at compile time. ATS Memory (-DUSE_ATS_MEMORY=ON) is
enabled by default. To use another memory model, you must first disable
ATS.
For a detailed explanation of these models, please refer to Section 3 of
the gpu_acceleration.md report.
vrtsim: add CIR DB playback with CLI flags
VRTSIM: CIR DB playback via CLI
This change adds an optional in-process CIR DB path to VRTSIM. When
enabled with --vrtsim.cirdb and optional --vrtsim.cirdb-path, VRTSIM
reads precomputed taps from cir_db.bin. Default behavior is unchanged
when the flag is not set.
Build
cmake ../../ -GNinja -DOAI_VRTSIM_CIRDB=ON -DOAI_VRTSIM_TAPS_CLIENT=OFF
cmake --build . --target vrtsim
How to run
gNB
sudo ./nr-softmodem \
-O ./targets/PROJECTS/GENERIC-NR-5GC/CONF/gnb.sa.band78.fr1.106PRB.usrpb210.conf \
--device.name vrtsim \
--vrtsim.role server \
--gNBs.[0].min_rxtxtime 3 \
--vrtsim.cirdb 1 \
--vrtsim.cirdb-path /home/merkeb/oai-cir-emitter/cir_db.bin
UE
sudo ./nr-uesoftmodem \
-C 3619200000 -r 106 --numerology 1 --ssb 516 --band 78 \
--device.name vrtsim
Expected logs
- VRTSIM prints that channel modelling is active and uses DB backed taps
- CIRDB ready: model=..., shape=Tx x Rx, L=..., S=..., fs=..., dt=...,
speed=...
CLI behavior and fallbacks
- TAPS client and chanmod paths are unchanged
- If --vrtsim.cirdb is not set or equals 0, VRTSIM uses the existing
channel model path
Add fallbacks for ARMv8a to make it compile
The current code, on ARM, uses ARMv8.1-A RDM instructions. Add a
fallback for processors that do not have this instruction set.
Add some specific compile options for this platform.
Add debian 12 to the list of architectures.
- Change gNB RRC enable_sdap default to true
- Update RRC and RUNMODEM docs to reflect new default behavior
- Clean up CI gNB configs by removing redundant enable_sdap=1
- Keep explicit SDAP disable config for NFAPI gNB VNF
According to TS 38.413:
* AMBR is conditional to PDU Session Resource to setup in the Initial UE
Context Setup procedure
* The UE Aggregate Maximum Bit Rate IE should be sent to the NG-RAN node
if the AMF has not sent it previously. If it is included in the PDU
SESSION RESOURCE SETUP REQUEST message, the NG-RAN node shall store
it in the UE context
* Range is INTEGER [0..4,000,000,000,000]
To this purpose, changes in the commit:
* Init AMBR to out of range value when creating UE context
* Store AMBR in UE context whenever is received
* Assert if not set at the time of Bearer Setup (should be known at that stage)
Fix parsing of DU configurations that include both RU and DU ethernet
addresses in the fhi_72 section. The extraneous DU addresses are used
in MPlane to configure the RU.
For now, test in the following configuration:
- 16 parallel channels
- channel length of 32 and 128
- with 61440 and 122880 samples
- 12 trials per each test (due to CUDA kernel loading delay in first iterations)
- multiple MIMO configurations: 2TX - 2/4/8/32 RX, 2/4/8/32 TX - 2RX
Add undocumented Aerial ERROR.indication vendor specific error codes
This MR adds the remaining vendor-specific ERROR.indication error codes
that are undocumented in the Aerial online documentation, but are
present in the cuPHY-CP/scfl2adapter/lib/scf_5g_fapi/scf_5g_fapi.h
header file.
This change stems from the appearance of an undocumented error code
during testing (SCF_ERROR_CODE_L1_MISSING_UL_IQ = 0x44) It aims to ease
the comprehension of the errors thrown by Aerial L1.
Extract common code from `fill_drb_to_be_setup_from_e1_resp` and
`rrc_fill_f1_drb_to_setup` into reusable helper functions to reduce
duplication and improve maintainability. This refactoring enables
multi-QoS flow handling per DRB and prepares helpers for upcoming
PDU Session Modify changes.
Impact:
- Enables multi-QoS flow support: `rrc_fill_f1_drb_to_setup` now
collects all QFIs mapped to a DRB instead of only the first one
- Reduces code duplication between E1 response and direct DRB setup
code paths
- Improves maintainability by centralizing F1 DRB setup logic
Major changes:
- Add `fill_f1_drb_info_nr`: fills DRB information IE including QoS
flows from QFI list
- Add `fill_f1_drb_to_be_setup`: creates F1 DRB to setup structure
from RRC DRB, PDU session, and QFI list
- Add `fill_f1_drbs_from_e1`: processes DRBs from E1 response and
extracts QFIs from drb_config->qosFlows
- Refactor `fill_drb_to_be_setup_from_e1_resp` to use
`fill_f1_drbs_from_e1`
- Refactor `rrc_fill_f1_drb_to_setup` to collect all QFIs mapped to
each DRB and use `fill_f1_drb_to_be_setup`
TS 38.413 8.2.3 PDU Session Modify:
> For each PDU session which failed to be modified, the PDU Session Resource Modify
> Unsuccessful Transfer IE shall be included containing the failure cause
> If the NG-RAN node receives a PDU SESSION RESOURCE MODIFY REQUEST message containing
> some PDU Session ID IEs (in the PDU Session Resource Modify Request List IE) that the
> NG-RAN node does not recognize, the NG-RAN node shall report the corresponding invalid
> PDU sessions as failed in the PDU SESSION RESOURCE MODIFY RESPONSE message
> with an appropriate cause value
Changes in the commit:
* add rrc_gNB_ngap_pdusession_mod_failure helper function to handle
abnormal conditions and send NGAP response with (all) failed PDU sessions
* improve error handling: send NGAP response instead of silently returning -1
when UE context lookup fails (unknown RAN UE NGAP ID)
* add AMF UE NGAP ID validation: check stored AMF UE NGAP ID matches request
and send failure response if mismatch (per 3GPP TS 38.413)
* handle unrecognized PDU Session IDs per 3GPP TS 38.413 8.2.3.4: add to
failed list to be sent in the reponse message
* simplify failure path: reuse rrc_gNB_send_NGAP_PDUSESSION_MODIFY_RESP
instead of manually constructing response message
* remove failed PDU sessions from RRC list in rrc_gNB_send_NGAP_PDUSESSION_MODIFY_RESP
* remove redundant code and improve code organization
* add bounds checking
* improve function signature
This commit refactors the rrc_gNB_send_NGAP_PDUSESSION_MODIFY_RESP function
to improve code clarity, fix bugs, and ensure compliance with 3GPP TS 38.413
section 9.2.1.6.
The main improvement is ensuring the PDU Session Resource Modify Response message
is always sent with mandatory IEs, even when there are no PDU sessions to report
in the optional lists.
Key changes:
- Initialize message structure parameters to 0 by leveraging structure
initialization, ensuring clean state
- Replace local counter variables (pdu_sessions_done, pdu_sessions_failed)
with direct access to resp->nb_of_pdusessions and resp->nb_of_pdusessions_failed,
eliminating redundant state tracking and improving code clarity
- Remove conditional message sending: always send the response message
(per 3GPP TS 38.413 9.2.1.6, PDUSessionResourceModifyResponse must always
be sent with mandatory IEs: AMF-UE-NGAP-ID and RAN-UE-NGAP-ID, even when
optional lists are empty)
- Fix duplicate assignment: remove redundant p->pdusession_id assignment
(was assigned twice in the PDU_SESSION_STATUS_DONE case)
- Improve failed PDU session handling: use direct structure assignment
(failed->cause = session->cause) instead of field-by-field copying,
making the code more maintainable
- Remove failed PDU session upon response
- Enhance logging: improve log messages for better debugging and clarity
- Add function documentation comment
Introduce fill_e1_qos_flow_to_setup to translate pdusession_level_qos_parameter_t (NGAP)
into qos_flow_to_setup_t (E1), including ARP and QoS characteristics (dynamic/non-dynamic).
Refactor fill_e1_drb_to_setup to use the new helper for building QoS flows.
* Only include QoS flows whose drb_id matches the target DRB.
* Add DevAssert(drb_ngran.numQosFlow2Setup > 0) to ensure at least one QoS flow is added.
This change centralizes and simplifies QoS mapping logic between NGAP inputs
and E1 DRB configuration. It ensures correctness by associating QoS flows
to the appropriate DRB.
Major changes:
- Extract QoS flow mapping logic into fill_e1_qos_flow_to_setup helper function
- Filter QoS flows by matching drb_id before adding to DRB setup list
- Replace inline mapping code with helper function call
- Add assertion to verify at least one QoS flow is added per DRB
This commit fixes two related issues:
1. SDAP Header Mapping Bug (cucp_cuup_handler.c):
The original code incorrectly mapped SDAP header configuration from
internal E1AP representation to ASN.1 enum values. The internal
representation (`bearer_context_sdap_config_t`) uses true=present/false=absent,
while ASN.1 enum (`E1AP_SDAP_Configuration_t`) uses
0=present/1=absent. The fix correctly inverts the mapping:
- Internal true (1) → ASN.1 0 (present)
- Internal false (0) → ASN.1 1 (absent)
The commit also clarifies internal representation in the stack.
2. Test Overhead Adjustment (nr-cuup-load-test.c):
The change in (1) led to a failure in nr_cuup_functional_test
with assertion: "size % 4 == 0" in recv_ng()
- Changed UL packet overhead from 3 bytes to 2 bytes in sender_thread()
- Before the SDAP fix, the test incorrectly accounted for 3 bytes:
* 2 bytes PDCP header (correct)
* 1 byte SDAP header (incorrectly included)
- After the SDAP fix, SDAP headers are correctly set to absent, so
only 2 bytes (PDCP header) are needed
This ensures the test sends packets with the correct header size,
matching what CU-UP expects after the SDAP configuration fix.
The load test does not enable SDAP headers, therefore the original 3-byte
overhead was possibly a workaround for the SDAP header bug.
With the bug fixed, the test correctly uses 2-byte overhead (PDCP header).
This commit refactors the PDU Session Resource Modify Response message
structures to improve naming, documentation, and compliance with 3GPP TS 38.413 9.2.1.6.
Changes in NGAP:
- Rename qos_flow_tobe_modified_t to qos_flow_addmod_response_item_t to better
reflect its purpose as a QoS Flow Add or Modify Response Item IE (as per specs)
- Add mandatory AMF UE NGAP ID field to ngap_pdusession_modify_resp_t structure
- Optimize field types: change nb_of_pdusessions and nb_of_pdusessions_failed
from uint16_t to uint8_t (NGAP_MAX_PDU_SESSION limit) (ref 3GPP TS 38.413, 9.2.1.6)
- Update PDU Session Modify Response encoder to use amf_ue_ngap_id from the
message structure (pdusession_modify_resp_p->amf_ue_ngap_id) instead of
retrieving it from UE context (ue_context_p->amf_ue_ngap_id)
- Add comprehensive comments with 3GPP TS 38.413 section references throughout
Changes in RRC/NGAP:
- Populate amf_ue_ngap_id field in PDU Session Modify Response message from
UE context
- Update type reference to qos_flow_addmod_response_item_t
A DRB To Modify Item contains a Flow Mapping Information,
which overrides previous mapping information. For consistency,
rename the count member to reflect a QoS Flow modification,
since it is not limited to setup only.
- Rename DRB flow mapping counters to `numQosFlowsMod`
for a DRB To Modify List
- Adjust encode/decode/equality + E1AP tests
Add DRB To Remove List IE to PDU Session To Modify list, which is sent
in the Bearer Context Modification request. This allows the CUCP to
inform the CUUP about DRBs to be released in a specific PDU session
during a MODIFY procedure.
Changes:
* Add struct definition
* Add enc/dec/eq/cp func
* Add test
- Delete drbs field from gNB_RrcConfigurationReq and its use in RCconfig_NRRRC()
- Remove GNB_CONFIG_STRING_DRBS, help text, index, and GNBPARAMS_CHECK entry from gnb_paramdef.h
- Drop DU-side verification and DRB count logging in gnb_config.c
This commit introduces a new, configurable benchmark,
`test_channel_scalability`, designed to measure the performance and
scalability of the CUDA channel simulation pipeline under various
conditions.
This benchmark provides several key features:
- It supports multiple GPU execution models, including a high-throughput
"batch" mode, an asynchronous "stream" mode, and a "serial" mode for
baseline measurements.
- It includes an optional interference simulation mode, activated by the
`--sum-outputs` flag, which sums the outputs of multiple parallel
channels.
- Simulation parameters (MIMO configuration, signal length,
channel length, and batch size) are configurable via command-line
arguments.
- It runs the full simulation pipeline on both the CPU and the GPU,
reporting the average execution time for each and calculating the
overall speedup.
This commit extends the CUDA API with new functions designed for
high-throughput scalability testing and advanced concurrency.
Key additions include:
- Batched versions of the kernels (`..._batched`) that use the
`blockIdx.z` grid dimension to process many independent channel
simulations in parallel.
- A C-callable wrapper, `run_channel_pipeline_cuda_batched`, to
orchestrate the execution of these new kernels.
- A "streamed" version of the pipeline,
`run_channel_pipeline_cuda_streamed`, which accepts a
`cudaStream_t` as an argument. This allows the calling
application to manage the execution stream, enabling overlap
with other GPU or CPU tasks.
This commit adds a performance benchmark for the end-to-end channel simulation pipeline, named `test_channel_simulation`. Its methodology is as follows:
- It systematically iterates through a matrix of test cases, varying
MIMO configuration (Tx/Rx antennas), signal length, and channel
length.
- For each case, it runs both the full CPU pipeline (`multipath_channel_float`
followed by `add_noise_float`) and the single `run_channel_pipeline_cuda`
function.
- It measures the average execution time for both implementations over
many trials and calculates the resulting speedup.
This provides a performance characterization of the GPU pipeline across different workloads and directly compares it to the CPU baseline.
This commit introduces the main C-callable pipeline function,
`run_channel_pipeline_cuda`, located in a new `channel_pipeline.cu`
file.
This function serves as the primary entry point for the GPU channel
simulation. It orchestrates the entire sequence of operations:
- Manages Host-to-Device data transfers for the input signal and
channel coefficients.
- Launches the `multipath_channel_kernel` and the
`add_noise_and_phase_noise_kernel` in the correct order.
- Manages the Device-to-Host data transfer for the final output signal.
- Synchronizes to ensure completion before returning.
This commit adds a new standalone benchmark, `test_noise`, to measure
and compare the performance of the CPU and CUDA implementations of noise
generation.
The benchmark's methodology is as follows:
- It iterates through various configurations, including different
numbers of receive antennas and signal lengths.
- For each configuration, it runs both the CPU (`add_noise_float`)
and CUDA (`add_noise_cuda`) functions multiple times to get a stable
average execution time.
- It prints a formatted table comparing the CPU and GPU latencies and
calculates the resulting speedup.
Robert Schmidt