[CI] Ensure multi-arch DockerHub Push after successful image build
This merge request improves the CI script to ensure that the DockerHub-Push
stage executes whenever:
- All Image Build stages succeed &
- doBuild is true &
- The event is a PUSH
This behavior ensures that Docker images are published whenever the image build
stages succeed, regardless of any failures in the Image Test stages.
Jetson ARM64/V8 and AMD64 images will now be published on DockerHub under the
same tag (develop or WEEK_TAG) to provide multi-architecture images.
Optimization of PDSCH precoding operation
x86/aarch64 optimizations for precoding, more complete timing reporting in
nr_dlsim. Timing report in nr_dlsim now shows the total PDSCH generation time
and is broken down into Layer mapping, resource mapping and precoding.
The optimizations for x86 are basically reorganizing loops (for AVX512,AVX2,
NEON but not SSE2) to minimize repetitive instructions when formatting of the
precoder weights, and creation of inline functions for complex multiply-add.
AVX512 optimization was added. requiring AVX512BW. Detection of up to 4 PRBs
with common PMI for efficiency with AVX512 (was 2 in develop for AVX2).
For aarch64, native coding of the precoder to avoid inefficient SIMDe
translations for this operation.
nr_dlsim now uses default precoding indices depending on the scenario instead
of 0 which doesn't do precoding at all.
See merge request description for nr_dlsim test results.
- On PUSH events, if image build stages succeed, trigger DockerHub push stage
- Build the images, then trigger the Push to DockerHub stage
- Publish Jetson (ARM64/v8) and AMD64 images under the same tag on DockerHub
Reduce throughput requirement, as this fails once two DUs are connected
to a UE, which creates more load on the machine and frequently fails
with low throughput.
CI: Enable Physim threshold check for tests on Caracal and GH
This MR adds support for configurable timing threshold files in PhySim
tests.
- Introduces a new CMake cache variable PHYSIM_CHECK_FILES to specify
one or more timing threshold files (semicolon-separated).
- Enables optional enforcement of timing thresholds check in PhySim
tests without changing existing test definitions.
- Defines timing thresholds for LDPC encoding/decoding in nr_ulsim,
nr_dlsim and ldpctest on Caracal and GH machines.
SRS noise power
This MR implements the calculation of the noise power based on SRS in a
different way, and check whether SNR is more accurate or not. This MR is
built on top of the MR: !3460
A graph was added in T_Tracer to observe UL-SNR calculated using SRS.
See merge request description for measurement results.
Replace the previous add_timed_physim_test() macro with add_physim_test()
to use a single, unified function for defining physim tests. This simplifies
the test configuration and removes redundant macro definitions while keeping
all existing functionality.
This change follows up on the earlier addition of support for including
custom timing threshold files (PHYSIM_CHECK_FILES).
Update existing tests accordingly.
Add missing timinig analysis scripts to RHEL9 physim docker image. These files
are now copied from the build stage to /oai-ran/openair1/SIMULATION/tests/,
where they are expected by ctest.
Resolves following issue when running ctest in docker container:
CMake Error: Error processing file: /oai-ran/openair1/SIMULATION/tests/RunTimedTest.cmake
For some physical simulators, we have multiple SNR iterations.
The previous version of the script correctly retained only the last
measurement, but if a previous measurement condition set RC=1, the
script would fail although the final "good" run might fulfil the
measurement condition (IOW, it would show SUCCESS, but still fail on the
check).
Instead, check if all measurement lines finish on SUCCESS, which handles
the above case.
The awk script in analyze-timing.sh used syntax that works with gawk but fails
on systems where mawk is the default implementation, producing errors such as:
awk: /dev/fd/3: line 6: syntax error at or near ,
-- test command finished with SIGPIPE
Concretely, the match(text, pattern, variable) function does not exist
in mawk, as it cannot capture matches in variables. Work around by
substituing strings, and force a numeric comparison.
Fixes to prevent assertion due to 0 PRB allocation
Assertion (NPRB>0 && (NPRB + RBstart <= BWPsize)) failed!
In PRBalloc_to_locationandbandwidth0() /home/user/openairinterface5g/common/utils/nr/nr_common.c:506
Illegal NPRB/RBstart Configuration (0,51) for BWPsize 51
Probably caused by missing safeguards to prevent computing TBS with 0 PRB
as input.
Fix OAI UE memory leaks
Fix 1: Replaced asn1cFreeStruc with ASN_STRUCT_FREE in MAC layer to properly
free nested ASN.1 structures in cell group configuration.
Fix 2: Removed premature nullification of dedicatedNAS_Message->buf in
dlInformationTransfer processing to allow proper ASN.1 cleanup.
Fix 3: Fixed memory ownership in nr_rrc_process_dedicatedNAS_MessageList by
copying OCTET_STRING data instead of transferring ownership, preventing
ASN.1 cleanup from losing track of allocated memory
Closes#828
fix RSRP computation at UE
Conflicts between !3595 (merged) and !3596 (merged) leading to double conversion
of RSRP into index in L1 and L2. With this fix it is only converted in L2.
Replace int16 min when obvious
INT16_MIN doesn't have a opposite value on 16 bits, so SIMD abs (absolute()), or
simple C operations doesn't make the expected result.
example:
(gdb) p (short)-(short)-32768
$4 = -32768
(gdb)
This MR replaces constants INT16_MIN by -INT16_MAX (so -32767) in places where
it used for mathematical operations. This is not perfect, nevertheless reduces
overflow probability.
We saw in some places a dirty case: code do first calls to _mm_adds_epi16
(or subs), so because of the saturated operations, the probability to reach
32767 and -32768 is higher than the other numbers but, _mm_abs_epi16(), or
negate operations (conjugate) can follow, that is making strong error as
abs(-32768)=-32768
Also, don't use "set -x" debugging for oc login command:
ci-scripts/cls_cmd.py:242 sets BASH_XTRACEFD=1, which makes that all
"set -x" output is in stdout, leading to a redirect to the file into
which we redirect. This would lead to the following error with
unsuccessfull oc login:
error: You are not a member of project "****-core-for-ci-ran".
Error from server (Forbidden): imagestreams.image.openshift.io "oai-physim"
is forbidden: User "system:anonymous" cannot get resource "imagestreams" in
API group "image.openshift.io" in the namespace "****-ran"
This fixes basically two problems. First, remove "set -e", which leads
to the script exiting on error. For instance, there was the sleep which
should account for the time of the pod being in "ContainerCreating", but
might not be enough. In that case, the script would exit, leaving the
pod deployed.
Second, a fixed time is not good, as it can be too short and unduly
prolong waiting time. Replace with a loop.
- Remove unnecessary NULL assignment for dedicatedNAS_Message->buf
- The original buffer is copied and transfered to NAS, so it can be properly freed by ASN.1 structure cleanup
Root cause: Code was nullifying dedicatedNAS_Message->buf pointer before ASN.1 cleanup could free the original allocation, causing the leak from OCTET_STRING_decode_uper.
AddressSanitizer leak trace (FIXED):
Direct leak of 119 byte(s) in 3 object(s) allocated from:
/#0 0x7ec2a42b4c38 in __interceptor_realloc
/#1 0x555fade852e0 in OCTET_STRING_decode_uper
/#9 0x555fadb48b11 in nr_rrc_ue_decode_dcch.
The leak was caused by improper memory ownership handling in the
dedicatedNAS_MessageList processing. The function was transferring
buffer ownership to NAS messages and then preventing ASN.1 cleanup
from freeing the original allocations by setting
list.count = 0, causing a memory leak.
Leak trace:
Direct leak of 40 byte(s) in 1 object(s) allocated from:
/#0 0x7f2ce3ab4a57 in __interceptor_calloc ../../../../src/libsanitizer/asan/asan_malloc_linux.cpp:154
/#1 0x567907083f0f in OCTET_STRING_decode_uper /cmake_targets/ran_build/build/openair2/RRC/LTE/MESSAGES/OCTET_STRING_uper.c:89
/#2 0x5679070ed5e6 in SET_OF_decode_uper /cmake_targets/ran_build/build/openair2/RRC/NR/MESSAGES/constr_SET_OF_uper.c:70
/#3 0x56790707b6dc in SEQUENCE_decode_uper /cmake_targets/ran_build/build/openair2/RRC/LTE/MESSAGES/constr_SEQUENCE_uper.c:118
/#4 0x56790707b6dc in SEQUENCE_decode_uper /cmake_targets/ran_build/build/openair2/RRC/LTE/MESSAGES/constr_SEQUENCE_uper.c:118
/#5 0x56790709a403 in CHOICE_decode_uper /cmake_targets/ran_build/build/openair2/RRC/LTE/MESSAGES/constr_CHOICE_uper.c:84
/#6 0x56790707b6dc in SEQUENCE_decode_uper /cmake_targets/ran_build/build/openair2/RRC/LTE/MESSAGES/constr_SEQUENCE_uper.c:118
/#7 0x56790709a403 in CHOICE_decode_uper /cmake_targets/ran_build/build/openair2/RRC/LTE/MESSAGES/constr_CHOICE_uper.c:84
/#8 0x56790709a403 in CHOICE_decode_uper /cmake_targets/ran_build/build/openair2/RRC/LTE/MESSAGES/constr_CHOICE_uper.c:84
/#9 0x56790707b6dc in SEQUENCE_decode_uper /cmake_targets/ran_build/build/openair2/RRC/LTE/MESSAGES/constr_SEQUENCE_uper.c:118
/#10 0x56790708b5f1 in uper_decode /cmake_targets/ran_build/build/openair2/RRC/LTE/MESSAGES/uper_decoder.c:83
/#11 0x567906d44b4b in nr_rrc_ue_decode_dcch /openair2/RRC/NR_UE/rrc_UE.c:2212
/#12 0x567906d520a3 in rrc_nrue /openair2/RRC/NR_UE/rrc_UE.c:2615
/#13 0x567906d5f77e in rrc_nrue_task /openair2/RRC/NR_UE/rrc_UE.c:2501
/#14 0x7f2ce2494ac2 in start_thread nptl/pthread_create.c:442
Fix:
- Copy OCTET_STRING data instead of transferring ownership
- Allow ASN.1 cleanup to properly free original allocations
This fixes the leak that occurred during rrcReconfiguration
message processing when dedicatedNAS_MessageList was present.
Closes#828
- Replace asn1cFreeStruc with ASN_STRUCT_FREE in MAC layer cleanup
- Fixes improper memory management of NR_CellGroupConfig structures
Root cause: asn1cFreeStruc only calls ASN_STRUCT_RESET + free(), which doesn't properly free nested ASN.1 allocations. ASN_STRUCT_FREE frees all nested structures completely.
AddressSanitizer leak trace:
Direct leak of 288 byte(s) in 1 object(s) allocated from:
/#0 0x7fb72bab4a57 in __interceptor_calloc
/#1 0x593752b28d0d in SEQUENCE_decode_uper
/#2 0x593752b9a546 in SET_OF_decode_uper
/#3 0x593752b2863c in SEQUENCE_decode_uper
/#4 0x593752b47363 in CHOICE_decode_uper
/#5 0x593752b2863c in SEQUENCE_decode_uper
/#6 0x593752b2863c in SEQUENCE_decode_uper
/#7 0x593752b2863c in SEQUENCE_decode_uper
/#8 0x593752b2863c in SEQUENCE_decode_uper
/#9 0x593752b2863c in SEQUENCE_decode_uper
/#10 0x593752b38551 in uper_decode
/#11 0x5937527d859f in nr_rrc_ue_process_masterCellGroup
/#12 0x5937527d9408 in nr_rrc_process_reconfiguration_v1530
/#13 0x5937527dd5bb in nr_rrc_ue_process_rrcReconfiguration
/#14 0x5937527f5e6b in nr_rrc_ue_decode_dcch
/#15 0x5937527ff003 in rrc_nrue
/#16 0x59375280c6de in rrc_nrue_task
Integration: `2025.w40`
* !3672 \[ci\] Fix MissingPropertyException - COTS-UE by binding OAI_Registry globally
* !3609 BWP with no PRACH occasions
* !3657 do not assert for mismatched TBS in phy-test mode
* !3667 CI: Updates for HO and AW2S pipeline
* !3674 NR UE: remove function get_nr_RSRP() and get_nr_rx_total_gain_dB() for cleanup
* !3478 NR UE: add UCI (HARQ-ACK/NACK) on PUSCH support at PHY
* !3664 modulation LUTs size was wrong, and remove inconsistent test of SSE2 enable
* !3665 Make all targets compile
* !3554 Handle failures in RRCReconfiguration encoding gracefully
* !3655 Improvements to N2 handover
* !3678 Add OAIUE longrun pipeline
* !3628 Update documentation
Closes#979, #726, #959, #967, and #930
See merge request oai/openairinterface5g!3676
Add OAIUE longrun pipeline
Add a new XML file for OAIUE longrun.
- Test description:
* 40 MHz BW, n78
* gNB with USRP N310, UE with USRP N310
* CN deployed on OC
- Test duration: 1 hour
- Test scenario: 3x DL 10 min test (20 Mbps) + 3x UL 10 min test (5
Mbps)
pods deployed through helm occasionally take more time, leading to
errors such as
Error from server (BadRequest): container "physim" in pod "oai-physims-4g-sczvl" is waiting to start: ContainerCreating
Give a bit more time before checking the logs. We cannot wait too much,
as the container will just end on FINISHED in running state, so the logs
will always be there.
Jaroslava Fiedlerova