- 05 Dec, 2023 1 commit
-
-
Robert Schmidt authored
-
- 04 Dec, 2023 4 commits
-
-
Robert Schmidt authored
-
Robert Schmidt authored
-
Robert Schmidt authored
-
Robert Schmidt authored
-
- 01 Dec, 2023 2 commits
-
-
Thomas Schlichter authored
-
Thomas Schlichter authored
-
- 30 Nov, 2023 18 commits
-
-
Robert Schmidt authored
The node ID (which corresponds to the gNB ID) should be read in all cases of execution, not only CU.
-
Robert Schmidt authored
-
Robert Schmidt authored
-
Robert Schmidt authored
The macrlc_has_f1 (static) variable was introduced to signal when we use F1 in the configuration. It is superfluous: the node_type already passes this information in a unique manner; hence, remove.
-
Robert Schmidt authored
-
Robert Schmidt authored
-
Robert Schmidt authored
-
Robert Schmidt authored
Correctly match UEs based on the NSSAI in MAC and RRC. There is a convenience function capture_sst_sd() that extracts SST and SD (which form the NSSAI). Because a previous version used integers to convey the slice identifier, this function gracefully handles this case, although it is not spec-compliant.
-
Robert Schmidt authored
Prior to this commit, the RAN function for KPM is leaking memory: it allocates using calloc() but never frees it. The fix consists of not allocating this memory on the heap in the first place, but on the stack. Therefore, there are changes to pass in a pointer to various functions to the structure on the stack, instead of passing the "container" for the heap memory by value. Further, the MAC was allocating an excessive amount of memory (N times sizeof(NR_UE_info_t)). Instead of copying the entire struct, the present code stores a pointer to the MAC UE container (NR_UE_info_t *). For the RRC, there is a similar simplification that consists in not storing the entire struct f1_ue_data_t (which is not required), but only the RRC UE ID. Note that both MAC and RRC might suffer from race conditions. For the MAC, we should lock the scheduler for the time we read statistics. For the RRC, the problem is more delicate as no mutex exists; the RRC assumes everything is executed synchronously in the ITTI task loop. In both cases, this commit does not address this problem
-
Robert Schmidt authored
Filtering UEs by NSSAI is broken, so remove it. A later commit will reestablish this functionality. Simplify the code: we only handle one condition, so we can check for it upfront.
-
Robert Schmidt authored
-
Robert Schmidt authored
This commit introduces an additional (implicit) option that signals if the E2 agent is to be activated or not. When reading the configuration, we now check if all fields are set, and activate the E2 agent only if this is the case. If not, we only print a warning and start the gNB without the agent. Note that previously, if not all fields are set correctly, the gNB stopped. To implement this behavior, it is necessary to change the defaults of the RIC IP and the directory for SMs to a (non-NULL) default value. Otherwise, the configuration system would not consider the field to be set to the default, and wrongly assumes the field is set, even though it is not; at least, this change is necessary to make config_isparamset() return NULL if the configuration is not set.
-
Robert Schmidt authored
3GPP has the concept of a gNB-CU-UP ID and a (separate) gNB ID (for gNB, CU-CP). This commit introduces the gNB-CU-UP as a separate ID that has to be set in the configuration file when running as a CU-UP. For the CU/monolithic gNB, it is optional (but needs to be the same as the gNB-ID if specified). The CU-UP ID is necessary, as some entities, e.g., the e2 agent, need to signal both IDs, e.g., to reconcile a CU-CP and (multiple) CU-UP(s) belonging together.
-
Robert Schmidt authored
3GPP has the concept of a gNB-DU ID and a (separate) gNB ID (for gNB, CU). This commit introduces the gNB-DU as a separate ID that has to be set in the configuration file. It is necessary, as some entities, e.g., the e2 agent, need to signal both IDs, e.g., to reconcile a CU and (multiple) DU(s) belonging together.
-
Robert Schmidt authored
-
Robert Schmidt authored
-
Robert Schmidt authored
-
Robert Schmidt authored
-
- 29 Nov, 2023 7 commits
-
-
Jaroslava Fiedlerova authored
-
Robert Schmidt authored
-
Robert Schmidt authored
-
Raphael Defosseux authored
* Dockerfile argument is now a generic BUILD_OPTION for any build_oai available option * docker-compose files are only modified at deployment time. Not needed for undeployment * fixes for oai-nr-cuup * cleaner section for the private registry push Signed-off-by: Raphael Defosseux <raphael.defosseux@eurecom.fr>
-
Raphael Defosseux authored
Signed-off-by: Raphael Defosseux <raphael.defosseux@eurecom.fr>
-
Raphael Defosseux authored
Signed-off-by: Raphael Defosseux <raphael.defosseux@eurecom.fr>
-
Raphael Defosseux authored
Signed-off-by: Raphael Defosseux <raphael.defosseux@eurecom.fr>
-
- 28 Nov, 2023 8 commits
-
-
Robert Schmidt authored
-
francescomani authored
-
francescomani authored
-
Jaroslava Fiedlerova authored
-
laurent authored
-
laurent authored
-
Jaroslava Fiedlerova authored
-
Jaroslava Fiedlerova authored
-