UE (OnePlus) gets disconnected in a few seconds

[alektoja]

Problem: OnePlus fails during RRC/SRB establishment, while another UE (Core-Z5) attaches and passes traffic on the exact same setup.

Environment:
O-RU: Benetel RAN650 - RAN650-1v1.4.2-NM-c48047d
DU/CU host: Ubuntu 22.04.5 LTS - the newest commit
Core: Amarisoft Callbox
NIC for OFH: Intel Ethernet Controller E810-C for SFP
UE: Crosscall Core-Z5, OnePlus 11 5G
Swtich beetwen gNB and RU

Configs:
du.txt
ru.txt

Logs (excerpt):
2025-08-31T15:58:13.530064 [PHY ] [I] [ 369.5] PDSCH: rnti=0x4604 h_id=2 k1=4 prb=[0, 5) symb=[2, 6) mod=QPSK rv=1 tbs=12 t=16.9us
2025-08-31T15:58:13.530141 [PHY ] [I] [ 368.19] PUCCH: rnti=0x4604 format=1 prb1=4 prb2=na symb=[0, 14) cs=0 occ=0 sr=no metric=0.1 sinr=-16.9dB t=19.2us
2025-08-31T15:58:13.535151 [PHY ] [I] [ 369.9] PUCCH: rnti=0x4604 format=1 prb1=2 prb2=na symb=[0, 14) cs=0 occ=0 ack=2 metric=0.1 sinr=-19.7dB t=25.3us
2025-08-31T15:58:13.535526 [SCHED ] [I] [ 369.16] Slot decisions pci=1 t=15us (1 PDSCH, 0 PUSCHs): DL: ue=0 c-rnti=0x4604 h_id=2 ss_id=2 rb=[0..5) k1=12 newtx=false rv=0 tbs=12
2025-08-31T15:58:13.535527 [DU-F1 ] [I] ue=0 du_ue=3 cu_ue=3 proc="UE Context Release": RRC container not ACKed within a time window of 120msec. Proceeding with the UE context release...
2025-08-31T15:58:13.535531 [DU-MNG ] [I] ue=0 proc="UE Delete": Procedure started....
2025-08-31T15:58:13.535540 [DU-MNG ] [I] ue=0: SRB and DRB traffic stopped
2025-08-31T15:58:13.535543 [DU-F1 ] [I] ue=0 c-rnti=0x4604 du_ue=3 cu_ue=3: F1 UE context removed.
2025-08-31T15:58:13.535555 [PHY ] [I] [ 369.16] PDCCH: rnti=0x4604 ss_id=2 format=1_1 cce=12 al=2 t=16.8us
2025-08-31T15:58:13.535570 [PHY ] [I] [ 369.16] PDSCH: rnti=0x4604 h_id=2 k1=12 prb=[0, 5) symb=[2, 6) mod=QPSK rv=0 tbs=12 t=19.2us
2025-08-31T15:58:13.544638 [PHY ] [I] [ 370.8] PUCCH: rnti=0x4604 format=1 prb1=2 prb2=na symb=[0, 14) cs=0 occ=0 ack=2 metric=0.2 sinr=-20.4dB t=28.3us
2025-08-31T15:58:13.550162 [PHY ] [I] [ 370.19] PUCCH: rnti=0x4604 format=1 prb1=4 prb2=na symb=[0, 14) cs=0 occ=0 sr=no metric=0.1 sinr=-22.9dB t=25.6us
2025-08-31T15:58:13.552057 [MAC ] [I] [ 371.9] ue=0 crnti=0x4604 proc="UE Delete Request": finished successfully
2025-08-31T15:58:13.552059 [DU-MNG ] [I] ue=0 proc="UE Delete": Procedure finished successfully.
2025-08-31T15:58:13.552064 [DU-F1 ] [I] Tx PDU du=0 cu_ue=3 du_ue=3: UEContextReleaseComplete
2025-08-31T15:58:13.552068 [CU-CP-F1] [I] Rx PDU du=0 ue=3 cu_ue=3 du_ue=3: UEContextReleaseComplete
2025-08-31T15:58:13.552169 [CU-CP ] [I] ue=3: "UE Removal Routine" finalized
2025-08-31T15:58:13.552180 [NGAP ] [I] Tx PDU ran_ue=3 amf_ue=202: UEContextReleaseComplete
2025-08-31T15:58:15.044719 [NGAP ] [I] Rx PDU: Paging
2025-08-31T15:58:15.044731 [CU-CP-F1] [I] Tx PDU du=0: Paging
2025-08-31T15:58:15.044737 [DU-F1 ] [I] Rx PDU du=0: Paging
2025-08-31T15:58:16.287527 [SCHED ] [I] [ 645.0] Slot decisions pci=1 t=8us (1 PDSCH, 0 PUSCHs): PG: rb=[21..22) tbs=9 ues: cn-pg-id=0x1012a0ad205
2025-08-31T15:58:17.506813 [NGAP ] [I] Rx PDU: Paging
2025-08-31T15:58:17.506824 [CU-CP-F1] [I] Tx PDU du=0: Paging
2025-08-31T15:58:17.506831 [DU-F1 ] [I] Rx PDU du=0: Paging
2025-08-31T15:58:17.567522 [SCHED ] [I] [ 773.0] Slot decisions pci=1 t=8us (1 PDSCH, 0 PUSCHs): PG: rb=[21..22) tbs=9 ues: cn-pg-id=0x1012a0ad205
2025-08-31T15:58:19.967779 [NGAP ] [I] Rx PDU: Paging
2025-08-31T15:58:19.967789 [CU-CP-F1] [I] Tx PDU du=0: Paging
2025-08-31T15:58:19.967793 [DU-F1 ] [I] Rx PDU du=0: Paging
2025-08-31T15:58:20.127533 [SCHED ] [I] [ 5.0] Slot decisions pci=1 t=8us (1 PDSCH, 0 PUSCHs): PG: rb=[21..22) tbs=9 ues: cn-pg-id=0x1012a0ad205
2025-08-31T15:58:21.407516 [SCHED ] [I] [ 133.0] Slot decisions pci=1 t=7us (1 PDSCH, 0 PUSCHs): PG: rb=[21..22) tbs=9 ues: cn-pg-id=0x1012a0ad205
2025-08-31T15:58:22.430783 [NGAP ] [I] Rx PDU: Paging
2025-08-31T15:58:22.430793 [CU-CP-F1] [I] Tx PDU du=0: Paging
2025-08-31T15:58:22.430799 [DU-F1 ] [I] Rx PDU du=0: Paging

Full logs:
gnb_issue.log
mme.log

What we observe:

• Repeated DL HARQ for tiny SRB grants leading to 4× discard.
• PUCCH returns ACK=2 (NACK/DTX) with very low SINR (≈ −6 to −25 dB).
• SRB1 poll-retransmit timer expiries → UE context release.
• Real-time “late grid/FAPI late” warnings disappeared after enabling srsRAN performance tuning, but the OnePlus still fails attach.

Tried (no change for OnePlus):

• srsRAN performance script on DU/CU; upgrade to latest master.
• Multiple configs
• Increased inactivity_timer

We would appreciate any tips/solutions.

[alektoja]

Has anyone encountered something similar in the current version? Many thanks in advance for help.