[josephnef]

#566, causes not established: the m6_daps_dl run with DL to the arriving UE ~111 µs late and 13 uplinks lost, recorded as "not reproduced".

The cause, from the original logs (regf-dl/fault-daps-1):

1. The CU's coordination schedule moved DU2's own UE from position 1 to 2. It was staged at net sf ~404, effective at 416.
2. Right after, the CU issued the handover with target position 2. It had taken that position as free from DU2's report, which predated the schedule.
3. DU2 accepted the command. Its free-position check read f1_view.ues, a map snapshot refreshed once per main-loop pass, not at the commit.
4. The arrival was published at position 2 as well. The arrival's DL frame then queued behind the other UE's frame in the same slot (~110 µs, one frame's airtime), and their uplinks collided.

It was not rare. Across the saved rig logs:

• 201 of 1550 handover commands (13 %) were refused by the target as invalid. That is the same race, caught when the target's view was fresh; the CU retried.
• 41–53 % of the same-channel A3 handover runs had one.
• 14 DU logs show the race missed: two UEs on one position.
• Cross-channel runs never had it, because coordination moves only co-channel positions.

Fix

• CU:
  • No handover while either DU has a schedule pushed and not yet applied. This is the reverse of the schedule's own in-flight guard.
  • The target id and position are chosen against the CU's own last schedule for that DU, as well as against the report.
• DU:
  • A staged schedule's positions count as taken (Stager::staged()).
  • The map view the handover check reads is refreshed at the commit.

Rig after (fixed build on all three boxes):

• m6_handover.sh 6 normal + 3 fault, plus m6_daps_dl.sh (all 16 runs): 25 runs, 29 commands.
• 0 refused, no position shared.
• The arrival's DL late_us p50 was 2–3 µs in every run.

ctest: 95/95 pass. docs/cell-m6.md "Bugs the rig found" records it.

🤖 Generated with Claude Code

[qodo-free-for-open-source-projects]

PR Summary by Qodo

Prevent handover arrivals from sharing staged schedule positions

🐞 Bug fix 📝 Documentation 🕐 20-40 Minutes

AI Description

• Delay handovers while either DU has an unapplied schedule, preventing selection from outdated
 reports.
• Reserve reported and staged positions on both CU and DU, and refresh the DU view at map commit.
• Document the race, its observed frequency, and post-fix rig results.

Diagram

sequenceDiagram
    participant CU as Controller
    participant Source as Source DU
    participant Target as Target DU
    participant Stager as Schedule stager
    participant Map as Map view
    Target->>CU: Report positions and versions
    Source->>CU: Report applied version
    CU->>CU: Wait for both schedules to apply
    CU->>CU: Reserve report and scheduled positions
    CU->>Target: Handover command
    Target->>Stager: Read staged positions
    Target->>Map: Read committed positions
    Target->>Target: Validate against both sets
    Map->>Map: Refresh at map commit

Loading

High-Level Assessment

The following are alternative approaches to this PR:

1. Target-side position reservation handshake

• ➕ Makes the target authoritative before the controller commits to an arrival position.
• ➖ Requires additional protocol state and round trips, complicating handover timing.

Recommendation: Keep the layered guard and validation in this PR. It uses existing schedule versions and target-side state to prevent the collision without introducing a new reservation protocol.

Files changed (4) +57 / -2

Bug fix (3) +27 / -2

main.cppValidate arrivals against staged and committed DU positions +10/-0

Validate arrivals against staged and committed DU positions

• Includes staged schedule assignments in the handover admission check. Refreshes the shared position view when a grant map commits rather than waiting for the next main-loop pass.

examples/cell/main.cpp

F1.hExpose the DU's staged schedule for admission checks +3/-0

Expose the DU's staged schedule for admission checks

• Adds a read-only accessor so handover validation can treat positions in an unpublished schedule as occupied.

src/cell/F1.h

main.cppAvoid handovers across pending schedules and reserve planned positions +14/-2

Avoid handovers across pending schedules and reserve planned positions

• Blocks handover issuance while either DU has a pushed, unapplied schedule. Selects the target UE ID and position against both the target report and the controller's last schedule.

tools/ran-controller/main.cpp

Documentation (1) +30 / -0

cell-m6.mdDocument the staged-schedule handover race +30/-0

Document the staged-schedule handover race

• Explains how an outdated target report and stale DU map view allowed two UEs to share a position. Records observed rig frequency, the fix, and post-fix run results.

docs/cell-m6.md

[qodo-free-for-open-source-projects]

Code Review by Qodo

🐞 Bugs (0) 📘 Rule violations (0) 📎 Requirement gaps (0)

Great, no issues found!

Qodo reviewed your code and found no material issues that require review

Tip of the day

💡 Did you know, you can ask Qodo to dismiss a finding you disagree with, with your reason on record

More tips ↗ | Customize Qodo ↗ | Qodo docs ↗