Skip to content

Latest commit

 

History

History
288 lines (275 loc) · 15 KB

backlog.md

File metadata and controls

288 lines (275 loc) · 15 KB
Raw

OpenAirInterface interop

  • F1 Setup Response not matching

NEXT UP

  • Open Air Interface interop

  • intermittent CI failure in two_workers_f1_before_e1 Jul 21 06:48:34.444 INFO f705b4a4-de48-4a18-81a4-7287bf2974b7 to get added to F1 interface amf: 1 Jul 21 06:48:34.446 INFO End listen 127.148.206.99:38412 Jul 21 06:48:34.446 INFO TNLA 24 closed cu-up: 127.86.101.151 Jul 21 06:48:34.447 INFO E1AP TNLA 22 closed amf: 1 Jul 21 06:48:34.447 INFO TNLA 31 closed cu-cp-w: f705b4a4-de48-4a18-81a4-7287bf2974b7 Jul 21 06:48:34.447 INFO End listen 127.0.25.178:38462 cu-up: 127.86.101.151 Jul 21 06:48:34.447 INFO E1AP TNLA 40 closed Error: Unexpected F1ap message InitiatingMessage(GnbCuConfigurationUpdate(GnbCuConfigurationUpdate { transaction_id: TransactionId(1), cells_to_be_activated_list: None, cells_to_be_deactivated_list: None, gnb_cu_tnl_association_to_add_list: Some(GnbCuTnlAssociationToAddList(NonEmpty { head: GnbCuTnlAssociationToAddItem { tnl_association_transport_layer_address: EndpointIpAddress(TransportLayerAddress(BitVec<u8, bitvec::order::Msb0> { addr: 0x7f6f3000a060, head: 000, bits: 32, capacity: 64 } [0, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 0, 0, 1, 1, 0, 1, 1, 0, 0, 1, 0])), tnl_association_usage: Both }, tail: [] })), gnb_cu_tnl_association_to_remove_list: None, gnb_cu_tnl_association_to_update_list: None, cells_to_be_barred_list: None, protected_eutra_resources_list: None, neighbour_cell_information_list: None, transport_layer_address_info: None, ul_bh_non_up_traffic_mapping: None, bap_address: None })) cu-cp-w: f705b4a4-de48-4a18-81a4-7287bf2974b7 Jul 21 06:48:34.447 INFO E1AP TNLA 41 closed cu-cp-w: 5508deba-7ca5-461d-b452-b4d0db412937 Jul 21 06:48:34.448 WARN Failed refresh worker ApiError("No response received: connection closed before message completed") cu-cp-w: f705b4a4-de48-4a18-81a4-7287bf2974b7 Jul 21 06:48:34.448 INFO End listen 127.0.25.178:38472 cu-cp-w: 5508deba-7ca5-461d-b452-b4d0db412937 Jul 21 06:48:34.448 INFO NGAP TNLA 23 closed cu-cp-w: f705b4a4-de48-4a18-81a4-7287bf2974b7 Jul 21 06:48:34.448 INFO NGAP TNLA 30 closed thread 'async-std/runtime' panicked at 'called Result::unwrap() on an Err value: SendError(..)',

  • RRC connection release

  • UE context release

  • userplane HA (two E1AP connections and userplane replication)

  • Testing of Session/context releases on different worker

  • Paging

  • Promotion of coordinator in single executable mode

TECH DEBT

BUGS

  • Ctrl-C is not reliably terminating

CU-UP and O-RAN O-DU interop

  • avoid need for recompile of ODU by enabling O1 (but we need to recompile it anyway to set ratio = 20)
  • document a method that other people could use to test ODU
  • don't set up SRB + 2 DRBs if all we need is one session = one DRB

SCALE OUT / MULTIPLE TNLA

  • Allow AMF to specify 2nd endpoint - ask worker 1
  • Allow AMF to specify 2nd endpoint - ask worker 2
  • Load balance, stickiness and switchover between TNLAs to AMF
  • Allow DU / UP to set up multiple connections to same worker
  • Restart and catchup of coordinator
  • Failure and retry to set up / join NG / F1 / E1
  • Switchover of UE on dead worker - RAN initiated
  • Switchover of UE on dead worker - AMF initiated
  • Connection API operations should be idempotent
  • Stickiness and switchover between TNLAs to DU / UP
  • AMF not started at point workers start
  • Both workers die - reset
  • All NGAP TNLAs drop - don't reset, ues-retained = true.
  • All F1AP TNLAs drop - reset.
  • All E1AP TNLAs drop - ?

CONNECTION MANAGEMENT

  • Regression test for tearing down requests when a connection dies
  • Retry connection to AMF if connection refused. (e.g. just run GNB-CU-CP on its own)
  • Fix hang on Ctrl-C when AMF connect doesn't complete
  • two worker enablement of DU interactions (share DU configuration between workers - see [documentation/design/State - DU.md])

FUNCTION

  • Proper graceful shutdown (waiting for / sending responses to pending requests)
  • Make values in NG Setup configurable rather than hard coded (Tac, Plmn Id, slices, etc)
  • Don't hang indefinitely waiting for response (e.g. NG Setup response)
  • Don't allow unlimited pending requests
  • Handle -ve response to InitialContextSetupRequest with bad RAN UE ID

MAINTAINABILITY + DIAGNOSTICS

  • standardize handler->workflow result handling and logging (who logs message, who forms failure) across CU-CP and CU-UP
  • Rather than saying "WARN Unsupported UlDcchMessage C1(RrcReconfigurationComplete" we should report that the messsage does not match a transaction.
  • in RequestProvider, pass a transaction context instead of a logger
    • transaction context provides logger
    • transaction context also provides "response action" to simplify?
    • also provides TNLA ID and remote IP address - can use this to remove double info! logging of TNLA setup
  • Remove slog from workflow module and access above transaction context via Workflow instead
  • Errors are too easy to miss - log_ue_error()? to optionally warn! on failure
    • e.g. "Inital access procedure failed - Connection refused (os error 111)" at debug
  • Ue logging level should be settable to allow warnings to show up. UE context should appear in logs / be stored in Logger.
  • Cleaner RRC interface in trait Gnbcu
  • Enforce Rust docs (see .cargo/config commented out compiler option)

TESTS

  • Failure to retrieve UE for each kind of request
  • Efficient monolithic GNB-DU + GNB-CU can be built without a F1AP Stack or TransportProvider
  • Efficient monolithic GNB-CU-CP + GNB-CU-UP can be built without an E1AP Stack or TransportProvider

ASN.1 GENERATOR

  • fixed size octet string should be [] not Vec?
  • inlining of lists to avoid newtypes of vecs (e.g. PDU-Session-Resource-Activity-List)
  • ENUMERATED{True} OPTIONAL (as seen in RRC) should appear in Rust as a bool
  • Cope with extension marker being set
  • Get rid of todo!() in top_pdu.rs and replace with a log
  • Implement Rrc setuprelease
  • Move to latest asn1-codecs crate version
  • Deduplicate inline definitions in RRC autogeneration
  • Generate procedures for Rrc and make F1AP a RequestProvider.
  • Move to latest version of specs
  • Fix clippy

REDIS

  • Don't create 1 Redis connection per access
  • Live redis test returns ok even after "# Failed listening on port 23491 (TCP), aborting."
  • Redis live test should not create Redis dump.rdb

SCTP

  • sock_opt.rs doesn't need to be a separate file
  • Remodel SCTP API to look like the one in the webrtc-sctp crate?

FREE5GC DEMO

  • free5GC demo can register 2 (N?) UEs
  • GNB-CU-UP executable can be started in demo and performs E1 Setup with GNB-CU-CP

MEDIUM TERM

  • Selection and stickiness of SCTP streams
  • Dockerfiles and Helm charts
  • Distributed timers and failure path cleanup mechanism

DONE

  • Pass SIB2 on F1 Setup Reponse
  • Encoding of IE extension container items
  • APER align after decoding IE
  • Add coordinator to single process GNB-CU
  • PDU session deletion
  • Retry connection to AMF
  • Single process GNB-CU with configurable MCC / MNC
  • SDAP / PDCP should be added/stripped
  • local-ip arg for mock-5gc and amf-ip arg for gnb-cu-cp
  • mock_amf becomes mock_5gc
  • dashmap
  • E1 Setup
  • CU-UP passes through userplane packets
  • ASN.1 generator makes use of NonEmpty
  • state.md flow "Eventually the AMF furnishes the GNB" wrongly shows DU context being created
  • Refactoring of pdu_session_resource_setup.rs
  • Add PduSessionId to xxap common
  • Improved case conversion of DLUPTNLInformation-ToBeSetup-List and similar
  • TransportLayerAddress has TryFrom<&str>
  • Form PDUSessionResourceSetupResponse correctly
  • Avoid missing NAS message Wireshark error on RrcReconfiguration
  • Find a way to get O-DU PHY stub to send ReconfigurationComplete in the right order (recompile with ratio = 20)
  • Common XXAP structures in Asn.1 generator (e.g. Snssai, GtpTunnel) to allow easy transfer between NGAP, F1AP, E1AP
  • Use proper 32 bit bitstrings for TransportAddress in E1AP messages (avoids Wireshark decode issue)
  • Support for BearerContextModificationRequest in CU-UP
  • Pass CellGroupConfig from DU to UE in Rrc Reconfiguration Request
  • Use proper NAS messages from AMF-SIM to improve Wireshark trace of ODU interop
  • Form CUtoDURRCInformation correctly in UeContextSetupRequest
  • Supply DRBs-to-be-Setup-List on UeContextSetupRequest
  • ASN.1 generator copes with choice IE extensions
  • Store cell (NR-CGI) on UE context and pass on UeContextSetupRequest
  • Tear down requests when a connection dies (e.g. if gnb-cu-up aborts while handling a request)
  • Add GNB-CU-UP
  • DL-CCCH-Message should not be PDCP encapsulated
  • 'Worker startup failure' (e.g. when specifying wrong --local-ip) should be fatal
  • RRC is UPER not APER
  • Call TnlaEventHandler serially for a given association allowing message ordering control by upper layers (...meaning that intermittent reordering in scripted tests can be avoided)
  • Tolerate missing mandatory IE TransactionId on InitialULRrcMessageTransfer for ORAN ODU interop
  • Respond to GNBDUConfigurationUpdate
  • Include CellsToBeActivated on F1SetupResponse.
  • Fix intermittent failures in mock connection checking
  • F1 Setup with O-RAN ODU
  • ASN.1 decoder should cope protocol extensions being present.
  • Allow local-ip to be passed in on command line
  • Triangular redirection
  • Get rid of revision number
  • Go public
  • Fix TNLA established event so that it is always processed before first packet
  • Add readmes + review as they appear on Github
  • Rename gnbcu to gnb-cu-cp
  • Simplify and add revision number to connection API
  • UP / DU connections in either order
  • Coordinator ensures time gap between attempts to add workers
  • 2nd worker receives UP / DU connection and adds 1st worker
  • Test one UE through each worker
  • Assoc should be in pool by the point that the connect call returns
  • 1st worker initializes NG interface and 2nd worker joins in
  • 1st worker receives UP connection and adds 2nd worker
  • 1st worker receives DU connection and adds 2nd worker
  • Timing bug causing cu_can_connect_to_amf() to sometimes hang when logging disabled. No repro
  • Log interleaving when RUST_LOG=debug and multiple tests run in parallel - noop - just do RUST_TEST_THREADS=1
  • Two workers up
  • Use IP address instead of ports to distinguish NGAP, E1 and F1 endpoints
  • Standalone single worker that runs built-in coordinator
  • Parallel registration of two UEs
  • Get build working on Github
  • Parallel tests
  • Don't run live redis test by default
  • PDU session resource setup - code review and tidy
  • Break procedures into small functions
  • NG Setup is in workflows but F1 and E1 Setup aren't
  • Ints should be Copy rather than Clone (e.g. AmfUeNgapId)
  • Registration sequence should not set up UE context in DU or invovle Rrc Reconfiguration
  • SRB set to 0 or 1 and checking of that in Mock DU
  • Add comments at top of files
  • GNB-CU-CP serves E1 and supports E1 Setup
  • Clean up Redis even if live redis test fails
  • E1AP generation
  • Update Amf Mock to support multiple Ues
  • Registration of multiple UEs in FV
  • Tidy code e.g. in procedures, GnbcuOps
  • GNBDU-SIM should produce a clear error message if mongo is not started
  • In demo, GNBDU SIM should not warn 'connection failed - will retry' on shutdown
  • Get live test and demo working
  • Store AMF UE Ngap Id on first downlink NAS transport
  • Uplink and downlink NAS transfer should be in the procedures module
  • Fix int coding bug
  • Redis live test
  • Single async fn / task that sends the Rrc Setup and waits for Rrc Setup Complete
  • Fix failing test
  • Code review (incl commented out code and TODOs) - gnbcu
  • Redo roadmap
  • In demo, gnbdu-sim has inconsistent message logging
  • At end of demo, gnbcu panicked on Ctrl-C with 'not yet implemented'. AMF was Ctrl-C'd first. AMF logs indicate it is sending AMF Status Indication. Crash is in ngap_gnb.rs. Add an FV for this.
  • Remove the weird TNLA 3, TNLA 53 IDs
  • Three lines for each log sucks
  • In demo, gnbcu logs strangely end with ">> UlRrcMessageTransfer". Where's the uplink message?
  • In demo, gnbcu logs saying UlRrcMessageTransfer should include (nas)
  • In demo, inconsistent debug log "Got Downlink Nas Transport - send RRC to UE via DU"
  • In demo, downgrade log Accepted connection from Some(127.0.0.1:60313), since we already have an INFO log for that
  • In demo, consider merging 'Accepted SCTP connection from 127.0.0.1:60313' and 'NGAP TNLA 3 established' into single INFO log
  • In demo, inconsistent debug log "InitialUeMessage >"
  • Write up reliable demo instructions starting with download of free5GC
  • Add quickstart instructions and notes on current status to top readme
  • Bump deps
  • Move to consistent message logging
  • Regression test integration covering same ground as live test with free5GC
  • Pretty Ue initial context and move into separate modules
  • Fix malformed CellGroupConfig issue seen in Wireshark in F1AP UEContextSetupResponse -- DUtoCURRCInformation, cellGroupConfig shown as
  • Build go code using cargo build.
  • Fix code that does match match
  • Registration (23.502, figure 4.2.2.2-1) wrapped in UE Initial Access procedure (38.401, figure 8.1-1) working against free5GC as seen in ngap_setup.pcap.
  • Go program that processes NAS information
  • Encapsulate RRC in PDCP PDU
  • gnbdu-sim
  • RRC autogeneration
  • Multi worker code (including coordinator) moved to a 'prototype' version leaving remaining code super simple
  • NG Setup working again against free5GC using the instructions in the free5gc howto.
  • Add error messages to ASN.1 library + get it upstreamed
  • GNB-CU code moved to a subdirectory (leaving code shared with DU or AMF at top level).
  • Autogeneration of procedures etc
  • Fix Address already in use warning seen in tests. Refactor so that the listen() fails more obviously.
  • Test should fail if the F1AP SCTP bind fails
  • 'Worker confirms successul TNLA initialization' is output before output of AMF sending setup response
  • send request instead of send pdu
  • Test script as DU can send Setup and get response
  • Test script as DU can connect to CU
  • Create callback client on coordinator
  • Drive NG Setup from callback server
  • Drive RAN configuration update from coordinator
  • Start callback server on worker
  • Develop callback in node control API.
  • test the sctp receive function
  • sort out TODOs and unwraps()
  • commonize TNLA pool
  • try out the Hampi library to decode the response at least?
  • start fake amf
  • get info / trace levels right
  • get tests passing cleanly
  • remove bindgen test cruft
  • on startup, the worker should connect to node controller, get AMF address, connect
  • fix typo refresh
  • factor out logging and signal handling to common library
  • treat that as a separate package within the same workspace for now
  • create a basic POST API (later we add callback server)
  • ok so we need to run OpenAPI generator
  • Implement the node controller / investigate OpenAPI
  • see if two alsorans can connect to one AMF
  • documentation of features and level of testing
  • socket OS errors getting hidden by anyhow
  • Maintain + Retry
  • Successful NG Setup with Free5G.
  • get rid of haphazard error conversion - see TODOs
  • clean up sctp_association.rs
  • Clippy
  • send a fake ng setup and test the sctp receive function
  • Properly close sockets when wrapper structs are dropped
  • Stay up until Ctrl-C
  • SCTP connect
  • test against a free5GC
  • add logging
  • explore HA + use of SCTP.