ieee80211: support immediate Compressed Block Ack

[mgonzalezlopezudc]

HCF previously rejected Compressed Block Ack, so users could not study its bitmap, response, and recovery behavior through a complete exchange. This PR adds immediate Compressed Block Ack to the supported individual-frame MAC path. Eligible HT infrastructure links and the supported ac path use Compressed by default. Explicit Basic remains available for historical compatibility.

The originator sends individual QoS DATA frames and requests receipt status with an explicit BAR. The recipient returns a 64-bit receipt bitmap. HCF uses the actual profile for response matching, rates, airtime, protection, and timeouts. The Compressed response contains 32 bytes with FCS; Basic contains 152 bytes. A smaller response alone does not prove a performance gain.

The main compatibility changes affect configuration and custom components. Both endpoints must select the same profile because ADDBA does not negotiate it. Links without usable HT state use Normal Ack. Custom MAC and PHY components can require interface changes. The supported path includes A-MSDU, but this PR does not add A-MPDU transfer or complete HT-immediate transfer coverage.

Which links use the new default

HCF resolves blockAckVariant once at initialization. A Block Ack agreement identifies one peer and traffic identifier, or TID. Compressed requires usable local HT support and peer HT state for the relevant direction. A PHY mode name alone does not supply that state. Associated ac peers can qualify through their inherited HT support.

flowchart TD
    E{"Is Block Ack enabled?"}
    E -->|No| N["Use ordinary acknowledgment behavior"]
    E -->|Yes| P{"Which profile does HCF select?"}
    P -->|Basic| B["Retain historical Basic eligibility<br/>Peer HT records are not required"]
    P -->|Compressed default| H{"Are local HT support<br/>and peer HT state usable?"}
    H -->|No| F["Use Normal Ack<br/>Do not establish Compressed"]
    H -->|Yes| A["Apply setup eligibility rules<br/>Use Compressed after ADDBA acceptance"]

Loading

For example, an HT ad hoc configuration can lack peer HT records. The default still selects Compressed, but DATA uses Normal Ack. To retain a historical Basic study, select mac.hcf.blockAckVariant = "basic" at both endpoints. Existing Basic demonstrations retain their explicit selectors and result rows.

For an active Compressed agreement, the originator suppresses fragmentation for that peer and TID. Normal Ack DATA within that agreement has the same restriction. The recipient rejects fragmented DATA before receipt or reorder updates. Basic retains fragment support. Outside a Compressed agreement, the existing fragmentation threshold applies.

Compressed supports immediate policy only. HCF rejects enabled delayedAckPolicySupported in either agreement policy at initialization, before ADDBA. Basic retains its existing configuration contract; that retention does not establish successful delayed Block Ack support.

How the exchange resolves DATA

The following hypothetical example starts with compatible profiles and an accepted agreement. DATA 101 is lost. The recipient retains DATA 102 until retransmission fills the gap. The example assumes no other gap-release condition applies.

sequenceDiagram
    participant O as Originator HCF
    participant R as Recipient HCF
    participant U as Recipient upper protocol
    O->>R: ADDBA Request for peer and TID
    R->>O: Accept immediate agreement
    O->>R: DATA 100 with Block Ack policy
    R->>U: Deliver 100
    O--xR: DATA 101 is lost
    O->>R: DATA 102 with Block Ack policy
    Note over R: Retain 102 behind the gap
    O->>R: Compressed BAR with start 100
    Note over R: Wait SIFS
    R->>O: Compressed Block Ack with bits 1, 0, 1
    Note over O: Resolve 100 and 102. Retain 101 for retry.
    O->>R: Retransmit DATA 101 before its deadline
    R->>U: Deliver 101, then 102

Loading

The response start can differ from the BAR start. The originator maps bits from the response start with modulo-4096 sequence arithmetic. Positive bits resolve only transmitted, unacknowledged frames inside the active transmit window. They cannot acknowledge unsent DATA. Zero bits leave frames unacknowledged.

A timeout or disjoint response releases the BAR wait so requested DATA can become retry eligible. The existing absolute DATA deadlines still apply.

The receive step checks profile, addresses, TID, Fragment Number zero, 64 bitmap bits, and supported length before success callbacks. Malformed responses cannot acknowledge DATA, refresh inactivity, or report success. A valid addressed BAR with the other profile causes a fatal configuration diagnostic when an agreement exists. HCF stops before state updates and sends no conflict-triggered DELBA. Users must correct endpoint profiles before a new run. Detection can follow earlier DATA delivery because ADDBA does not negotiate the profile.

Why receipt state survives delivery

The receipt record answers whether DATA arrived. The reorder buffer controls delivery order. Delivery must preserve receipt evidence for a later BAR. Scoreboard movement controls retirement of those bits. An absent receipt produces zero, including positions before the first received DATA.

flowchart TD
    D["Valid DATA for the active peer and TID"] --> S["Receipt record<br/>Record complete-unit reception"]
    S --> R["Reorder buffer<br/>Retain DATA until a release rule applies"]
    R --> U["Deliver eligible units in cyclic order"]
    U --> K["Preserve receipt bits<br/>until scoreboard retirement"]
    B["Valid Compressed BAR"] --> W["Adjust scoreboard window"]
    K --> W
    W --> A["Build 64 response bits<br/>Use zero for absent receipts"]

Loading

The recipient integration test receives DATA 100, 102, and 101. It delivers 100, 101, and 102 in order. A subsequent BAR still receives three positive bits and zero holes. Basic retains its separate fragment bitmap.

How physical timing permits gap release

The supported individual-MPDU release rule can permit delivery across a gap. It requires positive receipt evidence immediately before the scoreboard start. The DATA sequence must also lie inside the reorder window and above its start. The required SIFS or permitted RIFS relation must exist. RIFS means reduced interframe space. Without either receipt evidence or that relation, this rule permits no release.

Local receive timestamps alone cannot establish the interval when propagation delay exists. The medium supplies copied transmission and cached arrival endpoints. The PHY derives the interval at the current transmitter. HCF associates those facts with reception and ACK completion, without retained pointers to expired physical objects.

flowchart TD
    M["Radio medium<br/>Copy transmission and arrival endpoints"] --> P
    P["PHY<br/>Derive interval at current transmitter"] --> H
    E["Reception and ACK completion events"] --> H
    H["HCF<br/>Match event identity<br/>Validate SIFS or RIFS history"] --> C
    S["Receipt record<br/>Supply prior receipt evidence"] --> C
    C{"Do all release conditions hold?"}
    C -->|Yes| D["Reorder service releases eligible units<br/>Advance delivery window"]
    C -->|No| K["This rule permits no gap release"]

Loading

Unrelated, failed, abandoned, or overlapping receptions must not preserve stale eligibility. Local transmission clears prior history; completed ACK transmission can establish new history. HT management preserves advertised RIFS permission. VHT clears that permission. This PR adds no RIFS transmission scheduler.

The PHY also includes the six-microsecond HT signal extension at 2.4 GHz. The authoritative duration includes that extension, so HCF must not add it again. This correction can change airtime and simulation results.

How TXOP admission preserves progress

The initial Compressed candidate exposed a new supported continuation overrun. In General run 6, later A-MSDUs exceeded a one-millisecond TXOP by 310 microseconds. The exact S4 parent did not reproduce that case. The exchange selector checked positive time left without proof that the complete exchange fit.

A final transmit check alone could refuse an aggregate after source removal. The data service now creates a duplicate preview while original MSDUs remain queued. HCF selects one DATA mode and evaluates complete exchange cost. If the aggregate fails admission, HCF evaluates its first ordinary MSDU with the same mode.

flowchart TD
    Q["Original MSDUs remain queued"] --> A["Create aggregate preview<br/>Select one DATA mode"]
    A --> F{"Does the aggregate qualify?"}
    F -->|No| O["Evaluate ordinary MSDU preview<br/>Use the same DATA mode"]
    F -->|Yes| P["Retain preview and modes"]
    O --> G{"Does the ordinary form qualify?"}
    G -->|No| D["End sequence without DATA failure<br/>Preserve queue, deadlines, and sequence state"]
    G -->|Yes| P
    P --> R["Perform RTS/CTS if required"]
    R -->|RTS fails| X["Apply RTS recovery<br/>Discard unused preview"]
    R -->|No RTS or CTS succeeds| V{"Are sources and modes still valid?"}
    V -->|No| Z["Cancel preparation<br/>Preserve valid original sources"]
    V -->|Yes| C["Commit the same preview<br/>Submit DATA with retained mode"]

Loading

Commitment removes sources and assigns sequence state immediately before first DATA submission. Refusal preserves original queue frames, deadlines, and sequence state. One selected mode survives RTS/CTS. Failed RTS before commitment updates access-category recovery without a provisional sequence counter. Original lifetimes bound repeated attempts. Expiry or changed capabilities cancel invalid preparation.

Ordinary commitment uses normal dequeue or selected extraction and copies completed region metadata into the retained preview before source deletion. For example, an MSDU that waits 2 ms retains that queueing time and one departure after commitment. Retransmission retains the first measurement without another pull. Aggregation retains its existing source-removal behavior.

The corrected path retains permitted initial ordinary-MSDU and same-size retransmission exceptions. An oversized first-attempt A-MSDU cannot use the retransmission exception. Single protection reserves only the current exchange; a later BAR can require fresh contention. Basic and traffic outside active Compressed agreements retain their existing admission limits. A strict timing fixture preserves an inherited failure outside this scope.

Dependencies and complete commit order

This PR depends on #1343 and #1344. The comparison starts at master commit 675655fc3559231c56b12d651384cb8695255abc and ends at 1dd4ca23785b14b058693e9e20434a4257c3831f. The first seven commits are the exact synchronized #1344 prefix. The next ten supply S5. All 17 commits form the dependency prefix of #1363.

The shared queue contract permits selected service behind deferred DATA with normal pull effects. Agreement handlers and HCF coordinate setup, retirement, and safe physical completion. S4 supplies Basic BAR completion and shared window operations. S5 adds copied physical timing, wire formats, reversible preparation, and complete explicit Compressed exchanges.

1. 435fc0866d adds selected extraction with ordinary queue effects; it prepares the lifecycle consumer.
2. 2c6b1df761 connects Action delivery, setup, retirement, reset, and deferred DATA selection.
3. 832dc3672c gives the frame-sequence handler ownership of interruption outcomes.
4. 0c23400688 records S3b completion and its verification scope.
5. 23c1939f27 exposes DATA preview without extraction or sequence assignment.
6. 2c454a83d1 completes Basic BAR selection, retry feedback, reservation, and final transfer resolution.
7. 728d3dd911 gives the originator agreement ownership of window operations.
8. 1f3edba0c6 adds copied physical timing facts to the radio medium and communication caches.
9. ef9ecba78c includes the six-microsecond HT signal extension at 2.4 GHz.
10. 2c329da06c preserves advertised RIFS permission through HT management conversion.
11. b7fb7a52ff releases temporary Basic bitmap allocations before the wire-format change.
12. 96be3ee415 serializes complete Basic and Compressed Block Ack Types.
13. ca60df56ee adds reversible QoS DATA preparation and preserves ordinary queue effects at commitment.
14. 1f7b8f76ea adds Compressed receipt state without runtime activation.
15. 5b87231f41 adds legal Compressed control-rate selection without runtime activation.
16. 126c63f561 connects complete explicit immediate Compressed exchanges and their production fixtures.
17. 1dd4ca2378 selects Compressed by default after the complete runtime path exists.

Receipt and rate helpers precede activation. HCF still rejects Compressed in those two prerequisite commits. The activation commit connects agreement state, receipt state, physical intervals, response timing, and admission because a complete exchange needs those contracts together.

Architecture and migration

HCF owns the immutable profile and local interframe history. Agreement records own receipt state. The reorder service owns delivery order. The rate selector consumes committed MIB rates and distinguishes initial BAR from later BAR. Compressed BAR uses legal non-HT rates. Protection, timeouts, and Duration fields use the actual response profile and airtime.

The serializer independently decodes the complete four-bit wire Type. Type 0 retains Basic compatibility; Type 2 selects Compressed. Unsupported types follow the invalid-frame path. The Compressed header chunk contains 28 bytes; its trailer supplies the four-byte FCS. Compressed reserved bits write zero, and received reserved values do not change the result. Basic producers must keep overlapping Type bits zero.

Custom radio media and communication caches must supply copied timing facts without retained object pointers or cache mutation. Custom IEEE 802.11 modes must expose signal extension. Custom QoS rate selectors, recipient ACK policies, data services, and frame sequences must preserve the new exchange and preparation contracts. The migration guide gives the API obligations. The user guide and WHATSNEW explain configuration effects.

The replacement plan defines S5. The TXOP correction defines preparation and admission. The execution record gives detailed cases, commands, results, and limits.

Verification and limits

Fresh debug and release builds pass at 1dd4ca23785b14b058693e9e20434a4257c3831f with all 12 available CPUs. The debug campaign covers every changed test fixture in the master comparison. All selected fixtures report their expected successful result.

Check	Result
Debug and release compilation	Both PASS.
Module fixtures	33 PASS.
Protocol fixtures	4 PASS.
Serializer fixtures	1 PASS under its declared coverage and skip rules.
Unit fixtures	9 PASS.
Functional fingerprint cases	12 PASS, with unchanged expectations.
Series and description gates	All five PASS.
Source seals and scoped MAC architecture	PASS.
Global architecture, naming, and interface gates	The same inherited findings remain. These gates retain exit code 1.

The global comparison checks the finding text. Counts of unchanged interfaces and examined declarations can differ between stages. These counts do not introduce new findings.

Executed commands and selectors

The build commands run from this PR's checkout root:

make MODE=debug -j12
make MODE=release -j12
opp_repl --load /home/user/omnetpp-6.4.0aipre2/.venv/lib/python3.14/site-packages/opp_repl/opp/inet.opp -p inet

The opp_repl adapter selects this checkout's inet_project. It calls the checked-out test functions with these exact arguments. It rejects empty selections and unexpected results.

run_module_tests(simulation_project=inet_project, mode="debug", filter=r"(?:Ieee80211BlockAckActionDelivery_1\.test|Ieee80211BlockAckActionDispatch_1\.test|Ieee80211BlockAckActionDisposal_1\.test|Ieee80211BlockAckAdmission_1\.test|Ieee80211BlockAckAssociationReset_1\.test|Ieee80211BlockAckCompletionCost_1\.test|Ieee80211BlockAckDataSelection_1\.test|Ieee80211BlockAckDeferral_1\.test|Ieee80211BlockAckDeferredCollision_1\.test|Ieee80211BlockAckFragmentRetirement_1\.test|Ieee80211BlockAckLifecycle_1\.test|Ieee80211BlockAckNegotiation_1\.test|Ieee80211BlockAckReordering_1\.test|Ieee80211BlockAckResponseWait_1\.test|Ieee80211BlockAckRetiredData_1\.test|Ieee80211BlockAckScheduling_1\.test|Ieee80211BlockAckVariant_1\.test|Ieee80211CompressedBlockAckAdmission_1\.test|Ieee80211CompressedBlockAckExchange_1\.test|Ieee80211CompressedBlockAckInput_1\.test|Ieee80211CompressedBlockAckInterframe_1\.test|Ieee80211CompressedBlockAckMigration_1\.test|Ieee80211CompressedBlockAckOriginator_1\.test|Ieee80211HcfHeader_1\.test|Ieee80211HcfMultipleProtection_1\.test|Ieee80211HcfMultipleProtection_2\.test|Ieee80211HcfSingleProtection_1\.test|Ieee80211PreparedQosData_1\.test|Ieee80211RecipientNav_1\.test|Ieee80211SingleProtectionGroup_1\.test|Ieee80211Txnav_1\.test|Ieee80211VhtHtSupport_1\.test|PacketSelectiveExtraction_1\.test)$", concurrent=False)
run_protocol_tests(simulation_project=inet_project, mode="debug", filter=r"(?:Ac_BlockAck\.test|Ac_BlockAckFallback\.test|N_BlockAck\.test|N_CompressedBa\.test)$", concurrent=False)
run_serializer_tests(simulation_project=inet_project, mode="debug", filter=r"(?:serializer_chunk_roundtrip\.test)$", concurrent=False)
run_unit_tests(simulation_project=inet_project, mode="debug", filter=r"(?:Ieee80211BlockAckScheduling_1\.test|Ieee80211BlockAckWire_1\.test|Ieee80211CompressedBlockAckRate_1\.test|Ieee80211CompressedBlockAckRecord_1\.test|Ieee80211HtCapabilities_1\.test|Ieee80211HtModeSet_1\.test|Ieee80211MultipleProtection_1\.test|Ieee80211ResponseTimeout_1\.test|TransmissionTimingSnapshot_1\.test)$", concurrent=False)

The fingerprint command runs from tests/fingerprint:

./fingerprinttest -d -F tyf -t 3 -m 'showcases/wireless/(aggregation|qos|blockack)/|showcases/wireless/fragmentation/.*-c HCFfragblockack |showcases/wireless/txop/|examples/adhoc/qos/.*-c MacQos -r 1|examples/wireless/qos/.*-c MacQosWithBlockAck ' showcases.csv examples.csv

The mechanical checks use the pinned master base 675655fc3559231c56b12d651384cb8695255abc. They cover the complete series and description, source seals, scoped MAC architecture, and the three global gates. This campaign changes no recorded expectation.

All 17 commits retain their scoped debug results from the #1363 reconstruction. Compared with the former head, dedcdffb3696f0402eb4c1e9ba90e448e02cefe5, this prefix adds the already validated ordinary DATA commitment repair. Its four paths are the data service, two production fixtures, and WHATSNEW. The repair preserves normal pull effects and completed region metadata. It changes no wire bytes or baseline value.

S5 introduces no new fingerprint expectation. The seven lifecycle updates remain with 2c6b1df761; the seven S4 updates remain with 2c454a83d1. This synchronization retains their exact approved values. The execution records retain the causal evidence and original approval scope.

Evidence covers the selected deterministic cases and their configured seeds. Recipient prefix release retains the recorded interpretation of individual-MPDU release conditions. Production HT catalogs supply no distinct legal alternate response for the tested 32-byte frame. A selector unit checks that contract separately.

This PR excludes A-MPDU transfer, implicit BAR, delayed Compressed Block Ack, Multi-TID Block Ack, HE bitmap lengths, and dynamic fragmentation. Basic preserves historical compatibility without a complete historical IEEE conformance claim. The supported ac exchanges do not establish complete VHT transfer support. This PR adds no MAC stop/start recovery.

Release behavioral tests, graphical tyf fingerprints, and feature-isolation builds remain unverified. This synchronization supplies no new independent correctness audit. Fingerprint matches do not replace direct exchange or timing evidence.