Skip to content

ND Ethernet II emulator - RE session handoff (2026-07-08)

Target file (all C# edits): E:\Dev\Repos\Ronny\RetroCore\Emulated.HW\ND\CPU\NDBUS\NDBusEthernetII.cs

Goal: make the emulated ND Ethernet II controller (ND-110063, PCB 3094) work so SINTRAN START-NETWORK-SERVER ENNS0 completes and the COSMOS Ethernet server registers with XROUT. Current live failure: Error in communicating with XROUT ... Unknown name (of server or system).

User's standing directive (ABSOLUTE): the bug is in the C# Ethernet controller, NOT in XROUT/XMSG. Do NOT conclude "it's XROUT / the card is done". Keep hunting in the controller.

Oracles (use BOTH on every change)

  1. Regression oracle: TPE ETHERNET-TWO diagnostic (TPE> run, tests 1-33). Committed baseline: tests 1-11 PASS, 12 Timeout/no interrupt from MC68000, 22 varies, 24/25/26 FAIL, 33 DMA. Do NOT regress 1-11 (esp. test 11 = OPCOM/ND-calling-int).
  2. Goal oracle: RT ENNS0 then START-NETWORK-SERVER ENNS0 with debugger DebugTrace 2 4. Delete %LOCALAPPDATA%\trace\file-trace.txt before each capture (append-mode).

Full validated plan: ...\stripped\docs\ND_EthernetII_Fix_Plan.md. Reviews: ...\stripped\docs\ND_EthernetII_Emulator_Correctness_Analysis.md (code/spec/firmware) and D:\OCR\ai\ND-324534-G1-EN\NDBusEthernetII-Schematic-Review.md (schematic).

What is CONFIRMED CORRECT this session (do not touch)

  • Register map: status read at base+0/+2, control write base+1/+3; NO data register (0xFF stub removed).
  • SCIP doorbell: 0xEF0080-9F write -> ND-100 INT12.
  • MC68000 STOP wakes on IRQ (&= ~WAIT bug fixed) - see [[mc68k-stop-wait-bug]].
  • LANCE SwapByteLanes + big-endian DmaIn/DmaOut - LANCE now fully inits (IDON, MAC 08:00:26:64:00:00 = valid ND OUI, machine 100 - NOT a placeholder).
  • C-4 (committed): MFP timer clock 3.6864 -> 3.125 MHz (3CLK = 12.5MHz/4).
  • C-5 (committed): power-low (control bit 6) -> 68K level 7 rising-edge; LANRESET (0xEF00B0-B7) -> lance.Reset() via new OnLANReset action.
  • DMA touches (committed): DmaIn/DmaOut mask addr &= 0xFFFFFF + bounds-check -> lance.SimulateMemoryError() on overrun.

REVERTED this session (were my mistakes)

  • STC block (TPE test 12): the F138 decoder ignores addr bit 8, so EF01xx==EF00xx (EF0140=MERRSTAT, EF0160=EAREN, EF01A0=LANCE RDP). Test 12 FAILS on real ND-110 hardware. I had tuned the emulator to pass a test the real card fails - reverted. STC kept only as a harmless off-by-default diagnostic shim.
  • C-1 (IDENT keeps interruptEnabled): broadly regressed TPE 5/6/7/8/11 ("Unexpected error in KICK FROM PIOC"). Reverted; IDENT still does interruptEnabled = false; with a NOTE. The reviews say A1/C-1 must be done TOGETHER with C-2/C-3 holistically, not alone.

CURRENT LEAD - A4 (VALIDATED, READY TO COMMIT)

Change: control-word bit-2 handler (~line 1236 in NDBusEthernetII.cs). Replaced the broken rising-edge memoryMap.SetMFPInterrupt() with a proper LEVEL drive of MFP GPIP I6:

// A4: control bit 2 (ND interrupt) is a LEVEL on MFP GPIP I6 (active-low), NOT a one-shot.
// Old rising-edge SetMFPInterrupt() drove I6 LOW and never raised it -> pin stuck.
memoryMap.MFP.GPIO_6 = !ND_interrupt;   // bit2=1 -> I6 LOW (active); MFP gpio_input does AER+IER edges
if (ND_interrupt && !previousND_interrupt)
    Logger.Log("ND-100 -> 68000: GPIP I6 asserted (ND interrupt, vector 0x4E)", Logger.LogLevel.Device);

Rationale: ND-100's post-boot commands to the firmware go through GPIP I6 into nd_host_interrupt_handler (vector 0x4E, 68K level 3). The old code left I6 stuck LOW forever, so the firmware's GPDR polling never saw it deassert -> after the first OPCOM wake, subsequent ND-100 commands produced no fresh edge. This matches the observed stall symptom: "ND-100 responds/IDENTs once, firmware then floods unanswered SCIP."

TPE VALIDATION (2026-07-22): Tests 1-11 ALL PASS ✓ (NO REGRESSION on test 11). Tests 12/24/25/26 fail as expected (baseline failures, not A4-related).

STATUS: Commit A4 to NDBusEthernetII.cs immediately.

DO NOT TOUCH YET: the separate ND-interrupt path at ~line 848 (case 116 vector handler, does GPIO_6=false; MFP.TriggerInterrupt(6)). Leave until A4 is committed and ENNS0 trace is analyzed.

NEXT STEPS (Session 2026-07-22)

  1. AFTER WINDOWS REBOOT: Commit A4 to NDBusEthernetII.cs (no changes, tests validated). 
    // A4: control bit 2 (ND interrupt) is a LEVEL on MFP GPIP I6 (active-low)
memoryMap.MFP.GPIO_6 = !ND_interrupt;
if (ND_interrupt && !previousND_interrupt)
    Logger.Log("ND-100 -> 68000: GPIP I6 asserted (ND interrupt, vector 0x4E)", Logger.LogLevel.Device);
  2. Rebuild Emulated.HW, relaunch HOST.
  3. Boot menu option 8 "SINTRAN Ronny Ethernet II Test" added to ND100Script.ini.
    • Uses ND100-ETH config (Ethernet II + DMA Floppy + SMD).
    • Boot label: ND-BOOT-ETHII-SINTRAN.
    • Boot disk: D:\BIGDISK0-L.IMG (replaced with F:\RC\RonnyTest\HDLC1\BIGDISK0-L2-100.IMG, original backed up as .BACKUP).
  4. Run RT ENNS0 + START-NETWORK-SERVER ENNS0 with DebugTrace 2 4. Delete trace file first. Key question: does firmware process MORE than ONE ND-100 command? If yes -> handshake advanced.
  5. If A4 doesn't advance ENNS0: continue in controller. Candidates:
    • case-116 line-848 duplicate ND-interrupt path (reconcile with A4).
    • 68K<->ND-100 monitor postbox handshake (postbox ~0x40A: D0 ready signal).

Constraints (from user, non-negotiable)

  • NEVER mention Claude in commits. One change at a time. Both oracles each step.
  • Never kill/stop/terminate any process without explicit permission - ASK Ronny to relaunch the host for a fresh build.
  • No LINQ/foreach/FluentAssertions; keep all comments; ASCII only; full file paths.

[[ethii-emulator-fixes]], [[mc68k-stop-wait-bug]], [[encos-enns0-analysis]], [[enns0-monitor-and-xrout-protocol]], [[subtype-07-network-error]].