HANDOFF — ND 3112 floppy-DMA controller (C# + C)

Date: 2026-07-13. Status: all 10 floppy-DMA divergences fixed in both models; PIO fixes pending.

What was done

Docs (this folder NDInsight\SINTRAN\Devices\FloppyDMA\): full 11-file spec set for the 3106/3112 written from the ND-11.021.1 manual + reverse-engineered Z80 firmware 34300G.bin (overview, programming interface, commands, boot/autoload, formats, error codes, firmware internals, streamer, testing, implementation guide, README).

Ghidra project RetroGhidra\N100-FLOPPY-3112\ND-FLOPPY-3112.gpr: fully annotated — all FUN_ram_* renamed, all DAT_ram_* + I/O ports labelled, plate/EOL comments. Key anchors: HostCmd_ISR @030c, Autoload_BootstrapLoad @1ae8, status builder @06b3, RST 08h error table @1e8d, error-image construction @1f2f-1f6f, images @1a92 (LOAD-ERROR) / @1acc (WRONG-BOOTSTRAP).

RetroCore C# Emulated.HW\ND\CPU\NDBUS\NDBusFloppyDMA.cs (+ Emulated.Machines\...\ND100Machine.cs): - Autoload (ExecuteAutoload) implemented — reads floppy, parses BPUN via machine delegate OnAutoloadRequest → LoadFloppyBpunIntoMemory, DMAs to memory, completes. - Boot error image: DmaAutoloadErrorImage DMAs the byte-exact ROM LOAD-ERROR / WRONG-BOOTSTRAP image to the first page so the console prints ** LOAD-ERROR: nn ** on boot failure. - Write-protect enforced (WRITE_PROTECTED=14, error 16 oct, no write). - IDENTIFY + unknown commands now complete (queue ReadEnd) — no hang. - Bit-4 OR-of-errors set when errorCode != 0. - Enum gained WRITE_PROTECTED/FORMAT_NOT_FOUND/RAM_ERROR. - Two-status-word split: IOX +2/+4 → CalculateHardwareStatusWord (bit-15 dual-density, no code); CB+6 → CalculateStatusWord1 (code bits 9-14, bit-15 clear). - Tests Emulated.Tests.ND100\ControllerFloppyDMA\TestNDBusFloppyDMA.cs (26; full ND100 251/0).

nd100x C /home/ronny/repos/nd100x/src/devices/floppy/deviceFloppyDMA.{c,h}: - Error-image DMA on autoload failure (DmaAutoloadErrorImage, byte-exact) — but C still lacks a real BPUN autoload, so its autoload always shows LOAD-ERROR 50 (TODO in code). - FIX 1: error code .h bitfield moved to bits 9-14. FIX 2: IOX +4 → hardware word. FIX 3: split CalculateHardwareStatusWord / CalculateStatusWord1. Bit-4 includes errorCode!=0. Builds clean (cmake --build build_linux --target devices).

Not verified

End-to-end "console actually prints LOAD-ERROR" (no ND-100+console integration test). Image bytes verified vs ROM hexdump; DMA verified by unit test. Entry point word 0 inferred.

Pending — PIO (3027) controller, another day

7 tracked tasks from HANDOFF-floppy-pio-c-and-csharp-fixes.md (this is a DIFFERENT device): C-PIO-1 (C read-only image/dead WP guard), C-PIO-2 (FORMAT_TRACK off-by-one), CS-PIO-1 (C# Debug.Assert(false) landmines), SH-PIO-1/2/3 (control-reset RFT, device-clear+cmd ordering, sector auto-increment boundary — both models), cosmetics + SH-PIO-4 (autoload bit-2 unverifiable). Validate with the FLOPPY-FU-1986F test program. Spec: §B.4 in NDInsight\Reference-Manuals\ND-06.015.02 ND-100 Functional Description.md (~9747-9990).

Pending — analyse floppy driver vs FLOPPY-STREAM C03 (task #8, another day)

PRE-EXISTING (confirmed: reproduces on a pristine nd100x build, NOT caused by the fixes above). Running the TPE test program FLOPPY-STREAM Version C03 - 1988-11-08 (--boot=floppy --image=FLOPPY.IMG, then fl-s, sel-de unit 0) fails at SELECT-DEVICE with: ***** ERROR : Interrupt expected but not detected after transfer. The C floppy-DMA model (and likely C#) doesn't deliver the completion interrupt the test expects after a transfer. For another day: dig in with DEBUG_FLOPPY_DMA/DEBUG_DETAIL logging + DAP tracing of the ND-100 side — which command it issues, whether ExecuteFloppyGo→ReadEnd runs, why interruptEnabled && readyForTransfer doesn't fire when expected. (Test self-reports the card as "3027 Floppy DMA controller, Micro program B".)

To restore my C work if a session starts fresh: it's uncommitted in the WSL nd100x repo (git stash list / working tree src/devices/floppy/deviceFloppyDMA.{c,h}); rebuild with cmake --build build_linux --target nd100x.

Follow-ups (not blockers)

Status Word 2 §3.5.2.2 bit remap; C real BPUN autoload; error-51 trigger (BPUN parser can't distinguish 50 vs 51, so boot failures report 50).