The goal of this project is to provide an out-of-the-box CP/M-86 1.1 kernel (BIOS, BDOS and CCP) that incorporates all existing patches plus the capacity to run on modern hardware and virtualization. A distribution of the CP/M-86 OS is also provided fully patched as well as progressive reverse engineering/forensics of each of the basic commands. (tod, ed, help, submit, pip, asm86, ddt86, gencmd, stat have been reconstituted from sources of CP/M-80, CCP/M-86 whenever available). All the remaining tools will go through disassembly and a format command for disk and hard drive is also planned at a laterstage
The distribution also packages digital research assembler tools and various basic environments.
- visual y2k support and tod replacement (https://github.com/tsupplis/cpm86-hacking). The code is available in this project.
- AT support
- resilience to bios limits for video display
This is a raw dump right now with a way to compile the kernel starting from dissassembled sources. The compilation requires the (cross-development environment for CP/M-86)[https://github.com/tsupplis/cpm86-crossdev]
The CP/M-86 kernels (8088 and V20 Mixed 8080/8088 with CP/M-80 compatibility) for the V20 MBC are also produced. The sources of the bioses used are available at:
The 3 kernels use the same BDOS and CCP components.
To test, the PCE emulator and cpmtools are needed. PCE can be found at http://www.hampa.ch/pce/. the floppy images released also work with qemu.
Alternatively, you can use the excellent V20 MBC available at:
- https://hackaday.io/project/170924-v20-mbc-a-v20-8088-8080-cpu-homebrew-computer
- https://shop.mcjohn.it/en/diy-kit
- Patched CP/M-86 running in the PCE Emulator
- CP/M-86 BIOS Setup 1.2
- CP/M-86 Disk Maintenance 1.2
The repository holds three separate bodies of work:
| What | Where | |
|---|---|---|
| 1 | Three parallel source sets of the PC CP/M-86 1.1 kernel | repository root |
| 2 | The original tool binaries, patched and updated | base/, dev/, extra/ |
| 3 | Reconstruction of the tools from source | commands/ |
The kernel was recovered by disassembly, so the sources are kept in three
parallel sets that can be diffed and rebuilt against each other. ccp and
bdos are shared by all PC and V20 kernels; only the BIOS differs.
| Set | Sources | Kernel | Intent |
|---|---|---|---|
*org.* |
ccporg.a86 bdosorg.a86 pcbioorg.a86 |
cpmorg.sys |
Original patched release. Frozen reference, never edited. |
*.* |
ccp.a86 bdos.a86 pcbios.a86 |
cpm.sys |
Byte-identical to *org.*; labelling, commenting and documentation in progress. |
*exp.* |
ccpexp.a86 bdosexp.a86 pcbioexp.a86 |
cpmexp.sys |
Experimental cleaned-up and optimised variant. |
make check asserts that cpm.sys and cpmorg.sys are binary-identical, so
the annotation work cannot silently change behaviour.
Two further BIOSes target the V20 MBC:
| Kernel | Target |
|---|---|
cpmv20.sys |
MBC V20, 8088 mode (mbcv20.a86) |
cpm816.sys |
MBC V20, mixed 8080/8088 mode with CP/M-80 compatibility (mbc816.a86) |
Four sets are produced. The 160K set is the reference layout; the others repack
the same material to fit the media, except the experimental set which swaps in
the commands/ reconstruction.
160K single sided — 4 disks
| Image | Contents |
|---|---|
cpm86-160-1.img |
Bootable core CP/M-86 |
cpm86-160-1-at.img |
Bootable core, AT-compatible clock |
cpm86-160-2.img |
Assembler CP/M-86 tools |
cpm86-160-3.img |
Digital Research dev tools |
cpm86-160-4.img |
BASIC development |
320K double sided — 3 disks
| Image | Contents |
|---|---|
cpm86-320.img |
Bootable core + assembler tools |
cpm86-320-at.img |
Bootable core + assembler tools, AT-compatible clock |
cpm86-320-dev.img |
Digital Research dev tools + BASIC development |
720K / 1.44M via the "1.44 MB Feature" — 2 disks
| Image | Contents |
|---|---|
cpm86-720-at.img |
Everything above, AT-compatible clock |
cpm86-1440-at.img |
Same content on 1.44M media |
Both use Freek Heite's 1.44 MB Feature, rebuilt from source under extra/.
They carry identical content — the whole base/ command set plus the dev/
toolchain — and are produced by the same Makefile recipe, parameterised only by
geometry. 720K leaves about 100 of its 355 blocks free.
Experimental — 4 disks
| Image | Contents |
|---|---|
cpm86-exp-160-1.img |
Experimental kernel with the commands/ reconstruction in place of base/ |
cpm86-exp-160-1-at.img |
Same, AT-compatible clock |
cpm86-exp-720-at.img |
Experimental kernel plus the full reconstruction and the dev toolchain |
cpm86-exp-1440-at.img |
Same content on 1.44M media |
The experimental images boot cpmexp.sys and carry the rebuilt tools rather
than the base/ binaries, so a single boot exercises both halves of the work.
The 160K pair ships ed, help, pip, stat, submit and tod; the feature
images add asm86, ddt86 and gencmd for the complete set.
They also carry their own build of the feature: cpmexp.sys moves the CCP and
BDOS stack switches that the loader patches at boot, so the stock 144BLDR2
would fail its byte checks against it. See the expkrnl block in
extra/144pat2.asm.
PCE helper scripts exist per variant — ./cpm86, ./cpm86-320, ./cpm86-720,
./cpm86-1440, ./cpm86-exp, ./cpm86-exp-720, ./cpm86-exp-1440. make test, make test-320 and make test-exp are shortcuts for the common ones.
The 160K images carry a boot loader terminating with 55AA, which lets qemu
load CP/M-86 properly. Beware: formatting with dskmaint.cmd does not add the
signature.
The other formats cannot carry it, because byte 01FFh of the boot sector is
already a format byte and AAh would collide with it:
| Format | 01FFh |
Meaning |
|---|---|---|
| 160K | AAh |
free — second half of the 55AA signature |
| 320K | 01h |
selects 2K allocation blocks (DPBK2); any other value gives 1K (DPBK1) |
| 720K / 1.44M | 48h / 90h |
cpmedia media byte written by the feature's boot sector |
The 320K test lives in the second-stage loader, which reads the boot sector
still resident at 0000:7C00:
mov ax,0x400 ; default allocation block = 1024 (160K)
cmp byte [es:0x7dff],0x1 ; boot sector byte 01FFh == 1 ?
jnz keep
mov ax,0x800 ; yes -> allocation block = 2048 (320K)Stamping 55AA on a 320K image therefore makes the loader read a 2K-block
filesystem as if it used 1K blocks; it finds E5 fill instead of CPM.SYS and
hangs after printing one dot.
base/ holds the CP/M-86 command set. Some tools ship untouched, others are
patched or replaced.
| State | Tools |
|---|---|
| Original | asm86 assign config ddt86 function gencmd help print stat |
| Updated release | dskmaint (1.0 → 1.2), setup (1.0 → 1.2), hdmaint (1.0 → 1.1), help.hlp (fuller content) |
| Patched per DR recommendation | ed gendef pip submit |
| Locally patched | mform (interactive prompt removed) |
| Added | atinit (RTC sync), from cpm86-hacking |
tod is deliberately absent from base/: the original is superseded by the
CP/M-86-native rewrite, so every image takes it from commands/tod.
extra/ holds Freek Heite's "1.44 MB Feature", which adds 720K, 1.2M and 1.44M
diskette support to CP/M-86 1.1. It is built from source by make -C extra
using MASM 5.10 and LINK 5.13 under DOS emulation, replacing the original
144make2.bat:
| Output | Role |
|---|---|
144bldr2.cmd |
Secondary loader; patches the kernel in memory at boot |
144pat2.cmd |
Applies the same patches to an already-running system |
144prep2.cmd |
Prepares a diskette (boot sector + loader appended) |
144patx.cmd 144prepx.cmd 144bldrx.cmd |
The expkrnl builds, for cpmexp.sys |
The rebuilt 144pat2.cmd is byte-identical to the binary originally shipped
here, so the toolchain reproduces upstream exactly.
dev/ holds the Digital Research and Microsoft toolchains, as shipped and
unmodified:
| Tool | Product | Version | Date | Vendor |
|---|---|---|---|---|
rasm86.cmd |
RASM-86 Relocating Assembler | 1.2 | 1982-1983 | Digital Research |
link86.cmd |
LINK86 Linkage Editor | 2.02 | 02/Feb/87 | Digital Research |
lib86.cmd |
LIB-86 Library Manager (CDOS 86) | 1.3 | 2/23/86 | Digital Research |
xref86.cmd |
XREF-86 Cross Reference | 1.1 | 1982-1983 | Digital Research |
sid86.cmd |
SID-86 with 286 Disassembler | 2.4 | 07 May 1985 | Digital Research |
cbas86.cmd |
CBAS86 CBASIC Compiler | 1.4 | 1981-1983 | Digital Research |
crun86.cmd |
CRUN86 CBASIC Runtime | 1.4 | 1981-1983 | Digital Research |
pbasic.cmd |
Personal BASIC | 1.2 | 1983-1985 | Digital Research |
mbasic86.cmd |
BASIC-86 (CP/M-86 patched) | Rev. 5.22 | 5-Mar-82 | Microsoft |
commands/ rebuilds the tools from the surviving PL/M and ASM86 sources. Those
sources are Concurrent CP/M-86 3.1 vintage and assume BDOS 3.x, while this
kernel reports BDOS 2.2 with BH=0. Each tool was therefore audited for version
guards and for calls into BDOS functions this kernel does not implement.
| Tool | Outcome |
|---|---|
asm86 gencmd |
CCP/M-86 3.1 sources. Compatible with CP/M-86 |
ddt86 ed help |
CCP/M-86 3.1 sources. Compatible with CP/M-86 |
pip |
Reconstructed from CCP/M-86 3.1 sources |
stat |
Reconstructed from CP/M-80 2.2 sources |
submit |
Reconstructed from CP/M-80 2.2 sources |
tod |
Superseded by the CP/M-86-native rewrite; the only tool with no base/ binary |
make commands builds the whole set. The root build depends on it, since tod
is needed by every image and the rest by the experimental ones.
cpmtools 2.23 with libdsk is used.
cpmtools can be deployed with homebrew on mac of fetched at https://www.moria.de/~michael/cpmtools/.
All five formats live in the diskdefs file at the root of this repository.
cpmtools reads ./diskdefs in preference to the system-wide file and does
not merge the two, so run the cpm* tools from the repo root and expect
every format used here to be listed there.
| Format | Geometry | Block | Dir | Reserved |
|---|---|---|---|---|
ibmpc-514ss |
40 × 8 × 512 (160K SS) | 1K | 64 | 1 track |
cpm86-320 |
80 × 8 × 512 (320K DS) | 2K | 64 | 1 track |
cpm86-720feat |
160 × 9 × 512 (720K) | 2K | 256 | 2 tracks |
cpm86-144feat |
160 × 18 × 512 (1.44M) | 4K | 256 | 2 tracks |
cpm86-xt-hd |
272 × 63 × 512 (8M partition) | 4K | 2048 | 16 tracks |
All five use os 2.2, matching the kernel's BDOS level. That matters for the
feature formats: with os 3, mkfs.cpm writes an UNLABELED CP/M Plus disk
label into the directory, which this BDOS cannot use and which permanently
occupies a directory slot.
The two feature definitions come straight from the DPBs in extra/144tech.md — for 720K, SPT=36, BSH=4/BLM=15, DSM=354, DRM=255, OFF=2, i.e. 355 blocks of 2K with an 8K directory starting on track 2.
cpm86-xt-hd is different in kind: a hard disk's parameters are not compiled
into the BIOS. CHKHDDPAR scans the MBR for a partition of type DBh, reads
that partition's cylinder / head 0 / sector 4, checks that the 256 words of
that sector sum to zero, and lifts the DPB out of the Digital Research disk
label at offset 41. The label on the IBM XT image reads SPT=252 (63 physical
sectors), BSH=5/BLM=31/EXM=1, DSM=2015, DRM=2047, AL0=AL1=FFh, CKS=8000h
(bit 15 = fixed media), OFF=16.
CP/M tracks are counted from LBA 0 of the drive, not from the partition, so
OFF=16 spans the whole of cylinder 0 (16 heads × 63) and the directory begins
at LBA 1008. boottrk 16 therefore stands in for a partition offset, and the
definition applies to the raw disk image. Another drive will carry a different
label, so re-read it rather than reusing these numbers.
A blank feature diskette is only a boot sector plus an erased directory, so
base-720-at.img is generated rather than checked in:
mkfs.cpm -f cpm86-720feat -b <boot-sector> base-720-at.imgmkfs.cpm does not size the image, so it has to be pre-allocated to 737280
bytes of 0xE5 first. The boot sector is extra/144boot2.bin with its last
byte — the media byte, cpmedia in 144boot2.asm — set to 72 for 720K rather
than 144. Note this omits the DOS FAT camouflage and the CP/M-86.720
space-claiming file that a real 144PREP run writes; those only stop DOS and
stock CP/M-86 writing to the disk, and nothing in the boot path reads them.
Physical media still needs a DOS low-level format before 144PREP.
The stock ibmpc-514ds definition shipped with cpmtools corrupts 320K raw
images. The cause is in the BIOS, not in cpmtools.
SETUP_INTREG in pcbios.a86 maps a CP/M track to a physical cylinder/head
like this:
mov ch,al ; cylinder = CP/M track
cmp ch,28h ; below 40 ?
jb SETUP_INTREG_FIN
mov ch,4Fh ; 79
sub ch,al ; cylinder = 79 - track
mov dh,01h ; head 1So CP/M tracks 0-39 are cylinders 0-39 on head 0, and CP/M tracks 40-79 are
cylinders 39 down to 0 on head 1. Side 1 is laid down in reverse cylinder
order. A raw image is a plain CHS dump (cyl0/h0, cyl0/h1, cyl1/h0, ...), so the
CP/M track order and the image order do not match. cpmtools has no diskdef
syntax to express that, and ibmpc-514ds assumes a linear layout.
ibmpc-514ds also disagrees with the BIOS disk parameter block DPBK2
(SPT=32, BSH=4/BLM=15/EXM=1, DSM=157, DRM=63, OFF=1): boottrk 2 on a
linear image yields 156 blocks instead of 158.
The reason the corruption looks so odd is that the directory survives. With
OFF=1 the directory is block 0 at CP/M track 1, which is cylinder 1 head 0, at
byte 8192 of the image. ibmpc-514ds with boottrk 2 also lands on byte 8192.
Block 1 coincides too. Everything from block 2 onward does not, so cpmls
prints a perfect directory while every file reads back as garbage.
The 160K format is unaffected because it is single sided, so the reverse-order branch never runs.
The fix is the corrected cpm86-320 definition described above, plus
tools/cpm86twist.py to convert between physical CHS order and CP/M logical
track order:
python3 tools/cpm86twist.py untwist disk.img work.img
cpmcp -f cpm86-320 work.img base/pip.cmd 0:PIP.CMD
python3 tools/cpm86twist.py twist work.img disk.img
The Makefile targets for cpm86-320.img, cpm86-320-at.img and
cpm86-320-dev.img do this automatically.
--
| Project | Description |
|---|---|
| cpm86-kernel | CP/M-86 1.1 distribution rebuilt from patched and reconstituted sources |
| ccpm86-y2k | CCP/M-86 3.1 distribution rebuilt from patched and reconstituted sources |
| cpm86-crossdev | Unix CP/M-86 cross development project (compilers, emulation and tools) |
| cpm86-hacking | CP/M-86 miscellaneous tools and PCE emulator helpers |
| cpm86-cmdtools | CP/M-86 .cmd file manipulation tools |
| cpm86-ports | CP/M-86 application ports in C and assembler |
| cpm86-vi | STevie vi port for CP/M-86 and PC-DOS 1.1 |
| pcdos11-hacking | PC-DOS 1.1 distribution, tools and notes |
The main source for it is: http://www.cpm.z80.de
- Baseline: http://www.cpm.z80.de/download/cpm86src.zip
- Baseline: http://www.cpm.z80.de/download/cpmdev.zip
- Patching Source: http://www.cpm.z80.de/download/cpm86ann.zip
- Patching Source: http://www.cpm.z80.de/download/cpm86bug.zip
- 144FEAT2 from Freek Heite
To be continued...