<< Previous Message Main Index Next Message >>
<< Previous Message in Thread This Month Next Message in Thread >>
Date   : Tue, 09 Jul 1991 02:22:07 GMT
From   : iWarp.intel.com!ichips!inews!cad412!dbraun@uunet.uu.net (Doug Braun ~)
Subject: Re: UZI and ZCPR (was Re: small c, assemblers, linkers )

In article <921@spam.ua.oz> dcook@spam.adelaide.edu.au writes:
>>
>>>My objective is to port UZI (version 7) onto an Archives Model 3 just
for the
>>>hell of it. I would also like to investigate ZCPR3, but again it is
hard to see
>>>in the mass of stuff available a document describing exactly what is
required
>>>and what software/hardware prerequisites there are.
>>UZI was written for a fairly complete compiler. I forgot, sorry. Maybe
>>somebody else knows more (the author of UZI? ;-)).
>>
>
>I've seen UZI mentioned a lot in this group, but I've never
>seen anything that really indicates what it is.
>So, what is it ? 
>What sort of CP/M machine does it require ? Hard disk, 1 or 2 floppies ?
>56/128 k RAM ?
>


Well, here is my standard UZI plug, from about a year ago.  It mentions
the compiler I used.  Note that I am very close to releasing UZI-280,
which is ten times cooler than UZI, assuming you have a Z280 computer.

UZI is presumably still on SIMTEL20, and probably other places as well.





               UZI: UNIX Z-80 IMPLEMENTATION

                 Written by Douglas Braun


Introduction:

UZI is an implementation of the Unix kernel written for a Z-80 based
computer.  It implementts almost all of the functionality of the
7th Edition Unix kernel.  UZI was written to run on one specific
collection of custom-built hardware, but since it can easily have device
drivers added to it, and it does not use any memory management hardware,
it should be possible to port it to numerous computers that current use
the CP/M operating system.  The source code is written mostly in C,
and was compiled with The Code Works' Q/C compiler.  UZI's code was
written from scratch, and contains no AT&T code, so it is not subject
to any of AT&T's copyright or licensing restrictions.  Numerous 7th
Edition programs have been ported to UZI with little or no difficulty,
including the complete Bourne shell, ed, sed, dc, cpp, etc.


How it works:

Since there is no standard memory management hardware on 8080-family
computers, UZI uses "total swapping" to achieve multiprocessing.
This has two implications:  First, UZI requires a reasonably fast
hard disk.  Second, there is no point in running a different process
while a process is waiting for disk I/O.  This simplifies the design
of the block device drivers, since they do not have to be interrupt-based.

UZI itself occupies the upper 32K of memory, and the currently running
process occupies the lower 32K.   Since UZI currently barely fits in 32K,
a full 64K of RAM is necessary.

UZI does need some additional hardware support.  First, there must be
some sort of clock or timer that can provide a periodic interrupt.
Also, the current implementation uses an additional real-time clock
to get the time for file timestamps, etc.  The current TTY driver assumes
an interrupt-driven keyboard, which should exist on most systems.
The distribution contains code for hard and floppy disk drivers, but
since these were written for custom hardware, they are provided only
as templates to write new ones.


How UZI is different than real Unix:

UZI implements almost all of the 7th Edition functionality.
All file I/O, directories, mountable file systems, user and group IDs,
pipes, and applicable device I/O are supported.  Process control
(fork(), execve(), signal(), kill(), pause(), alarm(), and wait()) are fully
supported.  The number of processes is limited only by the swap space
available.  As mentioned above,  UZI implements Unix well enough to
run the Bourne shell in its full functionality.  The only changes made
to the shell's source code were to satisfy the limitations of the C compiler.

Here is a (possibly incomplete) list of missing features and limitations:

    The debugger- and profiler-related system calls do not exist.

    The old 6th edition seek() was implemented, instead of lseek().

    The supplied TTY driver is bare-bones.  It supports only one port,
    and most IOCTLs are not supported.

    Inode numbers are only 16-bit, so filesystems are 32 Meg or less.

    File dates are not in the standard format.  Instead they look like
    those used by MS-DOS.

    The 4.2BSD execve() was implemented.  Additional flavors of exec()
    are supported by the library.

    The format of the device driver switch table is unlike that of
    the 7th Edition.

    The necessary semaphores and locking mechanisms to implement 
    reentrant disk I/O are not there.  This would make it harder to
    implement interrupt-driven disk I/O without busy-waiting.


A Description of this Release:

Here is a list of the files supplied, and a brief description of each:


intro:         What you are reading

config.h:      Setup parameters, such as table sizes, and the device
               driver switch table.

unix.h:                All strcuture declarations, typedefs and defines.
               (Includes things like errno.h).

extern.h:      Declarations of all global variables and tables.

data.c:                Dummy to source extern.h and devine globals.

dispatch.c:    System call dispatch table.

scall1.c:      System calls, mostly file-related.

scall2.c:      Rest of system calls.

filesys.c:     Routines for managing file system.

process.c:     Routines for process management and context switching.
               Somewhat machine-dependent.

devio.c:       Generic I/O routines, including queue routines.

devtty.c:      Simple TTY driver, slightly-machine dependent.

devwd.c:       Hard disk driver.  Very machine-dependent.

devflop.c:     Floppy disk driver.  Very machine-dependent.

devmisc.c:     Simple device drivers, such as /dev/mem.

machdep.c:     Machine-dependent code, especially real-time-clock and
               interrupt handling code.

extras.c:      Procedures missing from the Q/C compiler's library.

filler.mac:    Dummy to make linker load UZI at correct address.

makeunix.sub:  CP/M SUBMIT file to compile everything.

loadunix.sub:  CP/M SUBMIT file to load everything.


Miscellaneous Notes:

UZI was compiled with the Code Works Q/C C compiler and the Microsoft
M80 assembler under the CP/M operating system, on the same hardware
it runs on.  Also used was a version of cpp ported to CP/M, since
the Q/C compiler does not handle macros with arguments.  However, there
are only a couple of these in the code, and they could easily be removed.

Because UZI occupies the upper 32K of memory, the standard L80 linker
could not be used to link it.  Instead, a homebrew L80 replacement linker
was used.  This generated a 64K-byte CP/M .COM file, which has the lower 
32K pruned by the CP/M PIP utility.  This is the reason for appearance
of the string "MOMBASSA" in filler.mac and loadunix.sub.

To boot UZI, a short CP/M program was run that reads in the UZI image,
copies it to the upper 32K of memory, and jumps to its start address.
Other CP/M programs were written to build, inspect, and check UZI filesystems
under CP/M.  These made it possible to have a root file system made before
starting up UZI.  If the demand exists, these programs can be included
in another release.


Running programs under UZI:

A number of 7th Edition, System V, and 4.2BSD programs were ported to
UZI.  Most notably, the Bourne shell and ed run fine under UZI.
In addition the 4.2BSD stdio library was also ported.  This, along
with the Code Works Q/C library and miscellaneous System V library 
functions, was used when porting programs.

Due to obvious legal reasons, the source or executables for most of these
programs cannot be released.  However, some kernel-dependent programs
such as ps and fsck were written from scratch and can be included in future
releases.  Also, a package was created that can be linked to CP/M .COM
files that will allow them to run under UZI.  This was used to get
the M80 assembler and L80 linker to run under UZI.  Cpp was also
ported to UZI.  However, it was not possible to fit the Q/C compiler
into 32K, so all programs (and UZI itself) were cross-compiled under CP/M.

The Minix operating system, written for PCs by Andrew Tanenbaum et al,
contains many programs that should compile and run under UZI.  Since
Minix is much less encumbered by licensing provisions than real Unix,
it would make sense to port Minix programs to UZI.  In fact, UZI itself
could be ported to the PC, and used as a replacement for the Minix kernel.

 Doug Braun                         Intel Design Technology
                                    408 765-4279

 dbraun@scdt.intel.com

 or maybe:

  / decwrl \
  | hplabs |
 -| oliveb |- !intelca!mipos3!cadev6!dbraun
  | amd    |
  \ qantel /

<< Previous Message Main Index Next Message >>
<< Previous Message in Thread This Month Next Message in Thread >>