Aren
/
NetBSD
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

New information about VAX port

This commit is contained in:
ragge 1994-10-08 15:41:42 +00:00
parent b23c0b1b23
commit e4b5499885
1 changed files with 26 additions and 26 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

52
sys/arch/vax/README
View File

@ -1,41 +1,41 @@
$Id: README,v 1.2 1994/08/16 23:39:52 ragge Exp $
$Id: README,v 1.3 1994/10/08 15:41:42 ragge Exp $
Some information and status of VAX port:
We are trying to make a more easy-to-port-to-new-vax-cpu structure than the old
Berkeley code provides. autoconf.c was a real mess! We will here try to keep
cpu-dependent autoconfig code in ka???.c, so we don't have to bother with
those old structures.
It runs single-user on 11/750 with not to much problem, but because
sendsig/sigreturn isn't written yet (the only important functions that
still is unwritten!) it isn't possible to use anything but /bin/sh.
It is possible to compile under NetBSD now, but there are an bug in
pmap that makes it reference deallocated pmaps sometimes.
The file skit.c shouldn't exist; it contains all currently undefined symbols
(even undefined pointers, bleh!) It is actually created from an awk script
taking input from ld :) Things are taken out of there as soon they are
implemented.
Device drivers for UDA50 and DELUA are slightly changed from the 4.4BSD
release, and uda/ra routines work well. Deuna/delua routines can receive
packets, but not send packets. haven't found out why yet.
Device drivers for UDA50 and DELUA are almost unchanged from the 4.4BSD
release, and at least uda/ra routines seem to work.
Config code isn't working yet, there are more important things to fix
Config code isn't working at all, there are more important things to fix
first :) To glue together IO devices there are a file glue.c. Config.new
just creates some header files right now.
Current state of VAX port of NetBSD:
This is how it look like just now:
- Some locore init works fine.
- Virtual memory starts ok.
- It may find some devices :)
- It mounts root filesystem from RA81 OK.
- It loads init and fork it to process 3 to exec /bin/sh, but
something is wrong, copy-on-write on stack area doesn't work...
- Only works on 11/750 yet.
- Runs single-user, compiling programs no problem.
- Locore is taken away, startup code totally written in C :)
- It is totally compatible with 4.3 Reno binaries, here I use
only 4.3Reno userland to be sure it works :)
Lots important things to do:
- Boot blocks and routines. (We currently use old 4.3 boot)
- Whole skit.c should be emptied.
- Clean the code almost everywhere from foolish printouts. :)
Important things to fix:
- find bug in pmap which makes it crash sometimes.
- write sendsig/sigreturn.
- fix autoconfig routines so that kernel isn't statically configed.
As base code for our work we have used 0.9b distribution. For development
we used gas 1.35 with patches for bit instructions and gcc 2.5.8.
Other things to do:
- Write boot blocks and /boot program.
As base code for our work we use 1.0b distribution. For development
we use gas 1.35 with patches for bit instructions and gcc 2.5.8.
(newer gas 2.X is fixed for these instructions already).
The kernel can currently only be compiled without -O option, because
we lacks some volatile definitions in device code. It anyway takes
about 2 1/2 hours to recompile the whole kernel...