23 Jun
2008
23 Jun
'08
2:27 a.m.
While fantasizing having a PDP-11 of my own, I was reading about TRIPOS
and decided that would be fun to play with. This line from the TRIPOS
wikipedia article caught my attention:
The most important TRIPOS concepts have been the non-memory-management
approach (meaning no checks are performed to stop programs from using
unallocated memory)
Does this mean that any process is allowed to scribble anywhere in memory
without provoking a segfault?
--
David Griffith
dgriffi at cs.csubak.edu
A: Because it fouls the order in which people normally read text.
Q: Why is top-posting such a bad thing?
A: Top-posting.
Q: What is the most annoying thing in e-mail?
22 Jun
22 Jun
5:32 p.m.
While fantasizing having a PDP-11 of my own, I was reading about TRIPOS
and decided that would be fun to play with. This line from the TRIPOS
wikipedia article caught my attention:
The most important TRIPOS concepts have been the non-memory-management
approach (meaning no checks are performed to stop programs from using
unallocated memory)
Does this mean that any process is allowed to scribble anywhere in memory
without provoking a segfault?
I beleive so. I'll have to look at the sources again, but I am pretty
sure that the original TRIPOS would run on a PDP11/03. And I seem to
rememebr mention of the PDP11/40 or PDP11/45 in the source, but that the
MMU was not used.
-tony
23 Jun
23 Jun
11:56 p.m.
MC68000 versions of TRIPOS (including AmigaDOS) definitely had no
interprocess memory protection. I don't know if later versions of
AmigaDOS implemented protection if a hardware MMU was available or
even whether they were still considered TRIPOS derivatives.
24 Jun
24 Jun
12:36 a.m.
On Tue, Jun 24, 2008 at 08:56:13AM +0200, Eric J Korpela wrote:
MC68000 versions of TRIPOS (including AmigaDOS)
definitely had no
interprocess memory protection. I don't know if later versions of
AmigaDOS implemented protection if a hardware MMU was available or
even whether they were still considered TRIPOS derivatives.
AmigaDOS itself never used the MMU for memory protection, but there
was a tool available _with_ AmigaDOS (first for developers, then for
everyone) called "Enforcer" that would set up some basic memory
protection between all processes and sensitive parts of the OS.
The basic memory protection scheme was to write-protect low
mem so a random process wouldn't scribble on $00000000 or $0000004
(reads/writes to $00000000 usually meant that your code was
trying to dereference a null pointer - the reaction to which
is implementation specific - it's usually safe to treat a
null pointer as if it were a pointer to a null (zero-length
string), but it's equally valid for the code to puke from
trying it; reads of $00000004 are absolutely required on
the Amiga because that's how you find all the libraries
and system structures. _Writes_ however are very bad and
can crash the machine in under a second).
I don't think there ever was any attempt to prevent one
process from scribbling on another process, though.
Enforcer just kept you out of the OS-required area
and kept you from accidentally writing to unallocated
memory (falling off the ends of buffers, etc). Malicious
writes into other process space wouldn't have been
stopped.
-ethan
--
Ethan Dicks, A-333-S Current South Pole Weather at 24-Jun-2008 at 07:30 Z
South Pole Station
PSC 468 Box 400 Temp -79.4 F (-61.9 C) Windchill -113.8 F (-81.0 C)
APO AP 96598 Wind 8.4 kts Grid 21 Barometer 676.1 mb (10776 ft)
Ethan.Dicks at usap.gov http://penguincentral.com/penguincentral.html
12:43 a.m.
On Tue, Jun 24, 2008 at 08:56:13AM +0200, Eric J Korpela wrote:
MC68000 versions of TRIPOS (including AmigaDOS)
definitely had no
interprocess memory protection. I don't know if later versions of
AmigaDOS implemented protection if a hardware MMU was available or
even whether they were still considered TRIPOS derivatives.
Oh... I should probably point out that AmigaDOS was pretty much always
tied to its TRIPOS roots, but that the DOS was only part of the software
running on the system - the underlying message-passing layer was "Exec",
then "Intuition" was the window manager. AmigaDOS was the part worried
about loading code segments and the filesystem and system directory
structure (where to put scripts and binaries and icons and such).
Device drivers and libraries (quite similar internally) and processes
and so on, was part of Exec.
It's quite a complex system, one reason for the multiple phonebooks
for a reference guide - very little of that, though, is AmigaDOS.
I think I was an Amiga developer for four or five years before I
ever saw the AmigaDOS manual. From day one, though, I had the
RKMs (ROM Kernel Manuals) and used them *lots*. You pretty much
needed them for any code that had to open a screen or a window,
or that would work with disks below the file layer (block access
and such) or that fiddled with audio. You could get away from that
level of involvement if you just did things with standard sorts
of C functions, curses, file I/O, etc., but not for anything deeper.
So while AmigaDOS may have been what you saw when you interacted
with a CLI window, and it certainly got invoked if you did stdio-
sorts of things, as a part of the whole, it wasn't a very big
chunk of "The Amiga".
I still hate BCPL and BPTRs, though.
-ethan
--
Ethan Dicks, A-333-S Current South Pole Weather at 24-Jun-2008 at 07:30 Z
South Pole Station
PSC 468 Box 400 Temp -79.4 F (-61.9 C) Windchill -113.8 F (-81.0 C)
APO AP 96598 Wind 8.4 kts Grid 21 Barometer 676.1 mb (10776 ft)
Ethan.Dicks at usap.gov http://penguincentral.com/penguincentral.html
6025
days inactive
6026
days old
4 comments
4 participants
participants (4)
-
ard＠p850ug1.demon.co.uk -
dgriffi＠cs.csubak.edu -
ethan.dicks＠usap.gov -
korpela＠ssl.berkeley.edu