The Memory Protection Debacle

Hi everyone,

Thank you for all the replies. This is a heavy subject and until now I didn't have free time to read this with my full attention.


Memory protection scares me.... Why? winxp is a big stinking pile of crap......and it HAS this "glorious all-necessary, 'I can't sleep at night without it' " memory protection, and I'd still rather be using AOS.

See, my problem is, I thought with it, might come a loss of what AOS was capable of, and win isn't, even though they use it.



Quote:
Hi tfrieden,

I was the original poster. I copy and pasted my comment that I wrote on AmigaWorld.Net a long time ago. My link in the first post goes right to that thread over on their site.

You misunderstood my question. I was asking, if there isn't memory protection in an OS, what SW can not be made?

Seems to me that there isn't any kind that couldn't be done.

I'm just sitting here wondering why every desktop OS has to have it.... just because, well, the other guys do? I counld use their OS then, because apparently without it OSs can't possibly function, even though we WERE for 15 years.


Quote:

Well, I don't know what that means, but I'd like to say what I am afraid of when it comes to memory protection.

I like snooping, so, if memory protection is on at "maximum strength", could someone, still write a program, even with basic, that could look through the entire ram, and find a string of characters anywhere of say "abcd" or any 8 bit binary combination?

If there are sections of ram that CAN'T be checked, then I am dead set against it, because it is MY computer and I OWN it and if I am looking for something I expect to be able to locate it.

Quote:
If just resource tracking is implemented, could AOS4.0 run all the SW that it's currently able to without SW failure cropping up? (That is, all current compiled AOS4.0 SW and all SW that works from previous versions of AOS1.0-3.9 that works, I guess that's the SW that runs on Petunia.)

I'd just like to say, I gotta hand it to "Team AOS-Hyperion", they've done it again..... There's their* way and then there's the "Amiga Way" of getting these things done.... resource tracking, memory protection, etc.


Hyperion, you are the best!!!!!
:clap:


* Mac(linux??), linux, windos(not)...

@Atheist

Quote:

Why not?
What you describe is a simple search function, which you should be able to perform on any whatsoever strict memory protection there will be.

It?s not that you going to write that string over with something else, is it?
Of course, if memory protection does a "protect against reading" from programs not owning that particular address, then it won?t work...but why would someone want to do that?

REMINDER: I don?t know nothing about memory protection...just trying to add to the noise

@Atheist

You wrote :

"The Memory Protection Debacle"

I think you wanted to write "Debate" ?

A "Debacle" is what the Amiga Community is experiencing in the last months

Luca

Quote:
Hi Raziel,

Actually, that's one of the things that made the Amiga truely as great as it ever was!

The unified memory model, where, if you could find a string in chip ram, and you could change the content, it would immediately be reflected on the screen, or through the speakers....

As much as the Amiga could be a serious computer, I liked the fun and peculiar aspects of it also, and am reluctant to see them go, even the trivial pieces of it.... Where is the other system that I can buy that offers this at the current time as well???

Only ONE exists, but it uses stuffy linux. (Guess what I'm talking about.)

Quote:
For fun..... and profit.



Hi Luca,

Errrrr, yeah, I agree.


Errr, OTOH, I think for me it may be a debacle, unfortiunately.

@Raziel

You are partially wrong here : generaly (not speaking about AOS4 here) memory protection comes with (="is implemented using") separated virtual memory space, that means that each task has it's own memory address space, thinking it's the only task running on the computer, then each used virtual memory page is mapped to a free physical page.
Telling this, you generally aren't able to snoop into other's tasks memory space on those systems (being for reading or writing) as if you are trying to access address 0xdeadbeef on task 1 you won't end in the same physical address than via address 0xdeadbeef on task 2...

Now speaking of AOS4, (current, as in AOS4 final) memory protection is implemented via two mechanisms (OS4 devs correct me if I'm wrong) :
- access to page '0' (i.e. via a null pointer) is catched and does not lead to a whole system crash
- tasks can have memory space protected against writing by other tasks (here some spaces are protected automatically by the compiler and/or system (for example constants, some code region, etc.), and other can be protected at allocation time by the programmer if he knows that this sapec won't be used by other tasks).

Hope that would have helped you both.

@abalaban

How can I get my process to have different memory space?
From address 0x00000000 to 0x00200000?

@LiveForIt

Using for example the MMU you can remap each memory page used by the process.
Thus this can 'fool' the process thinking it's accessing 0x00000000 whereas in fact it's accessing 0xDEADBEEF.

@abalaban

True, but LiveForIt can't do that himself, only the Exec can write to the MMU registers, only the Exec knows where LiveForIt's program is and what the mapping for his program is. And that's information that no other task knows or needs to know. Information that a hacker would LOVE to know.

@tonyw

Of course, such things are automatically dealt with by the kernel, not by application writers !

@tonyw

i don't wonto find out how memory layout of process is, i just wonto to use my own, i think i need it for native cpu support in a vm.

there is few wm out there that might be usefull to oss.

Hi Everyone,

I still didn't really get an answer to my question, so I'll start with one only.


IF only resource tracking was implemented in AOS4.0, would all, or some, or none of the old Amiga 1.0 to 3.9 and 4.0 SW break?

Hm, I think we make too many steps in advance here... I see lots of people writing things like "MP eases software development", "helps writing better code" et al, but (please excuse me) where the hell is these software?

Let?s face it: We discuss future features of the OS without having that much software that will definetly need these features right now (or even in near future).

Where is the need to publish software before a competitor does? Where?s the competition?

Where is the advantage of MP, if there?s even no actual SDK for the final update providing the "great" feature of shared objects (which I consider really unnecessary but for plain porting and "gcc pleasant" development)?

Well, we discuss the benefits of MP and just have a half usable development system for OS4.

I don?t want to fight the thoughts about OS future of anybody here, but I want to say that we face other, much more urgent, problems that need to be solved right now and that more energy should be put on them instead using all this energy for some nice dreams about future bloatware development (sorry, can?t resist).

Greetings,

@whose

you are making the mistake thinking there are 1000+ coders working full time on writing software for amigaos, i tell you what we are mostly part time develpers working on the week ends.

@Atheist

Quote:

i don't think, rosorcre tracking will not break old software.

smp support and memory protection will break old software, and only solution is to run an older os version in parallel in a virtual environment, as transparent as possible.

@Atheist

I'll ask you one question.. What is it that you absolutely
need that would break in a new system?
Mentioning any OS4 application doesn't apply because
those are actively being developed and can be changed
for a new API.

I personally don't have ANY classic applications installed.
Why should I? Other than a word processor maybe,
what's needed?
But I'm quite confident that when we have this new API
we'll also have both a competent browser aswell as a
word processor native.

EDIT: Sorry I have one classic application. A FTP client.

@LiveForIt

Resource tracking is inactive in OS4 just because it
would break software compatibilty. I believe it was
tfrieden who said this in another thread a while back.

@LiveForIt

No, I don?t make a mistake here... you mentioned the exact problem... there will be no benefit of MP and such, IF there will not be more programmers (regardless if fulltime or hobbyists). What?s the big use of a handful (possibly ported and bloated) programs that won?t break the system?

So, first stage should be making really useable build tools, fully working compilers, debuggers and such. MP is a future topic, but I fear that people use too much energy on it instead of doing all they can to get a good foundation for development again.

Greetings,

@whose

Quote:

No benefit?
What about applications crashing not taking down the
system, requiring a reboot?
What about making the system instable, if not crashing
at once?
What about the possibility to close down any program
that has crashed with out disturbing any other
application (Yes, this requires resource tracking aswell)?
What about making it easier for developers because not
the whole system goes down when testing applications?

I would guess some external developers won't touch a
OS that hasn't got MP because of the extra hassle
developing without it.
This doesn't only help when porting stuff. It will help
any developer making native stuff aswell.

@samwel

pls read carefully... it won't have benefit if we didn't manage to earn more devloper's interest for the platform.

This isn't possible yet, just because the development system is not even fully useable (latest SDK isn't there, gdb is more work to guess the failure than a help etc.).

All I say is that we should concentrate right now on the OS/development system before we put useless efforts on theories that will possibly last another 5-6 years before they become reality.

It's all the same when such discussions emerge: Everybody is talking about fine "must have" features that are just a pancake on the top, if they will appear one day in the far future, but the flaws of the system that exist yet will be forgotten every time.

Nobody is seriously using the possibilities AmigaOS bears right now. Where are the really good games everybody wants to see? Some SDL based efforts were made (ports), that's it. The often demanded development tools? Well, some half-finished ports, again. Very few AmigaOS only programs saw the light of the world until now.

OS4 is transformed to a never ending construction site, because lots of the new features and enhancements we just have don't get the attention they badly need because everybody is dreaming of things like MP and so on.

What about AmigaInput? What about the USB subsystem? What about 2D/3D hardware acceleration? More SCSI drivers? A real good debugger? More development documentation? I can ask the whole day on, there will be no satisfying answer.

I really know that there are not as much developers as needed, but to dream every day of new things to implement to the OS won't make the situation better.

All I ask is to start to use the things we have right now, polish them, make them more useful, expand them where it is badly needed (there are lots of fields to explore, polish and expand, believe me).

We have enough time to implement e.g. MP half a year or a year later. AmigaOS cannot compete in the actual market in this half finished state it is in, regardless if it bears MP facilties or not. It could, even without MP, if it would be in a better state than it is yet.

Greetings

P.S.: I own a OS4 system and I'm quite happy with it as a user, but not as a developer. I don't like command line developing using e.g. vi or NotePad, abcshell and a command line gcc with make. We all know that there were gui driven development tools in the past (and they weren't such bad as was said again and again) and that other platforms have these tools for years now. But for AmigaOS such tools are simply not possible, because lots of people don't see the point and keep on dreaming about things they don't need that urgently.