Aix Running on Clones?
Patrick Hurley
Has anyone out there ever gotten aix to run on clones?
What I experience was the following:
The first attempt at getting it going was an attempt at
installing it on an XYZ clone 386/20mhz with harddisk with
an ide disk controller. No luck , it complained about
not being able to find a harddisk.
Figuring that I decided to try to install it on
a 386 with a ESDI hard drive...like the Ps/2 . This was
tried on a Packard Bell 386/33mhz ( I'm not sure about the
wait state ) with a 120meg ESDI drive. The installation went
further up to the point of where the installation procedure
asked for the maintaince diskette. Then it bitched about
not being able to find information about the hard disk. But
the kernel came up ( So that was interesting ).
So at the point I took the maintaince diskette and mounted
its filesystem on another machine running AIX. I found using
strings command that the error was coming from the mounted /etc/maint
module. That is as far as I got.
Now after all this ranting.....has anyone gotten any further?
Is this the way IBM is copy protecting AIX? The way I figure it
the more versions of AIX out there the more hardware IBM will sell in the
long run if this is indeed the case.
Mike Jones
In article <1990Nov13.1...@watserv1.waterloo.edu>, hur...@watserv1.waterloo.edu (Patrick Hurley) writes:
>
> Has anyone out there ever gotten aix to run on clones?
>
>
> Now after all this ranting.....has anyone gotten any further?
>
> Is this the way IBM is copy protecting AIX? The way I figure it
> the more versions of AIX out there the more hardware IBM will sell in the
> long run if this is indeed the case.
Ummm...copy protecting? Not quite. IBM has never claimed that AIX will work
on anything other than PS/2's, and support has only been announced on
specific models of those. Most clones are clones only down to the BIOS
level, and have various bits of hardware underneath, so a "clone" is
actually a different machine to AIX. Would you be surprised if you tried to
install it on, say, a Sun i386 and it didn't work? The fact is that AIX is
optimized for the PS/2 hardware, not crippled to keep it from working on
other hardware.
My opinions, of course, are not IBM's official opinions.
Mike Jones | In the long run, every program becomes rococo, and
AIX Development | then rubble.
Kingston, NY | -- Alan Perlis
ibmps2!aix!mjones |
Steve Dyer
>The fact is that AIX is
>optimized for the PS/2 hardware, not crippled to keep it from working on
>other hardware.
Well, I'd agree except I'd say that "written for" and "only tested on"
isn't exactly the same as "optimized for".
I *HAVE* seen AIX PS/2 (or the very very close Locus equivalent)
running on Compaq 386's out at Locus. However, such support is
not in the product shipped by IBM.
--
Steve Dyer
dy...@ursa-major.spdcc.com aka {ima,harvard,rayssd,linus,m2c}!spdcc!dyer
dy...@arktouros.mit.edu, dy...@hstbme.mit.edu
HAAS
>I *HAVE* seen AIX PS/2 (or the very very close Locus equivalent)
>running on Compaq 386's out at Locus. However, such support is
>not in the product shipped by IBM.
>--
>Steve Dyer
>dy...@ursa-major.spdcc.com aka {ima,harvard,rayssd,linus,m2c}!spdcc!dyer
>dy...@arktouros.mit.edu, dy...@hstbme.mit.edu
I believe it was a man named Jack Vogel from Locus that claimed to have
AIX running on many clones within Locus, the LAST time this subject came
up. So we KNOW it's possible. I guess what all of us AIX folks that aren't
'Unix internals' guys want to know is: HOW? What changes are neccesary?
Does it require simple modifications to the installation diskettes, (even
if that might mean a binary patch) or do we simply not have the resources
neccesary?
IBM's AIX doesn't support any hard disk size EXCEPT for the sizes
that they make. This is absolutely STUPID, but can be 'fixed'. Good grief,
if IBM decides to sell another size disk, they'll have to release new
installation diskettes. All because they want to make it as dificult
as possible for anyone to use anything but (the small and expensive) IBM
hard disks.
So if anyone is willing to give us some clue as to what would be required
to run a LICENSED COPY of AIX on ANY machine other than one manufactured
by IBM, it would be greatly appreciated.
Thanks. . . .
hh
--
Harry Haas GTRI/RIDL/DB "What makes it DO that!?" - Bones
Georgia Institute of Technology, Atlanta Georgia, 30332
uucp: ...!{decvax,hplabs,ncar,purdue,rutgers}!gatech!prism!hh2
Internet: h...@prism.gatech.edu hhaas@{gtri01|rmadsun}.gatech.edu
Steve Dyer
>So if anyone is willing to give us some clue as to what would be required
>to run a LICENSED COPY of AIX on ANY machine other than one manufactured
>by IBM, it would be greatly appreciated.
I really doubt you could do this without AIX PS/2 source. It's just
that the system distributed by IBM is completely customized for a
microchannel PS/2 environment with certain-sized fixed disks attached
to the embedded IBM controllers, and so on. I'd even be suspicious of
trying this on a microchannel clone, not that there are many of them.
What you're looking for is essentially what SCO and ISC have done to
the original Intel/AT&T 386 UNIX distribution--take a distribution
which might be limited to a small subset of controllers/video adaptors
and add enough support that it will run on the vast majority of 386
systems constructed out of generic parts. While the work is "trivial",
the support isn't fast, easy or cheap. I surmise that Locus, the
source of AIX PS/2's primary technology, has not done this either
because of a marketing agreement with IBM (plus, IBM contributed large
parts of AIX PS/2), or, as likely, it simply isn't practical given the
dominance of Sys V.3 and soon, Sys V.4 systems in the 386/486 marketplace.
IBM can sell AIX PS/2 because it's taking advantage of a niche market.
A small company like Locus probably has its hands full already and can't
afford to be another SCO or ISC.
I agree that it's a real shame that you can't run AIX PS/2 on anything
other than an overpriced IBM machine, because it's really a quite
creditable version of UNIX, with many of the features which people
had been waiting for in SysV.4. It's also the only place you can currently
buy Locus distributed system technology (unless you have a spare 370
around to run AIX 370). The TCF facility seems a bit "weird" compared
to NFS or AFS until you get used to it, and then it's rather addictive.
Brian D. Horn
>>
>> Has anyone out there ever gotten aix to run on clones?
Actually AIX/PS2 started on a Compaq 386 as IBM couldn't get the PS/2s
shipped early enough. These worked at LCC until about a year ago.
...
>Ummm...copy protecting? Not quite. IBM has never claimed that AIX will work
>on anything other than PS/2's, and support has only been announced on
>specific models of those. Most clones are clones only down to the BIOS
>level, and have various bits of hardware underneath, so a "clone" is
>actually a different machine to AIX. Would you be surprised if you tried to
>install it on, say, a Sun i386 and it didn't work? The fact is that AIX is
>optimized for the PS/2 hardware, not crippled to keep it from working on
>other hardware.
There is no guarantee of AIX running on non-PS/2 machines, however
it will run on non-MCA machines (just don't complain when something
doesn't work). The biggest problem one might run into is the console
driver, but if the card is VGA compatible even that might work. Other
device drivers - you're on your own. The code is not especially optimized
for PS/2 hardware, but there probably are a few assumptions about the
hardware that are implicit (such as non-volatile RAM availability) that
might well cause problems on other boxes.
N.B.: I (and IBM and LCC) make NO promises or guarantees about AIX
running on non-supported hardware (or the accuracy of the above statements).
Newsgroups: comp.unix.aix
Subject: Re: AIX/PS2 Running on Clones?
Summary:
References: <1990Nov13.1...@watserv1.waterloo.edu> <1990Nov1...@fenway.aix.kingston.ibm.com>
Followup-To:
Distribution: comp
Organization: Locus Computing Corporation, Inglewood, CA
Keywords:
In article <1990Nov1...@fenway.aix.kingston.ibm.com> mjo...@fenway.aix.kingston.ibm.com writes:
>
>In article <1990Nov13.1...@watserv1.waterloo.edu>, hur...@watserv1.waterloo.edu (Patrick Hurley) writes:
>>
>> Has anyone out there ever gotten aix to run on clones?
Actually AIX/PS2 started on a Compaq 386 as IBM couldn't get the PS/2s
shipped early enough. These worked at LCC until about a year ago.
...
>Ummm...copy protecting? Not quite. IBM has never claimed that AIX will work
>on anything other than PS/2's, and support has only been announced on
>specific models of those. Most clones are clones only down to the BIOS
>level, and have various bits of hardware underneath, so a "clone" is
>actually a different machine to AIX. Would you be surprised if you tried to
>install it on, say, a Sun i386 and it didn't work? The fact is that AIX is
>optimized for the PS/2 hardware, not crippled to keep it from working on
>other hardware.
There is no guarantee of AIX running on non-PS/2 machines, however
it will run on non-MCA machines (just don't complain when something
doesn't work). The biggest problem one might run into is the console
driver, but if the card is VGA compatible even that might work. Other
device drivers - you're on your own. The code is not especially optimized
for PS/2 hardware, but there probably are a few assumptions about the
hardware that are implicit (such as non-volatile RAM availability) that
might well cause problems on other boxes.
N.B.: I (and IBM and LCC) make NO promises or guarantees about AIX
running on non-supported hardware (or the accuracy of the above statements).
Steve Dyer
> There is no guarantee of AIX running on non-PS/2 machines, however
>it will run on non-MCA machines (just don't complain when something
>doesn't work). The biggest problem one might run into is the console
>driver, but if the card is VGA compatible even that might work. Other
>device drivers - you're on your own.
I'd imagine that the disk driver would be a possible source of problems,
since IBM's embedded ESDI disks have a rather unusual controller
which does not look at all like your typical MFM/RLL AT-bus controller.
(I know AIX PS/2 1.2 added support for the Mod 55 with its MFM disk.
I don't know whether it looks like most AT-bus type controllers.)
Jack F. Vogel
>AIX running on many clones within Locus, the LAST time this subject came
>up. So we KNOW it's possible. I guess what all of us AIX folks that aren't
>'Unix internals' guys want to know is: HOW? What changes are neccesary?
>Does it require simple modifications to the installation diskettes, (even
>if that might mean a binary patch) or do we simply not have the resources
>neccesary?
However, as both Steve and Brian have noted in subsequent followups, we
were until very recently running AIX (a pre-GA version I might note) not
"on many clones", but on Compaqs. But neither Steve or Brian have gone
into detail on what was required to do this; as a matter of fact, we had
to build completely different kernels to run on these sites. So unfortunately
its not just a matter of diddling with your installation disks or anything
else that you might have the ability to fiddle with that will get it to
work.
Brian suggested that it would work on "clones" with VGA, actually I doubt
this is all that would be required. I think it would have to have an
identical NVRAM configuration as well which means ABIOS clone I believe.
And, as he noted, even if you could get it to run it would be unsupported.
Steve pondered on why Locus itself might not offer this option, well this
is an IBM product whose development was contracted to Locus so we wouldn't
have this option even if we wanted ( and, I might add, I make no claims
as to what the marketing types might want :-}). We also quit using these
systems and that code when AIX 1.2 GA'ed since there was no justification
for the work required to keep that code in sync, and since our production
environment is an IBM customer just like anyone else.
So the days of "clones" running AIX at LCC are gone. In its development
lifetime some distant relative of the current AIX 1.2 has run on many
different type cpu's, however IBM has chosen the present limitation and
you will be hard-pressed to get around it. Where this limitation might
change is with the offering from the OSF, or at least one can hope.
Disclaimer: Naturally, these opinions are mine, not LCC or IBM's.
--
Jack F. Vogel ja...@locus.com
AIX370 Technical Support - or -
Locus Computing Corp. ja...@turnkey.TCC.COM
Eliot Lim
>The TCF facility seems a bit "weird" compared
>to NFS or AFS until you get used to it, and then it's rather addictive.
Some people (hard core unix types) here think that TCF is an outdated
and laughable idea but I don't really agree. Is there any
justification for this other than simple bigotry?
(I don't expect Locus people to agree either :-))
Eliot
Steve Dyer
>I suspect there is also a problem with selling a product which can only be
>used by someone who is violating their license to use AIX on a specific
>machine.
Huh? There's nothing in my license agreement which says my copy of AIX PS/2
has to be running on a PS/2. Obviously you would only run AIX PS/2 on as
many machines as you had purchased licenses for, but that's just a standard
licensing issue.
Anyway, I doubt Locus could sell a "LPP" in the classical sense to get this
to work on clones, but I suppose an "clone installation kit" would work: new
boot/install disks with a kernel and utilities which knew about ISA-class
machines and common device controllers. Plus a selected few new *.o files
to go into /usr/sys/386/*.a. The new install disks would install AIX PS/2
as normal and finish by building a new kernel with the replacement drivers.
Hey, what a great idea! When will you come out with it? :-)
Richard M. Mathews
>But neither Steve or Brian have gone
>into detail on what was required to do this; as a matter of fact, we had
>to build completely different kernels to run on these sites. So unfortunately
>its not just a matter of diddling with your installation disks or anything
>else that you might have the ability to fiddle with that will get it to
>work.
We built different kernels, but did we have to? The old version didn't
have minidisks. If you could get minidisks to work on the clone, you
might be able to do this as an OCO change. Since, to my knowledge, this
has never been done, it is pure speculation.
>I think it would have to have an
>identical NVRAM configuration as well which means ABIOS clone I believe.
Don't we support at least one PS/2 model with no NVRAM?
>Steve pondered on why Locus itself might not offer this option, well this
>is an IBM product whose development was contracted to Locus so we wouldn't
>have this option even if we wanted ( and, I might add, I make no claims
>as to what the marketing types might want :-}).
I think Locus is free to create LPPs, but I doubt it would be in Locus's
best interest to upset IBM. There would certainly be a question in this
case over whether we had made legitimate use of confidential information.
I suspect there is also a problem with selling a product which can only be
used by someone who is violating their license to use AIX on a specific
machine.
>Disclaimer: Naturally, these opinions are mine, not LCC or IBM's.
You can say that again. In fact, I will:
Disclaimer: Naturally, these opinions are mine, not LCC or IBM's.
Richard M. Mathews D efend
Locus Computing Corporation E stonian-Latvian-Lithuanian
ric...@locus.com I ndependence
lcc!ric...@seas.ucla.edu
...!{uunet|ucla-se|turnkey}!lcc!richard
Richard M. Mathews
>>The TCF facility seems a bit "weird" compared
>>to NFS or AFS until you get used to it, and then it's rather addictive.
>Some people (hard core unix types) here think that TCF is an outdated
>and laughable idea but I don't really agree. Is there any
>justification for this other than simple bigotry?
I'd be really curious to find out what it is about TCF that some people
find "outdated and laughable." If TCF is outdated, we should work on
improving it. As a user of TCF for 6 years, I consider NFS and AFS to
be outdated and laughable. NFS and AFS give you file system access; TCF
gives you the whole machine.
Perhaps what they mean is that the version of AIX which includes TCF is
outdated because it doesn't include <fill in some feature from BSD4.3+,
SVR4, or where ever>? If that's the case, all I can say is that IBM
and LCC keep putting out updates -- if you want your favorite feature
to get on the list, try convincing support to put in a DCR or whatever
it's called.
Disclaimer: I speak only for myself (if that), not for LCC, IBM, or
anyone else.
Richard M. Mathews Freedom for Lithuania
Locus Computing Corporation Laisve!
ric...@locus.com
David Skeen
(Richard M. Mathews) writes:
> I'd be really curious to find out what it is about TCF that some people
> find "outdated and laughable." If TCF is outdated, we should work on
> improving it. As a user of TCF for 6 years, I consider NFS and AFS to
> be outdated and laughable. NFS and AFS give you file system access; TCF
> gives you the whole machine.
When OSF didn't accept the TCF parts of the DEcorum proposal, the future
of TCF became somewhat clouded. I would expect TCF functionality to be
added to OSF/DCE in the future, but when and how aren't understood. AFS
is an integral part of OSF/DCE; there is some degree of interoperability
between DCE and NFS. For the near term, that leaves TCF as odd man out.
(This is not an official IBM position, but it is my opinion ...)
Dave Skeen
IBM Internet: d...@dlsrt.austin.ibm.com / IBM VNET: SKEEN at AUSTIN
Steve Dyer
It's not in AIX 3.x, and I am beginning to fear that AIX PS/2
and AIX 370 are niche products which will eventually wither away
to be replaced by OSes with the same name but without the present
unique functionality.
I'm asking this in good faith. It's just that TCF never seems to
appear in any strategic direction reports from IBM.
Richard M. Mathews
>> find "outdated and laughable."
>of TCF became somewhat clouded. I would expect TCF functionality to be
>added to OSF/DCE in the future, but when and how aren't understood. AFS
>is an integral part of OSF/DCE; there is some degree of interoperability
>between DCE and NFS. For the near term, that leaves TCF as odd man out.
First, while OSF didn't accept TCF as part of DCE, TCF was not rejected
either. It was deemed to be beyond the scope of the DCE Request For
Technology. OSF may include part or all of it in the future. My
personal, biased opinion is that this is quite likely. You are right
that when and how are not known.
Even if TCF is never accepted by OSF, I would still not call it outdated.
Different, but not outdated. Perhaps even before its time. The success
of AFS does not preclude the success of TCF. The fact that HP and IBM
were interested in TCF to living side-by-side with AFS in the DEcorum
proposal indicates that the extensions TCF provides have some real
value. Even if OSF does not ever accept TCF, I personally expect TCF
to be available for some systems -- perhaps even OSF systems.
Finally, TCF is still part of AIX on the PS/2 and 370 platforms. I have
no reason to believe that IBM will turn its back on the customers who
have started to depend on this functionality.
By the way, to make sure my prejudices are clear, I was a manager on the
DEcorum project at Locus. I have also spent the last 6 years working on
TCF (mostly as a kernel hacker).
Disclaimer: Opinions are purely my own. I have no special knowledge of
the possible future of TCF which has biased my opinions. My opinions
were already sufficiently biased by my TCF chauvinism;-)