Re: obstacles to an iseries risc compiler

["Leif Svalgaard" <leif@xxxxxxxx>]

then you tell us :-)

no seriously, the first 4096 bytes of the program object is writeable.

The remaining pages have the PP (page protection bits) set to

disallow write (even in system state).

To write something there means that you first

have to mess with the page table entries (doable, but VERY

messy - for masochists only) or you have to call the lower-

level storage management routines that SST uses (also

not for the average Joe, even if he is "worth his salt")

So when you say that you have "hacked" AS/400 programs

you probably mean using SST and that is cheating.

Real programmer do this with restorable user-state

programs without adopted authority.   :-)   :-)  

 

----- Original Message -----

From: Walden H. Leverich

To: 'MI400@midrange.com'

Sent: Friday, June 08, 2001 10:06 PM

Subject: RE: obstacles to an iseries risc compiler

Leif,

 

In what way is a program write protected? I've hacked as/400 programs to do things they weren't originally compiled to do.

 

-Walden

-----Original Message-----
From: Leif Svalgaard [mailto:leif@leif.org]
Sent: Friday, June 08, 2001 8:34 AM
To: MI400@midrange.com
Subject: Re: obstacles to an iseries risc compiler

From: Steve Richter

 

An academic question.

 

What makes the writing of a non ibm powerpc compiler impossible on the iseries?

===> I don't think it is impossible. (as to the benefits, I'm not so sure).

 

The pgm validation value looks to be one. It prevents you from restoring your pgm after a system crash.

===> 1) the system doesn't crash

2) Even if it did, there is no need to restore your program

3) Computation of the PVV is possible

 

Is there sufficient undocument knowledge of the pgm hdr, passed on over the years from guru to guru, to create a working pgm object?

===> yes, but there are better ways (Attaching your code to an existing, dummy program).

 

The undocumented ibm extensions to the powerpc assembler code.  It should be doable to figure that out, right?

===> most them are already "figured out"

 

The system state, user state issue.  Would this rogue risc compiler be able to create the system state segments of the pgm object?

===> you probably mean "system domain" objects.

1) it is not necessary for the pgm to be system domain

2) you can easily give the pgm the system domain attribute

3) you probably don't want to as you cannot call your program from a user domain program.

 

You probably meant to ask if you can write protect the program as normal AS/400

programs are. This is not worth the trouble (for a rogue compiler!)