[HECnet] I have found how to recover the TOPS-10 Fortran V11 patch decryption keys

[G.]

Since here too there are some TOPS-10 users, I am pleased to announce that I
have found a way to recover the long-missing TOPS-10 Fortran V11 patch
decryption keys. Find below a copy of the post I just sent to alt.sys.pdp10.

HTH, :)
G.

=====

Dear all,

	as many of you may know, the TOPS-10 Fortran V11 tape image
available in the usual repositories is truncated because the original tape
reel from which it was sourced (BB-D480G-SB) was found to be damaged.

Luckily enough, the first four savesets (out of seven) were indeed readable,
therefore the V11 compiler is in fact available. Unfortunately, one of the
missing savesets (the seventh) contains the keys required to decode TSU tape
patches. Therefore, even if TSU tapes are available and completely readable,
their contents could not be used because they are encrypted.

Nonetheless, after much pondering and careful observing, and thanks to some
good guess, I was able to reconstruct the most important missing keys :)

Since both TOPS-10 and TOPS-20 Fortran compilers share the vast majority of
their sources, my first idea was to build the updated TOPS-10 Fortran
compiler from scratch with TOPS-20 sources from the latest TSU tape, but I
soon found it to be a too complex task because, among other things, some
TOPS-10-specific sources were missing. I would have had to use those from
some old field test tape available in the repository, but I was afraid that
the resulting compiler (if buildable at all) could have been a weird
bug-ridden hybrid. So I had to find a different route. Since I was there, I
also verified that TOPS-20 keys (which have always been available) do not
work with TOPS-10 products.

By observing the keys for other products I noticed that they are wildly
different in size and are usually pretty large, some are several hundred
blocks long, therefore I speculated that the encryption may have been just
some simple XOR of keys and sources. This means that XORing any cleartext
file with its corresponding encrypted version yields the key.

Since I could not directly recover the key for the patched compiler binaries
because I didn't have a suitable cleartext copy of any encrypted binary, I
had to follow a more circuitous route.

The solution comes from the fact that, as I said above, TOPS-10 and TOPS-20
Fortran compilers share the vast majority of their source files, which also
happen to be among the largest of the whole Fortran distribution. Also, most
important, the updated runtime sources come with their respective compiled
counterparts, i.e. REL files. Finally, runtime binary modules are contained
in the runtime sources saveset, whereas the resulting library is contained
in the binaries saveset.

So, here is a summary of the required steps.

1. First of all, on TOPS-20, restore the TOPS-20 Fortran V11 runtime key,
namely FORLIB.KEY, from the product distribution tape BB-4157J-BM.

2. Then, still on TOPS-20, restore the largest updated source of the Fortran
V11 runtime and the decryption tool, namely FOROPN.MAC and DCRYPT.EXE, from
TSU tape 04 (volume 1 of 2) BB-PENEA-BM.

3. Then, again on TOPS-20, use DCRYPT to obtain a plaintext version of
FOROPN.MAC, and call it e.g. FOROPN.T20.

4. Finally, transfer FOROPN.T20 to TOPS-10. If you do not have a working
DECnet network available you could use a tape image. The standard TOPS-20
backup program, DUMPER.EXE, could write (and read) TOPS-10 BACKUP tapes when
instructed to do so with the INTERCHANGE command. Note that the same
INTERCHANGE command in TOPS-10 BACKUP.EXE has another (unrelated) meaning.

5. Now on TOPS-10, restore the original (unpatched) TOPS-10 Fortran V11
runtime files from the truncated tape BB-D480G-SB. In fact they are in the
fourth saveset, i.e. the last readable one. Then, in the same directory,
restore the updated Fortran runtime files from TSU tape 04 (volume 3 of 3)
BB-PBDED-BB. The directory on tape is DSKB:[10,7,FTNOTS]. Several old files
will be overwritten with new patched and encrypted updated copies.

6. Restore tools APUTIL.EXE and DCRYPT.EXE from TSU tape 04 (volume 1 of 3)
BB-BT99V-BB. APUTIL and DCRYPT do basically the same thing, but APUTIL is
list-driven and could verify checksums, whereas DCRYPT does just one file at
a time and does not perform any check.

7. Then restore (or make available) the FOROPN.T20 file created on TOPS-20
and "misuse" DCRYPT to obtain the runtime key. Since the encryption is a
simple XOR and DCRYPT luckily does not distinguish among encrypted, plain,
and key files, just feed it with FOROPN.MAC (which is encrypted as it came
from the TOPS-10 TSU tape) when it asks for some encrypted file, and
FOROPN.T20 when it asks for a key. Call the resulting file FTNOTS.KEY.

|.r [,,tsu]dcrypt
| 
| [DCRYPT version 1(3)]
| Dcrypt>decrypt foropn.mac foropn.t20 ftnots.key
|   Decrypting DSK:FOROPN.MAC
|   with       DSK:FOROPN.T20
|   calling it DSK:FTNOTS.KEY ...[OK]
|
| Dcrypt>

8. Now, with our new key, we can use APUTIL to decrypt every runtime file.
APUTIL reads file names from a VFY file and is able to distinguish between
encrypted and non-encrypted files. Therefore in decrypt mode it will only
work on encrypted files, whereas in verify mode it will work on all files.
If the VFY file and the KEY file have the same name, it's all automatic.

| .r [,,tsu]aputil
| 
| APUTIL>read ftnots.vfy
| [Reading DSKB:FTNOTS.VFY[1,2,FTNOTS]]
| APUTIL>decrypt
| [Decrypting DSKB:[1,2,FTNOTS] using key file DSKB:FTNOTS.KEY[1,2,FTNOTS]]
|  B10FRS.CTL has been decrypted
|  FDDT.MAC has been decrypted
|  FORCHR.MAC has been decrypted
|  . . .
|  MTHDUM.REL has been decrypted
|  MTHPRM.UNV has been decrypted
|  FORDDT.REL has been decrypted
| [Decrypted 25 of 94 files, 0 errors, 0 checksum errors]
| APUTIL>verify
| [Verifying DSKB:[1,2,FTNOTS]]
|  B10FDT.CTL has been verified
|  B10FRS.CTL has been verified
|  F10LIB.CCL has been verified
|  . . .
|  MTHTRP.MAC has been verified
|  MTHTRP.REL has been verified
|  FORDDT.REL has been verified
| [Successfully verified 94 of 94 files, 0 errors, 0 checksum errors]
| APUTIL>

9. At this point we are ready to create our updated runtime library. Run
MAKLIB and feed it with F10LIB.CCL. This will create a new and updated
FORLIB.REL in the current directory, which will be the picklock to recover
the compiler binaries key. Let's rename it to something like FORLIB.NEW.

| .r maklib
| 
| *@f10lib.ccl
| 
| *^Z
| 
| .rename forlib.new=forlib.rel
| Files renamed:
| DSKB:FORLIB.REL

10. Now, in another directory, restore the updated and encrypted compiler
binary files from TSU tape 04 (volume 3 of 3) BB-PBDED-BB. The directory on
tape is DSKB:[10,7,FTNSYS]. Copy to the same directory FORLIB.NEW from the
previous step and feed it to DCRYPT.EXE together with FORLIB.REL (encrypted,
coming from the TSU tape) to obtain FTNSYS.KEY in the same way as above.

11. Now run APUTIL to decrypt and verify the binary compiler files, that is
our final target.

| APUTIL>read ftnsys.vfy
| [Reading DSKB:FTNSYS.VFY[1,2,FTNSYS]]
| APUTIL>decrypt
| [Decrypting DSKB:[1,2,FTNSYS] using key file DSKB:FTNSYS.KEY[1,2,FTNSYS]]
|  FORDDT.REL has been decrypted
|  FORLIB.REL has been decrypted
|  FORO11.EXE has been decrypted
|  FORTB.EXE has been decrypted
|  FORTC.EXE has been decrypted
|  FORTD.EXE has been decrypted
|  FORTE.EXE has been decrypted
|  FORTF.EXE has been decrypted
|  FORTG.EXE has been decrypted
|  FORTRA.EXE has been decrypted
| [Decrypted 10 of 10 files, 0 errors, 0 checksum errors]
| APUTIL>verify
| [Verifying DSKB:[1,2,FTNSYS]]
|  FORDDT.REL has been verified
|  FORLIB.REL has been verified
|  FORO11.EXE has been verified
|  FORTB.EXE has been verified
|  FORTC.EXE has been verified
|  FORTD.EXE has been verified
|  FORTE.EXE has been verified
|  FORTF.EXE has been verified
|  FORTG.EXE has been verified
|  FORTRA.EXE has been verified
| [Successfully verified 10 of 10 files, 0 errors, 0 checksum errors]
| APUTIL>

:)

The same trick, but with REVHST.MAC instead of FOROPN.MAC, could be used to
recover the compiler sources key, although it will not decrypt a couple of
files because they are longer than REVHST.MAC and some other unpatched files
will not pass the checksum verification because the originals are missing
(they were stored in the unreadable section of the product distribution
tape) and those from the field test tape I had to use are different. At the
moment I have no fresh data at hand, but if my notes are correct it worked
for 253 out of 258 files, i.e. only 5 files didn't pass the checksum test,
and of those only 2 are actual source files.

I hope you'll find all of this useful and amusing as much as I did, :)
G.