开云体育

ctrl + shift + ? for shortcuts
© 2025 开云体育
  1. H390-Vm
  2. Topics

How to save CMS system?

 

Here I am again to ask one last thing (I hope).
From what I understand, when you want to move a program to minidisk
MNT19E 19E Y/S to make it available to all other users, you have to (re)save the CMS system.
So:
1) acc 19E z, to make it available for writing
2) copyfile my program a = = z
3) read the MAINT MEMO B and
3.1) vmsetup cms
3.2) cp define stor 16m
3.3) ipl 190 clear
3.4) savesys cms
4) shutdown and reipl
5) with any user after logon I get:
DMSFRE161T INVALID DMSFRET CALL FROM F87AE8, ERROR NUMBER 5.
DMSFRE164T LOW-CORE NUCLEUS STORAGE POINTERS DESTROYED (INTERNAL ERROR CODE 5) RE-IPL CMS.? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ??
DMSFRE165T CHAIN HEADER AT 0036F0: 0001EC20 0000000B 000011B8 10F00200.
CP ENTERED; DISABLED WAIT PSW '00020000 00F812B6'
?
quite nasty and unfriendly...
?
I also tried the same things with a copy of VM/370 CE v1r1.2 just downloaded and therefore not "screwed" by me previously.
The result was the same.
?
Any advice is welcome, thanks!
?
Marco
 

Well it needs to mode 2 so the copy has to be = = z2...

.. I then think you need "access (noprof" after the IPL and before the savesyd but on my phone so can't check manuals

Dave


On Wed, 26 Feb 2025, 13:35 M Verpelli via , <marco.verpelli=[email protected]> wrote:
Here I am again to ask one last thing (I hope).
From what I understand, when you want to move a program to minidisk
MNT19E 19E Y/S to make it available to all other users, you have to (re)save the CMS system.
So:
1) acc 19E z, to make it available for writing
2) copyfile my program a = = z
3) read the MAINT MEMO B and
3.1) vmsetup cms
3.2) cp define stor 16m
3.3) ipl 190 clear
3.4) savesys cms
4) shutdown and reipl
5) with any user after logon I get:
DMSFRE161T INVALID DMSFRET CALL FROM F87AE8, ERROR NUMBER 5.
DMSFRE164T LOW-CORE NUCLEUS STORAGE POINTERS DESTROYED (INTERNAL ERROR CODE 5) RE-IPL CMS.? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ??
DMSFRE165T CHAIN HEADER AT 0036F0: 0001EC20 0000000B 000011B8 10F00200.
CP ENTERED; DISABLED WAIT PSW '00020000 00F812B6'
?
quite nasty and unfriendly...
?
I also tried the same things with a copy of VM/370 CE v1r1.2 just downloaded and therefore not "screwed" by me previously.
The result was the same.
?
Any advice is welcome, thanks!
?
Marco
 

Ok, I tryed
...
2) copyfile my program a = = z
...
3.4) access (noprof
3.5) savesys cms
....
Now I get:
DMSFRE161T INVALID DMSFRET CALL FROM F87AE8, ERROR NUMBER 5.
DMSFRE164T LOW-CORE NUCLEUS STORAGE POINTERS DESTROYED (INTERNAL ERROR CODE 5) RE-IPL CMS.? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ??
DMSFRE165T CHAIN HEADER AT 0036F0: 0001EC20 0000000B 000012A0 10F00200.
CP ENTERED; DISABLED WAIT PSW '00020000 00F812B6'
?
Marco
 

typing error
?
2) copyfile myprogram a == z2
?
sorry
?
Marco
 

On Wed, Feb 26, 2025 at 01:35 PM, M Verpelli wrote:
Here I am again to ask one last thing (I hope).
From what I understand, when you want to move a program to minidisk
MNT19E 19E Y/S to make it available to all other users, you have to (re)save the CMS system.
So:
1) acc 19E z, to make it available for writing
2) copyfile my program a = = z
3) read the MAINT MEMO B and
3.1) vmsetup cms
3.2) cp define stor 16m
3.3) ipl 190 clear
?
?
You didn't press enter at this point did you??
On checking it needs to be straight after the IPL

3.4) savesys cms
4) shutdown and reipl
5) with any user after logon I get:
DMSFRE161T INVALID DMSFRET CALL FROM F87AE8, ERROR NUMBER 5.
DMSFRE164T LOW-CORE NUCLEUS STORAGE POINTERS DESTROYED (INTERNAL ERROR CODE 5) RE-IPL CMS.? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ??
DMSFRE165T CHAIN HEADER AT 0036F0: 0001EC20 0000000B 000011B8 10F00200.
CP ENTERED; DISABLED WAIT PSW '00020000 00F812B6'
?
quite nasty and unfriendly...
?
I also tried the same things with a copy of VM/370 CE v1r1.2 just downloaded and therefore not "screwed" by me previously.
The result was the same.
?
Any advice is welcome, thanks!
?
Marco
 

Yes I press enter. Is this the mistake I made?
?
Marco
 

From what I understand, when you want to move a program to minidisk
> MNT19E 19E Y/S to make it available to all other users, you have to
> (re)save the CMS system.

A useful way to avoid having to resave CMS all the time is to have an
additional program disk to hold locally installed stuff that's not part
of the base CMS installation. We used a minidisk that was owned by
MAINT, and ensured that it was accessed by all users as 'P'.

An additional handy thing to know is that RENAME is atomic, so you can
add files to the P disk, or _update_ them there, with impunity, if you
do it right:

- Link the P disk M as some additional mode other than P.
- COPY the new or updated file there with a temp name.
- RENAME the old file (if any) to e.g. a versioned oldname.
- RENAME the new or updated file to the intended name.
- RELEASE the disk.
- Have all logged in users reACCESS P.

Until they reaccess, they will get the old version of an updated file.

If you use MR mode for the link, you may get an R link if someone else
has an M (or W, naughty) link, which is why I suggest M mode. (We had a
script for this, so it was desirable to avoid the issue rather than to
try to detect it and clean it up.)

De
 

The SAVESYS command *MUST* be *ONLY* entered at the 1st VM READ after the IPL command. If you wait until after the 1st ready message, then it won't work.
?
The problem is that DMSINS has some additional initialization to do after IPLing the saved system and if you enter the SAVESYS command after the ready message, that reinit work doesn't get done. One of the really important things that happen in DMSINS after ipling a saved system is to initialize all memory controls based on the memory size of the VM doing the IPL.
?
cheers,
William
 

On Wed, Feb 26, 2025 at 8:35?AM M Verpelli via <marco.verpelli=[email protected]> wrote:
Here I am again to ask one last thing (I hope).
From what I understand, when you want to move a program to minidisk
MNT19E 19E Y/S to make it available to all other users, you have to (re)save the CMS system.

Yes.? That's specifically because CMS keeps a copy of the file directory of the S and Y disks in the saved system, so if you don't do this, the CMS users won't see the change.? CMS refers to these directory copies as the "shared S-stat" and "shared Y-stat" accordingly.
?
So:
1) acc 19E z, to make it available for writing
2) copyfile my program a = = z

As Dave has already?mentioned, any file you put on the S or Y disks needs to be mode 2 (e.g., TYPE MODULE S2).? That's because when CMS creates the shared S and Y stats, it only puts the mode-2 files in them.

3) read the MAINT MEMO B and?
3.1) vmsetup cms

You don't need to do VMSETUP, and in fact the next command shuts CMS down and undoes it :-)\

3.2) cp define stor 16m

CP should respond:

STORAGE = 16384K
CP ENTERED; DISABLED WAIT PSW '00020000 00000000'
?
3.3) ipl 190 clear

At which point CMS puts out the "greeting message":

VM Community Edition V1 R1.2

3.4) savesys cms

You need to type that exactly as seen - no #CP, no other words or punctuation (but upper/lower case doesn't matter), and no pressing ENTER beforehand.? CMS checks the line entered in response to the greeting message for several special cases, including the SAVESYS response.

?Note that if you accidentally issue the CP SAVESYS command instead of responding SAVESYS to the greeting, it's anybody's guess what will happen.? When you do it correctly, CMS prepares itself for saving and then issues the actual CP SAVESYS command itself, via Diagnose 8.

4) shutdown and reipl
That's not necessary.? After your SAVESYS, any user (including MAINT) that does IPL CMS will get the newly-saved system.

5) with any user after logon I get:
DMSFRE161T INVALID DMSFRET CALL FROM F87AE8, ERROR NUMBER 5.
DMSFRE164T LOW-CORE NUCLEUS STORAGE POINTERS DESTROYED (INTERNAL ERROR CODE 5) RE-IPL CMS.? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ??
DMSFRE165T CHAIN HEADER AT 0036F0: 0001EC20 0000000B 000011B8 10F00200.
CP ENTERED; DISABLED WAIT PSW '00020000 00F812B6'

My best guess is that CP got the SAVESYS, not CMS, and that it was processed at an inopportune moment.? The proper sequence looks like this (just captured from my on VM/CE 1.1.2 system):

L MAINT CPCMS
DASD 19D LINKED R/W; R/O BY 004 USERS
DASD 19E LINKED R/W; R/O BY 004 USERS
DASD 290 LINKED R/W; R/O BY 004 USERS
DASD 190 LINKED R/W; R/O BY 004 USERS
DASD 194 LINKED R/W; R/O BY CPWATCH
LOGON AT 17:54:10 EDT WEDNESDAY 02/26/25
VM Community Edition V1 R1.2

CP DEF STOR 16M
STORAGE = 16384K
CP ENTERED; DISABLED WAIT PSW '00020000 00000000'
I 190 CLEAR
VM Community Edition V1 R1.2
savesys cms
SYSTEM SAVED
VM Community Edition V1 R1.2

Y (19E) R/O
Segment GCCLIB is not loaded because virtual machine memory is in use.
U (19D) R/O
B (5E5) R/O
Z (1FF) R/O
CP SET PF7 RETRIEVE
SET RDYMSG SMSG
Ready;

Note that "SYSTEM SAVED" message - CMS issues that when it saves the system.? If you don't get that, it didn't happen correctly.
Ross
 

开云体育



On 26/02/2025 21:18, M Verpelli via groups.io wrote:
Yes I press enter. Is this the mistake I made?
?

I think so. Thats why it pauses after IPL so you can save the CMS segment

Marco
_._
Dave
,_
 

Thanks, it worked perfectly.

I honestly never imagined that one <enter> too many could cause such a problem.
?
Thanks again, I'm going to stop playing with teddy bears and go back to turkeys.
?
Marco
 

While is in some ways a silly analogy it might help to think of CP (Control Program in VM/CMS environments) as BIOS (in traditional PCs) - hit the wrong key during a reboot and, ...the results are unpredictable.?
 

You're missing a crucial detail for an absolutely safe RENAME
  1. CP LINK sharedDisk xxxx 111 M
  2. ACCESS 111 N
  3. COPYFILE ABC EXEC A ABC NEWEXEC N (OLDDATE
  4. RENAME ABC EXEC N = OLDEXEC N (NOUPDIRT
  5. RENAME ABC NEWEXEC N = EXEC =
  6. RELEASE N (DETACH
The trick is NOUPD in step 4.? It makes that this RENAME is done in your storage only.? Any user just accessing this shared MDISK at this time will still see the old ABC EXEC.
The RENAME in step 5 causes the mdisk to be updated, both step 4 and 5 are then executed as an atomic event (and that was mentioned in the original message, without the NOUPD there's no such trick)
?
Beware: don't do any other update to disk N between step 4 and 5 or there is a while with only ABC OLDEXEC present on the mdisk.
And, do not ERASE ABC OLDEXEC before you are sure that all users of this shared disk have reACCESSed it.? ?Users that do not reACCESS have for ABC EXEC pointers in storage that point to what now is ABC OLDEXEC on disk.? If you erase if too soon, their pointers point to a place on the mdisk that can be overwritten at any moment
?
--
Kris Buelens
 

The VMREAD is there for the SAVESYS CMS as well as for an ACCESS (NOPROF.?
?
I remember the early days at IBM's VM test center in Belgium: people that left their 3270 logged on often suffered from someone including a CP LOGOFF in their PROFILE EXEC, harder to fix without this VM READ:
LOGON xxxx NOIPL
CP DEFINE 191 199
CP IPL CMS
ACCESS 199 A
Fix PROFILE EXEC
?
--
Kris Buelens
 

Ahh, yes.. people not playing nicely with other people's stuff.. that's why in the mid-70s my 2nd or 3rd CP mod was to force disconnect idle VMs after 10 minutes
?
cheers, William

1 - 15 of 15