CC-tea/org-hyperion-cules
1
0
Fork 0
mirror of https://github.com/SDL-Hercules-390/hyperion.git synced 2026-04-16 08:55:23 +02:00
Code Issues Packages Projects Releases Wiki Activity
Files
04cbb443da170f390093d7bc40015a8d451648a1
org-hyperion-cules/RELEASE.NOTES

437 lines
20 KiB
Plaintext
Raw Blame History

-------------------------------------------------------------------------------
** PROGRAMMING NOTE **
Don't forget to update "hercnot.html" with the below information too!
-------------------------------------------------------------------------------
Version 4.0 RELEASE NOTES
* Minimum supported Windows platform is now Windows XP
The new minimum supported Microsoft Windows platform is now Windows XP or
greater. Windows 98, Windows NT and Windows 2000 are no longer supported.
Support for Windows XP will likely be dropped in the very near future too
however, so all Hercules users are strongly encouraged to upgrade their
systems to at least Windows Vista. 64-bit Windows 7 Professional SP1
or greater is preferred.
* "MAXCPU" and "NUMCPU" statements
The combination of NUMCPU and MAXCPU controls the behavior of how many
CPU engines will be configured online upon startup and how many can be
configured online later. In previous versions this was controlled via
the NUMCPU statement and the compile-time constant 'MAX_CPU_ENGINES'.
For compatibility with previous versions if MAXCPU is not specified its
value defaults to NUMCPU. If neither is specified it defaults to 1.
* CMPSC Compression Call 2012 instruction implementation is now the default
The default implementation for the CMPSC Compression Call instruction is
now the new cmpsc_2012 implementation. The previous legacy implementation
is now the alternate implementation implemented as an optional dynamically
loadable module called "altcmpsc.dll". Refer to the README.CMPSC document
for more information.
* The sequence of certain system parameters may be important
The sequence of certain configuration file system parameter statements
may be important. Specifically, the ARCHLVL statement (previously called
ARCHMODE and defines the initial architectural mode of the system) should
generally precede (come before) any other ARCHLVL statement that enables
or disables a given architectural feature.
For example: on certain IBM operating systems the z/Architecture "ASN
and LX Reuse Facility" feature must be enabled or the system will not IPL.
The facility is enabled in Hercules via the ARCHLVL ENABLE ASN_LX_REUSE
configuration file statement, but since the facility is a z/Architecture-
only feature, your initial architecture must either be set to "z/Arch"
beforehand or else you need to specify the optional [archmode] parameter
on your ARCHLVL ENABLE ASN_LX_REUSE statement.
This means the ARCHLVL z/Arch< statement must usually come before your
ARCHLVL ENABLE ASN_LX_REUSE statement in your configuration file.
Otherwise the ASN and LX Reuse Facility would not be enabled.
The ARCHLVL statement is currently the only system parameter where the
sequence may matter. The order of all other system parameters does not
currently matter. This may change in the future however, so be sure to
always carefully read through the RELEASE NOTES (this document) with
each new release of Hercules.
* More architecturally compliant Channel Subsystem implementation
Hercules 4.0 Hyperion's channel subsystem implementation now more closely
adheres to published and unpublished architectural specifications. It does
not precisely adhere to the complete specification but it does adhere much
more closely than ever before (and definitely more closely than the legacy
implementation still used in the older spinhawk 3.xx series of Hercules).
If you experience any anomalies in the behavior of your guest operating
system, verify that you are using architecurally valid/correct configuration
settings that your guest operating system expects. Pay particular attention
to your LPARNUM, CPUMODEL, CPUSERIAL, CPUVERID, CPUIDFMT, MODEL, PLANT
and MANUFACTURER values since they typically have a direct impact on how
certain guest operating systems behave.
* Configuration file statements and panel commands handled identically
Most configuration file statements are now available as and processed as
panel commands. Most valid configuration file statements may now also be
entered as a panel command. Most panel commands are now also allowed as
configuration file statements. See the documentation for full details.
-------------------------------------------------------------------------------
Version 3.06 RELEASE NOTES
* Tape "AUTOMOUNT" support
Starting with Hercules version 3.06 a new AUTOMOUNT option is available
that allows guest operating systems to directly mount, unmount and query
tape device filenames for themselves, without any intervention on the
part of the Hercules operator.
Automount support is enabled via the AUTOMOUNT configuration file statement.
An example guest automount program for VSE called "TMOUNT" is provided in
the util subdirectory of the Hercules source code distribution.
Briefly, the 0x4B (Set Diagnose) CCW is used to mount (or unmount) a file
onto a tape drive, and the 0xE4 (Sense Id) CCW opcode is used to query the
name of the currently mounted file.
For mounts, the 0x4B CCW specifies the filename of the file to be mounted
onto the drive. The file MUST reside in the specified AUTOMOUNT directory
or the automount request will be rejected. To unmount the currently mounted
file, simply do a mount of the special filename "OFFLINE".
To query the name of the currently mounted file, the 0xE4 CCW is used. Note
however that the 0xE4 (Sense Id) CCW opcode cannot be used by itself since
the drive may also already natively support the Sense Id CCW opcode. Instead,
it must be preceded by (command-chained from) a 0x4B CCW with a data transfer
length of one byte. The following 0xE4 command is the one that then specifies
the i/o buffer and buffer length of where the query function is to place the
device's currently mounted host filename.
In summary:
MOUNT: X'4B', <filename>, X'20', <length>
UNMOUNT: (same thing but use filename "OFFLINE" instead)
QUERY: X'4B', <buffer>, X'60', 1
X'E4', <buffer>, X'20', <buffersize>
Again please refer to the provided TMOUNT program for a simple example
of how automount support might be implmented on a guest operating system.
-------------------------------------------------------------------------------
Version 3.05 RELEASE NOTES
* The 'conspawn' utility used to process 'sh' commands now recognizes a
specially designed keyword "startgui" to accomodate automatic starting
of Windows GUI applications via the .RC file or panel command-line.
If the first word following 'sh' is "startgui", then the "ShellExecute"
API is used to start the requested program rather than the 'system()'
API as otherwise.
The "startgui" keyword must always be used to start any Windows program
that is not a command-line program. Hercules, itself being a command-line
program, monitors the 'stderr' and 'stdout' pipes so it can log messages
received from either pipe directly to the Hercules console log. Programs
such as notepad however, because they are not command-line programs, do
not use stdout/stderr thus causing Hercules to hang if "start" is used
instead.
This rule applies regardless of how Hercules itself is started (i.e. via
HercGUI or directly via the command-line) and regardless of whether the
"start" command is wrapped in a batch file or not. That is to say, using
the Hercules command "sh batchfile foobar" to start your batch file which
then does "start notepad %1" still causes Hercules to hang until notepad
first exits. Instead, you should ask Hercules to "sh startgui batchfile",
and let the batchfile start notepad however it wants.
* Real SCSI tape drives used with Hercules must provide a certain minimum set
is "IBM compatible" support in their SCSI command set/behavior in order to
work properly with Hercules. Furthermore, the Hercules device-type used on
your device statement in your Hercules configuration file should match the
the level of support/behavior that they provide.
For example, all SCSI tape drives used with Hercules must provide the ability
to set variable-length blocks as well as long erase-gaps (long erase-gaps
allows new data to be appended to the end of existing data without having to
write a tape-mark to separate the new data from the old existing data first).
Another example would be using a model of SCSI tape drive that happens to
report physical block-id values in a format different from the way real IBM
mainframe tape drives report them. 3480/3490 tape drives for example report
their block-ids (used in Read Block Id and Locate CCWs) in a very specific
format wherein bits 1-7 of the high-order byte of the reported 4-byte block-
id indicates the tape's physical "segment" location of where the lower 22-
bit block number is physically located on the tape. (The block-id segment
is used to allow the tape drive to quickly position itself to the approximate
location where the desired block acually resides on the tape and thus allows
high-speed positioning for the Locate CCW).
If the model of SCSI tape drive you are actually using with Hercules does not
use this same block-id format however, then it cannot be used with Hercules
as a 3480 or 3490 model tape drive with specially defined options.
If the SCSI tape drive you are using reports its block-ids using a 32-bit
block-id value (the same way a 3590 model tape drive does), then similarly,
it should be defined to Hercules as a model 3590 device-type as well (since
that is how it is behaving with respect the format of the returned blockid
values). It you wish to define it in Hercules as a model 3480 or 3490, then
you will need to use specially defined options before it will work properly
as the model drive you wish it to emulate.
With all that being said, it should be noted that PARTIAL support for 3590
device emulation is possible with judicious use the aforementioned special
options, but full/complete 3590 support is unlikely due to lack of publicly
available documentation. Details regarding 3590 CCW handling is restricted
(confidential) IBM proprietary information, and is not normally available
outside of IBM. Not long ago IBM was required by US law to publish such
information, but unfortunately for Hercules, such is no longer the case.
For further information regarding use of SCSI attached tape drives with
Hercules and their associated specially defined options, please refer to
the section on SCSI tape drives in the Hercules's Device Configuration
documentation.
* In order to ensure proper functioning of the TOD clock with older versions
of guest operating systems, the default values of Hercules's internal thread
priorities for the Windows version of Hercules were changed to be identical
to those used by all other supported platforms. Originally, the default
thread priority values for the Windows version of Hercules were:
*** 3.04 (and prior) Default Priorities ***
Thread Priority Meaning
------- -------- ------------------------
HERCPRIO 0 Normal Process priority
DEVPRIO -8 Above Normal Thread priority
TODPRIO 0 Normal Thread priority
CPUPRIO 0 Normal Thread priority
which caused acceptable performance/functioning on most, but not all, guest
operating systems. Beginning with version 3.05 however, the prioriries now
default to:
*** 3.05 (and later) Default Priorities ***
Thread Priority Meaning
------- -------- ------------------------
HERCPRIO 0 Normal Process priority
TODPRIO -20 Time Critical Thread priority
DEVPRIO 8 Below Normal Thread priority
CPUPRIO 15 Lowest Thread priority
which may on more modern guest operating systems (which handle the TOD clock
differently than do older less sophticated versions) cause a slight decrease
in overall performance. If such is the case, the original default priorities
(and thus the original behavior) can be obtained via addition of appropriate
HERCPRIO, TODPRIO, DEVPRIO and CPUPRIO control file statements with values
identical to the original version 3.04 default values.
* Additional configuration file usability enhancements have been implemented
in the form of a new 'INCLUDE' (and associated 'IGNORE') statement, allowing
configuration files to "include" statements from a different named file.
Additonally, a new "enhanced" symbolic substitution syntax is now also
supported. Refer to the Hercules "Configuration File" documentation for
further information and details.
A rather nifty "Automatic Operator" facility has also been implemented in
the current release as well. While not exactly a "configuration file
usability enhancement", it is nevertheless something we hope might prove
to be more useful/helpful to our many users. See the "README.HAO" document
for more information.
-------------------------------------------------------------------------------
Version 3.01 RELEASE NOTES
- Support for z990 crypto instructions is conditional on the presence of the
glibcrypt library.
If Hercules is BUILT, the development files for glibcrypt should be available.
When hercules is RUN, the runtime files for glibcrypt should be installed.
Depending on the level of glibcrypt used to *build* hercules, the associated
level of glibcrypt should also be present on the target machine. On systems
supporting shared library versioning, multiple levels of the glibcrypt
runtime libraries can be installed simultaneously, ensuring availability
of the z990 crypto instructions, regardless of the level of glibcrypt with
which hercules was initially built.
- CTC and LCS devices now ++MUST++ specify ALL addresses on the configuration
statement.
Example:
0A00.2 LCS .....
0B00.2 CTCI ....
or
0A00.4 LCS -oat hercules.oat
or
0A00-0A03 LCS -oat hercules.oat
or
0A00 LCS -oat hercules.oat
0A01 LCS
0A02 LCS
0A03 LCS
Previously (i.e. with version 3.00), only the first (even numbered) device
needed to be defined and Herc would automatically define the odd numbered
device for you. Starting with Hercules version 3.01 however, you now need
to define BOTH devices, just like you did with versions prior to 3.00.
Once again, starting with version 3.01, you **MUST** define BOTH DEVICES.
-------------------------------------------------------------------------------
Version 3.00 RELEASE NOTES
- Reminder that CTCI device handling was changed as follows:
- The CTCI-W32 protocol is deprecated. You should use the CTCI protocol
instead.
- You MUST NOT define both even/odd CTCI device pairs in your configuration
file. You should ONLY define the first even numbered device. Hercules will
automatically define the odd numbered device for you. If you define the
odd numbered device by mistake, an open error will occur on that device.
This is by design. See the README.NETWORKING document for further details.
- Hercules Dynamic Loader support: starting with version 3.00, Hercules now
contains support for the dynamic loading of certain modules upon startup on
some platforms (e.g. Linux and Windows for example). As a result of this new
feature, "Hercules" itself now no longer consists of just the 'hercules.exe'
module by itself, but rather consists of BOTH the 'hercules.exe' program AS
WELL AS whatever dynamic modules (DLLs) that accompany it.
As a result if this change, whenever you install a new version of Hercules,
you must ensure that you ALSO install the accompanying new versions of the
new dynamic modules as well. Attempting to use a version of Hercules with a
dynamic module that was not specifically built for that version will cause
loading of that dynamic module to fail.
You CANNOT mix versions of Hercules with differing versions of dynamically
loaded modules.
Ensure that your library path LD_LIBRARY_PATH is set correctly such that it
includes the directory of your Hercules executables. Especially, message
HHCCF042E will indicate that system is unable to locate necessary loadable
modules.
- ECPS:VM: Do not use ECPS:VM (See README.ECPSVM) in an AP or MP environment
in VM/370. If AP=YES or MP=YES is coded in DMKSYS and the AP/MP control
file is used to build the CP nucleus and NUMCPU is set to more than 1 in
the hercules.cnf file, any of LOK001, LOK003 or other abends will occur.
This occurs because the Hercules ECPS:VM CP Assist implementation is not
MP safe, and particularily, attemps VM dispatching without holding necessary
AP or MP locks.
- Due to the change in Hercules' "mainstor" memory allocation technique to
address a "double memory consumption" bug in Cygwin's malloc implementation,
some Windows Hercules users may experience an "out of memory" error whenever
Hercules is started with a large MAINSIZE configuration file value:
"HHCCF031S Cannot obtain nnnMB main storage"
This error will most likely occur (if it does at all) for those users who
have manually adjusted their Cygwin "heap_chunk_in_mb" Windows registry
setting value (in order to allow them to specify a large MAINSIZE value
when running Hercules). If this problem does occur (i.e. if you DO happen
to experience the above mentioned "HHCCF031S Cannot obtain main storage"
error with this new release of Hercules), then either REDUCE your "heap_
chunk_in_mb" value (yes, that's correct: REDUCE it; i.e. change it to a
SMALLER value) or else remove it altogether (so as to let it default).
Detailed explanation:
History/background: Cygwin has a built-in limit to the amount of memory
that may be allocated in one chunk. If you try 'malloc'ing more than this
limit, you will receive an "out of memory" error. Since many Hercules users
specify large MAINSIZE values in their configuration file, they sometimes
experience this problem.
The suggested workaround to this "problem" was to add a "heap_chunk_in_mb"
registry value to Cygwin's registry settings with a large enough value such
that Cygwin would then be able to satisfy Hercules' 'malloc' request for
such a large MAINSIZE value.
This worked fine until sometime around version 1.3.15 of Cygwin, at which
time they began using a different 'malloc' technique that unfortunately
causes TWICE as much Windows virtual memory to be allocated for any large
memory allocation (the technical reasons of which are explained in comments
in source member config.c where mainsize is being allocated).
In order to address this double memory allocation issue in Cygwin's malloc
implementation, Hercules was changed to use mmap to allocate its mainstor
rather than malloc (which, unlike malloc, does NOT inadvertently allocate
twice as Windows virtual storage than was requested), which did indeed re-
solve the "double memory allocation" problem.
Unfortunately however, because Cygwin's malloc and mmap logic each consume
completely different portions of Windows' virtual memory, the more memory
that is reserved for malloc usage (via using a larger "heap_chunk_in_mb"
value), the LESS becomes available for mmap usage!
Thus, while increasing your "heap_chunk_in_mb" registry value USED to HELP
[to allow you to allocate larger amounts of mainstor (MAINSIZE)], it NOW
causes the complete OPPOSITE effect: it ends up PREVENTING you from being
able to 'mmap' as much storage as you'd like to have!
Conclusion:
Therefore, if you are currently using the "heap_chunk_in_mb" registry value
setting to allow you to allocate large MAINSIZE values, then starting with
version 3.00 of Hercules you need to DESCREASE your "heap_chunk_in_mb" value
(or remove it altogether and let it default) in order to leave enough memory
remaining for Hercules (Cygwin actually) to be able to satisfy its 'mmap'
request for your desired MAINSIZE amount.
-------------------------------------------------------------------------------
Reference in New Issue View Git Blame Copy Permalink