os4apic.psd is released

[OS4User]

it looks like it will be distributed with os4krnl
http://ru2.halfos.ru/core/downloads/kernel/os4/os2krnlSVN4066_unoff.zip

[R.M. Klippstein]

How does a person go about supporting this project?

klipp

[Roderick Klein]

While I know the OS/4 team does not have access to the current ACPI code.
Why code your own PSD ?

What I indicated in the past is that some things in eCS could benefit from OS/4 work.
But its not realy documented very well. Also some people complain that we (eCS team)
broke OS/4 kernel support. This was not done on purpose in ACPI.
But its simply difficult to find any information.


The readme file of the os4acpi.txt states for example states:

"OS4APIC.PSD hangs IBM kernel because of lack of service and some other
reasons. Unfortunately there is no way to avoid it due to the bug in the
IBM kernel. "

What is ment by that ? Could the details be more specific ?


MemFiles_en.txt states:
"The main goal is to run ACPI on AMD chipsets where int 13h cannot be used after switching
ACPI on."


From what I can tell about this above description is that its to generic (lacks details) or is just wrong.
I think its the last case. Possibly our PSD version still has some bugs in it.
But from what I can tell (also after haven spoken with David) there is no such thing as INT 13 going away
on AMD systems when ACPI is loaded. It was just that the orginal PSD code made wrong switched between real mode
and procted mode and that caused the hang! That is now fixed.  Has nothing todo with OS2LDR or OS2KRNL bugs.


Just like the orginal /FS switch that was required for Dell systems.
It turned out later it was not required. The PSD implemented
switching between real and protected mode in a wrong manner.
And even with /FS switch when it was still used you could not boot from USB CD ROM drives.
This was fixed with the new PSD's from David.

So while some of the bugfixes done to the OS/4 kernel I can not make a good judgement about them.
Other things the OS/4 team did I realy waonder if they did not went on the wrong technicall track.

But the basis is that technical information is missing as to why certain things where implemented.

Since David started working on ACPI I can only state the amount of random system freezes and locks has basicly gone
away. Of the 50 systems I have had access to test eCS on, only 1 box still fails with ACPI.

So that the current kernel from IBM is not perfect I would agree.

I'm posting this message because some people complained Mensys broke support for OS/4.
But where can I find some details as *why* certain things are said not to be working and why certain things where fixed.

Thanks

Roderick Klein

[walking_x]

Roderick, I have one question about loader, not kernel
There are two alternative OS2LDRs - one from OS/4 and one from separate QSINIT project. Both have a nice menu and a lot of functions.

But they became disfunctional with late ACPI revisions. Can David make some notes about: what ACPI.PSD uses from OS2LDR? Or how this can be "defined" in some kind of API?

[OS4User]

While I know the OS/4 team does not have access to the current ACPI code.

As far as I know ACPI code is closed and available only for eCS team.

The readme file of the os4acpi.txt states for example states:

"OS4APIC.PSD hangs IBM kernel because of lack of service and some other
reasons. Unfortunately there is no way to avoid it due to the bug in the
IBM kernel. "

What is ment by that ? Could the details be more specific ?

In fact it's not bug and it will not affect ACPI.psd in any way. What was meant probably is that in IBM kernel there is no scenario that when PSD is present in config.sys and returns error to INSTALL command. IBM kernel in this case will hang. As OS4APIC is meant only for OS/4 kernel and it uses its features it will not work on IBM kernel. Yes, OS4APIC can recognize IBM kernel, so what? IBM kernel still can't accept error to INSTALL.

MemFiles_en.txt states:
"The main goal is to run ACPI on AMD chipsets where int 13h cannot be used after switching
ACPI on."

From what I can tell about this above description is that its to generic (lacks details) or is just wrong.
I think its the last case. Possibly our PSD version still has some bugs in it.
But from what I can tell (also after haven spoken with David) there is no such thing as INT 13 going away
on AMD systems when ACPI is loaded. It was just that the orginal PSD code made wrong switched between real mode
and procted mode and that caused the hang! That is now fixed.  Has nothing todo with OS2LDR or OS2KRNL bugs.

Yes, you are right. It was a bug in earlier ACPI.PSD and it has nothing in common with kernel and ldr. But earlier there was an opinion that it was connected with AMD chipset. But later ACPI.PSD is working good where that bug used to appear.

As far as I know, there is no mode switch code in PSD, may be David meant switching between PIC and IOAPIC ?

Since David started working on ACPI I can only state the amount of random system freezes and locks has basicly gone
away. Of the 50 systems I have had access to test eCS on, only 1 box still fails with ACPI.

Yes, David did and is doing a great job for improving ACPI.PSD

[R.M. Klippstein]

Just released ! Updated OS4 Kernel

http://ru2.halfos.ru/core/downloads/kernel/os4/os2krnlSVN4070_unoff.zip

[walking_x]

Roderick, you don`t want to answer, but I`m still add a note...

May be you don`t know, but writing alternative OS2LDR it is a walking over the minefield. For example, brave Veit never checks the error code in his pcibus.snp (used in eCS) - and if you try to return error on missing PCI ID it will hang somethere in resource.sys... Another example - it is trying to change I/O Delay calulation... working fine, but all Intel network cards became broken. If you saw Linux drivers sources - you can understand why - it full of delay(5 mks) calls.... A small change in timings can broke all. And so on...
I.e., in addition to full emulation of OS2LDR API - all of "common software" bugs must be in mind.

And David adds a new one! I think he just patch something or read from hardcoded offset in IBM`s OS2LDR... But all, what he need can be received normally, if some kind of API will be defined.

And "alternative loaders" are really useful! Users can mix IBM and OS/4 kernels (OS/4 kernel allow to edit boot CONFIG.SYS in embedded editor - and system can be loaded without ACPI or trouble making driver). Using of multiple config.sys files, debug kernel SYM files, turning off logo from menu, ability to start another OS2LDR (debug version, for example) - are you sure, that David themself do not require some of those features?

[Doug Bissett]

Just released ! Updated OS4 Kernel

I tried that on my Lenovo ThinkPad L530. It is just as broken as the earlier version was. As soon as the actual boot starts (after the OS2  loader menu is responded to, or times out), the machine immediately reboots. This is long before OS2APIC, or OS4APIC would get involved.

I also see that there is a new QSINIT, which works a little better. At least the machine will boot.

I also tried the OS4 package on my antique Athlon machine. It seems to work there, but, from what I have seen, adds nothing that isn't in the IBM stuff.

Roderick, you don`t want to answer

I would hope that Roderick is working at getting eCS 2.2 out the door, in good shape. I would be surprised if he has been here since his last post. I also think it is not wise to discuss details with Roderick. He knows a bit about the overview, but I am sure that he knows very little about the details.