os4: KEE update notification

[OS4User]

OS4Team plans to add to KEE the following entries:

#define IAECPU_IN_BARRIER_ON  1
#define IAECPU_OUT_BARRIER_ON 2
#define IAECPU_EXCLUDECURRENT 4

typedef void KEEENTRY (*RendCallBack_t)(void*);

KEERET   KEEENTRY KernInvokeAtEachCpu(RendCallBack_t CallBack,void *arg,uint32_t flags);
KEERET   KEEENTRY KernInvokeAtSpecificCpu(RendCallBack_t CallBack,void *arg,uint32_t CPUnum);

These entries are to help CPU rendezvous.

Hope, it will be useful to VirtualBox porters,  test version is available upon request.

[Lars]

No it's not useful as it leaves everybody outside that wants to continue to use the original kernel.

[Martin Iturbide]

Hi

Why does everybody has to show off his frustration when someone does something that is not common?
If you find an effort/project/idea/intention useless, I think it is better for you to don't share your opinion if it is not constructive.

Why don't you suggest a better way to do it?

It may be useless for people using the original kernel, but it may be useful for people that uses the OS4 kernel... that is.

Regards

[Lars]

So you are looking for a community that splits up even more than it is today ?
That will be the moment in time when I will finally pull the plug on OS/2.

If you cannot install a standard eCS or AN Installation inside a virtual machine then that's it for me.
That's not frustration. That's common sense.

Ok, here is something constructive: don't put these routines in the kernel, put it in ACPI.PSD and ACPI.SYS (the ACPI driver from the OS4 team) respectively.
Then have the 2 parties agree on a common Interface.

[Martin Iturbide]

Ok, here is something constructive: don't put these routines in the kernel, put it in ACPI.PSD and ACPI.SYS (the ACPI driver from the OS4 team) respectively.
Then have the 2 parties agree on a common Interface.

That was a lot better.

People are free to choose the goal of their projects, if they want to do something crazy or different we can not stop them, since it is their time and their resources.
For example, I can not force the people at ArcaOS to do something specifically with their project, because I'm not a customer,  an investor or a developer investing time on it.  The only thing I can say is "please, check this out"... "my opinion is this one, because of blah blah and I think it will be good because of blah blah". It will not help at all to says some things like "IT IS USELESS", "IT IS BAD", "IT HAS NO FUTURE" without elaborating the reason and suggesting alternatives. (...and also I'm not on my 20's anymore

The same applies to the OS4 project, it is their project, they are the one investing their time on it. If you want to suggest things you have to elaborate. This is not twitter, we are not forced to say things on 140 character or less. And they also may reject the good ideas because of their limited resources just like the rest of the community developers.

We can not force other people to do what we want on their free time (and not being hired for) with incomplete arguments and just saying "it is common sense". It is required to convince and inspire the other party and explain why it is a good idea. Lars, you are in a better position than mine to inspire other developers at this technical level. I had failed on inspiring other OS/2 users/developers in the past and I don't think I'm accomplishing it these days.

"Common sense" may be great but it can not compete with the "free will" of each individual. "Free will" is why we have OS/2 users today, while some others slip away thanks to "Common sense".

Regards

[OS4User]

So you are looking for a community that splits up even more than it is today ?
That will be the moment in time when I will finally pull the plug on OS/2.

Lars, we respect you and your work on USB stack and other projects.
Sorry, for negative emotions which we bring you.

If you cannot install a standard eCS or AN Installation inside a virtual machine then that's it for me.
That's not frustration. That's common sense.

New KEE in OS/4 kernel is aimed to help a Host System, not a Guest System and thus it will in no way influence the installation of the standard eCS or AN inside a VM.

Ok, here is something constructive: don't put these routines in the kernel, put it in ACPI.PSD and ACPI.SYS (the ACPI driver from the OS4 team) respectively.
Then have the 2 parties agree on a common Interface.

Unfortunately it is impossible to properly provide such service within PSD or some other way, it is possible to do only in the kernel by adding new KEE (by the way, ACPI itself has absolutely no connection with the matter).

[Lars]

1) what happens if you want to install eCS/AN inside VirtualBox where eCS/AN is the host system ?

2) ACPI as such has no relation to it, yes. But a PSD is a good place to put additional platform functionality to leave the kernel unchanged.
There is even a function PSD_APP_COMM in the PSD that could exactly meet this purpose (in was originally intended to be used for Ring-3 apps but that was never/is no more implemented). By the way: Pasha had originally implemented those functions in ACPI.PSD.

Lars

[Herwig Bauernfeind]

Hello,

as the porters of Virtualbox for OS/2, we are certainly interested in providing proper hardware assisted VT. However, providing these in the kernel does limit the possible customer base considerably, as many customers simply will refuse to use OS/4 kernel (we do not need to repeat the discussion about that, it is a fact beyond our scope of influence).

Nevertheless, we are interested in testing and discussion of the whole issue.

Please, drop me a note, on how to procede.

Thank you!

Herwig Bauernfeind
bww bitwiseworks
 

[Oxyd]

2) ACPI as such has no relation to it, yes. But a PSD is a good place to put additional platform functionality to leave the kernel unchanged.

If the kernel does not need to be updated and all problems can be solved through PSD, why the same IBM released the updated kernels, with each new fixpack and each new version of the OS? Moreover, remember amount of test kernels on testcase? Probably architectural problems can not be solved with the external driver.

[Lars]

Good question why IBM did not follow. I guess they broke their own concept.
The kernel should be as lean as possible.
All platform specific stuff should be in the PSD. It already contains the specific code for "end of interrupt", setting interrupt status and in-service registers, sending inter-processor interrupts and a few more. Therefore it is the ideal place to put such code and as I said it has already been done.

[Oxyd]

Good question why IBM did not follow. I guess they broke their own concept.
The kernel should be as lean as possible.

We have monolitic kernel. :-( Many programming interfaces, extend API, and even the hardware support (HPET for example), unfortunately, can not be obtained through an external driver. Also legacy IBM kernels have many errors, and not optimized for modern hardware.

[Dave Yeo]

If the kernel does not need to be updated and all problems can be solved through PSD, why the same IBM released the updated kernels, with each new fixpack and each new version of the OS? Moreover, remember amount of test kernels on testcase? Probably architectural problems can not be solved with the external driver.

Remember that Warp V4 and later w kernels didn't use a PSD.

[Sergey Posokhov]

KEERET   KEEENTRY KernInvokeAtEachCpu(RendCallBack_t CallBack,void *arg,uint32_t flags);
KEERET   KEEENTRY KernInvokeAtSpecificCpu(RendCallBack_t CallBack,void *arg,uint32_t CPUnum);

I would like to see names like:

KEERET   KEEENTRY KernRegisterCallback(RendCallBack_t CallBack, void *arg, uint32_t flags, uint32_t CPUnum);

Guess why.

[Lars]

If the kernel does not need to be updated and all problems can be solved through PSD, why the same IBM released the updated kernels, with each new fixpack and each new version of the OS? Moreover, remember amount of test kernels on testcase? Probably architectural problems can not be solved with the external driver.

Remember that Warp V4 and later w kernels didn't use a PSD.

Then you will be delighted to hear that since about 14.104 or so the W4 kernels can also load a PSD !
You can try out (and switch to native FPU exception handling):

http://hobbes.nmsu.edu/download/pub/os2/system/drivers/misc/os2pcat3.wpi

It's a PSD that I have written to fix some prob with the W4 kernel and to switch over using native FPU exception reporting (instead of going via IRQ13 on the external interrupt controller). It was meant as a personal learning excercise.
It also comes with full source code (well there is not much ...).

Lars

[Oxyd]

If the kernel does not need to be updated and all problems can be solved through PSD, why the same IBM released the updated kernels, with each new fixpack and each new version of the OS? Moreover, remember amount of test kernels on testcase? Probably architectural problems can not be solved with the external driver.

Remember that Warp V4 and later w kernels didn't use a PSD.

UNI? Or SMP kernels?