Enabling SMP

On Sun, 7 Aug 2011 03:20:36 UTC in comp.os.os2.apps, Dave Yeo

I couldn't find a p5-pe-5m on google but a p5-p5-vm seems to be the one that
Asus still mention on their web site. With the right BIOS it seems that this can
support up to Core 2 Duo E6700 chips which would probably be a good upgrade from
the Pentium D.

Post by Dave Yeo
So naturally wanted to enable SMP. Got the 105 SMP kernel and loader and
corresponding doscall1 from FP 6 as well as OS2APIC.PSD and tried it.

My \os2\install\smp directory contains the following files (plus a bunch of
trace formatter files that aren't required)

29-12-04 11:25a 144631 0 DOSCALL1.DLL
29-12-04 11:23a 908505 54 OS2KRNL
29-12-04 12:18p 43008 0 OS2LDR

Post by Dave Yeo
The boot went fine, detected both processors, until the end when it
should have gone graphic where I got a TRAP E in DSPRES.

There should be a duplicate copy of dspres.dll in \os2\install\vga so it might
be worth looking at that to see if it is the same as the one you have installed
now in \os2\dll.

Post by Dave Yeo
everything APM related, also didn't help. The Snap problems.txt does say
it isn't SMP safe but lots of people don't seem to have a problem.

Which version of SNAP are you using? The latest is 3.1.8

--
Trevor Hemsley, Brighton, UK
Trevor dot Hemsley at ntlworld dot com

Dave Yeo

2011-08-07 18:23:30 UTC

Post by Trevor Hemsley
On Sun, 7 Aug 2011 03:20:36 UTC in comp.os.os2.apps, Dave Yeo

Yes, I made a mistake reading the boot screen, goes by fast.
And upgrading to a Core 2 Duo does sound like a good upgrade. This 2.8
Ghz P4 is barely faster then the 2 Ghz Athlon it replaced.

Post by Dave Yeo
So naturally wanted to enable SMP. Got the 105 SMP kernel and loader and
corresponding doscall1 from FP 6 as well as OS2APIC.PSD and tried it.

My \os2\install\smp directory contains the following files (plus a bunch of
trace formatter files that aren't required)
29-12-04 11:25a 144631 0 DOSCALL1.DLL
29-12-04 11:23a 908505 54 OS2KRNL
29-12-04 12:18p 43008 0 OS2LDR

I've got the corresponding files from the leaked fp6. Does DOSCALL1 have
to match the kernel?

Post by Dave Yeo
The boot went fine, detected both processors, until the end when it
should have gone graphic where I got a TRAP E in DSPRES.

There should be a duplicate copy of dspres.dll in \os2\install\vga so it might
be worth looking at that to see if it is the same as the one you have installed
now in \os2\dll.

The one in \os2\install\vga was the same but I did find an older one.
The system wouldn't go graphical with the W4 kernel and still trapped
with the SMP kernel.

Post by Dave Yeo
everything APM related, also didn't help. The Snap problems.txt does say
it isn't SMP safe but lots of people don't seem to have a problem.

Which version of SNAP are you using? The latest is 3.1.8

The latest.
I also tried a couple of different kernels, 104a smp and the OS/4 one.
Both hung during the boot screen with the hard drive light on.
I then got the idea of testing on my bootos2 partition as it is very
simple and just VGA. The system booted but as soon as I typed anything
it trapped E in PMSHELL.
Should probably update the BIOS next
Dave

Trevor Hemsley

2011-08-07 19:13:44 UTC

On Sun, 7 Aug 2011 18:23:30 UTC in comp.os.os2.apps, Dave Yeo

Post by Trevor Hemsley
29-12-04 11:25a 144631 0 DOSCALL1.DLL
29-12-04 11:23a 908505 54 OS2KRNL
29-12-04 12:18p 43008 0 OS2LDR

I've got the corresponding files from the leaked fp6. Does DOSCALL1 have
to match the kernel?

I don't think it matters but you do definitely need to have the right doscall1
and the right os2ldr to go with the SMP kernel, not sure they need to match
levels althought it's probably best if they do. The 104a kernel seems to be the
most reliable as far as I can see.

Post by Dave Yeo
I also tried a couple of different kernels, 104a smp and the OS/4 one.
Both hung during the boot screen with the hard drive light on.

What version of danis506.add do you have?

--
Trevor Hemsley, Brighton, UK
Trevor dot Hemsley at ntlworld dot com

Dave Yeo

2011-08-07 21:28:16 UTC

Post by Dave Yeo
I also tried a couple of different kernels, 104a smp and the OS/4 one.

Post by Dave Yeo
Both hung during the boot screen with the hard drive light on.

What version of danis506.add do you have?

E:\OS2\BOOT>bldlevel DaniS506.ADD
Build Level Display Facility Version 6.10.480 Oct 6 2000
(C) Copyright IBM Corporation 1993-2000
Signature: @#DANI:1.8#@##1## 18.4.2008 16:56:00
NACHTIGALL:
:::4::@@ Adapter Driver for PATA/SATA DASD
Vendor: DANI
Revision: 1.08
Date/Time: 18.4.2008 16:56:00
Build Machine: NACHTIGALL
File Version: 1.8.4
Description: Adapter Driver for PATA/SATA DASD

Dave

Trevor Hemsley

2011-08-09 22:57:36 UTC

On Sun, 7 Aug 2011 21:28:16 UTC in comp.os.os2.apps, Dave Yeo

Post by Dave Yeo
I also tried a couple of different kernels, 104a smp and the OS/4 one.

Post by Dave Yeo
Both hung during the boot screen with the hard drive light on.

What version of danis506.add do you have?

E:\OS2\BOOT>bldlevel DaniS506.ADD
Build Level Display Facility Version 6.10.480 Oct 6 2000
(C) Copyright IBM Corporation 1993-2000
Vendor: DANI
Revision: 1.08
Date/Time: 18.4.2008 16:56:00
Build Machine: NACHTIGALL
File Version: 1.8.4
Description: Adapter Driver for PATA/SATA DASD

There are definitely newer versions of that available. 1.8.7 is on hobbes, 1.8.9
is included in eCS 2.1 and I've just been reading the netlabs wiki pages that
says 1.8.11 of xATA has been released (which I think includes this driver as
well as os2ahci). No idea if this is your problem but it might be worth trying
an update.

--
Trevor Hemsley, Brighton, UK
Trevor dot Hemsley at ntlworld dot com

Dave Yeo

2011-08-10 03:39:35 UTC

Post by Trevor Hemsley
On Sun, 7 Aug 2011 21:28:16 UTC in comp.os.os2.apps, Dave Yeo

Post by Dave Yeo
I also tried a couple of different kernels, 104a smp and the OS/4 one.

Post by Dave Yeo
Both hung during the boot screen with the hard drive light on.

What version of danis506.add do you have?

I'll have to try updating as the wiki mentioned spin lock fixes for SMP
systems, though that may be unreleased yet.
I think that the DASD wouldn't result in such a quick crash. On my
maintainence partition all I have to do is touch a key to get an instant
trap E.
First will be updating the BIOS
Dave

Dave Yeo

2011-08-11 14:37:36 UTC

Post by Trevor Hemsley
On Sun, 7 Aug 2011 21:28:16 UTC in comp.os.os2.apps, Dave Yeo

Post by Dave Yeo
I also tried a couple of different kernels, 104a smp and the OS/4 one.

Post by Dave Yeo
Both hung during the boot screen with the hard drive light on.

What version of danis506.add do you have?

I updated to the one on Netlabs and updated my BIOS. No change, I have
to assume the tables are bad
Dave

Jonathan de Boyne Pollard

2011-08-09 08:45:09 UTC

Does DOSCALL1 have to match the kernel?

On general principles, given what that DLL is, the answer to that should
always be taken to be "Yes." unless one has positive information to the
contrary.

Dave Yeo

2011-08-10 03:42:27 UTC

Does DOSCALL1 have to match the kernel?

On general principles, given what that DLL is, the answer to that should
always be taken to be "Yes." unless one has positive information to the
contrary.

What I've found on this system is that only the 105 SMP kernel will get
past the boot logo, even with the PSD statement remmed out.
Others have reported mixing close versions together successfully and
IIRC when the testcase kernels were being regularly released for
testing, not once was a doscall.dll released.
Dave

Jonathan de Boyne Pollard

2011-08-10 09:26:14 UTC

Does DOSCALL1 have to match the kernel?

On general principles, given what that DLL is, the answer to that should
always be taken to be "Yes." unless one has positive information to the
contrary.

Others have reported mixing close versions together successfully and
IIRC when the testcase kernels were being regularly released for
testing, not once was a doscall.dll released.

Nonetheless the general principle stands. Yes, it *might* be the case
that things *happen to be* compatible. And yes, there's the implied
compatibility of a patched or updated kernel with the thing that it is
replacing. But on any system the lowest-level system call library and
the kernel providing that system call interface are best considered
tightly coupled to each other without positive information to the contrary.

But this is an aside from your situation, since the SciTech statement is
fairly definite and unequivocal. It translates, upon my reading it, to
"We haven't put the interlocks in." or "We have put the interlocks in
but we've found in testing that they aren't quite right.".

Post by Dave Yeo
What I've found on this system is that only the 105 SMP kernel
will get past the boot logo, even with the PSD statement remmed out.

Do you have the option of simply not using the SNAP drivers?

Steve Wendt

2011-08-11 02:54:26 UTC

Post by Dave Yeo
What I've found on this system is that only the 105 SMP kernel
will get past the boot logo, even with the PSD statement remmed out.

Do you have the option of simply not using the SNAP drivers?

What does that have to do with not getting past the boot logo?

Jonathan de Boyne Pollard

2011-08-11 07:18:48 UTC

Post by Steve Wendt

Post by Dave Yeo
What I've found on this system is that only the 105 SMP kernel
will get past the boot logo, even with the PSD statement remmed out.

Do you have the option of simply not using the SNAP drivers?

What does that have to do with not getting past the boot logo?

Go and read M. Yeo's first message in this very thread, and you'll find out.

Dave Yeo

2011-08-11 14:32:10 UTC

Post by Jonathan de Boyne Pollard
Do you have the option of simply not using the SNAP drivers?

Attempting to enable SMP on my maintainance partition, which is very
minimal including only VGA, the desktop comes up but as soon as I type
anything I get a trap E, often in PMSHELL. Pressing ctrl-alt-del causes
more trap E
Dave

Jonathan de Boyne Pollard

2011-08-12 10:46:24 UTC

Post by Jonathan de Boyne Pollard
Do you have the option of simply not using the SNAP drivers?

Ouch! That says that there's more than just the SNAP drivers at work
here. What does moving the mouse do? Is this with a USB keyboard? If
it's not with a USB keyboard, what happens if you use a USB keyboard and
USB drivers?

Dave Yeo

2011-08-12 14:54:33 UTC

Post by Jonathan de Boyne Pollard
Do you have the option of simply not using the SNAP drivers?

Currently no USB drivers, even though the BIOS is set to use the USB
mouse in compatible mode, which should emulate a PS/2 mouse, I don't
even get a mouse cursor.
It's a PS/2 keyboard. I'll test by adding the USB drivers and try the
USB keyboard. I'd like to know how it works under my regular system anyways.
Dave

Jonathan de Boyne Pollard

2011-08-15 15:07:34 UTC

I don't even get a mouse cursor.

Let's eliminate some possibilities. Try /P=1 on the OS2APIC.PSD command
line.

Dave Yeo

2011-08-15 23:45:41 UTC

I don't even get a mouse cursor.

Let's eliminate some possibilities. Try /P=1 on the OS2APIC.PSD command
line.

Works fine with the 105 kernel.
Dave

Jonathan de Boyne Pollard

2011-08-16 10:35:25 UTC

I don't even get a mouse cursor.

Let's eliminate some possibilities. Try /P=1 on the OS2APIC.PSD command
line.

Works fine with the 105 kernel.

I presume that that means that you get a mouse pointer, you don't get
any TRAPs, and the system works pretty much as it did when you used the
uniprocessor kernel. That's what I suspected. Blast!

My educated guess is that either spinlocks or IPIs are not working
properly on your system, because of some chip variation that IBM didn't
account for. The hypothesis still fits all of the observed behaviour.
But I cannot come up with a mechanism for _how_. I'm going to have a
look at the specifications for the Pentium D.

Dave Yeo

2012-01-07 21:08:53 UTC

Post by Trevor Hemsley
On Sun, 7 Aug 2011 03:20:36 UTC in comp.os.os2.apps, Dave Yeo

Yes, I meant p5-pe-vm. Anyways I had to upgrade to eCS 2.1 to get SMP
working and I just picked up a Core 2 Duo E6300 which was all I could
find locally for a reasonable price ($30). The E6300 halved the time to
compile Firefox, from 2.5 hours (3.5 hrs without SMP) to 1.25 hours.
Quite amazing considering the Core 2 Duo is only 1.86 Ghz compared to
the 2.8 Ghz P4
Dave

Shmuel (Seymour J.) Metz

2012-01-08 00:36:13 UTC

Post by Dave Yeo
Yes, I meant p5-pe-vm. Anyways I had to upgrade to eCS 2.1 to get SMP
working

I got it working[1] in 2.0 GA, but only by ignoring the recommendation
from the SMP wizard and using /SMP rather than /SMP /APIC.

On an AMD 2-core.

--
Shmuel (Seymour J.) Metz, SysProg and JOAT <http://patriot.net/~shmuel>

Unsolicited bulk E-mail subject to legal action. I reserve the
right to publicly post or ridicule any abusive E-mail. Reply to
domain Patriot dot net user shmuel+news to contact me. Do not
reply to ***@library.lspace.org

Dave Yeo

2012-01-08 01:51:05 UTC

Post by Shmuel (Seymour J.) Metz

Post by Dave Yeo
Yes, I meant p5-pe-vm. Anyways I had to upgrade to eCS 2.1 to get SMP
working

I got it working[1] in 2.0 GA, but only by ignoring the recommendation
from the SMP wizard and using /SMP rather than /SMP /APIC.

I was trying to get it working with Warp V4 and the old OS2APIC.PSD.
Always crashed when I did anything
Dave

Trevor Hemsley

2012-01-08 02:24:54 UTC

On Sat, 7 Jan 2012 21:08:53 UTC in comp.os.os2.apps, Dave Yeo

Post by Trevor Hemsley
On Sun, 7 Aug 2011 03:20:36 UTC in comp.os.os2.apps, Dave Yeo

Pentium D's had a dreadful reputation for being a bit crap :-) You've skipped
over the Pentium 4 that followed it (and was also a bit of a dud) and then
skipped to the model of chip that followed that which was the start of Intel's
recent dominance vs AMD. Shame you could only find the E6300 - both the E6600
and E6700 would have cut the time to under an hour if it scaled with clockspeed.

--
Trevor Hemsley, Brighton, UK
Trevor dot Hemsley at ntlworld dot com

MrG

2011-08-07 17:26:20 UTC

Dave, I have a 3.2 GHz Pentium D on an Asus P5B-E board. In os2apic.psd do you have the switch /apic in your config.sys. On my machine, this causes a trap also. Try it without the /apic switch. That worked for me.

Dave Yeo

2011-08-07 18:12:43 UTC

Post by MrG
Dave, I have a 3.2 GHz Pentium D on an Asus P5B-E board.
In os2apic.psd do you have the switch /apic in your config.sys.
On my machine, this causes a trap also. Try it without the /apic switch.
That worked for me.

I tried the /APIC switch which resulted in a trap D instead of a trap e
so removed it.
Dave
ps The newsserver I'm using threw an error about lines over 79 chars,
your message was all one line.

MrG

2011-08-07 17:36:32 UTC

Dave, I also found that Uniaud drivers would not load if plug and play is set to yes in bios. Set to no and Uniaud loads.

Dave Yeo

2011-08-07 18:14:23 UTC

Post by MrG
Dave, I also found that Uniaud drivers would not load if plug and play
is set to yes in bios. Set to no and Uniaud loads.

Interesting, the uniaud drivers here sometimes do not load but usually
do. There is also a long pause while the second driver loads. I'll try
changing the bios later
Dave

MrG

2011-08-07 17:33:03 UTC

After getting SMP running on my 3.2 GHz Pentium D, Asus P5B-E, I've run into two minor glitches. 1) Machine won't power off on shut down. 2) HDD activity light stays on constantly although there is no disk activity. Ran tests on the drives and they are good.
Any thoughts on how I can correct these items?

MrG

2011-08-07 22:12:24 UTC

Post by Dave Yeo
ps The newsserver I'm using threw an error about lines over 79 chars,
your message was all one line.

Sorry about that.
It's supposed to wrap automagically at 72 chars.
Right now I can only post through google groups and
they are really screwed up for the last week or so.
Posts are not showing up at all or 2 to 3 days later.
They show up on usenet though.
I read the articles from a non posting server using ProNews/2.
Sometimes I copy/paste the quoted message,
sometimes I'm lazy.
BTW, for what its worth, I'm using the 104a_SMP krnl
and the 104a_SMP OS2LDR with doscall1.dll from the unofficial
FP6, so I don't think it matters which doscall1 you use,
as long as it's the SMP version.

Dave Yeo

2011-08-07 22:51:15 UTC

Post by MrG

Post by Dave Yeo
ps The newsserver I'm using threw an error about lines over 79 chars,

Post by Dave Yeo
your message was all one line.

Do you need a posting news server? I've been using aioe (aioe.org) for
quite a while without problems. Some simple rules mostly etiquette
related like the line length thing but works fine.
Dave

Mr. G

2011-08-08 17:01:11 UTC

Post by MrG

Post by Dave Yeo
ps The newsserver I'm using threw an error about lines over 79 chars,

Post by Dave Yeo
your message was all one line.

Do you need a posting news server? I've been using aioe (aioe.org) for
quite a while without problems. Some simple rules mostly etiquette
related like the line length thing but works fine.
Dave

I used to use that server, but the reason I quit escapes me.
I'll give it another try. Thanks for the tip.

John Small

2011-09-17 10:38:46 UTC

On Sun, 7 Aug 2011 22:51:15 UTC, Dave Yeo <***@gmail.com>
wrote:

I am searching for new newsservers since my new ISP does not provide
one.

Post by Dave Yeo
Do you need a posting news server? I've been using aioe (aioe.org) for
quite a while without problems. Some simple rules mostly etiquette
related like the line length thing but works fine.

After reading this I tried out aioe.org.

I seem to be able to retrieve messages OK. But I tried posting through
aioe.org and my message was never relayed beyond aieo.org. So it only
users of aieo.org ever saw my message. Is this normal?

--
John Small

Peter Moylan

2011-09-17 13:15:55 UTC

Post by John Small
I am searching for new newsservers since my new ISP does not provide
one.

After reading this I tried out aioe.org.
I seem to be able to retrieve messages OK. But I tried posting through
aioe.org and my message was never relayed beyond aieo.org. So it only
users of aieo.org ever saw my message. Is this normal?

I use eternal-september.org whenever my ISP's news server has a glitch.
It has a good reputation. You have to register on their web site, but I
think that's just to block the spambots; there's no charge.

--
Peter Moylan, Newcastle, NSW, Australia. http://www.pmoylan.org
For an e-mail address, see my web page.

Steve

2011-09-17 13:43:05 UTC

Post by John Small
After reading this I tried out aioe.org.
I seem to be able to retrieve messages OK. But I tried posting through
aioe.org and my message was never relayed beyond aieo.org. So it only
users of aieo.org ever saw my message. Is this normal?

Hi,

I have been using aioe for news for quite a while. My
posts show up immediately on aioe even when posting to
a moderated group. However my posts do show up on
Google groups after some delay. Though I rarely see
responses to my posts, I have no idea if that is poor
content or poor dispersal.

Regards,

Steve N.

Allodoxaphobia

2011-09-17 21:56:39 UTC

Post by Steve

I have been using aioe for news for quite a while. My
posts show up immediately on aioe even when posting to
a moderated group. However my posts do show up on
Google groups after some delay. Though I rarely see
responses to my posts, I have no idea if that is poor
content or poor dispersal.

Add to that: poor acceptance. usenet has a low regard for aioe.org
due to the quantity of spam from there _and_ the number of
foaming-at-the-mouth trolls that do nym-shifting on those servers.
aioe.org no doubt appears in a lot of killfiles.

Jonesy

Dave Yeo

2011-09-17 16:15:14 UTC

Post by John Small
I am searching for new newsservers since my new ISP does not provide
one.

After reading this I tried out aioe.org.
I seem to be able to retrieve messages OK. But I tried posting through
aioe.org and my message was never relayed beyond aieo.org. So it only
users of aieo.org ever saw my message. Is this normal?

I've been posting through aioe.net for quite a while and the messages
seem to get through, at least others answer them.
Dave

Rich Walsh

2011-09-21 02:01:56 UTC

Post by John Small
I am searching for new newsservers since my new ISP does not provide
one.

After reading this I tried out aioe.org.
I seem to be able to retrieve messages OK. But I tried posting through
aioe.org and my message was never relayed beyond aieo.org. So it only
users of aieo.org ever saw my message. Is this normal?

Why don't you splurge and spend ten bucks on astraweb?

$10 buys 25gb of bandwidth with no time limit or speed limit, and up
to 50 simultaneous connections. If you're not into pirating movies
you'll have usenet for life. www.astraweb.com Love 'em...

--
== == almost usable email address: Rich AT E-vertise DOT Com == ==

Dariusz Piatkowski

2011-08-08 01:09:19 UTC

Dave,

Post by Dave Yeo
Probably the wrong newsgroup but after reading the doscall1 thread seams
good enough :)
Picked up a 2.8 Ghz Pentium D system (Asus P5-PE 5M MB) with a hosed
Windows install for $5 bucks. Booted a live Linux CD and stress tested
it a bit then ran memtest, the hardware seems fine. Moved my hard drives
to it and OS/2 is running fine besides a problem with sound and having
to change the USB drivers to suite the hardware.
So naturally wanted to enable SMP. Got the 105 SMP kernel and loader and
corresponding doscall1 from FP 6 as well as OS2APIC.PSD and tried it.
The boot went fine, detected both processors, until the end when it
should have gone graphic where I got a TRAP E in DSPRES.
Anyone have any ideas to work around this? I patched testcfg.sys as Rich
mentioned above and also removed it. Didn't help. Changed the snoop list
as Rich mentioned and lost my ps2 keyboard and Uniaud. Also remmed out
everything APM related, also didn't help. The Snap problems.txt does say
it isn't SMP safe but lots of people don't seem to have a problem.
Dave

Are you by any chance running JFS?

I ask b/c I have FP6 installed here. Am also running SMP on my 4 core AMD Phenom
II X956 CPU, however I'm using:

PSD=OS2APIC.PSD /P=4 /APIC

...and if I attempt to boot up with JFS as well that produces a TRAP.

I do have HPFS386 running, FAT32 as well, so a bunch of stuff going on at the
same time. My machine is NOT configured to support ACPI, however, setting my
BIOS to ACPI enabled and using MPS map 1.4 seems to produce an otherwise pretty
darn stable system. My only issue, for some time now, is that I will
occasionally get a soft-freeze only in certain applications, only when certain
function is being executed...CAD still shuts it down, but it it's a reboot
required solution. This I have not been able to track down yet.

Let's see...what else? Well, I do have latest SNAP drivers installed, the mb is
a MSI 790X-G45, using DANI 1.8.5 drivers. Peer drivers are loaded, but I am NOT
using this functionality...so one of my next steps is to try to rule this out as
a potential cause of the soft-freeze.

Dave Yeo

2011-08-08 04:22:49 UTC

Post by Dariusz Piatkowski
Are you by any chance running JFS?

Yes, I'll test without it, perhaps without HPFS as well (I've switched
to testing on my maintainence partition as it is minimal to begin with)

Post by Dariusz Piatkowski
I ask b/c I have FP6 installed here. Am also running SMP on my 4 core AMD Phenom
PSD=OS2APIC.PSD /P=4 /APIC
...and if I attempt to boot up with JFS as well that produces a TRAP.
I do have HPFS386 running, FAT32 as well, so a bunch of stuff going on at the
same time. My machine is NOT configured to support ACPI, however, setting my
BIOS to ACPI enabled and using MPS map 1.4 seems to produce an otherwise pretty
darn stable system. My only issue, for some time now, is that I will
occasionally get a soft-freeze only in certain applications, only when certain
function is being executed...CAD still shuts it down, but it it's a reboot
required solution. This I have not been able to track down yet.

So you have ACPI disabled even though having it enabled is more stable?
I've always had sot-freezes after a week to 10 days of uptime so I just
try to reboot weekly.
Dave

Dariusz Piatkowski

2011-08-09 03:09:16 UTC

Hi Dave!

Post by Dariusz Piatkowski
Are you by any chance running JFS?

Yes, I'll test without it, perhaps without HPFS as well (I've switched
to testing on my maintainence partition as it is minimal to begin with)

So you have ACPI disabled even though having it enabled is more stable?
I've always had sot-freezes after a week to 10 days of uptime so I just
try to reboot weekly.
Dave

Yeah...strangely enough as it is. The only way I can see why something may be
coming to play is that with ACPI enabled in BIOS the IOAPIC is also turned ON.
When I disable ACPI, the IOAPIC option disappears....so I'm suspecting it may be
actually be getting shut OFF internally...however I am not sure, nor do I know
how to check.

But...if IOAPIC is helping OS2 manage the SMP somehow (I am running with the
/APCI option in my OS2APIC driver) so be it.

Jonathan de Boyne Pollard

2011-08-09 09:39:25 UTC

Post by Dariusz Piatkowski
My machine is NOT configured to support ACPI, however, setting my
BIOS to ACPI enabled and using MPS map 1.4 seems to produce an
otherwise pretty darn stable system. [...]

So you have ACPI disabled even though having it enabled is more stable?

The last time around, M. Piatkowski said APIC, not ACPI. I pointed out
then that xe shouldn't be mixing up xyr initialisms. Disabling the
Advanced Configuration and Power Interface is a very different kettle of
fish to disabling any use of Advanced Programmable Interrupt
Controllers. It's actually the latter that M. Piatkowski is doing.

In an AMIBIOS machine, the setting that M. Piatkowski was talking about
earlier is named "ACPI APIC Support" and it deals with APICs. In
AWARD-Phoenix firmwares, the equivalent setting is named "APIC Mode".
It almost goes without saying that if one disables the use of APICs
entirely, by setting APIC mode to "disabled", then I/O APICs are as a
consequence unavailable. If all local APICs are switched off during
POST (as essentially all that this setting has to do), then whatever I/O
APICs may do is entirely irrelevant: there is no-one listening to them.

Peter Flass

2011-08-09 12:08:37 UTC

So you have ACPI disabled even though having it enabled is more stable?

The last time around, M. Piatkowski said APIC, not ACPI. I pointed out
then that xe shouldn't be mixing up xyr initialisms. Disabling the
Advanced Configuration and Power Interface is a very different kettle of
fish to disabling any use of Advanced Programmable Interrupt
Controllers. It's actually the latter that M. Piatkowski is doing.
In an AMIBIOS machine, the setting that M. Piatkowski was talking about
earlier is named "ACPI APIC Support" and it deals with APICs. In
AWARD-Phoenix firmwares, the equivalent setting is named "APIC Mode". It
almost goes without saying that if one disables the use of APICs
entirely, by setting APIC mode to "disabled", then I/O APICs are as a
consequence unavailable. If all local APICs are switched off during POST
(as essentially all that this setting has to do), then whatever I/O
APICs may do is entirely irrelevant: there is no-one listening to them.

It's unfortunate that these two acronyms are so easily confused.

Jonathan de Boyne Pollard

2011-08-10 09:31:08 UTC

Post by Peter Flass
It's unfortunate that these two acronyms are so easily confused.

Especially when, as it seems here, conversation is about to turn to the
way that Advanced Programmable Interrupt Controllers are dealt with in
the Advanced Configuration and Power Interface. (-:

Dariusz Piatkowski

2011-08-10 11:29:14 UTC

On Wed, 10 Aug 2011 09:31:08 UTC, Jonathan de Boyne Pollard

Post by Peter Flass
It's unfortunate that these two acronyms are so easily confused.

Especially when, as it seems here, conversation is about to turn to the
way that Advanced Programmable Interrupt Controllers are dealt with in

Alright now...LOL, my MSI BIOS does what it does...you may have whatever ideas
you have on how things should work...and I don't even necessairly disagree with
you, however, it is what it is.

In my MSI mb BIOS, as I previously described, the IOAPIC option ONLY shows up in
the menu when the ACPI mode is enabled. End of story.

C'mon man...accept that a vendor has chosen to implement something in a
particular way and deal with it...don't make that fact an overarching difficulty
for our conversation here. For the most part anyways you are merely pointing out
how you think BIOS settings should be, but I do not see any real examples you
are using to illustrate to us why and how it impacts OS2...

Jonathan de Boyne Pollard

2011-08-15 15:36:19 UTC

Post by Dariusz Piatkowski

Post by Peter Flass
It's unfortunate that these two acronyms are so easily confused.

Especially when, as it seems here, conversation is about to turn to the
way that Advanced Programmable Interrupt Controllers are dealt with in

Alright now...LOL, my MSI BIOS does what it does...you may have whatever ideas
you have on how things should work...and I don't even necessairly disagree with
you, however, it is what it is.
In my MSI mb BIOS, as I previously described, the IOAPIC option ONLY shows up in
the menu when the ACPI mode is enabled. End of story.
C'mon man...accept that a vendor has chosen to implement something in a
particular way and deal with it...don't make that fact an overarching difficulty
for our conversation here. For the most part anyways you are merely pointing out
how you think BIOS settings should be, but I do not see any real examples you
are using to illustrate to us why and how it impacts OS2...

That's a load of old twaddle, ascribing to me things that I never wrote.

Dariusz Piatkowski

2011-08-15 21:47:52 UTC

On Mon, 15 Aug 2011 15:36:19 UTC, Jonathan de Boyne Pollard

Post by Dariusz Piatkowski

Post by Peter Flass
It's unfortunate that these two acronyms are so easily confused.

Especially when, as it seems here, conversation is about to turn to the
way that Advanced Programmable Interrupt Controllers are dealt with in

Alright now...LOL, my MSI BIOS does what it does...you may have whatever ideas
you have on how things should work...and I don't even necessairly disagree with
you, however, it is what it is.
In my MSI mb BIOS, as I previously described, the IOAPIC option ONLY shows up in
the menu when the ACPI mode is enabled. End of story.
C'mon man...accept that a vendor has chosen to implement something in a
particular way and deal with it...don't make that fact an overarching difficulty
for our conversation here. For the most part anyways you are merely pointing out
how you think BIOS settings should be, but I do not see any real examples you
are using to illustrate to us why and how it impacts OS2...

That's a load of old twaddle, ascribing to me things that I never wrote.

Look...I have no quarrel with you...honestly, and I already said this, I
seriously appreciate the input folks like yourself can provide to the community.

My response was simply to address something you DID write in one of your posts
stating how things simply didn't work the way I was posting about
them...meanwhile, all I did was to try to re-create as closely as I could the
screens my BIOS showed...heck, I even found the best matching posts of such
screens from another MSI motherboard to illustrate what I was talking about and
included the URLs in my post for reference. Meanwhile your response was to say
how MSI got it wrong and we were therefore all confused...LOL!

YES...I accept the fact that if MSI had the wrong description of a BIOS function
then it certainly made me misunderstand the functionality...to correct this I
ask the OS community...I welcome all such feedback.

Jonathan de Boyne Pollard

2011-09-20 11:35:58 UTC

Post by Dariusz Piatkowski

Post by Jonathan de Boyne Pollard
That's a load of old twaddle, ascribing to me things that I never wrote.

Look...I have no quarrel with you...honestly, and I already said this, I
seriously appreciate the input folks like yourself can provide to the community.

I know. Nonetheless, you're ascribing to me things that I never wrote.
In fact, you're *still* doing that even now.

Post by Dariusz Piatkowski
Meanwhile your response was to say how MSI got it wrong and
we were therefore all confused...LOL!

This is a case in point. I never wrote any such thing. I wrote that
*you* interchanging ACPI and APIC confused people.

Dariusz Piatkowski

2011-08-09 21:36:12 UTC

Jonathan,

On Tue, 9 Aug 2011 09:39:25 UTC, Jonathan de Boyne Pollard

So you have ACPI disabled even though having it enabled is more stable?

The last time around, M. Piatkowski said APIC, not ACPI. I pointed out
then that xe shouldn't be mixing up xyr initialisms. Disabling the
Advanced Configuration and Power Interface is a very different kettle of
fish to disabling any use of Advanced Programmable Interrupt
Controllers. It's actually the latter that M. Piatkowski is doing.
In an AMIBIOS machine, the setting that M. Piatkowski was talking about
earlier is named "ACPI APIC Support" and it deals with APICs. In
AWARD-Phoenix firmwares, the equivalent setting is named "APIC Mode".
It almost goes without saying that if one disables the use of APICs
entirely, by setting APIC mode to "disabled", then I/O APICs are as a
consequence unavailable. If all local APICs are switched off during
POST (as essentially all that this setting has to do), then whatever I/O
APICs may do is entirely irrelevant: there is no-one listening to them.

Well...you did bring up an excellent point. I had read your original response to
my other post and had intended to respond...however I did encounter a
soft-freeze and wasn't able to find the posting anymore in my news reader...???

So here is the response, hopefully if I used the wrong terminology last time I
am getting it correct this time...I am in fact using the correct terms as
applicable to my hardware and in my descriptions I am addressing 2 separate
issues:

1) ACPI - replacing APM for example
2) APIC - dealing with SMP support and things like advanced programmable IO
controllers

In the BIOS of my motherboard, which is the MSI 790X-G45 piece, I do in fact
have 2 such separate settings.

1) Enabling ACPI also appears to enable the IOAPIC (that's correct), disabling
ACPI shuts OFF IOAPIC

2) SMP related stuff, I believe is really just the MPS table selection of 1.1 or
1.4

What I do not know is how enabling ACPI turns on the IOAPIC and how that in turn
relates to supporting SMP in OS2.

Here is a link to a mb review, and while it's not the matching model, it is very
close: MSI 790FX-GD70, the BIOS screenshots are very close (not the same
ofcourse for all settings) to what I see for my 790X-G45,
http://www.hardocp.com/article/2009/08/04/msi_790fxgd70_amd_motherboard/

Specifically, this is AMI BIOS and shows the following:

1)
http://www.hardocp.com/image.html?image=MTI0NTM3OTE2NFhVdWJNdTV0MzNfMl80X2wuZ2lm

...shows the IOAPIC function, it is ONLY present when ACPI has been turned ON

2)
http://www.hardocp.com/image.html?image=MTI0NTM3OTE2NFhVdWJNdTV0MzNfMl84X2wuZ2lm

...shows the ACPI controls

OK...so now given that information, I stand by my earlier conclusion...how does
ACPI control the IOAPIC??? Unless, the IOAPIC is only used to support the ACPI
functionality and has absolutely nothing to do with SMP.

Dariusz Piatkowski

2011-08-10 03:03:36 UTC

On Tue, 9 Aug 2011 21:36:12 UTC, "Dariusz Piatkowski"

Post by Dariusz Piatkowski
Jonathan,
On Tue, 9 Aug 2011 09:39:25 UTC, Jonathan de Boyne Pollard

So you have ACPI disabled even though having it enabled is more stable?

The last time around, M. Piatkowski said APIC, not ACPI. I pointed out
then that xe shouldn't be mixing up xyr initialisms. Disabling the
Advanced Configuration and Power Interface is a very different kettle of
fish to disabling any use of Advanced Programmable Interrupt
Controllers. It's actually the latter that M. Piatkowski is doing.
In an AMIBIOS machine, the setting that M. Piatkowski was talking about
earlier is named "ACPI APIC Support" and it deals with APICs. In
AWARD-Phoenix firmwares, the equivalent setting is named "APIC Mode".
It almost goes without saying that if one disables the use of APICs
entirely, by setting APIC mode to "disabled", then I/O APICs are as a
consequence unavailable. If all local APICs are switched off during
POST (as essentially all that this setting has to do), then whatever I/O
APICs may do is entirely irrelevant: there is no-one listening to them.

Well...you did bring up an excellent point. I had read your original response to
my other post and had intended to respond...however I did encounter a
soft-freeze and wasn't able to find the posting anymore in my news reader...???
So here is the response, hopefully if I used the wrong terminology last time I
am getting it correct this time...I am in fact using the correct terms as
applicable to my hardware and in my descriptions I am addressing 2 separate
1) ACPI - replacing APM for example
2) APIC - dealing with SMP support and things like advanced programmable IO
controllers
In the BIOS of my motherboard, which is the MSI 790X-G45 piece, I do in fact
have 2 such separate settings.
1) Enabling ACPI also appears to enable the IOAPIC (that's correct), disabling
ACPI shuts OFF IOAPIC
2) SMP related stuff, I believe is really just the MPS table selection of 1.1 or
1.4
What I do not know is how enabling ACPI turns on the IOAPIC and how that in turn
relates to supporting SMP in OS2.
Here is a link to a mb review, and while it's not the matching model, it is very
close: MSI 790FX-GD70, the BIOS screenshots are very close (not the same
ofcourse for all settings) to what I see for my 790X-G45,
http://www.hardocp.com/article/2009/08/04/msi_790fxgd70_amd_motherboard/
1)
http://www.hardocp.com/image.html?image=MTI0NTM3OTE2NFhVdWJNdTV0MzNfMl80X2wuZ2lm
...shows the IOAPIC function, it is ONLY present when ACPI has been turned ON
2)
http://www.hardocp.com/image.html?image=MTI0NTM3OTE2NFhVdWJNdTV0MzNfMl84X2wuZ2lm
...shows the ACPI controls
OK...so now given that information, I stand by my earlier conclusion...how does
ACPI control the IOAPIC??? Unless, the IOAPIC is only used to support the ACPI
functionality and has absolutely nothing to do with SMP.

...something else which I looked up in the BIOS this time, the description for
hte IOAPIC says the following:

'Include ACPI APIC table pointer to RSDT pointer list'

Now...does that make sense, and can you shed some light on this? Google is my
next stop...LOL...

Thanks!

Jonathan de Boyne Pollard

2011-08-10 07:20:16 UTC

Post by Dariusz Piatkowski
...something else which I looked up in the BIOS this time, the
'Include ACPI APIC table pointer to RSDT pointer list'
Now...does that make sense, and can you shed some light on this?

No, it doesn't make sense. It tells me that your SETUP utility was put
together in a fairly slapdash manner, with the wrong text attached to
the setting. Here is a correctly selected, albeit not quite English,

Post by Dariusz Piatkowski
IOAPIC Function
This field is used to enable or disable the APIC (Advanced
Programmable Interrupt Controller). Due to compliance with PC2001
design guide, the system is able to run in APIC mode. Enabling APIC
mode will expand available IRQ resources for the system.
Settings: [Enabled], [Disabled].

It expands it by two (three for some chipsets), and even then only if
the local APICs are enabled.

Lars Erdmann

2011-08-13 09:32:21 UTC

What switches did you try with OS2APIC.PSD ?

Try
OS2APIC.PSD /APIC (symmetric
mode)

or

OS2APIC /PIC=LINT0 /NMI=LINT1 /PREC=DLI (virtual wire mode)

Lars

Post by Dave Yeo
Probably the wrong newsgroup but after reading the doscall1 thread seams
good enough :)
Picked up a 2.8 Ghz Pentium D system (Asus P5-PE 5M MB) with a hosed
Windows install for $5 bucks. Booted a live Linux CD and stress tested it
a bit then ran memtest, the hardware seems fine. Moved my hard drives to
it and OS/2 is running fine besides a problem with sound and having to
change the USB drivers to suite the hardware.
So naturally wanted to enable SMP. Got the 105 SMP kernel and loader and
corresponding doscall1 from FP 6 as well as OS2APIC.PSD and tried it. The
boot went fine, detected both processors, until the end when it should
have gone graphic where I got a TRAP E in DSPRES.
Anyone have any ideas to work around this? I patched testcfg.sys as Rich
mentioned above and also removed it. Didn't help. Changed the snoop list
as Rich mentioned and lost my ps2 keyboard and Uniaud. Also remmed out
everything APM related, also didn't help. The Snap problems.txt does say
it isn't SMP safe but lots of people don't seem to have a problem.
Dave

Dave Yeo

2011-08-13 18:08:39 UTC

Post by Lars Erdmann
What switches did you try with OS2APIC.PSD ?

Everyone I could think of without guessing NMI numbers. PLus the ones
mentioned in this thread.

Post by Lars Erdmann
Try
OS2APIC.PSD /APIC (symmetric
mode)
or
OS2APIC /PIC=LINT0 /NMI=LINT1 /PREC=DLI (virtual wire mode)

Tried the APIC switch earlier, both with it turned on and off in BIOS.
No improvement.
Just tried the virtual wire one, same results
Dave

Jonathan de Boyne Pollard

2011-08-16 11:46:21 UTC