XP patches break Pspice

M

Malcolm Reeves

Guest
Hi,

I'm convinced that an XP update has broken pspice. The fault is
always in the same module, same dll. Results are the same with 9.2.3
and 10(SP1.1). Same with different sch file that has run ok, on XP,
with 9.23. Re-installs do not fix. Transient runs are ok. Problem
is a long Monte Carlo AC run. Error is RPC server unavailable. The
event viewer for the application error has this:

Faulting application simsrvr.exe, version 1.0.0.1, faulting module
psp_ppo1.dll, version 0.0.0.0, fault address 0x00006826.

Fault address changes but otherwise the same. I think it is the XP
patches since a lot of these are to do with RPC. Also, it might not
actually be a PSpice programming error as perhaps MC runs look a bit
like an attempt to cause a buffer overflow - lots of similar calls one
after another.

Is there anyone who has
XP home (or pro) with all the latest updates
pspice 9.2 or 10
and who has a large MC that works, still?

TIA (I'd like to know if it just me or what).


--

Malcolm Reeves BSc CEng MIEE MIRSE, Full Circuit Ltd, Chippenham, UK
(mreeves@fullcircuit.com, mreeves@fullcircuit.co.uk or mreeves@iee.org).
Design Service for Analogue/Digital H/W & S/W Railway Signalling and Power
electronics. More details plus freeware, Win95/98 DUN and Pspice tips, see:

http://www.fullcircuit.com or http://www.fullcircuit.co.uk

NEW - Desktop ToDo/Reminder program (free)
 
almost looks like you have another PC or a local proxy trying to hit you
with a buffer overrun..
from what you are describing i would check your PC alittle more..
either that the app really does have a problem that its not aware of.



Malcolm Reeves wrote:

Hi,

I'm convinced that an XP update has broken pspice. The fault is
always in the same module, same dll. Results are the same with 9.2.3
and 10(SP1.1). Same with different sch file that has run ok, on XP,
with 9.23. Re-installs do not fix. Transient runs are ok. Problem
is a long Monte Carlo AC run. Error is RPC server unavailable. The
event viewer for the application error has this:

Faulting application simsrvr.exe, version 1.0.0.1, faulting module
psp_ppo1.dll, version 0.0.0.0, fault address 0x00006826.

Fault address changes but otherwise the same. I think it is the XP
patches since a lot of these are to do with RPC. Also, it might not
actually be a PSpice programming error as perhaps MC runs look a bit
like an attempt to cause a buffer overflow - lots of similar calls one
after another.

Is there anyone who has
XP home (or pro) with all the latest updates
pspice 9.2 or 10
and who has a large MC that works, still?

TIA (I'd like to know if it just me or what).
Click to expand...
 
Jamie (jamie_5_not_valid_after_5_Please@charter.net) said those last words:
almost looks like you have another PC or a local proxy trying to hit you
with a buffer overrun..
from what you are describing i would check your PC alittle more..
either that the app really does have a problem that its not aware of.
Click to expand...
It's not his PC... it's just PSpice that is broken by the patch.
As he said in other post:

On Sun, 20 Jun 2004 17:19:43 -0300, Chaos Master
FreeBSD_Reloaded@yahoo.com.br> wrote:

Isn't this RPC the same RPC stuff that the Blaster virus (ab)used to attack?
If yes, then probably the Win XP patch may have fixed a security hole, but
broken PSpice.

Exactly right. I've installed on a Win98se and the MC ran to
completion so there is nothing wrong with the simulation (not that it
was that sort of error). I'll try setting simsvr to run in Win2k
compatibility mode to see if that helps. This will be an overnight
run so we'll see tomorrow if this helps.
Click to expand...
[]s
--
Chaos MasterŽ. Posting from Brazil.
ICQ: 126735906 MSN: Wizard_of_Yendor@hotmail.com
"* CREATED USING BLOOD,SWEAT AND TEARS ON 10/12/90 AT 09:03AM"
-- From a PSpice 5.2 model library
 
On Sun, 20 Jun 2004 16:59:45 +0100, Malcolm Reeves
<mreeves@fullcircuit.com> wrote:

Hi,

I'm convinced that an XP update has broken pspice. The fault is
always in the same module, same dll. Results are the same with 9.2.3
and 10(SP1.1). Same with different sch file that has run ok, on XP,
with 9.23. Re-installs do not fix. Transient runs are ok. Problem
is a long Monte Carlo AC run. Error is RPC server unavailable. The
event viewer for the application error has this:

Faulting application simsrvr.exe, version 1.0.0.1, faulting module
psp_ppo1.dll, version 0.0.0.0, fault address 0x00006826.

Fault address changes but otherwise the same. I think it is the XP
patches since a lot of these are to do with RPC. Also, it might not
actually be a PSpice programming error as perhaps MC runs look a bit
like an attempt to cause a buffer overflow - lots of similar calls one
after another.

Is there anyone who has
XP home (or pro) with all the latest updates
pspice 9.2 or 10
and who has a large MC that works, still?

TIA (I'd like to know if it just me or what).
Click to expand...
You do have the option of uninstalling patches and restoring the
previous driver the patch replaced. Control Panel > Add Remove
Programs.
If you know about when PSpice started barfing, you can look in your
Windows directory for directories named after the patch. The directory
date is when you installed the patch. This may reduce the number of
patches you need to remove. I have seen the patches break other
things, most notably, Windoze printer drivers if you didn't have the
proper service pack.
If you find the culprit, please tell us!

Mark
 
In article <MPG.1b41a7146fe872c6989b4d@130.133.1.4>,
Chaos Master <renan_tdbREMOVEME@yahoo.com.br> wrote:
[....]
It's not his PC... it's just PSpice that is broken by the patch.
Click to expand...
Has anyone suggested he try the "Windoz 98 compatibility" more for the
program. I found that it made non-XP programs a little more likely to run
on XP. It does slow things down a bit[1] and the screen writing on
graphics isn't really like windows 98 but it is more like 98 than XP is.

[1] The 2.2 GHz machine couldn't keep up with my typing.

--
--
kensmith@rahul.net forging knowledge
 
On Tue, 22 Jun 2004 18:40:45 -0700, qrk <SpamTrap@reson.com> wrote:


You do have the option of uninstalling patches and restoring the
previous driver the patch replaced. Control Panel > Add Remove
Programs.
Click to expand...
Unfortunately although my windows update history (on windows update
web page FYI) shows I've installed some patches that I think are
relevant, KB 823980 (ms03-026) and KB824146 (ms03-039), they are not
listed in add/remove, else I would have tried that.

If you find the culprit, please tell us!
Click to expand...
I have a theory and it would seem to stand up as I've fixed the
problem :)).

I think that IoPageLockLimit is implicated. The info on this registry
setting is conflicting. Some say it works on XP some only on NT (and
some NT pre SPx). Anyway I have XP-tweak (a windows optimising
program) which sets this. When I looked it was set to 32k. It should
have been set to 64k, given my RAM size, which is what some sites say
XP uses always, never looking at IoPageLockLimit.

I think the windows patch may have changed a component so that parts
of XP are looking at IoPageLockLimit and parts aren't. With
IoPageLockLimit set to 32k my run failed at 2559. With it set to 4k
it failed at 95. And with it set to 64k it complete all 5000. So it
would seem to be involved. In theory all settings should work since
it is meant vary depending on available RAM. The default state of
this key not existing would also seem to work.

It would fit that it is something to do with disk accesses since the
fail count depends on how many nodes probe looks at. I assume that
although the transient run that works produces a bigger file that the
i/o is different, perhaps more steady, or bigger chunks, than the MC
analysis.

So I think that MS has managed to mixed in some old NT code in the
patches and now parts of the rpc server use a fixed value for
IoPageLockLimit and parts look at the registry. And if they are not
the same - problem city!

You might also find other intensive applications (graphics,
rendering?) affected. Those who haven't tweaked IoPageLockLimit or
run a tweaking program should be ok.


--

Malcolm Reeves BSc CEng MIEE MIRSE, Full Circuit Ltd, Chippenham, UK
(mreeves@fullcircuit.com, mreeves@fullcircuit.co.uk or mreeves@iee.org).
Design Service for Analogue/Digital H/W & S/W Railway Signalling and Power
electronics. More details plus freeware, Win95/98 DUN and Pspice tips, see:

http://www.fullcircuit.com or http://www.fullcircuit.co.uk

NEW - Desktop ToDo/Reminder program (free)
 
In article <lgfid0tbffn6cvqvu9f0slnobda1j6iem3@4ax.com>,
Malcolm Reeves <mreeves@fullcircuit.com> wrote:
[...]
I think the windows patch may have changed a component so that parts
of XP are looking at IoPageLockLimit and parts aren't. With
IoPageLockLimit set to 32k my run failed at 2559. With it set to 4k
it failed at 95. And with it set to 64k it complete all 5000. So it
would seem to be involved.
Click to expand...
Sounds like you've found it.

In theory all settings should work since
it is meant vary depending on available RAM. The default state of
this key not existing would also seem to work.
Click to expand...
No, in theory only the right setting made while holding your left foot in
the air will work. I think IoPageLockLimit, is involved in the virtual
memory (mis)managment. Forcing the wrong value in can cause XP to have a
page fault it can't deal with.

[...]
So I think that MS has managed to mixed in some old NT code in the
patches and now parts of the rpc server use a fixed value for
IoPageLockLimit and parts look at the registry. And if they are not
the same - problem city!
Click to expand...
A lot of XP's code is NT code so you'd expect the XP and NT patches to
share the same code in some sections. XP also contains Win95->98->Me code
in some parts so those parts patches would be different than the NT one.

--
--
kensmith@rahul.net forging knowledge
 
On Wed, 23 Jun 2004 09:36:45 +0100, Malcolm Reeves
<mreeves@fullcircuit.com> wrote:

On Tue, 22 Jun 2004 18:40:45 -0700, qrk <SpamTrap@reson.com> wrote:


You do have the option of uninstalling patches and restoring the
previous driver the patch replaced. Control Panel > Add Remove
Programs.

Unfortunately although my windows update history (on windows update
web page FYI) shows I've installed some patches that I think are
relevant, KB 823980 (ms03-026) and KB824146 (ms03-039), they are not
listed in add/remove, else I would have tried that.

If you find the culprit, please tell us!

I have a theory and it would seem to stand up as I've fixed the
problem :)).

I think that IoPageLockLimit is implicated. The info on this registry
setting is conflicting. Some say it works on XP some only on NT (and
some NT pre SPx). Anyway I have XP-tweak (a windows optimising
program) which sets this. When I looked it was set to 32k. It should
have been set to 64k, given my RAM size, which is what some sites say
XP uses always, never looking at IoPageLockLimit.

I think the windows patch may have changed a component so that parts
of XP are looking at IoPageLockLimit and parts aren't. With
IoPageLockLimit set to 32k my run failed at 2559. With it set to 4k
it failed at 95. And with it set to 64k it complete all 5000. So it
would seem to be involved. In theory all settings should work since
it is meant vary depending on available RAM. The default state of
this key not existing would also seem to work.

It would fit that it is something to do with disk accesses since the
fail count depends on how many nodes probe looks at. I assume that
although the transient run that works produces a bigger file that the
i/o is different, perhaps more steady, or bigger chunks, than the MC
analysis.

So I think that MS has managed to mixed in some old NT code in the
patches and now parts of the rpc server use a fixed value for
IoPageLockLimit and parts look at the registry. And if they are not
the same - problem city!

You might also find other intensive applications (graphics,
rendering?) affected. Those who haven't tweaked IoPageLockLimit or
run a tweaking program should be ok.

I'm always amazed when someone can figure out that one little setting
Click to expand...
out of thousands can cure ills in Windoze! According to the first
article I pulled out of a Google News search
<http://groups.google.com/groups?&as_umsgid=ucWeKfCfBHA.2308@tkmsftngp04>,
IoPageLockLimit should be set to zero. This way the OS figures out
what it should use. Strange that it was set to a non-zero value.

---
Mark
 
On Wed, 23 Jun 2004 18:37:28 -0700, qrk <SpamTrap@reson.com> wrote:


I'm always amazed when someone can figure out that one little setting
out of thousands can cure ills in Windoze!
Click to expand...
So am I! Must have been the good fairy helping when I found that.

According to the first
article I pulled out of a Google News search
http://groups.google.com/groups?&as_umsgid=ucWeKfCfBHA.2308@tkmsftngp04>,
IoPageLockLimit should be set to zero. This way the OS figures out
what it should use. Strange that it was set to a non-zero value.
Click to expand...
Well windows does not have a good record of choosing the right setting
for everyone - that's why tweak programs exist :).

If you search the web for IoPageLockLimit (rather than groups) you
might find: http://www.winguides.com/registry/display.php/55/ (among
lots of others) which tells you to create it and set it (by default I
don't think it exists). To be fair that link says it is for winNT/2k
but others apply it to XP, like
http://www.pureperformance.com/js/showtip.asp?id=125 and so does my
Tweak-XP program.

The last link is interesting since it says 2k default setting is 0
(512k) where as XP it is 65k (1MB). Now if MS have mixed NT and XP
code there may be problems with a 0 setting. As it's working at the
moment so I'm not inclined to change :).


--

Malcolm Reeves BSc CEng MIEE MIRSE, Full Circuit Ltd, Chippenham, UK
(mreeves@fullcircuit.com, mreeves@fullcircuit.co.uk or mreeves@iee.org).
Design Service for Analogue/Digital H/W & S/W Railway Signalling and Power
electronics. More details plus freeware, Win95/98 DUN and Pspice tips, see:

http://www.fullcircuit.com or http://www.fullcircuit.co.uk

NEW - Desktop ToDo/Reminder program (free)
 
You must log in or register to reply here.

Welcome to EDABoard.com

Sponsor

Back
Top