disabling SMI

I have a real-time app that runs on a P4. The box we are running it on
has an Aopen i865PEa-7IF motherboard. We are using the On-Time RTOS.
We have found that every 37 seconds the app completely stops running
for 250 microseconds. Using a PCI bus analyzer we discovered that
between the last operation done by my app and when my app resumes
running shows 100's and 100's of I/O reads and writes to addresses B3,
EB, and 61. Googling on this has led me to believe that these accesses
are related to SMI's and SMM. I have disabled everything in the BIOS -
USB, LAN, power management, serial and parallel ports, modem, firewire,
MIDI and game ports - yet my app still stops and I still see these
accesses.

Does anyone know how I can totally disable the SMI's?

TIA!
-larry

 

Are you using an Intel processor with this, or a compatible non Intel part?
I have experienced a similar problem when we switched components once, but I
forget which (non Intel) part was giving us the grief. It turned out to be
some silicon bug which had been mentioned on the On-Time support archive.
Sorry my memory of it is too sketchy but I think there was some way of
disabling the SMI in software during the start-up routine - but this then
lead me to a different problem and in the end I just ditched the processor
for a different variant.

It might be that your problem is completely different but I suggest looking
through the On-Time support archive for SMI or SMM. You should hit some
info if you go back far enough.

Regards,
Richard.

http://www.FreeRTOS.org
*Now for ARM CORTEX M3!*

 

I am using an Intel 3.6 GHz P4 (775 socket)

I've been told this by others as well, but my searches in the archives
have not turned up any hits.

Thanks!
-larry

 

My bad - I found these posts, but they're all related to the AMD Geode
GX-1 CPU, not the P4. I did try what was suggested in those threads,
but it didn't help in my situation atl all.

-larry

 

We've seen this (150+ us delays in our case), and it was caused by
SMI's. Most BIOS'es do not allow you to disable SMI's. You'll have to
program the South Bridge to disable them yourself.

This is highly dependent on the particular chipset, so you should
download chipset documentation from Intel (freely available). In the
power management registers of the South Bridge, there's usually a "SMI
global enable" bit, which you want to set to zero.

 

Thanks very much for this reply Cyril. It was very helpful. It led me
to discover that my motherboard does not have a Northbridge/Southbridge
architecture; it has the newer IHA with a GMCH and ICH. I got the docs
on my ICH, found the SMI control and enable register, and modified my
program to clear the global SMI enable bit. This has stopped the SMI's.
Thanks again!

-larry

 

The code at this link is for the AMD Geode GX-1. It is not applicable
to the P4.

-larry

 

I also have a real-time app that runs on P4. It is running on AAEON MB-845GE
motherboard. CPU stalls for 250us for unknown reason. This event is not
periodic, sometimes it is after less than a minute, sometimes you have to
wait for few minutes. Motherboard is using Intel 845GE chipset. I downloaded
datasheet for South Bridge and wrote small program to read SMI enable
registers. All values are default (as specified in datasheet), that is, SMI
disabled.
After that, I disabled Thermal Monitor on CPU (TM1 & TM2) by writing to MSR
IA32_MISC_ENABLE (bits 3 and 13). No change. Still have 250us stall.

'cli' did not help, disabling APIC didn't help, disabling ALL interrupts on
PIC didn't help, writing to Intel Software Developers Forum didn't help...

Slavisa Zigic


: Armin Steinhoff wrote:
:> wrote:
:> > I have a real-time app that runs on a P4. The box we are running it on
:> > has an Aopen i865PEa-7IF motherboard. We are using the On-Time RTOS.
:> > We have found that every 37 seconds the app completely stops running
:> > for 250 microseconds. Using a PCI bus analyzer we discovered that
:> > between the last operation done by my app and when my app resumes
:> > running shows 100's and 100's of I/O reads and writes to addresses B3,
:> > EB, and 61. Googling on this has led me to believe that these accesses
:> > are related to SMI's and SMM. I have disabled everything in the BIOS -
:> > USB, LAN, power management, serial and parallel ports, modem, firewire,
:> > MIDI and game ports - yet my app still stops and I still see these
:> > accesses.
:> >
:> > Does anyone know how I can totally disable the SMI's?
:>
:> https://mail.rtai.org/pipermail/rtai/2005-July/012391.html
:> http://www.mail-archive.com//msg01507.html

: The code at this link is for the AMD Geode GX-1. It is not applicable
: to the P4.

: -larry

 

I found a way to disable SMI. Furthermore, I found that TCO_EN bit is
causing SMI#. I am using Intel 82801DB ICH4.

How safe it is to disable SMIs?
Is there any problem with overheating of CPU or any other side effect?

Thanks,

Slavisa Zigic
: I also have a real-time app that runs on P4. It is running on AAEON MB-845GE
: motherboard. CPU stalls for 250us for unknown reason. This event is not
: periodic, sometimes it is after less than a minute, sometimes you have to
: wait for few minutes. Motherboard is using Intel 845GE chipset. I downloaded
: datasheet for South Bridge and wrote small program to read SMI enable
: registers. All values are default (as specified in datasheet), that is, SMI
: disabled.
: After that, I disabled Thermal Monitor on CPU (TM1 & TM2) by writing to MSR
: IA32_MISC_ENABLE (bits 3 and 13). No change. Still have 250us stall.

: 'cli' did not help, disabling APIC didn't help, disabling ALL interrupts on
: PIC didn't help, writing to Intel Software Developers Forum didn't help...

: Slavisa Zigic


: : Armin Steinhoff wrote:
: :> wrote:
: :> > I have a real-time app that runs on a P4. The box we are running it on
: :> > has an Aopen i865PEa-7IF motherboard. We are using the On-Time RTOS.
: :> > We have found that every 37 seconds the app completely stops running
: :> > for 250 microseconds. Using a PCI bus analyzer we discovered that
: :> > between the last operation done by my app and when my app resumes
: :> > running shows 100's and 100's of I/O reads and writes to addresses B3,
: :> > EB, and 61. Googling on this has led me to believe that these accesses
: :> > are related to SMI's and SMM. I have disabled everything in the BIOS -
: :> > USB, LAN, power management, serial and parallel ports, modem, firewire,
: :> > MIDI and game ports - yet my app still stops and I still see these
: :> > accesses.
: :> >
: :> > Does anyone know how I can totally disable the SMI's?
: :>
: :> https://mail.rtai.org/pipermail/rtai/2005-July/012391.html
: :> http://www.mail-archive.com//msg01507.html

: : The code at this link is for the AMD Geode GX-1. It is not applicable
: : to the P4.

: : -larry

 

*** rampant top-posting fixed ***

Since this thread seems to contain valuable information, I have
gone to the trouble of correcting the top-posting and making it
readable. It is much easier to just avoid top-posting in the first
place. Your answer belongs after, or intermixed with, the material
to which you reply.

--
"Our enemies are innovative and resourceful, and so are we.
They never stop thinking about new ways to harm our country
and our people, and neither do we." -- G. W. Bush.
"The people can always be brought to the bidding of the
leaders. All you have to do is tell them they are being
attacked and denounce the pacifists for lack of patriotism
and exposing the country to danger. It works the same way
in any country." --Hermann Goering.

 

Depending on the specific MB, disabling SMI's could cause USB, power
management, audio, on-board NIC, MIDI, and joystick interfaces to not
work. On my platform (an Aopen i865PEa-7IF) the USB and NIC are still
working, and the other things are not needed by my app.

Thanks for doing this. Top-posting drives me crazy!

-larry

 

I have a similar issue as I write peripheral simulators for older machines (Ataris and Amigas) and I can't get the RDTSC to give me consistent results with various machines. On Toshiba laptops for example, it requires SMIs or some sort of watchdog timer to turn on/off the fan else the system heats up and shuts down. But with those timers on, the frequency of the processor changes with temperature and screws up the timing results I get from RDTSC. I have so far been using the original PIT (8253) to time things but it's slow and inefficient using ports 40h..43h and isn't as accurate.

Perhaps, someone can post code for disabling SMIs in general that they have discovered or read somewhere here. I know this won't help on some machines as it causes them to shut-down but it may help with other machines.

Thanks.

 

Anyone get more information about this?

It wouldn't be so bad if SMIs were the same on all machines but they vary with the machines and depend on BIOS settings as well. Really a nightmare for using PCs for accurate timing.

Anyone find a better way to time things. Besides the PIT on old machines which isn't affected by processor frequency changes, RDTSC is just "more accurate" in name only.