rtai_smi.ko stopped working

For years and years, I ran EMC2 (ups, LinuxCNC on an
IBM ThinkCentre, which had that cyclic, 32/64s realtime-problem.

In the past, loading the rtai_smi.ko module in rtapi.conf
was all I need to fix this problem.

Well, my latest update to 2.5.1 (or possibly already in 2.5.0) brought
up this problem again.
The module is loaded, but doesn't do anything...

Any suggestions?

Hi

Can you do a lspci --v and attach the print so I can see which particular chip you have.
Also which Ubuntu version are you using

I will then have a look at the code and see if I can spot anything.

regards

Just what I was about to say.

I checked 2.5 and 2.5.1 debs. They did not change any of the realtime modules, but both changed rtai.conf to a default version without rtai_smi loaded

I'm hitting this too.

rtai.org released 3.9 on 28.7 or thereabouts.

For some mysterious reason, with 3.9 against a patched 2.6.38.8, the latest kernel it is claimed to support, the module rtai_smi.ko is not being built?

Very strange? The same configuration was working with 3.8.1 + 2.6.35.7.

A more modern kernel opens up the Multitouch-HID interface, which would nicely provide a safety interlock against spurrious touch events.

Cheers
_jC
.

Thanks Andy

Yeah, I looked at that, and its as unclear as my notes from last April.

<snip from ref>
Find the smi-module.c inside rtai/base/arch/i386/calibration, and change the following line:
#define CONFIG_RTAI_HW_SMI_ALL 1

Once that is changed, you can compile the module and replace the version already installed.
</snip from ref>

changed how? changed to what?

I've looked at the code and 3.9(final) only adds ICH10

<smi-module.c lines 34-56>
/* set these as you need */
#define CONFIG_RTAI_HW_SMI_ALL 1
#define CONFIG_RTAI_HW_SMI_INTEL_USB2 0
#define CONFIG_RTAI_HW_SMI_LEGACY_USB2 0
#define CONFIG_RTAI_HW_SMI_PERIODIC 0
#define CONFIG_RTAI_HW_SMI_TCO 0
#define CONFIG_RTAI_HW_SMI_MC 0
#define CONFIG_RTAI_HW_SMI_APMC 0
#define CONFIG_RTAI_HW_SMI_LEGACY_USB 0
#define CONFIG_RTAI_HW_SMI_BIOS 0

#ifndef PCI_DEVICE_ID_INTEL_ICH7_0
#define PCI_DEVICE_ID_INTEL_ICH7_0 0x27b8
#endif
#ifndef PCI_DEVICE_ID_INTEL_ICH7_1
#define PCI_DEVICE_ID_INTEL_ICH7_1 0x27b9
#endif
#ifndef PCI_DEVICE_ID_INTEL_ICH8_4
#define PCI_DEVICE_ID_INTEL_ICH8_4 0x2815
#endif
#ifndef PCI_DEVICE_ID_INTEL_ICH10_1
#define PCI_DEVICE_ID_INTEL_ICH10_1 0x3a16
#endif
</smi-module.c lines 34-56>

Its gotta be somewhere in autoconf/GNUmakefile.in. Have you been there?

Thanks
_jC
.

<snip from ref>
Find the smi-module.c inside rtai/base/arch/i386/calibration, and change the following line:
#define CONFIG_RTAI_HW_SMI_ALL 1

You still haven't said what version of Ubuntu you are running , but those notes are for Dapper where rtai_smi was not present

The module in realtime has not changed over the versions and if you are running a stock rtai kernel for that version, you should just have to put it back into rtai.conf

regards

I am sorry for wasting your time on this, but as
it seems, the failure was caused by a corrupted module and was
in no way related to LinuxCNC or RTAI...

I replaced the file from a 2nd installation and milled some parts
without having any problems.

Sorry
(But good to know there's someone out there to help

I'm running a pretty stripped down & up to date Debian sid amd64.

so gcc 4.7.1 and automake 1.11

I've built the Kernel, RTAI, and EM^h^hLinuxCNC from the most recent sources.

Same machine as this thread linuxcnc.org/index.php/english/component...ew&catid=18&id=19919

This machine was working flawlessly.

This was an attempt to integrate the new 3.9 RTAI release with the 2.5.1CNC release with 2.6.38.8, especially because this kernel has multitouch support.

I just can't find my notes on rtai_smi.ko compilation, and that usually means it was no issue at the time.

Now its a problem. The Makefile Andy referenced is bitrotted too.

Thanks for your help.

If the 'Official' binary has this module out-of-the box my next step is to apt-get those sources and diff them against mine. Unless somebody can suggest exactly how it is that this module becomes included in that binary package.

Cheers
_jC
.