Help with SMI on 2.4.7

Thanks, I completely understand and its my problem if it damages the PC. Which it didnt.
Here's what the error log produced, we are much closer I think.

insmod: error inserting '/usr/realtime-2.6.32-122-rtai/modules/rtai_smi.ko': -1 Invalid module format

and

rtai_smi: version magic '2.6.32-RTAI SMP preempt mod_unload CORE2 ' should be '2.6.32-122-rtai SMP mod_unload 586TSC

Here's the whole emc2 startup debug file

Thank you very much for helping with this.
I dont know if it makes a difference or not but I dont have the isolcpus kernel mod running and I didnt shut 1 CPU off in the BIOS but I'll try that now. I did a fresh install last night and just doing the updates now.

Hi

The problem is exact compatibility between the 2 systems (yours and mine) and the rtai used.

I think I have now got it sorted using the other rtai libraries

Attached as a .zip this time in case any corruption occurred in the download, albeit it was byte length identical.

I am going to be away for a while, so won't be able to continue with it for that period, albeit I can monitor this thread.

If it still says the format is invalid, you could try compiling on your machine.

Ensure you have build-essentials package on your machine
Get the sources for kernel 2.6.32 and put in /usr/src along with the headers for 2.6.32-122 and 2.6.32-122-rtai
Copy the smi-module.c, rtai-wrappers.h and Makefile from the zip into /usr/src
cd to /usr/src and compile using make

good luck

I dont know if it makes a difference or not but I dont have the isolcpus kernel mod running and I didnt shut 1 CPU off in the BIOS but I'll try that now. I did a fresh install last night and just doing the updates now.

I think the default for the .rtai-config from Kconfig is for 2 cores anyway so I would leave it as it is.

The error message is to do with differences in between the rtai build, the second module I built should not have this.

regards

Thanks again, I will try the second module and if it fails I will follow your instructions and compile a module. It will be a great learning experience. Thank you for all your help. I will post results so you can follow along.

Cheers
Craig

Well your second module loaded, here is whats in the dmesg


[ 632.553735] RTAI: Intel chipset found, enabling SMI workaround.
[ 632.553749] SMI configuration has been set, mask used = 1.
[ 632.637028] config string '0x0378'
[ 655.921065] 23331: ERROR: Unexpected realtime delay: check dmesg for details.
[ 655.921072]
[ 655.921073] In recent history there were
[ 655.921075] 3378376, 3405678, 3378087, 3378155, and 3378359
[ 655.921077] elapsed clocks between calls to the motion controller.
[ 655.921082] This time, there were 4925359 which is so anomalously
[ 655.921094] large that it probably signifies a problem with your
[ 655.921097] realtime configuration. For the rest of this run of
[ 655.921099] EMC, this message will be suppressed.

I guess I'll try and compile a module but since yours worked mine will do the same thing.

Thanks
Craig

Hi

I'm back now, any joy with this?

As you said, the module loaded OK so there is no need to try compiling it yourself. The previous errors were just because I used a different flavour of rtai libraries and the version check was not as required.

What we don't know is whether this fix works for other ICH8 chips, there might be good reason why only ICH8_4 is specified, but it was worth a try.

Probably need to run without rtai-smi and evaluate the spikes, then run with it and see if the spikes have been suppressed.

What caused this RTAI error may be the module itself or may be something else.
You should have had an error like this every time you ran on the 32 second spike, but it is only output once and can be missed.

It might be worth trying the isolcpus=1 route and then loading the modified rtai-smi to see what happens.
I doubt the original MBs targeted by this module had dual core processor support.

If this still fails, well it was worth a try. You might have to consign that computer to mundane duties and consider an Intel Atom 525 Mini ITX MB which has been extensively tested and does work very well, as well as being very cheap.

regards

Hi ArcEye, Thanks very much for your help on this. I tried numerous things and couldn't get the huge spike stopped. We got close though, the software shutdown quit working so I had to push and hold the power button to actually shut the PC down, I had a HP Microserver N36L setup with PC-BSD running AFP time machine backups so I thought I'd give it a try. The little server produced numbers down around 5000 so I decided to use it. I had to purchase a PCI-E parallel port card ind install it and that went fine. Ubuntu found the card fine and I set Linuxcnc to use my parallel port at 0xe800 and away it goes. I tried Mach 3 but linuxcnc gives much smoother stepping and my steppers barely get warm at all. I use the KL-5056D (aka DM556) digital drivers and they work great. I'll use that other PC with Ubuntu to run FreeCAD / HeeksCad on. Now I need to try and obtain the step timing for these drivers I have.

Thanks again for all your help.

Craig

Yes I will do that for sure, I'm going to call Leadshine USA and ask them to provide me with the step timing for all their drivers. I will post them as soon as I can.

I would be interested too.

All my stepper drivers are Leadshine or re-badged Leadshine and they do work very well, just set the jumpers, install and forget.

I have never been able to find specific timing data for them so have used conservative estimates.

So I might have a bit more leeway for tuning!

regards