Re: RTAI-5 make hangs at self-calibrate.

Let's get rid of the user space calibration and see if RTAI works.
For that edit the file  rtai_instal_dir/calibrate/latencies and set the only record there as "1201 1700 100000".
Then try the user space latency test. Let's see if it runs and how.
Paolo
 
________________________________________
From: Duane AW. Jeffery [[email protected]]
Sent: Friday, April 15, 2016 8:36 PM
To: Paolo Mantegazza; Felix Frey; [email protected]
Subject: RE: [Rtai] RTAI-5 make hangs at self-calibrate.

Checking my linux .config, grep for NO_HZ returned

CONFIG_NO_HZ_COMMON=y
CONFIG_NO_HZ_IDLE=y
CONFIG_NO_HZ=y
# CONFIG_RCU_FAST_NO_HZ is not set

Grep for CPU_IDLE returned

# CONFIG_CPU_IDLE is not set

Grep for CPU_FREQ returned

# CONFIG_CPU_FREQ is not set

Attempted your mod to base/arch/x86/calibration/calibrate.c -- I don't see CONFIG_CPU_IDLE in there at all, but I think you meant DIAG_KF_LAT_EVAL, which was #define'd to 0, and used throughout to switch off several rt_printk's.  I changed its define to 1, recompiled rtai, rebooted, and tried again, and I found that the result seems to be the same.  Dmesg shows the kernel space cal complete ok, and the user space one not start -- no new lines in dmesg at all.  Ps aux still shows that calibrate script being called with the same parameters as you'd expect.

-Duane

-----Original Message-----
From: Paolo Mantegazza [mailto:[email protected]]
Sent: Friday, April 15, 2016 10:23 AM
To: Duane AW. Jeffery <[email protected]>; Felix Frey <[email protected]>; [email protected]
Subject: Re: [Rtai] RTAI-5 make hangs at self-calibrate.

With Linux be sure NO_HZ is on, it should by default. Then you do no need the BIOS to fix freq scaling and idlle.
CONFIG_CPU_FREQ and CONFIG_CPU_IDLE can be set while configuring Linux.
Even if I find it strange that it does not calibrate user space, after succeeding in kernel space, the method is the same, I'll be happy to have it better. In such a view having troubles on a machine could be a bless.
Unfortunately it's not mine. In any case there is still one thing you can try. In base/arch/x86/calibration/calibrate.c there is the macro CONFIG_CPU_IDLE. You should set it to 1 and retry. It will flood your dmesg with 10 thousands lines per second but you'll see what's going on.
Take into account that the loop must last, with the default calibration values of RTAI config, no more than 1 sec.

Paolo

On 04/15/2016 04:03 PM, Duane AW. Jeffery wrote:
> Processor on this machine is an Intel BYT-M 4core 3540, up to 2.66GHz, so says the sticker on it.
>
> Cat /proc/cpuinfo seems to call it a Pentium N3540 @ 2.16GHz, and
> shows its MHz as 2163.20 (which is the one parameter that seems to get passed to the calibrate script) The second parameter passed seems to be the result from the kernel latency test in the "LEAST = xxxx" field.
>
> Perhaps it's one of these chips that does the frequency scaling and stuff for battery life and that's making the calibrate hang?  On AMD machines I'm used to turning off some of those things in the BIOS, but this machine doesn't seem to have any such options.  The kernel config I'm running started as Shahbaz' default 64bit one, and then I added the things I needed for this machine, so it's pretty minimal.
>
> I tried manually calling the script from /usr/realtime/calibration with a ./run, and it hangs in the same place.  This time, the commands in ps aux are slightly different (the LEAST = parameter has changed a bit), dmesg looks similar, slightly different values of course.
>
> Not sure what else to try.
>
> -Duane
>
> -----Original Message-----
> From: Rtai [mailto:[email protected]] On Behalf Of Paolo
> Mantegazza
> Sent: Friday, April 15, 2016 5:52 AM
> To: Felix Frey <[email protected]>; [email protected]
> Subject: Re: [Rtai] RTAI-5 make hangs at self-calibrate.
>
> On 04/15/2016 11:25 AM, Felix Frey wrote:
>> Just checked with latest Vulcano and kernel 3.18.22 on my AMD GX-210.
>>
>> Insmoding rtai_sched tells:
>>
>> [  191.152594] USERMODE CHECK: OK.
>> [  191.152604] USERMODE CHECK PROVIDED (ns): KernelLatency -1, UserLatency -1.
>> [  191.277545] KERNEL SPACE LATENCY ENDED AT CYCLE: 1243, LATENCY = 4271, VARIANCE = 33761/10004569, GAIN = 843651/1000000000, LEAST = 3520.
>> [  191.357327] AFTER KERNEL CALIBRATION (WITH RTAI TASK, ns): KernelLatency 3526, UserLatency -1 (CALIBRATION: PERIOD 10000 (ns), TIME 204699629 (ns)).
>> [  191.480609] USER SPACE LATENCY ENDED AT CYCLE: 1203, LATENCY = 4899.93, VARIANCE = 0.00352333, GAIN = 0.000880833, LEAST = 4634.
>> [  191.482291] USERMODE USER SPACE CALIBRATION: OK.
>> [  191.482300] AFTER USER CALIBRATION (ns): KernelLatency 3526, UserLatency 4642 (CALIBRATION: PERIOD 10000 (ns), TIME 124954629 (ns)).
>> [  191.482306] FINAL CALIBRATION SUMMARY (ns): KernelLatency 3526, UserLatency 4642.
>>
>> Seems to be ok.
> Yes it is, thanks for your patience in pursuing the check.
> Paolo.
>> Felix
>>
>>
>>> I've no clue at the moment.
>>> I just know that on my 3 test machines,  gcc ranging from 3.something to 4.something, ubunt 8.somenthing to 14.something, P III to XEON, I get the following diagnostic summary:
>>> [  872.146524] USERMODE CHECK: OK.
>>> [  872.146531] USERMODE CHECK PROVIDED (ns): KernelLatency -1, UserLatency -1.
>>> [  872.271467] KERNEL SPACE LATENCY ENDED AT CYCLE: 1243, LATENCY = 5519, VARIANCE = 33761/10004569, GAIN = 843651/1000000000, LEAST = 5115.
>>> [  872.348028] AFTER KERNEL CALIBRATION (WITH RTAI TASK, ns): KernelLatency 1698, UserLatency -1 (CALIBRATION: PERIOD 10000 (ns), TIME 201475012 (ns)).
>>> [  872.477559] USER SPACE LATENCY ENDED AT CYCLE: 1203, LATENCY = 11950.4, VARIANCE = 0.00352333, GAIN = 0.000880833, LEAST = 11250.
>>> [  872.496137] USERMODE USER SPACE CALIBRATION: OK.
>>> [  872.496148] AFTER USER CALIBRATION (ns): KernelLatency 1698, UserLatency 3736 (CALIBRATION: PERIOD 10000 (ns), TIME 148104409 (ns)).
>>> [  872.496152] FINAL CALIBRATION SUMMARY (ns): KernelLatency 1698, UserLatency 3736.
>>> Figures can be differ for differing machines but I do not see why anything should hang.
>>> Your listing shows that the kernel part has finished, while ps aux
>>> shows the user space part running, even if I do not see where
>>> /usr/realtime/calibration/latencies 216320 2600 comes from. In any
>>> case the user space part, with the defaul frequency and calibration
>>> cycles, cannot but last 1 second What is the frequency of your CPU,
>>> something about I GHz? So
>>>
>>> Paolo.
>>>
>>> ________________________________________
>>> From: Duane AW. Jeffery [[email protected]]
>>> Sent: Thursday, April 14, 2016 7:24 PM
>>> To: Paolo Mantegazza
>>> Cc: '[email protected]'
>>> Subject: RE: [Rtai] RTAI-5 make hangs at self-calibrate.
>>>
>>> Tried Vulcano just now, after some fiddling, I was able to get it to compile, but then trying to insmod rtai_sched.ko seems to hang the terminal.
>>>
>>> Checking with another terminal, the last few lines of dmesg are like:
>>>
>>> USERMODE CHECK: OK
>>> USERMODE CHECK PROVIDED (ns): KernelLatency -1, UserLatency -1.
>>> KERNEL SPACE LATENCY ENDED AT CYCLE: 1243, LATENCY = 3058, VARIANCE = 33761/10004569, GAIN = 843651/1000000000, LEAST = 2600.
>>> AFTER KERNEL CALIBRATION (WITH RTAI TASK, ns): KernelLatency 1201, UserLatency -1 (CALIBRATION: PERIOD 10000 (ns), TIME 201353401 (ns)).
>>>
>>> Ps aux shows the following command running:
>>>
>>> /usr/realtime/calibration/calibrate
>>> /usr/realtime/calibration/latencies 216320 2600
>>>
>>> Any ideas?
>>>
>>> -Duane
>>>
>>> From: Paolo Mantegazza [mailto:[email protected]]
>>> Sent: Thursday, April 14, 2016 7:03 AM
>>> To: Duane AW. Jeffery <[email protected]>
>>> Subject: Re: [Rtai] RTAI-5 make hangs at self-calibrate.
>>>
>>> My suggestion is that you try VULCANO CVS. It wil be released as 5.0 test2 next week or so.
>>> One of the new things that come with it is exactly a new calibration procedure.
>>> I got a check, witnessed on this list, by just one user but two or three more fedback to me privatly it worked.
>>> So another check is welcomed.
>>>
>>> Paolo
>>>
>>> On 04/13/2016 10:28 PM, Duane AW. Jeffery wrote:
>>> Hi all,
>>>
>>> I'm working on a new laptop (ASUS E402M) and trying to get RTAI to compile on it.
>>>
>>> I am running kernel 3.18.22 - updated into what is otherwise a Slackware 14.1 install.  I have applied the hal patch and successfully compiled and booted my patched linux kernel.
>>>
>>> When I go to my RTAI source tree, unpacked into /usr/src/<folder> symlinked to /usr/src/rtai and try to compile it (first make menuconfig to setup, and then make), it begins ok, and then gets to the part where it calls "/bin/sh ../../base/sched/selfcalibrate.sh x86", and then it just hangs there forever.
>>>
>>> If I switch to another terminal, I can see that lsmod reports rtai_hal and rtai_tmpsched are loaded, and then that dmesg shows some messages that look ok from both the rtai_hal module and the rtai_tmpsched module.  Checking ps aux shows that the selfcalibrate.sh script is still running, and has called "/usr/src/rtai/base/arch/x86/calibration/.libs/lt-calibrate ../arch/x86/hal" which is also still running.  Along with a "tee tmp_output".
>>>
>>> Looking at the tmp_output file in /usr/src/rtai/base/sched/ reveals an empty file.  Also running is [RTAI_KTHRD_SRVR].
>>>
>>> I have tried both RTAI 5.0-test1.tar.bz2 as posted on the main page of www.rtai.org<http://www.rtai.org>, and also tried ShabbyX's github master tree.  Both fail the same way.
>>>
>>> Anybody have any idea what's going on here?  I've compiled RTAI many times before without this problem occurring, but never on this hardware.
>>>
>>> -Duane
>>>
>>>
>>>
>>>
>>> _______________________________________________
>>>
>>> Rtai mailing list
>>>
>>> [email protected]<mailto:[email protected]>
>>>
>>> https://mail.rtai.org/cgi-bin/mailman/listinfo/rtai
>>>
>>> _______________________________________________
>>> Rtai mailing list
>>> [email protected]
>>> https://mail.rtai.org/cgi-bin/mailman/listinfo/rtai
>>>
>> _______________________________________________
>> Rtai mailing list
>> [email protected]
>> https://mail.rtai.org/cgi-bin/mailman/listinfo/rtai
> _______________________________________________
> Rtai mailing list
> [email protected]
> https://mail.rtai.org/cgi-bin/mailman/listinfo/rtai

_______________________________________________
Rtai mailing list
[email protected]
https://mail.rtai.org/cgi-bin/mailman/listinfo/rtai