Re: Kernel Oops

Hi,

I disabled the high precision timer in my bios and recompiled a fresh patched 3.16 kernel not in preemtible, but in desktop mode.
Now the module seems to load, but the latency test hangs due to a non-responding module (?):

[   74.567982] RTAI[malloc]: global heap size = 2097152 bytes, <BSD>.
[   74.595871] , kstacks pool size = 524288 bytes.
[   74.595873] RTAI[sched]: hard timer type/freq = APIC/99748998(Hz); default timing: periodic; linear timed lists.
[   74.595875] RTAI[sched]: Linux timer freq = 250 (Hz), TimeBase freq = 3191992000 hz.
[   74.595877] RTAI[sched]: timer setup = 10 ns, resched latency = 3933 ns.

*
*
* Type ^C to stop this application.
*
*
Warning: Error opening /proc/rtai/latency_calibrate
         Couldn't get infos about the module's state.
RTAI Testsuite - KERNEL latency (all data in nanoseconds)
RTH|    lat min|    ovl min|    lat avg|    lat max|    ovl max|   overruns


What might be the reason for this and how can I debug?

Cheers,

Friedemann

> Am 30.03.2016 um 23:55 schrieb Paolo Mantegazza <[email protected]>:
> 
> Unfortunately this thread went on privately. Likely because of a wrong answer click of mine, apologies. 
> 
> In any case it gives me the chance of remarking a few things related to the support of Linux-3.x.x and now 4.x.x.
> - RTAI-4 onward) is based on the oneshot mode timing only. I still use periodic timers on data acquisition card to time specific tasks . whituout charging the scheduler with their pacing.
> - from RTAI-5, i.e the actual VULCANO to be released shortly,  there starting/stopping_rt_timer, it is done by rtai_sched onyhow.
> - in such a view the setting of the hard timer mode its start/stop, both in periodic and oneshot mode, will do nothing; they are just kept for compatibility, so that there is no need to change any existing code
> - HPET is not used by RTAI but 8259/54 is used in UP mode, even if it is strongly suggested to enable the APIC if available
> - RTAI does not use the second channel of 8259/54 to emulate a continuously running clock, so  a _TSC_is_needed
> 
> In practice what above restricts the newest RTAI versions to true 586s and onward. For fake 586s and 486 earlier version of RTAI should be used.
> 
> That said there is no need to disable NO_HZ and high precision/resolution timers if they are available in Linux.
> 
> Paolo
> 
> ________________________________________
> From: Rtai [[email protected]] on behalf of Alec Ari [[email protected]]
> Sent: Wednesday, March 30, 2016 10:37 PM
> To: RTAI Mailing List
> Subject: Re: [Rtai] Kernel Oops
> 
> RTAI should be able to use the 8259/8254 timers unless that's now broken? It's never _needed_ APIC or HPET in the past. If it's broken or not working (buggy hardware / BIOS) try disabling high precision + high resolution event timers in kernel config and BIOS if applicable. Also, to avoid NULL pointer issues, set CONFIG_LSM_MMAP_MIN_ADDR to value 65536.
> 
> Alec Ari
> _______________________________________________
> 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

-- 
Dr. rer. nat. Friedemann Bunjes
EDV Entwicklung und Koordination 
Hertie-Institut für Klinische Hirnforschung
Zentrum für Neurologie
Universitätsklinikum Tübingen
Otfried-Müller-Str. 25
72076 Tübingen
Tel. +49-7071-29-81999
Fax +49-7071-29-25010

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