[Xenomai-help] xeno-test can not run

[Xenomai-help] xeno-test can not run

[robert165]

[-- Attachment #1.1: Type: text/plain, Size: 1066 bytes --]

Hi all，
I installed xenomai-2.4.4 on AT91RM9200DK, all have done OK. But the board halted when running xeno-test.
would anyone help me, thanks a lot.
Besides, which version of xenomai is a good choice for real application with arm?
I used linux-2.6.20.21 + 2.6.20-at91.patch + adeos-ipipe-2.6.20-arm-1.8-03.patch + xenomai-2.4.4
The run info is as follows, and the full log is attached.
 
Thu Jan  1 00:07:57 UTC 1998
running: top
[H[JMem: 5692K used, 24184K free, 0K shrd, 0K buff, 1884K cached
Load average: 0.71, 0.17, 0.06    (State: S=sleeping R=running, W=waiting)
[7m  PID USER     STATUS   RSS  PPID %CPU %MEM COMMAND[0m
 1025 0        R        376     1 98.0  1.2 dd
 1212 0        R        676  1211  2.9  2.2 top
  738 0        S        760     1  0.0  2.5 sh
  743 0        S        612   738  0.0  2.0 xeno-test
    1 0        S        568     0  0.0  1.9 init
  798 0        S        472   743  0.0  1.5 xeno-test
Thu Jan  1 00:07:58 UTC 1998
running: ./run -- -sh -T 120 -t0 # latency
*
*
* Type ^C to stop this application.
*
*

Regards
 
Robert




[-- Attachment #1.2: Type: text/html, Size: 2031 bytes --]

[-- Attachment #2: test-2.6.20.21-980101.000730 --]
[-- Type: application/octet-stream, Size: 12260 bytes --]

Re: [Xenomai-help] xeno-test can not run

[Gilles Chanteperdrix]

robert165 wrote:
> Hi all，
> I installed xenomai-2.4.4 on AT91RM9200DK, all have done OK. But the
> board halted when running xeno-test.
> would anyone help me, thanks a lot.
> Besides, which version of xenomai is a good choice for real application
> with arm?

Any version is as good as we can make it, but if you want to be able to
report a bug, and not to stumble upon bugs which have been fixed, you
should be using the latest versoin.

There is also the recent discovery of a very old bug of basically all
the I-pipe patches on arm, so, you should be using Xenomai 2.4.10 and
the patches it provides. Add to that the integration of the FCSE patch
into the I-pipe patch which will give you much better real-time
performances.

> I used linux-2.6.20.21 + 2.6.20-at91.patch +
> adeos-ipipe-2.6.20-arm-1.8-03.patch + xenomai-2.4.4
> The run info is as follows, and the full log is attached.
>  
> Thu Jan  1 00:07:57 UTC 1998
> running: top
> [H[JMem: 5692K used, 24184K free, 0K shrd, 0K buff, 1884K cached
> Load average: 0.71, 0.17, 0.06    (State: S=sleeping R=running, W=waiting)
> [7m  PID USER     STATUS   RSS  PPID %CPU %MEM COMMAND[0m
>  1025 0        R        376     1 98.0  1.2 dd
>  1212 0        R        676  1211  2.9  2.2 top
>   738 0        S        760     1  0.0  2.5 sh
>   743 0        S        612   738  0.0  2.0 xeno-test
>     1 0        S        568     0  0.0  1.9 init
>   798 0        S        472   743  0.0  1.5 xeno-test
> Thu Jan  1 00:07:58 UTC 1998
> running: ./run -- -sh -T 120 -t0 # latency
> *
> *
> * Type ^C to stop this application.
> *
> *

That is a bug which has been around for too long due to my lazyness, but
which Philippe has finally fixed in 2.4.10. The default period of the
latency test is 100us, which is too short for an ARM. Run latency with
-p 1000 to ask for a 1ms period.

-- 
					    Gilles.

Re: [Xenomai-help] xeno-test can not run

[robert165]

[-- Attachment #1: Type: text/plain, Size: 1286 bytes --]



>> OK, I have found the FCSE option with kernel 2.6.28.
>> My working computer is not here, so I tested in a VMware and a old
>> kernel 2.6.20 at first. Then I download a 2.6.28 kernel and things be OK
>> now.
>
>I took care to tell you clearly that the FCSE option was available in
>the latest versions of the I-pipe patches, which were in the Xenomai
>2.4.10 release. Obviously the patch for kernel 2.6.20 is not part of
>Xenomai 2.4.10.
>
>> 
>>>Properly loading the system means hammering it real hard. And if you do
>>>that, you will get real results. Around 260us in user-space without
>>>FCSE, or 200us with FCSE.
>> 
>>  "hammering it real hard ", you mean work in real application?
>> Sorry, my English is so terrible, I just know that hammer is a tool for strike.
>
>Yes, so hammering means striking with a hammer... It is an image.

I used nfs filesystem in the test, that's the point?

>
>> 
>> By the way, shall we talk by MSN? My ID is robert165@domain.hid
>
>No. I had not noticed that you were sending me these mails privately.
>Xenomai has mailing lists, they are meant to be used so that other can
>find answers to the questions you are asking. I am not at your personal
>service.

I am so sorry, I did not noticed that too. 
Thanks for your help sincerely.

Robert

[-- Attachment #2: Type: text/html, Size: 1507 bytes --]

Re: [Xenomai-help] xeno-test can not run

[Gilles Chanteperdrix]

robert165 wrote:
> 
>>> OK, I have found the FCSE option with kernel 2.6.28.
>>> My working computer is not here, so I tested in a VMware and a old
>>> kernel 2.6.20 at first. Then I download a 2.6.28 kernel and things be OK
>>> now.
>>
>>I took care to tell you clearly that the FCSE option was available in
>>the latest versions of the I-pipe patches, which were in the Xenomai
>>2.4.10 release. Obviously the patch for kernel 2.6.20 is not part of
>>Xenomai 2.4.10.
>>
>>> 
>>>>Properly loading the system means hammering it real hard. And if you do
>>>>that, you will get real results. Around 260us in user-space without
>>>>FCSE, or 200us with FCSE.
>>> 
>>>  "hammering it real hard ", you mean work in real application?
>>> Sorry, my English is so terrible, I just know that hammer is a tool for strike.
>>
>>Yes, so hammering means striking with a hammer... It is an image.
> 
> I used nfs filesystem in the test, that's the point?

The point is to generate activity on your system, to stress it. But real
hard.


-- 
					    Gilles.

Re: [Xenomai-help] xeno-test can not run

[robert165]

[-- Attachment #1: Type: text/plain, Size: 1205 bytes --]






在2009-12-24，"Gilles Chanteperdrix" <gilles.chanteperdrix@xenomai.org> 写道：
>robert165 wrote:
>> 
>>>> OK, I have found the FCSE option with kernel 2.6.28.
>>>> My working computer is not here, so I tested in a VMware and a old
>>>> kernel 2.6.20 at first. Then I download a 2.6.28 kernel and things be OK
>>>> now.
>>>
>>>I took care to tell you clearly that the FCSE option was available in
>>>the latest versions of the I-pipe patches, which were in the Xenomai
>>>2.4.10 release. Obviously the patch for kernel 2.6.20 is not part of
>>>Xenomai 2.4.10.
>>>
>>>> 
>>>>>Properly loading the system means hammering it real hard. And if you do
>>>>>that, you will get real results. Around 260us in user-space without
>>>>>FCSE, or 200us with FCSE.
>>>> 
>>>>  "hammering it real hard ", you mean work in real application?
>>>> Sorry, my English is so terrible, I just know that hammer is a tool for strike.
>>>
>>>Yes, so hammering means striking with a hammer... It is an image.
>> 
>> I used nfs filesystem in the test, that's the point?
>
>The point is to generate activity on your system, to stress it. But real
>hard.

Yes, I understand now. 
Thank you again for your generous help!

Robert



[-- Attachment #2: Type: text/html, Size: 1603 bytes --]