[LTP] [PROBLEM_WITH_CPUACCT.SH_100_100_LTP_TESTSUITE]

[LTP] [PROBLEM_WITH_CPUACCT.SH_100_100_LTP_TESTSUITE]

[Na Nha]

Hello All,



My name is Le Minh Nhan from Vietnam. I am working on a Linux project which
is tested by LTP test suite.


I had run the test case “./cpuacct.sh 100 100” on my ARM board.

Afterward,  I have gotten a problem as bellow:



*1 TINFO: Creating 100 subgroups each with 100 processes*

*[69402.212655] imx2_wdt_interrupt: pretimer*

*[69402.212777] The following is the list of last 50 interrupts for each
CPU*

*[69402.212868] [69402.145532]: irqnumber 1023 cpuid 0*

*[69402.212960] [69402.145727]: irqnumber 99 cpuid 0*



And my ARM board restarted immediately.

I think the test case had made my board run out of CPU power(CPU
overloading ), so the watchdog timer has identified this problem and
restarted the board.

To resolve this problem, I have added a delay with 1 millisecond in
“do_test()” function of cpuacct.sh file as below, it will help to release
CPU resources

*for i in `seq 1 $max`; do*

*                        for j in `seq 1 $nbprocess`; do*

*                                    cpuacct_task
$testpath/subgroup_$i/tasks &*

*                                    sleep 0.001*

*                        done*

*            done*



But I worry that the solution makes the test case loss its meaning, whether
the test case is a type of stress test or not!



Could you please confirm that the solution (adding sleep function) is a
right way?

If not, could you please tell me the right way to resolve this problem!

Can I change its arguments from ./cpuacct.sh 100 100 to ./cpuacct.sh 50 50?



I look forward to your reply.



Thanks / Best regards,

*Le Minh Nhan*
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.linux.it/pipermail/ltp/attachments/20171122/cdfa3cd0/attachment-0001.html>

[LTP] [PROBLEM_WITH_CPUACCT.SH_100_100_LTP_TESTSUITE]

[Cyril Hrubis]

Hi!
> I had run the test case ???./cpuacct.sh 100 100??? on my ARM board.
> 
> Afterward,  I have gotten a problem as bellow:
> 
> 
> 
> *1 TINFO: Creating 100 subgroups each with 100 processes*
> 
> *[69402.212655] imx2_wdt_interrupt: pretimer*
> 
> *[69402.212777] The following is the list of last 50 interrupts for each
> CPU*
> 
> *[69402.212868] [69402.145532]: irqnumber 1023 cpuid 0*
> 
> *[69402.212960] [69402.145727]: irqnumber 99 cpuid 0*
> 
> 
> 
> And my ARM board restarted immediately.

If that happens the odds are that you hardware/kernel is buggy.

> I think the test case had made my board run out of CPU power(CPU
> overloading ), so the watchdog timer has identified this problem and
> restarted the board.

There is no such thing CPU overload as far as I can tell. For small
system it may be the case that you are running out of memory but if
adding sleep to the loop fixes the problem that shouldn't be the case.

So all in all this looks like buggy kernel to me.

-- 
Cyril Hrubis
chrubis@suse.cz

[LTP] [PROBLEM_WITH_CPUACCT.SH_100_100_LTP_TESTSUITE]

[Na Nha]

Hi!


Many thanks for your reply.

Could you explain further to me that why it’s not kind of CPU overload?

Actually, I had given it 2 tries for 2 cases: fresh cpuacct.sh
(non-modified) and cpuacct.sh with delay function.

Besides I had calculated the percentage of CPU usage from /proc/stat
from this project https://github.com/pcolby/scripts along with.

So what I have observed that the first case (without delay) always
consumes 100% CPU usage and my ARM board was restarted immediately
then.

On the other hand, after adding delay function into cpuacct.sh, it was
only 50% and nothing wrong happened.


So I think it’s truly kind of CPU overload.


Trân trọng / Best regards,

Nhan Le Minh

2017-11-22 21:31 GMT+07:00 Cyril Hrubis <chrubis@suse.cz>:
> Hi!
>> I had run the test case ???./cpuacct.sh 100 100??? on my ARM board.
>>
>> Afterward,  I have gotten a problem as bellow:
>>
>>
>>
>> *1 TINFO: Creating 100 subgroups each with 100 processes*
>>
>> *[69402.212655] imx2_wdt_interrupt: pretimer*
>>
>> *[69402.212777] The following is the list of last 50 interrupts for each
>> CPU*
>>
>> *[69402.212868] [69402.145532]: irqnumber 1023 cpuid 0*
>>
>> *[69402.212960] [69402.145727]: irqnumber 99 cpuid 0*
>>
>>
>>
>> And my ARM board restarted immediately.
>
> If that happens the odds are that you hardware/kernel is buggy.
>
>> I think the test case had made my board run out of CPU power(CPU
>> overloading ), so the watchdog timer has identified this problem and
>> restarted the board.
>
> There is no such thing CPU overload as far as I can tell. For small
> system it may be the case that you are running out of memory but if
> adding sleep to the loop fixes the problem that shouldn't be the case.
>
> So all in all this looks like buggy kernel to me.
>
> --
> Cyril Hrubis
> chrubis@suse.cz

[LTP] [PROBLEM_WITH_CPUACCT.SH_100_100_LTP_TESTSUITE]

[Cyril Hrubis]

Hi!
> Could you explain further to me that why it???s not kind of CPU overload?

If computer reboots under heavy CPU load it's broken.

-- 
Cyril Hrubis
chrubis@suse.cz