Re: [Xen-users] Call schedule set on arinc653 scheduler?

[Nathan Studer <nate.studer@xxxxxxxxx>]

On Wed, Jun 3, 2015 at 10:34 AM, Mr Idris <mr@xxxxxxxxxxxx> wrote:
> On Wed, Jun 3, 2015 at 4:28 PM, Mr Idris <mr@xxxxxxxxxxxx> wrote:
>>
>> Hi all,
>>
>> I have managed to call arinc653_scheduler_set.c without error.  The
>> message when i run it like this
>>
>> not error
>> not error
>> hypercall bounce and schedule set finish *
>> true
>>
>> * this message because i set on xc_sched_arinc653_schedule_set().
>>
>>
>> but when i try 'xl list -v' still VM is not running
>
>
> I'm sorry accidentally i press send but i haven't finished.
>
> I continue, but when i try 'xl list -v' still VM is not running like this :
> Name                                        ID   Mem VCPUs      State
> Time(s)   UUID                            Reason-Code   Security Label
> Domain-0                                     0  6771     1     r-----
> 10.0 00000000-0000-0000-0000-000000000000        -                -
> Debian                                       1   512     1     ------
> 0.0 938b9c5b-8d9d-402a-9be0-0e0cc4cf67dc        -                -
>
>
> Something weird after the small program run, the computer is becoming really
> slow. Is it something related to runtime?

That's how you know it's working!  The arinc653 scheduler is not work
conserving or pre-emptive, so you should expect some performance
degradation.  It probably should not be that bad, so I think it is a
symptom of the problem below.

> Does anyone have any idea what change I need to make to get the scheduler to
> run
> the VM?  I appreciate the help.

From the attached program, which is similar to your previous program:

sched.sched_entries[0].vcpu_id = 0;
sched.sched_entries[0].runtime = 30;
sched.major_frame += sched.sched_entries[0].runtime;

The runtime field is in units of nanoseconds.  30 nanoseconds is
orders of magnitude shorter than the context switch time.  I'm not
sure what the scheduler would do with a runtime this small, but it
would not be pretty.  For most configurations, the slice runtimes
should be in the milliseconds range, so multiple your runtimes by
1000000, and see if that fixes your issue.

sched.sched_entries[*].runtime = 10000000;  /* 10 ms */

     Nate

>
> Thanks,
> Idris
>
>
>

_______________________________________________
Xen-users mailing list
Xen-users@xxxxxxxxxxxxx
http://lists.xen.org/xen-users