Re: [RFC 2/6] Provide cgroup interface for manual jitter classification

[Parth Shah <parth015@linux.vnet.ibm.com>]

On 4/11/19 7:16 PM, Dietmar Eggemann wrote:
> On 3/22/19 7:06 AM, Parth Shah wrote:
>> Signed-off-by: Parth Shah <parth015@linux.vnet.ibm.com>
>> ---
>>   include/linux/sched.h |  8 ++++++++
>>   kernel/sched/core.c   | 46 +++++++++++++++++++++++++++++++++++++++++++
>>   kernel/sched/fair.c   | 15 +++++++++-----
>>   kernel/sched/sched.h  |  2 ++
>>   4 files changed, 66 insertions(+), 5 deletions(-)
>>
>> diff --git a/include/linux/sched.h b/include/linux/sched.h
>> index 903ef29b62c3..be8e8617e0cb 100644
>> --- a/include/linux/sched.h
>> +++ b/include/linux/sched.h
>> @@ -604,6 +604,14 @@ struct task_struct {
>>       unsigned int            flags;
>>       unsigned int            ptrace;
>>   +    /*
>> +     * task_characteristics:
>> +     * 0 = unknown. Follows regular CFS policy for task placement.
>> +     * 1 = Throughput intensive high utilization task
>> +     * 2 = Jitter tasks. Should be packed to reduce active core count.
> 
> This is very similar to the ideas in [PATCH v8 00/16] Add utilization clamping support https://lore.kernel.org/lkml/20190402104153.25404-1-patrick.bellasi@arm.com . Task classification from userspace ...
> 
> [...]
> 

Patches seems interesting and converges to the process of jitter classification.
I will try to re-base my patches on top of that.

Thanks