lkml.org 
[lkml]   [2020]   [Feb]   [25]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
Patch in this message
/
SubjectRe: [PATCH v4 3/4] sched: Allow sched_{get,set}attr to change latency_nice of the task
From
Date


On 2/25/20 12:24 PM, Pavan Kondeti wrote:
> On Mon, Feb 24, 2020 at 02:29:17PM +0530, Parth Shah wrote:
>
> [...]
>
>> diff --git a/kernel/sched/core.c b/kernel/sched/core.c
>> index 65b6c00d6dac..e1dc536d4ca3 100644
>> --- a/kernel/sched/core.c
>> +++ b/kernel/sched/core.c
>> @@ -4723,6 +4723,8 @@ static void __setscheduler_params(struct task_struct *p,
>> p->rt_priority = attr->sched_priority;
>> p->normal_prio = normal_prio(p);
>> set_load_weight(p, true);
>> +
>> + p->latency_nice = attr->sched_latency_nice;
>> }
>
> We don't want this when SCHED_FLAG_LATENCY_NICE is not set in
> attr->flags.
>
> The user may pass SCHED_FLAG_KEEP_PARAMS | SCHED_FLAG_LATENCY_NICE to
> change only latency nice value. So we have to update latency_nice
> outside __setscheduler_params(), I think.


AFAICT, passing SCHED_FLAG_KEEP_PARAMS with any other flag will prevent us
from changing the latency_nice value because of the below code flow:

__sched_setscheduler()
__setscheduler()
if (attr->sched_flags & SCHED_FLAG_KEEP_PARAMS) return;
__setscheduler_params()

whereas, one thing we still can do is add if condition when setting the
value, i.e.,

@@ -4724,7 +4724,8 @@ static void __setscheduler_params(struct task_struct *p,
p->normal_prio = normal_prio(p);
set_load_weight(p, true);

- p->latency_nice = attr->sched_latency_nice;
+ if (attr->sched_flags & SCHED_FLAG_LATENCY_NICE)
+ p->latency_nice = attr->sched_latency_nice;
}


>>
>> /* Actually do priority change: must hold pi & rq lock. */
>> @@ -4880,6 +4882,13 @@ static int __sched_setscheduler(struct task_struct *p,
>> return retval;
>> }
>>
>> + if (attr->sched_flags & SCHED_FLAG_LATENCY_NICE) {
>> + if (attr->sched_latency_nice > MAX_LATENCY_NICE)
>> + return -EINVAL;
>> + if (attr->sched_latency_nice < MIN_LATENCY_NICE)
>> + return -EINVAL;
>> + }
>> +
>> if (pi)
>> cpuset_read_lock();
>>
>> @@ -4914,6 +4923,9 @@ static int __sched_setscheduler(struct task_struct *p,
>> goto change;
>> if (attr->sched_flags & SCHED_FLAG_UTIL_CLAMP)
>> goto change;
>> + if (attr->sched_flags & SCHED_FLAG_LATENCY_NICE &&
>> + attr->sched_latency_nice != p->latency_nice)
>> + goto change;
>>
>> p->sched_reset_on_fork = reset_on_fork;
>> retval = 0;
>> @@ -5162,6 +5174,9 @@ static int sched_copy_attr(struct sched_attr __user *uattr, struct sched_attr *a
>> size < SCHED_ATTR_SIZE_VER1)
>> return -EINVAL;
>>
>> + if ((attr->sched_flags & SCHED_FLAG_LATENCY_NICE) &&
>> + size < SCHED_ATTR_SIZE_VER2)
>> + return -EINVAL;
>> /*
>> * XXX: Do we want to be lenient like existing syscalls; or do we want
>> * to be strict and return an error on out-of-bounds values?
>> @@ -5391,6 +5406,8 @@ SYSCALL_DEFINE4(sched_getattr, pid_t, pid, struct sched_attr __user *, uattr,
>> else
>> kattr.sched_nice = task_nice(p);
>>
>> + kattr.sched_latency_nice = p->latency_nice;
>> +
>
> Can you consider printing latency_nice value in proc_sched_show_task() in this
> patch/series?

Sure, I will add it.


Thanks,
Parth

\
 
 \ /
  Last update: 2020-02-25 16:04    [W:0.050 / U:7.080 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site