Messages in this thread |  | | | Date | Mon, 21 Jan 2008 08:16:20 -0800 | | From | Arjan van de Ven <> | | Subject | Re: LatencyTOP infrastructure patch | |
KOSAKI Motohiro wrote:
> Hi
>
>> +static void __sched
>> +account_global_scheduler_latency(struct task_struct *tsk, int usecs)
>> +{
>> + int i;
>> + int firstnonnull = MAXLR + 1;
>> +
>> + if (!tsk->latency_reason.reason)
>> + return;
>> +
>> + /* skip kernel threads for now */
>> + if (!tsk->mm)
>> + return;
>
> Why do you ignore kernel thread?
>
> may be, some network filesystem use kernel thread for local I/O.
> What do you think it?
kernel threads (by themselves) don't tend to result in user visible latency...
at least that's my assumption so far... if you think I'm wrong... I'm open to be shown
that we should count them.
|  |