Use of SCHED_SOFTIRQ

[xerofoify@gmail.com (nick)]

Vignesh,
You forgot to mention that schedule does not itself pick the next task.
The actual function called is pick_next_task which uses the leftmost
cached entry in the red black tree of able to run processes to run
next.
Nick

On 2015-01-28 12:53 PM, Vignesh Radhakrishnan wrote:
> Hey Sreejith,
> 
> 
> softirq's are run as bottom half processing (after an interrupt is handled
> or when there is no work to be done -
> https://www.kernel.org/doc/htmldocs/kernel-hacking/basics-softirqs.html )
> and sched_softirq is one such soft irq whose only function is confined to
> the routine run_rebalance_domain().
> 
> 
> __schedule() is the main scheduling function that tries to pick the next
> task and then perform context switch and other associated scheduling
> functions. Since run_rebalance_domain() need not be run at that exact time
> frame, it is scheduled for later time when the cpu can take it up using
> softirq's. Therefore __schedule() need not be called in this softirq
> context as such because these are independent operations.
> 
> Hope this helps.
> 
> Thanks and regards,
> Vignesh Radhakrishnan
> 
> On Wed, Jan 28, 2015 at 10:25 PM, Sreejith M M <sreejith.mm@gmail.com>
> wrote:
> 
>> Hi,
>>
>> I was reading LKD by Robert Love. I got the following idea from the book.
>>
>>
>> Correct me if I am worng
>> I was checking through source code and I found that on every timer
>> interrupt, through sched/fair.c we are raising the SCHED_SOFTIRQ().
>> I was checking the relation between SCHED_SOFTIRQ and actual
>> __schedule() function.
>>
>> My assumption:
>> schedule() function is the function which selects the processes which
>> are ready to run  in run queue. schedule() function is called in every
>> timer tick.
>>
>> What I was thinking is that schedule() will be called as part of
>> handling SCHED_SOFTIRQ() . But in source code, SCHED_SOFTIRQ is
>> handled through run_rebalance_domain() function (sched/fair.c) . I am
>> unable to trace __schedule()  from this function.
>>
>> Am I missing anything or my assumptions are wrong?
>> --
>> Regards,
>> Sreejith
>>
>> _______________________________________________
>> Kernelnewbies mailing list
>> Kernelnewbies at kernelnewbies.org
>> http://lists.kernelnewbies.org/mailman/listinfo/kernelnewbies
>>
> 
> 
> 
> 
> 
> _______________________________________________
> Kernelnewbies mailing list
> Kernelnewbies at kernelnewbies.org
> http://lists.kernelnewbies.org/mailman/listinfo/kernelnewbies
>