Re: NOHZ: local_softirq_pending

[Iratxo Pichel Ortiz <ipichel@albentia.com>]

Thomas,

Thomas Gleixner wrote:
> No, what happens is that the -RT starvation protector kicks in. That's
> a mechanism which checks whether the CPU is monopolized by rt tasks
> for longer than 950ms. It then blocks execution of rt tasks for
> 50ms. The silly thing is that it goes idle, but that's a different
> problem.
>   
> It's default on and can be disabled via:
>
> echo 0 > /proc/sys/kernel/rt_runtime_us
>
> Sorry that I did not notice that earlier, but when that triggers, then
> something is really wrong. There should never be a situation where -rt
> tasks (neither irq threads nor softirqs) should monopolize the CPU.
>   
No problem at all! Please let me know if the following assumption is 
right: This protection mechanism is dependent on the available CPU 
power. I mean, if the system has low CPU power, and if, for example, the 
network has very very high load, the total time taken to process all the 
data will be long. Generally much longer than in more powerful systems, 
so this protection arises. Is this right or it should not happen by 
design of the softirqs/RT tasks?

Anyway, seems to be correctly working. I still have to check why with  
HZ=100 it was periodically happening this events even with low load, but 
that could be very probably related to my software.

Again, thank you very much.

Best regards,

    Iratxo.
> Thanks,
>
> 	tglx
>
>   


-- 
Iratxo Pichel Ortiz
Software Development Manager

Albentia Systems S.A.
http://www.albentia.com

Tel: +34 914400567
Cel: +34 663808405
Fax: +34 914400569

C\Margarita Salas 22
Parque Tecnológico de Leganés
Leganés (28918)
Madrid
Spain


--
To unsubscribe from this list: send the line "unsubscribe linux-rt-users" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html