Re: [PATCH v6 1/2] nohz: add tick_nohz_full_add_cpus_to() and _remove_cpus_from() APIs

[Chris Metcalf <cmetcalf@ezchip.com>]

On 04/09/2015 09:34 PM, Frederic Weisbecker wrote:
> On Thu, Apr 09, 2015 at 02:01:45PM -0400, Chris Metcalf wrote:
>> --- a/include/linux/tick.h
>> +++ b/include/linux/tick.h
>> @@ -186,6 +186,18 @@ static inline bool tick_nohz_full_cpu(int cpu)
>>   	return cpumask_test_cpu(cpu, tick_nohz_full_mask);
>>   }
>>   
>> +static inline void tick_nohz_full_add_cpus_to(struct cpumask *mask)
> Or tick_nohz_full_affine() ?

I'd vote no to that, I think - the first problem I see is that it doesn't
make very clear that it modifies the argument, which is the problem
Peter Z has been having from the beginning with some suggestions.
The second problem is that it sounds like it might be setting the
argument unconditionally to the nohz_full set, when in fact it's doing
a cpumask_or() to increase the set of cpus further.

> +	/* nohz_full won't take effect without isolating the cpus. */
> +	tick_nohz_full_remove_cpus_from(cpu_isolated_map);
> This should be the other one I guess.
>

<Slaps self on forehead>  Thanks!

Now off to do some actual testing before sending the next patchset :-)

-- 
Chris Metcalf, EZChip Semiconductor
http://www.ezchip.com