Mark Haywood wrote:
> Eric Saxe wrote:
>
>> I'm thinking the cpu_change_speed and cstate_transition DTrace probes
>> (which are already in the PowerTop repo) are generally useful. So i'm
>> interested in putting them back to Nevada. Thoughts? Moving forward,
>> this could eliminate the PowerTop's dependence on the kernel.
>>
>>
>
> Yep. I think that the cpu_change_speed probe is a good idea. Didn't see
> the cstate_transition probe, but it seems like a good idea too. I'd
> rather we used Dtrace probes instead of kstats.
I've filed:
6635705 static DTrace probe points for CPU power management state
transitions
I'll send out a webrev when it's implemented. DTrace probes can have a
level of stability associated
with them. I might propose that they be considered "unstable" to start,
and then after we're convinced
that the interface is correct and sufficient we can go through ARC to
promote their stability. Does that
seem reasonable?
Thanks,
-Eric
