On Sat 18. Feb - 12:53:25, Stefan Seyfried wrote:
> Hi,
>
> by looking at the code and checking who generates huge system load on
> a hp machine that throws lots of thermal zone events i found out that
> ENABLE_THERMAL_MANAGEMENT now defaults to "kernel_passive".
> Is there a reason for this? I mean: the wohle thermal code does not
> (and can not) work on most machines and it might cause lots of
> /proc/acpi/thermal_zone/* polling which in turn causes cpu-load in the
> kernel - something i'd like to avoid when programming a power_SAVE_
> daemon ;-)
>
> So is there a reason for this not being "off" as a default?
Thomas (CC) said he has quite some good reasons for having
kernel_passive. But he also mentioned that there is still some work todo.
>
> What i wanted to do (and why i stumbled onto it) was:
> I wanted to put a check for "THERMAL_MANAGEMENT =! off" into the beginning
> of void PM_Interface::checkTemperatureStateChanges() in pm_interface.cpp
> because this function is called on every thermal_zone ACPI event and it
> is right now quite expensive. The downside would be that we would not get
> thermal events in powersaved if THERMAL_MANAGEMENT was set to "off", but
> there is not much useful we can do with them in that case anyway.
>
> The problem is really that on those HP machines, whenever you cross a trip
> point is crossed an acpi event is thrown and when the machine is cooling down,
> the cpu activity caused by this acpi event might be enough to cross the trip
> point again - upwards. This leaves the machine hovering around the trippoint
> all the time, with ~10-20% cpu load in kernel space, probably because of all
> that /proc/acpi polling which is grossly inefficient but we cannot do much
> about the ACPI braindamage.
>
> Any opinions?
I'm quite unemotional regarding all this and would go for setting it to
"off" by default if we can't fix it.
Thomas, please fix your code and put your changes in that you still want
to do. And this ASAP. And ASAP means during upcoming week ;-)
Seife, can you please create a bugreport for this? We are quite late for
for the next stable release.
Regards,
Holger
_______________________________________________
powersave-devel mailing list
[email protected]
http://forge.novell.com/mailman/listinfo/powersave-devel
