Re: CURRENT Kernel makes the system run very very hot

On Saturday 08 December 2007 21:57:18 韓家標 Bill Hacker wrote:
> Thomas Sparrevohn wrote:
> 
> Nothing readly visible, anyway.
> 
> Looks as if there is something looping as fast as the clock can push it, and at 
> a primitive or 'deep' enough level the tools available can't see it.
> 
> Are there any kernel build options aside from GENERIC?
> 
> Which kld's in use?
> 
> If no clues there, in the absence of panel switches for RUN/HALT/SS, I'd suggest 
> building or installing at least a new kernel, if not world.
> 
> Bill

There are no other options but GENERIC (AMD64) - I had "rtc.ko" loaded - But I have removed that
which helped a little bit (I can sit beside the machine when its running single job "make buildkernel" ).

On my standard make -j 8 it still gets very hot - which it never has done until very recently

World and Kernel was build this afternoon - I normally rebuild every day - kld loads

Id Refs Address            Size     Name
 1   14 0xffffffff80100000 b18a70   kernel (/boot/kernel/kernel)
 2    1 0xffffffff80c19000 f5f18    zfs.ko (/boot/kernel/zfs.ko)
 3    2 0xffffffffd9c0b000 1874f    linux.ko (/boot/kernel/linux.ko)
 4    1 0xffffffffd9c42000 346d     linprocfs.ko (/boot/kernel/linprocfs.ko)
 5    1 0xffffffffd9c96000 1076     daemon_saver.ko (/boot/kernel/daemon_saver.ko)
 6    1 0xffffffffd9c98000 729      coretemp.ko (/boot/kernel/coretemp.ko)


What are the most reliable way to get idle time? Running vmstat and sysctl dev.cpu | grep temper - seems
to indicate that the grow hotter even when idle is quite high e.g. 25-50% - but that could just be a measurement error