system 20% busy at all times?

Fleuriot Damien ml at my.gd
Tue Feb 19 09:54:38 UTC 2013


On Feb 19, 2013, at 10:44 AM, "Eggert, Lars" <lars at netapp.com> wrote:

> Hi,
> 
> On Feb 19, 2013, at 10:40, Fleuriot Damien <ml at my.gd>
> wrote:
>> What about reviewing top(1) ?
> 
> top shows the ~20% I mentioned:
> 
> last pid:  3176;  load averages:  0.79,  0.80,  0.84                                                                                     up 0+14:49:49  09:43:51
> 17 processes:  1 running, 16 sleeping
> CPU:  0.0% user,  0.0% nice, 18.7% system,  0.0% interrupt, 81.3% idle
> Mem: 32M Active, 9456K Inact, 196M Wired, 19M Buf, 15G Free
> Swap: 
> 


Ok the ~20% is from the system itself.


> 
>> or possibly ps(1) aufx
> 
> # ps -aufx
> USER   PID  %CPU %MEM    VSZ  RSS TT  STAT STARTED       TIME COMMAND
> root    10 346.8  0.0      0   64  -  RL    6:54PM 2862:46.43 [idle]
> root     0  64.1  0.0      0  496  -  DLs   6:54PM  694:47.32 [kernel]
> 

65% of a CPU core only for the kernel...


> 
> 
>> At least you should be able to see what takes up CPU:
>> - system
>> - user processes
>> - interrupts
> 
> # vmstat -i
> interrupt                          total       rate
> irq3: uart1                        11535          0
> irq4: uart0                         1227          0
> irq9: acpi0                   1989762564      37379
> irq16: uhci0 uhci1+                  393          0
> cpu0:timer                      32147924        603
> irq270: em4                      1907258         35
> cpu3:timer                      63027976       1184
> cpu2:timer                      56428246       1060
> cpu1:timer                      44799884        841
> Total                         2188087007      41104
> 
> So it seems that irq 9 is firing a whole lot. Why?


And indeed we find your answer here, acpi0 firing up a lot of interrupts.

Don't you get any message about that in dmesg -a or /var/log/messages ?

I'd expect something like "interrupt storm blabla… source throttled blabla.."



From man 4 acpi , in /boot/loader.conf :
hint.acpi.0.disabled=1
             Set this to 1 to disable all of ACPI.  If ACPI has been disabled
             on your system due to a blacklist entry for your BIOS, you can
             set this to 0 to re-enable ACPI for testing.


Any chance you could reboot the host with ACPI disabled ?


If that helps your CPU load, try setting this in /boot/loader.conf :
hw.acpi.verbose=1
	Turn on verbose debugging information about what ACPI is doing.


Hoping this gets some logs :)



More information about the freebsd-current mailing list