[dtrace-discuss] kcopy() untraceble on X4500/b65 ?

[Paul van der Zwan]

I am trying to run a benchmark on an X4500 and see the system 75%  
busy in kernel mode, lockstat output
shows:
# lockstat -Ikw sleep 10

Profiling interrupt: 4276 events in 11.017 seconds (388 events/sec)

Count indv cuml rcnt     nsec CPU+PIL                Hottest Caller
------------------------------------------------------------------------ 
-------
1010  24%  24% 0.00     3419 cpu[3]                 kcopy
1009  24%  47% 0.00     3926 cpu[1]                 kcopy
   974  23%  70% 0.00     4006 cpu[0]                 kcopy
   907  21%  91% 0.00     4001 cpu[2]                 (usermode)
    85   2%  93% 0.00     3762 cpu[2]+11              setbackdq
    84   2%  95% 0.00     3831 cpu[0]+11              setbackdq
    60   1%  97% 0.00     3614 cpu[1]+11              splr
    59   1%  98% 0.00     3382 cpu[3]+11              restorectx
    44   1%  99% 0.00     3923 cpu[2]+5               ddi_io_get32
    33   1% 100% 0.00     3920 cpu[2]+4               sdintr
    11   0% 100% 0.00     1684 cpu[0]+10              set_anoninfo
------------------------------------------------------------------------ 
-------


Trying the figure out who is calling kcopy() I tried:
#  dtrace -n 'fbt::kcopy:entry { @[stack(2)] = count()}'
But the only output I got was:

dtrace: description 'fbt::kcopy:entry ' matched 1 probe

And ^C after a while gave no output....


Any tips on how to figure out why kcopy() is saturating 3 opteron  
cores ??

	Paul