dtrace discuss - Apr 2006 - Is it possible to dtrace system clock interrupt handler?

if no, why?
if yes, how?
Thanks a lot
 
 
This message posted from opensolaris.org

it''s a  dumb q, I can trace clock, setsoftint etc.
 
 
This message posted from opensolaris.org

Qiang Liu wrote:
> if no, why?
> if yes, how?
> Thanks a lot
>  
>  
> This message posted from opensolaris.org
> _______________________________________________
> dtrace-discuss mailing list
> dtrace-discuss at opensolaris.org

Like this:

# dtrace -f clock''{stack(5);exit(0);}''
dtrace: description ''clock'' matched 2 probes
CPU     ID                    FUNCTION:NAME
   0   3437                      clock:entry
               genunix`cyclic_softint+0xc3
               unix`cbe_softclock+0x1e
               unix`av_dispatch_softvect+0x64
               unix`dosoftint+0x32


#


- Bart

Bart Smaalders			Solaris Kernel Performance
barts at cyber.eng.sun.com		http://blogs.sun.com/barts

On Fri 07 Apr 2006 at 05:27PM, Qiang Liu wrote:
> it''s a  dumb q, I can trace clock, setsoftint etc.
>  
If what you want, however, is a periodic probe firing, use the profile
(fires on all cpus) or tick (fires on 1 cpu) provider:

Here are some examples:

dtrace -n ''profile:::tick-173ms''

and

dtrace -q -n ''profile:::tick-1s{printf("the time is: %Y\n",
walltimestamp);}''

and

dtrace -n ''profile:::profile-997hz{@[execname]=count();}''

        -dp

-- 
Daniel Price - Solaris Kernel Engineering - dp at eng.sun.com - blogs.sun.com/dp

Hi Dan
I tried the following dtrace but no output  after 10 sec until I ctl-c in the
case of heady IO load in the system. Witout heavy IO load, I run the same
script, it returns in 10 sec automatically without using ctl-c. BTW both return
the same result.
why I have to use ctl-c in the first case.

#dtrace -n ''profile-100 { @[cpu] = count() } tick-10s
{exit(0)}''
 
 
This message posted from opensolaris.org