On Wed, 2006-06-07 at 12:58, Frank Ch. Eigler wrote:
> jkenisto wrote:
> 
...
> 
> > [...]  Yes, pretty slow.  In my prototype, my user-mode handler just
> > increments a counter.  On my Pentium M, overhead per probepoint hit
> > is ~14.2 usec, compared with 1.03 usec for the uprobes version last
> > posted to LKML.  [...]
> 
> It would be good to see your code.  I'm surprised the slowdown is
> *only* a factor of 14.

Code (and README, tests, bug report) attached.

The slowdown is still about 14x (over LKML uprobes), even after
generalizing for multiple probes and multiple processes.  It slows down
significantly (and often breaks -- see README) when probing multiple
heavily-probe-burdened processes simultaneously.

I'm out of the office 'til Monday, BTW.

Jim