Re: [patch 1/2] fork_connector: add a fork connector

[Paul Jackson <pj@engr.sgi.com>]

Dean wrote:
> by the time do_exit() occurs the parent may have disappeared

I don't think Jay was disagreeing with this.  I think he agrees
that there is to be collected:
 1) the classic bsd accounting data, in do_exit
 2) the fork time <parent pid, child pid> by some mechanism at
	fork time (perhaps just not the fork_connect mechanism)
 3) some additional data to be harvested at exit time, for CSA

I suspect you two are just tripping over words to describe this.

However, this does expose another possibility.  Record the original
forking parent pid in another task_struct field at fork time (didn't
someone else have a 'bio_pid' patch to this affect?), and add that task
struct value to the list of additional items to be written out, at exit
time.

I was skeptical that CBUS could have zero impact on fork, but recording
one more word in the task struct at fork gets about as close to zero
impact as one can get on fork.

-- 
                  I won't rest till it's the best ...
                  Programmer, Linux Scalability
                  Paul Jackson <pj@engr.sgi.com> 1.650.933.1373, 1.925.600.0401