Re: [PATCH for 2.5] preemptible kernel

[Keith Owens <kaos@ocs.com.au>]

Nigel Gamble wrote:
> A task that has been preempted is on the run queue and can be
> rescheduled on a different CPU, so I can't see how a per-CPU counter
> would work.  It seems to me that you would need a per run queue
> counter, like the example I gave in a previous posting.

Ouch.  What about all the per cpu structures in the kernel, how do you
handle them if a preempted task can be rescheduled on another cpu?

 int count[NR_CPUS], *p;
 p = count+smp_processor_id(); /* start on cpu 0, &count[0] */
 if (*p >= 1024) {
   /* preempt here, reschedule on cpu 1 */
   *p = 1;  /* update cpu 0 count from cpu 1, oops */
  }

Unless you find every use of a per cpu structure and wrap a spin lock
around it, migrating a task from one cpu to another is going to be a
source of wierd and wonderful errors.  Since the main use of per cpu
structures is to avoid locks, adding a spin lock to every structure
will be a killer.  Or have I missed something?