Add an above_background_load() function which can be used by other subsystems
to detect if there is anything besides niced tasks running. Place it in
sched.h to allow it to be compiled out if not used.
Signed-off-by: Con Kolivas <[email protected]>
---
include/linux/sched.h | 16 ++++++++++++++++
1 files changed, 16 insertions(+)
Index: linux-2.6.16-rc6-mm1/include/linux/sched.h
===================================================================
--- linux-2.6.16-rc6-mm1.orig/include/linux/sched.h 2006-03-13 17:05:14.000000000 +1100
+++ linux-2.6.16-rc6-mm1/include/linux/sched.h 2006-03-13 17:24:18.000000000 +1100
@@ -638,6 +638,22 @@ extern unsigned int max_cache_size;
#endif /* CONFIG_SMP */
+/*
+ * A runqueue laden with a single nice 0 task scores a weighted_cpuload of
+ * SCHED_LOAD_SCALE. This function returns 1 if any cpu is laden with a
+ * task of nice 0 or enough lower priority tasks to bring up the
+ * weighted_cpuload
+ */
+static inline int above_background_load(void)
+{
+ unsigned long cpu;
+
+ for_each_online_cpu(cpu) {
+ if (weighted_cpuload(cpu) >= SCHED_LOAD_SCALE)
+ return 1;
+ }
+ return 0;
+}
struct io_context; /* See blkdev.h */
void exit_io_context(void);
* Con Kolivas <[email protected]> wrote:
> Add an above_background_load() function which can be used by other
> subsystems to detect if there is anything besides niced tasks running.
> Place it in sched.h to allow it to be compiled out if not used.
>
> Signed-off-by: Con Kolivas <[email protected]>
Acked-by: Ingo Molnar <[email protected]>
Ingo
Con Kolivas wrote:
> Add an above_background_load() function which can be used by other subsystems
> to detect if there is anything besides niced tasks running. Place it in
> sched.h to allow it to be compiled out if not used.
>
> Signed-off-by: Con Kolivas <[email protected]>
>
> ---
> include/linux/sched.h | 16 ++++++++++++++++
> 1 files changed, 16 insertions(+)
>
> Index: linux-2.6.16-rc6-mm1/include/linux/sched.h
> ===================================================================
> --- linux-2.6.16-rc6-mm1.orig/include/linux/sched.h 2006-03-13 17:05:14.000000000 +1100
> +++ linux-2.6.16-rc6-mm1/include/linux/sched.h 2006-03-13 17:24:18.000000000 +1100
> @@ -638,6 +638,22 @@ extern unsigned int max_cache_size;
>
> #endif /* CONFIG_SMP */
>
> +/*
> + * A runqueue laden with a single nice 0 task scores a weighted_cpuload of
> + * SCHED_LOAD_SCALE. This function returns 1 if any cpu is laden with a
> + * task of nice 0 or enough lower priority tasks to bring up the
> + * weighted_cpuload
> + */
> +static inline int above_background_load(void)
> +{
> + unsigned long cpu;
> +
> + for_each_online_cpu(cpu) {
> + if (weighted_cpuload(cpu) >= SCHED_LOAD_SCALE)
> + return 1;
> + }
> + return 0;
> +}
>
> struct io_context; /* See blkdev.h */
> void exit_io_context(void);
I think that you may need to take into account the contribution to the
load by your swap prefetching thread when it calls this function
otherwise it could cause an incorrect (from your point of view) positive
return value. If the thread has a positive nice value this comment can
probably be ignored.
If not, then the adjustment could be made by subtracting the threads
load_weight from the weighted_cpuload() when looking at the CPU the
thread is running on. To minimize the effect on the overhead of the
function this refinement could be only performed when the current if
statement succeeds e.g. something like
if (weighted_cpuload(cpu) >= SCHED_LOAD_SCALE) {
if (smp_processor_id() != cpu || (weighted_cpuload(cpu) -
current->load_weight) >= SCHED_LOAD_SCALE)
return 1;
}
Peter
--
Peter Williams [email protected]
"Learning, n. The kind of ignorance distinguishing the studious."
-- Ambrose Bierce
Peter Williams writes:
> Con Kolivas wrote:
>> + if (weighted_cpuload(cpu) >= SCHED_LOAD_SCALE)
>> + return 1;
> I think that you may need to take into account the contribution to the
> load by your swap prefetching thread when it calls this function
> otherwise it could cause an incorrect (from your point of view) positive
> return value. If the thread has a positive nice value this comment can
> probably be ignored.
kprefetchd runs at nice 19.
Cheers,
Con