> - * Return 1 if free pages are above 'mark'. This takes into account the order
> + * Return true if free pages are above 'mark'. This takes into account the order
> * of the allocation.
> */
> -int zone_watermark_ok(struct zone *z, int order, unsigned long mark,
> - int classzone_idx, int alloc_flags)
> +bool __zone_watermark_ok(struct zone *z, int order, unsigned long mark,
> + int classzone_idx, int alloc_flags, long free_pages)
static?
> diff --git a/mm/vmscan.c b/mm/vmscan.c
> index c5dfabf..ba0c70a 100644
> --- a/mm/vmscan.c
> +++ b/mm/vmscan.c
> @@ -2082,7 +2082,7 @@ static int sleeping_prematurely(pg_data_t *pgdat, int order, long remaining)
> if (zone->all_unreclaimable)
> continue;
>
> - if (!zone_watermark_ok(zone, order, high_wmark_pages(zone),
> + if (!zone_watermark_ok_safe(zone, order, high_wmark_pages(zone),
> 0, 0))
> return 1;
> }
Do we need to change balance_pgdat() too?
Otherwise, balance_pgdat() return immediately and can make semi-infinite busy loop.
> diff --git a/mm/vmstat.c b/mm/vmstat.c
> index 355a9e6..ddee139 100644
> --- a/mm/vmstat.c
> +++ b/mm/vmstat.c
> @@ -81,6 +81,12 @@ EXPORT_SYMBOL(vm_stat);
>
> #ifdef CONFIG_SMP
>
> +static int calculate_pressure_threshold(struct zone *zone)
> +{
> + return max(1, (int)((high_wmark_pages(zone) - low_wmark_pages(zone) /
> + num_online_cpus())));
> +}
On Shaohua's machine,
CPU: 64
MEM: 8GBx4 (=32GB)
per-cpu vm-stat threashold: 98
zone->min = sqrt(32x1024x1024x16)/4 = 5792 KB = 1448 pages
zone->high - zone->low = zone->min/4 = 362pages
pressure-vm-threshold = 362/64 ~= 5
Hrm, this reduction seems slightly dramatically (98->5).
Shaohua, can you please rerun your problem workload on your 64cpus machine with
applying this patch?
Of cource, If there is no performance degression, I'm not against this one.
Thanks.
--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@xxxxxxxxxx For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@xxxxxxxxx";> email@xxxxxxxxx </a>