From: Andrey Ryabinin <aryabi...@virtuozzo.com>
Add 'lat' field to the mm_page_alloc tracepoint. It shows allocation
latency in microseconds (0.000001 second).
Signed-off-by: Andrey Ryabinin <aryabi...@virtuozzo.com>
Reviewed-by: Cyrill Gorcunov <gorcu...@virtuozzo.com>
++++++++++++
mm/page_alloc: fix latency in tracepoint.
Since transition from jiffies to sched_clock(), mm_page_alloc
garbadge since we substract from jiffies nanoseconds.
Fix this.
Fixes: e9e7602565ab ("mm/page_alloc: use sched_clock() instead of jiffies to
measure latency")
Signed-off-by: Andrey Ryabinin <aryabi...@virtuozzo.com>
++++++++++++
mm/page_alloc: fix latency in tracepoint - addon
Compilation warning fixed: use proper type.
Fixes: 7eff199a7a9e ("mm/page_alloc: fix latency in tracepoint.")
Signed-off-by: Andrey Ryabinin <aryabi...@virtuozzo.com>
++++++++++++
mm/trace: fix always 0 latency in page_alloc tracepoint
Since __entry->time wasn't assigned, mm_page_alloc trace point
shows always 0 lat.
Fixes: fc8961c87c00 ("mm/page_alloc: add latency to the page_alloc tracepoint")
Signed-off-by: Andrey Ryabinin <aryabi...@virtuozzo.com>
Rebased to vz8 and merged following patches:
- b3e2b2367630 ("mm/page_alloc: add latency to the page_alloc tracepoint")
- 5ae153305c03 ("mm/page_alloc: fix latency in tracepoint.")
- b6be69d05fd4 ("mm/page_alloc: fix latency in tracepoint - addon")
- 0662cba1feed ("mm/trace: fix always 0 latency in page_alloc tracepoint")
Signed-off-by: Andrey Zhadchenko <andrey.zhadche...@virtuozzo.com>
diff --git a/include/trace/events/kmem.h b/include/trace/events/kmem.h
index eb57e30..e6c62ea 100644
--- a/include/trace/events/kmem.h
+++ b/include/trace/events/kmem.h
@@ -192,15 +192,16 @@
TRACE_EVENT(mm_page_alloc,
TP_PROTO(struct page *page, unsigned int order,
- gfp_t gfp_flags, int migratetype),
+ gfp_t gfp_flags, int migratetype, u64 time),
- TP_ARGS(page, order, gfp_flags, migratetype),
+ TP_ARGS(page, order, gfp_flags, migratetype, time),
TP_STRUCT__entry(
__field( unsigned long, pfn )
__field( unsigned int, order )
__field( gfp_t, gfp_flags )
__field( int, migratetype )
+ __field( u64, time )
),
TP_fast_assign(
@@ -208,14 +209,16 @@
__entry->order = order;
__entry->gfp_flags = gfp_flags;
__entry->migratetype = migratetype;
+ __entry->time = time;
),
- TP_printk("page=%p pfn=%lu order=%d migratetype=%d gfp_flags=%s",
+ TP_printk("page=%p pfn=%lu order=%d migratetype=%d gfp_flags=%s
lat=%llu",
__entry->pfn != -1UL ? pfn_to_page(__entry->pfn) : NULL,
__entry->pfn != -1UL ? __entry->pfn : 0,
__entry->order,
__entry->migratetype,
- show_gfp_flags(__entry->gfp_flags))
+ show_gfp_flags(__entry->gfp_flags),
+ __entry->time)
);
DECLARE_EVENT_CLASS(mm_page,
diff --git a/mm/page_alloc.c b/mm/page_alloc.c
index b5afa2a..f66accf 100644
--- a/mm/page_alloc.c
+++ b/mm/page_alloc.c
@@ -4552,14 +4552,13 @@ void update_maxlat(struct kstat_lat_snap_struct
*alloc_lat,
}
static void __alloc_collect_stats(gfp_t gfp_mask, unsigned int order,
- struct page *page, u64 time)
+ struct page *page, u64 time, u64 current_clock)
{
#ifdef CONFIG_VE
unsigned long flags;
- u64 current_clock, delta;
+ u64 delta;
int ind, cpu;
- current_clock = sched_clock();
delta = current_clock - time;
if (!(gfp_mask & __GFP_RECLAIM))
ind = KSTAT_ALLOCSTAT_ATOMIC;
@@ -4596,7 +4595,7 @@ struct page *
unsigned int alloc_flags = ALLOC_WMARK_LOW;
gfp_t alloc_mask; /* The gfp_t that was actually used for allocation */
struct alloc_context ac = { };
- u64 start;
+ u64 start, current_clock;
gfp_mask &= gfp_allowed_mask;
alloc_mask = gfp_mask;
@@ -4639,8 +4638,10 @@ struct page *
page = NULL;
}
- __alloc_collect_stats(alloc_mask, order, page, start);
- trace_mm_page_alloc(page, order, alloc_mask, ac.migratetype);
+ current_clock = sched_clock();
+ __alloc_collect_stats(alloc_mask, order, page, start, current_clock);
+ trace_mm_page_alloc(page, order, alloc_mask, ac.migratetype,
+ (current_clock - start));
return page;
}
--
1.8.3.1
_______________________________________________
Devel mailing list
Devel@openvz.org
https://lists.openvz.org/mailman/listinfo/devel
