Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]

Groups > linux.kernel > #1547996 > unrolled thread

[PATCH 0/7] vm, vmscan: enahance vmscan tracepoints

Back to article view | Back to linux.kernel

Contents

  [PATCH 0/7] vm, vmscan: enahance vmscan tracepoints Michal Hocko <mhocko@kernel.org> - 2016-12-28 16:40 +0100
    [PATCH 7/7] mm, vmscan: add mm_vmscan_inactive_list_is_low tracepoint Michal Hocko <mhocko@kernel.org> - 2016-12-28 16:40 +0100
      Re: [PATCH 7/7] mm, vmscan: add mm_vmscan_inactive_list_is_low tracepoint "Hillf Danton" <hillf.zj@alibaba-inc.com> - 2016-12-29 09:30 +0100
    [PATCH 1/7] mm, vmscan: remove unused mm_vmscan_memcg_isolate Michal Hocko <mhocko@kernel.org> - 2016-12-28 16:40 +0100
      Re: [PATCH 1/7] mm, vmscan: remove unused mm_vmscan_memcg_isolate "Hillf Danton" <hillf.zj@alibaba-inc.com> - 2016-12-29 08:50 +0100
    [PATCH 5/7] mm, vmscan: extract shrink_page_list reclaim counters into a struct Michal Hocko <mhocko@kernel.org> - 2016-12-28 16:40 +0100
      Re: [PATCH 5/7] mm, vmscan: extract shrink_page_list reclaim counters into a struct "Hillf Danton" <hillf.zj@alibaba-inc.com> - 2016-12-29 09:10 +0100
    [PATCH 6/7] mm, vmscan: enhance mm_vmscan_lru_shrink_inactive tracepoint Michal Hocko <mhocko@kernel.org> - 2016-12-28 16:40 +0100
      Re: [PATCH 6/7] mm, vmscan: enhance mm_vmscan_lru_shrink_inactive tracepoint "Hillf Danton" <hillf.zj@alibaba-inc.com> - 2016-12-29 09:20 +0100
    [PATCH 4/7] mm, vmscan: show LRU name in mm_vmscan_lru_isolate tracepoint Michal Hocko <mhocko@kernel.org> - 2016-12-28 16:40 +0100
      Re: [PATCH 4/7] mm, vmscan: show LRU name in mm_vmscan_lru_isolate  tracepoint Nikolay Borisov <n.borisov.lkml@gmail.com> - 2016-12-28 17:00 +0100
        Re: [PATCH 4/7] mm, vmscan: show LRU name in mm_vmscan_lru_isolate  tracepoint Michal Hocko <mhocko@kernel.org> - 2016-12-28 17:10 +0100
          Re: [PATCH 4/7] mm, vmscan: show LRU name in mm_vmscan_lru_isolate  tracepoint Nikolay Borisov <n.borisov.lkml@gmail.com> - 2016-12-28 17:50 +0100
            Re: [PATCH 4/7] mm, vmscan: show LRU name in mm_vmscan_lru_isolate  tracepoint Michal Hocko <mhocko@kernel.org> - 2016-12-28 18:00 +0100
      Re: [PATCH 4/7] mm, vmscan: show LRU name in mm_vmscan_lru_isolate  tracepoint Minchan Kim <minchan@kernel.org> - 2016-12-29 07:10 +0100
        Re: [PATCH 4/7] mm, vmscan: show LRU name in mm_vmscan_lru_isolate  tracepoint Michal Hocko <mhocko@kernel.org> - 2016-12-29 09:00 +0100
          Re: [PATCH 4/7] mm, vmscan: show LRU name in mm_vmscan_lru_isolate  tracepoint Minchan Kim <minchan@kernel.org> - 2016-12-30 03:00 +0100
            Re: [PATCH 4/7] mm, vmscan: show LRU name in mm_vmscan_lru_isolate  tracepoint Michal Hocko <mhocko@kernel.org> - 2016-12-30 10:40 +0100
      Re: [PATCH 4/7] mm, vmscan: show LRU name in mm_vmscan_lru_isolate tracepoint "Hillf Danton" <hillf.zj@alibaba-inc.com> - 2016-12-29 09:00 +0100
      Re: [PATCH 4/7] mm, vmscan: show LRU name in mm_vmscan_lru_isolate  tracepoint Vlastimil Babka <vbabka@suse.cz> - 2017-01-03 18:10 +0100
    [PATCH 3/7] mm, vmscan: show the number of skipped pages in mm_vmscan_lru_isolate Michal Hocko <mhocko@kernel.org> - 2016-12-28 16:40 +0100
      Re: [PATCH 3/7] mm, vmscan: show the number of skipped pages in  mm_vmscan_lru_isolate Minchan Kim <minchan@kernel.org> - 2016-12-29 07:00 +0100
      Re: [PATCH 3/7] mm, vmscan: show the number of skipped pages in mm_vmscan_lru_isolate "Hillf Danton" <hillf.zj@alibaba-inc.com> - 2016-12-29 09:00 +0100
      Re: [PATCH 3/7] mm, vmscan: show the number of skipped pages in  mm_vmscan_lru_isolate Vlastimil Babka <vbabka@suse.cz> - 2017-01-03 18:30 +0100
    [PATCH 2/7] mm, vmscan: add active list aging tracepoint Michal Hocko <mhocko@kernel.org> - 2016-12-28 16:40 +0100
      Re: [PATCH 2/7] mm, vmscan: add active list aging tracepoint Minchan Kim <minchan@kernel.org> - 2016-12-29 06:40 +0100
        Re: [PATCH 2/7] mm, vmscan: add active list aging tracepoint Michal Hocko <mhocko@kernel.org> - 2016-12-29 09:00 +0100
          Re: [PATCH 2/7] mm, vmscan: add active list aging tracepoint Minchan Kim <minchan@kernel.org> - 2016-12-30 02:50 +0100
            Re: [PATCH 2/7] mm, vmscan: add active list aging tracepoint Michal Hocko <mhocko@kernel.org> - 2016-12-30 10:30 +0100
              Re: [PATCH 2/7] mm, vmscan: add active list aging tracepoint "Hillf Danton" <hillf.zj@alibaba-inc.com> - 2016-12-30 10:50 +0100
              Re: [PATCH 2/7] mm, vmscan: add active list aging tracepoint Minchan Kim <minchan@kernel.org> - 2016-12-30 17:10 +0100
                Re: [PATCH 2/7] mm, vmscan: add active list aging tracepoint Michal Hocko <mhocko@kernel.org> - 2016-12-30 17:40 +0100
                  Re: [PATCH 2/7] mm, vmscan: add active list aging tracepoint Michal Hocko <mhocko@kernel.org> - 2016-12-30 18:40 +0100
                  Re: [PATCH 2/7] mm, vmscan: add active list aging tracepoint Minchan Kim <minchan@kernel.org> - 2017-01-03 06:10 +0100
                    Re: [PATCH 2/7] mm, vmscan: add active list aging tracepoint Michal Hocko <mhocko@kernel.org> - 2017-01-03 09:30 +0100
      Re: [PATCH 2/7] mm, vmscan: add active list aging tracepoint "Hillf Danton" <hillf.zj@alibaba-inc.com> - 2016-12-29 09:00 +0100
    Re: [PATCH 0/7] vm, vmscan: enahance vmscan tracepoints Mel Gorman <mgorman@suse.de> - 2016-12-30 10:20 +0100
      Re: [PATCH 0/7] vm, vmscan: enahance vmscan tracepoints Michal Hocko <mhocko@kernel.org> - 2016-12-30 10:40 +0100
        Re: [PATCH 0/7] vm, vmscan: enahance vmscan tracepoints Mel Gorman <mgorman@suse.de> - 2016-12-30 11:30 +0100

Page 2 of 2 — ← Prev page 1 [2]

#1548004 — [PATCH 3/7] mm, vmscan: show the number of skipped pages in mm_vmscan_lru_isolate

From: Michal Hocko <mhocko@suse.com>

mm_vmscan_lru_isolate shows the number of requested, scanned and taken
pages. This is mostly OK but on 32b systems the number of scanned pages
is quite misleading because it includes both the scanned and skipped
pages.  Moreover the skipped part is scaled based on the number of taken
pages. Let's report the exact numbers without any additional logic and
add the number of skipped pages. This should make the reported data much
more easier to interpret.

Signed-off-by: Michal Hocko <mhocko@suse.com>
---
 include/trace/events/vmscan.h |  8 ++++++--
 mm/vmscan.c                   | 10 +++++-----
 2 files changed, 11 insertions(+), 7 deletions(-)

diff --git a/include/trace/events/vmscan.h b/include/trace/events/vmscan.h
index d34cc0ced2be..6af4dae46db2 100644
--- a/include/trace/events/vmscan.h
+++ b/include/trace/events/vmscan.h
@@ -274,17 +274,19 @@ TRACE_EVENT(mm_vmscan_lru_isolate,
 		int order,
 		unsigned long nr_requested,
 		unsigned long nr_scanned,
+		unsigned long nr_skipped,
 		unsigned long nr_taken,
 		isolate_mode_t isolate_mode,
 		int file),
 
-	TP_ARGS(classzone_idx, order, nr_requested, nr_scanned, nr_taken, isolate_mode, file),
+	TP_ARGS(classzone_idx, order, nr_requested, nr_scanned, nr_skipped, nr_taken, isolate_mode, file),
 
 	TP_STRUCT__entry(
 		__field(int, classzone_idx)
 		__field(int, order)
 		__field(unsigned long, nr_requested)
 		__field(unsigned long, nr_scanned)
+		__field(unsigned long, nr_skipped)
 		__field(unsigned long, nr_taken)
 		__field(isolate_mode_t, isolate_mode)
 		__field(int, file)
@@ -295,17 +297,19 @@ TRACE_EVENT(mm_vmscan_lru_isolate,
 		__entry->order = order;
 		__entry->nr_requested = nr_requested;
 		__entry->nr_scanned = nr_scanned;
+		__entry->nr_skipped = nr_skipped;
 		__entry->nr_taken = nr_taken;
 		__entry->isolate_mode = isolate_mode;
 		__entry->file = file;
 	),
 
-	TP_printk("isolate_mode=%d classzone=%d order=%d nr_requested=%lu nr_scanned=%lu nr_taken=%lu file=%d",
+	TP_printk("isolate_mode=%d classzone=%d order=%d nr_requested=%lu nr_scanned=%lu nr_skipped=%lu nr_taken=%lu file=%d",
 		__entry->isolate_mode,
 		__entry->classzone_idx,
 		__entry->order,
 		__entry->nr_requested,
 		__entry->nr_scanned,
+		__entry->nr_skipped,
 		__entry->nr_taken,
 		__entry->file)
 );
diff --git a/mm/vmscan.c b/mm/vmscan.c
index 2302a1a58c6e..4f7c0d66d629 100644
--- a/mm/vmscan.c
+++ b/mm/vmscan.c
@@ -1428,6 +1428,7 @@ static unsigned long isolate_lru_pages(unsigned long nr_to_scan,
 	unsigned long nr_taken = 0;
 	unsigned long nr_zone_taken[MAX_NR_ZONES] = { 0 };
 	unsigned long nr_skipped[MAX_NR_ZONES] = { 0, };
+	unsigned long skipped = 0, total_skipped = 0;
 	unsigned long scan, nr_pages;
 	LIST_HEAD(pages_skipped);
 
@@ -1479,14 +1480,13 @@ static unsigned long isolate_lru_pages(unsigned long nr_to_scan,
 	 */
 	if (!list_empty(&pages_skipped)) {
 		int zid;
-		unsigned long total_skipped = 0;
 
 		for (zid = 0; zid < MAX_NR_ZONES; zid++) {
 			if (!nr_skipped[zid])
 				continue;
 
 			__count_zid_vm_events(PGSCAN_SKIP, zid, nr_skipped[zid]);
-			total_skipped += nr_skipped[zid];
+			skipped += nr_skipped[zid];
 		}
 
 		/*
@@ -1494,13 +1494,13 @@ static unsigned long isolate_lru_pages(unsigned long nr_to_scan,
 		 * close to unreclaimable. If the LRU list is empty, account
 		 * skipped pages as a full scan.
 		 */
-		scan += list_empty(src) ? total_skipped : total_skipped >> 2;
+		total_skipped = list_empty(src) ? skipped : skipped >> 2;
 
 		list_splice(&pages_skipped, src);
 	}
-	*nr_scanned = scan;
+	*nr_scanned = scan + total_skipped;
 	trace_mm_vmscan_lru_isolate(sc->reclaim_idx, sc->order, nr_to_scan, scan,
-				    nr_taken, mode, is_file_lru(lru));
+				    skipped, nr_taken, mode, is_file_lru(lru));
 	update_lru_sizes(lruvec, lru, nr_zone_taken, nr_taken);
 	return nr_taken;
 }
-- 
2.10.2

[toc] | [prev] | [next] | [standalone]

#1548192 — Re: [PATCH 3/7] mm, vmscan: show the number of skipped pages in mm_vmscan_lru_isolate

On Wed, Dec 28, 2016 at 04:30:28PM +0100, Michal Hocko wrote:
> From: Michal Hocko <mhocko@suse.com>
> 
> mm_vmscan_lru_isolate shows the number of requested, scanned and taken
> pages. This is mostly OK but on 32b systems the number of scanned pages
> is quite misleading because it includes both the scanned and skipped
> pages.  Moreover the skipped part is scaled based on the number of taken
> pages. Let's report the exact numbers without any additional logic and
> add the number of skipped pages. This should make the reported data much
> more easier to interpret.
> 
> Signed-off-by: Michal Hocko <mhocko@suse.com>
Acked-by: Minchan Kim <minchan@kernel.org>

[toc] | [prev] | [next] | [standalone]

#1548222 — Re: [PATCH 3/7] mm, vmscan: show the number of skipped pages in mm_vmscan_lru_isolate

On Wednesday, December 28, 2016 11:30 PM Michal Hocko wrote:
> From: Michal Hocko <mhocko@suse.com>
> 
> mm_vmscan_lru_isolate shows the number of requested, scanned and taken
> pages. This is mostly OK but on 32b systems the number of scanned pages
> is quite misleading because it includes both the scanned and skipped
> pages.  Moreover the skipped part is scaled based on the number of taken
> pages. Let's report the exact numbers without any additional logic and
> add the number of skipped pages. This should make the reported data much
> more easier to interpret.
> 
> Signed-off-by: Michal Hocko <mhocko@suse.com>
> ---
Acked-by: Hillf Danton <hillf.zj@alibaba-inc.com> 

[toc] | [prev] | [next] | [standalone]

#1549990 — Re: [PATCH 3/7] mm, vmscan: show the number of skipped pages in mm_vmscan_lru_isolate

On 12/28/2016 04:30 PM, Michal Hocko wrote:
> --- a/mm/vmscan.c
> +++ b/mm/vmscan.c
> @@ -1428,6 +1428,7 @@ static unsigned long isolate_lru_pages(unsigned long nr_to_scan,
>  	unsigned long nr_taken = 0;
>  	unsigned long nr_zone_taken[MAX_NR_ZONES] = { 0 };
>  	unsigned long nr_skipped[MAX_NR_ZONES] = { 0, };
> +	unsigned long skipped = 0, total_skipped = 0;
>  	unsigned long scan, nr_pages;
>  	LIST_HEAD(pages_skipped);
>
> @@ -1479,14 +1480,13 @@ static unsigned long isolate_lru_pages(unsigned long nr_to_scan,
>  	 */
>  	if (!list_empty(&pages_skipped)) {
>  		int zid;
> -		unsigned long total_skipped = 0;
>
>  		for (zid = 0; zid < MAX_NR_ZONES; zid++) {
>  			if (!nr_skipped[zid])
>  				continue;
>
>  			__count_zid_vm_events(PGSCAN_SKIP, zid, nr_skipped[zid]);
> -			total_skipped += nr_skipped[zid];
> +			skipped += nr_skipped[zid];
>  		}
>
>  		/*
> @@ -1494,13 +1494,13 @@ static unsigned long isolate_lru_pages(unsigned long nr_to_scan,
>  		 * close to unreclaimable. If the LRU list is empty, account
>  		 * skipped pages as a full scan.
>  		 */
> -		scan += list_empty(src) ? total_skipped : total_skipped >> 2;
> +		total_skipped = list_empty(src) ? skipped : skipped >> 2;

Should the tracepoint output reflect this halving heuristic or rather report the 
raw data? Or is each variant inferrable from the other?

>
>  		list_splice(&pages_skipped, src);
>  	}
> -	*nr_scanned = scan;
> +	*nr_scanned = scan + total_skipped;
>  	trace_mm_vmscan_lru_isolate(sc->reclaim_idx, sc->order, nr_to_scan, scan,
> -				    nr_taken, mode, is_file_lru(lru));
> +				    skipped, nr_taken, mode, is_file_lru(lru));
>  	update_lru_sizes(lruvec, lru, nr_zone_taken, nr_taken);
>  	return nr_taken;
>  }
>

[toc] | [prev] | [next] | [standalone]

#1548005 — [PATCH 2/7] mm, vmscan: add active list aging tracepoint

From: Michal Hocko <mhocko@suse.com>

Our reclaim process has several tracepoints to tell us more about how
things are progressing. We are, however, missing a tracepoint to track
active list aging. Introduce mm_vmscan_lru_shrink_active which reports
the number of scanned, rotated, deactivated and freed pages from the
particular node's active list.

Signed-off-by: Michal Hocko <mhocko@suse.com>
---
 include/linux/gfp.h           |  2 +-
 include/trace/events/vmscan.h | 38 ++++++++++++++++++++++++++++++++++++++
 mm/page_alloc.c               |  6 +++++-
 mm/vmscan.c                   | 22 +++++++++++++++++-----
 4 files changed, 61 insertions(+), 7 deletions(-)

diff --git a/include/linux/gfp.h b/include/linux/gfp.h
index 4175dca4ac39..61aa9b49e86d 100644
--- a/include/linux/gfp.h
+++ b/include/linux/gfp.h
@@ -503,7 +503,7 @@ void * __meminit alloc_pages_exact_nid(int nid, size_t size, gfp_t gfp_mask);
 extern void __free_pages(struct page *page, unsigned int order);
 extern void free_pages(unsigned long addr, unsigned int order);
 extern void free_hot_cold_page(struct page *page, bool cold);
-extern void free_hot_cold_page_list(struct list_head *list, bool cold);
+extern int free_hot_cold_page_list(struct list_head *list, bool cold);
 
 struct page_frag_cache;
 extern void __page_frag_drain(struct page *page, unsigned int order,
diff --git a/include/trace/events/vmscan.h b/include/trace/events/vmscan.h
index 39bad8921ca1..d34cc0ced2be 100644
--- a/include/trace/events/vmscan.h
+++ b/include/trace/events/vmscan.h
@@ -363,6 +363,44 @@ TRACE_EVENT(mm_vmscan_lru_shrink_inactive,
 		show_reclaim_flags(__entry->reclaim_flags))
 );
 
+TRACE_EVENT(mm_vmscan_lru_shrink_active,
+
+	TP_PROTO(int nid, unsigned long nr_scanned, unsigned long nr_freed,
+		unsigned long nr_unevictable, unsigned long nr_deactivated,
+		unsigned long nr_rotated, int priority, int file),
+
+	TP_ARGS(nid, nr_scanned, nr_freed, nr_unevictable, nr_deactivated, nr_rotated, priority, file),
+
+	TP_STRUCT__entry(
+		__field(int, nid)
+		__field(unsigned long, nr_scanned)
+		__field(unsigned long, nr_freed)
+		__field(unsigned long, nr_unevictable)
+		__field(unsigned long, nr_deactivated)
+		__field(unsigned long, nr_rotated)
+		__field(int, priority)
+		__field(int, reclaim_flags)
+	),
+
+	TP_fast_assign(
+		__entry->nid = nid;
+		__entry->nr_scanned = nr_scanned;
+		__entry->nr_freed = nr_freed;
+		__entry->nr_unevictable = nr_unevictable;
+		__entry->nr_deactivated = nr_deactivated;
+		__entry->nr_rotated = nr_rotated;
+		__entry->priority = priority;
+		__entry->reclaim_flags = trace_shrink_flags(file);
+	),
+
+	TP_printk("nid=%d nr_scanned=%ld nr_freed=%ld nr_unevictable=%ld nr_deactivated=%ld nr_rotated=%ld priority=%d flags=%s",
+		__entry->nid,
+		__entry->nr_scanned, __entry->nr_freed, __entry->nr_unevictable,
+		__entry->nr_deactivated, __entry->nr_rotated,
+		__entry->priority,
+		show_reclaim_flags(__entry->reclaim_flags))
+);
+
 #endif /* _TRACE_VMSCAN_H */
 
 /* This part must be outside protection */
diff --git a/mm/page_alloc.c b/mm/page_alloc.c
index 1c24112308d6..77d204660857 100644
--- a/mm/page_alloc.c
+++ b/mm/page_alloc.c
@@ -2487,14 +2487,18 @@ void free_hot_cold_page(struct page *page, bool cold)
 /*
  * Free a list of 0-order pages
  */
-void free_hot_cold_page_list(struct list_head *list, bool cold)
+int free_hot_cold_page_list(struct list_head *list, bool cold)
 {
 	struct page *page, *next;
+	int ret = 0;
 
 	list_for_each_entry_safe(page, next, list, lru) {
 		trace_mm_page_free_batched(page, cold);
 		free_hot_cold_page(page, cold);
+		ret++;
 	}
+
+	return ret;
 }
 
 /*
diff --git a/mm/vmscan.c b/mm/vmscan.c
index c4abf08861d2..2302a1a58c6e 100644
--- a/mm/vmscan.c
+++ b/mm/vmscan.c
@@ -1846,9 +1846,11 @@ shrink_inactive_list(unsigned long nr_to_scan, struct lruvec *lruvec,
  *
  * The downside is that we have to touch page->_refcount against each page.
  * But we had to alter page->flags anyway.
+ *
+ * Returns the number of pages moved to the given lru.
  */
 
-static void move_active_pages_to_lru(struct lruvec *lruvec,
+static int move_active_pages_to_lru(struct lruvec *lruvec,
 				     struct list_head *list,
 				     struct list_head *pages_to_free,
 				     enum lru_list lru)
@@ -1857,6 +1859,7 @@ static void move_active_pages_to_lru(struct lruvec *lruvec,
 	unsigned long pgmoved = 0;
 	struct page *page;
 	int nr_pages;
+	int nr_moved = 0;
 
 	while (!list_empty(list)) {
 		page = lru_to_page(list);
@@ -1882,11 +1885,15 @@ static void move_active_pages_to_lru(struct lruvec *lruvec,
 				spin_lock_irq(&pgdat->lru_lock);
 			} else
 				list_add(&page->lru, pages_to_free);
+		} else {
+			nr_moved++;
 		}
 	}
 
 	if (!is_active_lru(lru))
 		__count_vm_events(PGDEACTIVATE, pgmoved);
+
+	return nr_moved;
 }
 
 static void shrink_active_list(unsigned long nr_to_scan,
@@ -1902,7 +1909,8 @@ static void shrink_active_list(unsigned long nr_to_scan,
 	LIST_HEAD(l_inactive);
 	struct page *page;
 	struct zone_reclaim_stat *reclaim_stat = &lruvec->reclaim_stat;
-	unsigned long nr_rotated = 0;
+	unsigned long nr_rotated = 0, nr_unevictable = 0;
+	unsigned long nr_freed, nr_deactivate, nr_activate;
 	isolate_mode_t isolate_mode = 0;
 	int file = is_file_lru(lru);
 	struct pglist_data *pgdat = lruvec_pgdat(lruvec);
@@ -1935,6 +1943,7 @@ static void shrink_active_list(unsigned long nr_to_scan,
 
 		if (unlikely(!page_evictable(page))) {
 			putback_lru_page(page);
+			nr_unevictable++;
 			continue;
 		}
 
@@ -1980,13 +1989,16 @@ static void shrink_active_list(unsigned long nr_to_scan,
 	 */
 	reclaim_stat->recent_rotated[file] += nr_rotated;
 
-	move_active_pages_to_lru(lruvec, &l_active, &l_hold, lru);
-	move_active_pages_to_lru(lruvec, &l_inactive, &l_hold, lru - LRU_ACTIVE);
+	nr_activate = move_active_pages_to_lru(lruvec, &l_active, &l_hold, lru);
+	nr_deactivate = move_active_pages_to_lru(lruvec, &l_inactive, &l_hold, lru - LRU_ACTIVE);
 	__mod_node_page_state(pgdat, NR_ISOLATED_ANON + file, -nr_taken);
 	spin_unlock_irq(&pgdat->lru_lock);
 
 	mem_cgroup_uncharge_list(&l_hold);
-	free_hot_cold_page_list(&l_hold, true);
+	nr_freed = free_hot_cold_page_list(&l_hold, true);
+	trace_mm_vmscan_lru_shrink_active(pgdat->node_id, nr_scanned, nr_freed,
+			nr_unevictable, nr_deactivate, nr_rotated,
+			sc->priority, file);
 }
 
 /*
-- 
2.10.2

[toc] | [prev] | [next] | [standalone]

#1548189 — Re: [PATCH 2/7] mm, vmscan: add active list aging tracepoint

On Wed, Dec 28, 2016 at 04:30:27PM +0100, Michal Hocko wrote:
> From: Michal Hocko <mhocko@suse.com>
> 
> Our reclaim process has several tracepoints to tell us more about how
> things are progressing. We are, however, missing a tracepoint to track
> active list aging. Introduce mm_vmscan_lru_shrink_active which reports
> the number of scanned, rotated, deactivated and freed pages from the
> particular node's active list.
> 
> Signed-off-by: Michal Hocko <mhocko@suse.com>
> ---
>  include/linux/gfp.h           |  2 +-
>  include/trace/events/vmscan.h | 38 ++++++++++++++++++++++++++++++++++++++
>  mm/page_alloc.c               |  6 +++++-
>  mm/vmscan.c                   | 22 +++++++++++++++++-----
>  4 files changed, 61 insertions(+), 7 deletions(-)
> 
> diff --git a/include/linux/gfp.h b/include/linux/gfp.h
> index 4175dca4ac39..61aa9b49e86d 100644
> --- a/include/linux/gfp.h
> +++ b/include/linux/gfp.h
> @@ -503,7 +503,7 @@ void * __meminit alloc_pages_exact_nid(int nid, size_t size, gfp_t gfp_mask);
>  extern void __free_pages(struct page *page, unsigned int order);
>  extern void free_pages(unsigned long addr, unsigned int order);
>  extern void free_hot_cold_page(struct page *page, bool cold);
> -extern void free_hot_cold_page_list(struct list_head *list, bool cold);
> +extern int free_hot_cold_page_list(struct list_head *list, bool cold);
>  
>  struct page_frag_cache;
>  extern void __page_frag_drain(struct page *page, unsigned int order,
> diff --git a/include/trace/events/vmscan.h b/include/trace/events/vmscan.h
> index 39bad8921ca1..d34cc0ced2be 100644
> --- a/include/trace/events/vmscan.h
> +++ b/include/trace/events/vmscan.h
> @@ -363,6 +363,44 @@ TRACE_EVENT(mm_vmscan_lru_shrink_inactive,
>  		show_reclaim_flags(__entry->reclaim_flags))
>  );
>  
> +TRACE_EVENT(mm_vmscan_lru_shrink_active,
> +
> +	TP_PROTO(int nid, unsigned long nr_scanned, unsigned long nr_freed,
> +		unsigned long nr_unevictable, unsigned long nr_deactivated,
> +		unsigned long nr_rotated, int priority, int file),
> +
> +	TP_ARGS(nid, nr_scanned, nr_freed, nr_unevictable, nr_deactivated, nr_rotated, priority, file),

I agree it is helpful. And it was when I investigated aging problem of 32bit
when node-lru was introduced. However, the question is we really need all those
kinds of information? just enough with nr_taken, nr_deactivated, priority, file?

Also, look at minor thing below.

Thanks.

> +
> +	TP_STRUCT__entry(
> +		__field(int, nid)
> +		__field(unsigned long, nr_scanned)
> +		__field(unsigned long, nr_freed)
> +		__field(unsigned long, nr_unevictable)
> +		__field(unsigned long, nr_deactivated)
> +		__field(unsigned long, nr_rotated)
> +		__field(int, priority)
> +		__field(int, reclaim_flags)
> +	),
> +
> +	TP_fast_assign(
> +		__entry->nid = nid;
> +		__entry->nr_scanned = nr_scanned;
> +		__entry->nr_freed = nr_freed;
> +		__entry->nr_unevictable = nr_unevictable;
> +		__entry->nr_deactivated = nr_deactivated;
> +		__entry->nr_rotated = nr_rotated;
> +		__entry->priority = priority;
> +		__entry->reclaim_flags = trace_shrink_flags(file);
> +	),
> +
> +	TP_printk("nid=%d nr_scanned=%ld nr_freed=%ld nr_unevictable=%ld nr_deactivated=%ld nr_rotated=%ld priority=%d flags=%s",
> +		__entry->nid,
> +		__entry->nr_scanned, __entry->nr_freed, __entry->nr_unevictable,
> +		__entry->nr_deactivated, __entry->nr_rotated,
> +		__entry->priority,
> +		show_reclaim_flags(__entry->reclaim_flags))
> +);
> +
>  #endif /* _TRACE_VMSCAN_H */
>  
>  /* This part must be outside protection */
> diff --git a/mm/page_alloc.c b/mm/page_alloc.c
> index 1c24112308d6..77d204660857 100644
> --- a/mm/page_alloc.c
> +++ b/mm/page_alloc.c
> @@ -2487,14 +2487,18 @@ void free_hot_cold_page(struct page *page, bool cold)
>  /*
>   * Free a list of 0-order pages
>   */
> -void free_hot_cold_page_list(struct list_head *list, bool cold)
> +int free_hot_cold_page_list(struct list_head *list, bool cold)
>  {
>  	struct page *page, *next;
> +	int ret = 0;
>  
>  	list_for_each_entry_safe(page, next, list, lru) {
>  		trace_mm_page_free_batched(page, cold);
>  		free_hot_cold_page(page, cold);
> +		ret++;
>  	}
> +
> +	return ret;
>  }
>  
>  /*
> diff --git a/mm/vmscan.c b/mm/vmscan.c
> index c4abf08861d2..2302a1a58c6e 100644
> --- a/mm/vmscan.c
> +++ b/mm/vmscan.c
> @@ -1846,9 +1846,11 @@ shrink_inactive_list(unsigned long nr_to_scan, struct lruvec *lruvec,
>   *
>   * The downside is that we have to touch page->_refcount against each page.
>   * But we had to alter page->flags anyway.
> + *
> + * Returns the number of pages moved to the given lru.
>   */
>  
> -static void move_active_pages_to_lru(struct lruvec *lruvec,
> +static int move_active_pages_to_lru(struct lruvec *lruvec,
>  				     struct list_head *list,
>  				     struct list_head *pages_to_free,
>  				     enum lru_list lru)
> @@ -1857,6 +1859,7 @@ static void move_active_pages_to_lru(struct lruvec *lruvec,
>  	unsigned long pgmoved = 0;
>  	struct page *page;
>  	int nr_pages;
> +	int nr_moved = 0;
>  
>  	while (!list_empty(list)) {
>  		page = lru_to_page(list);
> @@ -1882,11 +1885,15 @@ static void move_active_pages_to_lru(struct lruvec *lruvec,
>  				spin_lock_irq(&pgdat->lru_lock);
>  			} else
>  				list_add(&page->lru, pages_to_free);
> +		} else {
> +			nr_moved++;
>  		}
>  	}
>  
>  	if (!is_active_lru(lru))
>  		__count_vm_events(PGDEACTIVATE, pgmoved);
> +
> +	return nr_moved;
>  }
>  
>  static void shrink_active_list(unsigned long nr_to_scan,
> @@ -1902,7 +1909,8 @@ static void shrink_active_list(unsigned long nr_to_scan,
>  	LIST_HEAD(l_inactive);
>  	struct page *page;
>  	struct zone_reclaim_stat *reclaim_stat = &lruvec->reclaim_stat;
> -	unsigned long nr_rotated = 0;
> +	unsigned long nr_rotated = 0, nr_unevictable = 0;
> +	unsigned long nr_freed, nr_deactivate, nr_activate;
>  	isolate_mode_t isolate_mode = 0;
>  	int file = is_file_lru(lru);
>  	struct pglist_data *pgdat = lruvec_pgdat(lruvec);
> @@ -1935,6 +1943,7 @@ static void shrink_active_list(unsigned long nr_to_scan,
>  
>  		if (unlikely(!page_evictable(page))) {
>  			putback_lru_page(page);
> +			nr_unevictable++;
>  			continue;
>  		}
>  
> @@ -1980,13 +1989,16 @@ static void shrink_active_list(unsigned long nr_to_scan,
>  	 */
>  	reclaim_stat->recent_rotated[file] += nr_rotated;
>  
> -	move_active_pages_to_lru(lruvec, &l_active, &l_hold, lru);
> -	move_active_pages_to_lru(lruvec, &l_inactive, &l_hold, lru - LRU_ACTIVE);
> +	nr_activate = move_active_pages_to_lru(lruvec, &l_active, &l_hold, lru);

Who use nr_active in here?

> +	nr_deactivate = move_active_pages_to_lru(lruvec, &l_inactive, &l_hold, lru - LRU_ACTIVE);
>  	__mod_node_page_state(pgdat, NR_ISOLATED_ANON + file, -nr_taken);
>  	spin_unlock_irq(&pgdat->lru_lock);
>  
>  	mem_cgroup_uncharge_list(&l_hold);
> -	free_hot_cold_page_list(&l_hold, true);
> +	nr_freed = free_hot_cold_page_list(&l_hold, true);
> +	trace_mm_vmscan_lru_shrink_active(pgdat->node_id, nr_scanned, nr_freed,
> +			nr_unevictable, nr_deactivate, nr_rotated,
> +			sc->priority, file);
>  }
>  
>  /*
> -- 
> 2.10.2
> 
> --
> To unsubscribe, send a message with 'unsubscribe linux-mm' in
> the body to majordomo@kvack.org.  For more info on Linux MM,
> see: http://www.linux-mm.org/ .
> Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

[toc] | [prev] | [next] | [standalone]

#1548226 — Re: [PATCH 2/7] mm, vmscan: add active list aging tracepoint

On Thu 29-12-16 14:33:59, Minchan Kim wrote:
> On Wed, Dec 28, 2016 at 04:30:27PM +0100, Michal Hocko wrote:
> > From: Michal Hocko <mhocko@suse.com>
> > 
> > Our reclaim process has several tracepoints to tell us more about how
> > things are progressing. We are, however, missing a tracepoint to track
> > active list aging. Introduce mm_vmscan_lru_shrink_active which reports
> > the number of scanned, rotated, deactivated and freed pages from the
> > particular node's active list.
> > 
> > Signed-off-by: Michal Hocko <mhocko@suse.com>
> > ---
> >  include/linux/gfp.h           |  2 +-
> >  include/trace/events/vmscan.h | 38 ++++++++++++++++++++++++++++++++++++++
> >  mm/page_alloc.c               |  6 +++++-
> >  mm/vmscan.c                   | 22 +++++++++++++++++-----
> >  4 files changed, 61 insertions(+), 7 deletions(-)
> > 
> > diff --git a/include/linux/gfp.h b/include/linux/gfp.h
> > index 4175dca4ac39..61aa9b49e86d 100644
> > --- a/include/linux/gfp.h
> > +++ b/include/linux/gfp.h
> > @@ -503,7 +503,7 @@ void * __meminit alloc_pages_exact_nid(int nid, size_t size, gfp_t gfp_mask);
> >  extern void __free_pages(struct page *page, unsigned int order);
> >  extern void free_pages(unsigned long addr, unsigned int order);
> >  extern void free_hot_cold_page(struct page *page, bool cold);
> > -extern void free_hot_cold_page_list(struct list_head *list, bool cold);
> > +extern int free_hot_cold_page_list(struct list_head *list, bool cold);
> >  
> >  struct page_frag_cache;
> >  extern void __page_frag_drain(struct page *page, unsigned int order,
> > diff --git a/include/trace/events/vmscan.h b/include/trace/events/vmscan.h
> > index 39bad8921ca1..d34cc0ced2be 100644
> > --- a/include/trace/events/vmscan.h
> > +++ b/include/trace/events/vmscan.h
> > @@ -363,6 +363,44 @@ TRACE_EVENT(mm_vmscan_lru_shrink_inactive,
> >  		show_reclaim_flags(__entry->reclaim_flags))
> >  );
> >  
> > +TRACE_EVENT(mm_vmscan_lru_shrink_active,
> > +
> > +	TP_PROTO(int nid, unsigned long nr_scanned, unsigned long nr_freed,
> > +		unsigned long nr_unevictable, unsigned long nr_deactivated,
> > +		unsigned long nr_rotated, int priority, int file),
> > +
> > +	TP_ARGS(nid, nr_scanned, nr_freed, nr_unevictable, nr_deactivated, nr_rotated, priority, file),
> 
> I agree it is helpful. And it was when I investigated aging problem of 32bit
> when node-lru was introduced. However, the question is we really need all those
> kinds of information? just enough with nr_taken, nr_deactivated, priority, file?

Dunno. Is it harmful to add this information? I like it more when the
numbers just add up and you have a clear picture. You never know what
might be useful when debugging a weird behavior. 

[...]
> > -	move_active_pages_to_lru(lruvec, &l_active, &l_hold, lru);
> > -	move_active_pages_to_lru(lruvec, &l_inactive, &l_hold, lru - LRU_ACTIVE);
> > +	nr_activate = move_active_pages_to_lru(lruvec, &l_active, &l_hold, lru);
> 
> Who use nr_active in here?

this is an omission. I just forgot to add it... Thanks for noticing.

-- 
Michal Hocko
SUSE Labs

[toc] | [prev] | [next] | [standalone]

#1548426 — Re: [PATCH 2/7] mm, vmscan: add active list aging tracepoint

On Thu, Dec 29, 2016 at 08:52:46AM +0100, Michal Hocko wrote:
> On Thu 29-12-16 14:33:59, Minchan Kim wrote:
> > On Wed, Dec 28, 2016 at 04:30:27PM +0100, Michal Hocko wrote:
> > > From: Michal Hocko <mhocko@suse.com>
> > > 
> > > Our reclaim process has several tracepoints to tell us more about how
> > > things are progressing. We are, however, missing a tracepoint to track
> > > active list aging. Introduce mm_vmscan_lru_shrink_active which reports
> > > the number of scanned, rotated, deactivated and freed pages from the
> > > particular node's active list.
> > > 
> > > Signed-off-by: Michal Hocko <mhocko@suse.com>
> > > ---
> > >  include/linux/gfp.h           |  2 +-
> > >  include/trace/events/vmscan.h | 38 ++++++++++++++++++++++++++++++++++++++
> > >  mm/page_alloc.c               |  6 +++++-
> > >  mm/vmscan.c                   | 22 +++++++++++++++++-----
> > >  4 files changed, 61 insertions(+), 7 deletions(-)
> > > 
> > > diff --git a/include/linux/gfp.h b/include/linux/gfp.h
> > > index 4175dca4ac39..61aa9b49e86d 100644
> > > --- a/include/linux/gfp.h
> > > +++ b/include/linux/gfp.h
> > > @@ -503,7 +503,7 @@ void * __meminit alloc_pages_exact_nid(int nid, size_t size, gfp_t gfp_mask);
> > >  extern void __free_pages(struct page *page, unsigned int order);
> > >  extern void free_pages(unsigned long addr, unsigned int order);
> > >  extern void free_hot_cold_page(struct page *page, bool cold);
> > > -extern void free_hot_cold_page_list(struct list_head *list, bool cold);
> > > +extern int free_hot_cold_page_list(struct list_head *list, bool cold);
> > >  
> > >  struct page_frag_cache;
> > >  extern void __page_frag_drain(struct page *page, unsigned int order,
> > > diff --git a/include/trace/events/vmscan.h b/include/trace/events/vmscan.h
> > > index 39bad8921ca1..d34cc0ced2be 100644
> > > --- a/include/trace/events/vmscan.h
> > > +++ b/include/trace/events/vmscan.h
> > > @@ -363,6 +363,44 @@ TRACE_EVENT(mm_vmscan_lru_shrink_inactive,
> > >  		show_reclaim_flags(__entry->reclaim_flags))
> > >  );
> > >  
> > > +TRACE_EVENT(mm_vmscan_lru_shrink_active,
> > > +
> > > +	TP_PROTO(int nid, unsigned long nr_scanned, unsigned long nr_freed,
> > > +		unsigned long nr_unevictable, unsigned long nr_deactivated,
> > > +		unsigned long nr_rotated, int priority, int file),
> > > +
> > > +	TP_ARGS(nid, nr_scanned, nr_freed, nr_unevictable, nr_deactivated, nr_rotated, priority, file),
> > 
> > I agree it is helpful. And it was when I investigated aging problem of 32bit
> > when node-lru was introduced. However, the question is we really need all those
> > kinds of information? just enough with nr_taken, nr_deactivated, priority, file?
> 
> Dunno. Is it harmful to add this information? I like it more when the
> numbers just add up and you have a clear picture. You never know what
> might be useful when debugging a weird behavior. 

Michal, I'm not huge fan of "might be useful" although it's a small piece of code.
It adds just all of kinds overheads (memory footprint, runtime performance,
maintainance) without any proved benefit.

If we allow such things, people would start adding more things with just "why not,
it might be useful. you never know the future" and it ends up making linux fiction
novel mess.

If it's necessary, someday, someone will catch up and will send or ask patch with
detailed description "why the stat is important and how it is good for us to solve
some problem". From that, we can learn workload, way to solve the problem and git
history has the valuable description so new comers can keep the community up easily.
So, finally, overheads are justified and get merged.

Please add must-have for your goal described.

[toc] | [prev] | [next] | [standalone]

#1548507 — Re: [PATCH 2/7] mm, vmscan: add active list aging tracepoint

On Fri 30-12-16 10:48:53, Minchan Kim wrote:
> On Thu, Dec 29, 2016 at 08:52:46AM +0100, Michal Hocko wrote:
> > On Thu 29-12-16 14:33:59, Minchan Kim wrote:
> > > On Wed, Dec 28, 2016 at 04:30:27PM +0100, Michal Hocko wrote:
[...]
> > > > +TRACE_EVENT(mm_vmscan_lru_shrink_active,
> > > > +
> > > > +	TP_PROTO(int nid, unsigned long nr_scanned, unsigned long nr_freed,
> > > > +		unsigned long nr_unevictable, unsigned long nr_deactivated,
> > > > +		unsigned long nr_rotated, int priority, int file),
> > > > +
> > > > +	TP_ARGS(nid, nr_scanned, nr_freed, nr_unevictable, nr_deactivated, nr_rotated, priority, file),
> > > 
> > > I agree it is helpful. And it was when I investigated aging problem of 32bit
> > > when node-lru was introduced. However, the question is we really need all those
> > > kinds of information? just enough with nr_taken, nr_deactivated, priority, file?
> > 
> > Dunno. Is it harmful to add this information? I like it more when the
> > numbers just add up and you have a clear picture. You never know what
> > might be useful when debugging a weird behavior. 
> 
> Michal, I'm not huge fan of "might be useful" although it's a small piece of code.

But these are tracepoints. One of their primary reasons to exist is
to help debug things.  And it is really hard to predict what might be
useful in advance. It is not like the patch would export numbers which
would be irrelevant to the reclaim.

> It adds just all of kinds overheads (memory footprint, runtime performance,
> maintainance) without any proved benefit.

Does it really add any measurable overhead or the maintenance burden? I
think the only place we could argue about is free_hot_cold_page_list
which is used in hot paths.

I think we can sacrifice it. The same for culled unevictable
pages. We wouldn't know what is the missing part
nr_taken-(nr_activate+nr_deactivate) because it could be either freed or
moved to the unevictable list but that could be handled in a separate
tracepoint in putback_lru_page which sounds like a useful thing I guess.
 
> If we allow such things, people would start adding more things with just "why not,
> it might be useful. you never know the future" and it ends up making linux fiction
> novel mess.

I agree with this concern in general, but is this the case in this
particular case?

> If it's necessary, someday, someone will catch up and will send or ask patch with
> detailed description "why the stat is important and how it is good for us to solve
> some problem".

I can certainly enhance the changelog. See below.

> From that, we can learn workload, way to solve the problem and git
> history has the valuable description so new comers can keep the community up easily.
> So, finally, overheads are justified and get merged.
> 
> Please add must-have for your goal described.

My primary point is that tracepoints which do not give us a good picture
are quite useless and force us to add trace_printk or other means to
give us further information. Then I wonder why to have an incomplete
tracepoint at all.

Anyway, what do you think about this updated patch? I have kept Hillf's
A-b so please let me know if it is no longer valid.
--- 
From 5f1bc22ad1e54050b4da3228d68945e70342ebb6 Mon Sep 17 00:00:00 2001
From: Michal Hocko <mhocko@suse.com>
Date: Tue, 27 Dec 2016 13:18:20 +0100
Subject: [PATCH] mm, vmscan: add active list aging tracepoint

Our reclaim process has several tracepoints to tell us more about how
things are progressing. We are, however, missing a tracepoint to track
active list aging. Introduce mm_vmscan_lru_shrink_active which reports
the number of
	- nr_scanned, nr_taken pages to tell us the LRU isolation
	  effectiveness.
	- nr_rotated pages which tells us that we are hitting referenced
	  pages which are deactivated. If this is a large part of the
	  reported nr_deactivated pages then the active list is too small
	- nr_activated pages which tells us how many pages are keept on the
	  active list - mostly exec pages. A high number can indicate
	  that we might be trashing on executables.

Changes since v1
- report nr_taken pages as per Minchan
- report nr_activated as per Minchan
- do not report nr_freed pages because that would add a tiny overhead to
  free_hot_cold_page_list which is a hot path
- do not report nr_unevictable because we can report this number via a
  different and more generic tracepoint in putback_lru_page
- fix move_active_pages_to_lru to report proper page count when we hit
  into large pages

Acked-by: Hillf Danton <hillf.zj@alibaba-inc.com>
Signed-off-by: Michal Hocko <mhocko@suse.com>
---
 include/trace/events/vmscan.h | 38 ++++++++++++++++++++++++++++++++++++++
 mm/vmscan.c                   | 18 ++++++++++++++----
 2 files changed, 52 insertions(+), 4 deletions(-)

diff --git a/include/trace/events/vmscan.h b/include/trace/events/vmscan.h
index 39bad8921ca1..f9ef242ece1b 100644
--- a/include/trace/events/vmscan.h
+++ b/include/trace/events/vmscan.h
@@ -363,6 +363,44 @@ TRACE_EVENT(mm_vmscan_lru_shrink_inactive,
 		show_reclaim_flags(__entry->reclaim_flags))
 );
 
+TRACE_EVENT(mm_vmscan_lru_shrink_active,
+
+	TP_PROTO(int nid, unsigned long nr_scanned, unsigned long nr_taken,
+		unsigned long nr_activate, unsigned long nr_deactivated,
+		unsigned long nr_rotated, int priority, int file),
+
+	TP_ARGS(nid, nr_scanned, nr_taken, nr_activate, nr_deactivated, nr_rotated, priority, file),
+
+	TP_STRUCT__entry(
+		__field(int, nid)
+		__field(unsigned long, nr_scanned)
+		__field(unsigned long, nr_taken)
+		__field(unsigned long, nr_activate)
+		__field(unsigned long, nr_deactivated)
+		__field(unsigned long, nr_rotated)
+		__field(int, priority)
+		__field(int, reclaim_flags)
+	),
+
+	TP_fast_assign(
+		__entry->nid = nid;
+		__entry->nr_scanned = nr_scanned;
+		__entry->nr_taken = nr_taken;
+		__entry->nr_activate = nr_activate;
+		__entry->nr_deactivated = nr_deactivated;
+		__entry->nr_rotated = nr_rotated;
+		__entry->priority = priority;
+		__entry->reclaim_flags = trace_shrink_flags(file);
+	),
+
+	TP_printk("nid=%d nr_scanned=%ld nr_taken=%ld nr_activated=%ld nr_deactivated=%ld nr_rotated=%ld priority=%d flags=%s",
+		__entry->nid,
+		__entry->nr_scanned, __entry->nr_taken,
+		__entry->nr_activate, __entry->nr_deactivated, __entry->nr_rotated,
+		__entry->priority,
+		show_reclaim_flags(__entry->reclaim_flags))
+);
+
 #endif /* _TRACE_VMSCAN_H */
 
 /* This part must be outside protection */
diff --git a/mm/vmscan.c b/mm/vmscan.c
index c4abf08861d2..4da4d8d0496c 100644
--- a/mm/vmscan.c
+++ b/mm/vmscan.c
@@ -1846,9 +1846,11 @@ shrink_inactive_list(unsigned long nr_to_scan, struct lruvec *lruvec,
  *
  * The downside is that we have to touch page->_refcount against each page.
  * But we had to alter page->flags anyway.
+ *
+ * Returns the number of pages moved to the given lru.
  */
 
-static void move_active_pages_to_lru(struct lruvec *lruvec,
+static unsigned move_active_pages_to_lru(struct lruvec *lruvec,
 				     struct list_head *list,
 				     struct list_head *pages_to_free,
 				     enum lru_list lru)
@@ -1857,6 +1859,7 @@ static void move_active_pages_to_lru(struct lruvec *lruvec,
 	unsigned long pgmoved = 0;
 	struct page *page;
 	int nr_pages;
+	int nr_moved = 0;
 
 	while (!list_empty(list)) {
 		page = lru_to_page(list);
@@ -1882,11 +1885,15 @@ static void move_active_pages_to_lru(struct lruvec *lruvec,
 				spin_lock_irq(&pgdat->lru_lock);
 			} else
 				list_add(&page->lru, pages_to_free);
+		} else {
+			nr_moved += nr_pages;
 		}
 	}
 
 	if (!is_active_lru(lru))
 		__count_vm_events(PGDEACTIVATE, pgmoved);
+
+	return nr_moved;
 }
 
 static void shrink_active_list(unsigned long nr_to_scan,
@@ -1902,7 +1909,8 @@ static void shrink_active_list(unsigned long nr_to_scan,
 	LIST_HEAD(l_inactive);
 	struct page *page;
 	struct zone_reclaim_stat *reclaim_stat = &lruvec->reclaim_stat;
-	unsigned long nr_rotated = 0;
+	unsigned nr_deactivate, nr_activate;
+	unsigned nr_rotated = 0;
 	isolate_mode_t isolate_mode = 0;
 	int file = is_file_lru(lru);
 	struct pglist_data *pgdat = lruvec_pgdat(lruvec);
@@ -1980,13 +1988,15 @@ static void shrink_active_list(unsigned long nr_to_scan,
 	 */
 	reclaim_stat->recent_rotated[file] += nr_rotated;
 
-	move_active_pages_to_lru(lruvec, &l_active, &l_hold, lru);
-	move_active_pages_to_lru(lruvec, &l_inactive, &l_hold, lru - LRU_ACTIVE);
+	nr_activate = move_active_pages_to_lru(lruvec, &l_active, &l_hold, lru);
+	nr_deactivate = move_active_pages_to_lru(lruvec, &l_inactive, &l_hold, lru - LRU_ACTIVE);
 	__mod_node_page_state(pgdat, NR_ISOLATED_ANON + file, -nr_taken);
 	spin_unlock_irq(&pgdat->lru_lock);
 
 	mem_cgroup_uncharge_list(&l_hold);
 	free_hot_cold_page_list(&l_hold, true);
+	trace_mm_vmscan_lru_shrink_active(pgdat->node_id, nr_scanned, nr_taken,
+			nr_activate, nr_deactivate, nr_rotated, sc->priority, file);
 }
 
 /*
-- 
2.10.2


-- 
Michal Hocko
SUSE Labs

[toc] | [prev] | [next] | [standalone]

#1548512 — Re: [PATCH 2/7] mm, vmscan: add active list aging tracepoint

On Friday, December 30, 2016 5:27 PM Michal Hocko wrote: 
> Anyway, what do you think about this updated patch? I have kept Hillf's
> A-b so please let me know if it is no longer valid.
> 
My mind is not changed:)

Happy new year folks!

Hillf

[toc] | [prev] | [next] | [standalone]

#1548651 — Re: [PATCH 2/7] mm, vmscan: add active list aging tracepoint

On Fri, Dec 30, 2016 at 10:26:37AM +0100, Michal Hocko wrote:
> On Fri 30-12-16 10:48:53, Minchan Kim wrote:
> > On Thu, Dec 29, 2016 at 08:52:46AM +0100, Michal Hocko wrote:
> > > On Thu 29-12-16 14:33:59, Minchan Kim wrote:
> > > > On Wed, Dec 28, 2016 at 04:30:27PM +0100, Michal Hocko wrote:
> [...]
> > > > > +TRACE_EVENT(mm_vmscan_lru_shrink_active,
> > > > > +
> > > > > +	TP_PROTO(int nid, unsigned long nr_scanned, unsigned long nr_freed,
> > > > > +		unsigned long nr_unevictable, unsigned long nr_deactivated,
> > > > > +		unsigned long nr_rotated, int priority, int file),
> > > > > +
> > > > > +	TP_ARGS(nid, nr_scanned, nr_freed, nr_unevictable, nr_deactivated, nr_rotated, priority, file),
> > > > 
> > > > I agree it is helpful. And it was when I investigated aging problem of 32bit
> > > > when node-lru was introduced. However, the question is we really need all those
> > > > kinds of information? just enough with nr_taken, nr_deactivated, priority, file?
> > > 
> > > Dunno. Is it harmful to add this information? I like it more when the
> > > numbers just add up and you have a clear picture. You never know what
> > > might be useful when debugging a weird behavior. 
> > 
> > Michal, I'm not huge fan of "might be useful" although it's a small piece of code.
> 
> But these are tracepoints. One of their primary reasons to exist is
> to help debug things.  And it is really hard to predict what might be
> useful in advance. It is not like the patch would export numbers which
> would be irrelevant to the reclaim.

What's different?

Please think over if everyone says like that they want to add something
with the reason "it's tracepoint which helps dubug and we cannot assume
what might be useful in the future."

> 
> > It adds just all of kinds overheads (memory footprint, runtime performance,
> > maintainance) without any proved benefit.
> 
> Does it really add any measurable overhead or the maintenance burden? I

Don't limit your thought in this particular case and expand the idea to
others who want to see random value via tracepoint with just "might-be-
good". We will lose the reason to prevent that trend if we merge any
tracepoint expansion patch with just "might-be-useful" reason.
Finally, that would bite us.

> think the only place we could argue about is free_hot_cold_page_list
> which is used in hot paths.

The point of view about shrinking active list, what we want to know
is just (nr_taken|nr_deactivated|priority|file) and it's enough,
I think. So, if you want to add nr_freed, nr_unevictable, nr_rotated
please, describe "what problem we can solve with those each numbers".

> 
> I think we can sacrifice it. The same for culled unevictable
> pages. We wouldn't know what is the missing part
> nr_taken-(nr_activate+nr_deactivate) because it could be either freed or
> moved to the unevictable list but that could be handled in a separate
> tracepoint in putback_lru_page which sounds like a useful thing I guess.
>  
> > If we allow such things, people would start adding more things with just "why not,
> > it might be useful. you never know the future" and it ends up making linux fiction
> > novel mess.
> 
> I agree with this concern in general, but is this the case in this
> particular case?

I believe it's not different.

> 
> > If it's necessary, someday, someone will catch up and will send or ask patch with
> > detailed description "why the stat is important and how it is good for us to solve
> > some problem".
> 
> I can certainly enhance the changelog. See below.
> 
> > From that, we can learn workload, way to solve the problem and git
> > history has the valuable description so new comers can keep the community up easily.
> > So, finally, overheads are justified and get merged.
> > 
> > Please add must-have for your goal described.
> 
> My primary point is that tracepoints which do not give us a good picture
> are quite useless and force us to add trace_printk or other means to
> give us further information. Then I wonder why to have an incomplete
> tracepoint at all.
> 
> Anyway, what do you think about this updated patch? I have kept Hillf's
> A-b so please let me know if it is no longer valid.
> --- 
> From 5f1bc22ad1e54050b4da3228d68945e70342ebb6 Mon Sep 17 00:00:00 2001
> From: Michal Hocko <mhocko@suse.com>
> Date: Tue, 27 Dec 2016 13:18:20 +0100
> Subject: [PATCH] mm, vmscan: add active list aging tracepoint
> 
> Our reclaim process has several tracepoints to tell us more about how
> things are progressing. We are, however, missing a tracepoint to track
> active list aging. Introduce mm_vmscan_lru_shrink_active which reports

I agree this part.

> the number of
> 	- nr_scanned, nr_taken pages to tell us the LRU isolation
> 	  effectiveness.

I agree nr_taken for knowing shrinking effectiveness but don't
agree nr_scanned. If we want to know LRU isolation effectiveness
with nr_scanned and nr_taken, isolate_lru_pages will do.

> 	- nr_rotated pages which tells us that we are hitting referenced
> 	  pages which are deactivated. If this is a large part of the
> 	  reported nr_deactivated pages then the active list is too small

It might be but not exactly. If your goal is to know LRU size, it can be
done in get_scan_count. I tend to agree LRU size is helpful for
performance analysis because decreased LRU size signals memory shortage
then performance drop.

> 	- nr_activated pages which tells us how many pages are keept on the
                                                               kept

> 	  active list - mostly exec pages. A high number can indicate

                               file-based exec pages

> 	  that we might be trashing on executables.

And welcome to drop nr_unevictable, nr_freed.

I will be off until next week monday so please understand if my response
is slow.

Thanks.

> 
> Changes since v1
> - report nr_taken pages as per Minchan
> - report nr_activated as per Minchan
> - do not report nr_freed pages because that would add a tiny overhead to
>   free_hot_cold_page_list which is a hot path
> - do not report nr_unevictable because we can report this number via a
>   different and more generic tracepoint in putback_lru_page
> - fix move_active_pages_to_lru to report proper page count when we hit
>   into large pages
> 
> Acked-by: Hillf Danton <hillf.zj@alibaba-inc.com>
> Signed-off-by: Michal Hocko <mhocko@suse.com>
> ---
>  include/trace/events/vmscan.h | 38 ++++++++++++++++++++++++++++++++++++++
>  mm/vmscan.c                   | 18 ++++++++++++++----
>  2 files changed, 52 insertions(+), 4 deletions(-)
> 
> diff --git a/include/trace/events/vmscan.h b/include/trace/events/vmscan.h
> index 39bad8921ca1..f9ef242ece1b 100644
> --- a/include/trace/events/vmscan.h
> +++ b/include/trace/events/vmscan.h
> @@ -363,6 +363,44 @@ TRACE_EVENT(mm_vmscan_lru_shrink_inactive,
>  		show_reclaim_flags(__entry->reclaim_flags))
>  );
>  
> +TRACE_EVENT(mm_vmscan_lru_shrink_active,
> +
> +	TP_PROTO(int nid, unsigned long nr_scanned, unsigned long nr_taken,
> +		unsigned long nr_activate, unsigned long nr_deactivated,
> +		unsigned long nr_rotated, int priority, int file),
> +
> +	TP_ARGS(nid, nr_scanned, nr_taken, nr_activate, nr_deactivated, nr_rotated, priority, file),
> +
> +	TP_STRUCT__entry(
> +		__field(int, nid)
> +		__field(unsigned long, nr_scanned)
> +		__field(unsigned long, nr_taken)
> +		__field(unsigned long, nr_activate)
> +		__field(unsigned long, nr_deactivated)
> +		__field(unsigned long, nr_rotated)
> +		__field(int, priority)
> +		__field(int, reclaim_flags)
> +	),
> +
> +	TP_fast_assign(
> +		__entry->nid = nid;
> +		__entry->nr_scanned = nr_scanned;
> +		__entry->nr_taken = nr_taken;
> +		__entry->nr_activate = nr_activate;
> +		__entry->nr_deactivated = nr_deactivated;
> +		__entry->nr_rotated = nr_rotated;
> +		__entry->priority = priority;
> +		__entry->reclaim_flags = trace_shrink_flags(file);
> +	),
> +
> +	TP_printk("nid=%d nr_scanned=%ld nr_taken=%ld nr_activated=%ld nr_deactivated=%ld nr_rotated=%ld priority=%d flags=%s",
> +		__entry->nid,
> +		__entry->nr_scanned, __entry->nr_taken,
> +		__entry->nr_activate, __entry->nr_deactivated, __entry->nr_rotated,
> +		__entry->priority,
> +		show_reclaim_flags(__entry->reclaim_flags))
> +);
> +
>  #endif /* _TRACE_VMSCAN_H */
>  
>  /* This part must be outside protection */
> diff --git a/mm/vmscan.c b/mm/vmscan.c
> index c4abf08861d2..4da4d8d0496c 100644
> --- a/mm/vmscan.c
> +++ b/mm/vmscan.c
> @@ -1846,9 +1846,11 @@ shrink_inactive_list(unsigned long nr_to_scan, struct lruvec *lruvec,
>   *
>   * The downside is that we have to touch page->_refcount against each page.
>   * But we had to alter page->flags anyway.
> + *
> + * Returns the number of pages moved to the given lru.
>   */
>  
> -static void move_active_pages_to_lru(struct lruvec *lruvec,
> +static unsigned move_active_pages_to_lru(struct lruvec *lruvec,
>  				     struct list_head *list,
>  				     struct list_head *pages_to_free,
>  				     enum lru_list lru)
> @@ -1857,6 +1859,7 @@ static void move_active_pages_to_lru(struct lruvec *lruvec,
>  	unsigned long pgmoved = 0;
>  	struct page *page;
>  	int nr_pages;
> +	int nr_moved = 0;
>  
>  	while (!list_empty(list)) {
>  		page = lru_to_page(list);
> @@ -1882,11 +1885,15 @@ static void move_active_pages_to_lru(struct lruvec *lruvec,
>  				spin_lock_irq(&pgdat->lru_lock);
>  			} else
>  				list_add(&page->lru, pages_to_free);
> +		} else {
> +			nr_moved += nr_pages;
>  		}
>  	}
>  
>  	if (!is_active_lru(lru))
>  		__count_vm_events(PGDEACTIVATE, pgmoved);
> +
> +	return nr_moved;
>  }
>  
>  static void shrink_active_list(unsigned long nr_to_scan,
> @@ -1902,7 +1909,8 @@ static void shrink_active_list(unsigned long nr_to_scan,
>  	LIST_HEAD(l_inactive);
>  	struct page *page;
>  	struct zone_reclaim_stat *reclaim_stat = &lruvec->reclaim_stat;
> -	unsigned long nr_rotated = 0;
> +	unsigned nr_deactivate, nr_activate;
> +	unsigned nr_rotated = 0;
>  	isolate_mode_t isolate_mode = 0;
>  	int file = is_file_lru(lru);
>  	struct pglist_data *pgdat = lruvec_pgdat(lruvec);
> @@ -1980,13 +1988,15 @@ static void shrink_active_list(unsigned long nr_to_scan,
>  	 */
>  	reclaim_stat->recent_rotated[file] += nr_rotated;
>  
> -	move_active_pages_to_lru(lruvec, &l_active, &l_hold, lru);
> -	move_active_pages_to_lru(lruvec, &l_inactive, &l_hold, lru - LRU_ACTIVE);
> +	nr_activate = move_active_pages_to_lru(lruvec, &l_active, &l_hold, lru);
> +	nr_deactivate = move_active_pages_to_lru(lruvec, &l_inactive, &l_hold, lru - LRU_ACTIVE);
>  	__mod_node_page_state(pgdat, NR_ISOLATED_ANON + file, -nr_taken);
>  	spin_unlock_irq(&pgdat->lru_lock);
>  
>  	mem_cgroup_uncharge_list(&l_hold);
>  	free_hot_cold_page_list(&l_hold, true);
> +	trace_mm_vmscan_lru_shrink_active(pgdat->node_id, nr_scanned, nr_taken,
> +			nr_activate, nr_deactivate, nr_rotated, sc->priority, file);
>  }
>  
>  /*
> -- 
> 2.10.2
> 
> 
> -- 
> Michal Hocko
> SUSE Labs
> 
> --
> To unsubscribe, send a message with 'unsubscribe linux-mm' in
> the body to majordomo@kvack.org.  For more info on Linux MM,
> see: http://www.linux-mm.org/ .
> Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

[toc] | [prev] | [next] | [standalone]

#1548656 — Re: [PATCH 2/7] mm, vmscan: add active list aging tracepoint

On Sat 31-12-16 01:04:56, Minchan Kim wrote:
[...]
> > From 5f1bc22ad1e54050b4da3228d68945e70342ebb6 Mon Sep 17 00:00:00 2001
> > From: Michal Hocko <mhocko@suse.com>
> > Date: Tue, 27 Dec 2016 13:18:20 +0100
> > Subject: [PATCH] mm, vmscan: add active list aging tracepoint
> > 
> > Our reclaim process has several tracepoints to tell us more about how
> > things are progressing. We are, however, missing a tracepoint to track
> > active list aging. Introduce mm_vmscan_lru_shrink_active which reports
> 
> I agree this part.
> 
> > the number of
> > 	- nr_scanned, nr_taken pages to tell us the LRU isolation
> > 	  effectiveness.
> 
> I agree nr_taken for knowing shrinking effectiveness but don't
> agree nr_scanned. If we want to know LRU isolation effectiveness
> with nr_scanned and nr_taken, isolate_lru_pages will do.

Yes it will. On the other hand the number is there and there is no
additional overhead, maintenance or otherwise, to provide that number.
The inactive counterpart does that for quite some time already. So why
exactly does that matter? Don't take me wrong but isn't this more on a
nit picking side than necessary? Or do I just misunderstand your
concenrs? It is not like we are providing a stable user API as the
tracepoint is clearly implementation specific and not something to be
used for anything other than debugging.

> > 	- nr_rotated pages which tells us that we are hitting referenced
> > 	  pages which are deactivated. If this is a large part of the
> > 	  reported nr_deactivated pages then the active list is too small
> 
> It might be but not exactly. If your goal is to know LRU size, it can be
> done in get_scan_count. I tend to agree LRU size is helpful for
> performance analysis because decreased LRU size signals memory shortage
> then performance drop.

No, I am not really interested in the exact size but rather to allow to
find whether we are aging the active list too early...

> 
> > 	- nr_activated pages which tells us how many pages are keept on the
>                                                                kept

fixed

> 
> > 	  active list - mostly exec pages. A high number can indicate
> 
>                                file-based exec pages

OK, fixed

> 
> > 	  that we might be trashing on executables.
> 
> And welcome to drop nr_unevictable, nr_freed.
> 
> I will be off until next week monday so please understand if my response
> is slow.

There is no reason to hurry...
-- 
Michal Hocko
SUSE Labs

[toc] | [prev] | [next] | [standalone]

#1548664 — Re: [PATCH 2/7] mm, vmscan: add active list aging tracepoint

On Fri 30-12-16 17:37:42, Michal Hocko wrote:
> On Sat 31-12-16 01:04:56, Minchan Kim wrote:
[...]
> > > 	- nr_rotated pages which tells us that we are hitting referenced
> > > 	  pages which are deactivated. If this is a large part of the
> > > 	  reported nr_deactivated pages then the active list is too small
> > 
> > It might be but not exactly. If your goal is to know LRU size, it can be
> > done in get_scan_count. I tend to agree LRU size is helpful for
> > performance analysis because decreased LRU size signals memory shortage
> > then performance drop.
> 
> No, I am not really interested in the exact size but rather to allow to
> find whether we are aging the active list too early...

But thinking about that some more, maybe sticking with the nr_rotated
terminology is rather confusing and displaying the value as nr_referenced
would be more clear.
-- 
Michal Hocko
SUSE Labs

[toc] | [prev] | [next] | [standalone]

#1549533 — Re: [PATCH 2/7] mm, vmscan: add active list aging tracepoint

Hi Michal,

On Fri, Dec 30, 2016 at 05:37:42PM +0100, Michal Hocko wrote:
> On Sat 31-12-16 01:04:56, Minchan Kim wrote:
> [...]
> > > From 5f1bc22ad1e54050b4da3228d68945e70342ebb6 Mon Sep 17 00:00:00 2001
> > > From: Michal Hocko <mhocko@suse.com>
> > > Date: Tue, 27 Dec 2016 13:18:20 +0100
> > > Subject: [PATCH] mm, vmscan: add active list aging tracepoint
> > > 
> > > Our reclaim process has several tracepoints to tell us more about how
> > > things are progressing. We are, however, missing a tracepoint to track
> > > active list aging. Introduce mm_vmscan_lru_shrink_active which reports
> > 
> > I agree this part.
> > 
> > > the number of
> > > 	- nr_scanned, nr_taken pages to tell us the LRU isolation
> > > 	  effectiveness.
> > 
> > I agree nr_taken for knowing shrinking effectiveness but don't
> > agree nr_scanned. If we want to know LRU isolation effectiveness
> > with nr_scanned and nr_taken, isolate_lru_pages will do.
> 
> Yes it will. On the other hand the number is there and there is no
> additional overhead, maintenance or otherwise, to provide that number.

You are adding some instructions, how can you imagine it's no overhead?
Let's say whether it's measurable. Although it's not big in particular case,
it would be measurable if everyone start to say like that "it's trivial so
what's the problem adding a few instructions although it was duplicated?"

You already said "LRU isolate effectiveness". It should be done in there,
isolate_lru_pages and we have been. You need another reasons if you want to
add the duplicated work, strongly.

> The inactive counterpart does that for quite some time already. So why

It couldn't be a reason. If it was duplicated in there, it would be
better to fix it rather than adding more duplciated work to match both
sides.

> exactly does that matter? Don't take me wrong but isn't this more on a
> nit picking side than necessary? Or do I just misunderstand your
> concenrs? It is not like we are providing a stable user API as the

My concern is that I don't see what we can get benefit from those
duplicated work. If it doesn't give benefit to us, I don't want to add.
I hope you think another reasonable reasons.

> tracepoint is clearly implementation specific and not something to be
> used for anything other than debugging.

My point is we already had things "LRU isolation effectivness". Namely,
isolate_lru_pages.

> 
> > > 	- nr_rotated pages which tells us that we are hitting referenced
> > > 	  pages which are deactivated. If this is a large part of the
> > > 	  reported nr_deactivated pages then the active list is too small
> > 
> > It might be but not exactly. If your goal is to know LRU size, it can be
> > done in get_scan_count. I tend to agree LRU size is helpful for
> > performance analysis because decreased LRU size signals memory shortage
> > then performance drop.
> 
> No, I am not really interested in the exact size but rather to allow to
> find whether we are aging the active list too early...

Could you elaborate it more that how we can get active list early aging
with nr_rotated?

Thanks.

[toc] | [prev] | [next] | [standalone]

#1549601 — Re: [PATCH 2/7] mm, vmscan: add active list aging tracepoint

On Tue 03-01-17 14:03:28, Minchan Kim wrote:
> Hi Michal,
> 
> On Fri, Dec 30, 2016 at 05:37:42PM +0100, Michal Hocko wrote:
> > On Sat 31-12-16 01:04:56, Minchan Kim wrote:
> > [...]
> > > > From 5f1bc22ad1e54050b4da3228d68945e70342ebb6 Mon Sep 17 00:00:00 2001
> > > > From: Michal Hocko <mhocko@suse.com>
> > > > Date: Tue, 27 Dec 2016 13:18:20 +0100
> > > > Subject: [PATCH] mm, vmscan: add active list aging tracepoint
> > > > 
> > > > Our reclaim process has several tracepoints to tell us more about how
> > > > things are progressing. We are, however, missing a tracepoint to track
> > > > active list aging. Introduce mm_vmscan_lru_shrink_active which reports
> > > 
> > > I agree this part.
> > > 
> > > > the number of
> > > > 	- nr_scanned, nr_taken pages to tell us the LRU isolation
> > > > 	  effectiveness.
> > > 
> > > I agree nr_taken for knowing shrinking effectiveness but don't
> > > agree nr_scanned. If we want to know LRU isolation effectiveness
> > > with nr_scanned and nr_taken, isolate_lru_pages will do.
> > 
> > Yes it will. On the other hand the number is there and there is no
> > additional overhead, maintenance or otherwise, to provide that number.
> 
> You are adding some instructions, how can you imagine it's no overhead?

There should be close to zero overhead when the tracepoint is disabled
(we pay only one more argument when the function is called). Is this
really worth discussing in this cold path? We are talking about the
reclaim here.

> Let's say whether it's measurable. Although it's not big in particular case,
> it would be measurable if everyone start to say like that "it's trivial so
> what's the problem adding a few instructions although it was duplicated?"
> 
> You already said "LRU isolate effectiveness". It should be done in there,
> isolate_lru_pages and we have been. You need another reasons if you want to
> add the duplicated work, strongly.

isolate_lru_pages is certainly there but you have to enable a trace
point for that. Sometimes it is quite useful to get a reasonably good
picture even without all the vmscan tracepoints enabled because they
can generate quite a lot of output. So if the counter is available I
see no reason to exclude it, especially when it can provide a useful
information. One of the most frustrating debugging experience is when
you are missing some part of the information and have to guess which
part is that and patch, rebuild the kernel and hope to reproduce it
again in the same/similar way.

There are two things about this and other tracepoint patches in general
I believe. 1) Is the tracepoint useful? and 2) Do we have to go over
extra hops to show tracepoint data?

I guess we are in an agreement that the answer for 1 is yes. And
regarding 2, all the data we are showing are there or trivially
retrieved without touching _any_ hot path. Som of it might be duplicated
with other tracepoints but that can be helpful because you do not have
all the tracepoints enabled all the time. So unless you see this
particular thing as a road block I would rather keep it.
 
> > The inactive counterpart does that for quite some time already. So why
> 
> It couldn't be a reason. If it was duplicated in there, it would be
> better to fix it rather than adding more duplciated work to match both
> sides.

I really do not see this as a bad thing.

> > exactly does that matter? Don't take me wrong but isn't this more on a
> > nit picking side than necessary? Or do I just misunderstand your
> > concenrs? It is not like we are providing a stable user API as the
> 
> My concern is that I don't see what we can get benefit from those
> duplicated work. If it doesn't give benefit to us, I don't want to add.
> I hope you think another reasonable reasons.
> 
> > tracepoint is clearly implementation specific and not something to be
> > used for anything other than debugging.
> 
> My point is we already had things "LRU isolation effectivness". Namely,
> isolate_lru_pages.
> 
> > 
> > > > 	- nr_rotated pages which tells us that we are hitting referenced
> > > > 	  pages which are deactivated. If this is a large part of the
> > > > 	  reported nr_deactivated pages then the active list is too small
> > > 
> > > It might be but not exactly. If your goal is to know LRU size, it can be
> > > done in get_scan_count. I tend to agree LRU size is helpful for
> > > performance analysis because decreased LRU size signals memory shortage
> > > then performance drop.
> > 
> > No, I am not really interested in the exact size but rather to allow to
> > find whether we are aging the active list too early...
> 
> Could you elaborate it more that how we can get active list early aging
> with nr_rotated?

If you see too many referenced pages on the active list then they have
been used since promoted and that is an indication that they might be
reclaimed too early. If you are debugging a performance issue and see
this happening then it might be a good indication to look at.

Thanks
-- 
Michal Hocko
SUSE Labs

[toc] | [prev] | [next] | [standalone]

#1548227 — Re: [PATCH 2/7] mm, vmscan: add active list aging tracepoint

On Wednesday, December 28, 2016 11:30 PM Michal Hocko wrote:
> From: Michal Hocko <mhocko@suse.com>
> 
> Our reclaim process has several tracepoints to tell us more about how
> things are progressing. We are, however, missing a tracepoint to track
> active list aging. Introduce mm_vmscan_lru_shrink_active which reports
> the number of scanned, rotated, deactivated and freed pages from the
> particular node's active list.
> 
> Signed-off-by: Michal Hocko <mhocko@suse.com>
> ---
Acked-by: Hillf Danton <hillf.zj@alibaba-inc.com> 

[toc] | [prev] | [next] | [standalone]

#1548506

On Wed, Dec 28, 2016 at 04:30:25PM +0100, Michal Hocko wrote:
> Hi,
> while debugging [1] I've realized that there is some room for
> improvements in the tracepoints set we offer currently. I had hard times
> to make any conclusion from the existing ones. The resulting problem
> turned out to be active list aging [2] and we are missing at least two
> tracepoints to debug such a problem.
> 
> Some existing tracepoints could export more information to see _why_ the
> reclaim progress cannot be made not only _how much_ we could reclaim.
> The later could be seen quite reasonably from the vmstat counters
> already. It can be argued that we are showing too many implementation
> details in those tracepoints but I consider them way too lowlevel
> already to be usable by any kernel independent userspace. I would be
> _really_ surprised if anything but debugging tools have used them.
> 
> Any feedback is highly appreciated.
> 

There is some minor overhead introduced in some paths regardless of
whether the tracepoints are active or not but I suspect it's negligible
in the context of the overhead of reclaim in general so;

Acked-by: Mel Gorman <mgorman@suse.de>

-- 
Mel Gorman
SUSE Labs

[toc] | [prev] | [next] | [standalone]

#1548510

On Fri 30-12-16 09:11:17, Mel Gorman wrote:
> On Wed, Dec 28, 2016 at 04:30:25PM +0100, Michal Hocko wrote:
> > Hi,
> > while debugging [1] I've realized that there is some room for
> > improvements in the tracepoints set we offer currently. I had hard times
> > to make any conclusion from the existing ones. The resulting problem
> > turned out to be active list aging [2] and we are missing at least two
> > tracepoints to debug such a problem.
> > 
> > Some existing tracepoints could export more information to see _why_ the
> > reclaim progress cannot be made not only _how much_ we could reclaim.
> > The later could be seen quite reasonably from the vmstat counters
> > already. It can be argued that we are showing too many implementation
> > details in those tracepoints but I consider them way too lowlevel
> > already to be usable by any kernel independent userspace. I would be
> > _really_ surprised if anything but debugging tools have used them.
> > 
> > Any feedback is highly appreciated.
> > 
> 
> There is some minor overhead introduced in some paths regardless of
> whether the tracepoints are active or not but I suspect it's negligible
> in the context of the overhead of reclaim in general so;

I will work on improving some of them. E.g. I've dropped the change to
free_hot_cold_page_list because that is indeed a hot path but other than
that there shouldn't be any even medium hot path which should see any
overhead I can see. If you are aware of any, please let me know and I
will think about how to improve it.
 
> Acked-by: Mel Gorman <mgorman@suse.de>

Thanks!
-- 
Michal Hocko
SUSE Labs

[toc] | [prev] | [next] | [standalone]

#1548524

On Fri, Dec 30, 2016 at 10:36:55AM +0100, Michal Hocko wrote:
> On Fri 30-12-16 09:11:17, Mel Gorman wrote:
> > On Wed, Dec 28, 2016 at 04:30:25PM +0100, Michal Hocko wrote:
> > > Hi,
> > > while debugging [1] I've realized that there is some room for
> > > improvements in the tracepoints set we offer currently. I had hard times
> > > to make any conclusion from the existing ones. The resulting problem
> > > turned out to be active list aging [2] and we are missing at least two
> > > tracepoints to debug such a problem.
> > > 
> > > Some existing tracepoints could export more information to see _why_ the
> > > reclaim progress cannot be made not only _how much_ we could reclaim.
> > > The later could be seen quite reasonably from the vmstat counters
> > > already. It can be argued that we are showing too many implementation
> > > details in those tracepoints but I consider them way too lowlevel
> > > already to be usable by any kernel independent userspace. I would be
> > > _really_ surprised if anything but debugging tools have used them.
> > > 
> > > Any feedback is highly appreciated.
> > > 
> > 
> > There is some minor overhead introduced in some paths regardless of
> > whether the tracepoints are active or not but I suspect it's negligible
> > in the context of the overhead of reclaim in general so;
> 
> I will work on improving some of them. E.g. I've dropped the change to
> free_hot_cold_page_list because that is indeed a hot path but other than
> that there shouldn't be any even medium hot path which should see any
> overhead I can see. If you are aware of any, please let me know and I
> will think about how to improve it.
>  

I didn't spot one. The path where I saw the most overhead is already
quite heavy.

-- 
Mel Gorman
SUSE Labs

[toc] | [prev] | [standalone]

Page 2 of 2 — ← Prev page 1 [2]

Back to top | Article view | linux.kernel

csiph-web