Skip to contentSkip to conversation

compaction: handle active and inactive fairly in too_many_isolated

PATCH Superseded Minchan Kim 2 participants 6 messages Aug 24, 2010 – Aug 26, 2010

Subsystems (6)
lore
Share & tools
lore
#0 lore raw

PATCH v2 1/2 compaction: handle active and inactive fairly in too_many_isolated

Iram reported compaction's too_many_isolated loops forever.
(http://www.spinics.net/lists/linux-mm/msg08123.html)

The meminfo of situation happened was inactive anon is zero. That's because the system has no memory pressure until then. While all anon pages was in active lru, compaction could select active lru as well as inactive lru. That's different things with vmscan's isolated. So we has been two too_many_isolated.

While compaction can isolated pages in both active and inactive, current implementation of too_many_isolated only considers inactive. It made Iram's problem.

This patch handles active and inactive with fair.
That's because we can't expect where from and how many compaction would isolated pages.

This patch changes (nr_isolated > nr_inactive) with nr_isolated > (nr_active + nr_inactive) / 2.

---
 mm/compaction.c |    7 ++++---
 1 files changed, 4 insertions(+), 3 deletions(-)
mm/compaction.c +4 -3
index 94cce51..4d709ee 100644
@@ -214,15 +214,16 @@ static void acct_isolated(struct zone *zone, struct compact_control *cc)
 /* Similar to reclaim, but different enough that they don't share logic */
 static bool too_many_isolated(struct zone *zone)
 {
-
-	unsigned long inactive, isolated;
+	unsigned long active, inactive, isolated;
 
 	inactive = zone_page_state(zone, NR_INACTIVE_FILE) +
 					zone_page_state(zone, NR_INACTIVE_ANON);
+	active = zone_page_state(zone, NR_ACTIVE_FILE) +
+					zone_page_state(zone, NR_ACTIVE_ANON);
 	isolated = zone_page_state(zone, NR_ISOLATED_FILE) +
 					zone_page_state(zone, NR_ISOLATED_ANON);
 
-	return isolated > inactive;
+	return isolated > (inactive + active) / 2;
 }
 
 /*
Signature · 7 lines
1.7.0.5

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to [email protected].  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"[email protected]"> [email protected] </a>
#1 lore raw

PATCH v2 2/2 compaction: fix COMPACTPAGEFAILED counting

Now update_nr_listpages doesn't have a role. That's because lists passed is always empty just after calling migrate_pages. The migrate_pages cleans up page list which have failed to migrate before returning by aaa994b3.

[PATCH] page migration: handle freeing of pages in migrate_pages()

Do not leave pages on the lists passed to migrate_pages(). Seems that we will not need any postprocessing of pages. This will simplify the handling of pages by the callers of migrate_pages().

At that time, we thought we don't need any postprocessing of pages. But the situation is changed. The compaction need to know the number of failed to migrate for COMPACTPAGEFAILED stat

This patch introduces new argument 'cleanup' to migrate_pages.
This patch make new rule for caller of migrate_pages to call putback_lru_pages. So caller need to clean up the lists so it has a chance to postprocess the pages.

Cc Hugh Dickins <[email protected]>
Cc Andi Kleen <[email protected]>
Cc Christoph Lameter <[email protected]>
Cc Mel Gorman <[email protected]>
---
 mm/memory-failure.c |    1 +
 mm/memory_hotplug.c |    2 ++
 mm/mempolicy.c      |   10 ++++++++--
 mm/migrate.c        |   12 +++++++-----
 4 files changed, 18 insertions(+), 7 deletions(-)
mm/memory-failure.c +1 -0
index 9c26eec..5267861 100644
@@ -1339,6 +1339,7 @@ int soft_offline_page(struct page *page, int flags)
 		list_add(&page->lru, &pagelist);
 		ret = migrate_pages(&pagelist, new_page, MPOL_MF_MOVE_ALL, 0);
 		if (ret) {
+			putback_lru_pages(&pagelist);
 			pr_debug("soft offline: %#lx: migration failed %d, type %lx\n",
 				pfn, ret, page->flags);
 			if (ret > 0)
mm/memory_hotplug.c +2 -0
index a4cfcdc..2638079 100644
@@ -731,6 +731,8 @@ do_migrate_range(unsigned long start_pfn, unsigned long end_pfn)
 		goto out;
 	/* this function returns # of failed pages */
 	ret = migrate_pages(&source, hotremove_migrate_alloc, 0, 1);
+	if (ret)
+		putback_lru_pages(&source);
 
 out:
 	return ret;
mm/mempolicy.c +8 -2
index f969da5..21243b2 100644
@@ -931,8 +931,11 @@ static int migrate_to_node(struct mm_struct *mm, int source, int dest,
 	check_range(mm, mm->mmap->vm_start, mm->task_size, &nmask,
 			flags | MPOL_MF_DISCONTIG_OK, &pagelist);
 
-	if (!list_empty(&pagelist))
+	if (!list_empty(&pagelist)) {
 		err = migrate_pages(&pagelist, new_node_page, dest, 0);
+		if (err)
+			putback_lru_pages(&pagelist);
+	}
 
 	return err;
 }
@@ -1147,9 +1150,12 @@ static long do_mbind(unsigned long start, unsigned long len,
 
 		err = mbind_range(mm, start, end, new);
 
-		if (!list_empty(&pagelist))
+		if (!list_empty(&pagelist)) {
 			nr_failed = migrate_pages(&pagelist, new_vma_page,
 						(unsigned long)vma, 0);
+			if (nr_failed)
+				putback_lru_pages(&pagelist);
+		}
 
 		if (!err && nr_failed && (flags & MPOL_MF_STRICT))
 			err = -EIO;
mm/migrate.c +7 -5
index 38e7cad..ed38c22 100644
@@ -732,8 +732,9 @@ move_newpage:
  *
  * The function returns after 10 attempts or if no pages
  * are movable anymore because to has become empty
- * or no retryable pages exist anymore. All pages will be
- * returned to the LRU or freed.
+ * or no retryable pages exist anymore.
+ * Caller should call putback_lru_pages to return pages to the LRU
+ * or free list.
  *
  * Return: Number of pages not migrated or error code.
  */
@@ -780,8 +781,6 @@ out:
 	if (!swapwrite)
 		current->flags &= ~PF_SWAPWRITE;
 
-	putback_lru_pages(from);
-
 	if (rc)
 		return rc;
 
@@ -890,9 +889,12 @@ set_status:
 	}
 
 	err = 0;
-	if (!list_empty(&pagelist))
+	if (!list_empty(&pagelist)) {
 		err = migrate_pages(&pagelist, new_page_node,
 				(unsigned long)pm, 0);
+		if (err)
+			putback_lru_pages(&pagelist);
+	}
 
 	up_read(&mm->mmap_sem);
 	return err;
Signature · 7 lines
1.7.0.5

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to [email protected].  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"[email protected]"> [email protected] </a>
#2 lore raw

Re: [PATCH v2 2/2] compaction: fix COMPACTPAGEFAILED counting

Broken description. Will resend

Signature · 8 lines
Kind regards,
Minchan Kim

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to [email protected].  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"[email protected]"> [email protected] </a>
#3 lore raw
In reply to Minchan Kim
Quoted 17 lines · Minchan Kim

Iram reported compaction's too_many_isolated loops forever.
(http://www.spinics.net/lists/linux-mm/msg08123.html)

The meminfo of situation happened was inactive anon is zero. That's because the system has no memory pressure until then. While all anon pages was in active lru, compaction could select active lru as well as inactive lru. That's different things with vmscan's isolated. So we has been two too_many_isolated.

While compaction can isolated pages in both active and inactive, current implementation of too_many_isolated only considers inactive. It made Iram's problem.

This patch handles active and inactive with fair.
That's because we can't expect where from and how many compaction would isolated pages.

Quoted 7 lines

This patch changes (nr_isolated > nr_inactive) with nr_isolated > (nr_active + nr_inactive) / 2.

Please send this patch on its own as it looks like it should be merged and arguably is a stable candidate for 2.6.35. Alternatively, Andrew, can you pick up just this patch? It seems unrelated to the second patch on COMPACTPAGEFAILED.

Signature · 9 lines
Mel Gorman
Part-time Phd Student                          Linux Technology Center
University of Limerick                         IBM Dublin Software Lab

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to [email protected].  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"[email protected]"> [email protected] </a>
#4 lore raw
Quoted 21 lines · Mel Gorman
Quoted 20 lines · Minchan Kim

Iram reported compaction's too_many_isolated loops forever.
(http://www.spinics.net/lists/linux-mm/msg08123.html)

The meminfo of situation happened was inactive anon is zero. That's because the system has no memory pressure until then. While all anon pages was in active lru, compaction could select active lru as well as inactive lru. That's different things with vmscan's isolated. So we has been two too_many_isolated.

While compaction can isolated pages in both active and inactive, current implementation of too_many_isolated only considers inactive. It made Iram's problem.

This patch handles active and inactive with fair.
That's because we can't expect where from and how many compaction would isolated pages.

This patch changes (nr_isolated > nr_inactive) with nr_isolated > (nr_active + nr_inactive) / 2.

Quoted 8 lines
Quoted 4 lines

Please send this patch on its own as it looks like it should be merged and arguably is a stable candidate for 2.6.35. Alternatively, Andrew, can you pick up just this patch? It seems unrelated to the second patch on COMPACTPAGEFAILED.

I thought it's not urgent and next patch would apply based on this patch without HUNK.
If Andrew doesn't have a response, I will resend as a standalone. Thanks.

Signature · 8 lines
Kind regards,
Minchan Kim

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to [email protected].  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"[email protected]"> [email protected] </a>
#5 lore raw
Quoted 27 lines · Minchan Kim
Quoted 26 lines · Mel Gorman
Quoted 24 lines · Minchan Kim

Iram reported compaction's too_many_isolated loops forever.
(http://www.spinics.net/lists/linux-mm/msg08123.html)

The meminfo of situation happened was inactive anon is zero. That's because the system has no memory pressure until then. While all anon pages was in active lru, compaction could select active lru as well as inactive lru. That's different things with vmscan's isolated. So we has been two too_many_isolated.

While compaction can isolated pages in both active and inactive, current implementation of too_many_isolated only considers inactive. It made Iram's problem.

This patch handles active and inactive with fair.
That's because we can't expect where from and how many compaction would isolated pages.

This patch changes (nr_isolated > nr_inactive) with nr_isolated > (nr_active + nr_inactive) / 2.

Quoted 6 lines
Quoted 3 lines

Please send this patch on its own as it looks like it should be merged and arguably is a stable candidate for 2.6.35. Alternatively, Andrew, can you pick up just this patch? It seems unrelated to the second patch on COMPACTPAGEFAILED.

I thought it's not urgent and next patch would apply based on this patch without HUNK.

Well it fixes a known bug so I would consider it a little urgent. It's not earth shattering or anything but why delay unnecessarily?

Quoted 3 lines

If Andrew doesn't have a response, I will resend as a standalone. Thanks.

Thanks.

Signature · 9 lines
Mel Gorman
Part-time Phd Student                          Linux Technology Center
University of Limerick                         IBM Dublin Software Lab

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to [email protected].  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"[email protected]"> [email protected] </a>