[patch -mm 1/2] mm/compaction: split freepages without holding the zone lock fix

[patch -mm 1/2] mm/compaction: split freepages without holding the zone lock fix

[David Rientjes]

If __isolate_free_page() fails, avoid adding to freelist so we don't call
map_pages() with it.

Signed-off-by: David Rientjes <rientjes@google.com>
---
 Fix for mm-compaction-split-freepages-without-holding-the-zone-lock.patch in
 -mm.

 mm/compaction.c | 29 +++++++++++++----------------
 1 file changed, 13 insertions(+), 16 deletions(-)

diff --git a/mm/compaction.c b/mm/compaction.c
--- a/mm/compaction.c
+++ b/mm/compaction.c
@@ -494,24 +494,21 @@ static unsigned long isolate_freepages_block(struct compact_control *cc,
 
 		/* Found a free page, will break it into order-0 pages */
 		order = page_order(page);
-		isolated = __isolate_free_page(page, page_order(page));
+		isolated = __isolate_free_page(page, order);
+		if (!isolated)
+			goto isolate_fail;
 		set_page_private(page, order);
 		total_isolated += isolated;
 		list_add_tail(&page->lru, freelist);
-
-		/* If a page was split, advance to the end of it */
-		if (isolated) {
-			cc->nr_freepages += isolated;
-			if (!strict &&
-				cc->nr_migratepages <= cc->nr_freepages) {
-				blockpfn += isolated;
-				break;
-			}
-
-			blockpfn += isolated - 1;
-			cursor += isolated - 1;
-			continue;
+		cc->nr_freepages += isolated;
+		if (!strict && cc->nr_migratepages <= cc->nr_freepages) {
+			blockpfn += isolated;
+			break;
 		}
+		/* Advance to end of split page */
+		blockpfn += isolated - 1;
+		cursor += isolated - 1;
+		continue;
 
 isolate_fail:
 		if (strict)
@@ -622,7 +619,7 @@ isolate_freepages_range(struct compact_control *cc,
 		 */
 	}
 
-	/* split_free_page does not map the pages */
+	/* __isolate_free_page() does not map the pages */
 	map_pages(&freelist);
 
 	if (pfn < end_pfn) {
@@ -1124,7 +1121,7 @@ static void isolate_freepages(struct compact_control *cc)
 		}
 	}
 
-	/* split_free_page does not map the pages */
+	/* __isolate_free_page() does not map the pages */
 	map_pages(freelist);
 
 	/*

[patch -mm 2/2] mm, compaction: abort free scanner if split fails

[David Rientjes]

If the memory compaction free scanner cannot successfully split a free
page (only possible due to per-zone low watermark), terminate the free 
scanner rather than continuing to scan memory needlessly.

If the per-zone watermark is insufficient for a free page of 
order <= cc->order, then terminate the scanner since future splits will 
also likely fail.

This prevents the compaction freeing scanner from scanning all memory on 
very large zones (very noticeable for zones > 128GB, for instance) when 
all splits will likely fail.

Signed-off-by: David Rientjes <rientjes@google.com>
---
 Note: I think we may want to backport this to -stable since this problem
 has existed since at least 3.11.  This patch won't cleanly apply to any
 stable tree, though.  If people think it should be backported, let me know
 and I'll handle the failures as they arise and rebase.

 mm/compaction.c | 17 +++++++++++------
 1 file changed, 11 insertions(+), 6 deletions(-)

diff --git a/mm/compaction.c b/mm/compaction.c
--- a/mm/compaction.c
+++ b/mm/compaction.c
@@ -496,7 +496,7 @@ static unsigned long isolate_freepages_block(struct compact_control *cc,
 		order = page_order(page);
 		isolated = __isolate_free_page(page, order);
 		if (!isolated)
-			goto isolate_fail;
+			break;
 		set_page_private(page, order);
 		total_isolated += isolated;
 		list_add_tail(&page->lru, freelist);
@@ -518,6 +518,9 @@ isolate_fail:
 
 	}
 
+	if (locked)
+		spin_unlock_irqrestore(&cc->zone->lock, flags);
+
 	/*
 	 * There is a tiny chance that we have read bogus compound_order(),
 	 * so be careful to not go outside of the pageblock.
@@ -539,9 +542,6 @@ isolate_fail:
 	if (strict && blockpfn < end_pfn)
 		total_isolated = 0;
 
-	if (locked)
-		spin_unlock_irqrestore(&cc->zone->lock, flags);
-
 	/* Update the pageblock-skip if the whole pageblock was scanned */
 	if (blockpfn == end_pfn)
 		update_pageblock_skip(cc, valid_page, total_isolated, false);
@@ -1068,6 +1068,7 @@ static void isolate_freepages(struct compact_control *cc)
 				block_end_pfn = block_start_pfn,
 				block_start_pfn -= pageblock_nr_pages,
 				isolate_start_pfn = block_start_pfn) {
+		unsigned long isolated;
 
 		/*
 		 * This can iterate a massively long zone without finding any
@@ -1092,8 +1093,12 @@ static void isolate_freepages(struct compact_control *cc)
 			continue;
 
 		/* Found a block suitable for isolating free pages from. */
-		isolate_freepages_block(cc, &isolate_start_pfn,
-					block_end_pfn, freelist, false);
+		isolated = isolate_freepages_block(cc, &isolate_start_pfn,
+						block_end_pfn, freelist, false);
+		/* If isolation failed, do not continue needlessly */
+		if (!isolated && isolate_start_pfn < block_end_pfn &&
+		    cc->nr_freepages <= cc->nr_migratepages)
+			break;
 
 		/*
 		 * If we isolated enough freepages, or aborted due to async

[patch] mm, compaction: abort free scanner if split fails

[David Rientjes]

If the memory compaction free scanner cannot successfully split a free
page (only possible due to per-zone low watermark), terminate the free 
scanner rather than continuing to scan memory needlessly.  If the 
watermark is insufficient for a free page of order <= cc->order, then 
terminate the scanner since all future splits will also likely fail.

This prevents the compaction freeing scanner from scanning all memory on 
very large zones (very noticeable for zones > 128GB, for instance) when 
all splits will likely fail while holding zone->lock.

Cc: stable@vger.kernel.org
Signed-off-by: David Rientjes <rientjes@google.com>
---
 Based on Linus's tree

 Suggest including in 4.7 if anybody else agrees?

 mm/compaction.c | 39 +++++++++++++++++++++------------------
 1 file changed, 21 insertions(+), 18 deletions(-)

diff --git a/mm/compaction.c b/mm/compaction.c
--- a/mm/compaction.c
+++ b/mm/compaction.c
@@ -441,25 +441,23 @@ static unsigned long isolate_freepages_block(struct compact_control *cc,
 
 		/* Found a free page, break it into order-0 pages */
 		isolated = split_free_page(page);
+		if (!isolated)
+			break;
+
 		total_isolated += isolated;
+		cc->nr_freepages += isolated;
 		for (i = 0; i < isolated; i++) {
 			list_add(&page->lru, freelist);
 			page++;
 		}
-
-		/* If a page was split, advance to the end of it */
-		if (isolated) {
-			cc->nr_freepages += isolated;
-			if (!strict &&
-				cc->nr_migratepages <= cc->nr_freepages) {
-				blockpfn += isolated;
-				break;
-			}
-
-			blockpfn += isolated - 1;
-			cursor += isolated - 1;
-			continue;
+		if (!strict && cc->nr_migratepages <= cc->nr_freepages) {
+			blockpfn += isolated;
+			break;
 		}
+		/* Advance to the end of split page */
+		blockpfn += isolated - 1;
+		cursor += isolated - 1;
+		continue;
 
 isolate_fail:
 		if (strict)
@@ -469,6 +467,9 @@ isolate_fail:
 
 	}
 
+	if (locked)
+		spin_unlock_irqrestore(&cc->zone->lock, flags);
+
 	/*
 	 * There is a tiny chance that we have read bogus compound_order(),
 	 * so be careful to not go outside of the pageblock.
@@ -490,9 +491,6 @@ isolate_fail:
 	if (strict && blockpfn < end_pfn)
 		total_isolated = 0;
 
-	if (locked)
-		spin_unlock_irqrestore(&cc->zone->lock, flags);
-
 	/* Update the pageblock-skip if the whole pageblock was scanned */
 	if (blockpfn == end_pfn)
 		update_pageblock_skip(cc, valid_page, total_isolated, false);
@@ -1011,6 +1009,7 @@ static void isolate_freepages(struct compact_control *cc)
 				block_end_pfn = block_start_pfn,
 				block_start_pfn -= pageblock_nr_pages,
 				isolate_start_pfn = block_start_pfn) {
+		unsigned long isolated;
 
 		/*
 		 * This can iterate a massively long zone without finding any
@@ -1035,8 +1034,12 @@ static void isolate_freepages(struct compact_control *cc)
 			continue;
 
 		/* Found a block suitable for isolating free pages from. */
-		isolate_freepages_block(cc, &isolate_start_pfn,
-					block_end_pfn, freelist, false);
+		isolated = isolate_freepages_block(cc, &isolate_start_pfn,
+						block_end_pfn, freelist, false);
+		/* If isolation failed early, do not continue needlessly */
+		if (!isolated && isolate_start_pfn < block_end_pfn &&
+		    cc->nr_migratepages > cc->nr_freepages)
+			break;
 
 		/*
 		 * If we isolated enough freepages, or aborted due to async

Re: [patch] mm, compaction: abort free scanner if split fails

[Vlastimil Babka]

On 06/22/2016 03:22 AM, David Rientjes wrote:
> If the memory compaction free scanner cannot successfully split a free
> page (only possible due to per-zone low watermark), terminate the free
> scanner rather than continuing to scan memory needlessly.  If the
> watermark is insufficient for a free page of order <= cc->order, then
> terminate the scanner since all future splits will also likely fail.
>
> This prevents the compaction freeing scanner from scanning all memory on
> very large zones (very noticeable for zones > 128GB, for instance) when
> all splits will likely fail while holding zone->lock.
>
> Cc: stable@vger.kernel.org
> Signed-off-by: David Rientjes <rientjes@google.com>

Acked-by: Vlastimil Babka <vbabka@suse.cz>

> ---
>  Based on Linus's tree
>
>  Suggest including in 4.7 if anybody else agrees?

4.7 definitely. Stable is less clear especially if you say it won't apply 
cleanly, but if you're ready to handle it, sure. The rules now allow fixing 
glaring performance bugs.

Thanks.

Re: [patch] mm, compaction: abort free scanner if split fails

[Andrew Morton]

On Tue, 21 Jun 2016 18:22:49 -0700 (PDT) David Rientjes <rientjes@google.com> wrote:

> If the memory compaction free scanner cannot successfully split a free
> page (only possible due to per-zone low watermark), terminate the free 
> scanner rather than continuing to scan memory needlessly.  If the 
> watermark is insufficient for a free page of order <= cc->order, then 
> terminate the scanner since all future splits will also likely fail.
> 
> This prevents the compaction freeing scanner from scanning all memory on 
> very large zones (very noticeable for zones > 128GB, for instance) when 
> all splits will likely fail while holding zone->lock.
> 

This collides pretty heavily with Joonsoo's "mm/compaction: split
freepages without holding the zone lock".

I ended up with this, in isolate_freepages_block():

		/* Found a free page, will break it into order-0 pages */
		order = page_order(page);
		isolated = __isolate_free_page(page, page_order(page));
		set_page_private(page, order);

		total_isolated += isolated;
		cc->nr_freepages += isolated;
		list_add_tail(&page->lru, freelist);

		if (!strict && cc->nr_migratepages <= cc->nr_freepages) {
			blockpfn += isolated;
			break;
		}
		/* Advance to the end of split page */
		blockpfn += isolated - 1;
		cursor += isolated - 1;
		continue;

isolate_fail:

Re: [patch] mm, compaction: abort free scanner if split fails

[Andrew Morton]

On Wed, 22 Jun 2016 14:56:17 -0700 Andrew Morton <akpm@linux-foundation.org> wrote:

> On Tue, 21 Jun 2016 18:22:49 -0700 (PDT) David Rientjes <rientjes@google.com> wrote:
> 
> > If the memory compaction free scanner cannot successfully split a free
> > page (only possible due to per-zone low watermark), terminate the free 
> > scanner rather than continuing to scan memory needlessly.  If the 
> > watermark is insufficient for a free page of order <= cc->order, then 
> > terminate the scanner since all future splits will also likely fail.
> > 
> > This prevents the compaction freeing scanner from scanning all memory on 
> > very large zones (very noticeable for zones > 128GB, for instance) when 
> > all splits will likely fail while holding zone->lock.
> > 
> 
> This collides pretty heavily with Joonsoo's "mm/compaction: split
> freepages without holding the zone lock".
> 
> I ended up with this, in isolate_freepages_block():
> 
> 		/* Found a free page, will break it into order-0 pages */
> 		order = page_order(page);
> 		isolated = __isolate_free_page(page, page_order(page));
> 		set_page_private(page, order);
> 
> 		total_isolated += isolated;
> 		cc->nr_freepages += isolated;
> 		list_add_tail(&page->lru, freelist);
> 
> 		if (!strict && cc->nr_migratepages <= cc->nr_freepages) {
> 			blockpfn += isolated;
> 			break;
> 		}
> 		/* Advance to the end of split page */
> 		blockpfn += isolated - 1;
> 		cursor += isolated - 1;
> 		continue;
> 
> isolate_fail:
> 

And
mm-compaction-split-freepages-without-holding-the-zone-lock-fix.patch
churns things around some more.  Now this:


		/* Found a free page, will break it into order-0 pages */
		order = page_order(page);
		isolated = __isolate_free_page(page, order);
		set_page_private(page, order);
		total_isolated += isolated;
		list_add_tail(&page->lru, freelist);
		cc->nr_freepages += isolated;
		if (!strict && cc->nr_migratepages <= cc->nr_freepages) {
			blockpfn += isolated;
			break;
		}
		/* Advance to the end of split page */
		blockpfn += isolated - 1;
		cursor += isolated - 1;
		continue;

isolate_fail:

and things are looking a bit better...

Re: [patch] mm, compaction: abort free scanner if split fails

[David Rientjes]

On Wed, 22 Jun 2016, Andrew Morton wrote:

> And
> mm-compaction-split-freepages-without-holding-the-zone-lock-fix.patch
> churns things around some more.  Now this:
> 
> 
> 		/* Found a free page, will break it into order-0 pages */
> 		order = page_order(page);
> 		isolated = __isolate_free_page(page, order);
> 		set_page_private(page, order);
> 		total_isolated += isolated;
> 		list_add_tail(&page->lru, freelist);
> 		cc->nr_freepages += isolated;
> 		if (!strict && cc->nr_migratepages <= cc->nr_freepages) {
> 			blockpfn += isolated;
> 			break;
> 		}
> 		/* Advance to the end of split page */
> 		blockpfn += isolated - 1;
> 		cursor += isolated - 1;
> 		continue;
> 
> isolate_fail:
> 
> and things are looking a bit better...
> 

This looks like it's missing the

	if (!isolated)
		break;

check from mm-compaction-abort-free-scanner-if-split-fails.patch which is 
needed to properly terminate when the low watermark fails (and adding to 
freelist as Minchan mentioned before I saw this patch).

I rebased 
mm-compaction-split-freepages-without-holding-the-zone-lock.patch as I 
thought it should be done and folded 
mm-compaction-split-freepages-without-holding-the-zone-lock-fix.patch into 
it for simplicity.  I think this should replace 
mm-compaction-split-freepages-without-holding-the-zone-lock.patch in -mm.


From: Joonsoo Kim <iamjoonsoo.kim@lge.com>

We don't need to split freepages with holding the zone lock.  It will
cause more contention on zone lock so not desirable.

Signed-off-by: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
---
diff --git a/include/linux/mm.h b/include/linux/mm.h
--- a/include/linux/mm.h
+++ b/include/linux/mm.h
@@ -537,7 +537,6 @@ void __put_page(struct page *page);
 void put_pages_list(struct list_head *pages);
 
 void split_page(struct page *page, unsigned int order);
-int split_free_page(struct page *page);
 
 /*
  * Compound pages have a destructor function.  Provide a
diff --git a/mm/compaction.c b/mm/compaction.c
index ab21497..9d17b21 100644
--- a/mm/compaction.c
+++ b/mm/compaction.c
@@ -65,13 +65,31 @@ static unsigned long release_freepages(struct list_head *freelist)
 
 static void map_pages(struct list_head *list)
 {
-	struct page *page;
+	unsigned int i, order, nr_pages;
+	struct page *page, *next;
+	LIST_HEAD(tmp_list);
+
+	list_for_each_entry_safe(page, next, list, lru) {
+		list_del(&page->lru);
 
-	list_for_each_entry(page, list, lru) {
-		arch_alloc_page(page, 0);
-		kernel_map_pages(page, 1, 1);
-		kasan_alloc_pages(page, 0);
+		order = page_private(page);
+		nr_pages = 1 << order;
+		set_page_private(page, 0);
+		set_page_refcounted(page);
+
+		arch_alloc_page(page, order);
+		kernel_map_pages(page, nr_pages, 1);
+		kasan_alloc_pages(page, order);
+		if (order)
+			split_page(page, order);
+
+		for (i = 0; i < nr_pages; i++) {
+			list_add(&page->lru, &tmp_list);
+			page++;
+		}
 	}
+
+	list_splice(&tmp_list, list);
 }
 
 static inline bool migrate_async_suitable(int migratetype)
@@ -406,12 +424,13 @@ static unsigned long isolate_freepages_block(struct compact_control *cc,
 	unsigned long flags = 0;
 	bool locked = false;
 	unsigned long blockpfn = *start_pfn;
+	unsigned int order;
 
 	cursor = pfn_to_page(blockpfn);
 
 	/* Isolate free pages. */
 	for (; blockpfn < end_pfn; blockpfn++, cursor++) {
-		int isolated, i;
+		int isolated;
 		struct page *page = cursor;
 
 		/*
@@ -477,17 +496,17 @@ static unsigned long isolate_freepages_block(struct compact_control *cc,
 				goto isolate_fail;
 		}
 
-		/* Found a free page, break it into order-0 pages */
-		isolated = split_free_page(page);
+		/* Found a free page, will break it into order-0 pages */
+		order = page_order(page);
+		isolated = __isolate_free_page(page, order);
 		if (!isolated)
 			break;
+		set_page_private(page, order);
 
 		total_isolated += isolated;
 		cc->nr_freepages += isolated;
-		for (i = 0; i < isolated; i++) {
-			list_add(&page->lru, freelist);
-			page++;
-		}
+		list_add_tail(&page->lru, freelist);
+
 		if (!strict && cc->nr_migratepages <= cc->nr_freepages) {
 			blockpfn += isolated;
 			break;
@@ -606,7 +625,7 @@ isolate_freepages_range(struct compact_control *cc,
 		 */
 	}
 
-	/* split_free_page does not map the pages */
+	/* __isolate_free_page() does not map the pages */
 	map_pages(&freelist);
 
 	if (pfn < end_pfn) {
@@ -1113,7 +1132,7 @@ static void isolate_freepages(struct compact_control *cc)
 		}
 	}
 
-	/* split_free_page does not map the pages */
+	/* __isolate_free_page() does not map the pages */
 	map_pages(freelist);
 
 	/*
diff --git a/mm/page_alloc.c b/mm/page_alloc.c
--- a/mm/page_alloc.c
+++ b/mm/page_alloc.c
@@ -2560,33 +2560,6 @@ int __isolate_free_page(struct page *page, unsigned int order)
 }
 
 /*
- * Similar to split_page except the page is already free. As this is only
- * being used for migration, the migratetype of the block also changes.
- * As this is called with interrupts disabled, the caller is responsible
- * for calling arch_alloc_page() and kernel_map_page() after interrupts
- * are enabled.
- *
- * Note: this is probably too low level an operation for use in drivers.
- * Please consult with lkml before using this in your driver.
- */
-int split_free_page(struct page *page)
-{
-	unsigned int order;
-	int nr_pages;
-
-	order = page_order(page);
-
-	nr_pages = __isolate_free_page(page, order);
-	if (!nr_pages)
-		return 0;
-
-	/* Split into individual pages */
-	set_page_refcounted(page);
-	split_page(page, order);
-	return nr_pages;
-}
-
-/*
  * Update NUMA hit/miss statistics
  *
  * Must be called with interrupts disabled.

Re: [patch] mm, compaction: abort free scanner if split fails

[Vlastimil Babka]

On 06/23/2016 01:40 AM, David Rientjes wrote:
> On Wed, 22 Jun 2016, Andrew Morton wrote:
>
>> And
>> mm-compaction-split-freepages-without-holding-the-zone-lock-fix.patch
>> churns things around some more.  Now this:
>>
>>
>> 		/* Found a free page, will break it into order-0 pages */
>> 		order = page_order(page);
>> 		isolated = __isolate_free_page(page, order);
>> 		set_page_private(page, order);
>> 		total_isolated += isolated;
>> 		list_add_tail(&page->lru, freelist);
>> 		cc->nr_freepages += isolated;
>> 		if (!strict && cc->nr_migratepages <= cc->nr_freepages) {
>> 			blockpfn += isolated;
>> 			break;
>> 		}
>> 		/* Advance to the end of split page */
>> 		blockpfn += isolated - 1;
>> 		cursor += isolated - 1;
>> 		continue;
>>
>> isolate_fail:
>>
>> and things are looking a bit better...
>>
>
> This looks like it's missing the
>
> 	if (!isolated)
> 		break;
>
> check from mm-compaction-abort-free-scanner-if-split-fails.patch which is
> needed to properly terminate when the low watermark fails (and adding to
> freelist as Minchan mentioned before I saw this patch).

Agreed.

>
> I rebased
> mm-compaction-split-freepages-without-holding-the-zone-lock.patch as I
> thought it should be done and folded
> mm-compaction-split-freepages-without-holding-the-zone-lock-fix.patch into
> it for simplicity.  I think this should replace
> mm-compaction-split-freepages-without-holding-the-zone-lock.patch in -mm.

Yes, it should replace both the .patch and the -fix.patch.

Thanks!

>
> From: Joonsoo Kim <iamjoonsoo.kim@lge.com>
>
> We don't need to split freepages with holding the zone lock.  It will
> cause more contention on zone lock so not desirable.
>
> Signed-off-by: Joonsoo Kim <iamjoonsoo.kim@lge.com>
> Acked-by: Vlastimil Babka <vbabka@suse.cz>
> ---
> diff --git a/include/linux/mm.h b/include/linux/mm.h
> --- a/include/linux/mm.h
> +++ b/include/linux/mm.h
> @@ -537,7 +537,6 @@ void __put_page(struct page *page);
>  void put_pages_list(struct list_head *pages);
>
>  void split_page(struct page *page, unsigned int order);
> -int split_free_page(struct page *page);
>
>  /*
>   * Compound pages have a destructor function.  Provide a
> diff --git a/mm/compaction.c b/mm/compaction.c
> index ab21497..9d17b21 100644
> --- a/mm/compaction.c
> +++ b/mm/compaction.c
> @@ -65,13 +65,31 @@ static unsigned long release_freepages(struct list_head *freelist)
>
>  static void map_pages(struct list_head *list)
>  {
> -	struct page *page;
> +	unsigned int i, order, nr_pages;
> +	struct page *page, *next;
> +	LIST_HEAD(tmp_list);
> +
> +	list_for_each_entry_safe(page, next, list, lru) {
> +		list_del(&page->lru);
>
> -	list_for_each_entry(page, list, lru) {
> -		arch_alloc_page(page, 0);
> -		kernel_map_pages(page, 1, 1);
> -		kasan_alloc_pages(page, 0);
> +		order = page_private(page);
> +		nr_pages = 1 << order;
> +		set_page_private(page, 0);
> +		set_page_refcounted(page);
> +
> +		arch_alloc_page(page, order);
> +		kernel_map_pages(page, nr_pages, 1);
> +		kasan_alloc_pages(page, order);
> +		if (order)
> +			split_page(page, order);
> +
> +		for (i = 0; i < nr_pages; i++) {
> +			list_add(&page->lru, &tmp_list);
> +			page++;
> +		}
>  	}
> +
> +	list_splice(&tmp_list, list);
>  }
>
>  static inline bool migrate_async_suitable(int migratetype)
> @@ -406,12 +424,13 @@ static unsigned long isolate_freepages_block(struct compact_control *cc,
>  	unsigned long flags = 0;
>  	bool locked = false;
>  	unsigned long blockpfn = *start_pfn;
> +	unsigned int order;
>
>  	cursor = pfn_to_page(blockpfn);
>
>  	/* Isolate free pages. */
>  	for (; blockpfn < end_pfn; blockpfn++, cursor++) {
> -		int isolated, i;
> +		int isolated;
>  		struct page *page = cursor;
>
>  		/*
> @@ -477,17 +496,17 @@ static unsigned long isolate_freepages_block(struct compact_control *cc,
>  				goto isolate_fail;
>  		}
>
> -		/* Found a free page, break it into order-0 pages */
> -		isolated = split_free_page(page);
> +		/* Found a free page, will break it into order-0 pages */
> +		order = page_order(page);
> +		isolated = __isolate_free_page(page, order);
>  		if (!isolated)
>  			break;
> +		set_page_private(page, order);
>
>  		total_isolated += isolated;
>  		cc->nr_freepages += isolated;
> -		for (i = 0; i < isolated; i++) {
> -			list_add(&page->lru, freelist);
> -			page++;
> -		}
> +		list_add_tail(&page->lru, freelist);
> +
>  		if (!strict && cc->nr_migratepages <= cc->nr_freepages) {
>  			blockpfn += isolated;
>  			break;
> @@ -606,7 +625,7 @@ isolate_freepages_range(struct compact_control *cc,
>  		 */
>  	}
>
> -	/* split_free_page does not map the pages */
> +	/* __isolate_free_page() does not map the pages */
>  	map_pages(&freelist);
>
>  	if (pfn < end_pfn) {
> @@ -1113,7 +1132,7 @@ static void isolate_freepages(struct compact_control *cc)
>  		}
>  	}
>
> -	/* split_free_page does not map the pages */
> +	/* __isolate_free_page() does not map the pages */
>  	map_pages(freelist);
>
>  	/*
> diff --git a/mm/page_alloc.c b/mm/page_alloc.c
> --- a/mm/page_alloc.c
> +++ b/mm/page_alloc.c
> @@ -2560,33 +2560,6 @@ int __isolate_free_page(struct page *page, unsigned int order)
>  }
>
>  /*
> - * Similar to split_page except the page is already free. As this is only
> - * being used for migration, the migratetype of the block also changes.
> - * As this is called with interrupts disabled, the caller is responsible
> - * for calling arch_alloc_page() and kernel_map_page() after interrupts
> - * are enabled.
> - *
> - * Note: this is probably too low level an operation for use in drivers.
> - * Please consult with lkml before using this in your driver.
> - */
> -int split_free_page(struct page *page)
> -{
> -	unsigned int order;
> -	int nr_pages;
> -
> -	order = page_order(page);
> -
> -	nr_pages = __isolate_free_page(page, order);
> -	if (!nr_pages)
> -		return 0;
> -
> -	/* Split into individual pages */
> -	set_page_refcounted(page);
> -	split_page(page, order);
> -	return nr_pages;
> -}
> -
> -/*
>   * Update NUMA hit/miss statistics
>   *
>   * Must be called with interrupts disabled.
>

Re: [patch] mm, compaction: abort free scanner if split fails

[David Rientjes]

On Wed, 22 Jun 2016, Andrew Morton wrote:

> On Tue, 21 Jun 2016 18:22:49 -0700 (PDT) David Rientjes <rientjes@google.com> wrote:
> 
> > If the memory compaction free scanner cannot successfully split a free
> > page (only possible due to per-zone low watermark), terminate the free 
> > scanner rather than continuing to scan memory needlessly.  If the 
> > watermark is insufficient for a free page of order <= cc->order, then 
> > terminate the scanner since all future splits will also likely fail.
> > 
> > This prevents the compaction freeing scanner from scanning all memory on 
> > very large zones (very noticeable for zones > 128GB, for instance) when 
> > all splits will likely fail while holding zone->lock.
> > 
> 
> This collides pretty heavily with Joonsoo's "mm/compaction: split
> freepages without holding the zone lock".
> 

Sorry if it wasn't clear, but I was proposing this patch for 4.7 
inclusion and Vlastimil agreed we should ask for that.  Joonsoo said he 
was prepared to rebase on top of that.  Is 
mm-compaction-split-freepages-without-holding-the-zone-lock.patch and 
friends going into 4.7 or are we deferring this fix until 4.8?

compaction_alloc() iterating a 128GB zone has been benchmarked to take 
over 400ms on some systems whereas any free page isolated and ready to be 
split ends up failing in split_free_page() because of the low watermark 
check and thus the iteration continues.

The next time compaction occurs, the freeing scanner will likely start at 
the end of the zone again since no success was made previously and we get 
the same lengthy iteration until the zone is brought above the low 
watermark.  All thp page faults can take >400ms in such a state without 
this fix.

Re: [patch] mm, compaction: abort free scanner if split fails

[Andrew Morton]

On Wed, 22 Jun 2016 15:06:29 -0700 (PDT) David Rientjes <rientjes@google.com> wrote:

> On Wed, 22 Jun 2016, Andrew Morton wrote:
> 
> > On Tue, 21 Jun 2016 18:22:49 -0700 (PDT) David Rientjes <rientjes@google.com> wrote:
> > 
> > > If the memory compaction free scanner cannot successfully split a free
> > > page (only possible due to per-zone low watermark), terminate the free 
> > > scanner rather than continuing to scan memory needlessly.  If the 
> > > watermark is insufficient for a free page of order <= cc->order, then 
> > > terminate the scanner since all future splits will also likely fail.
> > > 
> > > This prevents the compaction freeing scanner from scanning all memory on 
> > > very large zones (very noticeable for zones > 128GB, for instance) when 
> > > all splits will likely fail while holding zone->lock.
> > > 
> > 
> > This collides pretty heavily with Joonsoo's "mm/compaction: split
> > freepages without holding the zone lock".
> > 
> 
> Sorry if it wasn't clear, but I was proposing this patch for 4.7 
> inclusion and Vlastimil agreed we should ask for that.  Joonsoo said he 
> was prepared to rebase on top of that.  Is 
> mm-compaction-split-freepages-without-holding-the-zone-lock.patch and 
> friends going into 4.7 or are we deferring this fix until 4.8?

I have this patch lined up for 4.7 so I was rebasing Joonsoo's patches
on top.

[patch] mm, compaction: ignore watermarks when isolating free pages

[David Rientjes]

The goal of memory compaction is to defragment memory by moving migratable 
pages to free pages at the end of the zone.  No additional memory is being 
allocated.

Ignore per-zone low watermarks in __isolate_free_page() because memory is 
either fully migrated or isolated free pages are returned when migration 
fails.

This fixes an issue where the compaction freeing scanner can isolate 
memory but the zone drops below its low watermark for that page order, so 
the scanner must continue to scan all memory pointlessly.

Signed-off-by: David Rientjes <rientjes@google.com>
---
 mm/page_alloc.c | 14 ++------------
 1 file changed, 2 insertions(+), 12 deletions(-)

diff --git a/mm/page_alloc.c b/mm/page_alloc.c
--- a/mm/page_alloc.c
+++ b/mm/page_alloc.c
@@ -2484,23 +2484,14 @@ EXPORT_SYMBOL_GPL(split_page);
 
 int __isolate_free_page(struct page *page, unsigned int order)
 {
-	unsigned long watermark;
 	struct zone *zone;
-	int mt;
+	const int mt = get_pageblock_migratetype(page);
 
 	BUG_ON(!PageBuddy(page));
-
 	zone = page_zone(page);
-	mt = get_pageblock_migratetype(page);
-
-	if (!is_migrate_isolate(mt)) {
-		/* Obey watermarks as if the page was being allocated */
-		watermark = low_wmark_pages(zone) + (1 << order);
-		if (!zone_watermark_ok(zone, 0, watermark, 0, 0))
-			return 0;
 
+	if (!is_migrate_isolate(mt))
 		__mod_zone_freepage_state(zone, -(1UL << order), mt);
-	}
 
 	/* Remove page from free list */
 	list_del(&page->lru);
@@ -2520,7 +2511,6 @@ int __isolate_free_page(struct page *page, unsigned int order)
 		}
 	}
 
-
 	return 1UL << order;
 }
 

Re: [patch] mm, compaction: ignore watermarks when isolating free pages

[Vlastimil Babka]

On 06/16/2016 12:34 AM, David Rientjes wrote:
> The goal of memory compaction is to defragment memory by moving migratable
> pages to free pages at the end of the zone.  No additional memory is being
> allocated.
>
> Ignore per-zone low watermarks in __isolate_free_page() because memory is
> either fully migrated or isolated free pages are returned when migration
> fails.

Michal Hocko suggested that too, but I didn't think it safe that 
compaction should go below the min watermark, even temporarily. It means 
the system is struggling with order-0 allocations, so making it worse 
for the benefit of high-order allocations doesn't make sense. The 
high-order allocation would likely fail anyway due to watermark checks, 
even if the page of sufficient order was formed by compaction. So in my 
series, I just changed the low watermark check to min [1].

> This fixes an issue where the compaction freeing scanner can isolate
> memory but the zone drops below its low watermark for that page order, so
> the scanner must continue to scan all memory pointlessly.

Good point, looks like failing the watermark is the only reason when 
__isolate_free_page() can fail. isolate_freepages_block() and its 
callers should take this as an indication that compaction should return 
with failure immediately.

[1] http://article.gmane.org/gmane.linux.kernel/2231369

> Signed-off-by: David Rientjes <rientjes@google.com>
> ---
>  mm/page_alloc.c | 14 ++------------
>  1 file changed, 2 insertions(+), 12 deletions(-)
>
> diff --git a/mm/page_alloc.c b/mm/page_alloc.c
> --- a/mm/page_alloc.c
> +++ b/mm/page_alloc.c
> @@ -2484,23 +2484,14 @@ EXPORT_SYMBOL_GPL(split_page);
>
>  int __isolate_free_page(struct page *page, unsigned int order)
>  {
> -	unsigned long watermark;
>  	struct zone *zone;
> -	int mt;
> +	const int mt = get_pageblock_migratetype(page);
>
>  	BUG_ON(!PageBuddy(page));
> -
>  	zone = page_zone(page);
> -	mt = get_pageblock_migratetype(page);
> -
> -	if (!is_migrate_isolate(mt)) {
> -		/* Obey watermarks as if the page was being allocated */
> -		watermark = low_wmark_pages(zone) + (1 << order);
> -		if (!zone_watermark_ok(zone, 0, watermark, 0, 0))
> -			return 0;
>
> +	if (!is_migrate_isolate(mt))
>  		__mod_zone_freepage_state(zone, -(1UL << order), mt);
> -	}
>
>  	/* Remove page from free list */
>  	list_del(&page->lru);
> @@ -2520,7 +2511,6 @@ int __isolate_free_page(struct page *page, unsigned int order)
>  		}
>  	}
>
> -
>  	return 1UL << order;
>  }
>
>

[patch] mm, compaction: abort free scanner if split fails

[David Rientjes]

If the memory compaction free scanner cannot successfully split a free
page (only possible due to per-zone low watermark), terminate the free 
scanner rather than continuing to scan memory needlessly.

If the per-zone watermark is insufficient for a free page of 
order <= cc->order, then terminate the scanner since future splits will 
also likely fail.

This prevents the compaction freeing scanner from scanning all memory on 
very large zones (very noticeable for zones > 128GB, for instance) when 
all splits will likely fail.

Signed-off-by: David Rientjes <rientjes@google.com>
---
 mm/compaction.c | 49 +++++++++++++++++++++++++++++--------------------
 1 file changed, 29 insertions(+), 20 deletions(-)

diff --git a/mm/compaction.c b/mm/compaction.c
--- a/mm/compaction.c
+++ b/mm/compaction.c
@@ -494,24 +494,22 @@ static unsigned long isolate_freepages_block(struct compact_control *cc,
 
 		/* Found a free page, will break it into order-0 pages */
 		order = page_order(page);
-		isolated = __isolate_free_page(page, page_order(page));
+		isolated = __isolate_free_page(page, order);
+		if (!isolated)
+			break;
 		set_page_private(page, order);
 		total_isolated += isolated;
 		list_add_tail(&page->lru, freelist);
 
-		/* If a page was split, advance to the end of it */
-		if (isolated) {
-			cc->nr_freepages += isolated;
-			if (!strict &&
-				cc->nr_migratepages <= cc->nr_freepages) {
-				blockpfn += isolated;
-				break;
-			}
-
-			blockpfn += isolated - 1;
-			cursor += isolated - 1;
-			continue;
+		/* Advance to the end of split page */
+		cc->nr_freepages += isolated;
+		if (!strict && cc->nr_migratepages <= cc->nr_freepages) {
+			blockpfn += isolated;
+			break;
 		}
+		blockpfn += isolated - 1;
+		cursor += isolated - 1;
+		continue;
 
 isolate_fail:
 		if (strict)
@@ -521,6 +519,9 @@ isolate_fail:
 
 	}
 
+	if (locked)
+		spin_unlock_irqrestore(&cc->zone->lock, flags);
+
 	/*
 	 * There is a tiny chance that we have read bogus compound_order(),
 	 * so be careful to not go outside of the pageblock.
@@ -542,9 +543,6 @@ isolate_fail:
 	if (strict && blockpfn < end_pfn)
 		total_isolated = 0;
 
-	if (locked)
-		spin_unlock_irqrestore(&cc->zone->lock, flags);
-
 	/* Update the pageblock-skip if the whole pageblock was scanned */
 	if (blockpfn == end_pfn)
 		update_pageblock_skip(cc, valid_page, total_isolated, false);
@@ -622,7 +620,7 @@ isolate_freepages_range(struct compact_control *cc,
 		 */
 	}
 
-	/* split_free_page does not map the pages */
+	/* __isolate_free_page() does not map the pages */
 	map_pages(&freelist);
 
 	if (pfn < end_pfn) {
@@ -1071,6 +1069,7 @@ static void isolate_freepages(struct compact_control *cc)
 				block_end_pfn = block_start_pfn,
 				block_start_pfn -= pageblock_nr_pages,
 				isolate_start_pfn = block_start_pfn) {
+		unsigned long isolated;
 
 		/*
 		 * This can iterate a massively long zone without finding any
@@ -1095,8 +1094,12 @@ static void isolate_freepages(struct compact_control *cc)
 			continue;
 
 		/* Found a block suitable for isolating free pages from. */
-		isolate_freepages_block(cc, &isolate_start_pfn,
-					block_end_pfn, freelist, false);
+		isolated = isolate_freepages_block(cc, &isolate_start_pfn,
+						block_end_pfn, freelist, false);
+		/* If free page split failed, do not continue needlessly */
+		if (!isolated && isolate_start_pfn < block_end_pfn &&
+		    cc->nr_freepages <= cc->nr_migratepages)
+			break;
 
 		/*
 		 * If we isolated enough freepages, or aborted due to async
@@ -1124,7 +1127,7 @@ static void isolate_freepages(struct compact_control *cc)
 		}
 	}
 
-	/* split_free_page does not map the pages */
+	/* __isolate_free_page() does not map the pages */
 	map_pages(freelist);
 
 	/*
@@ -1703,6 +1706,12 @@ enum compact_result try_to_compact_pages(gfp_t gfp_mask, unsigned int order,
 			continue;
 		}
 
+		/* Don't attempt compaction if splitting free page will fail */
+		if (!zone_watermark_ok(zone, 0,
+				       low_wmark_pages(zone) + (1 << order),
+				       0, 0))
+			continue;
+
 		status = compact_zone_order(zone, order, gfp_mask, mode,
 				&zone_contended, alloc_flags,
 				ac_classzone_idx(ac));

Re: [patch] mm, compaction: abort free scanner if split fails

[Vlastimil Babka]

On 06/21/2016 12:27 AM, David Rientjes wrote:
> If the memory compaction free scanner cannot successfully split a free
> page (only possible due to per-zone low watermark), terminate the free
> scanner rather than continuing to scan memory needlessly.
>
> If the per-zone watermark is insufficient for a free page of
> order <= cc->order, then terminate the scanner since future splits will
> also likely fail.
>
> This prevents the compaction freeing scanner from scanning all memory on
> very large zones (very noticeable for zones > 128GB, for instance) when
> all splits will likely fail.
>
> Signed-off-by: David Rientjes <rientjes@google.com>

Acked-by: Vlastimil Babka <vbabka@suse.cz>

But some notes below.

> ---
>  mm/compaction.c | 49 +++++++++++++++++++++++++++++--------------------
>  1 file changed, 29 insertions(+), 20 deletions(-)
>
> diff --git a/mm/compaction.c b/mm/compaction.c
> --- a/mm/compaction.c
> +++ b/mm/compaction.c
> @@ -494,24 +494,22 @@ static unsigned long isolate_freepages_block(struct compact_control *cc,
>
>  		/* Found a free page, will break it into order-0 pages */
>  		order = page_order(page);
> -		isolated = __isolate_free_page(page, page_order(page));
> +		isolated = __isolate_free_page(page, order);
> +		if (!isolated)
> +			break;

This seems to fix as a side-effect a bug in Joonsoo's mmotm patch 
mm-compaction-split-freepages-without-holding-the-zone-lock.patch, that 
Minchan found: http://marc.info/?l=linux-mm&m=146607176528495&w=2

So it should be noted somewhere so they are merged together. Or Joonsoo 
posts an isolated fix and this patch has to rebase.

>  		set_page_private(page, order);
>  		total_isolated += isolated;
>  		list_add_tail(&page->lru, freelist);
>
> -		/* If a page was split, advance to the end of it */
> -		if (isolated) {
> -			cc->nr_freepages += isolated;
> -			if (!strict &&
> -				cc->nr_migratepages <= cc->nr_freepages) {
> -				blockpfn += isolated;
> -				break;
> -			}
> -
> -			blockpfn += isolated - 1;
> -			cursor += isolated - 1;
> -			continue;
> +		/* Advance to the end of split page */
> +		cc->nr_freepages += isolated;
> +		if (!strict && cc->nr_migratepages <= cc->nr_freepages) {
> +			blockpfn += isolated;
> +			break;
>  		}
> +		blockpfn += isolated - 1;
> +		cursor += isolated - 1;
> +		continue;
>
>  isolate_fail:
>  		if (strict)
> @@ -521,6 +519,9 @@ isolate_fail:
>
>  	}
>
> +	if (locked)
> +		spin_unlock_irqrestore(&cc->zone->lock, flags);
> +
>  	/*
>  	 * There is a tiny chance that we have read bogus compound_order(),
>  	 * so be careful to not go outside of the pageblock.
> @@ -542,9 +543,6 @@ isolate_fail:
>  	if (strict && blockpfn < end_pfn)
>  		total_isolated = 0;
>
> -	if (locked)
> -		spin_unlock_irqrestore(&cc->zone->lock, flags);
> -
>  	/* Update the pageblock-skip if the whole pageblock was scanned */
>  	if (blockpfn == end_pfn)
>  		update_pageblock_skip(cc, valid_page, total_isolated, false);
> @@ -622,7 +620,7 @@ isolate_freepages_range(struct compact_control *cc,
>  		 */
>  	}
>
> -	/* split_free_page does not map the pages */
> +	/* __isolate_free_page() does not map the pages */
>  	map_pages(&freelist);
>
>  	if (pfn < end_pfn) {
> @@ -1071,6 +1069,7 @@ static void isolate_freepages(struct compact_control *cc)
>  				block_end_pfn = block_start_pfn,
>  				block_start_pfn -= pageblock_nr_pages,
>  				isolate_start_pfn = block_start_pfn) {
> +		unsigned long isolated;
>
>  		/*
>  		 * This can iterate a massively long zone without finding any
> @@ -1095,8 +1094,12 @@ static void isolate_freepages(struct compact_control *cc)
>  			continue;
>
>  		/* Found a block suitable for isolating free pages from. */
> -		isolate_freepages_block(cc, &isolate_start_pfn,
> -					block_end_pfn, freelist, false);
> +		isolated = isolate_freepages_block(cc, &isolate_start_pfn,
> +						block_end_pfn, freelist, false);
> +		/* If free page split failed, do not continue needlessly */

More accurately, free page isolation failed?

> +		if (!isolated && isolate_start_pfn < block_end_pfn &&
> +		    cc->nr_freepages <= cc->nr_migratepages)
> +			break;
>
>  		/*
>  		 * If we isolated enough freepages, or aborted due to async
> @@ -1124,7 +1127,7 @@ static void isolate_freepages(struct compact_control *cc)
>  		}
>  	}
>
> -	/* split_free_page does not map the pages */
> +	/* __isolate_free_page() does not map the pages */
>  	map_pages(freelist);
>
>  	/*
> @@ -1703,6 +1706,12 @@ enum compact_result try_to_compact_pages(gfp_t gfp_mask, unsigned int order,
>  			continue;
>  		}
>
> +		/* Don't attempt compaction if splitting free page will fail */
> +		if (!zone_watermark_ok(zone, 0,
> +				       low_wmark_pages(zone) + (1 << order),
> +				       0, 0))
> +			continue;
> +

Please don't add this, compact_zone already checks this via 
compaction_suitable() (and the usual 2 << order gap), so this is adding 
yet another watermark check with a different kind of gap.

Thanks.

>  		status = compact_zone_order(zone, order, gfp_mask, mode,
>  				&zone_contended, alloc_flags,
>  				ac_classzone_idx(ac));
>

Re: [patch] mm, compaction: abort free scanner if split fails

[David Rientjes]

On Tue, 21 Jun 2016, Vlastimil Babka wrote:

> > diff --git a/mm/compaction.c b/mm/compaction.c
> > --- a/mm/compaction.c
> > +++ b/mm/compaction.c
> > @@ -494,24 +494,22 @@ static unsigned long isolate_freepages_block(struct
> > compact_control *cc,
> > 
> >  		/* Found a free page, will break it into order-0 pages */
> >  		order = page_order(page);
> > -		isolated = __isolate_free_page(page, page_order(page));
> > +		isolated = __isolate_free_page(page, order);
> > +		if (!isolated)
> > +			break;
> 
> This seems to fix as a side-effect a bug in Joonsoo's mmotm patch
> mm-compaction-split-freepages-without-holding-the-zone-lock.patch, that
> Minchan found: http://marc.info/?l=linux-mm&m=146607176528495&w=2
> 
> So it should be noted somewhere so they are merged together. Or Joonsoo posts
> an isolated fix and this patch has to rebase.
> 

Indeed, I hadn't noticed the differences between Linus's tree and -mm.  
Thanks very much for pointing it out.

My interest is to eventually backport this to a much older kernel where we 
suffer from the same issue: it seems that we have always not terminated 
the freeing scanner when splitting the free page fails and we feel it 
because some of our systems have 128GB zones and migrate_pages() can call 
compaction_alloc() several times if it keeps getting -EAGAIN.  It's very 
expensive.

I'm not sure we should label it as a -fix for
mm-compaction-split-freepages-without-holding-the-zone-lock.patch since 
the problem this patch is addressing has seemingly existed for years.  
Perhaps it would be better to have two patches, one as a -fix and then the 
abort on page split failure on top.  I'll send out a two patch series in 
this form.

> >  		set_page_private(page, order);
> >  		total_isolated += isolated;
> >  		list_add_tail(&page->lru, freelist);
> > 
> > -		/* If a page was split, advance to the end of it */
> > -		if (isolated) {
> > -			cc->nr_freepages += isolated;
> > -			if (!strict &&
> > -				cc->nr_migratepages <= cc->nr_freepages) {
> > -				blockpfn += isolated;
> > -				break;
> > -			}
> > -
> > -			blockpfn += isolated - 1;
> > -			cursor += isolated - 1;
> > -			continue;
> > +		/* Advance to the end of split page */
> > +		cc->nr_freepages += isolated;
> > +		if (!strict && cc->nr_migratepages <= cc->nr_freepages) {
> > +			blockpfn += isolated;
> > +			break;
> >  		}
> > +		blockpfn += isolated - 1;
> > +		cursor += isolated - 1;
> > +		continue;
> > 
> >  isolate_fail:
> >  		if (strict)
> > @@ -521,6 +519,9 @@ isolate_fail:
> > 
> >  	}
> > 
> > +	if (locked)
> > +		spin_unlock_irqrestore(&cc->zone->lock, flags);
> > +
> >  	/*
> >  	 * There is a tiny chance that we have read bogus compound_order(),
> >  	 * so be careful to not go outside of the pageblock.
> > @@ -542,9 +543,6 @@ isolate_fail:
> >  	if (strict && blockpfn < end_pfn)
> >  		total_isolated = 0;
> > 
> > -	if (locked)
> > -		spin_unlock_irqrestore(&cc->zone->lock, flags);
> > -
> >  	/* Update the pageblock-skip if the whole pageblock was scanned */
> >  	if (blockpfn == end_pfn)
> >  		update_pageblock_skip(cc, valid_page, total_isolated, false);
> > @@ -622,7 +620,7 @@ isolate_freepages_range(struct compact_control *cc,
> >  		 */
> >  	}
> > 
> > -	/* split_free_page does not map the pages */
> > +	/* __isolate_free_page() does not map the pages */
> >  	map_pages(&freelist);
> > 
> >  	if (pfn < end_pfn) {
> > @@ -1071,6 +1069,7 @@ static void isolate_freepages(struct compact_control
> > *cc)
> >  				block_end_pfn = block_start_pfn,
> >  				block_start_pfn -= pageblock_nr_pages,
> >  				isolate_start_pfn = block_start_pfn) {
> > +		unsigned long isolated;
> > 
> >  		/*
> >  		 * This can iterate a massively long zone without finding any
> > @@ -1095,8 +1094,12 @@ static void isolate_freepages(struct compact_control
> > *cc)
> >  			continue;
> > 
> >  		/* Found a block suitable for isolating free pages from. */
> > -		isolate_freepages_block(cc, &isolate_start_pfn,
> > -					block_end_pfn, freelist, false);
> > +		isolated = isolate_freepages_block(cc, &isolate_start_pfn,
> > +						block_end_pfn, freelist,
> > false);
> > +		/* If free page split failed, do not continue needlessly */
> 
> More accurately, free page isolation failed?
> 

Eek, maybe.  The condition should only work if we terminated early because 

 - need_resched() or zone->lock contention for MIGRATE_ASYNC, or

 - __isolate_free_page() fails.

And the latter can only fail because of this (somewhat arbitrary) split 
watermark check.  I'll rename it because it includes both, but I thought 
the next immediate condition check for cc->contended and its comment was 
explanatory enough.

> > +		if (!isolated && isolate_start_pfn < block_end_pfn &&
> > +		    cc->nr_freepages <= cc->nr_migratepages)
> > +			break;
> > 
> >  		/*
> >  		 * If we isolated enough freepages, or aborted due to async
> > @@ -1124,7 +1127,7 @@ static void isolate_freepages(struct compact_control
> > *cc)
> >  		}
> >  	}
> > 
> > -	/* split_free_page does not map the pages */
> > +	/* __isolate_free_page() does not map the pages */
> >  	map_pages(freelist);
> > 
> >  	/*
> > @@ -1703,6 +1706,12 @@ enum compact_result try_to_compact_pages(gfp_t
> > gfp_mask, unsigned int order,
> >  			continue;
> >  		}
> > 
> > +		/* Don't attempt compaction if splitting free page will fail
> > */
> > +		if (!zone_watermark_ok(zone, 0,
> > +				       low_wmark_pages(zone) + (1 << order),
> > +				       0, 0))
> > +			continue;
> > +
> 
> Please don't add this, compact_zone already checks this via
> compaction_suitable() (and the usual 2 << order gap), so this is adding yet
> another watermark check with a different kind of gap.
> 

Good point, thanks.