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

Groups > linux.kernel > #1519786 > unrolled thread

[PATCH] z3fold: discourage use of pages that weren't compacted

Back to article view | Back to linux.kernel

Contents

  [PATCH] z3fold: discourage use of pages that weren't compacted Vitaly Wool <vitalywool@gmail.com> - 2016-11-11 14:20 +0100
    Re: [PATCH] z3fold: discourage use of pages that weren't compacted Andrew Morton <akpm@linux-foundation.org> - 2016-11-15 01:40 +0100
      Re: [PATCH] z3fold: discourage use of pages that weren't compacted Vitaly Wool <vitalywool@gmail.com> - 2016-11-15 14:40 +0100

#1519786 — [PATCH] z3fold: discourage use of pages that weren't compacted

If a z3fold page couldn't be compacted, we don't want it to be
used for next object allocation in the first place. It makes more
sense to add it to the end of the relevant unbuddied list. If that
page gets compacted later, it will be added to the beginning of
the list then.

This simple idea gives 5-7% improvement in randrw fio tests and
about 10% improvement in fio sequential read/write.

Signed-off-by: Vitaly Wool <vitalywool@gmail.com>
---
 mm/z3fold.c | 32 ++++++++++++++++++++++----------
 1 file changed, 22 insertions(+), 10 deletions(-)

diff --git a/mm/z3fold.c b/mm/z3fold.c
index 5fe2652..eb8f9a0 100644
--- a/mm/z3fold.c
+++ b/mm/z3fold.c
@@ -277,10 +277,10 @@ static inline void *mchunk_memmove(struct z3fold_header *zhdr,
 
 #define BIG_CHUNK_GAP	3
 /* Has to be called with lock held */
-static int z3fold_compact_page(struct z3fold_header *zhdr)
+static bool z3fold_compact_page(struct z3fold_header *zhdr)
 {
 	struct page *page = virt_to_page(zhdr);
-	int ret = 0;
+	bool ret = false;
 
 	if (test_bit(MIDDLE_CHUNK_MAPPED, &page->private))
 		goto out;
@@ -292,7 +292,7 @@ static int z3fold_compact_page(struct z3fold_header *zhdr)
 			zhdr->middle_chunks = 0;
 			zhdr->start_middle = 0;
 			zhdr->first_num++;
-			ret = 1;
+			ret = true;
 			goto out;
 		}
 
@@ -304,7 +304,7 @@ static int z3fold_compact_page(struct z3fold_header *zhdr)
 		    zhdr->start_middle > zhdr->first_chunks + BIG_CHUNK_GAP) {
 			mchunk_memmove(zhdr, zhdr->first_chunks + 1);
 			zhdr->start_middle = zhdr->first_chunks + 1;
-			ret = 1;
+			ret = true;
 			goto out;
 		}
 		if (zhdr->last_chunks != 0 && zhdr->first_chunks == 0 &&
@@ -314,7 +314,7 @@ static int z3fold_compact_page(struct z3fold_header *zhdr)
 				zhdr->middle_chunks;
 			mchunk_memmove(zhdr, new_start);
 			zhdr->start_middle = new_start;
-			ret = 1;
+			ret = true;
 			goto out;
 		}
 	}
@@ -535,11 +535,19 @@ static void z3fold_free(struct z3fold_pool *pool, unsigned long handle)
 		free_z3fold_page(zhdr);
 		atomic64_dec(&pool->pages_nr);
 	} else {
-		z3fold_compact_page(zhdr);
+		bool compacted = z3fold_compact_page(zhdr);
 		/* Add to the unbuddied list */
 		spin_lock(&pool->lock);
 		freechunks = num_free_chunks(zhdr);
-		list_add(&zhdr->buddy, &pool->unbuddied[freechunks]);
+		/*
+		 * If the page has been compacted, we want to use it
+		 * in the first place.
+		 */
+		if (compacted)
+			list_add(&zhdr->buddy, &pool->unbuddied[freechunks]);
+		else
+			list_add_tail(&zhdr->buddy,
+				      &pool->unbuddied[freechunks]);
 		spin_unlock(&pool->lock);
 		z3fold_page_unlock(zhdr);
 	}
@@ -668,12 +676,16 @@ static int z3fold_reclaim_page(struct z3fold_pool *pool, unsigned int retries)
 				spin_lock(&pool->lock);
 				list_add(&zhdr->buddy, &pool->buddied);
 			} else {
-				z3fold_compact_page(zhdr);
+				bool compacted = z3fold_compact_page(zhdr);
 				/* add to unbuddied list */
 				spin_lock(&pool->lock);
 				freechunks = num_free_chunks(zhdr);
-				list_add(&zhdr->buddy,
-					 &pool->unbuddied[freechunks]);
+				if (compacted)
+					list_add(&zhdr->buddy,
+						&pool->unbuddied[freechunks]);
+				else
+					list_add_tail(&zhdr->buddy,
+						&pool->unbuddied[freechunks]);
 			}
 		}
 
-- 
2.4.2

[toc] | [next] | [standalone]

#1522208

On Fri, 11 Nov 2016 14:02:07 +0100 Vitaly Wool <vitalywool@gmail.com> wrote:

> If a z3fold page couldn't be compacted, we don't want it to be
> used for next object allocation in the first place. It makes more
> sense to add it to the end of the relevant unbuddied list. If that
> page gets compacted later, it will be added to the beginning of
> the list then.
> 
> This simple idea gives 5-7% improvement in randrw fio tests and
> about 10% improvement in fio sequential read/write.

This patch appears to require "z3fold: use per-page spinlock", and
"z3fold: use per-page spinlock" doesn't apply properly.

So things are in a bit of a mess.

I presently have

z3fold-limit-first_num-to-the-actual-range-of-possible-buddy-indexes.patch
z3fold-make-pages_nr-atomic.patch
z3fold-extend-compaction-function.patch

Please take a look, figure out what we should do.  Perhaps do it all as
a coherent series rather than an interdependent dribble?

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

#1522726

On Tue, Nov 15, 2016 at 1:33 AM, Andrew Morton
<akpm@linux-foundation.org> wrote:
> On Fri, 11 Nov 2016 14:02:07 +0100 Vitaly Wool <vitalywool@gmail.com> wrote:
>
>> If a z3fold page couldn't be compacted, we don't want it to be
>> used for next object allocation in the first place. It makes more
>> sense to add it to the end of the relevant unbuddied list. If that
>> page gets compacted later, it will be added to the beginning of
>> the list then.
>>
>> This simple idea gives 5-7% improvement in randrw fio tests and
>> about 10% improvement in fio sequential read/write.
>
> This patch appears to require "z3fold: use per-page spinlock", and
> "z3fold: use per-page spinlock" doesn't apply properly.
>
> So things are in a bit of a mess.

Yep, sorry about that.

> I presently have
>
> z3fold-limit-first_num-to-the-actual-range-of-possible-buddy-indexes.patch
> z3fold-make-pages_nr-atomic.patch
> z3fold-extend-compaction-function.patch

These are not interdependent.

> Please take a look, figure out what we should do.  Perhaps do it all as
> a coherent series rather than an interdependent dribble?

I'll come up with a 2-patch update in a short while.

~vitaly

[toc] | [prev] | [standalone]

Back to top | Article view | linux.kernel

csiph-web