Linux bcache driver list
 help / color / mirror / Atom feed
From: mingzhe.zou@easystack.cn
To: colyli@suse.de
Cc: linux-bcache@vger.kernel.org, dongsheng.yang@easystack.cn,
	zoumingzhe@qq.com
Subject: [PATCH 1/3] bcache: avoid invalidating buckets in use
Date: Tue, 19 Nov 2024 11:28:50 +0800	[thread overview]
Message-ID: <20241119032852.2511-1-mingzhe.zou@easystack.cn> (raw)

From: Mingzhe Zou <mingzhe.zou@easystack.cn>

If the bucket was reused while our bio was in flight, we might
have read the wrong data. Currently, we will reread the data from
the backing device. This not only reduces performance, but also
makes the process more complex.

When the bucket is in use, we hope not to reclaim it.

Signed-off-by: Mingzhe Zou <mingzhe.zou@easystack.cn>
---
 drivers/md/bcache/alloc.c  | 30 +++++++++++++++++++++---------
 drivers/md/bcache/bcache.h |  3 ++-
 2 files changed, 23 insertions(+), 10 deletions(-)

diff --git a/drivers/md/bcache/alloc.c b/drivers/md/bcache/alloc.c
index da50f6661bae..32f65d6fc906 100644
--- a/drivers/md/bcache/alloc.c
+++ b/drivers/md/bcache/alloc.c
@@ -134,25 +134,39 @@ bool bch_can_invalidate_bucket(struct cache *ca, struct bucket *b)
 	       !atomic_read(&b->pin) && can_inc_bucket_gen(b));
 }
 
-void __bch_invalidate_one_bucket(struct cache *ca, struct bucket *b)
+bool __bch_invalidate_one_bucket(struct cache *ca, struct bucket *b)
 {
 	lockdep_assert_held(&ca->set->bucket_lock);
 	BUG_ON(GC_MARK(b) && GC_MARK(b) != GC_MARK_RECLAIMABLE);
 
+	/*
+	 * If the bucket was reused while read bio was in flight, it will
+	 * reread the data from the backing device. This will increase latency
+	 * and cause other errors. When b->pin is not 0, do not invalidate
+	 * the bucket.
+	 */
+
+	b->invalidating = 1;
+
+	if (atomic_inc_return(&b->pin) > 1) {
+		atomic_dec(&b->pin);
+		return false;
+	}
+
 	if (GC_SECTORS_USED(b))
 		trace_bcache_invalidate(ca, b - ca->buckets);
 
 	bch_inc_gen(ca, b);
 	b->prio = INITIAL_PRIO;
-	atomic_inc(&b->pin);
 	b->reclaimable_in_gc = 0;
+	b->invalidating = 0;
+	return true;
 }
 
 static void bch_invalidate_one_bucket(struct cache *ca, struct bucket *b)
 {
-	__bch_invalidate_one_bucket(ca, b);
-
-	fifo_push(&ca->free_inc, b - ca->buckets);
+	if (bch_can_invalidate_bucket(ca, b) && __bch_invalidate_one_bucket(ca, b))
+		fifo_push(&ca->free_inc, b - ca->buckets);
 }
 
 /*
@@ -253,8 +267,7 @@ static void invalidate_buckets_fifo(struct cache *ca)
 
 		b = ca->buckets + ca->fifo_last_bucket++;
 
-		if (bch_can_invalidate_bucket(ca, b))
-			bch_invalidate_one_bucket(ca, b);
+		bch_invalidate_one_bucket(ca, b);
 
 		if (++checked >= ca->sb.nbuckets) {
 			ca->invalidate_needs_gc = 1;
@@ -279,8 +292,7 @@ static void invalidate_buckets_random(struct cache *ca)
 
 		b = ca->buckets + n;
 
-		if (bch_can_invalidate_bucket(ca, b))
-			bch_invalidate_one_bucket(ca, b);
+		bch_invalidate_one_bucket(ca, b);
 
 		if (++checked >= ca->sb.nbuckets / 2) {
 			ca->invalidate_needs_gc = 1;
diff --git a/drivers/md/bcache/bcache.h b/drivers/md/bcache/bcache.h
index 785b0d9008fa..2777d72e1038 100644
--- a/drivers/md/bcache/bcache.h
+++ b/drivers/md/bcache/bcache.h
@@ -201,6 +201,7 @@ struct bucket {
 	uint8_t		last_gc; /* Most out of date gen in the btree */
 	uint16_t	gc_mark; /* Bitfield used by GC. See below for field */
 	uint16_t	reclaimable_in_gc:1;
+	uint16_t	invalidating:1;
 };
 
 /*
@@ -981,7 +982,7 @@ uint8_t bch_inc_gen(struct cache *ca, struct bucket *b);
 void bch_rescale_priorities(struct cache_set *c, int sectors);
 
 bool bch_can_invalidate_bucket(struct cache *ca, struct bucket *b);
-void __bch_invalidate_one_bucket(struct cache *ca, struct bucket *b);
+bool __bch_invalidate_one_bucket(struct cache *ca, struct bucket *b);
 
 void __bch_bucket_free(struct cache *ca, struct bucket *b);
 void bch_bucket_free(struct cache_set *c, struct bkey *k);
-- 
2.34.1


             reply	other threads:[~2024-11-19  9:12 UTC|newest]

Thread overview: 7+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2024-11-19  3:28 mingzhe.zou [this message]
2024-11-19  3:28 ` [PATCH 2/3] bcache: fix io error during cache read race mingzhe.zou
2024-11-22  0:21   ` kernel test robot
2024-11-22  0:21   ` kernel test robot
2024-11-19  3:28 ` [PATCH 3/3] bcache: remove unused parameters mingzhe.zou
2024-11-22  1:46   ` kernel test robot
  -- strict thread matches above, loose matches on Subject: below --
2024-11-18  9:26 [PATCH 1/3] bcache: avoid invalidating buckets in use mingzhe.zou

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20241119032852.2511-1-mingzhe.zou@easystack.cn \
    --to=mingzhe.zou@easystack.cn \
    --cc=colyli@suse.de \
    --cc=dongsheng.yang@easystack.cn \
    --cc=linux-bcache@vger.kernel.org \
    --cc=zoumingzhe@qq.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox