[f2fs-dev] [PATCH v2 0/3] f2fs: reduce checkpoint write latency under metadata-intensive workloads

[f2fs-dev] [PATCH v2 0/3] f2fs: reduce checkpoint write latency under metadata-intensive workloads

[Yongpeng Yang]

From: Yongpeng Yang <yangyongpeng@xiaomi.com>

This patch series addresses long checkpoint write latency observed under
workloads with frequent metadata operations. Analysis shows that the main
bottleneck is high synchronous read latency of NAT blocks during checkpoint
processing.

The series implements the following improvements:

1. Add flush time stats for NAT and SIT blocks. This stats helps
identify which phase of the checkpoint thread contributes most to the
latency.

2. Readahead all NAT blocks required for the read-modify-write phase
before actual writes. This reduces the synchronous wait time for
individual NAT block reads, significantly lowering checkpoint duration.

Test results show that with NAT block readahead, the total flush time
of NAT blocks is reduced by more than 90% under similar access patterns.

---
v2:
- Replace list_for_each_entry_safe with list_for_each_entry in patch
3/3.

Yongpeng Yang (3):
  f2fs: add write latency stats for NAT and SIT blocks in
    f2fs_write_checkpoint
  f2fs: change size parameter of __has_cursum_space() to unsigned int
  f2fs: optimize NAT block loading during checkpoint write

 fs/f2fs/checkpoint.c |  4 +++-
 fs/f2fs/f2fs.h       |  6 ++++--
 fs/f2fs/node.c       | 13 ++++++++++++-
 3 files changed, 19 insertions(+), 4 deletions(-)

-- 
2.43.0



_______________________________________________
Linux-f2fs-devel mailing list
Linux-f2fs-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/linux-f2fs-devel

[f2fs-dev] [PATCH v2 1/3] f2fs: add write latency stats for NAT and SIT blocks in f2fs_write_checkpoint

[Yongpeng Yang]

From: Yongpeng Yang <yangyongpeng@xiaomi.com>

This patch adds separate write latency accounting for NAT and SIT blocks
in f2fs_write_checkpoint().

Signed-off-by: Yongpeng Yang <yangyongpeng@xiaomi.com>
---
 fs/f2fs/checkpoint.c | 4 +++-
 fs/f2fs/f2fs.h       | 4 +++-
 2 files changed, 6 insertions(+), 2 deletions(-)

diff --git a/fs/f2fs/checkpoint.c b/fs/f2fs/checkpoint.c
index 3dfc83a0813e..7a67e23d0e33 100644
--- a/fs/f2fs/checkpoint.c
+++ b/fs/f2fs/checkpoint.c
@@ -1840,6 +1840,7 @@ int f2fs_write_checkpoint(struct f2fs_sb_info *sbi, struct cp_control *cpc)
 			goto out;
 		}
 	}
+	stat_cp_time(cpc, CP_TIME_MERGE_WRITE);
 
 	/*
 	 * update checkpoint pack index
@@ -1856,10 +1857,11 @@ int f2fs_write_checkpoint(struct f2fs_sb_info *sbi, struct cp_control *cpc)
 		f2fs_bug_on(sbi, !f2fs_cp_error(sbi));
 		goto stop;
 	}
+	stat_cp_time(cpc, CP_TIME_FLUSH_NAT);
 
 	f2fs_flush_sit_entries(sbi, cpc);
 
-	stat_cp_time(cpc, CP_TIME_FLUSH_META);
+	stat_cp_time(cpc, CP_TIME_FLUSH_SIT);
 
 	/* save inmem log status */
 	f2fs_save_inmem_curseg(sbi);
diff --git a/fs/f2fs/f2fs.h b/fs/f2fs/f2fs.h
index ded41b416ed7..bc666cfa83d4 100644
--- a/fs/f2fs/f2fs.h
+++ b/fs/f2fs/f2fs.h
@@ -319,7 +319,9 @@ enum cp_time {
 	CP_TIME_START,		/* begin */
 	CP_TIME_LOCK,		/* after cp_global_sem */
 	CP_TIME_OP_LOCK,	/* after block_operation */
-	CP_TIME_FLUSH_META,	/* after flush sit/nat */
+	CP_TIME_MERGE_WRITE,	/* after flush DATA/NODE/META */
+	CP_TIME_FLUSH_NAT,	/* after flush nat */
+	CP_TIME_FLUSH_SIT,	/* after flush sit */
 	CP_TIME_SYNC_META,	/* after sync_meta_pages */
 	CP_TIME_SYNC_CP_META,	/* after sync cp meta pages */
 	CP_TIME_WAIT_DIRTY_META,/* after wait on dirty meta */
-- 
2.43.0



_______________________________________________
Linux-f2fs-devel mailing list
Linux-f2fs-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/linux-f2fs-devel

Re: [f2fs-dev] [PATCH v2 1/3] f2fs: add write latency stats for NAT and SIT blocks in f2fs_write_checkpoint

[Chao Yu via Linux-f2fs-devel]

On 1/13/2026 11:21 PM, Yongpeng Yang wrote:
> From: Yongpeng Yang <yangyongpeng@xiaomi.com>
> 
> This patch adds separate write latency accounting for NAT and SIT blocks
> in f2fs_write_checkpoint().
> 
> Signed-off-by: Yongpeng Yang <yangyongpeng@xiaomi.com>

Reviewed-by: Chao Yu <chao@kernel.org>

Thanks,


_______________________________________________
Linux-f2fs-devel mailing list
Linux-f2fs-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/linux-f2fs-devel

[f2fs-dev] [PATCH v2 2/3] f2fs: change size parameter of __has_cursum_space() to unsigned int

[Yongpeng Yang]

From: Yongpeng Yang <yangyongpeng@xiaomi.com>

All callers of __has_cursum_space() pass an unsigned int value as the
size parameter. Change the parameter type to unsigned int accordingly.

Signed-off-by: Yongpeng Yang <yangyongpeng@xiaomi.com>
---
 fs/f2fs/f2fs.h | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/fs/f2fs/f2fs.h b/fs/f2fs/f2fs.h
index bc666cfa83d4..1f4698a7b72f 100644
--- a/fs/f2fs/f2fs.h
+++ b/fs/f2fs/f2fs.h
@@ -572,7 +572,7 @@ static inline int update_sits_in_cursum(struct f2fs_journal *journal, int i)
 }
 
 static inline bool __has_cursum_space(struct f2fs_journal *journal,
-							int size, int type)
+						unsigned int size, int type)
 {
 	if (type == NAT_JOURNAL)
 		return size <= MAX_NAT_JENTRIES(journal);
-- 
2.43.0



_______________________________________________
Linux-f2fs-devel mailing list
Linux-f2fs-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/linux-f2fs-devel

Re: [f2fs-dev] [PATCH v2 2/3] f2fs: change size parameter of __has_cursum_space() to unsigned int

[Chao Yu via Linux-f2fs-devel]

On 1/13/2026 11:21 PM, Yongpeng Yang wrote:
> From: Yongpeng Yang <yangyongpeng@xiaomi.com>
> 
> All callers of __has_cursum_space() pass an unsigned int value as the
> size parameter. Change the parameter type to unsigned int accordingly.
> 
> Signed-off-by: Yongpeng Yang <yangyongpeng@xiaomi.com>

Reviewed-by: Chao Yu <chao@kernel.org>

Thanks,


_______________________________________________
Linux-f2fs-devel mailing list
Linux-f2fs-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/linux-f2fs-devel

[f2fs-dev] [PATCH v2 3/3] f2fs: optimize NAT block loading during checkpoint write

[Yongpeng Yang]

From: Yongpeng Yang <yangyongpeng@xiaomi.com>

Under stress tests with frequent metadata operations, checkpoint write
time can become excessively long. Analysis shows that the slowdown is
caused by synchronous, one-by-one reads of NAT blocks during checkpoint
processing.

The issue can be reproduced with the following workload:
1. seq 1 650000 | xargs -P 16 -n 1 touch
2. sync # avoid checkpoint write during deleting
3. delete 1 file every 455 files
4. echo 3 > /proc/sys/vm/drop_caches
5. sync # trigger checkpoint write

This patch submits read I/O for all NAT blocks required in the
__flush_nat_entry_set() phase in advance, reducing the overhead of
synchronous waiting for individual NAT block reads.

The NAT block flush latency before and after the change is as below:

|             |NAT blocks accessed|NAT blocks read|Flush time (ms)|
|-------------|-------------------|---------------|---------------|
|Before change|1205               |1191           |158            |
|After change |1264               |1242           |11             |

With a similar number of NAT blocks accessed and read from disk, adding
NAT block readahead reduces the total NAT block flush time by more than
90%.

Signed-off-by: Yongpeng Yang <yangyongpeng@xiaomi.com>
---
 fs/f2fs/node.c | 13 ++++++++++++-
 1 file changed, 12 insertions(+), 1 deletion(-)

diff --git a/fs/f2fs/node.c b/fs/f2fs/node.c
index 99e425e8c00a..fa1ddfd6633f 100644
--- a/fs/f2fs/node.c
+++ b/fs/f2fs/node.c
@@ -3164,7 +3164,7 @@ int f2fs_flush_nat_entries(struct f2fs_sb_info *sbi, struct cp_control *cpc)
 	struct f2fs_journal *journal = curseg->journal;
 	struct nat_entry_set *setvec[NAT_VEC_SIZE];
 	struct nat_entry_set *set, *tmp;
-	unsigned int found;
+	unsigned int found, entry_count = 0;
 	nid_t set_idx = 0;
 	LIST_HEAD(sets);
 	int err = 0;
@@ -3204,6 +3204,17 @@ int f2fs_flush_nat_entries(struct f2fs_sb_info *sbi, struct cp_control *cpc)
 						MAX_NAT_JENTRIES(journal));
 	}
 
+	/*
+	 * Readahead the current NAT block to prevent read requests from
+	 * being issued and waited on one by one.
+	 */
+	list_for_each_entry(set, &sets, set_list) {
+		entry_count += set->entry_cnt;
+		if (!enabled_nat_bits(sbi, cpc) &&
+			__has_cursum_space(journal, entry_count, NAT_JOURNAL))
+			continue;
+		f2fs_ra_meta_pages(sbi, set->set, 1, META_NAT, true);
+	}
 	/* flush dirty nats in nat entry set */
 	list_for_each_entry_safe(set, tmp, &sets, set_list) {
 		err = __flush_nat_entry_set(sbi, set, cpc);
-- 
2.43.0



_______________________________________________
Linux-f2fs-devel mailing list
Linux-f2fs-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/linux-f2fs-devel

Re: [f2fs-dev] [PATCH v2 3/3] f2fs: optimize NAT block loading during checkpoint write

[Chao Yu via Linux-f2fs-devel]

On 1/13/2026 11:23 PM, Yongpeng Yang wrote:
> From: Yongpeng Yang <yangyongpeng@xiaomi.com>
> 
> Under stress tests with frequent metadata operations, checkpoint write
> time can become excessively long. Analysis shows that the slowdown is
> caused by synchronous, one-by-one reads of NAT blocks during checkpoint
> processing.
> 
> The issue can be reproduced with the following workload:
> 1. seq 1 650000 | xargs -P 16 -n 1 touch
> 2. sync # avoid checkpoint write during deleting
> 3. delete 1 file every 455 files
> 4. echo 3 > /proc/sys/vm/drop_caches
> 5. sync # trigger checkpoint write
> 
> This patch submits read I/O for all NAT blocks required in the
> __flush_nat_entry_set() phase in advance, reducing the overhead of
> synchronous waiting for individual NAT block reads.
> 
> The NAT block flush latency before and after the change is as below:
> 
> |             |NAT blocks accessed|NAT blocks read|Flush time (ms)|
> |-------------|-------------------|---------------|---------------|
> |Before change|1205               |1191           |158            |
> |After change |1264               |1242           |11             |
> 
> With a similar number of NAT blocks accessed and read from disk, adding
> NAT block readahead reduces the total NAT block flush time by more than
> 90%.
> 
> Signed-off-by: Yongpeng Yang <yangyongpeng@xiaomi.com>

Reviewed-by: Chao Yu <chao@kernel.org>

Thanks,


_______________________________________________
Linux-f2fs-devel mailing list
Linux-f2fs-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/linux-f2fs-devel

Re: [f2fs-dev] [PATCH v2 0/3] f2fs: reduce checkpoint write latency under metadata-intensive workloads

[patchwork-bot+f2fs--- via Linux-f2fs-devel]

Hello:

This series was applied to jaegeuk/f2fs.git (dev)
by Jaegeuk Kim <jaegeuk@kernel.org>:

On Tue, 13 Jan 2026 23:21:36 +0800 you wrote:
> From: Yongpeng Yang <yangyongpeng@xiaomi.com>
> 
> This patch series addresses long checkpoint write latency observed under
> workloads with frequent metadata operations. Analysis shows that the main
> bottleneck is high synchronous read latency of NAT blocks during checkpoint
> processing.
> 
> [...]

Here is the summary with links:
  - [f2fs-dev,v2,1/3] f2fs: add write latency stats for NAT and SIT blocks in f2fs_write_checkpoint
    https://git.kernel.org/jaegeuk/f2fs/c/10f96e94b38d
  - [f2fs-dev,v2,2/3] f2fs: change size parameter of __has_cursum_space() to unsigned int
    (no matching commit)
  - [f2fs-dev,v2,3/3] f2fs: optimize NAT block loading during checkpoint write
    (no matching commit)

You are awesome, thank you!
-- 
Deet-doot-dot, I am a bot.
https://korg.docs.kernel.org/patchwork/pwbot.html




_______________________________________________
Linux-f2fs-devel mailing list
Linux-f2fs-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/linux-f2fs-devel