[PATCH 1/2] f2fs: improve write performance under frequent fsync calls

[PATCH 1/2] f2fs: improve write performance under frequent fsync calls

[Jaegeuk Kim]

When considering a bunch of data writes with very frequent fsync calls, we
are able to think the following performance regression.

N: Node IO, D: Data IO, IO scheduler: cfq

Issue    pending IOs
	 D1 D2 D3 D4
 D1         D2 D3 D4 N1
 D2            D3 D4 N1 N2
 N1            D3 D4 N2 D1
 --> N1 can be selected by cfq becase of the same priority of N and D.
     Then D3 and D4 would be delayed, resuling in performance degradation.

So, when processing the fsync call, it'd better give higher priority to data IOs
than node IOs by assigning WRITE and WRITE_SYNC respectively.
This patch improves the random wirte performance with frequent fsync calls by up
to 10%.

Signed-off-by: Jaegeuk Kim <jaegeuk.kim@samsung.com>
---
 fs/f2fs/f2fs.h    | 4 ++--
 fs/f2fs/file.c    | 2 +-
 fs/f2fs/node.c    | 7 ++++++-
 fs/f2fs/segment.c | 8 ++------
 4 files changed, 11 insertions(+), 10 deletions(-)

diff --git a/fs/f2fs/f2fs.h b/fs/f2fs/f2fs.h
index 07a7ae0..b69f190 100644
--- a/fs/f2fs/f2fs.h
+++ b/fs/f2fs/f2fs.h
@@ -1118,8 +1118,8 @@ int npages_for_summary_flush(struct f2fs_sb_info *);
 void allocate_new_segments(struct f2fs_sb_info *);
 struct page *get_sum_page(struct f2fs_sb_info *, unsigned int);
 void write_meta_page(struct f2fs_sb_info *, struct page *);
-void write_node_page(struct f2fs_sb_info *, struct page *, unsigned int,
-					block_t, block_t *);
+void write_node_page(struct f2fs_sb_info *, struct page *,
+		struct f2fs_io_info *, unsigned int, block_t, block_t *);
 void write_data_page(struct page *, struct dnode_of_data *, block_t *,
 					struct f2fs_io_info *);
 void rewrite_data_page(struct page *, block_t, struct f2fs_io_info *);
diff --git a/fs/f2fs/file.c b/fs/f2fs/file.c
index c77ad4d..14511b0 100644
--- a/fs/f2fs/file.c
+++ b/fs/f2fs/file.c
@@ -115,7 +115,7 @@ int f2fs_sync_file(struct file *file, loff_t start, loff_t end, int datasync)
 	int ret = 0;
 	bool need_cp = false;
 	struct writeback_control wbc = {
-		.sync_mode = WB_SYNC_ALL,
+		.sync_mode = WB_SYNC_NONE,
 		.nr_to_write = LONG_MAX,
 		.for_reclaim = 0,
 	};
diff --git a/fs/f2fs/node.c b/fs/f2fs/node.c
index 0230326..b8c9301 100644
--- a/fs/f2fs/node.c
+++ b/fs/f2fs/node.c
@@ -1194,6 +1194,10 @@ static int f2fs_write_node_page(struct page *page,
 	nid_t nid;
 	block_t new_addr;
 	struct node_info ni;
+	struct f2fs_io_info fio = {
+		.type = NODE,
+		.rw = (wbc->sync_mode == WB_SYNC_ALL) ? WRITE_SYNC: WRITE,
+	};
 
 	if (unlikely(sbi->por_doing))
 		goto redirty_out;
@@ -1218,7 +1222,7 @@ static int f2fs_write_node_page(struct page *page,
 
 	mutex_lock(&sbi->node_write);
 	set_page_writeback(page);
-	write_node_page(sbi, page, nid, ni.blk_addr, &new_addr);
+	write_node_page(sbi, page, &fio, nid, ni.blk_addr, &new_addr);
 	set_node_addr(sbi, &ni, new_addr);
 	dec_page_count(sbi, F2FS_DIRTY_NODES);
 	mutex_unlock(&sbi->node_write);
@@ -1253,6 +1257,7 @@ static int f2fs_write_node_pages(struct address_space *mapping,
 
 	/* if mounting is failed, skip writing node pages */
 	wbc->nr_to_write = 3 * max_hw_blocks(sbi);
+	wbc->sync_mode = WB_SYNC_NONE;
 	sync_node_pages(sbi, 0, wbc);
 	wbc->nr_to_write = nr_to_write - (3 * max_hw_blocks(sbi) -
 						wbc->nr_to_write);
diff --git a/fs/f2fs/segment.c b/fs/f2fs/segment.c
index 555ae76..5f84639 100644
--- a/fs/f2fs/segment.c
+++ b/fs/f2fs/segment.c
@@ -924,16 +924,12 @@ void write_meta_page(struct f2fs_sb_info *sbi, struct page *page)
 }
 
 void write_node_page(struct f2fs_sb_info *sbi, struct page *page,
+		struct f2fs_io_info *fio,
 		unsigned int nid, block_t old_blkaddr, block_t *new_blkaddr)
 {
 	struct f2fs_summary sum;
-	struct f2fs_io_info fio = {
-		.type = NODE,
-		.rw = WRITE_SYNC,
-	};
-
 	set_summary(&sum, nid, 0, 0);
-	do_write_page(sbi, page, old_blkaddr, new_blkaddr, &sum, &fio);
+	do_write_page(sbi, page, old_blkaddr, new_blkaddr, &sum, fio);
 }
 
 void write_data_page(struct page *page, struct dnode_of_data *dn,
-- 
1.8.4.474.g128a96c

[PATCH 2/2] f2fs: add a sysfs entry to control max_victim_search

[Jaegeuk Kim]

Previously during SSR and GC, the maximum number of retrials to find a victim
segment was hard-coded by MAX_VICTIM_SEARCH, 4096 by default.

This number makes an effect on IO locality, when SSR mode is activated, which
results in performance fluctuation on some low-end devices.

If max_victim_search = 4, the victim will be searched like below.
("D" represents a dirty segment, and "*" indicates a selected victim segment.)

 D1 D2 D3 D4 D5 D6 D7 D8 D9
[   *       ]
      [   *    ]
            [         * ]
	                [ ....]

This patch adds a sysfs entry to control the number dynamically through:
  /sys/fs/f2fs/$dev/max_victim_search

Signed-off-by: Jaegeuk Kim <jaegeuk.kim@samsung.com>
---
 fs/f2fs/f2fs.h  | 3 +++
 fs/f2fs/gc.c    | 4 ++--
 fs/f2fs/gc.h    | 2 +-
 fs/f2fs/super.c | 6 ++++++
 4 files changed, 12 insertions(+), 3 deletions(-)

diff --git a/fs/f2fs/f2fs.h b/fs/f2fs/f2fs.h
index b69f190..5f7dc4f 100644
--- a/fs/f2fs/f2fs.h
+++ b/fs/f2fs/f2fs.h
@@ -451,6 +451,9 @@ struct f2fs_sb_info {
 	struct f2fs_gc_kthread	*gc_thread;	/* GC thread */
 	unsigned int cur_victim_sec;		/* current victim section num */
 
+	/* maximum # of trials to find a victim segment for SSR and GC */
+	unsigned int max_victim_search;
+
 	/*
 	 * for stat information.
 	 * one is for the LFS mode, and the other is for the SSR mode.
diff --git a/fs/f2fs/gc.c b/fs/f2fs/gc.c
index 599f546..9117cca 100644
--- a/fs/f2fs/gc.c
+++ b/fs/f2fs/gc.c
@@ -163,8 +163,8 @@ static void select_policy(struct f2fs_sb_info *sbi, int gc_type,
 		p->ofs_unit = sbi->segs_per_sec;
 	}
 
-	if (p->max_search > MAX_VICTIM_SEARCH)
-		p->max_search = MAX_VICTIM_SEARCH;
+	if (p->max_search > sbi->max_victim_search)
+		p->max_search = sbi->max_victim_search;
 
 	p->offset = sbi->last_victim[p->gc_mode];
 }
diff --git a/fs/f2fs/gc.h b/fs/f2fs/gc.h
index 507056d..5d5eb60 100644
--- a/fs/f2fs/gc.h
+++ b/fs/f2fs/gc.h
@@ -20,7 +20,7 @@
 #define LIMIT_FREE_BLOCK	40 /* percentage over invalid + free space */
 
 /* Search max. number of dirty segments to select a victim segment */
-#define MAX_VICTIM_SEARCH 4096 /* covers 8GB */
+#define DEF_MAX_VICTIM_SEARCH 4096 /* covers 8GB */
 
 struct f2fs_gc_kthread {
 	struct task_struct *f2fs_gc_task;
diff --git a/fs/f2fs/super.c b/fs/f2fs/super.c
index f16da92..b070f30 100644
--- a/fs/f2fs/super.c
+++ b/fs/f2fs/super.c
@@ -74,6 +74,7 @@ static match_table_t f2fs_tokens = {
 enum {
 	GC_THREAD,	/* struct f2fs_gc_thread */
 	SM_INFO,	/* struct f2fs_sm_info */
+	F2FS_SBI,	/* struct f2fs_sb_info */
 };
 
 struct f2fs_attr {
@@ -91,6 +92,8 @@ static unsigned char *__struct_ptr(struct f2fs_sb_info *sbi, int struct_type)
 		return (unsigned char *)sbi->gc_thread;
 	else if (struct_type == SM_INFO)
 		return (unsigned char *)SM_I(sbi);
+	else if (struct_type == F2FS_SBI)
+		return (unsigned char *)sbi;
 	return NULL;
 }
 
@@ -180,6 +183,7 @@ F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, reclaim_segments, rec_prefree_segments);
 F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, max_small_discards, max_discards);
 F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, ipu_policy, ipu_policy);
 F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, min_ipu_util, min_ipu_util);
+F2FS_RW_ATTR(F2FS_SBI, f2fs_sb_info, max_victim_search, max_victim_search);
 
 #define ATTR_LIST(name) (&f2fs_attr_##name.attr)
 static struct attribute *f2fs_attrs[] = {
@@ -191,6 +195,7 @@ static struct attribute *f2fs_attrs[] = {
 	ATTR_LIST(max_small_discards),
 	ATTR_LIST(ipu_policy),
 	ATTR_LIST(min_ipu_util),
+	ATTR_LIST(max_victim_search),
 	NULL,
 };
 
@@ -775,6 +780,7 @@ static void init_sb_info(struct f2fs_sb_info *sbi)
 	sbi->node_ino_num = le32_to_cpu(raw_super->node_ino);
 	sbi->meta_ino_num = le32_to_cpu(raw_super->meta_ino);
 	sbi->cur_victim_sec = NULL_SECNO;
+	sbi->max_victim_search = DEF_MAX_VICTIM_SEARCH;
 
 	for (i = 0; i < NR_COUNT_TYPE; i++)
 		atomic_set(&sbi->nr_pages[i], 0);
-- 
1.8.4.474.g128a96c