f2fs: adjust readahead block number during recovery

In a fragmented image, entries in dnode block list may locate in
incontiguous physical block address space, however, in recovery flow,
we will always readahead BIO_MAX_VECS size blocks, so in such case,
current readahead policy is low efficient, let's adjust readahead
window size dynamically based on consecutiveness of dnode blocks.

Signed-off-by: Chao Yu <chao@kernel.org>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>

3 files changed, 35 insertions, 6 deletions

diff --git a/fs/f2fs/checkpoint.c b/fs/f2fs/checkpoint.c
index deeda95688f0..a13b6b4af220 100644
--- a/fs/f2fs/checkpoint.c
+++ b/fs/f2fs/checkpoint.c
@@ -282,18 +282,22 @@ out:
 	return blkno - start;
 }
 
-void f2fs_ra_meta_pages_cond(struct f2fs_sb_info *sbi, pgoff_t index)
+void f2fs_ra_meta_pages_cond(struct f2fs_sb_info *sbi, pgoff_t index,
+							unsigned int ra_blocks)
 {
 	struct page *page;
 	bool readahead = false;
 
+	if (ra_blocks == RECOVERY_MIN_RA_BLOCKS)
+		return;
+
 	page = find_get_page(META_MAPPING(sbi), index);
 	if (!page || !PageUptodate(page))
 		readahead = true;
 	f2fs_put_page(page, 0);
 
 	if (readahead)
-		f2fs_ra_meta_pages(sbi, index, BIO_MAX_VECS, META_POR, true);
+		f2fs_ra_meta_pages(sbi, index, ra_blocks, META_POR, true);
 }
 
 static int __f2fs_write_meta_page(struct page *page,
diff --git a/fs/f2fs/f2fs.h b/fs/f2fs/f2fs.h
index 63c90416364b..51c1392708e6 100644
--- a/fs/f2fs/f2fs.h
+++ b/fs/f2fs/f2fs.h
@@ -590,6 +590,9 @@ enum {
 /* number of extent info in extent cache we try to shrink */
 #define EXTENT_CACHE_SHRINK_NUMBER	128
 
+#define RECOVERY_MAX_RA_BLOCKS		BIO_MAX_VECS
+#define RECOVERY_MIN_RA_BLOCKS		1
+
 struct rb_entry {
 	struct rb_node rb_node;		/* rb node located in rb-tree */
 	union {
@@ -3651,7 +3654,8 @@ bool f2fs_is_valid_blkaddr(struct f2fs_sb_info *sbi,
 					block_t blkaddr, int type);
 int f2fs_ra_meta_pages(struct f2fs_sb_info *sbi, block_t start, int nrpages,
 			int type, bool sync);
-void f2fs_ra_meta_pages_cond(struct f2fs_sb_info *sbi, pgoff_t index);
+void f2fs_ra_meta_pages_cond(struct f2fs_sb_info *sbi, pgoff_t index,
+							unsigned int ra_blocks);
 long f2fs_sync_meta_pages(struct f2fs_sb_info *sbi, enum page_type type,
 			long nr_to_write, enum iostat_type io_type);
 void f2fs_add_ino_entry(struct f2fs_sb_info *sbi, nid_t ino, int type);
diff --git a/fs/f2fs/recovery.c b/fs/f2fs/recovery.c
index 10d152cfa58d..2af503f75b4f 100644
--- a/fs/f2fs/recovery.c
+++ b/fs/f2fs/recovery.c
@@ -343,6 +343,19 @@ static int recover_inode(struct inode *inode, struct page *page)
 	return 0;
 }
 
+static unsigned int adjust_por_ra_blocks(struct f2fs_sb_info *sbi,
+				unsigned int ra_blocks, unsigned int blkaddr,
+				unsigned int next_blkaddr)
+{
+	if (blkaddr + 1 == next_blkaddr)
+		ra_blocks = min_t(unsigned int, RECOVERY_MAX_RA_BLOCKS,
+							ra_blocks * 2);
+	else if (next_blkaddr % sbi->blocks_per_seg)
+		ra_blocks = max_t(unsigned int, RECOVERY_MIN_RA_BLOCKS,
+							ra_blocks / 2);
+	return ra_blocks;
+}
+
 static int find_fsync_dnodes(struct f2fs_sb_info *sbi, struct list_head *head,
 				bool check_only)
 {
@@ -350,6 +363,7 @@ static int find_fsync_dnodes(struct f2fs_sb_info *sbi, struct list_head *head,
 	struct page *page = NULL;
 	block_t blkaddr;
 	unsigned int loop_cnt = 0;
+	unsigned int ra_blocks = RECOVERY_MAX_RA_BLOCKS;
 	unsigned int free_blocks = MAIN_SEGS(sbi) * sbi->blocks_per_seg -
 						valid_user_blocks(sbi);
 	int err = 0;
@@ -424,11 +438,14 @@ next:
 			break;
 		}
 
+		ra_blocks = adjust_por_ra_blocks(sbi, ra_blocks, blkaddr,
+						next_blkaddr_of_node(page));
+
 		/* check next segment */
 		blkaddr = next_blkaddr_of_node(page);
 		f2fs_put_page(page, 1);
 
-		f2fs_ra_meta_pages_cond(sbi, blkaddr);
+		f2fs_ra_meta_pages_cond(sbi, blkaddr, ra_blocks);
 	}
 	return err;
 }
@@ -704,6 +721,7 @@ static int recover_data(struct f2fs_sb_info *sbi, struct list_head *inode_list,
 	struct page *page = NULL;
 	int err = 0;
 	block_t blkaddr;
+	unsigned int ra_blocks = RECOVERY_MAX_RA_BLOCKS;
 
 	/* get node pages in the current segment */
 	curseg = CURSEG_I(sbi, CURSEG_WARM_NODE);
@@ -715,8 +733,6 @@ static int recover_data(struct f2fs_sb_info *sbi, struct list_head *inode_list,
 		if (!f2fs_is_valid_blkaddr(sbi, blkaddr, META_POR))
 			break;
 
-		f2fs_ra_meta_pages_cond(sbi, blkaddr);
-
 		page = f2fs_get_tmp_page(sbi, blkaddr);
 		if (IS_ERR(page)) {
 			err = PTR_ERR(page);
@@ -759,9 +775,14 @@ static int recover_data(struct f2fs_sb_info *sbi, struct list_head *inode_list,
 		if (entry->blkaddr == blkaddr)
 			list_move_tail(&entry->list, tmp_inode_list);
 next:
+		ra_blocks = adjust_por_ra_blocks(sbi, ra_blocks, blkaddr,
+						next_blkaddr_of_node(page));
+
 		/* check next segment */
 		blkaddr = next_blkaddr_of_node(page);
 		f2fs_put_page(page, 1);
+
+		f2fs_ra_meta_pages_cond(sbi, blkaddr, ra_blocks);
 	}
 	if (!err)
 		f2fs_allocate_new_segments(sbi);