[PATCH] f2fs: avoid readahead race condition

[Jaegeuk Kim]

If two readahead threads having same offset enter in readpages, every read
IOs are split and issued to the disk which giving lower bandwidth.

This patch tries to avoid redundant readahead calls.

Fixes one build error reported by Randy.
Fix build error when F2FS_FS_COMPRESSION is not set/enabled.
This label is needed in either case.

../fs/f2fs/data.c: In function âf2fs_mpage_readpagesâ:
../fs/f2fs/data.c:2327:5: error: label ânext_pageâ used but not defined
     goto next_page;

Reviewed-by: Chao Yu <yuchao0@xxxxxxxxxx>
Signed-off-by: Jaegeuk Kim <jaegeuk@xxxxxxxxxx>
---
 v4:
  - fix missing unlock_page
 v3:
  - use READ|WRITE_ONCE
 v2:
   - add missing code to bypass read

 fs/f2fs/data.c  | 20 ++++++++++++++++++++
 fs/f2fs/f2fs.h  |  1 +
 fs/f2fs/super.c |  2 ++
 3 files changed, 23 insertions(+)

diff --git a/fs/f2fs/data.c b/fs/f2fs/data.c
index 995cf78b23c5e..066d29938c03a 100644
--- a/fs/f2fs/data.c
+++ b/fs/f2fs/data.c
@@ -2296,6 +2296,7 @@ static int f2fs_mpage_readpages(struct inode *inode,
 	unsigned nr_pages = rac ? readahead_count(rac) : 1;
 	unsigned max_nr_pages = nr_pages;
 	int ret = 0;
+	bool drop_ra = false;
 
 	map.m_pblk = 0;
 	map.m_lblk = 0;
@@ -2306,10 +2307,26 @@ static int f2fs_mpage_readpages(struct inode *inode,
 	map.m_seg_type = NO_CHECK_TYPE;
 	map.m_may_create = false;
 
+	/*
+	 * Two readahead threads for same address range can cause race condition
+	 * which fragments sequential read IOs. So let's avoid each other.
+	 */
+	if (rac && readahead_count(rac)) {
+		if (READ_ONCE(F2FS_I(inode)->ra_offset) == readahead_index(rac))
+			drop_ra = true;
+		else
+			WRITE_ONCE(F2FS_I(inode)->ra_offset,
+						readahead_index(rac));
+	}
+
 	for (; nr_pages; nr_pages--) {
 		if (rac) {
 			page = readahead_page(rac);
 			prefetchw(&page->flags);
+			if (drop_ra) {
+				f2fs_put_page(page, 1);
+				continue;
+			}
 		}
 
 #ifdef CONFIG_F2FS_FS_COMPRESSION
@@ -2372,6 +2389,9 @@ static int f2fs_mpage_readpages(struct inode *inode,
 	}
 	if (bio)
 		__submit_bio(F2FS_I_SB(inode), bio, DATA);
+
+	if (rac && readahead_count(rac) && !drop_ra)
+		WRITE_ONCE(F2FS_I(inode)->ra_offset, -1);
 	return ret;
 }
 
diff --git a/fs/f2fs/f2fs.h b/fs/f2fs/f2fs.h
index 6a655edeb522f..e6e47618a3576 100644
--- a/fs/f2fs/f2fs.h
+++ b/fs/f2fs/f2fs.h
@@ -809,6 +809,7 @@ struct f2fs_inode_info {
 	struct list_head inmem_pages;	/* inmemory pages managed by f2fs */
 	struct task_struct *inmem_task;	/* store inmemory task */
 	struct mutex inmem_lock;	/* lock for inmemory pages */
+	pgoff_t ra_offset;		/* ongoing readahead offset */
 	struct extent_tree *extent_tree;	/* cached extent_tree entry */
 
 	/* avoid racing between foreground op and gc */
diff --git a/fs/f2fs/super.c b/fs/f2fs/super.c
index 7326522057378..80cb7cd358f84 100644
--- a/fs/f2fs/super.c
+++ b/fs/f2fs/super.c
@@ -1015,6 +1015,8 @@ static struct inode *f2fs_alloc_inode(struct super_block *sb)
 	/* Will be used by directory only */
 	fi->i_dir_level = F2FS_SB(sb)->dir_level;
 
+	fi->ra_offset = -1;
+
 	return &fi->vfs_inode;
 }
 
-- 
2.27.0.212.ge8ba1cc988-goog