f2fs: introduce a new global lock scheme

In the previous version, f2fs uses global locks according to the usage types,
such as directory operations, block allocation, block write, and so on.

Reference the following lock types in f2fs.h.
enum lock_type {
	RENAME,		/* for renaming operations */
	DENTRY_OPS,	/* for directory operations */
	DATA_WRITE,	/* for data write */
	DATA_NEW,	/* for data allocation */
	DATA_TRUNC,	/* for data truncate */
	NODE_NEW,	/* for node allocation */
	NODE_TRUNC,	/* for node truncate */
	NODE_WRITE,	/* for node write */
	NR_LOCK_TYPE,
};

In that case, we lose the performance under the multi-threading environment,
since every types of operations must be conducted one at a time.

In order to address the problem, let's share the locks globally with a mutex
array regardless of any types.
So, let users grab a mutex and perform their jobs in parallel as much as
possbile.

For this, I propose a new global lock scheme as follows.

0. Data structure
 - f2fs_sb_info -> mutex_lock[NR_GLOBAL_LOCKS]
 - f2fs_sb_info -> node_write

1. mutex_lock_op(sbi)
 - try to get an avaiable lock from the array.
 - returns the index of the gottern lock variable.

2. mutex_unlock_op(sbi, index of the lock)
 - unlock the given index of the lock.

3. mutex_lock_all(sbi)
 - grab all the locks in the array before the checkpoint.

4. mutex_unlock_all(sbi)
 - release all the locks in the array after checkpoint.

5. block_operations()
 - call mutex_lock_all()
 - sync_dirty_dir_inodes()
 - grab node_write
 - sync_node_pages()

Note that,
 the pairs of mutex_lock_op()/mutex_unlock_op() and
 mutex_lock_all()/mutex_unlock_all() should be used together.

Signed-off-by: Jaegeuk Kim <jaegeuk.kim@samsung.com>

1 files changed, 17 insertions, 27 deletions

diff --git a/fs/f2fs/node.c b/fs/f2fs/node.c
index ad3adbee842a..5a7edf90ca45 100644
--- a/fs/f2fs/node.c
+++ b/fs/f2fs/node.c
@@ -385,6 +385,9 @@ got:
 
 /*
  * Caller should call f2fs_put_dnode(dn).
+ * Also, it should grab and release a mutex by calling mutex_lock_op() and
+ * mutex_unlock_op() only if ro is not set RDONLY_NODE.
+ * In the case of RDONLY_NODE, we don't need to care about mutex.
  */
 int get_dnode_of_data(struct dnode_of_data *dn, pgoff_t index, int mode)
 {
@@ -415,11 +418,8 @@ int get_dnode_of_data(struct dnode_of_data *dn, pgoff_t index, int mode)
 		bool done = false;
 
 		if (!nids[i] && mode == ALLOC_NODE) {
-			mutex_lock_op(sbi, NODE_NEW);
-
 			/* alloc new node */
 			if (!alloc_nid(sbi, &(nids[i]))) {
-				mutex_unlock_op(sbi, NODE_NEW);
 				err = -ENOSPC;
 				goto release_pages;
 			}
@@ -428,14 +428,12 @@ int get_dnode_of_data(struct dnode_of_data *dn, pgoff_t index, int mode)
 			npage[i] = new_node_page(dn, noffset[i]);
 			if (IS_ERR(npage[i])) {
 				alloc_nid_failed(sbi, nids[i]);
-				mutex_unlock_op(sbi, NODE_NEW);
 				err = PTR_ERR(npage[i]);
 				goto release_pages;
 			}
 
 			set_nid(parent, offset[i - 1], nids[i], i == 1);
 			alloc_nid_done(sbi, nids[i]);
-			mutex_unlock_op(sbi, NODE_NEW);
 			done = true;
 		} else if (mode == LOOKUP_NODE_RA && i == level && level > 1) {
 			npage[i] = get_node_page_ra(parent, offset[i - 1]);
@@ -745,6 +743,10 @@ fail:
 	return err > 0 ? 0 : err;
 }
 
+/*
+ * Caller should grab and release a mutex by calling mutex_lock_op() and
+ * mutex_unlock_op().
+ */
 int remove_inode_page(struct inode *inode)
 {
 	struct f2fs_sb_info *sbi = F2FS_SB(inode->i_sb);
@@ -752,21 +754,16 @@ int remove_inode_page(struct inode *inode)
 	nid_t ino = inode->i_ino;
 	struct dnode_of_data dn;
 
-	mutex_lock_op(sbi, NODE_TRUNC);
 	page = get_node_page(sbi, ino);
-	if (IS_ERR(page)) {
-		mutex_unlock_op(sbi, NODE_TRUNC);
+	if (IS_ERR(page))
 		return PTR_ERR(page);
-	}
 
 	if (F2FS_I(inode)->i_xattr_nid) {
 		nid_t nid = F2FS_I(inode)->i_xattr_nid;
 		struct page *npage = get_node_page(sbi, nid);
 
-		if (IS_ERR(npage)) {
-			mutex_unlock_op(sbi, NODE_TRUNC);
+		if (IS_ERR(npage))
 			return PTR_ERR(npage);
-		}
 
 		F2FS_I(inode)->i_xattr_nid = 0;
 		set_new_dnode(&dn, inode, page, npage, nid);
@@ -778,23 +775,18 @@ int remove_inode_page(struct inode *inode)
 	BUG_ON(inode->i_blocks != 0 && inode->i_blocks != 1);
 	set_new_dnode(&dn, inode, page, page, ino);
 	truncate_node(&dn);
-
-	mutex_unlock_op(sbi, NODE_TRUNC);
 	return 0;
 }
 
 int new_inode_page(struct inode *inode, const struct qstr *name)
 {
-	struct f2fs_sb_info *sbi = F2FS_SB(inode->i_sb);
 	struct page *page;
 	struct dnode_of_data dn;
 
 	/* allocate inode page for new inode */
 	set_new_dnode(&dn, inode, NULL, NULL, inode->i_ino);
-	mutex_lock_op(sbi, NODE_NEW);
 	page = new_node_page(&dn, 0);
 	init_dent_inode(name, page);
-	mutex_unlock_op(sbi, NODE_NEW);
 	if (IS_ERR(page))
 		return PTR_ERR(page);
 	f2fs_put_page(page, 1);
@@ -985,7 +977,7 @@ void sync_inode_page(struct dnode_of_data *dn)
 		if (!dn->inode_page_locked)
 			unlock_page(dn->inode_page);
 	} else {
-		f2fs_write_inode(dn->inode, NULL);
+		update_inode_page(dn->inode);
 	}
 }
 
@@ -1102,8 +1094,6 @@ static int f2fs_write_node_page(struct page *page,
 
 	wait_on_page_writeback(page);
 
-	mutex_lock_op(sbi, NODE_WRITE);
-
 	/* get old block addr of this node page */
 	nid = nid_of_node(page);
 	BUG_ON(page->index != nid);
@@ -1111,25 +1101,25 @@ static int f2fs_write_node_page(struct page *page,
 	get_node_info(sbi, nid, &ni);
 
 	/* This page is already truncated */
-	if (ni.blk_addr == NULL_ADDR)
-		goto out;
+	if (ni.blk_addr == NULL_ADDR) {
+		dec_page_count(sbi, F2FS_DIRTY_NODES);
+		unlock_page(page);
+		return 0;
+	}
 
 	if (wbc->for_reclaim) {
 		dec_page_count(sbi, F2FS_DIRTY_NODES);
 		wbc->pages_skipped++;
 		set_page_dirty(page);
-		mutex_unlock_op(sbi, NODE_WRITE);
 		return AOP_WRITEPAGE_ACTIVATE;
 	}
 
+	mutex_lock(&sbi->node_write);
 	set_page_writeback(page);
-
-	/* insert node offset */
 	write_node_page(sbi, page, nid, ni.blk_addr, &new_addr);
 	set_node_addr(sbi, &ni, new_addr);
-out:
 	dec_page_count(sbi, F2FS_DIRTY_NODES);
-	mutex_unlock_op(sbi, NODE_WRITE);
+	mutex_unlock(&sbi->node_write);
 	unlock_page(page);
 	return 0;
 }