* [PATCH 1/3] f2fs: introduce fi->i_sem to protect fi's info
@ 2014-03-21 2:01 ` Jaegeuk Kim
0 siblings, 0 replies; 6+ messages in thread
From: Jaegeuk Kim @ 2014-03-21 2:01 UTC (permalink / raw)
Cc: Jaegeuk Kim, linux-fsdevel, linux-kernel, linux-f2fs-devel
This patch introduces fi->i_sem to protect fi's info that includes xattr_ver,
pino, i_nlink.
This enables to remove i_mutex during f2fs_sync_file, resulting in performance
improvement when a number of fsync calls are triggered from many concurrent
threads.
Signed-off-by: Jaegeuk Kim <jaegeuk.kim@samsung.com>
---
fs/f2fs/dir.c | 6 ++++++
fs/f2fs/f2fs.h | 1 +
fs/f2fs/file.c | 14 ++++++++++----
fs/f2fs/namei.c | 7 +++++++
fs/f2fs/super.c | 1 +
fs/f2fs/xattr.c | 3 +++
6 files changed, 28 insertions(+), 4 deletions(-)
diff --git a/fs/f2fs/dir.c b/fs/f2fs/dir.c
index 7c9b17c..972fd0e 100644
--- a/fs/f2fs/dir.c
+++ b/fs/f2fs/dir.c
@@ -493,6 +493,7 @@ start:
add_dentry:
f2fs_wait_on_page_writeback(dentry_page, DATA);
+ down_write(&F2FS_I(inode)->i_sem);
page = init_inode_metadata(inode, dir, name);
if (IS_ERR(page)) {
err = PTR_ERR(page);
@@ -515,6 +516,8 @@ add_dentry:
update_parent_metadata(dir, inode, current_depth);
fail:
+ up_write(&F2FS_I(inode)->i_sem);
+
if (is_inode_flag_set(F2FS_I(dir), FI_UPDATE_DIR)) {
update_inode_page(dir);
clear_inode_flag(F2FS_I(dir), FI_UPDATE_DIR);
@@ -559,6 +562,8 @@ void f2fs_delete_entry(struct f2fs_dir_entry *dentry, struct page *page,
if (inode) {
struct f2fs_sb_info *sbi = F2FS_SB(dir->i_sb);
+ down_write(&F2FS_I(inode)->i_sem);
+
if (S_ISDIR(inode->i_mode)) {
drop_nlink(dir);
update_inode_page(dir);
@@ -569,6 +574,7 @@ void f2fs_delete_entry(struct f2fs_dir_entry *dentry, struct page *page,
drop_nlink(inode);
i_size_write(inode, 0);
}
+ up_write(&F2FS_I(inode)->i_sem);
update_inode_page(inode);
if (inode->i_nlink == 0)
diff --git a/fs/f2fs/f2fs.h b/fs/f2fs/f2fs.h
index 05c6524..469779a 100644
--- a/fs/f2fs/f2fs.h
+++ b/fs/f2fs/f2fs.h
@@ -210,6 +210,7 @@ struct f2fs_inode_info {
/* Use below internally in f2fs*/
unsigned long flags; /* use to pass per-file flags */
+ struct rw_semaphore i_sem; /* protect fi info */
atomic_t dirty_dents; /* # of dirty dentry pages */
f2fs_hash_t chash; /* hash value of given file name */
unsigned int clevel; /* maximum level of given file name */
diff --git a/fs/f2fs/file.c b/fs/f2fs/file.c
index e755ee5..a9474cd 100644
--- a/fs/f2fs/file.c
+++ b/fs/f2fs/file.c
@@ -111,6 +111,7 @@ static int get_parent_ino(struct inode *inode, nid_t *pino)
int f2fs_sync_file(struct file *file, loff_t start, loff_t end, int datasync)
{
struct inode *inode = file->f_mapping->host;
+ struct f2fs_inode_info *fi = F2FS_I(inode);
struct f2fs_sb_info *sbi = F2FS_SB(inode->i_sb);
int ret = 0;
bool need_cp = false;
@@ -133,7 +134,7 @@ int f2fs_sync_file(struct file *file, loff_t start, loff_t end, int datasync)
/* guarantee free sections for fsync */
f2fs_balance_fs(sbi);
- mutex_lock(&inode->i_mutex);
+ down_read(&fi->i_sem);
/*
* Both of fdatasync() and fsync() are able to be recovered from
@@ -150,21 +151,27 @@ int f2fs_sync_file(struct file *file, loff_t start, loff_t end, int datasync)
else if (F2FS_I(inode)->xattr_ver == cur_cp_version(F2FS_CKPT(sbi)))
need_cp = true;
+ up_read(&fi->i_sem);
+
if (need_cp) {
nid_t pino;
- F2FS_I(inode)->xattr_ver = 0;
-
/* all the dirty node pages should be flushed for POR */
ret = f2fs_sync_fs(inode->i_sb, 1);
+
+ down_write(&fi->i_sem);
+ F2FS_I(inode)->xattr_ver = 0;
if (file_wrong_pino(inode) && inode->i_nlink == 1 &&
get_parent_ino(inode, &pino)) {
F2FS_I(inode)->i_pino = pino;
file_got_pino(inode);
+ up_write(&fi->i_sem);
mark_inode_dirty_sync(inode);
ret = f2fs_write_inode(inode, NULL);
if (ret)
goto out;
+ } else {
+ up_write(&fi->i_sem);
}
} else {
/* if there is no written node page, write its inode page */
@@ -180,7 +187,6 @@ int f2fs_sync_file(struct file *file, loff_t start, loff_t end, int datasync)
ret = blkdev_issue_flush(inode->i_sb->s_bdev, GFP_KERNEL, NULL);
}
out:
- mutex_unlock(&inode->i_mutex);
trace_f2fs_sync_file_exit(inode, need_cp, datasync, ret);
return ret;
}
diff --git a/fs/f2fs/namei.c b/fs/f2fs/namei.c
index 397d459..0cea874 100644
--- a/fs/f2fs/namei.c
+++ b/fs/f2fs/namei.c
@@ -424,12 +424,17 @@ static int f2fs_rename(struct inode *old_dir, struct dentry *old_dentry,
}
f2fs_set_link(new_dir, new_entry, new_page, old_inode);
+ down_write(&F2FS_I(old_inode)->i_sem);
F2FS_I(old_inode)->i_pino = new_dir->i_ino;
+ up_write(&F2FS_I(old_inode)->i_sem);
new_inode->i_ctime = CURRENT_TIME;
+ down_write(&F2FS_I(new_inode)->i_sem);
if (old_dir_entry)
drop_nlink(new_inode);
drop_nlink(new_inode);
+ up_write(&F2FS_I(new_inode)->i_sem);
+
mark_inode_dirty(new_inode);
if (!new_inode->i_nlink)
@@ -459,7 +464,9 @@ static int f2fs_rename(struct inode *old_dir, struct dentry *old_dentry,
if (old_dir != new_dir) {
f2fs_set_link(old_inode, old_dir_entry,
old_dir_page, new_dir);
+ down_write(&F2FS_I(old_inode)->i_sem);
F2FS_I(old_inode)->i_pino = new_dir->i_ino;
+ up_write(&F2FS_I(old_inode)->i_sem);
update_inode_page(old_inode);
} else {
kunmap(old_dir_page);
diff --git a/fs/f2fs/super.c b/fs/f2fs/super.c
index 34c47b2..89ea046 100644
--- a/fs/f2fs/super.c
+++ b/fs/f2fs/super.c
@@ -360,6 +360,7 @@ static struct inode *f2fs_alloc_inode(struct super_block *sb)
fi->i_current_depth = 1;
fi->i_advise = 0;
rwlock_init(&fi->ext.ext_lock);
+ init_rwsem(&fi->i_sem);
set_inode_flag(fi, FI_NEW_INODE);
diff --git a/fs/f2fs/xattr.c b/fs/f2fs/xattr.c
index 89d0422..8419130 100644
--- a/fs/f2fs/xattr.c
+++ b/fs/f2fs/xattr.c
@@ -590,7 +590,10 @@ int f2fs_setxattr(struct inode *inode, int name_index, const char *name,
f2fs_balance_fs(sbi);
f2fs_lock_op(sbi);
+ /* protect xattr_ver */
+ down_write(&F2FS_I(inode)->i_sem);
err = __f2fs_setxattr(inode, name_index, name, value, value_len, ipage);
+ up_write(&F2FS_I(inode)->i_sem);
f2fs_unlock_op(sbi);
return err;
--
1.8.4.474.g128a96c
^ permalink raw reply related [flat|nested] 6+ messages in thread
* [PATCH 1/3] f2fs: introduce fi->i_sem to protect fi's info
@ 2014-03-21 2:01 ` Jaegeuk Kim
0 siblings, 0 replies; 6+ messages in thread
From: Jaegeuk Kim @ 2014-03-21 2:01 UTC (permalink / raw)
Cc: linux-fsdevel, linux-kernel, linux-f2fs-devel
This patch introduces fi->i_sem to protect fi's info that includes xattr_ver,
pino, i_nlink.
This enables to remove i_mutex during f2fs_sync_file, resulting in performance
improvement when a number of fsync calls are triggered from many concurrent
threads.
Signed-off-by: Jaegeuk Kim <jaegeuk.kim@samsung.com>
---
fs/f2fs/dir.c | 6 ++++++
fs/f2fs/f2fs.h | 1 +
fs/f2fs/file.c | 14 ++++++++++----
fs/f2fs/namei.c | 7 +++++++
fs/f2fs/super.c | 1 +
fs/f2fs/xattr.c | 3 +++
6 files changed, 28 insertions(+), 4 deletions(-)
diff --git a/fs/f2fs/dir.c b/fs/f2fs/dir.c
index 7c9b17c..972fd0e 100644
--- a/fs/f2fs/dir.c
+++ b/fs/f2fs/dir.c
@@ -493,6 +493,7 @@ start:
add_dentry:
f2fs_wait_on_page_writeback(dentry_page, DATA);
+ down_write(&F2FS_I(inode)->i_sem);
page = init_inode_metadata(inode, dir, name);
if (IS_ERR(page)) {
err = PTR_ERR(page);
@@ -515,6 +516,8 @@ add_dentry:
update_parent_metadata(dir, inode, current_depth);
fail:
+ up_write(&F2FS_I(inode)->i_sem);
+
if (is_inode_flag_set(F2FS_I(dir), FI_UPDATE_DIR)) {
update_inode_page(dir);
clear_inode_flag(F2FS_I(dir), FI_UPDATE_DIR);
@@ -559,6 +562,8 @@ void f2fs_delete_entry(struct f2fs_dir_entry *dentry, struct page *page,
if (inode) {
struct f2fs_sb_info *sbi = F2FS_SB(dir->i_sb);
+ down_write(&F2FS_I(inode)->i_sem);
+
if (S_ISDIR(inode->i_mode)) {
drop_nlink(dir);
update_inode_page(dir);
@@ -569,6 +574,7 @@ void f2fs_delete_entry(struct f2fs_dir_entry *dentry, struct page *page,
drop_nlink(inode);
i_size_write(inode, 0);
}
+ up_write(&F2FS_I(inode)->i_sem);
update_inode_page(inode);
if (inode->i_nlink == 0)
diff --git a/fs/f2fs/f2fs.h b/fs/f2fs/f2fs.h
index 05c6524..469779a 100644
--- a/fs/f2fs/f2fs.h
+++ b/fs/f2fs/f2fs.h
@@ -210,6 +210,7 @@ struct f2fs_inode_info {
/* Use below internally in f2fs*/
unsigned long flags; /* use to pass per-file flags */
+ struct rw_semaphore i_sem; /* protect fi info */
atomic_t dirty_dents; /* # of dirty dentry pages */
f2fs_hash_t chash; /* hash value of given file name */
unsigned int clevel; /* maximum level of given file name */
diff --git a/fs/f2fs/file.c b/fs/f2fs/file.c
index e755ee5..a9474cd 100644
--- a/fs/f2fs/file.c
+++ b/fs/f2fs/file.c
@@ -111,6 +111,7 @@ static int get_parent_ino(struct inode *inode, nid_t *pino)
int f2fs_sync_file(struct file *file, loff_t start, loff_t end, int datasync)
{
struct inode *inode = file->f_mapping->host;
+ struct f2fs_inode_info *fi = F2FS_I(inode);
struct f2fs_sb_info *sbi = F2FS_SB(inode->i_sb);
int ret = 0;
bool need_cp = false;
@@ -133,7 +134,7 @@ int f2fs_sync_file(struct file *file, loff_t start, loff_t end, int datasync)
/* guarantee free sections for fsync */
f2fs_balance_fs(sbi);
- mutex_lock(&inode->i_mutex);
+ down_read(&fi->i_sem);
/*
* Both of fdatasync() and fsync() are able to be recovered from
@@ -150,21 +151,27 @@ int f2fs_sync_file(struct file *file, loff_t start, loff_t end, int datasync)
else if (F2FS_I(inode)->xattr_ver == cur_cp_version(F2FS_CKPT(sbi)))
need_cp = true;
+ up_read(&fi->i_sem);
+
if (need_cp) {
nid_t pino;
- F2FS_I(inode)->xattr_ver = 0;
-
/* all the dirty node pages should be flushed for POR */
ret = f2fs_sync_fs(inode->i_sb, 1);
+
+ down_write(&fi->i_sem);
+ F2FS_I(inode)->xattr_ver = 0;
if (file_wrong_pino(inode) && inode->i_nlink == 1 &&
get_parent_ino(inode, &pino)) {
F2FS_I(inode)->i_pino = pino;
file_got_pino(inode);
+ up_write(&fi->i_sem);
mark_inode_dirty_sync(inode);
ret = f2fs_write_inode(inode, NULL);
if (ret)
goto out;
+ } else {
+ up_write(&fi->i_sem);
}
} else {
/* if there is no written node page, write its inode page */
@@ -180,7 +187,6 @@ int f2fs_sync_file(struct file *file, loff_t start, loff_t end, int datasync)
ret = blkdev_issue_flush(inode->i_sb->s_bdev, GFP_KERNEL, NULL);
}
out:
- mutex_unlock(&inode->i_mutex);
trace_f2fs_sync_file_exit(inode, need_cp, datasync, ret);
return ret;
}
diff --git a/fs/f2fs/namei.c b/fs/f2fs/namei.c
index 397d459..0cea874 100644
--- a/fs/f2fs/namei.c
+++ b/fs/f2fs/namei.c
@@ -424,12 +424,17 @@ static int f2fs_rename(struct inode *old_dir, struct dentry *old_dentry,
}
f2fs_set_link(new_dir, new_entry, new_page, old_inode);
+ down_write(&F2FS_I(old_inode)->i_sem);
F2FS_I(old_inode)->i_pino = new_dir->i_ino;
+ up_write(&F2FS_I(old_inode)->i_sem);
new_inode->i_ctime = CURRENT_TIME;
+ down_write(&F2FS_I(new_inode)->i_sem);
if (old_dir_entry)
drop_nlink(new_inode);
drop_nlink(new_inode);
+ up_write(&F2FS_I(new_inode)->i_sem);
+
mark_inode_dirty(new_inode);
if (!new_inode->i_nlink)
@@ -459,7 +464,9 @@ static int f2fs_rename(struct inode *old_dir, struct dentry *old_dentry,
if (old_dir != new_dir) {
f2fs_set_link(old_inode, old_dir_entry,
old_dir_page, new_dir);
+ down_write(&F2FS_I(old_inode)->i_sem);
F2FS_I(old_inode)->i_pino = new_dir->i_ino;
+ up_write(&F2FS_I(old_inode)->i_sem);
update_inode_page(old_inode);
} else {
kunmap(old_dir_page);
diff --git a/fs/f2fs/super.c b/fs/f2fs/super.c
index 34c47b2..89ea046 100644
--- a/fs/f2fs/super.c
+++ b/fs/f2fs/super.c
@@ -360,6 +360,7 @@ static struct inode *f2fs_alloc_inode(struct super_block *sb)
fi->i_current_depth = 1;
fi->i_advise = 0;
rwlock_init(&fi->ext.ext_lock);
+ init_rwsem(&fi->i_sem);
set_inode_flag(fi, FI_NEW_INODE);
diff --git a/fs/f2fs/xattr.c b/fs/f2fs/xattr.c
index 89d0422..8419130 100644
--- a/fs/f2fs/xattr.c
+++ b/fs/f2fs/xattr.c
@@ -590,7 +590,10 @@ int f2fs_setxattr(struct inode *inode, int name_index, const char *name,
f2fs_balance_fs(sbi);
f2fs_lock_op(sbi);
+ /* protect xattr_ver */
+ down_write(&F2FS_I(inode)->i_sem);
err = __f2fs_setxattr(inode, name_index, name, value, value_len, ipage);
+ up_write(&F2FS_I(inode)->i_sem);
f2fs_unlock_op(sbi);
return err;
--
1.8.4.474.g128a96c
------------------------------------------------------------------------------
Learn Graph Databases - Download FREE O'Reilly Book
"Graph Databases" is the definitive new guide to graph databases and their
applications. Written by three acclaimed leaders in the field,
this first edition is now available. Download your free book today!
http://p.sf.net/sfu/13534_NeoTech
^ permalink raw reply related [flat|nested] 6+ messages in thread
* [PATCH 2/3] f2fs: skip unnecessary node writes during fsync
2014-03-21 2:01 ` Jaegeuk Kim
@ 2014-03-21 2:01 ` Jaegeuk Kim
-1 siblings, 0 replies; 6+ messages in thread
From: Jaegeuk Kim @ 2014-03-21 2:01 UTC (permalink / raw)
Cc: Jaegeuk Kim, linux-fsdevel, linux-kernel, linux-f2fs-devel
If multiple redundant fsync calls are triggered, we don't need to write its
node pages with fsync mark continuously.
So, this patch adds FI_NEED_FSYNC to track whether the latest node block is
written with the fsync mark or not.
If the mark was set, a new fsync doesn't need to write a node block.
Otherwise, we should do a new node block with the mark for roll-forward
recovery.
Signed-off-by: Jaegeuk Kim <jaegeuk.kim@samsung.com>
---
fs/f2fs/f2fs.h | 1 +
fs/f2fs/file.c | 2 ++
fs/f2fs/node.c | 37 ++++++++++++++++++++++++++++---------
fs/f2fs/node.h | 1 +
4 files changed, 32 insertions(+), 9 deletions(-)
diff --git a/fs/f2fs/f2fs.h b/fs/f2fs/f2fs.h
index 469779a..f83433e 100644
--- a/fs/f2fs/f2fs.h
+++ b/fs/f2fs/f2fs.h
@@ -1126,6 +1126,7 @@ struct dnode_of_data;
struct node_info;
int is_checkpointed_node(struct f2fs_sb_info *, nid_t);
+bool fsync_mark_done(struct f2fs_sb_info *, nid_t);
void get_node_info(struct f2fs_sb_info *, nid_t, struct node_info *);
int get_dnode_of_data(struct dnode_of_data *, pgoff_t, int);
int truncate_inode_blocks(struct inode *, pgoff_t);
diff --git a/fs/f2fs/file.c b/fs/f2fs/file.c
index a9474cd..6ba2668 100644
--- a/fs/f2fs/file.c
+++ b/fs/f2fs/file.c
@@ -176,6 +176,8 @@ int f2fs_sync_file(struct file *file, loff_t start, loff_t end, int datasync)
} else {
/* if there is no written node page, write its inode page */
while (!sync_node_pages(sbi, inode->i_ino, &wbc)) {
+ if (fsync_mark_done(sbi, inode->i_ino))
+ goto out;
mark_inode_dirty_sync(inode);
ret = f2fs_write_inode(inode, NULL);
if (ret)
diff --git a/fs/f2fs/node.c b/fs/f2fs/node.c
index daf644c..eced8d7 100644
--- a/fs/f2fs/node.c
+++ b/fs/f2fs/node.c
@@ -133,6 +133,20 @@ int is_checkpointed_node(struct f2fs_sb_info *sbi, nid_t nid)
return is_cp;
}
+bool fsync_mark_done(struct f2fs_sb_info *sbi, nid_t nid)
+{
+ struct f2fs_nm_info *nm_i = NM_I(sbi);
+ struct nat_entry *e;
+ bool fsync_done = false;
+
+ read_lock(&nm_i->nat_tree_lock);
+ e = __lookup_nat_cache(nm_i, nid);
+ if (e)
+ fsync_done = e->fsync_done;
+ read_unlock(&nm_i->nat_tree_lock);
+ return fsync_done;
+}
+
static struct nat_entry *grab_nat_entry(struct f2fs_nm_info *nm_i, nid_t nid)
{
struct nat_entry *new;
@@ -173,7 +187,7 @@ retry:
}
static void set_node_addr(struct f2fs_sb_info *sbi, struct node_info *ni,
- block_t new_blkaddr)
+ block_t new_blkaddr, bool fsync_done)
{
struct f2fs_nm_info *nm_i = NM_I(sbi);
struct nat_entry *e;
@@ -217,6 +231,11 @@ retry:
/* change address */
nat_set_blkaddr(e, new_blkaddr);
__set_nat_cache_dirty(nm_i, e);
+
+ /* update fsync_mark if its inode nat entry is still alive */
+ e = __lookup_nat_cache(nm_i, ni->ino);
+ if (e)
+ e->fsync_done = fsync_done;
write_unlock(&nm_i->nat_tree_lock);
}
@@ -483,7 +502,7 @@ static void truncate_node(struct dnode_of_data *dn)
/* Deallocate node address */
invalidate_blocks(sbi, ni.blk_addr);
dec_valid_node_count(sbi, dn->inode);
- set_node_addr(sbi, &ni, NULL_ADDR);
+ set_node_addr(sbi, &ni, NULL_ADDR, false);
if (dn->nid == dn->inode->i_ino) {
remove_orphan_inode(sbi, dn->nid);
@@ -846,7 +865,7 @@ struct page *new_node_page(struct dnode_of_data *dn,
f2fs_bug_on(old_ni.blk_addr != NULL_ADDR);
new_ni = old_ni;
new_ni.ino = dn->inode->i_ino;
- set_node_addr(sbi, &new_ni, NEW_ADDR);
+ set_node_addr(sbi, &new_ni, NEW_ADDR, false);
fill_node_footer(page, dn->nid, dn->inode->i_ino, ofs, true);
set_cold_node(dn->inode, page);
@@ -1202,7 +1221,7 @@ static int f2fs_write_node_page(struct page *page,
mutex_lock(&sbi->node_write);
set_page_writeback(page);
write_node_page(sbi, page, &fio, nid, ni.blk_addr, &new_addr);
- set_node_addr(sbi, &ni, new_addr);
+ set_node_addr(sbi, &ni, new_addr, is_fsync_dnode(page));
dec_page_count(sbi, F2FS_DIRTY_NODES);
mutex_unlock(&sbi->node_write);
unlock_page(page);
@@ -1503,7 +1522,7 @@ void recover_node_page(struct f2fs_sb_info *sbi, struct page *page,
block_t new_blkaddr)
{
rewrite_node_page(sbi, page, sum, ni->blk_addr, new_blkaddr);
- set_node_addr(sbi, ni, new_blkaddr);
+ set_node_addr(sbi, ni, new_blkaddr, false);
clear_node_page_dirty(page);
}
@@ -1559,7 +1578,7 @@ bool recover_xattr_data(struct inode *inode, struct page *page, block_t blkaddr)
f2fs_bug_on(ni.blk_addr == NULL_ADDR);
invalidate_blocks(sbi, ni.blk_addr);
dec_valid_node_count(sbi, inode);
- set_node_addr(sbi, &ni, NULL_ADDR);
+ set_node_addr(sbi, &ni, NULL_ADDR, false);
recover_xnid:
/* 2: allocate new xattr nid */
@@ -1569,12 +1588,12 @@ recover_xnid:
remove_free_nid(NM_I(sbi), new_xnid);
get_node_info(sbi, new_xnid, &ni);
ni.ino = inode->i_ino;
- set_node_addr(sbi, &ni, NEW_ADDR);
+ set_node_addr(sbi, &ni, NEW_ADDR, false);
F2FS_I(inode)->i_xattr_nid = new_xnid;
/* 3: update xattr blkaddr */
refresh_sit_entry(sbi, NEW_ADDR, blkaddr);
- set_node_addr(sbi, &ni, blkaddr);
+ set_node_addr(sbi, &ni, blkaddr, false);
update_inode_page(inode);
return true;
@@ -1612,7 +1631,7 @@ int recover_inode_page(struct f2fs_sb_info *sbi, struct page *page)
if (unlikely(!inc_valid_node_count(sbi, NULL)))
WARN_ON(1);
- set_node_addr(sbi, &new_ni, NEW_ADDR);
+ set_node_addr(sbi, &new_ni, NEW_ADDR, false);
inc_valid_inode_count(sbi);
f2fs_put_page(ipage, 1);
return 0;
diff --git a/fs/f2fs/node.h b/fs/f2fs/node.h
index c2015b7..5decc1a 100644
--- a/fs/f2fs/node.h
+++ b/fs/f2fs/node.h
@@ -42,6 +42,7 @@ struct node_info {
struct nat_entry {
struct list_head list; /* for clean or dirty nat list */
bool checkpointed; /* whether it is checkpointed or not */
+ bool fsync_done; /* whether the latest node has fsync mark */
struct node_info ni; /* in-memory node information */
};
--
1.8.4.474.g128a96c
^ permalink raw reply related [flat|nested] 6+ messages in thread
* [PATCH 2/3] f2fs: skip unnecessary node writes during fsync
@ 2014-03-21 2:01 ` Jaegeuk Kim
0 siblings, 0 replies; 6+ messages in thread
From: Jaegeuk Kim @ 2014-03-21 2:01 UTC (permalink / raw)
Cc: linux-fsdevel, linux-kernel, linux-f2fs-devel
If multiple redundant fsync calls are triggered, we don't need to write its
node pages with fsync mark continuously.
So, this patch adds FI_NEED_FSYNC to track whether the latest node block is
written with the fsync mark or not.
If the mark was set, a new fsync doesn't need to write a node block.
Otherwise, we should do a new node block with the mark for roll-forward
recovery.
Signed-off-by: Jaegeuk Kim <jaegeuk.kim@samsung.com>
---
fs/f2fs/f2fs.h | 1 +
fs/f2fs/file.c | 2 ++
fs/f2fs/node.c | 37 ++++++++++++++++++++++++++++---------
fs/f2fs/node.h | 1 +
4 files changed, 32 insertions(+), 9 deletions(-)
diff --git a/fs/f2fs/f2fs.h b/fs/f2fs/f2fs.h
index 469779a..f83433e 100644
--- a/fs/f2fs/f2fs.h
+++ b/fs/f2fs/f2fs.h
@@ -1126,6 +1126,7 @@ struct dnode_of_data;
struct node_info;
int is_checkpointed_node(struct f2fs_sb_info *, nid_t);
+bool fsync_mark_done(struct f2fs_sb_info *, nid_t);
void get_node_info(struct f2fs_sb_info *, nid_t, struct node_info *);
int get_dnode_of_data(struct dnode_of_data *, pgoff_t, int);
int truncate_inode_blocks(struct inode *, pgoff_t);
diff --git a/fs/f2fs/file.c b/fs/f2fs/file.c
index a9474cd..6ba2668 100644
--- a/fs/f2fs/file.c
+++ b/fs/f2fs/file.c
@@ -176,6 +176,8 @@ int f2fs_sync_file(struct file *file, loff_t start, loff_t end, int datasync)
} else {
/* if there is no written node page, write its inode page */
while (!sync_node_pages(sbi, inode->i_ino, &wbc)) {
+ if (fsync_mark_done(sbi, inode->i_ino))
+ goto out;
mark_inode_dirty_sync(inode);
ret = f2fs_write_inode(inode, NULL);
if (ret)
diff --git a/fs/f2fs/node.c b/fs/f2fs/node.c
index daf644c..eced8d7 100644
--- a/fs/f2fs/node.c
+++ b/fs/f2fs/node.c
@@ -133,6 +133,20 @@ int is_checkpointed_node(struct f2fs_sb_info *sbi, nid_t nid)
return is_cp;
}
+bool fsync_mark_done(struct f2fs_sb_info *sbi, nid_t nid)
+{
+ struct f2fs_nm_info *nm_i = NM_I(sbi);
+ struct nat_entry *e;
+ bool fsync_done = false;
+
+ read_lock(&nm_i->nat_tree_lock);
+ e = __lookup_nat_cache(nm_i, nid);
+ if (e)
+ fsync_done = e->fsync_done;
+ read_unlock(&nm_i->nat_tree_lock);
+ return fsync_done;
+}
+
static struct nat_entry *grab_nat_entry(struct f2fs_nm_info *nm_i, nid_t nid)
{
struct nat_entry *new;
@@ -173,7 +187,7 @@ retry:
}
static void set_node_addr(struct f2fs_sb_info *sbi, struct node_info *ni,
- block_t new_blkaddr)
+ block_t new_blkaddr, bool fsync_done)
{
struct f2fs_nm_info *nm_i = NM_I(sbi);
struct nat_entry *e;
@@ -217,6 +231,11 @@ retry:
/* change address */
nat_set_blkaddr(e, new_blkaddr);
__set_nat_cache_dirty(nm_i, e);
+
+ /* update fsync_mark if its inode nat entry is still alive */
+ e = __lookup_nat_cache(nm_i, ni->ino);
+ if (e)
+ e->fsync_done = fsync_done;
write_unlock(&nm_i->nat_tree_lock);
}
@@ -483,7 +502,7 @@ static void truncate_node(struct dnode_of_data *dn)
/* Deallocate node address */
invalidate_blocks(sbi, ni.blk_addr);
dec_valid_node_count(sbi, dn->inode);
- set_node_addr(sbi, &ni, NULL_ADDR);
+ set_node_addr(sbi, &ni, NULL_ADDR, false);
if (dn->nid == dn->inode->i_ino) {
remove_orphan_inode(sbi, dn->nid);
@@ -846,7 +865,7 @@ struct page *new_node_page(struct dnode_of_data *dn,
f2fs_bug_on(old_ni.blk_addr != NULL_ADDR);
new_ni = old_ni;
new_ni.ino = dn->inode->i_ino;
- set_node_addr(sbi, &new_ni, NEW_ADDR);
+ set_node_addr(sbi, &new_ni, NEW_ADDR, false);
fill_node_footer(page, dn->nid, dn->inode->i_ino, ofs, true);
set_cold_node(dn->inode, page);
@@ -1202,7 +1221,7 @@ static int f2fs_write_node_page(struct page *page,
mutex_lock(&sbi->node_write);
set_page_writeback(page);
write_node_page(sbi, page, &fio, nid, ni.blk_addr, &new_addr);
- set_node_addr(sbi, &ni, new_addr);
+ set_node_addr(sbi, &ni, new_addr, is_fsync_dnode(page));
dec_page_count(sbi, F2FS_DIRTY_NODES);
mutex_unlock(&sbi->node_write);
unlock_page(page);
@@ -1503,7 +1522,7 @@ void recover_node_page(struct f2fs_sb_info *sbi, struct page *page,
block_t new_blkaddr)
{
rewrite_node_page(sbi, page, sum, ni->blk_addr, new_blkaddr);
- set_node_addr(sbi, ni, new_blkaddr);
+ set_node_addr(sbi, ni, new_blkaddr, false);
clear_node_page_dirty(page);
}
@@ -1559,7 +1578,7 @@ bool recover_xattr_data(struct inode *inode, struct page *page, block_t blkaddr)
f2fs_bug_on(ni.blk_addr == NULL_ADDR);
invalidate_blocks(sbi, ni.blk_addr);
dec_valid_node_count(sbi, inode);
- set_node_addr(sbi, &ni, NULL_ADDR);
+ set_node_addr(sbi, &ni, NULL_ADDR, false);
recover_xnid:
/* 2: allocate new xattr nid */
@@ -1569,12 +1588,12 @@ recover_xnid:
remove_free_nid(NM_I(sbi), new_xnid);
get_node_info(sbi, new_xnid, &ni);
ni.ino = inode->i_ino;
- set_node_addr(sbi, &ni, NEW_ADDR);
+ set_node_addr(sbi, &ni, NEW_ADDR, false);
F2FS_I(inode)->i_xattr_nid = new_xnid;
/* 3: update xattr blkaddr */
refresh_sit_entry(sbi, NEW_ADDR, blkaddr);
- set_node_addr(sbi, &ni, blkaddr);
+ set_node_addr(sbi, &ni, blkaddr, false);
update_inode_page(inode);
return true;
@@ -1612,7 +1631,7 @@ int recover_inode_page(struct f2fs_sb_info *sbi, struct page *page)
if (unlikely(!inc_valid_node_count(sbi, NULL)))
WARN_ON(1);
- set_node_addr(sbi, &new_ni, NEW_ADDR);
+ set_node_addr(sbi, &new_ni, NEW_ADDR, false);
inc_valid_inode_count(sbi);
f2fs_put_page(ipage, 1);
return 0;
diff --git a/fs/f2fs/node.h b/fs/f2fs/node.h
index c2015b7..5decc1a 100644
--- a/fs/f2fs/node.h
+++ b/fs/f2fs/node.h
@@ -42,6 +42,7 @@ struct node_info {
struct nat_entry {
struct list_head list; /* for clean or dirty nat list */
bool checkpointed; /* whether it is checkpointed or not */
+ bool fsync_done; /* whether the latest node has fsync mark */
struct node_info ni; /* in-memory node information */
};
--
1.8.4.474.g128a96c
------------------------------------------------------------------------------
Learn Graph Databases - Download FREE O'Reilly Book
"Graph Databases" is the definitive new guide to graph databases and their
applications. Written by three acclaimed leaders in the field,
this first edition is now available. Download your free book today!
http://p.sf.net/sfu/13534_NeoTech
^ permalink raw reply related [flat|nested] 6+ messages in thread
* [PATCH 3/3] f2fs: avoid RECLAIM_FS-ON-W warning
2014-03-21 2:01 ` Jaegeuk Kim
@ 2014-03-21 2:01 ` Jaegeuk Kim
-1 siblings, 0 replies; 6+ messages in thread
From: Jaegeuk Kim @ 2014-03-21 2:01 UTC (permalink / raw)
Cc: Jaegeuk Kim, linux-fsdevel, linux-kernel, linux-f2fs-devel
This patch should resolve the following possible bug.
RECLAIM_FS-ON-W at:
mark_held_locks+0xb9/0x140
lockdep_trace_alloc+0x85/0xf0
__kmalloc+0x53/0x1d0
read_all_xattrs+0x3d1/0x3f0 [f2fs]
f2fs_getxattr+0x4f/0x100 [f2fs]
f2fs_get_acl+0x4c/0x290 [f2fs]
get_acl+0x4f/0x80
posix_acl_create+0x72/0x180
f2fs_init_acl+0x29/0xcc [f2fs]
__f2fs_add_link+0x259/0x710 [f2fs]
f2fs_create+0xad/0x1c0 [f2fs]
vfs_create+0xed/0x150
do_last+0xd36/0xed0
path_openat+0xc5/0x680
do_filp_open+0x43/0xa0
do_sys_open+0x13c/0x230
SyS_creat+0x1e/0x20
system_call_fastpath+0x16/0x1b
Signed-off-by: Jaegeuk Kim <jaegeuk.kim@samsung.com>
---
fs/f2fs/acl.c | 2 +-
fs/f2fs/xattr.c | 2 +-
2 files changed, 2 insertions(+), 2 deletions(-)
diff --git a/fs/f2fs/acl.c b/fs/f2fs/acl.c
index fa8da4c..a285715 100644
--- a/fs/f2fs/acl.c
+++ b/fs/f2fs/acl.c
@@ -174,7 +174,7 @@ struct posix_acl *f2fs_get_acl(struct inode *inode, int type)
retval = f2fs_getxattr(inode, name_index, "", NULL, 0);
if (retval > 0) {
- value = kmalloc(retval, GFP_KERNEL);
+ value = kmalloc(retval, GFP_F2FS_ZERO);
if (!value)
return ERR_PTR(-ENOMEM);
retval = f2fs_getxattr(inode, name_index, "", value, retval);
diff --git a/fs/f2fs/xattr.c b/fs/f2fs/xattr.c
index 8419130..0121e45 100644
--- a/fs/f2fs/xattr.c
+++ b/fs/f2fs/xattr.c
@@ -275,7 +275,7 @@ static void *read_all_xattrs(struct inode *inode, struct page *ipage)
inline_size = inline_xattr_size(inode);
- txattr_addr = kzalloc(inline_size + size, GFP_KERNEL);
+ txattr_addr = kzalloc(inline_size + size, GFP_F2FS_ZERO);
if (!txattr_addr)
return NULL;
--
1.8.4.474.g128a96c
^ permalink raw reply related [flat|nested] 6+ messages in thread
* [PATCH 3/3] f2fs: avoid RECLAIM_FS-ON-W warning
@ 2014-03-21 2:01 ` Jaegeuk Kim
0 siblings, 0 replies; 6+ messages in thread
From: Jaegeuk Kim @ 2014-03-21 2:01 UTC (permalink / raw)
Cc: linux-fsdevel, linux-kernel, linux-f2fs-devel
This patch should resolve the following possible bug.
RECLAIM_FS-ON-W at:
mark_held_locks+0xb9/0x140
lockdep_trace_alloc+0x85/0xf0
__kmalloc+0x53/0x1d0
read_all_xattrs+0x3d1/0x3f0 [f2fs]
f2fs_getxattr+0x4f/0x100 [f2fs]
f2fs_get_acl+0x4c/0x290 [f2fs]
get_acl+0x4f/0x80
posix_acl_create+0x72/0x180
f2fs_init_acl+0x29/0xcc [f2fs]
__f2fs_add_link+0x259/0x710 [f2fs]
f2fs_create+0xad/0x1c0 [f2fs]
vfs_create+0xed/0x150
do_last+0xd36/0xed0
path_openat+0xc5/0x680
do_filp_open+0x43/0xa0
do_sys_open+0x13c/0x230
SyS_creat+0x1e/0x20
system_call_fastpath+0x16/0x1b
Signed-off-by: Jaegeuk Kim <jaegeuk.kim@samsung.com>
---
fs/f2fs/acl.c | 2 +-
fs/f2fs/xattr.c | 2 +-
2 files changed, 2 insertions(+), 2 deletions(-)
diff --git a/fs/f2fs/acl.c b/fs/f2fs/acl.c
index fa8da4c..a285715 100644
--- a/fs/f2fs/acl.c
+++ b/fs/f2fs/acl.c
@@ -174,7 +174,7 @@ struct posix_acl *f2fs_get_acl(struct inode *inode, int type)
retval = f2fs_getxattr(inode, name_index, "", NULL, 0);
if (retval > 0) {
- value = kmalloc(retval, GFP_KERNEL);
+ value = kmalloc(retval, GFP_F2FS_ZERO);
if (!value)
return ERR_PTR(-ENOMEM);
retval = f2fs_getxattr(inode, name_index, "", value, retval);
diff --git a/fs/f2fs/xattr.c b/fs/f2fs/xattr.c
index 8419130..0121e45 100644
--- a/fs/f2fs/xattr.c
+++ b/fs/f2fs/xattr.c
@@ -275,7 +275,7 @@ static void *read_all_xattrs(struct inode *inode, struct page *ipage)
inline_size = inline_xattr_size(inode);
- txattr_addr = kzalloc(inline_size + size, GFP_KERNEL);
+ txattr_addr = kzalloc(inline_size + size, GFP_F2FS_ZERO);
if (!txattr_addr)
return NULL;
--
1.8.4.474.g128a96c
------------------------------------------------------------------------------
Learn Graph Databases - Download FREE O'Reilly Book
"Graph Databases" is the definitive new guide to graph databases and their
applications. Written by three acclaimed leaders in the field,
this first edition is now available. Download your free book today!
http://p.sf.net/sfu/13534_NeoTech
^ permalink raw reply related [flat|nested] 6+ messages in thread
end of thread, other threads:[~2014-03-21 2:04 UTC | newest]
Thread overview: 6+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2014-03-21 2:01 [PATCH 1/3] f2fs: introduce fi->i_sem to protect fi's info Jaegeuk Kim
2014-03-21 2:01 ` Jaegeuk Kim
2014-03-21 2:01 ` [PATCH 2/3] f2fs: skip unnecessary node writes during fsync Jaegeuk Kim
2014-03-21 2:01 ` Jaegeuk Kim
2014-03-21 2:01 ` [PATCH 3/3] f2fs: avoid RECLAIM_FS-ON-W warning Jaegeuk Kim
2014-03-21 2:01 ` Jaegeuk Kim
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.