f2fs: remove checking dirty_exceed

PATCH New Jaegeuk Kim 1 participant 7 messages Dec 15, 2014

Subsystems (2)

VFS F2FS file system

lore

Share & tools

lore

Jaegeuk Kim

author maintainer

2014-12-15 07:27 UTC #0 lore raw

PATCH 1/7 f2fs: remove checking dirty_exceed

We don't need to force to write dirty_exceeded for f2fs_balance_fs_bg. This flag was only meaningful to write bypassing conditions.

Signed-off-by Jaegeuk Kim <[email protected]>

---
 fs/f2fs/node.c | 2 --
 1 file changed, 2 deletions(-)

fs/f2fs/node.c +0 -2

index f83326c..d6f073e 100644
@@ -57,8 +57,6 @@ bool available_free_memory(struct f2fs_sb_info *sbi, int type)
 	} else if (type == INO_ENTRIES) {
 		int i;
 
-		if (sbi->sb->s_bdi->dirty_exceeded)
-			return false;
 		for (i = 0; i <= UPDATE_INO; i++)
 			mem_size += (sbi->im[i].ino_num *
 				sizeof(struct ino_entry)) >> PAGE_CACHE_SHIFT;

Signature · 1 line

2.1.1

Jaegeuk Kim

author maintainer

2014-12-15 07:27 UTC #1 lore raw

PATCH 2/7 f2fs: fix wrong condition check to trigger f2fs_sync_fs

If there is not enough available memory, we need to trigger f2fs_sync_fs.

Signed-off-by Jaegeuk Kim <[email protected]>

---
 fs/f2fs/segment.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

fs/f2fs/segment.c +1 -1

index 42607a6..11e4b5c 100644
@@ -290,7 +290,7 @@ void f2fs_balance_fs_bg(struct f2fs_sb_info *sbi)
 	/* check the # of cached NAT entries and prefree segments */
 	if (try_to_free_nats(sbi, NAT_ENTRY_PER_BLOCK) ||
 			excess_prefree_segs(sbi) ||
-			available_free_memory(sbi, INO_ENTRIES))
+			!available_free_memory(sbi, INO_ENTRIES))
 		f2fs_sync_fs(sbi->sb, true);
 }

Signature · 9 lines

2.1.1


------------------------------------------------------------------------------
Download BIRT iHub F-Type - The Free Enterprise-Grade BIRT Server
from Actuate! Instantly Supercharge Your Business Reports and Dashboards
with Interactivity, Sharing, Native Excel Exports, App Integration & more
Get technology previously reserved for billion-dollar corporations, FREE
http://pubads.g.doubleclick.net/gampad/clk?id=164703151&iu=/4140/ostg.clktrk

Jaegeuk Kim

author maintainer

2014-12-15 07:27 UTC #2 lore raw

PATCH 3/7 f2fs: don't need to call lock_op and lock_page for abort

In reply to Jaegeuk Kim

We don't need to call lock_op and lock_page at the aborting path.

Signed-off-by Jaegeuk Kim <[email protected]>

---
 fs/f2fs/segment.c | 35 ++++++++++++++++++++---------------
 1 file changed, 20 insertions(+), 15 deletions(-)

fs/f2fs/segment.c +20 ~~-15~~

index 11e4b5c..3ce86c5 100644
@@ -240,33 +240,38 @@ void commit_inmem_pages(struct inode *inode, bool abort)
 	 * Otherwise, f2fs_gc in f2fs_balance_fs can wait forever until this
 	 * inode becomes free by iget_locked in f2fs_iget.
 	 */
-	if (!abort)
+	if (!abort) {
 		f2fs_balance_fs(sbi);
-
-	f2fs_lock_op(sbi);
+		f2fs_lock_op(sbi);
+	}
 
 	mutex_lock(&fi->inmem_lock);
 	list_for_each_entry_safe(cur, tmp, &fi->inmem_pages, list) {
-		lock_page(cur->page);
-		if (!abort && cur->page->mapping == inode->i_mapping) {
-			f2fs_wait_on_page_writeback(cur->page, DATA);
-			if (clear_page_dirty_for_io(cur->page))
-				inode_dec_dirty_pages(inode);
-			do_write_data_page(cur->page, &fio);
-			submit_bio = true;
+		if (!abort) {
+			lock_page(cur->page);
+			if (cur->page->mapping == inode->i_mapping) {
+				f2fs_wait_on_page_writeback(cur->page, DATA);
+				if (clear_page_dirty_for_io(cur->page))
+					inode_dec_dirty_pages(inode);
+				do_write_data_page(cur->page, &fio);
+				submit_bio = true;
+			}
+			f2fs_put_page(cur->page, 1);
+		} else {
+			put_page(cur->page);
 		}
 		radix_tree_delete(&fi->inmem_root, cur->page->index);
-		f2fs_put_page(cur->page, 1);
 		list_del(&cur->list);
 		kmem_cache_free(inmem_entry_slab, cur);
 		dec_page_count(F2FS_I_SB(inode), F2FS_INMEM_PAGES);
 	}
-	if (submit_bio)
-		f2fs_submit_merged_bio(sbi, DATA, WRITE);
 	mutex_unlock(&fi->inmem_lock);
 
-	filemap_fdatawait_range(inode->i_mapping, 0, LLONG_MAX);
-	f2fs_unlock_op(sbi);
+	if (!abort) {
+		f2fs_unlock_op(sbi);
+		if (submit_bio)
+			f2fs_submit_merged_bio(sbi, DATA, WRITE);
+	}
 }
 
 /*

Signature · 9 lines

2.1.1


------------------------------------------------------------------------------
Download BIRT iHub F-Type - The Free Enterprise-Grade BIRT Server
from Actuate! Instantly Supercharge Your Business Reports and Dashboards
with Interactivity, Sharing, Native Excel Exports, App Integration & more
Get technology previously reserved for billion-dollar corporations, FREE
http://pubads.g.doubleclick.net/gampad/clk?id=164703151&iu=/4140/ostg.clktrk

Jaegeuk Kim

author maintainer

2014-12-15 07:27 UTC #3 lore raw

PATCH 4/7 f2fs: change atomic and volatile write policies

In reply to Jaegeuk Kim

This patch adds two new ioctls to release inmemory pages grabbed by atomic writes.
o f2fs_ioc_abort_volatile_write
- If transaction was failed, all the grabbed pages and data should be written.
o f2fs_ioc_release_volatile_write
- This is to enhance the performance of PERSIST mode in sqlite.

In order to avoid huge memory consumption which causes OOM, this patch changes volatile writes to use normal dirty pages, instead blocked flushing to the disk as long as system does not suffer from memory pressure.

Signed-off-by Jaegeuk Kim <[email protected]>

---
 fs/f2fs/data.c    |  9 +++++--
 fs/f2fs/f2fs.h    |  8 ++++++
 fs/f2fs/file.c    | 77 +++++++++++++++++++++++++++++++++++++++++++++++--------
 fs/f2fs/inode.c   |  2 +-
 fs/f2fs/node.c    |  3 +++
 fs/f2fs/node.h    |  1 +
 fs/f2fs/segment.c |  2 +-
 7 files changed, 88 insertions(+), 14 deletions(-)

fs/f2fs/data.c +7 -2

index 7ec697b..32264e3 100644
@@ -814,6 +814,11 @@ static int f2fs_write_data_page(struct page *page,
 write:
 	if (unlikely(sbi->por_doing))
 		goto redirty_out;
+	if (f2fs_is_drop_cache(inode))
+		goto out;
+	if (f2fs_is_volatile_file(inode) && !wbc->for_reclaim &&
+			available_free_memory(sbi, BASE_CHECK))
+		goto redirty_out;
 
 	/* Dentry blocks are controlled by checkpoint */
 	if (S_ISDIR(inode->i_mode)) {

@@ -1109,7 +1114,7 @@ static void f2fs_invalidate_data_page(struct page *page, unsigned int offset,
 	if (offset % PAGE_CACHE_SIZE || length != PAGE_CACHE_SIZE)
 		return;
 
-	if (f2fs_is_atomic_file(inode) || f2fs_is_volatile_file(inode))
+	if (f2fs_is_atomic_file(inode))
 		invalidate_inmem_page(inode, page);
 
 	if (PageDirty(page))

@@ -1132,7 +1137,7 @@ static int f2fs_set_data_page_dirty(struct page *page)
 
 	SetPageUptodate(page);
 
-	if (f2fs_is_atomic_file(inode) || f2fs_is_volatile_file(inode)) {
+	if (f2fs_is_atomic_file(inode)) {
 		register_inmem_page(inode, page);
 		return 1;
 	}

fs/f2fs/f2fs.h +8 -0

index ec58bb2..8c9bf3d 100644
@@ -201,6 +201,8 @@ static inline bool __has_cursum_space(struct f2fs_summary_block *sum, int size,
 #define F2FS_IOC_START_ATOMIC_WRITE	_IO(F2FS_IOCTL_MAGIC, 1)
 #define F2FS_IOC_COMMIT_ATOMIC_WRITE	_IO(F2FS_IOCTL_MAGIC, 2)
 #define F2FS_IOC_START_VOLATILE_WRITE	_IO(F2FS_IOCTL_MAGIC, 3)
+#define F2FS_IOC_RELEASE_VOLATILE_WRITE	_IO(F2FS_IOCTL_MAGIC, 4)
+#define F2FS_IOC_ABORT_VOLATILE_WRITE	_IO(F2FS_IOCTL_MAGIC, 5)
 
 #if defined(__KERNEL__) && defined(CONFIG_COMPAT)
 /*

@@ -1113,6 +1115,7 @@ enum {
 	FI_NEED_IPU,		/* used for ipu per file */
 	FI_ATOMIC_FILE,		/* indicate atomic file */
 	FI_VOLATILE_FILE,	/* indicate volatile file */
+	FI_DROP_CACHE,		/* drop dirty page cache */
 	FI_DATA_EXIST,		/* indicate data exists */
 };

@@ -1220,6 +1223,11 @@ static inline bool f2fs_is_volatile_file(struct inode *inode)
 	return is_inode_flag_set(F2FS_I(inode), FI_VOLATILE_FILE);
 }
 
+static inline bool f2fs_is_drop_cache(struct inode *inode)
+{
+	return is_inode_flag_set(F2FS_I(inode), FI_DROP_CACHE);
+}
+
 static inline void *inline_data_addr(struct page *page)
 {
 	struct f2fs_inode *ri = F2FS_INODE(page);

fs/f2fs/file.c +67 ~~-10~~

122 lines not on this page · raw message

fs/f2fs/inode.c +1 -1

index 196cc78..3a8958d 100644
@@ -304,7 +304,7 @@ void f2fs_evict_inode(struct inode *inode)
 	nid_t xnid = F2FS_I(inode)->i_xattr_nid;
 
 	/* some remained atomic pages should discarded */
-	if (f2fs_is_atomic_file(inode) || f2fs_is_volatile_file(inode))
+	if (f2fs_is_atomic_file(inode))
 		commit_inmem_pages(inode, true);
 
 	trace_f2fs_evict_inode(inode);

fs/f2fs/node.c +3 -0

index d6f073e..cabecee 100644
@@ -61,6 +61,9 @@ bool available_free_memory(struct f2fs_sb_info *sbi, int type)
 			mem_size += (sbi->im[i].ino_num *
 				sizeof(struct ino_entry)) >> PAGE_CACHE_SHIFT;
 		res = mem_size < ((avail_ram * nm_i->ram_thresh / 100) >> 1);
+	} else {
+		if (sbi->sb->s_bdi->dirty_exceeded)
+			return false;
 	}
 	return res;
 }

fs/f2fs/node.h +1 -0

index d10b644..036b887 100644
@@ -108,6 +108,7 @@ enum mem_type {
 	NAT_ENTRIES,	/* indicates the cached nat entry */
 	DIRTY_DENTS,	/* indicates dirty dentry pages */
 	INO_ENTRIES,	/* indicates inode entries */
+	BASE_CHECK,	/* check kernel status */
 };
 
 struct nat_entry_set {

fs/f2fs/segment.c +1 -1

index 3ce86c5..de9c070 100644
@@ -230,7 +230,7 @@ void commit_inmem_pages(struct inode *inode, bool abort)
 	bool submit_bio = false;
 	struct f2fs_io_info fio = {
 		.type = DATA,
-		.rw = WRITE_SYNC,
+		.rw = WRITE_SYNC | REQ_PRIO,
 	};
 
 	/*

Signature · 9 lines

2.1.1


------------------------------------------------------------------------------
Download BIRT iHub F-Type - The Free Enterprise-Grade BIRT Server
from Actuate! Instantly Supercharge Your Business Reports and Dashboards
with Interactivity, Sharing, Native Excel Exports, App Integration & more
Get technology previously reserved for billion-dollar corporations, FREE
http://pubads.g.doubleclick.net/gampad/clk?id=164703151&iu=/4140/ostg.clktrk

Jaegeuk Kim

author maintainer

2014-12-15 07:27 UTC #4 lore raw

PATCH 5/7 f2fs: fix small discards not to issue redundantly

In reply to Jaegeuk Kim

The ckpt_valid_map and cur_valid_map are synced by seg_info_to_raw_sit.

In the case of small discards, the candidates are selected before sync, while fitrim selects candidates after sync.

So, for small discards, we need to add candidates only just being obsoleted.

Signed-off-by Jaegeuk Kim <[email protected]>

---
 fs/f2fs/segment.c | 8 +++++---
 1 file changed, 5 insertions(+), 3 deletions(-)

fs/f2fs/segment.c +5 -3

index de9c070..335418c 100644
@@ -525,7 +525,8 @@ static void add_discard_addrs(struct f2fs_sb_info *sbi, struct cp_control *cpc)
 	bool force = (cpc->reason == CP_DISCARD);
 	int i;
 
-	if (!force && !test_opt(sbi, DISCARD))
+	if (!force && (!test_opt(sbi, DISCARD) ||
+			SM_I(sbi)->nr_discards >= SM_I(sbi)->max_discards))
 		return;
 
 	if (force && !se->valid_blocks) {

@@ -553,7 +554,8 @@ static void add_discard_addrs(struct f2fs_sb_info *sbi, struct cp_control *cpc)
 
 	/* SIT_VBLOCK_MAP_SIZE should be multiple of sizeof(unsigned long) */
 	for (i = 0; i < entries; i++)
-		dmap[i] = ~(cur_map[i] | ckpt_map[i]);
+		dmap[i] = force ? ~ckpt_map[i] :
+				(cur_map[i] ^ ckpt_map[i]) & ckpt_map[i];
 
 	while (force || SM_I(sbi)->nr_discards <= SM_I(sbi)->max_discards) {
 		start = __find_rev_next_bit(dmap, max_blocks, end + 1);

@@ -1759,7 +1761,7 @@ void flush_sit_entries(struct f2fs_sb_info *sbi, struct cp_control *cpc)
 			se = get_seg_entry(sbi, segno);
 
 			/* add discard candidates */
-			if (SM_I(sbi)->nr_discards < SM_I(sbi)->max_discards) {
+			if (cpc->reason != CP_DISCARD) {
 				cpc->trim_start = segno;
 				add_discard_addrs(sbi, cpc);
 			}

Signature · 1 line

2.1.1

Jaegeuk Kim

author maintainer

2014-12-15 07:27 UTC #5 lore raw

PATCH 6/7 f2fs: remove unnecessary call to invalidate inmemory pages

In reply to Jaegeuk Kim

Now we use inmemory pages for atomic write only and provide abort procedure, we don't need to truncate them explicitly.

Signed-off-by Jaegeuk Kim <[email protected]>

---
 fs/f2fs/data.c    |  3 ---
 fs/f2fs/f2fs.h    |  1 -
 fs/f2fs/segment.c | 17 -----------------
 3 files changed, 21 deletions(-)

fs/f2fs/data.c +0 -3

index 32264e3..caa08e4 100644
@@ -1114,9 +1114,6 @@ static void f2fs_invalidate_data_page(struct page *page, unsigned int offset,
 	if (offset % PAGE_CACHE_SIZE || length != PAGE_CACHE_SIZE)
 		return;
 
-	if (f2fs_is_atomic_file(inode))
-		invalidate_inmem_page(inode, page);
-
 	if (PageDirty(page))
 		inode_dec_dirty_pages(inode);
 	ClearPagePrivate(page);

fs/f2fs/f2fs.h +0 -1

index 8c9bf3d..d3699da 100644
@@ -1397,7 +1397,6 @@ void destroy_node_manager_caches(void);
  * segment.c
  */
 void register_inmem_page(struct inode *, struct page *);
-void invalidate_inmem_page(struct inode *, struct page *);
 void commit_inmem_pages(struct inode *, bool);
 void f2fs_balance_fs(struct f2fs_sb_info *);
 void f2fs_balance_fs_bg(struct f2fs_sb_info *);

fs/f2fs/segment.c +0 ~~-17~~

index 335418c..3791fa9 100644
@@ -205,23 +205,6 @@ retry:
 	mutex_unlock(&fi->inmem_lock);
 }
 
-void invalidate_inmem_page(struct inode *inode, struct page *page)
-{
-	struct f2fs_inode_info *fi = F2FS_I(inode);
-	struct inmem_pages *cur;
-
-	mutex_lock(&fi->inmem_lock);
-	cur = radix_tree_lookup(&fi->inmem_root, page->index);
-	if (cur) {
-		radix_tree_delete(&fi->inmem_root, cur->page->index);
-		f2fs_put_page(cur->page, 0);
-		list_del(&cur->list);
-		kmem_cache_free(inmem_entry_slab, cur);
-		dec_page_count(F2FS_I_SB(inode), F2FS_INMEM_PAGES);
-	}
-	mutex_unlock(&fi->inmem_lock);
-}
-
 void commit_inmem_pages(struct inode *inode, bool abort)
 {
 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);

Signature · 9 lines

2.1.1


------------------------------------------------------------------------------
Download BIRT iHub F-Type - The Free Enterprise-Grade BIRT Server
from Actuate! Instantly Supercharge Your Business Reports and Dashboards
with Interactivity, Sharing, Native Excel Exports, App Integration & more
Get technology previously reserved for billion-dollar corporations, FREE
http://pubads.g.doubleclick.net/gampad/clk?id=164703151&iu=/4140/ostg.clktrk

Jaegeuk Kim

author maintainer

2014-12-15 07:27 UTC #6 lore raw

PATCH 7/7 f2fs: use missing the use of f2fs_kunmap_page

In reply to Jaegeuk Kim

This patch calls f2fs_kunmap_page which I missed before.

Signed-off-by Jaegeuk Kim <[email protected]>

---
 fs/f2fs/dir.c | 3 +--
 1 file changed, 1 insertion(+), 2 deletions(-)

fs/f2fs/dir.c +1 -2

index b1a7d57..b74097a 100644
@@ -286,8 +286,7 @@ void f2fs_set_link(struct inode *dir, struct f2fs_dir_entry *de,
 	f2fs_wait_on_page_writeback(page, type);
 	de->ino = cpu_to_le32(inode->i_ino);
 	set_de_type(de, inode);
-	if (!f2fs_has_inline_dentry(dir))
-		kunmap(page);
+	f2fs_dentry_kunmap(dir, page);
 	set_page_dirty(page);
 	dir->i_mtime = dir->i_ctime = CURRENT_TIME;
 	mark_inode_dirty(dir);

Signature · 9 lines

2.1.1


------------------------------------------------------------------------------
Download BIRT iHub F-Type - The Free Enterprise-Grade BIRT Server
from Actuate! Instantly Supercharge Your Business Reports and Dashboards
with Interactivity, Sharing, Native Excel Exports, App Integration & more
Get technology previously reserved for billion-dollar corporations, FREE
http://pubads.g.doubleclick.net/gampad/clk?id=164703151&iu=/4140/ostg.clktrk