ext4: online defrag-- Defragmentation for the relevant files (-r mode)
From: Akira Fujita <[email protected]>
Relevant file fragmentation could be solved by moving
the files under the specified directory close together with
the block containing the directory data.
Signed-off-by: Akira Fujita <[email protected]>
Signed-off-by: Takashi Sato <[email protected]>
---
fs/ext4/defrag.c | 48 +++++++++++++++++++++++++++++++++++-------------
fs/ext4/ext4.h | 4 +++-
fs/ext4/inode.c | 2 +-
fs/ext4/ioctl.c | 1 +
4 files changed, 40 insertions(+), 15 deletions(-)
diff --git a/fs/ext4/defrag.c b/fs/ext4/defrag.c
index 8ae5a48..61f577b 100644
--- a/fs/ext4/defrag.c
+++ b/fs/ext4/defrag.c
@@ -95,13 +95,26 @@ int ext4_defrag_ioctl(struct inode *inode, struct file *filp, unsigned int cmd,
{
int err = 0;
- if (!(EXT4_I(inode)->i_flags & EXT4_EXTENTS_FL)) {
+ if (!(EXT4_I(inode)->i_flags & EXT4_EXTENTS_FL ||
+ cmd == EXT4_IOC_FIBMAP)) {
printk(KERN_ERR "ext4 defrag: ino[%lu] is not extents "
"based file\n", inode->i_ino);
return -EOPNOTSUPP;
}
- if (cmd == EXT4_IOC_DEFRAG) {
+ if (cmd == EXT4_IOC_FIBMAP) {
+ ext4_fsblk_t __user *p = (ext4_fsblk_t __user *)arg;
+ ext4_fsblk_t block = 0;
+ struct address_space *mapping = filp->f_mapping;
+
+ if (copy_from_user(&block, (ext4_fsblk_t __user *)arg,
+ sizeof(block)))
+ return -EFAULT;
+
+ block = ext4_bmap(mapping, block);
+
+ return put_user(block, p);
+ } else if (cmd == EXT4_IOC_DEFRAG) {
struct ext4_ext_defrag_data defrag;
struct ext4_super_block *es = EXT4_SB(inode->i_sb)->s_es;
@@ -127,7 +140,7 @@ int ext4_defrag_ioctl(struct inode *inode, struct file *filp, unsigned int cmd,
}
err = ext4_defrag(filp, defrag.start_offset,
- defrag.defrag_size);
+ defrag.defrag_size, defrag.goal);
}
return err;
@@ -745,6 +758,7 @@ out:
* @dest_path: indicating the temporary inode's extent
* @req_blocks: contiguous blocks count we need
* @iblock: target file offset
+ * @goal: goal offset
*
*/
static void
@@ -752,7 +766,8 @@ ext4_defrag_fill_ar(struct inode *org_inode, struct inode *dest_inode,
struct ext4_allocation_request *ar,
struct ext4_ext_path *org_path,
struct ext4_ext_path *dest_path,
- ext4_fsblk_t req_blocks, ext4_lblk_t iblock)
+ ext4_fsblk_t req_blocks, ext4_lblk_t iblock,
+ ext4_fsblk_t goal)
{
ar->inode = dest_inode;
ar->len = req_blocks;
@@ -764,7 +779,10 @@ ext4_defrag_fill_ar(struct inode *org_inode, struct inode *dest_inode,
ar->lright = 0;
ar->pright = 0;
- ar->goal = ext4_ext_find_goal(dest_inode, dest_path, iblock);
+ if (goal)
+ ar->goal = goal;
+ else
+ ar->goal = ext4_ext_find_goal(dest_inode, dest_path, iblock);
}
/**
@@ -941,6 +959,7 @@ out:
* original extent tree
* @tar_end: the last block number of the allocated blocks
* @sum_tmp: the extents count in the allocated blocks
+ * @goal: block offset for allocaton
*
*
* This function returns the values as below.
@@ -951,7 +970,7 @@ out:
static int
ext4_defrag_comp_ext_count(struct inode *org_inode,
struct ext4_ext_path *org_path, ext4_lblk_t tar_end,
- int sum_tmp)
+ int sum_tmp, ext4_fsblk_t goal)
{
struct ext4_extent *ext = NULL;
int depth = ext_depth(org_inode);
@@ -975,7 +994,7 @@ ext4_defrag_comp_ext_count(struct inode *org_inode,
* Fail if goal is not set and the fragmentation
* is not improved.
*/
- if (sum_org == sum_tmp) {
+ if (sum_org == sum_tmp && !goal) {
/* Not improved */
ret = 1;
} else if (sum_org < sum_tmp) {
@@ -1006,6 +1025,7 @@ ext4_defrag_comp_ext_count(struct inode *org_inode,
* @tar_start: starting offset to allocate in blocks
* @tar_blocks: the number of blocks to allocate
* @iblock: file related offset
+ * @goal: block offset for allocaton
*
*
* This function returns the value as below:
@@ -1016,7 +1036,8 @@ ext4_defrag_comp_ext_count(struct inode *org_inode,
static int
ext4_defrag_new_extent_tree(struct inode *org_inode, struct inode *tmp_inode,
struct ext4_ext_path *org_path, ext4_lblk_t tar_start,
- ext4_lblk_t tar_blocks, ext4_lblk_t iblock)
+ ext4_lblk_t tar_blocks, ext4_lblk_t iblock,
+ ext4_fsblk_t goal)
{
handle_t *handle;
struct ext4_extent_header *eh = NULL;
@@ -1042,7 +1063,7 @@ ext4_defrag_new_extent_tree(struct inode *org_inode, struct inode *tmp_inode,
/* Fill struct ext4_allocation_request with necessary info */
ext4_defrag_fill_ar(org_inode, tmp_inode, &ar, org_path,
- dest_path, tar_blocks, iblock);
+ dest_path, tar_blocks, iblock, goal);
handle = ext4_journal_start(tmp_inode, 0);
if (IS_ERR(handle)) {
@@ -1074,7 +1095,7 @@ ext4_defrag_new_extent_tree(struct inode *org_inode, struct inode *tmp_inode,
}
ret = ext4_defrag_comp_ext_count(org_inode, org_path, tar_end,
- sum_tmp);
+ sum_tmp, goal);
out:
if (ret < 0 || ret == 1) {
@@ -1204,13 +1225,14 @@ out:
* @filp: pointer to file
* @block_start: starting offset to defrag in blocks
* @defrag_size: size of defrag in blocks
+ * @goal: block offset for allocation
*
* This function returns the number of blocks if succeed, otherwise
* returns error value.
*/
int
ext4_defrag(struct file *filp, ext4_lblk_t block_start,
- ext4_lblk_t defrag_size)
+ ext4_lblk_t defrag_size, ext4_fsblk_t goal)
{
struct inode *org_inode = filp->f_dentry->d_inode, *tmp_inode = NULL;
struct ext4_ext_path *org_path = NULL, *holecheck_path = NULL;
@@ -1329,14 +1351,14 @@ ext4_defrag(struct file *filp, ext4_lblk_t block_start,
}
/* Found an isolated block */
- if (seq_extents == 1) {
+ if (seq_extents == 1 && !goal) {
seq_start = le32_to_cpu(ext_cur->ee_block);
goto CLEANUP;
}
ret = ext4_defrag_new_extent_tree(org_inode, tmp_inode,
org_path, seq_start, seq_blocks,
- block_start);
+ block_start, goal);
if (ret < 0) {
break;
diff --git a/fs/ext4/ext4.h b/fs/ext4/ext4.h
index 1e9ce39..12b3fea 100644
--- a/fs/ext4/ext4.h
+++ b/fs/ext4/ext4.h
@@ -298,6 +298,7 @@ struct ext4_new_group_data {
#define EXT4_IOC_GETRSVSZ _IOR('f', 5, long)
#define EXT4_IOC_SETRSVSZ _IOW('f', 6, long)
#define EXT4_IOC_MIGRATE _IO('f', 7)
+#define EXT4_IOC_FIBMAP _IOW('f', 9, ext4_fsblk_t)
#define EXT4_IOC_DEFRAG _IOW('f', 10, struct ext4_ext_defrag_data)
/*
@@ -1018,6 +1019,7 @@ extern int ext4_htree_store_dirent(struct file *dir_file, __u32 hash,
__u32 minor_hash,
struct ext4_dir_entry_2 *dirent);
extern void ext4_htree_free_dir_info(struct dir_private_info *p);
+extern sector_t ext4_bmap(struct address_space *mapping, sector_t block);
/* fsync.c */
extern int ext4_sync_file (struct file *, struct dentry *, int);
@@ -1133,7 +1135,7 @@ extern void ext4_inode_table_set(struct super_block *sb,
extern handle_t *ext4_ext_journal_restart(handle_t *handle, int needed);
/* defrag.c */
extern int ext4_defrag(struct file *filp, ext4_lblk_t block_start,
- ext4_lblk_t defrag_size);
+ ext4_lblk_t defrag_size, ext4_fsblk_t goal);
extern int ext4_defrag_ioctl(struct inode *, struct file *, unsigned int,
unsigned long);
diff --git a/fs/ext4/inode.c b/fs/ext4/inode.c
index a30f56c..27f3d2c 100644
--- a/fs/ext4/inode.c
+++ b/fs/ext4/inode.c
@@ -1648,7 +1648,7 @@ out:
* So, if we see any bmap calls here on a modified, data-journaled file,
* take extra steps to flush any blocks which might be in the cache.
*/
-static sector_t ext4_bmap(struct address_space *mapping, sector_t block)
+sector_t ext4_bmap(struct address_space *mapping, sector_t block)
{
struct inode *inode = mapping->host;
journal_t *journal;
diff --git a/fs/ext4/ioctl.c b/fs/ext4/ioctl.c
index 4c7fca1..e1b9c10 100644
--- a/fs/ext4/ioctl.c
+++ b/fs/ext4/ioctl.c
@@ -241,6 +241,7 @@ setversion_out:
return err;
}
+ case EXT4_IOC_FIBMAP:
case EXT4_IOC_DEFRAG: {
return ext4_defrag_ioctl(inode, filp, cmd, arg);
}