[PATCH v2 3/10] ext4: Add support FALLOC_FL_INSERT_RANGE for fallocate

[Namjae Jeon]

This patch implements fallocate's FALLOC_FL_INSERT_RANGE for Ext4.

1) Make sure that both offset and len are block size aligned.
2) Update the i_size of inode by len bytes.
3) Compute the file's logical block number against offset. If the computed
   block number is not the starting block of the extent, split the extent
   such that the block number is the starting block of the extent.
4) Shift all the extents which are lying bewteen [offset, last allocated extent]
   towards right by len bytes. This step will make a hole of len bytes
   at offset.
5) Allocate unwritten extents for the hole created in step 4.

Signed-off-by: Namjae Jeon <namjae.jeon@xxxxxxxxxxx>
Signed-off-by: Ashish Sangwan <a.sangwan@xxxxxxxxxxx>
---
 fs/ext4/ext4.h              |   1 +
 fs/ext4/extents.c           | 333 ++++++++++++++++++++++++++++++++++++++++++--
 include/trace/events/ext4.h |  25 ++++
 3 files changed, 350 insertions(+), 9 deletions(-)

diff --git a/fs/ext4/ext4.h b/fs/ext4/ext4.h
index 78fed7b..c8e074a 100644
--- a/fs/ext4/ext4.h
+++ b/fs/ext4/ext4.h
@@ -2745,6 +2745,7 @@ extern int ext4_fiemap(struct inode *inode, struct fiemap_extent_info *fieinfo,
 			__u64 start, __u64 len);
 extern int ext4_ext_precache(struct inode *inode);
 extern int ext4_collapse_range(struct inode *inode, loff_t offset, loff_t len);
+extern int ext4_insert_range(struct file *file, loff_t offset, loff_t len);
 
 /* move_extent.c */
 extern void ext4_double_down_write_data_sem(struct inode *first,
diff --git a/fs/ext4/extents.c b/fs/ext4/extents.c
index 086baa9..17321c6 100644
--- a/fs/ext4/extents.c
+++ b/fs/ext4/extents.c
@@ -4900,7 +4900,8 @@ long ext4_fallocate(struct file *file, int mode, loff_t offset, loff_t len)
 
 	/* Return error if mode is not supported */
 	if (mode & ~(FALLOC_FL_KEEP_SIZE | FALLOC_FL_PUNCH_HOLE |
-		     FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_ZERO_RANGE))
+		     FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_ZERO_RANGE |
+		     FALLOC_FL_INSERT_RANGE))
 		return -EOPNOTSUPP;
 
 	if (mode & FALLOC_FL_PUNCH_HOLE)
@@ -4923,6 +4924,9 @@ long ext4_fallocate(struct file *file, int mode, loff_t offset, loff_t len)
 	if (mode & FALLOC_FL_ZERO_RANGE)
 		return ext4_zero_range(file, offset, len, mode);
 
+	if (mode & FALLOC_FL_INSERT_RANGE)
+		return ext4_insert_range(file, offset, len);
+
 	trace_ext4_fallocate_enter(inode, offset, len, mode);
 	lblk = offset >> blkbits;
 	/*
@@ -5225,13 +5229,13 @@ ext4_access_path(handle_t *handle, struct inode *inode,
 }
 
 /*
- * ext4_ext_shift_path_extents:
+ * ext4_ext_shift_path_extents_left:
  * Shift the extents of a path structure lying between path[depth].p_ext
- * and EXT_LAST_EXTENT(path[depth].p_hdr) downwards, by subtracting shift
+ * and EXT_LAST_EXTENT(path[depth].p_hdr) to the left, by subtracting shift
  * from starting block for each extent.
  */
 static int
-ext4_ext_shift_path_extents(struct ext4_ext_path *path, ext4_lblk_t shift,
+ext4_ext_shift_path_extents_left(struct ext4_ext_path *path, ext4_lblk_t shift,
 			    struct inode *inode, handle_t *handle,
 			    ext4_lblk_t *start)
 {
@@ -5301,13 +5305,13 @@ out:
 }
 
 /*
- * ext4_ext_shift_extents:
+ * ext4_ext_shift_extents_left:
  * All the extents which lies in the range from start to the last allocated
- * block for the file are shifted downwards by shift blocks.
+ * block for the file are shifted to the left by shift blocks.
  * On success, 0 is returned, error otherwise.
  */
 static int
-ext4_ext_shift_extents(struct inode *inode, handle_t *handle,
+ext4_ext_shift_extents_left(struct inode *inode, handle_t *handle,
 		       ext4_lblk_t start, ext4_lblk_t shift)
 {
 	struct ext4_ext_path *path;
@@ -5387,7 +5391,7 @@ ext4_ext_shift_extents(struct inode *inode, handle_t *handle,
 				break;
 			}
 		}
-		ret = ext4_ext_shift_path_extents(path, shift, inode,
+		ret = ext4_ext_shift_path_extents_left(path, shift, inode,
 				handle, &start);
 		ext4_ext_drop_refs(path);
 		kfree(path);
@@ -5495,7 +5499,7 @@ int ext4_collapse_range(struct inode *inode, loff_t offset, loff_t len)
 	}
 	ext4_discard_preallocations(inode);
 
-	ret = ext4_ext_shift_extents(inode, handle, punch_stop,
+	ret = ext4_ext_shift_extents_left(inode, handle, punch_stop,
 				     punch_stop - punch_start);
 	if (ret) {
 		up_write(&EXT4_I(inode)->i_data_sem);
@@ -5520,3 +5524,314 @@ out_mutex:
 	mutex_unlock(&inode->i_mutex);
 	return ret;
 }
+
+/*
+ * ext4_ext_shift_path_extents_right:
+ * Shift the extents of a path structure towards right, by adding shift_lblk
+ * to the starting ee_block of each extent. Shifting is done from
+ * the last extent in the path till we reach first extent OR hit start_lblk.
+ * In case the first extent in the path is updated, extent index will also be
+ * updated if it is present.
+ * On success, 0 is returned, error otherwise.
+ */
+static int
+ext4_ext_shift_path_extents_right(struct ext4_ext_path *path,
+		struct inode *inode, handle_t *handle,
+		ext4_lblk_t start_lblk, ext4_lblk_t shift_lblk)
+{
+	int depth, err = 0;
+	struct ext4_extent *ex_start, *ex_last;
+
+	depth = ext_depth(inode);
+	while (depth >= 0) {
+		if (depth == path->p_depth) {
+			ex_start = EXT_FIRST_EXTENT(path[depth].p_hdr);
+
+			ex_last = EXT_LAST_EXTENT(path[depth].p_hdr);
+			if (!ex_last)
+				return -EIO;
+
+			err = ext4_access_path(handle, inode, path + depth);
+			if (err)
+				goto out;
+
+			while ((ex_start <= ex_last) &&
+			       (le32_to_cpu(ex_last->ee_block) >= start_lblk)) {
+				le32_add_cpu(&ex_last->ee_block, shift_lblk);
+				ext4_ext_try_to_merge_right(inode, path,
+						ex_last);
+				ex_last--;
+			}
+			err = ext4_ext_dirty(handle, inode, path + depth);
+			if (err)
+				goto out;
+
+			if (--depth < 0 || ex_start <= ex_last)
+				break;
+		}
+
+		/* Update index too */
+		err = ext4_access_path(handle, inode, path + depth);
+		if (err)
+			goto out;
+		le32_add_cpu(&path[depth].p_idx->ei_block, shift_lblk);
+		err = ext4_ext_dirty(handle, inode, path + depth);
+		if (err)
+			goto out;
+
+		/* we are done if current index is not a starting index */
+		if (path[depth].p_idx != EXT_FIRST_INDEX(path[depth].p_hdr))
+			break;
+
+		depth--;
+	}
+
+out:
+	return err;
+}
+
+/*
+ * ext4_ext_shift_extents_right:
+ * All the extents of an inode which lies in the range from start_lblk
+ * to the last allocated block are shifted right by @shift_lblk blocks.
+ * As we will be shifitng complete extents, @start_lblk should be the
+ * starting block of an extent OR it can lie in a hole.
+ * On success, 0 is returned, error otherwise.
+ */
+static int
+ext4_ext_shift_extents_right(struct inode *inode, handle_t *handle,
+			     ext4_lblk_t start_lblk, ext4_lblk_t shift_lblk)
+{
+	struct ext4_ext_path *path;
+	struct ext4_extent *ex_start;
+	int ret = 0, depth;
+	ext4_lblk_t current_block = EXT_MAX_BLOCKS - 1;
+
+	/* Its safe to start updating extents */
+	while (start_lblk < current_block) {
+		path = ext4_ext_find_extent(inode, current_block, NULL, 0);
+		if (IS_ERR(path))
+			return PTR_ERR(path);
+
+		depth = ext_depth(inode);
+		if (unlikely(path[depth].p_hdr == NULL)) {
+			ret = -EIO;
+			goto out_stop;
+		}
+
+		ex_start = EXT_FIRST_EXTENT(path[depth].p_hdr);
+		if (!ex_start) {
+			ret = -EIO;
+			goto out_stop;
+		}
+
+		current_block = ex_start->ee_block;
+		ret = ext4_ext_shift_path_extents_right(path, inode, handle,
+							start_lblk, shift_lblk);
+out_stop:
+		ext4_ext_drop_refs(path);
+		kfree(path);
+		if (ret)
+			break;
+	}
+
+	return ret;
+}
+
+/*
+ * ext4_insert_range:
+ * This function implements the FALLOC_FL_INSERT_RANGE flag of fallocate.
+ * Firstly, the data blocks starting from @offset to the EOF are shifted by
+ * @len towards right to create a hole in the @inode. Secondly, the hole is
+ * filled with uninit extent(s). Inode size is increased by len bytes.
+ * Returns 0 on success, error otherwise.
+ */
+int ext4_insert_range(struct file *file, loff_t offset, loff_t len)
+{
+	struct inode *inode = file_inode(file);
+	struct super_block *sb = inode->i_sb;
+	handle_t *handle;
+	struct ext4_ext_path *path;
+	struct ext4_extent *extent;
+	ext4_lblk_t offset_lblk, len_lblk, ee_start_lblk, ee_last_lblk;
+	unsigned int credits, ee_len;
+	int ret = 0, depth, split_flag = 0;
+	loff_t ioffset;
+
+	/* Insert range works only on fs block size aligned offsets. */
+	if (offset & (EXT4_BLOCK_SIZE(sb) - 1) ||
+			len & (EXT4_BLOCK_SIZE(sb) - 1))
+		return -EINVAL;
+
+	if (!S_ISREG(inode->i_mode))
+		return -EOPNOTSUPP;
+
+	if (EXT4_SB(inode->i_sb)->s_cluster_ratio > 1)
+		return -EOPNOTSUPP;
+
+	trace_ext4_insert_range(inode, offset, len);
+
+	offset_lblk = offset >> EXT4_BLOCK_SIZE_BITS(sb);
+	len_lblk = len >> EXT4_BLOCK_SIZE_BITS(sb);
+
+	/* Call ext4_force_commit to flush all data in case of data=journal */
+	if (ext4_should_journal_data(inode)) {
+		ret = ext4_force_commit(inode->i_sb);
+		if (ret)
+			return ret;
+	}
+
+	/*
+	 * Need to round down to align start offset to page size boundary
+	 * for page size > block size.
+	 */
+	ioffset = round_down(offset, PAGE_SIZE);
+
+	/* Write out all dirty pages */
+	ret = filemap_write_and_wait_range(inode->i_mapping, ioffset,
+		LLONG_MAX);
+	if (ret)
+		return ret;
+
+	/* Take mutex lock */
+	mutex_lock(&inode->i_mutex);
+
+	/* Currently just for extent based files */
+	if (!ext4_test_inode_flag(inode, EXT4_INODE_EXTENTS)) {
+		ret = -EOPNOTSUPP;
+		goto out_mutex;
+	}
+
+	/* Check for wrap through zero */
+	if (inode->i_size + len > inode->i_sb->s_maxbytes) {
+		ret = -EFBIG;
+		goto out_mutex;
+	}
+
+	/* Offset should be less than i_size */
+	if (offset >= i_size_read(inode)) {
+		ret = -EINVAL;
+		goto out_mutex;
+	}
+
+	path = ext4_ext_find_extent(inode, EXT_MAX_BLOCKS - 1, NULL, 0);
+	if (IS_ERR(path)) {
+		ret = PTR_ERR(path);
+		goto out_mutex;
+	}
+
+	depth = path->p_depth;
+	extent = path[depth].p_ext;
+	if (extent) {
+		/*
+		 * Check if number of blocks of file shifted by insert range
+		 * can be bigger than EXT_MAX_BLOCKS at first.
+		 */
+		ee_last_lblk = le32_to_cpu(extent->ee_block) +
+			ext4_ext_get_actual_len(extent);
+		if (ee_last_lblk + len_lblk > EXT_MAX_BLOCKS - 1)
+			ret = -EINVAL;
+	}
+	ext4_ext_drop_refs(path);
+	kfree(path);
+	if (ret)
+		goto out_mutex;
+
+	truncate_pagecache(inode, ioffset);
+
+	/* Wait for existing dio to complete */
+	ext4_inode_block_unlocked_dio(inode);
+	inode_dio_wait(inode);
+
+	credits = ext4_writepage_trans_blocks(inode);
+	handle = ext4_journal_start(inode, EXT4_HT_TRUNCATE, credits);
+	if (IS_ERR(handle)) {
+		ret = PTR_ERR(handle);
+		goto out_dio;
+	}
+
+	/* Expand file to avoid data loss if there is error while shifting */
+	inode->i_size += len;
+	EXT4_I(inode)->i_disksize += len;
+	ret = ext4_mark_inode_dirty(handle, inode);
+	if (ret)
+		goto out_dio;
+
+	if (!extent)
+		/* Just allocate unwritten blocks and exit */
+		goto alloc_blocks;
+
+	down_write(&EXT4_I(inode)->i_data_sem);
+	ext4_discard_preallocations(inode);
+
+	path = ext4_ext_find_extent(inode, offset_lblk, NULL, 0);
+	if (IS_ERR(path))
+		goto out_sem;
+
+	depth = ext_depth(inode);
+	extent = path[depth].p_ext;
+	ee_start_lblk = le32_to_cpu(extent->ee_block);
+	ee_len = ext4_ext_get_actual_len(extent);
+
+	/*
+	 * If offset_lblk is not the starting block of extent, split
+	 * the extent @offset_lblk
+	 */
+	if ((offset_lblk > ee_start_lblk) &&
+	    (offset_lblk < (ee_start_lblk + ee_len))) {
+		if (ext4_ext_is_unwritten(extent))
+			split_flag = EXT4_EXT_MARK_UNWRIT1 |
+				     EXT4_EXT_MARK_UNWRIT2;
+		ret = ext4_split_extent_at(handle, inode, path, offset_lblk,
+					   split_flag, EXT4_EX_NOCACHE |
+					   EXT4_GET_BLOCKS_PRE_IO |
+					   EXT4_GET_BLOCKS_METADATA_NOFAIL);
+	}
+
+	ext4_ext_drop_refs(path);
+	kfree(path);
+	if (ret)
+		goto out_sem;
+
+	ret = ext4_es_remove_extent(inode, offset_lblk,
+			EXT_MAX_BLOCKS - offset_lblk);
+	if (ret)
+		goto out_sem;
+
+	/*
+	 * if offset_lblk lies in a hole which is at start of file, use
+	 * ee_start_lblk to shift extents
+	 */
+	ret = ext4_ext_shift_extents_right(inode, handle,
+		ee_start_lblk > offset_lblk ? ee_start_lblk : offset_lblk,
+		len_lblk);
+	if (ret)
+		goto out_sem;
+
+	up_write(&EXT4_I(inode)->i_data_sem);
+
+alloc_blocks:
+	inode->i_mtime = inode->i_ctime = ext4_current_time(inode);
+	ret = ext4_mark_inode_dirty(handle, inode);
+	if (ret)
+		goto out_stop;
+
+	if (IS_SYNC(inode))
+		ext4_handle_sync(handle);
+
+	ext4_journal_stop(handle);
+
+	ret = ext4_alloc_file_blocks(file, offset_lblk, len_lblk,
+			EXT4_GET_BLOCKS_CREATE_UNWRIT_EXT, 0);
+	goto out_dio;
+
+out_sem:
+	up_write(&EXT4_I(inode)->i_data_sem);
+out_stop:
+	ext4_journal_stop(handle);
+out_dio:
+	ext4_inode_resume_unlocked_dio(inode);
+out_mutex:
+	mutex_unlock(&inode->i_mutex);
+	return ret;
+}
diff --git a/include/trace/events/ext4.h b/include/trace/events/ext4.h
index d4f70a7..0b90106 100644
--- a/include/trace/events/ext4.h
+++ b/include/trace/events/ext4.h
@@ -2438,6 +2438,31 @@ TRACE_EVENT(ext4_collapse_range,
 		  __entry->offset, __entry->len)
 );
 
+TRACE_EVENT(ext4_insert_range,
+	TP_PROTO(struct inode *inode, loff_t offset, loff_t len),
+
+	TP_ARGS(inode, offset, len),
+
+	TP_STRUCT__entry(
+		__field(dev_t,	dev)
+		__field(ino_t,	ino)
+		__field(loff_t,	offset)
+		__field(loff_t, len)
+	),
+
+	TP_fast_assign(
+		__entry->dev	= inode->i_sb->s_dev;
+		__entry->ino	= inode->i_ino;
+		__entry->offset	= offset;
+		__entry->len	= len;
+	),
+
+	TP_printk("dev %d,%d ino %lu offset %lld len %lld",
+		  MAJOR(__entry->dev), MINOR(__entry->dev),
+		  (unsigned long) __entry->ino,
+		  __entry->offset, __entry->len)
+);
+
 #endif /* _TRACE_EXT4_H */
 
 /* This part must be outside protection */
-- 
1.7.11-rc0

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/