ocfs2 refcount tree is stored as an extent tree while
the leaf ocfs2_refcount_rec points to a refcount block.

The following step can trip a kernel panic.
mkfs.ocfs2 -b 512 -C 1M --fs-features=refcount $DEVICE
mount -t ocfs2 $DEVICE $MNT_DIR
FILE_NAME=$RANDOM
FILE_NAME_1=$RANDOM
FILE_REF="${FILE_NAME}_ref"
FILE_REF_1="${FILE_NAME}_ref_1"
for((i=0;i<305;i++))
do
# /mnt/1048576 is a file with 1048576 sizes.
cat /mnt/1048576 >> $MNT_DIR/$FILE_NAME
cat /mnt/1048576 >> $MNT_DIR/$FILE_NAME_1
done
for((i=0;i<3;i++))
do
cat /mnt/1048576 >> $MNT_DIR/$FILE_NAME
done

for((i=0;i<2;i++))
do
cat /mnt/1048576 >> $MNT_DIR/$FILE_NAME
cat /mnt/1048576 >> $MNT_DIR/$FILE_NAME_1
done

cat /mnt/1048576 >> $MNT_DIR/$FILE_NAME

for((i=0;i<11;i++))
do
cat /mnt/1048576 >> $MNT_DIR/$FILE_NAME
cat /mnt/1048576 >> $MNT_DIR/$FILE_NAME_1
done
reflink $MNT_DIR/$FILE_NAME $MNT_DIR/$FILE_REF
# write_f is a program which will write some bytes to a file at offset.
# write_f -f file_name -l offset -w write_bytes.
./write_f -f $MNT_DIR/$FILE_REF -l $[310*1048576] -w 4096
./write_f -f $MNT_DIR/$FILE_REF -l $[306*1048576] -w 4096
./write_f -f $MNT_DIR/$FILE_REF -l $[311*1048576] -w 4096
./write_f -f $MNT_DIR/$FILE_NAME -l $[310*1048576] -w 4096
./write_f -f $MNT_DIR/$FILE_NAME -l $[311*1048576] -w 4096
reflink $MNT_DIR/$FILE_NAME $MNT_DIR/$FILE_REF_1
./write_f -f $MNT_DIR/$FILE_NAME -l $[311*1048576] -w 4096
#kernel panic here.

The reason is that if the ocfs2_extent_rec is the last record
in a leaf extent block, the old solution fails to find the
suitable end cpos. So this patch try to walk through the b-tree,
find the next sub root and get the c_pos the next sub-tree starts
from.

btw, I have runned tristan's test case against the patched kernel
for several days and this type of kernel panic never happens again.
Signed-off-by: NTao Ma <tao.ma@oracle.com>
Signed-off-by: NJoel Becker <joel.becker@oracle.com>

The old reflink fails to handle inodes with inline data and will oops
if it encounters them.  This patch copies inline data to the new inode.
Extended attributes may still be refcounted.
Signed-off-by: NTao Ma <tao.ma@oracle.com>
Signed-off-by: NJoel Becker <joel.becker@oracle.com>
Tested-by: NTristan Ye <tristan.ye@oracle.com>

As its name ocfs2_complete_reflink indicates, it should
be called after all the work for reflink is done, so
it really should be called after we reflink xattr
successfully.
Signed-off-by: NTao Ma <tao.ma@oracle.com>
Signed-off-by: NJoel Becker <joel.becker@oracle.com>
Tested-by: NTristan Ye <tristan.ye@oracle.com>

The ioctl will take 3 parameters: old_path, new_path and
preserve and call vfs_reflink. It is useful when we backport
reflink features to old kernels.
Signed-off-by: NTao Ma <tao.ma@oracle.com>

Implement ocfs2_reflink.
Signed-off-by: NTao Ma <tao.ma@oracle.com>

reflink has 2 options for the destination file:
1. snapshot: reflink will attempt to preserve ownership, permissions,
   and all other security state in order to create a full snapshot.
2. new file: it will acquire the data extent sharing but will see the
   file's security state and attributes initialized as a new file.

So add the option to ocfs2.
Signed-off-by: NTao Ma <tao.ma@oracle.com>

Actually the whole reflink will touch refcount tree 2 times:
1. It will add the clusters in the extent record to the tree if it
   isn't refcounted before.
2. It will add 1 refcount to these clusters when it add these
   extent records to the tree.

So actually we shouldn't do merge in the 1st operation since the 2nd
one will soon be called and we may have to split it again. Do a merge
first and split soon is a waste of time. So we only merge in the 2nd
round. This is done by adding a new internal __ocfs2_increase_refcount
and call it with "not-merge" for 1st refcount operation in reflink.

This also has a side-effect that we don't need to worry too much about
the metadata allocation in the 2nd round since it will only merge and
no split will happen for those records.
Signed-off-by: NTao Ma <tao.ma@oracle.com>

Signed-off-by: NTao Ma <tao.ma@oracle.com>

Now with xattr refcount support, we need to check whether
we have xattr refcounted before we remove the refcount tree.

Now the mechanism is:
1) Check whether i_clusters == 0, if no, exit.
2) check whether we have i_xattr_loc in dinode. if yes, exit.
2) Check whether we have inline xattr stored outside, if yes, exit.
4) Remove the tree.
Signed-off-by: NTao Ma <tao.ma@oracle.com>

In ocfs2, when xattr's value is larger than OCFS2_XATTR_INLINE_SIZE,
it will be kept outside of the blocks we store xattr entry. And they
are stored in a b-tree also. So this patch try to attach all these
clusters to refcount tree also.
Signed-off-by: NTao Ma <tao.ma@oracle.com>

In order to make 2 transcation(xattr and cow) independent with each other,
we CoW the whole xattr out in case we are setting them.
Signed-off-by: NTao Ma <tao.ma@oracle.com>

We currently use pagecache to duplicate clusters in CoW,
but it isn't suitable for xattr case. So abstract it out
so that the caller can decide which method it use.
Signed-off-by: NTao Ma <tao.ma@oracle.com>

A reflink creates a snapshot of a file, that means the attributes
must be identical except for three exceptions - nlink, ino, and ctime.

As for time changes, Here is a brief description:

1. Source file:
   1) atime: Ignore. Let the lazy atime code handle that.
   2) mtime: don't touch.
   3) ctime: If we change the tree (adding REFCOUNTED to at least one
             extent), update it.
2. Destination file:
   1) atime: ignore.
   2) mtime: we want it to appear identical to the source.
   3) ctime: update.

The idea here is that an ls -l will show the same time for the
src and target - it shows mtime.  Backup software like rsync and tar
will treat the new file correctly too.
Signed-off-by: NTao Ma <tao.ma@oracle.com>

2 major functions are added in this patch.

ocfs2_attach_refcount_tree will create a new refcount tree to the
old file if it doesn't have one and insert all the extent records
to the tree if they are not refcounted.

ocfs2_create_reflink_node will:
1. set the refcount tree to the new file.
2. call ocfs2_duplicate_extent_list which will iterate all the
   extents for the old file, insert it to the new file and increase
   the corresponding referennce count.
Signed-off-by: NTao Ma <tao.ma@oracle.com>

When we truncate a file to a specific size which resides in a reflinked
cluster, we need to CoW it since ocfs2_zero_range_for_truncate will
zero the space after the size(just another type of write).

So we add a "max_cpos" in ocfs2_refcount_cow so that it will stop when
it hit the max cluster offset.
Signed-off-by: NTao Ma <tao.ma@oracle.com>

During CoW, if the old extent record is refcounted, we allocate
som new clusters and do CoW. Actually we can have some improvement
here. If the old extent has refcount=1, that means now it is only
used by this file. So we don't need to allocate new clusters, just
remove the refcounted flag and it is OK. We also have to remove
it from the refcount tree while not deleting it.
Signed-off-by: NTao Ma <tao.ma@oracle.com>

This patch try CoW support for a refcounted record.

the whole process will be:
1. Calculate how many clusters we need to CoW and where we start.
   Extents that are not completely encompassed by the write will
   be broken on 1MB boundaries.
2. Do CoW for the clusters with the help of page cache.
3. Change the b-tree structure with the new allocated clusters.
Signed-off-by: NTao Ma <tao.ma@oracle.com>

Add 'Decrement refcount for delete' in to the normal truncate
process. So for a refcounted extent record, call refcount rec
decrementation instead of cluster free.
Signed-off-by: NTao Ma <tao.ma@oracle.com>

Add function ocfs2_mark_extent_refcounted which can mark
an extent refcounted.
Signed-off-by: NTao Ma <tao.ma@oracle.com>

    Given a physical cpos and length, decrement the refcount
in the tree. If the refcount for any portion of the extent goes
to zero, that portion is queued for freeing.
Signed-off-by: NTao Ma <tao.ma@oracle.com>

    Given a physical cpos and length, increment the refcount
in the tree. If the extent has not been seen before, a refcount
record is created for it. Refcount records may be merged or
split by this operation.
Signed-off-by: NTao Ma <tao.ma@oracle.com>

Add basic refcount tree root operation.
Signed-off-by: NTao Ma <tao.ma@oracle.com>

Implement locking around struct ocfs2_refcount_tree.  This protects
all read/write operations on refcount trees.  ocfs2_refcount_tree
has its own lock and its own caching_info, protecting buffers among
multiple nodes.

User must call ocfs2_lock_refcount_tree before his operation on
the tree and unlock it after that.

ocfs2_refcount_trees are referenced by the block number of the
refcount tree root block, So we create an rb-tree on the ocfs2_super
to look them up.
Signed-off-by: NTao Ma <tao.ma@oracle.com>

refcount tree should use its own caching info so that when
we downconvert the refcount tree lock, we can drop all the
cached buffer head.
Signed-off-by: NTao Ma <tao.ma@oracle.com>

Signed-off-by: NTao Ma <tao.ma@oracle.com>