Btrfs: Allow subvolumes and snapshots anywhere in the directory tree
Before, all snapshots and subvolumes lived in a single flat directory. This
was awkward and confusing because the single flat directory was only writable
with the ioctls.
This commit changes the ioctls to create subvols and snapshots at any
point in the directory tree. This requires making separate ioctls for
snapshot and subvol creation instead of a combining them into one.
The subvol ioctl does:
btrfsctl -S subvol_name parent_dir
After the ioctl is done subvol_name lives inside parent_dir.
The snapshot ioctl does:
btrfsctl -s path_for_snapshot root_to_snapshot
path_for_snapshot can be an absolute or relative path. btrfsctl breaks it up
into directory and basename components.
root_to_snapshot can be any file or directory in the FS. The snapshot
is taken of the entire root where that file lives.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h
index 5ff7428..5611f8e 100644
--- a/fs/btrfs/ctree.h
+++ b/fs/btrfs/ctree.h
@@ -606,6 +606,7 @@
struct btrfs_root *tree_root;
struct btrfs_root *chunk_root;
struct btrfs_root *dev_root;
+ struct btrfs_root *fs_root;
/* the log root tree is a directory of all the other log roots */
struct btrfs_root *log_root_tree;
@@ -758,7 +759,6 @@
struct btrfs_root_item root_item;
struct btrfs_key root_key;
struct btrfs_fs_info *fs_info;
- struct inode *inode;
struct extent_io_tree dirty_log_pages;
struct kobject root_kobj;
@@ -1876,6 +1876,8 @@
#define PageChecked PageFsMisc
#endif
+struct inode *btrfs_lookup_dentry(struct inode *dir, struct dentry *dentry);
+int btrfs_set_inode_index(struct inode *dir, u64 *index);
int btrfs_unlink_inode(struct btrfs_trans_handle *trans,
struct btrfs_root *root,
struct inode *dir, struct inode *inode,
@@ -1896,9 +1898,6 @@
struct btrfs_trans_handle *trans, u64 new_dirid,
struct btrfs_block_group_cache *block_group);
-void btrfs_invalidate_dcache_root(struct btrfs_root *root, char *name,
- int namelen);
-
int btrfs_merge_bio_hook(struct page *page, unsigned long offset,
size_t size, struct bio *bio, unsigned long bio_flags);
diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c
index 82833e5..0a53505 100644
--- a/fs/btrfs/disk-io.c
+++ b/fs/btrfs/disk-io.c
@@ -838,7 +838,6 @@
u64 objectid)
{
root->node = NULL;
- root->inode = NULL;
root->commit_root = NULL;
root->ref_tree = NULL;
root->sectorsize = sectorsize;
@@ -1430,6 +1429,7 @@
u32 blocksize;
u32 stripesize;
u64 generation;
+ struct btrfs_key location;
struct buffer_head *bh;
struct btrfs_root *extent_root = kzalloc(sizeof(struct btrfs_root),
GFP_NOFS);
@@ -1729,7 +1729,7 @@
goto fail_cleaner;
if (sb->s_flags & MS_RDONLY)
- return tree_root;
+ goto read_fs_root;
if (btrfs_super_log_root(disk_super) != 0) {
u32 blocksize;
@@ -1755,6 +1755,14 @@
ret = btrfs_cleanup_reloc_trees(tree_root);
BUG_ON(ret);
+ location.objectid = BTRFS_FS_TREE_OBJECTID;
+ location.type = BTRFS_ROOT_ITEM_KEY;
+ location.offset = (u64)-1;
+
+read_fs_root:
+ fs_info->fs_root = btrfs_read_fs_root_no_name(fs_info, &location);
+ if (!fs_info->fs_root)
+ goto fail_cleaner;
return tree_root;
fail_cleaner:
@@ -1944,8 +1952,6 @@
(unsigned long)root->root_key.objectid);
if (root->in_sysfs)
btrfs_sysfs_del_root(root);
- if (root->inode)
- iput(root->inode);
if (root->node)
free_extent_buffer(root->node);
if (root->commit_root)
diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c
index 3e3620e..e163b1b 100644
--- a/fs/btrfs/inode.c
+++ b/fs/btrfs/inode.c
@@ -3038,8 +3038,7 @@
return inode;
}
-static struct dentry *btrfs_lookup(struct inode *dir, struct dentry *dentry,
- struct nameidata *nd)
+struct inode *btrfs_lookup_dentry(struct inode *dir, struct dentry *dentry)
{
struct inode * inode;
struct btrfs_inode *bi = BTRFS_I(dir);
@@ -3067,13 +3066,21 @@
inode = btrfs_iget(dir->i_sb, &location, sub_root, &new);
if (IS_ERR(inode))
return ERR_CAST(inode);
-
- /* the inode and parent dir are two different roots */
- if (new && root != sub_root) {
- igrab(inode);
- sub_root->inode = inode;
- }
}
+ return inode;
+}
+
+static struct dentry *btrfs_lookup(struct inode *dir, struct dentry *dentry,
+ struct nameidata *nd)
+{
+ struct inode *inode;
+
+ if (dentry->d_name.len > BTRFS_NAME_LEN)
+ return ERR_PTR(-ENAMETOOLONG);
+
+ inode = btrfs_lookup_dentry(dir, dentry);
+ if (IS_ERR(inode))
+ return ERR_CAST(inode);
return d_splice_alias(inode, dentry);
}
@@ -3129,7 +3136,6 @@
return 0;
filp->f_pos = 2;
}
-
path = btrfs_alloc_path();
path->reada = 2;
@@ -3159,6 +3165,7 @@
path->slots[0]++;
}
}
+
advance = 1;
item = btrfs_item_nr(leaf, slot);
btrfs_item_key_to_cpu(leaf, &found_key, slot);
@@ -3194,16 +3201,25 @@
d_type = btrfs_filetype_table[btrfs_dir_type(leaf, di)];
btrfs_dir_item_key_to_cpu(leaf, di, &location);
+
+ /* is this a reference to our own snapshot? If so
+ * skip it
+ */
+ if (location.type == BTRFS_ROOT_ITEM_KEY &&
+ location.objectid == root->root_key.objectid) {
+ over = 0;
+ goto skip;
+ }
over = filldir(dirent, name_ptr, name_len,
found_key.offset, location.objectid,
d_type);
+skip:
if (name_ptr != tmp_name)
kfree(name_ptr);
if (over)
goto nopos;
-
di_len = btrfs_dir_name_len(leaf, di) +
btrfs_dir_data_len(leaf, di) + sizeof(*di);
di_cur += di_len;
@@ -3318,8 +3334,7 @@
* helper to find a free sequence number in a given directory. This current
* code is very simple, later versions will do smarter things in the btree
*/
-static int btrfs_set_inode_index(struct inode *dir, struct inode *inode,
- u64 *index)
+int btrfs_set_inode_index(struct inode *dir, u64 *index)
{
int ret = 0;
@@ -3365,7 +3380,7 @@
return ERR_PTR(-ENOMEM);
if (dir) {
- ret = btrfs_set_inode_index(dir, inode, index);
+ ret = btrfs_set_inode_index(dir, index);
if (ret)
return ERR_PTR(ret);
}
@@ -3651,7 +3666,7 @@
err = btrfs_check_free_space(root, 1, 0);
if (err)
goto fail;
- err = btrfs_set_inode_index(dir, inode, &index);
+ err = btrfs_set_inode_index(dir, &index);
if (err)
goto fail;
@@ -4349,13 +4364,13 @@
* Invalidate a single dcache entry at the root of the filesystem.
* Needed after creation of snapshot or subvolume.
*/
-void btrfs_invalidate_dcache_root(struct btrfs_root *root, char *name,
+void btrfs_invalidate_dcache_root(struct inode *dir, char *name,
int namelen)
{
struct dentry *alias, *entry;
struct qstr qstr;
- alias = d_find_alias(root->fs_info->sb->s_root->d_inode);
+ alias = d_find_alias(dir);
if (alias) {
qstr.name = name;
qstr.len = namelen;
@@ -4387,7 +4402,6 @@
return PTR_ERR(inode);
inode->i_op = &btrfs_dir_inode_operations;
inode->i_fop = &btrfs_dir_file_operations;
- new_root->inode = inode;
inode->i_nlink = 1;
btrfs_i_size_write(inode, 0);
@@ -4590,7 +4604,7 @@
}
}
- ret = btrfs_set_inode_index(new_dir, old_inode, &index);
+ ret = btrfs_set_inode_index(new_dir, &index);
if (ret)
goto out_fail;
diff --git a/fs/btrfs/ioctl.c b/fs/btrfs/ioctl.c
index f43df72..ec45b30 100644
--- a/fs/btrfs/ioctl.c
+++ b/fs/btrfs/ioctl.c
@@ -67,6 +67,7 @@
int err;
u64 objectid;
u64 new_dirid = BTRFS_FIRST_FREE_OBJECTID;
+ u64 index = 0;
unsigned long nr = 1;
ret = btrfs_check_free_space(root, 1, 0);
@@ -126,6 +127,7 @@
key.objectid = objectid;
key.offset = 1;
btrfs_set_key_type(&key, BTRFS_ROOT_ITEM_KEY);
+printk("inserting root objectid %Lu\n", objectid);
ret = btrfs_insert_root(trans, root->fs_info->tree_root, &key,
&root_item);
if (ret)
@@ -135,24 +137,27 @@
* insert the directory item
*/
key.offset = (u64)-1;
- dir = root->fs_info->sb->s_root->d_inode;
- ret = btrfs_insert_dir_item(trans, root->fs_info->tree_root,
+ dir = dentry->d_parent->d_inode;
+ ret = btrfs_set_inode_index(dir, &index);
+ BUG_ON(ret);
+
+ ret = btrfs_insert_dir_item(trans, root,
name, namelen, dir->i_ino, &key,
- BTRFS_FT_DIR, 0);
+ BTRFS_FT_DIR, index);
if (ret)
goto fail;
-
+#if 0
ret = btrfs_insert_inode_ref(trans, root->fs_info->tree_root,
name, namelen, objectid,
root->fs_info->sb->s_root->d_inode->i_ino, 0);
if (ret)
goto fail;
-
+#endif
ret = btrfs_commit_transaction(trans, root);
if (ret)
goto fail_commit;
- new_root = btrfs_read_fs_root(root->fs_info, &key, name, namelen);
+ new_root = btrfs_read_fs_root_no_name(root->fs_info, &key);
BUG_ON(!new_root);
trans = btrfs_start_transaction(new_root, 1);
@@ -170,14 +175,16 @@
ret = err;
fail_commit:
btrfs_btree_balance_dirty(root, nr);
+printk("all done ret %d\n", ret);
return ret;
}
-static int create_snapshot(struct btrfs_root *root, char *name, int namelen)
+static int create_snapshot(struct btrfs_root *root, struct dentry *dentry,
+ char *name, int namelen)
{
struct btrfs_pending_snapshot *pending_snapshot;
struct btrfs_trans_handle *trans;
- int ret;
+ int ret = 0;
int err;
unsigned long nr = 0;
@@ -188,7 +195,7 @@
if (ret)
goto fail_unlock;
- pending_snapshot = kmalloc(sizeof(*pending_snapshot), GFP_NOFS);
+ pending_snapshot = kzalloc(sizeof(*pending_snapshot), GFP_NOFS);
if (!pending_snapshot) {
ret = -ENOMEM;
goto fail_unlock;
@@ -201,12 +208,12 @@
}
memcpy(pending_snapshot->name, name, namelen);
pending_snapshot->name[namelen] = '\0';
+ pending_snapshot->dentry = dentry;
trans = btrfs_start_transaction(root, 1);
BUG_ON(!trans);
pending_snapshot->root = root;
list_add(&pending_snapshot->list,
&trans->transaction->pending_snapshots);
- ret = btrfs_update_inode(trans, root, root->inode);
err = btrfs_commit_transaction(trans, root);
fail_unlock:
@@ -230,7 +237,8 @@
* inside this filesystem so it's quite a bit simpler.
*/
static noinline int btrfs_mksubvol(struct path *parent, char *name,
- int mode, int namelen)
+ int mode, int namelen,
+ struct btrfs_root *snap_src)
{
struct dentry *dentry;
int error;
@@ -248,6 +256,7 @@
if (!IS_POSIXACL(parent->dentry->d_inode))
mode &= ~current->fs->umask;
+
error = mnt_want_write(parent->mnt);
if (error)
goto out_dput;
@@ -266,8 +275,12 @@
* Also we should pass on the mode eventually to allow creating new
* subvolume with specific mode bits.
*/
- error = create_subvol(BTRFS_I(parent->dentry->d_inode)->root, dentry,
- name, namelen);
+ if (snap_src) {
+ error = create_snapshot(snap_src, dentry, name, namelen);
+ } else {
+ error = create_subvol(BTRFS_I(parent->dentry->d_inode)->root,
+ dentry, name, namelen);
+ }
if (error)
goto out_drop_write;
@@ -471,15 +484,16 @@
}
static noinline int btrfs_ioctl_snap_create(struct file *file,
- void __user *arg)
+ void __user *arg, int subvol)
{
struct btrfs_root *root = BTRFS_I(fdentry(file)->d_inode)->root;
struct btrfs_ioctl_vol_args *vol_args;
struct btrfs_dir_item *di;
struct btrfs_path *path;
+ struct file *src_file;
u64 root_dirid;
int namelen;
- int ret;
+ int ret = 0;
if (root->fs_info->sb->s_flags & MS_RDONLY)
return -EROFS;
@@ -523,12 +537,29 @@
goto out;
}
- if (root == root->fs_info->tree_root) {
+ if (subvol) {
ret = btrfs_mksubvol(&file->f_path, vol_args->name,
file->f_path.dentry->d_inode->i_mode,
- namelen);
+ namelen, NULL);
} else {
- ret = create_snapshot(root, vol_args->name, namelen);
+ struct inode *src_inode;
+ src_file = fget(vol_args->fd);
+ if (!src_file) {
+ ret = -EINVAL;
+ goto out;
+ }
+
+ src_inode = src_file->f_path.dentry->d_inode;
+ if (src_inode->i_sb != file->f_path.dentry->d_inode->i_sb) {
+ printk("btrfs: Snapshot src from another FS\n");
+ ret = -EINVAL;
+ fput(src_file);
+ goto out;
+ }
+ ret = btrfs_mksubvol(&file->f_path, vol_args->name,
+ file->f_path.dentry->d_inode->i_mode,
+ namelen, BTRFS_I(src_inode)->root);
+ fput(src_file);
}
out:
@@ -1030,7 +1061,9 @@
switch (cmd) {
case BTRFS_IOC_SNAP_CREATE:
- return btrfs_ioctl_snap_create(file, (void __user *)arg);
+ return btrfs_ioctl_snap_create(file, (void __user *)arg, 0);
+ case BTRFS_IOC_SUBVOL_CREATE:
+ return btrfs_ioctl_snap_create(file, (void __user *)arg, 1);
case BTRFS_IOC_DEFRAG:
return btrfs_ioctl_defrag(file);
case BTRFS_IOC_RESIZE:
diff --git a/fs/btrfs/ioctl.h b/fs/btrfs/ioctl.h
index 989ba8a..78049ea 100644
--- a/fs/btrfs/ioctl.h
+++ b/fs/btrfs/ioctl.h
@@ -22,9 +22,10 @@
#define BTRFS_IOCTL_MAGIC 0x94
#define BTRFS_VOL_NAME_MAX 255
-#define BTRFS_PATH_NAME_MAX 4095
+#define BTRFS_PATH_NAME_MAX 3072
struct btrfs_ioctl_vol_args {
+ __s64 fd;
char name[BTRFS_PATH_NAME_MAX + 1];
};
@@ -51,7 +52,6 @@
struct btrfs_ioctl_vol_args)
#define BTRFS_IOC_BALANCE _IOW(BTRFS_IOCTL_MAGIC, 12, \
struct btrfs_ioctl_vol_args)
-
struct btrfs_ioctl_clone_range_args {
__s64 src_fd;
__u64 src_offset, src_length;
@@ -61,4 +61,7 @@
#define BTRFS_IOC_CLONE_RANGE _IOW(BTRFS_IOCTL_MAGIC, 13, \
struct btrfs_ioctl_clone_range_args)
+#define BTRFS_IOC_SUBVOL_CREATE _IOW(BTRFS_IOCTL_MAGIC, 14, \
+ struct btrfs_ioctl_vol_args)
+
#endif
diff --git a/fs/btrfs/super.c b/fs/btrfs/super.c
index 92393cc..77c5eff 100644
--- a/fs/btrfs/super.c
+++ b/fs/btrfs/super.c
@@ -285,11 +285,11 @@
out:
/*
* If no subvolume name is specified we use the default one. Allocate
- * a copy of the string "default" here so that code later in the
+ * a copy of the string "." here so that code later in the
* mount path doesn't care if it's the default volume or another one.
*/
if (!*subvol_name) {
- *subvol_name = kstrdup("default", GFP_KERNEL);
+ *subvol_name = kstrdup(".", GFP_KERNEL);
if (!*subvol_name)
return -ENOMEM;
}
@@ -323,12 +323,12 @@
}
sb->s_fs_info = tree_root;
disk_super = &tree_root->fs_info->super_copy;
- inode = btrfs_iget_locked(sb, btrfs_super_root_dir(disk_super),
- tree_root);
+ inode = btrfs_iget_locked(sb, BTRFS_FIRST_FREE_OBJECTID,
+ tree_root->fs_info->fs_root);
bi = BTRFS_I(inode);
bi->location.objectid = inode->i_ino;
bi->location.offset = 0;
- bi->root = tree_root;
+ bi->root = tree_root->fs_info->fs_root;
btrfs_set_key_type(&bi->location, BTRFS_INODE_ITEM_KEY);
diff --git a/fs/btrfs/transaction.c b/fs/btrfs/transaction.c
index 202c1b6..eec8b24 100644
--- a/fs/btrfs/transaction.c
+++ b/fs/btrfs/transaction.c
@@ -779,7 +779,6 @@
struct extent_buffer *tmp;
struct extent_buffer *old;
int ret;
- int namelen;
u64 objectid;
new_root_item = kmalloc(sizeof(*new_root_item), GFP_NOFS);
@@ -816,28 +815,48 @@
if (ret)
goto fail;
+ key.offset = (u64)-1;
+ memcpy(&pending->root_key, &key, sizeof(key));
+fail:
+ kfree(new_root_item);
+ return ret;
+}
+
+static noinline int finish_pending_snapshot(struct btrfs_fs_info *fs_info,
+ struct btrfs_pending_snapshot *pending)
+{
+ int ret;
+ int namelen;
+ u64 index = 0;
+ struct btrfs_trans_handle *trans;
+ struct inode *parent_inode;
+ struct inode *inode;
+
+ trans = btrfs_start_transaction(fs_info->fs_root, 1);
+
/*
* insert the directory item
*/
- key.offset = (u64)-1;
namelen = strlen(pending->name);
- ret = btrfs_insert_dir_item(trans, root->fs_info->tree_root,
- pending->name, namelen,
- root->fs_info->sb->s_root->d_inode->i_ino,
- &key, BTRFS_FT_DIR, 0);
+ parent_inode = pending->dentry->d_parent->d_inode;
+ ret = btrfs_set_inode_index(parent_inode, &index);
+ ret = btrfs_insert_dir_item(trans,
+ BTRFS_I(parent_inode)->root,
+ pending->name, namelen,
+ parent_inode->i_ino,
+ &pending->root_key, BTRFS_FT_DIR, index);
if (ret)
goto fail;
-
+#if 0
ret = btrfs_insert_inode_ref(trans, root->fs_info->tree_root,
pending->name, strlen(pending->name), objectid,
root->fs_info->sb->s_root->d_inode->i_ino, 0);
-
- /* Invalidate existing dcache entry for new snapshot. */
- btrfs_invalidate_dcache_root(root, pending->name, namelen);
-
+#endif
+ inode = btrfs_lookup_dentry(parent_inode, pending->dentry);
+ d_instantiate(pending->dentry, inode);
fail:
- kfree(new_root_item);
+ btrfs_end_transaction(trans, fs_info->fs_root);
return ret;
}
@@ -849,12 +868,28 @@
{
struct btrfs_pending_snapshot *pending;
struct list_head *head = &trans->transaction->pending_snapshots;
+ struct list_head *cur;
+ int ret;
+
+ list_for_each(cur, head) {
+ pending = list_entry(cur, struct btrfs_pending_snapshot, list);
+ ret = create_pending_snapshot(trans, fs_info, pending);
+ BUG_ON(ret);
+ }
+ return 0;
+}
+
+static noinline int finish_pending_snapshots(struct btrfs_trans_handle *trans,
+ struct btrfs_fs_info *fs_info)
+{
+ struct btrfs_pending_snapshot *pending;
+ struct list_head *head = &trans->transaction->pending_snapshots;
int ret;
while(!list_empty(head)) {
pending = list_entry(head->next,
struct btrfs_pending_snapshot, list);
- ret = create_pending_snapshot(trans, fs_info, pending);
+ ret = finish_pending_snapshot(fs_info, pending);
BUG_ON(ret);
list_del(&pending->list);
kfree(pending->name);
@@ -1033,11 +1068,15 @@
btrfs_drop_dead_reloc_roots(root);
mutex_unlock(&root->fs_info->tree_reloc_mutex);
+ /* do the directory inserts of any pending snapshot creations */
+ finish_pending_snapshots(trans, root->fs_info);
+
mutex_lock(&root->fs_info->trans_mutex);
cur_trans->commit_done = 1;
root->fs_info->last_trans_committed = cur_trans->transid;
wake_up(&cur_trans->commit_wait);
+
put_transaction(cur_trans);
put_transaction(cur_trans);
@@ -1046,6 +1085,7 @@
list_splice_init(&root->fs_info->dead_roots, &dirty_fs_roots);
mutex_unlock(&root->fs_info->trans_mutex);
+
kmem_cache_free(btrfs_trans_handle_cachep, trans);
if (root->fs_info->closing) {
diff --git a/fs/btrfs/transaction.h b/fs/btrfs/transaction.h
index eef2cb7..202c8be 100644
--- a/fs/btrfs/transaction.h
+++ b/fs/btrfs/transaction.h
@@ -47,8 +47,10 @@
};
struct btrfs_pending_snapshot {
+ struct dentry *dentry;
struct btrfs_root *root;
char *name;
+ struct btrfs_key root_key;
struct list_head list;
};