Btrfs: transaction ioctls

These ioctls let a user application hold a transaction open while it
performs a series of operations.  A final ioctl does a sync on the fs
(closing the current transaction).  This is the main requirement for
Ceph's OSD to be able to keep the data it's storing in a btrfs volume
consistent, and AFAICS it works just fine.  The application would do
something like

	fd = ::open("some/file", O_RDONLY);
	::ioctl(fd, BTRFS_IOC_TRANS_START);
	/* do a bunch of stuff */
	::ioctl(fd, BTRFS_IOC_TRANS_END);
or just
	::close(fd);

And to ensure it commits to disk,

	::ioctl(fd, BTRFS_IOC_SYNC);

When a transaction is held open, the trans_handle is attached to the
struct file (via private_data) so that it will get cleaned up if the
process dies unexpectedly.  A held transaction is also ended on fsync() to
avoid a deadlock.

A misbehaving application could also deliberately hold a transaction open,
effectively locking up the FS, so it may make sense to restrict something
like this to root or something.
Signed-off-by: Chris Mason <chris.mason@oracle.com>

fs/btrfs/ctree.h

@@ -1575,6 +1575,7 @@ struct inode *btrfs_alloc_inode(struct super_block *sb);
 void btrfs_destroy_inode(struct inode *inode);
 int btrfs_init_cachep(void);
 void btrfs_destroy_cachep(void);
+long btrfs_ioctl_trans_end(struct file *file);
 long btrfs_ioctl(struct file *file, unsigned int cmd, unsigned long arg);
 struct inode *btrfs_iget_locked(struct super_block *s, u64 objectid,
 				struct btrfs_root *root);
@@ -1595,6 +1596,8 @@ extern struct file_operations btrfs_file_operations;
 int btrfs_drop_extents(struct btrfs_trans_handle *trans,
 		       struct btrfs_root *root, struct inode *inode,
 		       u64 start, u64 end, u64 inline_limit, u64 *hint_block);
+int btrfs_release_file(struct inode *inode, struct file *file);
+
 /* tree-defrag.c */
 int btrfs_defrag_leaves(struct btrfs_trans_handle *trans,
 			struct btrfs_root *root, int cache_only);
@@ -1615,4 +1618,5 @@ int btrfs_delete_xattrs(struct btrfs_trans_handle *trans,
 u64 btrfs_parse_size(char *str);
 int btrfs_parse_options(char *options, struct btrfs_root *root,
 			char **subvol_name);
+int btrfs_sync_fs(struct super_block *sb, int wait);
 #endif

fs/btrfs/file.c

@@ -978,9 +978,11 @@ out_nolock:
 	return num_written ? num_written : err;
 }
 
-static int btrfs_release_file (struct inode * inode, struct file * filp)
+int btrfs_release_file(struct inode * inode, struct file * filp)
 {
 	btrfs_del_ordered_inode(inode);
+	if (filp->private_data)
+		btrfs_ioctl_trans_end(filp);
 	return 0;
 }
 
@@ -1011,6 +1013,9 @@ static int btrfs_sync_file(struct file *file,
 	/*
 	 * ok we haven't committed the transaction yet, lets do a commit
 	 */
+	if (file->private_data)
+		btrfs_ioctl_trans_end(file);
+
 	trans = btrfs_start_transaction(root, 1);
 	if (!trans) {
 		ret = -ENOMEM;

fs/btrfs/inode.c

@@ -3336,6 +3336,61 @@ out_fput:
 	return ret;
 }
 
+/*
+ * there are many ways the trans_start and trans_end ioctls can lead
+ * to deadlocks.  They should only be used by applications that
+ * basically own the machine, and have a very in depth understanding
+ * of all the possible deadlocks and enospc problems.
+ */
+long btrfs_ioctl_trans_start(struct file *file)
+{
+	struct inode *inode = fdentry(file)->d_inode;
+	struct btrfs_root *root = BTRFS_I(inode)->root;
+	struct btrfs_trans_handle *trans;
+	int ret = 0;
+
+	mutex_lock(&root->fs_info->fs_mutex);
+	if (file->private_data) {
+		ret = -EINPROGRESS;
+		goto out;
+	}
+	trans = btrfs_start_transaction(root, 0);
+	if (trans)
+		file->private_data = trans;
+	else
+		ret = -ENOMEM;
+	/*printk(KERN_INFO "btrfs_ioctl_trans_start on %p\n", file);*/
+out:
+	mutex_unlock(&root->fs_info->fs_mutex);
+	return ret;
+}
+
+/*
+ * there are many ways the trans_start and trans_end ioctls can lead
+ * to deadlocks.  They should only be used by applications that
+ * basically own the machine, and have a very in depth understanding
+ * of all the possible deadlocks and enospc problems.
+ */
+long btrfs_ioctl_trans_end(struct file *file)
+{
+	struct inode *inode = fdentry(file)->d_inode;
+	struct btrfs_root *root = BTRFS_I(inode)->root;
+	struct btrfs_trans_handle *trans;
+	int ret = 0;
+
+	mutex_lock(&root->fs_info->fs_mutex);
+	trans = file->private_data;
+	if (!trans) {
+		ret = -EINVAL;
+		goto out;
+	}
+	btrfs_end_transaction(trans, root);
+	file->private_data = 0;
+out:
+	mutex_unlock(&root->fs_info->fs_mutex);
+	return ret;
+}
+
 long btrfs_ioctl(struct file *file, unsigned int
 		cmd, unsigned long arg)
 {
@@ -3356,6 +3411,13 @@ long btrfs_ioctl(struct file *file, unsigned int
 		return btrfs_balance(root->fs_info->dev_root);
 	case BTRFS_IOC_CLONE:
 		return btrfs_ioctl_clone(file, arg);
+	case BTRFS_IOC_TRANS_START:
+		return btrfs_ioctl_trans_start(file);
+	case BTRFS_IOC_TRANS_END:
+		return btrfs_ioctl_trans_end(file);
+	case BTRFS_IOC_SYNC:
+		btrfs_sync_fs(file->f_dentry->d_sb, 1);
+		return 0;
 	}
 
 	return -ENOTTY;
@@ -3679,6 +3741,7 @@ static struct file_operations btrfs_dir_file_operations = {
 #ifdef CONFIG_COMPAT
 	.compat_ioctl	= btrfs_ioctl,
 #endif
+	.release        = btrfs_release_file,
 };
 
 static struct extent_io_ops btrfs_extent_io_ops = {

fs/btrfs/ioctl.h

@@ -36,6 +36,14 @@ struct btrfs_ioctl_vol_args {
 				   struct btrfs_ioctl_vol_args)
 #define BTRFS_IOC_SCAN_DEV _IOW(BTRFS_IOCTL_MAGIC, 4, \
 				   struct btrfs_ioctl_vol_args)
+/* trans start and trans end are dangerous, and only for
+ * use by applications that know how to avoid the
+ * resulting deadlocks
+ */
+#define BTRFS_IOC_TRANS_START  _IO(BTRFS_IOCTL_MAGIC, 6)
+#define BTRFS_IOC_TRANS_END    _IO(BTRFS_IOCTL_MAGIC, 7)
+#define BTRFS_IOC_SYNC         _IO(BTRFS_IOCTL_MAGIC, 8)
+
 #define BTRFS_IOC_CLONE        _IOW(BTRFS_IOCTL_MAGIC, 9, int)
 #define BTRFS_IOC_ADD_DEV _IOW(BTRFS_IOCTL_MAGIC, 10, \
 				   struct btrfs_ioctl_vol_args)
@@ -43,4 +51,5 @@ struct btrfs_ioctl_vol_args {
 				   struct btrfs_ioctl_vol_args)
 #define BTRFS_IOC_BALANCE _IOW(BTRFS_IOCTL_MAGIC, 12, \
 				   struct btrfs_ioctl_vol_args)
+
 #endif

fs/btrfs/super.c

@@ -293,7 +293,7 @@ fail_close:
 	return err;
 }
 
-static int btrfs_sync_fs(struct super_block *sb, int wait)
+int btrfs_sync_fs(struct super_block *sb, int wait)
 {
 	struct btrfs_trans_handle *trans;
 	struct btrfs_root *root;