[deliverable/linux.git] / fs / btrfs / ioctl.c

/*
 * Copyright (C) 2007 Oracle.  All rights reserved.
 *
 * This program is free software; you can redistribute it and/or
 * modify it under the terms of the GNU General Public
 * License v2 as published by the Free Software Foundation.
 *
 * This program is distributed in the hope that it will be useful,
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
 * General Public License for more details.
 *
 * You should have received a copy of the GNU General Public
 * License along with this program; if not, write to the
 * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
 * Boston, MA 021110-1307, USA.
 */

#include <linux/kernel.h>
#include <linux/bio.h>
#include <linux/buffer_head.h>
#include <linux/file.h>
#include <linux/fs.h>
#include <linux/pagemap.h>
#include <linux/highmem.h>
#include <linux/time.h>
#include <linux/init.h>
#include <linux/string.h>
#include <linux/smp_lock.h>
#include <linux/backing-dev.h>
#include <linux/mpage.h>
#include <linux/swap.h>
#include <linux/writeback.h>
#include <linux/statfs.h>
#include <linux/compat.h>
#include <linux/bit_spinlock.h>
#include <linux/version.h>
#include <linux/xattr.h>
#include "ctree.h"
#include "disk-io.h"
#include "transaction.h"
#include "btrfs_inode.h"
#include "ioctl.h"
#include "print-tree.h"
#include "volumes.h"
#include "locking.h"


static noinline int create_subvol(struct btrfs_root *root, char *name,
				  int namelen)
{
	struct btrfs_trans_handle *trans;
	struct btrfs_key key;
	struct btrfs_root_item root_item;
	struct btrfs_inode_item *inode_item;
	struct extent_buffer *leaf;
	struct btrfs_root *new_root = root;
	struct inode *dir;
	int ret;
	int err;
	u64 objectid;
	u64 new_dirid = BTRFS_FIRST_FREE_OBJECTID;
	unsigned long nr = 1;

	ret = btrfs_check_free_space(root, 1, 0);
	if (ret)
		goto fail_commit;

	trans = btrfs_start_transaction(root, 1);
	BUG_ON(!trans);

	ret = btrfs_find_free_objectid(trans, root->fs_info->tree_root,
				       0, &objectid);
	if (ret)
		goto fail;

	leaf = btrfs_alloc_free_block(trans, root, root->leafsize,
				      objectid, trans->transid, 0, 0,
				      0, 0);
	if (IS_ERR(leaf))
		return PTR_ERR(leaf);

	btrfs_set_header_nritems(leaf, 0);
	btrfs_set_header_level(leaf, 0);
	btrfs_set_header_bytenr(leaf, leaf->start);
	btrfs_set_header_generation(leaf, trans->transid);
	btrfs_set_header_owner(leaf, objectid);

	write_extent_buffer(leaf, root->fs_info->fsid,
			    (unsigned long)btrfs_header_fsid(leaf),
			    BTRFS_FSID_SIZE);
	btrfs_mark_buffer_dirty(leaf);

	inode_item = &root_item.inode;
	memset(inode_item, 0, sizeof(*inode_item));
	inode_item->generation = cpu_to_le64(1);
	inode_item->size = cpu_to_le64(3);
	inode_item->nlink = cpu_to_le32(1);
	inode_item->nblocks = cpu_to_le64(1);
	inode_item->mode = cpu_to_le32(S_IFDIR | 0755);

	btrfs_set_root_bytenr(&root_item, leaf->start);
	btrfs_set_root_level(&root_item, 0);
	btrfs_set_root_refs(&root_item, 1);
	btrfs_set_root_used(&root_item, 0);

	memset(&root_item.drop_progress, 0, sizeof(root_item.drop_progress));
	root_item.drop_level = 0;

	btrfs_tree_unlock(leaf);
	free_extent_buffer(leaf);
	leaf = NULL;

	btrfs_set_root_dirid(&root_item, new_dirid);

	key.objectid = objectid;
	key.offset = 1;
	btrfs_set_key_type(&key, BTRFS_ROOT_ITEM_KEY);
	ret = btrfs_insert_root(trans, root->fs_info->tree_root, &key,
				&root_item);
	if (ret)
		goto fail;

	/*
	 * 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,
				    name, namelen, dir->i_ino, &key,
				    BTRFS_FT_DIR);
	if (ret)
		goto fail;

	ret = btrfs_insert_inode_ref(trans, root->fs_info->tree_root,
			     name, namelen, objectid,
			     root->fs_info->sb->s_root->d_inode->i_ino);
	if (ret)
		goto fail;

	ret = btrfs_commit_transaction(trans, root);
	if (ret)
		goto fail_commit;

	new_root = btrfs_read_fs_root(root->fs_info, &key, name, namelen);
	BUG_ON(!new_root);

	trans = btrfs_start_transaction(new_root, 1);
	BUG_ON(!trans);

	ret = btrfs_create_subvol_root(new_root, trans, new_dirid,
				       BTRFS_I(dir)->block_group);
	if (ret)
		goto fail;

	/* Invalidate existing dcache entry for new subvolume. */
	btrfs_invalidate_dcache_root(root, name, namelen);

fail:
	nr = trans->blocks_used;
	err = btrfs_commit_transaction(trans, new_root);
	if (err && !ret)
		ret = err;
fail_commit:
	btrfs_btree_balance_dirty(root, nr);
	btrfs_throttle(root);
	return ret;
}

static int create_snapshot(struct btrfs_root *root, char *name, int namelen)
{
	struct btrfs_pending_snapshot *pending_snapshot;
	struct btrfs_trans_handle *trans;
	int ret;
	int err;
	unsigned long nr = 0;

	if (!root->ref_cows)
		return -EINVAL;

	ret = btrfs_check_free_space(root, 1, 0);
	if (ret)
		goto fail_unlock;

	pending_snapshot = kmalloc(sizeof(*pending_snapshot), GFP_NOFS);
	if (!pending_snapshot) {
		ret = -ENOMEM;
		goto fail_unlock;
	}
	pending_snapshot->name = kmalloc(namelen + 1, GFP_NOFS);
	if (!pending_snapshot->name) {
		ret = -ENOMEM;
		kfree(pending_snapshot);
		goto fail_unlock;
	}
	memcpy(pending_snapshot->name, name, namelen);
	pending_snapshot->name[namelen] = '\0';
	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:
	btrfs_btree_balance_dirty(root, nr);
	btrfs_throttle(root);
	return ret;
}

int btrfs_defrag_file(struct file *file)
{
	struct inode *inode = fdentry(file)->d_inode;
	struct btrfs_root *root = BTRFS_I(inode)->root;
	struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
	struct page *page;
	unsigned long last_index;
	unsigned long ra_pages = root->fs_info->bdi.ra_pages;
	unsigned long total_read = 0;
	u64 page_start;
	u64 page_end;
	unsigned long i;
	int ret;

	ret = btrfs_check_free_space(root, inode->i_size, 0);
	if (ret)
		return -ENOSPC;

	mutex_lock(&inode->i_mutex);
	last_index = inode->i_size >> PAGE_CACHE_SHIFT;
	for (i = 0; i <= last_index; i++) {
		if (total_read % ra_pages == 0) {
			btrfs_force_ra(inode->i_mapping, &file->f_ra, file, i,
				       min(last_index, i + ra_pages - 1));
		}
		total_read++;
		page = grab_cache_page(inode->i_mapping, i);
		if (!page)
			goto out_unlock;
		if (!PageUptodate(page)) {
			btrfs_readpage(NULL, page);
			lock_page(page);
			if (!PageUptodate(page)) {
				unlock_page(page);
				page_cache_release(page);
				goto out_unlock;
			}
		}

#if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,18)
		ClearPageDirty(page);
#else
		cancel_dirty_page(page, PAGE_CACHE_SIZE);
#endif
		wait_on_page_writeback(page);
		set_page_extent_mapped(page);

		page_start = (u64)page->index << PAGE_CACHE_SHIFT;
		page_end = page_start + PAGE_CACHE_SIZE - 1;

		lock_extent(io_tree, page_start, page_end, GFP_NOFS);
		set_extent_delalloc(io_tree, page_start,
				    page_end, GFP_NOFS);

		unlock_extent(io_tree, page_start, page_end, GFP_NOFS);
		set_page_dirty(page);
		unlock_page(page);
		page_cache_release(page);
		balance_dirty_pages_ratelimited_nr(inode->i_mapping, 1);
	}

out_unlock:
	mutex_unlock(&inode->i_mutex);
	return 0;
}

/*
 * Called inside transaction, so use GFP_NOFS
 */

static int btrfs_ioctl_resize(struct btrfs_root *root, void __user *arg)
{
	u64 new_size;
	u64 old_size;
	u64 devid = 1;
	struct btrfs_ioctl_vol_args *vol_args;
	struct btrfs_trans_handle *trans;
	struct btrfs_device *device = NULL;
	char *sizestr;
	char *devstr = NULL;
	int ret = 0;
	int namelen;
	int mod = 0;

	vol_args = kmalloc(sizeof(*vol_args), GFP_NOFS);

	if (!vol_args)
		return -ENOMEM;

	if (copy_from_user(vol_args, arg, sizeof(*vol_args))) {
		ret = -EFAULT;
		goto out;
	}
	namelen = strlen(vol_args->name);
	if (namelen > BTRFS_VOL_NAME_MAX) {
		ret = -EINVAL;
		goto out;
	}

	mutex_lock(&root->fs_info->alloc_mutex);
	mutex_lock(&root->fs_info->chunk_mutex);
	sizestr = vol_args->name;
	devstr = strchr(sizestr, ':');
	if (devstr) {
		char *end;
		sizestr = devstr + 1;
		*devstr = '\0';
		devstr = vol_args->name;
		devid = simple_strtoull(devstr, &end, 10);
		printk(KERN_INFO "resizing devid %llu\n", devid);
	}
	device = btrfs_find_device(root, devid, NULL);
	if (!device) {
		printk(KERN_INFO "resizer unable to find device %llu\n", devid);
		ret = -EINVAL;
		goto out_unlock;
	}
	if (!strcmp(sizestr, "max"))
		new_size = device->bdev->bd_inode->i_size;
	else {
		if (sizestr[0] == '-') {
			mod = -1;
			sizestr++;
		} else if (sizestr[0] == '+') {
			mod = 1;
			sizestr++;
		}
		new_size = btrfs_parse_size(sizestr);
		if (new_size == 0) {
			ret = -EINVAL;
			goto out_unlock;
		}
	}

	old_size = device->total_bytes;

	if (mod < 0) {
		if (new_size > old_size) {
			ret = -EINVAL;
			goto out_unlock;
		}
		new_size = old_size - new_size;
	} else if (mod > 0) {
		new_size = old_size + new_size;
	}

	if (new_size < 256 * 1024 * 1024) {
		ret = -EINVAL;
		goto out_unlock;
	}
	if (new_size > device->bdev->bd_inode->i_size) {
		ret = -EFBIG;
		goto out_unlock;
	}

	do_div(new_size, root->sectorsize);
	new_size *= root->sectorsize;

	printk(KERN_INFO "new size for %s is %llu\n",
		device->name, (unsigned long long)new_size);

	if (new_size > old_size) {
		trans = btrfs_start_transaction(root, 1);
		ret = btrfs_grow_device(trans, device, new_size);
		btrfs_commit_transaction(trans, root);
	} else {
		ret = btrfs_shrink_device(device, new_size);
	}

out_unlock:
	mutex_lock(&root->fs_info->alloc_mutex);
	mutex_lock(&root->fs_info->chunk_mutex);
out:
	kfree(vol_args);
	return ret;
}

static noinline int btrfs_ioctl_snap_create(struct btrfs_root *root,
					    void __user *arg)
{
	struct btrfs_ioctl_vol_args *vol_args;
	struct btrfs_dir_item *di;
	struct btrfs_path *path;
	u64 root_dirid;
	int namelen;
	int ret;

	vol_args = kmalloc(sizeof(*vol_args), GFP_NOFS);

	if (!vol_args)
		return -ENOMEM;

	if (copy_from_user(vol_args, arg, sizeof(*vol_args))) {
		ret = -EFAULT;
		goto out;
	}

	namelen = strlen(vol_args->name);
	if (namelen > BTRFS_VOL_NAME_MAX) {
		ret = -EINVAL;
		goto out;
	}
	if (strchr(vol_args->name, '/')) {
		ret = -EINVAL;
		goto out;
	}

	path = btrfs_alloc_path();
	if (!path) {
		ret = -ENOMEM;
		goto out;
	}

	root_dirid = root->fs_info->sb->s_root->d_inode->i_ino,
	di = btrfs_lookup_dir_item(NULL, root->fs_info->tree_root,
			    path, root_dirid,
			    vol_args->name, namelen, 0);
	btrfs_free_path(path);

	if (di && !IS_ERR(di)) {
		ret = -EEXIST;
		goto out;
	}

	if (IS_ERR(di)) {
		ret = PTR_ERR(di);
		goto out;
	}

	mutex_lock(&root->fs_info->drop_mutex);
	if (root == root->fs_info->tree_root)
		ret = create_subvol(root, vol_args->name, namelen);
	else
		ret = create_snapshot(root, vol_args->name, namelen);
	mutex_unlock(&root->fs_info->drop_mutex);
out:
	kfree(vol_args);
	return ret;
}

static int btrfs_ioctl_defrag(struct file *file)
{
	struct inode *inode = fdentry(file)->d_inode;
	struct btrfs_root *root = BTRFS_I(inode)->root;

	switch (inode->i_mode & S_IFMT) {
	case S_IFDIR:
		btrfs_defrag_root(root, 0);
		btrfs_defrag_root(root->fs_info->extent_root, 0);
		break;
	case S_IFREG:
		btrfs_defrag_file(file);
		break;
	}

	return 0;
}

long btrfs_ioctl_add_dev(struct btrfs_root *root, void __user *arg)
{
	struct btrfs_ioctl_vol_args *vol_args;
	int ret;

	vol_args = kmalloc(sizeof(*vol_args), GFP_NOFS);

	if (!vol_args)
		return -ENOMEM;

	if (copy_from_user(vol_args, arg, sizeof(*vol_args))) {
		ret = -EFAULT;
		goto out;
	}
	ret = btrfs_init_new_device(root, vol_args->name);

out:
	kfree(vol_args);
	return ret;
}

long btrfs_ioctl_rm_dev(struct btrfs_root *root, void __user *arg)
{
	struct btrfs_ioctl_vol_args *vol_args;
	int ret;

	vol_args = kmalloc(sizeof(*vol_args), GFP_NOFS);

	if (!vol_args)
		return -ENOMEM;

	if (copy_from_user(vol_args, arg, sizeof(*vol_args))) {
		ret = -EFAULT;
		goto out;
	}
	ret = btrfs_rm_device(root, vol_args->name);

out:
	kfree(vol_args);
	return ret;
}

int dup_item_to_inode(struct btrfs_trans_handle *trans,
		       struct btrfs_root *root,
		       struct btrfs_path *path,
		       struct extent_buffer *leaf,
		       int slot,
		       struct btrfs_key *key,
		       u64 destino)
{
	char *dup;
	int len = btrfs_item_size_nr(leaf, slot);
	struct btrfs_key ckey = *key;
	int ret = 0;

	dup = kmalloc(len, GFP_NOFS);
	if (!dup)
		return -ENOMEM;

	read_extent_buffer(leaf, dup, btrfs_item_ptr_offset(leaf, slot), len);
	btrfs_release_path(root, path);

	ckey.objectid = destino;
	ret = btrfs_insert_item(trans, root, &ckey, dup, len);
	kfree(dup);
	return ret;
}

long btrfs_ioctl_clone(struct file *file, unsigned long src_fd)
{
	struct inode *inode = fdentry(file)->d_inode;
	struct btrfs_root *root = BTRFS_I(inode)->root;
	struct file *src_file;
	struct inode *src;
	struct btrfs_trans_handle *trans;
	int ret;
	u64 pos;
	struct btrfs_path *path;
	struct btrfs_key key;
	struct extent_buffer *leaf;
	u32 nritems;
	int slot;

	src_file = fget(src_fd);
	if (!src_file)
		return -EBADF;
	src = src_file->f_dentry->d_inode;

	ret = -EXDEV;
	if (src->i_sb != inode->i_sb)
		goto out_fput;

	if (inode < src) {
		mutex_lock(&inode->i_mutex);
		mutex_lock(&src->i_mutex);
	} else {
		mutex_lock(&src->i_mutex);
		mutex_lock(&inode->i_mutex);
	}

	ret = -ENOTEMPTY;
	if (inode->i_size)
		goto out_unlock;

	/* do any pending delalloc/csum calc on src, one way or
	   another, and lock file content */
	while (1) {
		filemap_write_and_wait(src->i_mapping);
		lock_extent(&BTRFS_I(src)->io_tree, 0, (u64)-1, GFP_NOFS);
		if (BTRFS_I(src)->delalloc_bytes == 0)
			break;
		unlock_extent(&BTRFS_I(src)->io_tree, 0, (u64)-1, GFP_NOFS);
	}

	trans = btrfs_start_transaction(root, 0);
	path = btrfs_alloc_path();
	if (!path) {
		ret = -ENOMEM;
		goto out;
	}
	key.offset = 0;
	key.type = BTRFS_EXTENT_DATA_KEY;
	key.objectid = src->i_ino;
	pos = 0;
	path->reada = 2;

	while (1) {
		/*
		 * note the key will change type as we walk through the
		 * tree.
		 */
		ret = btrfs_search_slot(trans, root, &key, path, 0, 0);
		if (ret < 0)
			goto out;

		if (path->slots[0] >= btrfs_header_nritems(path->nodes[0])) {
			ret = btrfs_next_leaf(root, path);
			if (ret < 0)
				goto out;
			if (ret > 0)
				break;
		}
		leaf = path->nodes[0];
		slot = path->slots[0];
		btrfs_item_key_to_cpu(leaf, &key, slot);
		nritems = btrfs_header_nritems(leaf);

		if (btrfs_key_type(&key) > BTRFS_CSUM_ITEM_KEY ||
		    key.objectid != src->i_ino)
			break;

		if (btrfs_key_type(&key) == BTRFS_EXTENT_DATA_KEY) {
			struct btrfs_file_extent_item *extent;
			int found_type;
			pos = key.offset;
			extent = btrfs_item_ptr(leaf, slot,
						struct btrfs_file_extent_item);
			found_type = btrfs_file_extent_type(leaf, extent);
			if (found_type == BTRFS_FILE_EXTENT_REG) {
				u64 len = btrfs_file_extent_num_bytes(leaf,
								      extent);
				u64 ds = btrfs_file_extent_disk_bytenr(leaf,
								       extent);
				u64 dl = btrfs_file_extent_disk_num_bytes(leaf,
								 extent);
				u64 off = btrfs_file_extent_offset(leaf,
								   extent);
				btrfs_insert_file_extent(trans, root,
							 inode->i_ino, pos,
							 ds, dl, len, off);
				/* ds == 0 means there's a hole */
				if (ds != 0) {
					btrfs_inc_extent_ref(trans, root,
						     ds, dl,
						     root->root_key.objectid,
						     trans->transid,
						     inode->i_ino, pos);
				}
				pos = key.offset + len;
			} else if (found_type == BTRFS_FILE_EXTENT_INLINE) {
				ret = dup_item_to_inode(trans, root, path,
							leaf, slot, &key,
							inode->i_ino);
				if (ret)
					goto out;
				pos = key.offset + btrfs_item_size_nr(leaf,
								      slot);
			}
		} else if (btrfs_key_type(&key) == BTRFS_CSUM_ITEM_KEY) {
			ret = dup_item_to_inode(trans, root, path, leaf,
						slot, &key, inode->i_ino);

			if (ret)
				goto out;
		}
		key.offset++;
		btrfs_release_path(root, path);
	}

	ret = 0;
out:
	btrfs_free_path(path);

	inode->i_blocks = src->i_blocks;
	i_size_write(inode, src->i_size);
	btrfs_update_inode(trans, root, inode);

	unlock_extent(&BTRFS_I(src)->io_tree, 0, (u64)-1, GFP_NOFS);

	btrfs_end_transaction(trans, root);

out_unlock:
	mutex_unlock(&src->i_mutex);
	mutex_unlock(&inode->i_mutex);
out_fput:
	fput(src_file);
	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;

	if (!capable(CAP_SYS_ADMIN))
		return -EPERM;

	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:
	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;

	trans = file->private_data;
	if (!trans) {
		ret = -EINVAL;
		goto out;
	}
	btrfs_end_transaction(trans, root);
	file->private_data = 0;
out:
	return ret;
}

long btrfs_ioctl(struct file *file, unsigned int
		cmd, unsigned long arg)
{
	struct btrfs_root *root = BTRFS_I(fdentry(file)->d_inode)->root;

	switch (cmd) {
	case BTRFS_IOC_SNAP_CREATE:
		return btrfs_ioctl_snap_create(root, (void __user *)arg);
	case BTRFS_IOC_DEFRAG:
		return btrfs_ioctl_defrag(file);
	case BTRFS_IOC_RESIZE:
		return btrfs_ioctl_resize(root, (void __user *)arg);
	case BTRFS_IOC_ADD_DEV:
		return btrfs_ioctl_add_dev(root, (void __user *)arg);
	case BTRFS_IOC_RM_DEV:
		return btrfs_ioctl_rm_dev(root, (void __user *)arg);
	case BTRFS_IOC_BALANCE:
		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;
}
Commit	Line	Data
f46b5a66 CH	1	/*
	2	* Copyright (C) 2007 Oracle. All rights reserved.
	3	*
	4	* This program is free software; you can redistribute it and/or
	5	* modify it under the terms of the GNU General Public
	6	* License v2 as published by the Free Software Foundation.
	7	*
	8	* This program is distributed in the hope that it will be useful,
	9	* but WITHOUT ANY WARRANTY; without even the implied warranty of
	10	* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
	11	* General Public License for more details.
	12	*
	13	* You should have received a copy of the GNU General Public
	14	* License along with this program; if not, write to the
	15	* Free Software Foundation, Inc., 59 Temple Place - Suite 330,
	16	* Boston, MA 021110-1307, USA.
	17	*/
	18
	19	#include <linux/kernel.h>
	20	#include <linux/bio.h>
	21	#include <linux/buffer_head.h>
	22	#include <linux/file.h>
	23	#include <linux/fs.h>
	24	#include <linux/pagemap.h>
	25	#include <linux/highmem.h>
	26	#include <linux/time.h>
	27	#include <linux/init.h>
	28	#include <linux/string.h>
	29	#include <linux/smp_lock.h>
	30	#include <linux/backing-dev.h>
	31	#include <linux/mpage.h>
	32	#include <linux/swap.h>
	33	#include <linux/writeback.h>
	34	#include <linux/statfs.h>
	35	#include <linux/compat.h>
	36	#include <linux/bit_spinlock.h>
	37	#include <linux/version.h>
	38	#include <linux/xattr.h>
	39	#include "ctree.h"
	40	#include "disk-io.h"
	41	#include "transaction.h"
	42	#include "btrfs_inode.h"
	43	#include "ioctl.h"
	44	#include "print-tree.h"
	45	#include "volumes.h"
925baedd	46	#include "locking.h"
f46b5a66 CH	47
	48
	49
	50	static noinline int create_subvol(struct btrfs_root root, char name,
	51	int namelen)
	52	{
	53	struct btrfs_trans_handle *trans;
	54	struct btrfs_key key;
	55	struct btrfs_root_item root_item;
	56	struct btrfs_inode_item *inode_item;
	57	struct extent_buffer *leaf;
	58	struct btrfs_root *new_root = root;
	59	struct inode *dir;
	60	int ret;
	61	int err;
	62	u64 objectid;
	63	u64 new_dirid = BTRFS_FIRST_FREE_OBJECTID;
	64	unsigned long nr = 1;
	65
f46b5a66 CH	66	ret = btrfs_check_free_space(root, 1, 0);
	67	if (ret)
	68	goto fail_commit;
	69
	70	trans = btrfs_start_transaction(root, 1);
	71	BUG_ON(!trans);
	72
	73	ret = btrfs_find_free_objectid(trans, root->fs_info->tree_root,
	74	0, &objectid);
	75	if (ret)
	76	goto fail;
	77
925baedd CM	78	leaf = btrfs_alloc_free_block(trans, root, root->leafsize,
	79	objectid, trans->transid, 0, 0,
	80	0, 0);
f46b5a66 CH	81	if (IS_ERR(leaf))
	82	return PTR_ERR(leaf);
	83
	84	btrfs_set_header_nritems(leaf, 0);
	85	btrfs_set_header_level(leaf, 0);
	86	btrfs_set_header_bytenr(leaf, leaf->start);
	87	btrfs_set_header_generation(leaf, trans->transid);
	88	btrfs_set_header_owner(leaf, objectid);
	89
	90	write_extent_buffer(leaf, root->fs_info->fsid,
	91	(unsigned long)btrfs_header_fsid(leaf),
	92	BTRFS_FSID_SIZE);
	93	btrfs_mark_buffer_dirty(leaf);
	94
	95	inode_item = &root_item.inode;
	96	memset(inode_item, 0, sizeof(*inode_item));
	97	inode_item->generation = cpu_to_le64(1);
	98	inode_item->size = cpu_to_le64(3);
	99	inode_item->nlink = cpu_to_le32(1);
	100	inode_item->nblocks = cpu_to_le64(1);
	101	inode_item->mode = cpu_to_le32(S_IFDIR \| 0755);
	102
	103	btrfs_set_root_bytenr(&root_item, leaf->start);
	104	btrfs_set_root_level(&root_item, 0);
	105	btrfs_set_root_refs(&root_item, 1);
	106	btrfs_set_root_used(&root_item, 0);
	107
	108	memset(&root_item.drop_progress, 0, sizeof(root_item.drop_progress));
	109	root_item.drop_level = 0;
	110
925baedd	111	btrfs_tree_unlock(leaf);
f46b5a66 CH	112	free_extent_buffer(leaf);
	113	leaf = NULL;
	114
	115	btrfs_set_root_dirid(&root_item, new_dirid);
	116
	117	key.objectid = objectid;
	118	key.offset = 1;
	119	btrfs_set_key_type(&key, BTRFS_ROOT_ITEM_KEY);
	120	ret = btrfs_insert_root(trans, root->fs_info->tree_root, &key,
	121	&root_item);
	122	if (ret)
	123	goto fail;
	124
	125	/*
	126	* insert the directory item
	127	*/
	128	key.offset = (u64)-1;
	129	dir = root->fs_info->sb->s_root->d_inode;
	130	ret = btrfs_insert_dir_item(trans, root->fs_info->tree_root,
	131	name, namelen, dir->i_ino, &key,
	132	BTRFS_FT_DIR);
	133	if (ret)
	134	goto fail;
	135
	136	ret = btrfs_insert_inode_ref(trans, root->fs_info->tree_root,
	137	name, namelen, objectid,
	138	root->fs_info->sb->s_root->d_inode->i_ino);
	139	if (ret)
	140	goto fail;
	141
	142	ret = btrfs_commit_transaction(trans, root);
	143	if (ret)
	144	goto fail_commit;
	145
	146	new_root = btrfs_read_fs_root(root->fs_info, &key, name, namelen);
	147	BUG_ON(!new_root);
	148
	149	trans = btrfs_start_transaction(new_root, 1);
	150	BUG_ON(!trans);
	151
	152	ret = btrfs_create_subvol_root(new_root, trans, new_dirid,
	153	BTRFS_I(dir)->block_group);
	154	if (ret)
	155	goto fail;
	156
	157	/* Invalidate existing dcache entry for new subvolume. */
	158	btrfs_invalidate_dcache_root(root, name, namelen);
	159
	160	fail:
	161	nr = trans->blocks_used;
	162	err = btrfs_commit_transaction(trans, new_root);
	163	if (err && !ret)
	164	ret = err;
	165	fail_commit:
f46b5a66 CH	166	btrfs_btree_balance_dirty(root, nr);
	167	btrfs_throttle(root);
	168	return ret;
	169	}
	170
	171	static int create_snapshot(struct btrfs_root root, char name, int namelen)
	172	{
	173	struct btrfs_pending_snapshot *pending_snapshot;
	174	struct btrfs_trans_handle *trans;
	175	int ret;
	176	int err;
	177	unsigned long nr = 0;
	178
	179	if (!root->ref_cows)
	180	return -EINVAL;
	181
f46b5a66 CH	182	ret = btrfs_check_free_space(root, 1, 0);
	183	if (ret)
	184	goto fail_unlock;
	185
	186	pending_snapshot = kmalloc(sizeof(*pending_snapshot), GFP_NOFS);
	187	if (!pending_snapshot) {
	188	ret = -ENOMEM;
	189	goto fail_unlock;
	190	}
	191	pending_snapshot->name = kmalloc(namelen + 1, GFP_NOFS);
	192	if (!pending_snapshot->name) {
	193	ret = -ENOMEM;
	194	kfree(pending_snapshot);
	195	goto fail_unlock;
	196	}
	197	memcpy(pending_snapshot->name, name, namelen);
	198	pending_snapshot->name[namelen] = '\0';
	199	trans = btrfs_start_transaction(root, 1);
	200	BUG_ON(!trans);
	201	pending_snapshot->root = root;
	202	list_add(&pending_snapshot->list,
	203	&trans->transaction->pending_snapshots);
	204	ret = btrfs_update_inode(trans, root, root->inode);
	205	err = btrfs_commit_transaction(trans, root);
	206
	207	fail_unlock:
f46b5a66 CH	208	btrfs_btree_balance_dirty(root, nr);
	209	btrfs_throttle(root);
	210	return ret;
	211	}
	212
	213	int btrfs_defrag_file(struct file *file)
	214	{
	215	struct inode *inode = fdentry(file)->d_inode;
	216	struct btrfs_root *root = BTRFS_I(inode)->root;
	217	struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
	218	struct page *page;
	219	unsigned long last_index;
	220	unsigned long ra_pages = root->fs_info->bdi.ra_pages;
	221	unsigned long total_read = 0;
	222	u64 page_start;
	223	u64 page_end;
	224	unsigned long i;
	225	int ret;
	226
f46b5a66	227	ret = btrfs_check_free_space(root, inode->i_size, 0);
f46b5a66 CH	228	if (ret)
	229	return -ENOSPC;
	230
	231	mutex_lock(&inode->i_mutex);
	232	last_index = inode->i_size >> PAGE_CACHE_SHIFT;
	233	for (i = 0; i <= last_index; i++) {
	234	if (total_read % ra_pages == 0) {
	235	btrfs_force_ra(inode->i_mapping, &file->f_ra, file, i,
	236	min(last_index, i + ra_pages - 1));
	237	}
	238	total_read++;
	239	page = grab_cache_page(inode->i_mapping, i);
	240	if (!page)
	241	goto out_unlock;
	242	if (!PageUptodate(page)) {
	243	btrfs_readpage(NULL, page);
	244	lock_page(page);
	245	if (!PageUptodate(page)) {
	246	unlock_page(page);
	247	page_cache_release(page);
	248	goto out_unlock;
	249	}
	250	}
	251
	252	#if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,18)
	253	ClearPageDirty(page);
	254	#else
	255	cancel_dirty_page(page, PAGE_CACHE_SIZE);
	256	#endif
	257	wait_on_page_writeback(page);
	258	set_page_extent_mapped(page);
	259
	260	page_start = (u64)page->index << PAGE_CACHE_SHIFT;
	261	page_end = page_start + PAGE_CACHE_SIZE - 1;
	262
	263	lock_extent(io_tree, page_start, page_end, GFP_NOFS);
	264	set_extent_delalloc(io_tree, page_start,
	265	page_end, GFP_NOFS);
	266
	267	unlock_extent(io_tree, page_start, page_end, GFP_NOFS);
	268	set_page_dirty(page);
	269	unlock_page(page);
	270	page_cache_release(page);
	271	balance_dirty_pages_ratelimited_nr(inode->i_mapping, 1);
	272	}
	273
	274	out_unlock:
	275	mutex_unlock(&inode->i_mutex);
	276	return 0;
	277	}
	278
	279	/*
	280	* Called inside transaction, so use GFP_NOFS
	281	*/
	282
	283	static int btrfs_ioctl_resize(struct btrfs_root root, void __user arg)
	284	{
	285	u64 new_size;
	286	u64 old_size;
	287	u64 devid = 1;
	288	struct btrfs_ioctl_vol_args *vol_args;
	289	struct btrfs_trans_handle *trans;
	290	struct btrfs_device *device = NULL;
	291	char *sizestr;
292	char *devstr = NULL;
293	int ret = 0;
294	int namelen;
295	int mod = 0;
296
297	vol_args = kmalloc(sizeof(*vol_args), GFP_NOFS);
298
299	if (!vol_args)
300	return -ENOMEM;
301
302	if (copy_from_user(vol_args, arg, sizeof(*vol_args))) {
303	ret = -EFAULT;
304	goto out;
305	}
306	namelen = strlen(vol_args->name);
307	if (namelen > BTRFS_VOL_NAME_MAX) {
308	ret = -EINVAL;
309	goto out;
310	}
311
a2135011 CM	312	mutex_lock(&root->fs_info->alloc_mutex);
a2135011 CM	313	mutex_lock(&root->fs_info->chunk_mutex);
f46b5a66 CH	314	sizestr = vol_args->name;
	315	devstr = strchr(sizestr, ':');
	316	if (devstr) {
	317	char *end;
	318	sizestr = devstr + 1;
	319	*devstr = '\0';
	320	devstr = vol_args->name;
	321	devid = simple_strtoull(devstr, &end, 10);
	322	printk(KERN_INFO "resizing devid %llu\n", devid);
	323	}
	324	device = btrfs_find_device(root, devid, NULL);
	325	if (!device) {
	326	printk(KERN_INFO "resizer unable to find device %llu\n", devid);
	327	ret = -EINVAL;
	328	goto out_unlock;
	329	}
	330	if (!strcmp(sizestr, "max"))
	331	new_size = device->bdev->bd_inode->i_size;
	332	else {
	333	if (sizestr[0] == '-') {
	334	mod = -1;
	335	sizestr++;
	336	} else if (sizestr[0] == '+') {
	337	mod = 1;
	338	sizestr++;
	339	}
	340	new_size = btrfs_parse_size(sizestr);
	341	if (new_size == 0) {
	342	ret = -EINVAL;
	343	goto out_unlock;
	344	}
	345	}
	346
	347	old_size = device->total_bytes;
	348
	349	if (mod < 0) {
	350	if (new_size > old_size) {
	351	ret = -EINVAL;
	352	goto out_unlock;
	353	}
	354	new_size = old_size - new_size;
	355	} else if (mod > 0) {
	356	new_size = old_size + new_size;
	357	}
	358
	359	if (new_size < 256 * 1024 * 1024) {
	360	ret = -EINVAL;
	361	goto out_unlock;
	362	}
	363	if (new_size > device->bdev->bd_inode->i_size) {
	364	ret = -EFBIG;
	365	goto out_unlock;
	366	}
	367
	368	do_div(new_size, root->sectorsize);
	369	new_size *= root->sectorsize;
	370
	371	printk(KERN_INFO "new size for %s is %llu\n",
	372	device->name, (unsigned long long)new_size);
	373
	374	if (new_size > old_size) {
	375	trans = btrfs_start_transaction(root, 1);
	376	ret = btrfs_grow_device(trans, device, new_size);
	377	btrfs_commit_transaction(trans, root);
378	} else {
379	ret = btrfs_shrink_device(device, new_size);
380	}
381
382	out_unlock:
a2135011 CM	383	mutex_lock(&root->fs_info->alloc_mutex);
a2135011 CM	384	mutex_lock(&root->fs_info->chunk_mutex);
f46b5a66 CH	385	out:
	386	kfree(vol_args);
	387	return ret;
	388	}
	389
	390	static noinline int btrfs_ioctl_snap_create(struct btrfs_root *root,
	391	void __user *arg)
	392	{
	393	struct btrfs_ioctl_vol_args *vol_args;
	394	struct btrfs_dir_item *di;
	395	struct btrfs_path *path;
	396	u64 root_dirid;
	397	int namelen;
	398	int ret;
	399
	400	vol_args = kmalloc(sizeof(*vol_args), GFP_NOFS);
	401
	402	if (!vol_args)
	403	return -ENOMEM;
	404
	405	if (copy_from_user(vol_args, arg, sizeof(*vol_args))) {
	406	ret = -EFAULT;
	407	goto out;
	408	}
	409
	410	namelen = strlen(vol_args->name);
	411	if (namelen > BTRFS_VOL_NAME_MAX) {
	412	ret = -EINVAL;
	413	goto out;
	414	}
	415	if (strchr(vol_args->name, '/')) {
	416	ret = -EINVAL;
	417	goto out;
	418	}
	419
	420	path = btrfs_alloc_path();
	421	if (!path) {
	422	ret = -ENOMEM;
	423	goto out;
	424	}
	425
	426	root_dirid = root->fs_info->sb->s_root->d_inode->i_ino,
f46b5a66 CH	427	di = btrfs_lookup_dir_item(NULL, root->fs_info->tree_root,
	428	path, root_dirid,
	429	vol_args->name, namelen, 0);
f46b5a66 CH	430	btrfs_free_path(path);
	431
	432	if (di && !IS_ERR(di)) {
	433	ret = -EEXIST;
	434	goto out;
	435	}
	436
	437	if (IS_ERR(di)) {
	438	ret = PTR_ERR(di);
	439	goto out;
	440	}
	441
a2135011	442	mutex_lock(&root->fs_info->drop_mutex);
f46b5a66 CH	443	if (root == root->fs_info->tree_root)
	444	ret = create_subvol(root, vol_args->name, namelen);
	445	else
	446	ret = create_snapshot(root, vol_args->name, namelen);
a2135011	447	mutex_unlock(&root->fs_info->drop_mutex);
f46b5a66 CH	448	out:
	449	kfree(vol_args);
	450	return ret;
	451	}
	452
	453	static int btrfs_ioctl_defrag(struct file *file)
	454	{
	455	struct inode *inode = fdentry(file)->d_inode;
	456	struct btrfs_root *root = BTRFS_I(inode)->root;
	457
	458	switch (inode->i_mode & S_IFMT) {
	459	case S_IFDIR:
f46b5a66 CH	460	btrfs_defrag_root(root, 0);
f46b5a66 CH	461	btrfs_defrag_root(root->fs_info->extent_root, 0);
f46b5a66 CH	462	break;
	463	case S_IFREG:
	464	btrfs_defrag_file(file);
	465	break;
	466	}
	467
	468	return 0;
	469	}
	470
	471	long btrfs_ioctl_add_dev(struct btrfs_root root, void __user arg)
	472	{
	473	struct btrfs_ioctl_vol_args *vol_args;
	474	int ret;
	475
	476	vol_args = kmalloc(sizeof(*vol_args), GFP_NOFS);
	477
	478	if (!vol_args)
	479	return -ENOMEM;
	480
	481	if (copy_from_user(vol_args, arg, sizeof(*vol_args))) {
	482	ret = -EFAULT;
	483	goto out;
	484	}
	485	ret = btrfs_init_new_device(root, vol_args->name);
	486
	487	out:
	488	kfree(vol_args);
	489	return ret;
	490	}
	491
	492	long btrfs_ioctl_rm_dev(struct btrfs_root root, void __user arg)
	493	{
	494	struct btrfs_ioctl_vol_args *vol_args;
	495	int ret;
	496
	497	vol_args = kmalloc(sizeof(*vol_args), GFP_NOFS);
	498
	499	if (!vol_args)
	500	return -ENOMEM;
	501
	502	if (copy_from_user(vol_args, arg, sizeof(*vol_args))) {
	503	ret = -EFAULT;
	504	goto out;
	505	}
	506	ret = btrfs_rm_device(root, vol_args->name);
	507
	508	out:
	509	kfree(vol_args);
	510	return ret;
	511	}
	512
	513	int dup_item_to_inode(struct btrfs_trans_handle *trans,
	514	struct btrfs_root *root,
	515	struct btrfs_path *path,
	516	struct extent_buffer *leaf,
	517	int slot,
	518	struct btrfs_key *key,
	519	u64 destino)
	520	{
	521	char *dup;
	522	int len = btrfs_item_size_nr(leaf, slot);
	523	struct btrfs_key ckey = *key;
	524	int ret = 0;
	525
526	dup = kmalloc(len, GFP_NOFS);
527	if (!dup)
528	return -ENOMEM;
529
530	read_extent_buffer(leaf, dup, btrfs_item_ptr_offset(leaf, slot), len);
531	btrfs_release_path(root, path);
532
533	ckey.objectid = destino;
534	ret = btrfs_insert_item(trans, root, &ckey, dup, len);
535	kfree(dup);
536	return ret;
537	}
538
539	long btrfs_ioctl_clone(struct file *file, unsigned long src_fd)
540	{
541	struct inode *inode = fdentry(file)->d_inode;
542	struct btrfs_root *root = BTRFS_I(inode)->root;
543	struct file *src_file;
544	struct inode *src;
545	struct btrfs_trans_handle *trans;
546	int ret;
547	u64 pos;
548	struct btrfs_path *path;
549	struct btrfs_key key;
550	struct extent_buffer *leaf;
551	u32 nritems;
552	int slot;
553
554	src_file = fget(src_fd);
555	if (!src_file)
556	return -EBADF;
557	src = src_file->f_dentry->d_inode;
558
559	ret = -EXDEV;
560	if (src->i_sb != inode->i_sb)
561	goto out_fput;
562
563	if (inode < src) {
564	mutex_lock(&inode->i_mutex);
565	mutex_lock(&src->i_mutex);
566	} else {
567	mutex_lock(&src->i_mutex);
568	mutex_lock(&inode->i_mutex);
569	}
570
571	ret = -ENOTEMPTY;
572	if (inode->i_size)
573	goto out_unlock;
574
575	/* do any pending delalloc/csum calc on src, one way or
576	another, and lock file content */
577	while (1) {
578	filemap_write_and_wait(src->i_mapping);
579	lock_extent(&BTRFS_I(src)->io_tree, 0, (u64)-1, GFP_NOFS);
580	if (BTRFS_I(src)->delalloc_bytes == 0)
581	break;
582	unlock_extent(&BTRFS_I(src)->io_tree, 0, (u64)-1, GFP_NOFS);
583	}
584
f46b5a66 CH	585	trans = btrfs_start_transaction(root, 0);
	586	path = btrfs_alloc_path();
	587	if (!path) {
	588	ret = -ENOMEM;
	589	goto out;
	590	}
	591	key.offset = 0;
	592	key.type = BTRFS_EXTENT_DATA_KEY;
	593	key.objectid = src->i_ino;
	594	pos = 0;
	595	path->reada = 2;
	596
	597	while (1) {
	598	/*
	599	* note the key will change type as we walk through the
	600	* tree.
	601	*/
	602	ret = btrfs_search_slot(trans, root, &key, path, 0, 0);
	603	if (ret < 0)
	604	goto out;
	605
	606	if (path->slots[0] >= btrfs_header_nritems(path->nodes[0])) {
	607	ret = btrfs_next_leaf(root, path);
	608	if (ret < 0)
	609	goto out;
	610	if (ret > 0)
	611	break;
	612	}
	613	leaf = path->nodes[0];
	614	slot = path->slots[0];
	615	btrfs_item_key_to_cpu(leaf, &key, slot);
	616	nritems = btrfs_header_nritems(leaf);
	617
	618	if (btrfs_key_type(&key) > BTRFS_CSUM_ITEM_KEY \|\|
	619	key.objectid != src->i_ino)
	620	break;
	621
	622	if (btrfs_key_type(&key) == BTRFS_EXTENT_DATA_KEY) {
	623	struct btrfs_file_extent_item *extent;
	624	int found_type;
	625	pos = key.offset;
	626	extent = btrfs_item_ptr(leaf, slot,
	627	struct btrfs_file_extent_item);
	628	found_type = btrfs_file_extent_type(leaf, extent);
	629	if (found_type == BTRFS_FILE_EXTENT_REG) {
	630	u64 len = btrfs_file_extent_num_bytes(leaf,
	631	extent);
	632	u64 ds = btrfs_file_extent_disk_bytenr(leaf,
	633	extent);
	634	u64 dl = btrfs_file_extent_disk_num_bytes(leaf,
	635	extent);
	636	u64 off = btrfs_file_extent_offset(leaf,
	637	extent);
	638	btrfs_insert_file_extent(trans, root,
	639	inode->i_ino, pos,
	640	ds, dl, len, off);
	641	/* ds == 0 means there's a hole */
	642	if (ds != 0) {
	643	btrfs_inc_extent_ref(trans, root,
	644	ds, dl,
	645	root->root_key.objectid,
	646	trans->transid,
	647	inode->i_ino, pos);
	648	}
649	pos = key.offset + len;
650	} else if (found_type == BTRFS_FILE_EXTENT_INLINE) {
651	ret = dup_item_to_inode(trans, root, path,
652	leaf, slot, &key,
653	inode->i_ino);
654	if (ret)
655	goto out;
656	pos = key.offset + btrfs_item_size_nr(leaf,
657	slot);
658	}
659	} else if (btrfs_key_type(&key) == BTRFS_CSUM_ITEM_KEY) {
660	ret = dup_item_to_inode(trans, root, path, leaf,
661	slot, &key, inode->i_ino);
662
663	if (ret)
664	goto out;
665	}
666	key.offset++;
667	btrfs_release_path(root, path);
668	}
669
670	ret = 0;
671	out:
672	btrfs_free_path(path);
673
674	inode->i_blocks = src->i_blocks;
675	i_size_write(inode, src->i_size);
676	btrfs_update_inode(trans, root, inode);
677
678	unlock_extent(&BTRFS_I(src)->io_tree, 0, (u64)-1, GFP_NOFS);
679
680	btrfs_end_transaction(trans, root);
f46b5a66 CH	681
	682	out_unlock:
	683	mutex_unlock(&src->i_mutex);
	684	mutex_unlock(&inode->i_mutex);
	685	out_fput:
	686	fput(src_file);
	687	return ret;
	688	}
	689
	690	/*
	691	* there are many ways the trans_start and trans_end ioctls can lead
	692	* to deadlocks. They should only be used by applications that
	693	* basically own the machine, and have a very in depth understanding
	694	* of all the possible deadlocks and enospc problems.
	695	*/
	696	long btrfs_ioctl_trans_start(struct file *file)
	697	{
	698	struct inode *inode = fdentry(file)->d_inode;
	699	struct btrfs_root *root = BTRFS_I(inode)->root;
	700	struct btrfs_trans_handle *trans;
	701	int ret = 0;
	702
df5b5520 CH	703	if (!capable(CAP_SYS_ADMIN))
	704	return -EPERM;
	705
f46b5a66 CH	706	if (file->private_data) {
	707	ret = -EINPROGRESS;
	708	goto out;
	709	}
	710	trans = btrfs_start_transaction(root, 0);
	711	if (trans)
	712	file->private_data = trans;
	713	else
	714	ret = -ENOMEM;
	715	/printk(KERN_INFO "btrfs_ioctl_trans_start on %p\n", file);/
	716	out:
f46b5a66 CH	717	return ret;
	718	}
	719
	720	/*
	721	* there are many ways the trans_start and trans_end ioctls can lead
	722	* to deadlocks. They should only be used by applications that
	723	* basically own the machine, and have a very in depth understanding
	724	* of all the possible deadlocks and enospc problems.
	725	*/
	726	long btrfs_ioctl_trans_end(struct file *file)
	727	{
	728	struct inode *inode = fdentry(file)->d_inode;
	729	struct btrfs_root *root = BTRFS_I(inode)->root;
	730	struct btrfs_trans_handle *trans;
	731	int ret = 0;
	732
f46b5a66 CH	733	trans = file->private_data;
	734	if (!trans) {
	735	ret = -EINVAL;
	736	goto out;
	737	}
	738	btrfs_end_transaction(trans, root);
	739	file->private_data = 0;
	740	out:
f46b5a66 CH	741	return ret;
	742	}
	743
	744	long btrfs_ioctl(struct file *file, unsigned int
	745	cmd, unsigned long arg)
	746	{
	747	struct btrfs_root *root = BTRFS_I(fdentry(file)->d_inode)->root;
	748
	749	switch (cmd) {
	750	case BTRFS_IOC_SNAP_CREATE:
	751	return btrfs_ioctl_snap_create(root, (void __user *)arg);
	752	case BTRFS_IOC_DEFRAG:
	753	return btrfs_ioctl_defrag(file);
	754	case BTRFS_IOC_RESIZE:
	755	return btrfs_ioctl_resize(root, (void __user *)arg);
	756	case BTRFS_IOC_ADD_DEV:
	757	return btrfs_ioctl_add_dev(root, (void __user *)arg);
	758	case BTRFS_IOC_RM_DEV:
	759	return btrfs_ioctl_rm_dev(root, (void __user *)arg);
	760	case BTRFS_IOC_BALANCE:
	761	return btrfs_balance(root->fs_info->dev_root);
	762	case BTRFS_IOC_CLONE:
	763	return btrfs_ioctl_clone(file, arg);
	764	case BTRFS_IOC_TRANS_START:
	765	return btrfs_ioctl_trans_start(file);
	766	case BTRFS_IOC_TRANS_END:
	767	return btrfs_ioctl_trans_end(file);
	768	case BTRFS_IOC_SYNC:
	769	btrfs_sync_fs(file->f_dentry->d_sb, 1);
	770	return 0;
	771	}
	772
	773	return -ENOTTY;
	774	}