2 * Copyright (C) 2007 Oracle. All rights reserved.
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.
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.
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.
19 #include <linux/kernel.h>
20 #include <linux/bio.h>
21 #include <linux/buffer_head.h>
22 #include <linux/file.h>
24 #include <linux/fsnotify.h>
25 #include <linux/pagemap.h>
26 #include <linux/highmem.h>
27 #include <linux/time.h>
28 #include <linux/init.h>
29 #include <linux/string.h>
30 #include <linux/backing-dev.h>
31 #include <linux/mount.h>
32 #include <linux/mpage.h>
33 #include <linux/namei.h>
34 #include <linux/swap.h>
35 #include <linux/writeback.h>
36 #include <linux/statfs.h>
37 #include <linux/compat.h>
38 #include <linux/bit_spinlock.h>
39 #include <linux/security.h>
40 #include <linux/xattr.h>
41 #include <linux/vmalloc.h>
45 #include "transaction.h"
46 #include "btrfs_inode.h"
48 #include "print-tree.h"
52 /* Mask out flags that are inappropriate for the given type of inode. */
53 static inline __u32 btrfs_mask_flags(umode_t mode, __u32 flags)
57 else if (S_ISREG(mode))
58 return flags & ~FS_DIRSYNC_FL;
60 return flags & (FS_NODUMP_FL | FS_NOATIME_FL);
64 * Export inode flags to the format expected by the FS_IOC_GETFLAGS ioctl.
66 static unsigned int btrfs_flags_to_ioctl(unsigned int flags)
68 unsigned int iflags = 0;
70 if (flags & BTRFS_INODE_SYNC)
72 if (flags & BTRFS_INODE_IMMUTABLE)
73 iflags |= FS_IMMUTABLE_FL;
74 if (flags & BTRFS_INODE_APPEND)
75 iflags |= FS_APPEND_FL;
76 if (flags & BTRFS_INODE_NODUMP)
77 iflags |= FS_NODUMP_FL;
78 if (flags & BTRFS_INODE_NOATIME)
79 iflags |= FS_NOATIME_FL;
80 if (flags & BTRFS_INODE_DIRSYNC)
81 iflags |= FS_DIRSYNC_FL;
87 * Update inode->i_flags based on the btrfs internal flags.
89 void btrfs_update_iflags(struct inode *inode)
91 struct btrfs_inode *ip = BTRFS_I(inode);
93 inode->i_flags &= ~(S_SYNC|S_APPEND|S_IMMUTABLE|S_NOATIME|S_DIRSYNC);
95 if (ip->flags & BTRFS_INODE_SYNC)
96 inode->i_flags |= S_SYNC;
97 if (ip->flags & BTRFS_INODE_IMMUTABLE)
98 inode->i_flags |= S_IMMUTABLE;
99 if (ip->flags & BTRFS_INODE_APPEND)
100 inode->i_flags |= S_APPEND;
101 if (ip->flags & BTRFS_INODE_NOATIME)
102 inode->i_flags |= S_NOATIME;
103 if (ip->flags & BTRFS_INODE_DIRSYNC)
104 inode->i_flags |= S_DIRSYNC;
108 * Inherit flags from the parent inode.
110 * Unlike extN we don't have any flags we don't want to inherit currently.
112 void btrfs_inherit_iflags(struct inode *inode, struct inode *dir)
119 flags = BTRFS_I(dir)->flags;
121 if (S_ISREG(inode->i_mode))
122 flags &= ~BTRFS_INODE_DIRSYNC;
123 else if (!S_ISDIR(inode->i_mode))
124 flags &= (BTRFS_INODE_NODUMP | BTRFS_INODE_NOATIME);
126 BTRFS_I(inode)->flags = flags;
127 btrfs_update_iflags(inode);
130 static int btrfs_ioctl_getflags(struct file *file, void __user *arg)
132 struct btrfs_inode *ip = BTRFS_I(file->f_path.dentry->d_inode);
133 unsigned int flags = btrfs_flags_to_ioctl(ip->flags);
135 if (copy_to_user(arg, &flags, sizeof(flags)))
140 static int btrfs_ioctl_setflags(struct file *file, void __user *arg)
142 struct inode *inode = file->f_path.dentry->d_inode;
143 struct btrfs_inode *ip = BTRFS_I(inode);
144 struct btrfs_root *root = ip->root;
145 struct btrfs_trans_handle *trans;
146 unsigned int flags, oldflags;
149 if (copy_from_user(&flags, arg, sizeof(flags)))
152 if (flags & ~(FS_IMMUTABLE_FL | FS_APPEND_FL | \
153 FS_NOATIME_FL | FS_NODUMP_FL | \
154 FS_SYNC_FL | FS_DIRSYNC_FL))
157 if (!is_owner_or_cap(inode))
160 mutex_lock(&inode->i_mutex);
162 flags = btrfs_mask_flags(inode->i_mode, flags);
163 oldflags = btrfs_flags_to_ioctl(ip->flags);
164 if ((flags ^ oldflags) & (FS_APPEND_FL | FS_IMMUTABLE_FL)) {
165 if (!capable(CAP_LINUX_IMMUTABLE)) {
171 ret = mnt_want_write(file->f_path.mnt);
175 if (flags & FS_SYNC_FL)
176 ip->flags |= BTRFS_INODE_SYNC;
178 ip->flags &= ~BTRFS_INODE_SYNC;
179 if (flags & FS_IMMUTABLE_FL)
180 ip->flags |= BTRFS_INODE_IMMUTABLE;
182 ip->flags &= ~BTRFS_INODE_IMMUTABLE;
183 if (flags & FS_APPEND_FL)
184 ip->flags |= BTRFS_INODE_APPEND;
186 ip->flags &= ~BTRFS_INODE_APPEND;
187 if (flags & FS_NODUMP_FL)
188 ip->flags |= BTRFS_INODE_NODUMP;
190 ip->flags &= ~BTRFS_INODE_NODUMP;
191 if (flags & FS_NOATIME_FL)
192 ip->flags |= BTRFS_INODE_NOATIME;
194 ip->flags &= ~BTRFS_INODE_NOATIME;
195 if (flags & FS_DIRSYNC_FL)
196 ip->flags |= BTRFS_INODE_DIRSYNC;
198 ip->flags &= ~BTRFS_INODE_DIRSYNC;
201 trans = btrfs_join_transaction(root, 1);
204 ret = btrfs_update_inode(trans, root, inode);
207 btrfs_update_iflags(inode);
208 inode->i_ctime = CURRENT_TIME;
209 btrfs_end_transaction(trans, root);
211 mnt_drop_write(file->f_path.mnt);
213 mutex_unlock(&inode->i_mutex);
217 static int btrfs_ioctl_getversion(struct file *file, int __user *arg)
219 struct inode *inode = file->f_path.dentry->d_inode;
221 return put_user(inode->i_generation, arg);
224 static noinline int create_subvol(struct btrfs_root *root,
225 struct dentry *dentry,
226 char *name, int namelen)
228 struct btrfs_trans_handle *trans;
229 struct btrfs_key key;
230 struct btrfs_root_item root_item;
231 struct btrfs_inode_item *inode_item;
232 struct extent_buffer *leaf;
233 struct btrfs_root *new_root;
234 struct inode *dir = dentry->d_parent->d_inode;
238 u64 new_dirid = BTRFS_FIRST_FREE_OBJECTID;
240 unsigned long nr = 1;
242 ret = btrfs_check_metadata_free_space(root);
246 trans = btrfs_start_transaction(root, 1);
249 ret = btrfs_find_free_objectid(trans, root->fs_info->tree_root,
254 leaf = btrfs_alloc_free_block(trans, root, root->leafsize,
255 0, objectid, NULL, 0, 0, 0);
261 memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header));
262 btrfs_set_header_bytenr(leaf, leaf->start);
263 btrfs_set_header_generation(leaf, trans->transid);
264 btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV);
265 btrfs_set_header_owner(leaf, objectid);
267 write_extent_buffer(leaf, root->fs_info->fsid,
268 (unsigned long)btrfs_header_fsid(leaf),
270 write_extent_buffer(leaf, root->fs_info->chunk_tree_uuid,
271 (unsigned long)btrfs_header_chunk_tree_uuid(leaf),
273 btrfs_mark_buffer_dirty(leaf);
275 inode_item = &root_item.inode;
276 memset(inode_item, 0, sizeof(*inode_item));
277 inode_item->generation = cpu_to_le64(1);
278 inode_item->size = cpu_to_le64(3);
279 inode_item->nlink = cpu_to_le32(1);
280 inode_item->nbytes = cpu_to_le64(root->leafsize);
281 inode_item->mode = cpu_to_le32(S_IFDIR | 0755);
283 btrfs_set_root_bytenr(&root_item, leaf->start);
284 btrfs_set_root_generation(&root_item, trans->transid);
285 btrfs_set_root_level(&root_item, 0);
286 btrfs_set_root_refs(&root_item, 1);
287 btrfs_set_root_used(&root_item, 0);
288 btrfs_set_root_last_snapshot(&root_item, 0);
290 memset(&root_item.drop_progress, 0, sizeof(root_item.drop_progress));
291 root_item.drop_level = 0;
293 btrfs_tree_unlock(leaf);
294 free_extent_buffer(leaf);
297 btrfs_set_root_dirid(&root_item, new_dirid);
299 key.objectid = objectid;
301 btrfs_set_key_type(&key, BTRFS_ROOT_ITEM_KEY);
302 ret = btrfs_insert_root(trans, root->fs_info->tree_root, &key,
307 key.offset = (u64)-1;
308 new_root = btrfs_read_fs_root_no_name(root->fs_info, &key);
309 BUG_ON(IS_ERR(new_root));
311 btrfs_record_root_in_trans(trans, new_root);
313 ret = btrfs_create_subvol_root(trans, new_root, new_dirid,
314 BTRFS_I(dir)->block_group);
316 * insert the directory item
318 ret = btrfs_set_inode_index(dir, &index);
321 ret = btrfs_insert_dir_item(trans, root,
322 name, namelen, dir->i_ino, &key,
323 BTRFS_FT_DIR, index);
327 btrfs_i_size_write(dir, dir->i_size + namelen * 2);
328 ret = btrfs_update_inode(trans, root, dir);
331 ret = btrfs_add_root_ref(trans, root->fs_info->tree_root,
332 objectid, root->root_key.objectid,
333 dir->i_ino, index, name, namelen);
337 d_instantiate(dentry, btrfs_lookup_dentry(dir, dentry));
339 nr = trans->blocks_used;
340 err = btrfs_commit_transaction(trans, root);
346 static int create_snapshot(struct btrfs_root *root, struct dentry *dentry,
347 char *name, int namelen)
349 struct btrfs_pending_snapshot *pending_snapshot;
350 struct btrfs_trans_handle *trans;
353 unsigned long nr = 0;
358 ret = btrfs_check_metadata_free_space(root);
362 pending_snapshot = kzalloc(sizeof(*pending_snapshot), GFP_NOFS);
363 if (!pending_snapshot) {
367 pending_snapshot->name = kmalloc(namelen + 1, GFP_NOFS);
368 if (!pending_snapshot->name) {
370 kfree(pending_snapshot);
373 memcpy(pending_snapshot->name, name, namelen);
374 pending_snapshot->name[namelen] = '\0';
375 pending_snapshot->dentry = dentry;
376 trans = btrfs_start_transaction(root, 1);
378 pending_snapshot->root = root;
379 list_add(&pending_snapshot->list,
380 &trans->transaction->pending_snapshots);
381 err = btrfs_commit_transaction(trans, root);
384 btrfs_btree_balance_dirty(root, nr);
388 /* copy of may_create in fs/namei.c() */
389 static inline int btrfs_may_create(struct inode *dir, struct dentry *child)
395 return inode_permission(dir, MAY_WRITE | MAY_EXEC);
399 * Create a new subvolume below @parent. This is largely modeled after
400 * sys_mkdirat and vfs_mkdir, but we only do a single component lookup
401 * inside this filesystem so it's quite a bit simpler.
403 static noinline int btrfs_mksubvol(struct path *parent,
404 char *name, int namelen,
405 struct btrfs_root *snap_src)
407 struct inode *dir = parent->dentry->d_inode;
408 struct dentry *dentry;
411 mutex_lock_nested(&dir->i_mutex, I_MUTEX_PARENT);
413 dentry = lookup_one_len(name, parent->dentry, namelen);
414 error = PTR_ERR(dentry);
422 error = mnt_want_write(parent->mnt);
426 error = btrfs_may_create(dir, dentry);
430 down_read(&BTRFS_I(dir)->root->fs_info->subvol_sem);
432 if (btrfs_root_refs(&BTRFS_I(dir)->root->root_item) == 0)
436 error = create_snapshot(snap_src, dentry,
439 error = create_subvol(BTRFS_I(dir)->root, dentry,
443 fsnotify_mkdir(dir, dentry);
445 up_read(&BTRFS_I(dir)->root->fs_info->subvol_sem);
447 mnt_drop_write(parent->mnt);
451 mutex_unlock(&dir->i_mutex);
455 static int btrfs_defrag_file(struct file *file)
457 struct inode *inode = fdentry(file)->d_inode;
458 struct btrfs_root *root = BTRFS_I(inode)->root;
459 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
460 struct btrfs_ordered_extent *ordered;
462 unsigned long last_index;
463 unsigned long ra_pages = root->fs_info->bdi.ra_pages;
464 unsigned long total_read = 0;
470 ret = btrfs_check_data_free_space(root, inode, inode->i_size);
474 mutex_lock(&inode->i_mutex);
475 last_index = inode->i_size >> PAGE_CACHE_SHIFT;
476 for (i = 0; i <= last_index; i++) {
477 if (total_read % ra_pages == 0) {
478 btrfs_force_ra(inode->i_mapping, &file->f_ra, file, i,
479 min(last_index, i + ra_pages - 1));
483 page = grab_cache_page(inode->i_mapping, i);
486 if (!PageUptodate(page)) {
487 btrfs_readpage(NULL, page);
489 if (!PageUptodate(page)) {
491 page_cache_release(page);
496 wait_on_page_writeback(page);
498 page_start = (u64)page->index << PAGE_CACHE_SHIFT;
499 page_end = page_start + PAGE_CACHE_SIZE - 1;
500 lock_extent(io_tree, page_start, page_end, GFP_NOFS);
502 ordered = btrfs_lookup_ordered_extent(inode, page_start);
504 unlock_extent(io_tree, page_start, page_end, GFP_NOFS);
506 page_cache_release(page);
507 btrfs_start_ordered_extent(inode, ordered, 1);
508 btrfs_put_ordered_extent(ordered);
511 set_page_extent_mapped(page);
514 * this makes sure page_mkwrite is called on the
515 * page if it is dirtied again later
517 clear_page_dirty_for_io(page);
519 btrfs_set_extent_delalloc(inode, page_start, page_end);
520 set_page_dirty(page);
521 unlock_extent(io_tree, page_start, page_end, GFP_NOFS);
523 page_cache_release(page);
524 balance_dirty_pages_ratelimited_nr(inode->i_mapping, 1);
528 mutex_unlock(&inode->i_mutex);
532 static noinline int btrfs_ioctl_resize(struct btrfs_root *root,
538 struct btrfs_ioctl_vol_args *vol_args;
539 struct btrfs_trans_handle *trans;
540 struct btrfs_device *device = NULL;
547 if (root->fs_info->sb->s_flags & MS_RDONLY)
550 if (!capable(CAP_SYS_ADMIN))
553 vol_args = memdup_user(arg, sizeof(*vol_args));
554 if (IS_ERR(vol_args))
555 return PTR_ERR(vol_args);
557 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
558 namelen = strlen(vol_args->name);
560 mutex_lock(&root->fs_info->volume_mutex);
561 sizestr = vol_args->name;
562 devstr = strchr(sizestr, ':');
565 sizestr = devstr + 1;
567 devstr = vol_args->name;
568 devid = simple_strtoull(devstr, &end, 10);
569 printk(KERN_INFO "resizing devid %llu\n",
570 (unsigned long long)devid);
572 device = btrfs_find_device(root, devid, NULL, NULL);
574 printk(KERN_INFO "resizer unable to find device %llu\n",
575 (unsigned long long)devid);
579 if (!strcmp(sizestr, "max"))
580 new_size = device->bdev->bd_inode->i_size;
582 if (sizestr[0] == '-') {
585 } else if (sizestr[0] == '+') {
589 new_size = btrfs_parse_size(sizestr);
596 old_size = device->total_bytes;
599 if (new_size > old_size) {
603 new_size = old_size - new_size;
604 } else if (mod > 0) {
605 new_size = old_size + new_size;
608 if (new_size < 256 * 1024 * 1024) {
612 if (new_size > device->bdev->bd_inode->i_size) {
617 do_div(new_size, root->sectorsize);
618 new_size *= root->sectorsize;
620 printk(KERN_INFO "new size for %s is %llu\n",
621 device->name, (unsigned long long)new_size);
623 if (new_size > old_size) {
624 trans = btrfs_start_transaction(root, 1);
625 ret = btrfs_grow_device(trans, device, new_size);
626 btrfs_commit_transaction(trans, root);
628 ret = btrfs_shrink_device(device, new_size);
632 mutex_unlock(&root->fs_info->volume_mutex);
637 static noinline int btrfs_ioctl_snap_create(struct file *file,
638 void __user *arg, int subvol)
640 struct btrfs_root *root = BTRFS_I(fdentry(file)->d_inode)->root;
641 struct btrfs_ioctl_vol_args *vol_args;
642 struct file *src_file;
646 if (root->fs_info->sb->s_flags & MS_RDONLY)
649 vol_args = memdup_user(arg, sizeof(*vol_args));
650 if (IS_ERR(vol_args))
651 return PTR_ERR(vol_args);
653 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
654 namelen = strlen(vol_args->name);
655 if (strchr(vol_args->name, '/')) {
661 ret = btrfs_mksubvol(&file->f_path, vol_args->name, namelen,
664 struct inode *src_inode;
665 src_file = fget(vol_args->fd);
671 src_inode = src_file->f_path.dentry->d_inode;
672 if (src_inode->i_sb != file->f_path.dentry->d_inode->i_sb) {
673 printk(KERN_INFO "btrfs: Snapshot src from "
679 ret = btrfs_mksubvol(&file->f_path, vol_args->name, namelen,
680 BTRFS_I(src_inode)->root);
689 * helper to check if the subvolume references other subvolumes
691 static noinline int may_destroy_subvol(struct btrfs_root *root)
693 struct btrfs_path *path;
694 struct btrfs_key key;
697 path = btrfs_alloc_path();
701 key.objectid = root->root_key.objectid;
702 key.type = BTRFS_ROOT_REF_KEY;
703 key.offset = (u64)-1;
705 ret = btrfs_search_slot(NULL, root->fs_info->tree_root,
712 if (path->slots[0] > 0) {
714 btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
715 if (key.objectid == root->root_key.objectid &&
716 key.type == BTRFS_ROOT_REF_KEY)
720 btrfs_free_path(path);
724 static noinline int btrfs_ioctl_snap_destroy(struct file *file,
727 struct dentry *parent = fdentry(file);
728 struct dentry *dentry;
729 struct inode *dir = parent->d_inode;
731 struct btrfs_root *root = BTRFS_I(dir)->root;
732 struct btrfs_root *dest = NULL;
733 struct btrfs_ioctl_vol_args *vol_args;
734 struct btrfs_trans_handle *trans;
739 if (!capable(CAP_SYS_ADMIN))
742 vol_args = memdup_user(arg, sizeof(*vol_args));
743 if (IS_ERR(vol_args))
744 return PTR_ERR(vol_args);
746 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
747 namelen = strlen(vol_args->name);
748 if (strchr(vol_args->name, '/') ||
749 strncmp(vol_args->name, "..", namelen) == 0) {
754 err = mnt_want_write(file->f_path.mnt);
758 mutex_lock_nested(&dir->i_mutex, I_MUTEX_PARENT);
759 dentry = lookup_one_len(vol_args->name, parent, namelen);
760 if (IS_ERR(dentry)) {
761 err = PTR_ERR(dentry);
765 if (!dentry->d_inode) {
770 inode = dentry->d_inode;
771 if (inode->i_ino != BTRFS_FIRST_FREE_OBJECTID) {
776 dest = BTRFS_I(inode)->root;
778 mutex_lock(&inode->i_mutex);
779 err = d_invalidate(dentry);
783 down_write(&root->fs_info->subvol_sem);
785 err = may_destroy_subvol(dest);
789 trans = btrfs_start_transaction(root, 1);
790 ret = btrfs_unlink_subvol(trans, root, dir,
791 dest->root_key.objectid,
796 btrfs_record_root_in_trans(trans, dest);
798 memset(&dest->root_item.drop_progress, 0,
799 sizeof(dest->root_item.drop_progress));
800 dest->root_item.drop_level = 0;
801 btrfs_set_root_refs(&dest->root_item, 0);
803 ret = btrfs_insert_orphan_item(trans,
804 root->fs_info->tree_root,
805 dest->root_key.objectid);
808 ret = btrfs_commit_transaction(trans, root);
810 inode->i_flags |= S_DEAD;
812 up_write(&root->fs_info->subvol_sem);
814 mutex_unlock(&inode->i_mutex);
816 btrfs_invalidate_inodes(dest);
822 mutex_unlock(&dir->i_mutex);
823 mnt_drop_write(file->f_path.mnt);
829 static int btrfs_ioctl_defrag(struct file *file)
831 struct inode *inode = fdentry(file)->d_inode;
832 struct btrfs_root *root = BTRFS_I(inode)->root;
835 ret = mnt_want_write(file->f_path.mnt);
839 switch (inode->i_mode & S_IFMT) {
841 if (!capable(CAP_SYS_ADMIN)) {
845 btrfs_defrag_root(root, 0);
846 btrfs_defrag_root(root->fs_info->extent_root, 0);
849 if (!(file->f_mode & FMODE_WRITE)) {
853 btrfs_defrag_file(file);
857 mnt_drop_write(file->f_path.mnt);
861 static long btrfs_ioctl_add_dev(struct btrfs_root *root, void __user *arg)
863 struct btrfs_ioctl_vol_args *vol_args;
866 if (!capable(CAP_SYS_ADMIN))
869 vol_args = memdup_user(arg, sizeof(*vol_args));
870 if (IS_ERR(vol_args))
871 return PTR_ERR(vol_args);
873 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
874 ret = btrfs_init_new_device(root, vol_args->name);
880 static long btrfs_ioctl_rm_dev(struct btrfs_root *root, void __user *arg)
882 struct btrfs_ioctl_vol_args *vol_args;
885 if (!capable(CAP_SYS_ADMIN))
888 if (root->fs_info->sb->s_flags & MS_RDONLY)
891 vol_args = memdup_user(arg, sizeof(*vol_args));
892 if (IS_ERR(vol_args))
893 return PTR_ERR(vol_args);
895 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
896 ret = btrfs_rm_device(root, vol_args->name);
902 static noinline long btrfs_ioctl_clone(struct file *file, unsigned long srcfd,
903 u64 off, u64 olen, u64 destoff)
905 struct inode *inode = fdentry(file)->d_inode;
906 struct btrfs_root *root = BTRFS_I(inode)->root;
907 struct file *src_file;
909 struct btrfs_trans_handle *trans;
910 struct btrfs_path *path;
911 struct extent_buffer *leaf;
913 struct btrfs_key key;
918 u64 bs = root->fs_info->sb->s_blocksize;
923 * - split compressed inline extents. annoying: we need to
924 * decompress into destination's address_space (the file offset
925 * may change, so source mapping won't do), then recompress (or
926 * otherwise reinsert) a subrange.
927 * - allow ranges within the same file to be cloned (provided
928 * they don't overlap)?
931 /* the destination must be opened for writing */
932 if (!(file->f_mode & FMODE_WRITE))
935 ret = mnt_want_write(file->f_path.mnt);
939 src_file = fget(srcfd);
944 src = src_file->f_dentry->d_inode;
951 if (S_ISDIR(src->i_mode) || S_ISDIR(inode->i_mode))
955 if (src->i_sb != inode->i_sb || BTRFS_I(src)->root != root)
959 buf = vmalloc(btrfs_level_size(root, 0));
963 path = btrfs_alloc_path();
971 mutex_lock(&inode->i_mutex);
972 mutex_lock(&src->i_mutex);
974 mutex_lock(&src->i_mutex);
975 mutex_lock(&inode->i_mutex);
978 /* determine range to clone */
980 if (off >= src->i_size || off + len > src->i_size)
983 olen = len = src->i_size - off;
984 /* if we extend to eof, continue to block boundary */
985 if (off + len == src->i_size)
986 len = ((src->i_size + bs-1) & ~(bs-1))
989 /* verify the end result is block aligned */
990 if ((off & (bs-1)) ||
991 ((off + len) & (bs-1)))
994 /* do any pending delalloc/csum calc on src, one way or
995 another, and lock file content */
997 struct btrfs_ordered_extent *ordered;
998 lock_extent(&BTRFS_I(src)->io_tree, off, off+len, GFP_NOFS);
999 ordered = btrfs_lookup_first_ordered_extent(inode, off+len);
1000 if (BTRFS_I(src)->delalloc_bytes == 0 && !ordered)
1002 unlock_extent(&BTRFS_I(src)->io_tree, off, off+len, GFP_NOFS);
1004 btrfs_put_ordered_extent(ordered);
1005 btrfs_wait_ordered_range(src, off, off+len);
1008 trans = btrfs_start_transaction(root, 1);
1011 /* punch hole in destination first */
1012 btrfs_drop_extents(trans, root, inode, off, off + len,
1013 off + len, 0, &hint_byte, 1);
1016 key.objectid = src->i_ino;
1017 key.type = BTRFS_EXTENT_DATA_KEY;
1022 * note the key will change type as we walk through the
1025 ret = btrfs_search_slot(trans, root, &key, path, 0, 0);
1029 nritems = btrfs_header_nritems(path->nodes[0]);
1030 if (path->slots[0] >= nritems) {
1031 ret = btrfs_next_leaf(root, path);
1036 nritems = btrfs_header_nritems(path->nodes[0]);
1038 leaf = path->nodes[0];
1039 slot = path->slots[0];
1041 btrfs_item_key_to_cpu(leaf, &key, slot);
1042 if (btrfs_key_type(&key) > BTRFS_EXTENT_DATA_KEY ||
1043 key.objectid != src->i_ino)
1046 if (btrfs_key_type(&key) == BTRFS_EXTENT_DATA_KEY) {
1047 struct btrfs_file_extent_item *extent;
1050 struct btrfs_key new_key;
1051 u64 disko = 0, diskl = 0;
1052 u64 datao = 0, datal = 0;
1055 size = btrfs_item_size_nr(leaf, slot);
1056 read_extent_buffer(leaf, buf,
1057 btrfs_item_ptr_offset(leaf, slot),
1060 extent = btrfs_item_ptr(leaf, slot,
1061 struct btrfs_file_extent_item);
1062 comp = btrfs_file_extent_compression(leaf, extent);
1063 type = btrfs_file_extent_type(leaf, extent);
1064 if (type == BTRFS_FILE_EXTENT_REG ||
1065 type == BTRFS_FILE_EXTENT_PREALLOC) {
1066 disko = btrfs_file_extent_disk_bytenr(leaf,
1068 diskl = btrfs_file_extent_disk_num_bytes(leaf,
1070 datao = btrfs_file_extent_offset(leaf, extent);
1071 datal = btrfs_file_extent_num_bytes(leaf,
1073 } else if (type == BTRFS_FILE_EXTENT_INLINE) {
1074 /* take upper bound, may be compressed */
1075 datal = btrfs_file_extent_ram_bytes(leaf,
1078 btrfs_release_path(root, path);
1080 if (key.offset + datal < off ||
1081 key.offset >= off+len)
1084 memcpy(&new_key, &key, sizeof(new_key));
1085 new_key.objectid = inode->i_ino;
1086 new_key.offset = key.offset + destoff - off;
1088 if (type == BTRFS_FILE_EXTENT_REG ||
1089 type == BTRFS_FILE_EXTENT_PREALLOC) {
1090 ret = btrfs_insert_empty_item(trans, root, path,
1095 leaf = path->nodes[0];
1096 slot = path->slots[0];
1097 write_extent_buffer(leaf, buf,
1098 btrfs_item_ptr_offset(leaf, slot),
1101 extent = btrfs_item_ptr(leaf, slot,
1102 struct btrfs_file_extent_item);
1104 if (off > key.offset) {
1105 datao += off - key.offset;
1106 datal -= off - key.offset;
1108 if (key.offset + datao + datal + key.offset >
1110 datal = off + len - key.offset - datao;
1111 /* disko == 0 means it's a hole */
1115 btrfs_set_file_extent_offset(leaf, extent,
1117 btrfs_set_file_extent_num_bytes(leaf, extent,
1120 inode_add_bytes(inode, datal);
1121 ret = btrfs_inc_extent_ref(trans, root,
1123 root->root_key.objectid,
1125 new_key.offset - datao);
1128 } else if (type == BTRFS_FILE_EXTENT_INLINE) {
1131 if (off > key.offset) {
1132 skip = off - key.offset;
1133 new_key.offset += skip;
1136 if (key.offset + datal > off+len)
1137 trim = key.offset + datal - (off+len);
1139 if (comp && (skip || trim)) {
1143 size -= skip + trim;
1144 datal -= skip + trim;
1145 ret = btrfs_insert_empty_item(trans, root, path,
1152 btrfs_file_extent_calc_inline_size(0);
1153 memmove(buf+start, buf+start+skip,
1157 leaf = path->nodes[0];
1158 slot = path->slots[0];
1159 write_extent_buffer(leaf, buf,
1160 btrfs_item_ptr_offset(leaf, slot),
1162 inode_add_bytes(inode, datal);
1165 btrfs_mark_buffer_dirty(leaf);
1169 btrfs_release_path(root, path);
1174 btrfs_release_path(root, path);
1176 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
1177 if (destoff + olen > inode->i_size)
1178 btrfs_i_size_write(inode, destoff + olen);
1179 BTRFS_I(inode)->flags = BTRFS_I(src)->flags;
1180 ret = btrfs_update_inode(trans, root, inode);
1182 btrfs_end_transaction(trans, root);
1183 unlock_extent(&BTRFS_I(src)->io_tree, off, off+len, GFP_NOFS);
1185 vmtruncate(inode, 0);
1187 mutex_unlock(&src->i_mutex);
1188 mutex_unlock(&inode->i_mutex);
1190 btrfs_free_path(path);
1194 mnt_drop_write(file->f_path.mnt);
1198 static long btrfs_ioctl_clone_range(struct file *file, void __user *argp)
1200 struct btrfs_ioctl_clone_range_args args;
1202 if (copy_from_user(&args, argp, sizeof(args)))
1204 return btrfs_ioctl_clone(file, args.src_fd, args.src_offset,
1205 args.src_length, args.dest_offset);
1209 * there are many ways the trans_start and trans_end ioctls can lead
1210 * to deadlocks. They should only be used by applications that
1211 * basically own the machine, and have a very in depth understanding
1212 * of all the possible deadlocks and enospc problems.
1214 static long btrfs_ioctl_trans_start(struct file *file)
1216 struct inode *inode = fdentry(file)->d_inode;
1217 struct btrfs_root *root = BTRFS_I(inode)->root;
1218 struct btrfs_trans_handle *trans;
1221 if (!capable(CAP_SYS_ADMIN))
1224 if (file->private_data) {
1229 ret = mnt_want_write(file->f_path.mnt);
1233 mutex_lock(&root->fs_info->trans_mutex);
1234 root->fs_info->open_ioctl_trans++;
1235 mutex_unlock(&root->fs_info->trans_mutex);
1237 trans = btrfs_start_ioctl_transaction(root, 0);
1239 file->private_data = trans;
1242 /*printk(KERN_INFO "btrfs_ioctl_trans_start on %p\n", file);*/
1248 * there are many ways the trans_start and trans_end ioctls can lead
1249 * to deadlocks. They should only be used by applications that
1250 * basically own the machine, and have a very in depth understanding
1251 * of all the possible deadlocks and enospc problems.
1253 long btrfs_ioctl_trans_end(struct file *file)
1255 struct inode *inode = fdentry(file)->d_inode;
1256 struct btrfs_root *root = BTRFS_I(inode)->root;
1257 struct btrfs_trans_handle *trans;
1260 trans = file->private_data;
1265 btrfs_end_transaction(trans, root);
1266 file->private_data = NULL;
1268 mutex_lock(&root->fs_info->trans_mutex);
1269 root->fs_info->open_ioctl_trans--;
1270 mutex_unlock(&root->fs_info->trans_mutex);
1272 mnt_drop_write(file->f_path.mnt);
1278 long btrfs_ioctl(struct file *file, unsigned int
1279 cmd, unsigned long arg)
1281 struct btrfs_root *root = BTRFS_I(fdentry(file)->d_inode)->root;
1282 void __user *argp = (void __user *)arg;
1285 case FS_IOC_GETFLAGS:
1286 return btrfs_ioctl_getflags(file, argp);
1287 case FS_IOC_SETFLAGS:
1288 return btrfs_ioctl_setflags(file, argp);
1289 case FS_IOC_GETVERSION:
1290 return btrfs_ioctl_getversion(file, argp);
1291 case BTRFS_IOC_SNAP_CREATE:
1292 return btrfs_ioctl_snap_create(file, argp, 0);
1293 case BTRFS_IOC_SUBVOL_CREATE:
1294 return btrfs_ioctl_snap_create(file, argp, 1);
1295 case BTRFS_IOC_SNAP_DESTROY:
1296 return btrfs_ioctl_snap_destroy(file, argp);
1297 case BTRFS_IOC_DEFRAG:
1298 return btrfs_ioctl_defrag(file);
1299 case BTRFS_IOC_RESIZE:
1300 return btrfs_ioctl_resize(root, argp);
1301 case BTRFS_IOC_ADD_DEV:
1302 return btrfs_ioctl_add_dev(root, argp);
1303 case BTRFS_IOC_RM_DEV:
1304 return btrfs_ioctl_rm_dev(root, argp);
1305 case BTRFS_IOC_BALANCE:
1306 return btrfs_balance(root->fs_info->dev_root);
1307 case BTRFS_IOC_CLONE:
1308 return btrfs_ioctl_clone(file, arg, 0, 0, 0);
1309 case BTRFS_IOC_CLONE_RANGE:
1310 return btrfs_ioctl_clone_range(file, argp);
1311 case BTRFS_IOC_TRANS_START:
1312 return btrfs_ioctl_trans_start(file);
1313 case BTRFS_IOC_TRANS_END:
1314 return btrfs_ioctl_trans_end(file);
1315 case BTRFS_IOC_SYNC:
1316 btrfs_sync_fs(file->f_dentry->d_sb, 1);