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;
247 ret = btrfs_reserve_metadata_space(root, 6);
251 trans = btrfs_start_transaction(root, 1);
254 ret = btrfs_find_free_objectid(trans, root->fs_info->tree_root,
259 leaf = btrfs_alloc_free_block(trans, root, root->leafsize,
260 0, objectid, NULL, 0, 0, 0);
266 memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header));
267 btrfs_set_header_bytenr(leaf, leaf->start);
268 btrfs_set_header_generation(leaf, trans->transid);
269 btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV);
270 btrfs_set_header_owner(leaf, objectid);
272 write_extent_buffer(leaf, root->fs_info->fsid,
273 (unsigned long)btrfs_header_fsid(leaf),
275 write_extent_buffer(leaf, root->fs_info->chunk_tree_uuid,
276 (unsigned long)btrfs_header_chunk_tree_uuid(leaf),
278 btrfs_mark_buffer_dirty(leaf);
280 inode_item = &root_item.inode;
281 memset(inode_item, 0, sizeof(*inode_item));
282 inode_item->generation = cpu_to_le64(1);
283 inode_item->size = cpu_to_le64(3);
284 inode_item->nlink = cpu_to_le32(1);
285 inode_item->nbytes = cpu_to_le64(root->leafsize);
286 inode_item->mode = cpu_to_le32(S_IFDIR | 0755);
289 root_item.byte_limit = 0;
290 inode_item->flags = cpu_to_le64(BTRFS_INODE_ROOT_ITEM_INIT);
292 btrfs_set_root_bytenr(&root_item, leaf->start);
293 btrfs_set_root_generation(&root_item, trans->transid);
294 btrfs_set_root_level(&root_item, 0);
295 btrfs_set_root_refs(&root_item, 1);
296 btrfs_set_root_used(&root_item, leaf->len);
297 btrfs_set_root_last_snapshot(&root_item, 0);
299 memset(&root_item.drop_progress, 0, sizeof(root_item.drop_progress));
300 root_item.drop_level = 0;
302 btrfs_tree_unlock(leaf);
303 free_extent_buffer(leaf);
306 btrfs_set_root_dirid(&root_item, new_dirid);
308 key.objectid = objectid;
310 btrfs_set_key_type(&key, BTRFS_ROOT_ITEM_KEY);
311 ret = btrfs_insert_root(trans, root->fs_info->tree_root, &key,
316 key.offset = (u64)-1;
317 new_root = btrfs_read_fs_root_no_name(root->fs_info, &key);
318 BUG_ON(IS_ERR(new_root));
320 btrfs_record_root_in_trans(trans, new_root);
322 ret = btrfs_create_subvol_root(trans, new_root, new_dirid,
323 BTRFS_I(dir)->block_group);
325 * insert the directory item
327 ret = btrfs_set_inode_index(dir, &index);
330 ret = btrfs_insert_dir_item(trans, root,
331 name, namelen, dir->i_ino, &key,
332 BTRFS_FT_DIR, index);
336 btrfs_i_size_write(dir, dir->i_size + namelen * 2);
337 ret = btrfs_update_inode(trans, root, dir);
340 ret = btrfs_add_root_ref(trans, root->fs_info->tree_root,
341 objectid, root->root_key.objectid,
342 dir->i_ino, index, name, namelen);
346 d_instantiate(dentry, btrfs_lookup_dentry(dir, dentry));
348 err = btrfs_commit_transaction(trans, root);
352 btrfs_unreserve_metadata_space(root, 6);
356 static int create_snapshot(struct btrfs_root *root, struct dentry *dentry,
357 char *name, int namelen)
360 struct btrfs_pending_snapshot *pending_snapshot;
361 struct btrfs_trans_handle *trans;
373 ret = btrfs_reserve_metadata_space(root, 6);
377 pending_snapshot = kzalloc(sizeof(*pending_snapshot), GFP_NOFS);
378 if (!pending_snapshot) {
380 btrfs_unreserve_metadata_space(root, 6);
383 pending_snapshot->name = kmalloc(namelen + 1, GFP_NOFS);
384 if (!pending_snapshot->name) {
386 kfree(pending_snapshot);
387 btrfs_unreserve_metadata_space(root, 6);
390 memcpy(pending_snapshot->name, name, namelen);
391 pending_snapshot->name[namelen] = '\0';
392 pending_snapshot->dentry = dentry;
393 trans = btrfs_start_transaction(root, 1);
395 pending_snapshot->root = root;
396 list_add(&pending_snapshot->list,
397 &trans->transaction->pending_snapshots);
398 ret = btrfs_commit_transaction(trans, root);
400 btrfs_unreserve_metadata_space(root, 6);
402 inode = btrfs_lookup_dentry(dentry->d_parent->d_inode, dentry);
404 ret = PTR_ERR(inode);
408 d_instantiate(dentry, inode);
414 /* copy of may_create in fs/namei.c() */
415 static inline int btrfs_may_create(struct inode *dir, struct dentry *child)
421 return inode_permission(dir, MAY_WRITE | MAY_EXEC);
425 * Create a new subvolume below @parent. This is largely modeled after
426 * sys_mkdirat and vfs_mkdir, but we only do a single component lookup
427 * inside this filesystem so it's quite a bit simpler.
429 static noinline int btrfs_mksubvol(struct path *parent,
430 char *name, int namelen,
431 struct btrfs_root *snap_src)
433 struct inode *dir = parent->dentry->d_inode;
434 struct dentry *dentry;
437 mutex_lock_nested(&dir->i_mutex, I_MUTEX_PARENT);
439 dentry = lookup_one_len(name, parent->dentry, namelen);
440 error = PTR_ERR(dentry);
448 error = mnt_want_write(parent->mnt);
452 error = btrfs_may_create(dir, dentry);
456 down_read(&BTRFS_I(dir)->root->fs_info->subvol_sem);
458 if (btrfs_root_refs(&BTRFS_I(dir)->root->root_item) == 0)
462 error = create_snapshot(snap_src, dentry,
465 error = create_subvol(BTRFS_I(dir)->root, dentry,
469 fsnotify_mkdir(dir, dentry);
471 up_read(&BTRFS_I(dir)->root->fs_info->subvol_sem);
473 mnt_drop_write(parent->mnt);
477 mutex_unlock(&dir->i_mutex);
481 static int btrfs_defrag_file(struct file *file)
483 struct inode *inode = fdentry(file)->d_inode;
484 struct btrfs_root *root = BTRFS_I(inode)->root;
485 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
486 struct btrfs_ordered_extent *ordered;
488 unsigned long last_index;
489 unsigned long ra_pages = root->fs_info->bdi.ra_pages;
490 unsigned long total_read = 0;
496 ret = btrfs_check_data_free_space(root, inode, inode->i_size);
500 mutex_lock(&inode->i_mutex);
501 last_index = inode->i_size >> PAGE_CACHE_SHIFT;
502 for (i = 0; i <= last_index; i++) {
503 if (total_read % ra_pages == 0) {
504 btrfs_force_ra(inode->i_mapping, &file->f_ra, file, i,
505 min(last_index, i + ra_pages - 1));
509 page = grab_cache_page(inode->i_mapping, i);
512 if (!PageUptodate(page)) {
513 btrfs_readpage(NULL, page);
515 if (!PageUptodate(page)) {
517 page_cache_release(page);
522 wait_on_page_writeback(page);
524 page_start = (u64)page->index << PAGE_CACHE_SHIFT;
525 page_end = page_start + PAGE_CACHE_SIZE - 1;
526 lock_extent(io_tree, page_start, page_end, GFP_NOFS);
528 ordered = btrfs_lookup_ordered_extent(inode, page_start);
530 unlock_extent(io_tree, page_start, page_end, GFP_NOFS);
532 page_cache_release(page);
533 btrfs_start_ordered_extent(inode, ordered, 1);
534 btrfs_put_ordered_extent(ordered);
537 set_page_extent_mapped(page);
540 * this makes sure page_mkwrite is called on the
541 * page if it is dirtied again later
543 clear_page_dirty_for_io(page);
545 btrfs_set_extent_delalloc(inode, page_start, page_end);
546 set_page_dirty(page);
547 unlock_extent(io_tree, page_start, page_end, GFP_NOFS);
549 page_cache_release(page);
550 balance_dirty_pages_ratelimited_nr(inode->i_mapping, 1);
554 mutex_unlock(&inode->i_mutex);
558 static noinline int btrfs_ioctl_resize(struct btrfs_root *root,
564 struct btrfs_ioctl_vol_args *vol_args;
565 struct btrfs_trans_handle *trans;
566 struct btrfs_device *device = NULL;
573 if (root->fs_info->sb->s_flags & MS_RDONLY)
576 if (!capable(CAP_SYS_ADMIN))
579 vol_args = memdup_user(arg, sizeof(*vol_args));
580 if (IS_ERR(vol_args))
581 return PTR_ERR(vol_args);
583 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
584 namelen = strlen(vol_args->name);
586 mutex_lock(&root->fs_info->volume_mutex);
587 sizestr = vol_args->name;
588 devstr = strchr(sizestr, ':');
591 sizestr = devstr + 1;
593 devstr = vol_args->name;
594 devid = simple_strtoull(devstr, &end, 10);
595 printk(KERN_INFO "resizing devid %llu\n",
596 (unsigned long long)devid);
598 device = btrfs_find_device(root, devid, NULL, NULL);
600 printk(KERN_INFO "resizer unable to find device %llu\n",
601 (unsigned long long)devid);
605 if (!strcmp(sizestr, "max"))
606 new_size = device->bdev->bd_inode->i_size;
608 if (sizestr[0] == '-') {
611 } else if (sizestr[0] == '+') {
615 new_size = btrfs_parse_size(sizestr);
622 old_size = device->total_bytes;
625 if (new_size > old_size) {
629 new_size = old_size - new_size;
630 } else if (mod > 0) {
631 new_size = old_size + new_size;
634 if (new_size < 256 * 1024 * 1024) {
638 if (new_size > device->bdev->bd_inode->i_size) {
643 do_div(new_size, root->sectorsize);
644 new_size *= root->sectorsize;
646 printk(KERN_INFO "new size for %s is %llu\n",
647 device->name, (unsigned long long)new_size);
649 if (new_size > old_size) {
650 trans = btrfs_start_transaction(root, 1);
651 ret = btrfs_grow_device(trans, device, new_size);
652 btrfs_commit_transaction(trans, root);
654 ret = btrfs_shrink_device(device, new_size);
658 mutex_unlock(&root->fs_info->volume_mutex);
663 static noinline int btrfs_ioctl_snap_create(struct file *file,
664 void __user *arg, int subvol)
666 struct btrfs_root *root = BTRFS_I(fdentry(file)->d_inode)->root;
667 struct btrfs_ioctl_vol_args *vol_args;
668 struct file *src_file;
672 if (root->fs_info->sb->s_flags & MS_RDONLY)
675 vol_args = memdup_user(arg, sizeof(*vol_args));
676 if (IS_ERR(vol_args))
677 return PTR_ERR(vol_args);
679 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
680 namelen = strlen(vol_args->name);
681 if (strchr(vol_args->name, '/')) {
687 ret = btrfs_mksubvol(&file->f_path, vol_args->name, namelen,
690 struct inode *src_inode;
691 src_file = fget(vol_args->fd);
697 src_inode = src_file->f_path.dentry->d_inode;
698 if (src_inode->i_sb != file->f_path.dentry->d_inode->i_sb) {
699 printk(KERN_INFO "btrfs: Snapshot src from "
705 ret = btrfs_mksubvol(&file->f_path, vol_args->name, namelen,
706 BTRFS_I(src_inode)->root);
715 * helper to check if the subvolume references other subvolumes
717 static noinline int may_destroy_subvol(struct btrfs_root *root)
719 struct btrfs_path *path;
720 struct btrfs_key key;
723 path = btrfs_alloc_path();
727 key.objectid = root->root_key.objectid;
728 key.type = BTRFS_ROOT_REF_KEY;
729 key.offset = (u64)-1;
731 ret = btrfs_search_slot(NULL, root->fs_info->tree_root,
738 if (path->slots[0] > 0) {
740 btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
741 if (key.objectid == root->root_key.objectid &&
742 key.type == BTRFS_ROOT_REF_KEY)
746 btrfs_free_path(path);
750 static noinline int btrfs_ioctl_snap_destroy(struct file *file,
753 struct dentry *parent = fdentry(file);
754 struct dentry *dentry;
755 struct inode *dir = parent->d_inode;
757 struct btrfs_root *root = BTRFS_I(dir)->root;
758 struct btrfs_root *dest = NULL;
759 struct btrfs_ioctl_vol_args *vol_args;
760 struct btrfs_trans_handle *trans;
765 if (!capable(CAP_SYS_ADMIN))
768 vol_args = memdup_user(arg, sizeof(*vol_args));
769 if (IS_ERR(vol_args))
770 return PTR_ERR(vol_args);
772 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
773 namelen = strlen(vol_args->name);
774 if (strchr(vol_args->name, '/') ||
775 strncmp(vol_args->name, "..", namelen) == 0) {
780 err = mnt_want_write(file->f_path.mnt);
784 mutex_lock_nested(&dir->i_mutex, I_MUTEX_PARENT);
785 dentry = lookup_one_len(vol_args->name, parent, namelen);
786 if (IS_ERR(dentry)) {
787 err = PTR_ERR(dentry);
791 if (!dentry->d_inode) {
796 inode = dentry->d_inode;
797 if (inode->i_ino != BTRFS_FIRST_FREE_OBJECTID) {
802 dest = BTRFS_I(inode)->root;
804 mutex_lock(&inode->i_mutex);
805 err = d_invalidate(dentry);
809 down_write(&root->fs_info->subvol_sem);
811 err = may_destroy_subvol(dest);
815 trans = btrfs_start_transaction(root, 1);
816 ret = btrfs_unlink_subvol(trans, root, dir,
817 dest->root_key.objectid,
822 btrfs_record_root_in_trans(trans, dest);
824 memset(&dest->root_item.drop_progress, 0,
825 sizeof(dest->root_item.drop_progress));
826 dest->root_item.drop_level = 0;
827 btrfs_set_root_refs(&dest->root_item, 0);
829 ret = btrfs_insert_orphan_item(trans,
830 root->fs_info->tree_root,
831 dest->root_key.objectid);
834 ret = btrfs_commit_transaction(trans, root);
836 inode->i_flags |= S_DEAD;
838 up_write(&root->fs_info->subvol_sem);
840 mutex_unlock(&inode->i_mutex);
842 shrink_dcache_sb(root->fs_info->sb);
843 btrfs_invalidate_inodes(dest);
849 mutex_unlock(&dir->i_mutex);
850 mnt_drop_write(file->f_path.mnt);
856 static int btrfs_ioctl_defrag(struct file *file)
858 struct inode *inode = fdentry(file)->d_inode;
859 struct btrfs_root *root = BTRFS_I(inode)->root;
862 ret = mnt_want_write(file->f_path.mnt);
866 switch (inode->i_mode & S_IFMT) {
868 if (!capable(CAP_SYS_ADMIN)) {
872 btrfs_defrag_root(root, 0);
873 btrfs_defrag_root(root->fs_info->extent_root, 0);
876 if (!(file->f_mode & FMODE_WRITE)) {
880 btrfs_defrag_file(file);
884 mnt_drop_write(file->f_path.mnt);
888 static long btrfs_ioctl_add_dev(struct btrfs_root *root, void __user *arg)
890 struct btrfs_ioctl_vol_args *vol_args;
893 if (!capable(CAP_SYS_ADMIN))
896 vol_args = memdup_user(arg, sizeof(*vol_args));
897 if (IS_ERR(vol_args))
898 return PTR_ERR(vol_args);
900 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
901 ret = btrfs_init_new_device(root, vol_args->name);
907 static long btrfs_ioctl_rm_dev(struct btrfs_root *root, void __user *arg)
909 struct btrfs_ioctl_vol_args *vol_args;
912 if (!capable(CAP_SYS_ADMIN))
915 if (root->fs_info->sb->s_flags & MS_RDONLY)
918 vol_args = memdup_user(arg, sizeof(*vol_args));
919 if (IS_ERR(vol_args))
920 return PTR_ERR(vol_args);
922 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
923 ret = btrfs_rm_device(root, vol_args->name);
929 static noinline long btrfs_ioctl_clone(struct file *file, unsigned long srcfd,
930 u64 off, u64 olen, u64 destoff)
932 struct inode *inode = fdentry(file)->d_inode;
933 struct btrfs_root *root = BTRFS_I(inode)->root;
934 struct file *src_file;
936 struct btrfs_trans_handle *trans;
937 struct btrfs_path *path;
938 struct extent_buffer *leaf;
940 struct btrfs_key key;
945 u64 bs = root->fs_info->sb->s_blocksize;
950 * - split compressed inline extents. annoying: we need to
951 * decompress into destination's address_space (the file offset
952 * may change, so source mapping won't do), then recompress (or
953 * otherwise reinsert) a subrange.
954 * - allow ranges within the same file to be cloned (provided
955 * they don't overlap)?
958 /* the destination must be opened for writing */
959 if (!(file->f_mode & FMODE_WRITE) || (file->f_flags & O_APPEND))
962 ret = mnt_want_write(file->f_path.mnt);
966 src_file = fget(srcfd);
972 src = src_file->f_dentry->d_inode;
978 /* the src must be open for reading */
979 if (!(src_file->f_mode & FMODE_READ))
983 if (S_ISDIR(src->i_mode) || S_ISDIR(inode->i_mode))
987 if (src->i_sb != inode->i_sb || BTRFS_I(src)->root != root)
991 buf = vmalloc(btrfs_level_size(root, 0));
995 path = btrfs_alloc_path();
1003 mutex_lock(&inode->i_mutex);
1004 mutex_lock(&src->i_mutex);
1006 mutex_lock(&src->i_mutex);
1007 mutex_lock(&inode->i_mutex);
1010 /* determine range to clone */
1012 if (off + len > src->i_size || off + len < off)
1015 olen = len = src->i_size - off;
1016 /* if we extend to eof, continue to block boundary */
1017 if (off + len == src->i_size)
1018 len = ((src->i_size + bs-1) & ~(bs-1))
1021 /* verify the end result is block aligned */
1022 if ((off & (bs-1)) ||
1023 ((off + len) & (bs-1)))
1026 /* do any pending delalloc/csum calc on src, one way or
1027 another, and lock file content */
1029 struct btrfs_ordered_extent *ordered;
1030 lock_extent(&BTRFS_I(src)->io_tree, off, off+len, GFP_NOFS);
1031 ordered = btrfs_lookup_first_ordered_extent(inode, off+len);
1032 if (BTRFS_I(src)->delalloc_bytes == 0 && !ordered)
1034 unlock_extent(&BTRFS_I(src)->io_tree, off, off+len, GFP_NOFS);
1036 btrfs_put_ordered_extent(ordered);
1037 btrfs_wait_ordered_range(src, off, off+len);
1040 trans = btrfs_start_transaction(root, 1);
1043 /* punch hole in destination first */
1044 btrfs_drop_extents(trans, inode, off, off + len, &hint_byte, 1);
1047 key.objectid = src->i_ino;
1048 key.type = BTRFS_EXTENT_DATA_KEY;
1053 * note the key will change type as we walk through the
1056 ret = btrfs_search_slot(trans, root, &key, path, 0, 0);
1060 nritems = btrfs_header_nritems(path->nodes[0]);
1061 if (path->slots[0] >= nritems) {
1062 ret = btrfs_next_leaf(root, path);
1067 nritems = btrfs_header_nritems(path->nodes[0]);
1069 leaf = path->nodes[0];
1070 slot = path->slots[0];
1072 btrfs_item_key_to_cpu(leaf, &key, slot);
1073 if (btrfs_key_type(&key) > BTRFS_EXTENT_DATA_KEY ||
1074 key.objectid != src->i_ino)
1077 if (btrfs_key_type(&key) == BTRFS_EXTENT_DATA_KEY) {
1078 struct btrfs_file_extent_item *extent;
1081 struct btrfs_key new_key;
1082 u64 disko = 0, diskl = 0;
1083 u64 datao = 0, datal = 0;
1086 size = btrfs_item_size_nr(leaf, slot);
1087 read_extent_buffer(leaf, buf,
1088 btrfs_item_ptr_offset(leaf, slot),
1091 extent = btrfs_item_ptr(leaf, slot,
1092 struct btrfs_file_extent_item);
1093 comp = btrfs_file_extent_compression(leaf, extent);
1094 type = btrfs_file_extent_type(leaf, extent);
1095 if (type == BTRFS_FILE_EXTENT_REG ||
1096 type == BTRFS_FILE_EXTENT_PREALLOC) {
1097 disko = btrfs_file_extent_disk_bytenr(leaf,
1099 diskl = btrfs_file_extent_disk_num_bytes(leaf,
1101 datao = btrfs_file_extent_offset(leaf, extent);
1102 datal = btrfs_file_extent_num_bytes(leaf,
1104 } else if (type == BTRFS_FILE_EXTENT_INLINE) {
1105 /* take upper bound, may be compressed */
1106 datal = btrfs_file_extent_ram_bytes(leaf,
1109 btrfs_release_path(root, path);
1111 if (key.offset + datal < off ||
1112 key.offset >= off+len)
1115 memcpy(&new_key, &key, sizeof(new_key));
1116 new_key.objectid = inode->i_ino;
1117 new_key.offset = key.offset + destoff - off;
1119 if (type == BTRFS_FILE_EXTENT_REG ||
1120 type == BTRFS_FILE_EXTENT_PREALLOC) {
1121 ret = btrfs_insert_empty_item(trans, root, path,
1126 leaf = path->nodes[0];
1127 slot = path->slots[0];
1128 write_extent_buffer(leaf, buf,
1129 btrfs_item_ptr_offset(leaf, slot),
1132 extent = btrfs_item_ptr(leaf, slot,
1133 struct btrfs_file_extent_item);
1135 if (off > key.offset) {
1136 datao += off - key.offset;
1137 datal -= off - key.offset;
1140 if (key.offset + datal > off + len)
1141 datal = off + len - key.offset;
1143 /* disko == 0 means it's a hole */
1147 btrfs_set_file_extent_offset(leaf, extent,
1149 btrfs_set_file_extent_num_bytes(leaf, extent,
1152 inode_add_bytes(inode, datal);
1153 ret = btrfs_inc_extent_ref(trans, root,
1155 root->root_key.objectid,
1157 new_key.offset - datao);
1160 } else if (type == BTRFS_FILE_EXTENT_INLINE) {
1163 if (off > key.offset) {
1164 skip = off - key.offset;
1165 new_key.offset += skip;
1168 if (key.offset + datal > off+len)
1169 trim = key.offset + datal - (off+len);
1171 if (comp && (skip || trim)) {
1175 size -= skip + trim;
1176 datal -= skip + trim;
1177 ret = btrfs_insert_empty_item(trans, root, path,
1184 btrfs_file_extent_calc_inline_size(0);
1185 memmove(buf+start, buf+start+skip,
1189 leaf = path->nodes[0];
1190 slot = path->slots[0];
1191 write_extent_buffer(leaf, buf,
1192 btrfs_item_ptr_offset(leaf, slot),
1194 inode_add_bytes(inode, datal);
1197 btrfs_mark_buffer_dirty(leaf);
1201 btrfs_release_path(root, path);
1206 btrfs_release_path(root, path);
1208 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
1209 if (destoff + olen > inode->i_size)
1210 btrfs_i_size_write(inode, destoff + olen);
1211 BTRFS_I(inode)->flags = BTRFS_I(src)->flags;
1212 ret = btrfs_update_inode(trans, root, inode);
1214 btrfs_end_transaction(trans, root);
1215 unlock_extent(&BTRFS_I(src)->io_tree, off, off+len, GFP_NOFS);
1217 vmtruncate(inode, 0);
1219 mutex_unlock(&src->i_mutex);
1220 mutex_unlock(&inode->i_mutex);
1222 btrfs_free_path(path);
1226 mnt_drop_write(file->f_path.mnt);
1230 static long btrfs_ioctl_clone_range(struct file *file, void __user *argp)
1232 struct btrfs_ioctl_clone_range_args args;
1234 if (copy_from_user(&args, argp, sizeof(args)))
1236 return btrfs_ioctl_clone(file, args.src_fd, args.src_offset,
1237 args.src_length, args.dest_offset);
1241 * there are many ways the trans_start and trans_end ioctls can lead
1242 * to deadlocks. They should only be used by applications that
1243 * basically own the machine, and have a very in depth understanding
1244 * of all the possible deadlocks and enospc problems.
1246 static long btrfs_ioctl_trans_start(struct file *file)
1248 struct inode *inode = fdentry(file)->d_inode;
1249 struct btrfs_root *root = BTRFS_I(inode)->root;
1250 struct btrfs_trans_handle *trans;
1254 if (!capable(CAP_SYS_ADMIN))
1258 if (file->private_data)
1261 ret = mnt_want_write(file->f_path.mnt);
1265 mutex_lock(&root->fs_info->trans_mutex);
1266 root->fs_info->open_ioctl_trans++;
1267 mutex_unlock(&root->fs_info->trans_mutex);
1270 trans = btrfs_start_ioctl_transaction(root, 0);
1274 file->private_data = trans;
1278 mutex_lock(&root->fs_info->trans_mutex);
1279 root->fs_info->open_ioctl_trans--;
1280 mutex_unlock(&root->fs_info->trans_mutex);
1281 mnt_drop_write(file->f_path.mnt);
1287 * there are many ways the trans_start and trans_end ioctls can lead
1288 * to deadlocks. They should only be used by applications that
1289 * basically own the machine, and have a very in depth understanding
1290 * of all the possible deadlocks and enospc problems.
1292 long btrfs_ioctl_trans_end(struct file *file)
1294 struct inode *inode = fdentry(file)->d_inode;
1295 struct btrfs_root *root = BTRFS_I(inode)->root;
1296 struct btrfs_trans_handle *trans;
1298 trans = file->private_data;
1301 file->private_data = NULL;
1303 btrfs_end_transaction(trans, root);
1305 mutex_lock(&root->fs_info->trans_mutex);
1306 root->fs_info->open_ioctl_trans--;
1307 mutex_unlock(&root->fs_info->trans_mutex);
1309 mnt_drop_write(file->f_path.mnt);
1313 long btrfs_ioctl(struct file *file, unsigned int
1314 cmd, unsigned long arg)
1316 struct btrfs_root *root = BTRFS_I(fdentry(file)->d_inode)->root;
1317 void __user *argp = (void __user *)arg;
1320 case FS_IOC_GETFLAGS:
1321 return btrfs_ioctl_getflags(file, argp);
1322 case FS_IOC_SETFLAGS:
1323 return btrfs_ioctl_setflags(file, argp);
1324 case FS_IOC_GETVERSION:
1325 return btrfs_ioctl_getversion(file, argp);
1326 case BTRFS_IOC_SNAP_CREATE:
1327 return btrfs_ioctl_snap_create(file, argp, 0);
1328 case BTRFS_IOC_SUBVOL_CREATE:
1329 return btrfs_ioctl_snap_create(file, argp, 1);
1330 case BTRFS_IOC_SNAP_DESTROY:
1331 return btrfs_ioctl_snap_destroy(file, argp);
1332 case BTRFS_IOC_DEFRAG:
1333 return btrfs_ioctl_defrag(file);
1334 case BTRFS_IOC_RESIZE:
1335 return btrfs_ioctl_resize(root, argp);
1336 case BTRFS_IOC_ADD_DEV:
1337 return btrfs_ioctl_add_dev(root, argp);
1338 case BTRFS_IOC_RM_DEV:
1339 return btrfs_ioctl_rm_dev(root, argp);
1340 case BTRFS_IOC_BALANCE:
1341 return btrfs_balance(root->fs_info->dev_root);
1342 case BTRFS_IOC_CLONE:
1343 return btrfs_ioctl_clone(file, arg, 0, 0, 0);
1344 case BTRFS_IOC_CLONE_RANGE:
1345 return btrfs_ioctl_clone_range(file, argp);
1346 case BTRFS_IOC_TRANS_START:
1347 return btrfs_ioctl_trans_start(file);
1348 case BTRFS_IOC_TRANS_END:
1349 return btrfs_ioctl_trans_end(file);
1350 case BTRFS_IOC_SYNC:
1351 btrfs_sync_fs(file->f_dentry->d_sb, 1);