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>
42 #include <linux/slab.h>
43 #include <linux/blkdev.h>
47 #include "transaction.h"
48 #include "btrfs_inode.h"
50 #include "print-tree.h"
53 #include "inode-map.h"
56 /* Mask out flags that are inappropriate for the given type of inode. */
57 static inline __u32 btrfs_mask_flags(umode_t mode, __u32 flags)
61 else if (S_ISREG(mode))
62 return flags & ~FS_DIRSYNC_FL;
64 return flags & (FS_NODUMP_FL | FS_NOATIME_FL);
68 * Export inode flags to the format expected by the FS_IOC_GETFLAGS ioctl.
70 static unsigned int btrfs_flags_to_ioctl(unsigned int flags)
72 unsigned int iflags = 0;
74 if (flags & BTRFS_INODE_SYNC)
76 if (flags & BTRFS_INODE_IMMUTABLE)
77 iflags |= FS_IMMUTABLE_FL;
78 if (flags & BTRFS_INODE_APPEND)
79 iflags |= FS_APPEND_FL;
80 if (flags & BTRFS_INODE_NODUMP)
81 iflags |= FS_NODUMP_FL;
82 if (flags & BTRFS_INODE_NOATIME)
83 iflags |= FS_NOATIME_FL;
84 if (flags & BTRFS_INODE_DIRSYNC)
85 iflags |= FS_DIRSYNC_FL;
86 if (flags & BTRFS_INODE_NODATACOW)
87 iflags |= FS_NOCOW_FL;
89 if ((flags & BTRFS_INODE_COMPRESS) && !(flags & BTRFS_INODE_NOCOMPRESS))
90 iflags |= FS_COMPR_FL;
91 else if (flags & BTRFS_INODE_NOCOMPRESS)
92 iflags |= FS_NOCOMP_FL;
98 * Update inode->i_flags based on the btrfs internal flags.
100 void btrfs_update_iflags(struct inode *inode)
102 struct btrfs_inode *ip = BTRFS_I(inode);
104 inode->i_flags &= ~(S_SYNC|S_APPEND|S_IMMUTABLE|S_NOATIME|S_DIRSYNC);
106 if (ip->flags & BTRFS_INODE_SYNC)
107 inode->i_flags |= S_SYNC;
108 if (ip->flags & BTRFS_INODE_IMMUTABLE)
109 inode->i_flags |= S_IMMUTABLE;
110 if (ip->flags & BTRFS_INODE_APPEND)
111 inode->i_flags |= S_APPEND;
112 if (ip->flags & BTRFS_INODE_NOATIME)
113 inode->i_flags |= S_NOATIME;
114 if (ip->flags & BTRFS_INODE_DIRSYNC)
115 inode->i_flags |= S_DIRSYNC;
119 * Inherit flags from the parent inode.
121 * Currently only the compression flags and the cow flags are inherited.
123 void btrfs_inherit_iflags(struct inode *inode, struct inode *dir)
130 flags = BTRFS_I(dir)->flags;
132 if (flags & BTRFS_INODE_NOCOMPRESS) {
133 BTRFS_I(inode)->flags &= ~BTRFS_INODE_COMPRESS;
134 BTRFS_I(inode)->flags |= BTRFS_INODE_NOCOMPRESS;
135 } else if (flags & BTRFS_INODE_COMPRESS) {
136 BTRFS_I(inode)->flags &= ~BTRFS_INODE_NOCOMPRESS;
137 BTRFS_I(inode)->flags |= BTRFS_INODE_COMPRESS;
140 if (flags & BTRFS_INODE_NODATACOW)
141 BTRFS_I(inode)->flags |= BTRFS_INODE_NODATACOW;
143 btrfs_update_iflags(inode);
146 static int btrfs_ioctl_getflags(struct file *file, void __user *arg)
148 struct btrfs_inode *ip = BTRFS_I(file->f_path.dentry->d_inode);
149 unsigned int flags = btrfs_flags_to_ioctl(ip->flags);
151 if (copy_to_user(arg, &flags, sizeof(flags)))
156 static int check_flags(unsigned int flags)
158 if (flags & ~(FS_IMMUTABLE_FL | FS_APPEND_FL | \
159 FS_NOATIME_FL | FS_NODUMP_FL | \
160 FS_SYNC_FL | FS_DIRSYNC_FL | \
161 FS_NOCOMP_FL | FS_COMPR_FL |
165 if ((flags & FS_NOCOMP_FL) && (flags & FS_COMPR_FL))
171 static int btrfs_ioctl_setflags(struct file *file, void __user *arg)
173 struct inode *inode = file->f_path.dentry->d_inode;
174 struct btrfs_inode *ip = BTRFS_I(inode);
175 struct btrfs_root *root = ip->root;
176 struct btrfs_trans_handle *trans;
177 unsigned int flags, oldflags;
180 unsigned int i_oldflags;
182 if (btrfs_root_readonly(root))
185 if (copy_from_user(&flags, arg, sizeof(flags)))
188 ret = check_flags(flags);
192 if (!inode_owner_or_capable(inode))
195 mutex_lock(&inode->i_mutex);
197 ip_oldflags = ip->flags;
198 i_oldflags = inode->i_flags;
200 flags = btrfs_mask_flags(inode->i_mode, flags);
201 oldflags = btrfs_flags_to_ioctl(ip->flags);
202 if ((flags ^ oldflags) & (FS_APPEND_FL | FS_IMMUTABLE_FL)) {
203 if (!capable(CAP_LINUX_IMMUTABLE)) {
209 ret = mnt_want_write_file(file);
213 if (flags & FS_SYNC_FL)
214 ip->flags |= BTRFS_INODE_SYNC;
216 ip->flags &= ~BTRFS_INODE_SYNC;
217 if (flags & FS_IMMUTABLE_FL)
218 ip->flags |= BTRFS_INODE_IMMUTABLE;
220 ip->flags &= ~BTRFS_INODE_IMMUTABLE;
221 if (flags & FS_APPEND_FL)
222 ip->flags |= BTRFS_INODE_APPEND;
224 ip->flags &= ~BTRFS_INODE_APPEND;
225 if (flags & FS_NODUMP_FL)
226 ip->flags |= BTRFS_INODE_NODUMP;
228 ip->flags &= ~BTRFS_INODE_NODUMP;
229 if (flags & FS_NOATIME_FL)
230 ip->flags |= BTRFS_INODE_NOATIME;
232 ip->flags &= ~BTRFS_INODE_NOATIME;
233 if (flags & FS_DIRSYNC_FL)
234 ip->flags |= BTRFS_INODE_DIRSYNC;
236 ip->flags &= ~BTRFS_INODE_DIRSYNC;
237 if (flags & FS_NOCOW_FL)
238 ip->flags |= BTRFS_INODE_NODATACOW;
240 ip->flags &= ~BTRFS_INODE_NODATACOW;
243 * The COMPRESS flag can only be changed by users, while the NOCOMPRESS
244 * flag may be changed automatically if compression code won't make
247 if (flags & FS_NOCOMP_FL) {
248 ip->flags &= ~BTRFS_INODE_COMPRESS;
249 ip->flags |= BTRFS_INODE_NOCOMPRESS;
250 } else if (flags & FS_COMPR_FL) {
251 ip->flags |= BTRFS_INODE_COMPRESS;
252 ip->flags &= ~BTRFS_INODE_NOCOMPRESS;
254 ip->flags &= ~(BTRFS_INODE_COMPRESS | BTRFS_INODE_NOCOMPRESS);
257 trans = btrfs_start_transaction(root, 1);
259 ret = PTR_ERR(trans);
263 btrfs_update_iflags(inode);
264 inode->i_ctime = CURRENT_TIME;
265 ret = btrfs_update_inode(trans, root, inode);
267 btrfs_end_transaction(trans, root);
270 ip->flags = ip_oldflags;
271 inode->i_flags = i_oldflags;
274 mnt_drop_write_file(file);
276 mutex_unlock(&inode->i_mutex);
280 static int btrfs_ioctl_getversion(struct file *file, int __user *arg)
282 struct inode *inode = file->f_path.dentry->d_inode;
284 return put_user(inode->i_generation, arg);
287 static noinline int btrfs_ioctl_fitrim(struct file *file, void __user *arg)
289 struct btrfs_fs_info *fs_info = btrfs_sb(fdentry(file)->d_sb);
290 struct btrfs_device *device;
291 struct request_queue *q;
292 struct fstrim_range range;
293 u64 minlen = ULLONG_MAX;
295 u64 total_bytes = btrfs_super_total_bytes(fs_info->super_copy);
298 if (!capable(CAP_SYS_ADMIN))
302 list_for_each_entry_rcu(device, &fs_info->fs_devices->devices,
306 q = bdev_get_queue(device->bdev);
307 if (blk_queue_discard(q)) {
309 minlen = min((u64)q->limits.discard_granularity,
317 if (copy_from_user(&range, arg, sizeof(range)))
319 if (range.start > total_bytes)
322 range.len = min(range.len, total_bytes - range.start);
323 range.minlen = max(range.minlen, minlen);
324 ret = btrfs_trim_fs(fs_info->tree_root, &range);
328 if (copy_to_user(arg, &range, sizeof(range)))
334 static noinline int create_subvol(struct btrfs_root *root,
335 struct dentry *dentry,
336 char *name, int namelen,
339 struct btrfs_trans_handle *trans;
340 struct btrfs_key key;
341 struct btrfs_root_item root_item;
342 struct btrfs_inode_item *inode_item;
343 struct extent_buffer *leaf;
344 struct btrfs_root *new_root;
345 struct dentry *parent = dentry->d_parent;
350 u64 new_dirid = BTRFS_FIRST_FREE_OBJECTID;
353 ret = btrfs_find_free_objectid(root->fs_info->tree_root, &objectid);
357 dir = parent->d_inode;
365 trans = btrfs_start_transaction(root, 6);
367 return PTR_ERR(trans);
369 leaf = btrfs_alloc_free_block(trans, root, root->leafsize,
370 0, objectid, NULL, 0, 0, 0, 0);
376 memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header));
377 btrfs_set_header_bytenr(leaf, leaf->start);
378 btrfs_set_header_generation(leaf, trans->transid);
379 btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV);
380 btrfs_set_header_owner(leaf, objectid);
382 write_extent_buffer(leaf, root->fs_info->fsid,
383 (unsigned long)btrfs_header_fsid(leaf),
385 write_extent_buffer(leaf, root->fs_info->chunk_tree_uuid,
386 (unsigned long)btrfs_header_chunk_tree_uuid(leaf),
388 btrfs_mark_buffer_dirty(leaf);
390 inode_item = &root_item.inode;
391 memset(inode_item, 0, sizeof(*inode_item));
392 inode_item->generation = cpu_to_le64(1);
393 inode_item->size = cpu_to_le64(3);
394 inode_item->nlink = cpu_to_le32(1);
395 inode_item->nbytes = cpu_to_le64(root->leafsize);
396 inode_item->mode = cpu_to_le32(S_IFDIR | 0755);
399 root_item.byte_limit = 0;
400 inode_item->flags = cpu_to_le64(BTRFS_INODE_ROOT_ITEM_INIT);
402 btrfs_set_root_bytenr(&root_item, leaf->start);
403 btrfs_set_root_generation(&root_item, trans->transid);
404 btrfs_set_root_level(&root_item, 0);
405 btrfs_set_root_refs(&root_item, 1);
406 btrfs_set_root_used(&root_item, leaf->len);
407 btrfs_set_root_last_snapshot(&root_item, 0);
409 memset(&root_item.drop_progress, 0, sizeof(root_item.drop_progress));
410 root_item.drop_level = 0;
412 btrfs_tree_unlock(leaf);
413 free_extent_buffer(leaf);
416 btrfs_set_root_dirid(&root_item, new_dirid);
418 key.objectid = objectid;
420 btrfs_set_key_type(&key, BTRFS_ROOT_ITEM_KEY);
421 ret = btrfs_insert_root(trans, root->fs_info->tree_root, &key,
426 key.offset = (u64)-1;
427 new_root = btrfs_read_fs_root_no_name(root->fs_info, &key);
428 BUG_ON(IS_ERR(new_root));
430 btrfs_record_root_in_trans(trans, new_root);
432 ret = btrfs_create_subvol_root(trans, new_root, new_dirid);
434 * insert the directory item
436 ret = btrfs_set_inode_index(dir, &index);
439 ret = btrfs_insert_dir_item(trans, root,
440 name, namelen, dir, &key,
441 BTRFS_FT_DIR, index);
445 btrfs_i_size_write(dir, dir->i_size + namelen * 2);
446 ret = btrfs_update_inode(trans, root, dir);
449 ret = btrfs_add_root_ref(trans, root->fs_info->tree_root,
450 objectid, root->root_key.objectid,
451 btrfs_ino(dir), index, name, namelen);
455 d_instantiate(dentry, btrfs_lookup_dentry(dir, dentry));
458 *async_transid = trans->transid;
459 err = btrfs_commit_transaction_async(trans, root, 1);
461 err = btrfs_commit_transaction(trans, root);
468 static int create_snapshot(struct btrfs_root *root, struct dentry *dentry,
469 char *name, int namelen, u64 *async_transid,
473 struct btrfs_pending_snapshot *pending_snapshot;
474 struct btrfs_trans_handle *trans;
480 pending_snapshot = kzalloc(sizeof(*pending_snapshot), GFP_NOFS);
481 if (!pending_snapshot)
484 btrfs_init_block_rsv(&pending_snapshot->block_rsv);
485 pending_snapshot->dentry = dentry;
486 pending_snapshot->root = root;
487 pending_snapshot->readonly = readonly;
489 trans = btrfs_start_transaction(root->fs_info->extent_root, 5);
491 ret = PTR_ERR(trans);
495 ret = btrfs_snap_reserve_metadata(trans, pending_snapshot);
498 spin_lock(&root->fs_info->trans_lock);
499 list_add(&pending_snapshot->list,
500 &trans->transaction->pending_snapshots);
501 spin_unlock(&root->fs_info->trans_lock);
503 *async_transid = trans->transid;
504 ret = btrfs_commit_transaction_async(trans,
505 root->fs_info->extent_root, 1);
507 ret = btrfs_commit_transaction(trans,
508 root->fs_info->extent_root);
512 ret = pending_snapshot->error;
516 ret = btrfs_orphan_cleanup(pending_snapshot->snap);
520 inode = btrfs_lookup_dentry(dentry->d_parent->d_inode, dentry);
522 ret = PTR_ERR(inode);
526 d_instantiate(dentry, inode);
529 kfree(pending_snapshot);
533 /* copy of check_sticky in fs/namei.c()
534 * It's inline, so penalty for filesystems that don't use sticky bit is
537 static inline int btrfs_check_sticky(struct inode *dir, struct inode *inode)
539 uid_t fsuid = current_fsuid();
541 if (!(dir->i_mode & S_ISVTX))
543 if (inode->i_uid == fsuid)
545 if (dir->i_uid == fsuid)
547 return !capable(CAP_FOWNER);
550 /* copy of may_delete in fs/namei.c()
551 * Check whether we can remove a link victim from directory dir, check
552 * whether the type of victim is right.
553 * 1. We can't do it if dir is read-only (done in permission())
554 * 2. We should have write and exec permissions on dir
555 * 3. We can't remove anything from append-only dir
556 * 4. We can't do anything with immutable dir (done in permission())
557 * 5. If the sticky bit on dir is set we should either
558 * a. be owner of dir, or
559 * b. be owner of victim, or
560 * c. have CAP_FOWNER capability
561 * 6. If the victim is append-only or immutable we can't do antyhing with
562 * links pointing to it.
563 * 7. If we were asked to remove a directory and victim isn't one - ENOTDIR.
564 * 8. If we were asked to remove a non-directory and victim isn't one - EISDIR.
565 * 9. We can't remove a root or mountpoint.
566 * 10. We don't allow removal of NFS sillyrenamed files; it's handled by
567 * nfs_async_unlink().
570 static int btrfs_may_delete(struct inode *dir,struct dentry *victim,int isdir)
574 if (!victim->d_inode)
577 BUG_ON(victim->d_parent->d_inode != dir);
578 audit_inode_child(victim, dir);
580 error = inode_permission(dir, MAY_WRITE | MAY_EXEC);
585 if (btrfs_check_sticky(dir, victim->d_inode)||
586 IS_APPEND(victim->d_inode)||
587 IS_IMMUTABLE(victim->d_inode) || IS_SWAPFILE(victim->d_inode))
590 if (!S_ISDIR(victim->d_inode->i_mode))
594 } else if (S_ISDIR(victim->d_inode->i_mode))
598 if (victim->d_flags & DCACHE_NFSFS_RENAMED)
603 /* copy of may_create in fs/namei.c() */
604 static inline int btrfs_may_create(struct inode *dir, struct dentry *child)
610 return inode_permission(dir, MAY_WRITE | MAY_EXEC);
614 * Create a new subvolume below @parent. This is largely modeled after
615 * sys_mkdirat and vfs_mkdir, but we only do a single component lookup
616 * inside this filesystem so it's quite a bit simpler.
618 static noinline int btrfs_mksubvol(struct path *parent,
619 char *name, int namelen,
620 struct btrfs_root *snap_src,
621 u64 *async_transid, bool readonly)
623 struct inode *dir = parent->dentry->d_inode;
624 struct dentry *dentry;
627 mutex_lock_nested(&dir->i_mutex, I_MUTEX_PARENT);
629 dentry = lookup_one_len(name, parent->dentry, namelen);
630 error = PTR_ERR(dentry);
638 error = mnt_want_write(parent->mnt);
642 error = btrfs_may_create(dir, dentry);
646 down_read(&BTRFS_I(dir)->root->fs_info->subvol_sem);
648 if (btrfs_root_refs(&BTRFS_I(dir)->root->root_item) == 0)
652 error = create_snapshot(snap_src, dentry,
653 name, namelen, async_transid, readonly);
655 error = create_subvol(BTRFS_I(dir)->root, dentry,
656 name, namelen, async_transid);
659 fsnotify_mkdir(dir, dentry);
661 up_read(&BTRFS_I(dir)->root->fs_info->subvol_sem);
663 mnt_drop_write(parent->mnt);
667 mutex_unlock(&dir->i_mutex);
672 * When we're defragging a range, we don't want to kick it off again
673 * if it is really just waiting for delalloc to send it down.
674 * If we find a nice big extent or delalloc range for the bytes in the
675 * file you want to defrag, we return 0 to let you know to skip this
678 static int check_defrag_in_cache(struct inode *inode, u64 offset, int thresh)
680 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
681 struct extent_map *em = NULL;
682 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
685 read_lock(&em_tree->lock);
686 em = lookup_extent_mapping(em_tree, offset, PAGE_CACHE_SIZE);
687 read_unlock(&em_tree->lock);
690 end = extent_map_end(em);
692 if (end - offset > thresh)
695 /* if we already have a nice delalloc here, just stop */
697 end = count_range_bits(io_tree, &offset, offset + thresh,
698 thresh, EXTENT_DELALLOC, 1);
705 * helper function to walk through a file and find extents
706 * newer than a specific transid, and smaller than thresh.
708 * This is used by the defragging code to find new and small
711 static int find_new_extents(struct btrfs_root *root,
712 struct inode *inode, u64 newer_than,
713 u64 *off, int thresh)
715 struct btrfs_path *path;
716 struct btrfs_key min_key;
717 struct btrfs_key max_key;
718 struct extent_buffer *leaf;
719 struct btrfs_file_extent_item *extent;
722 u64 ino = btrfs_ino(inode);
724 path = btrfs_alloc_path();
728 min_key.objectid = ino;
729 min_key.type = BTRFS_EXTENT_DATA_KEY;
730 min_key.offset = *off;
732 max_key.objectid = ino;
733 max_key.type = (u8)-1;
734 max_key.offset = (u64)-1;
736 path->keep_locks = 1;
739 ret = btrfs_search_forward(root, &min_key, &max_key,
740 path, 0, newer_than);
743 if (min_key.objectid != ino)
745 if (min_key.type != BTRFS_EXTENT_DATA_KEY)
748 leaf = path->nodes[0];
749 extent = btrfs_item_ptr(leaf, path->slots[0],
750 struct btrfs_file_extent_item);
752 type = btrfs_file_extent_type(leaf, extent);
753 if (type == BTRFS_FILE_EXTENT_REG &&
754 btrfs_file_extent_num_bytes(leaf, extent) < thresh &&
755 check_defrag_in_cache(inode, min_key.offset, thresh)) {
756 *off = min_key.offset;
757 btrfs_free_path(path);
761 if (min_key.offset == (u64)-1)
765 btrfs_release_path(path);
768 btrfs_free_path(path);
772 static int should_defrag_range(struct inode *inode, u64 start, u64 len,
773 int thresh, u64 *last_len, u64 *skip,
776 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
777 struct extent_map *em = NULL;
778 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
782 * make sure that once we start defragging an extent, we keep on
785 if (start < *defrag_end)
791 * hopefully we have this extent in the tree already, try without
792 * the full extent lock
794 read_lock(&em_tree->lock);
795 em = lookup_extent_mapping(em_tree, start, len);
796 read_unlock(&em_tree->lock);
799 /* get the big lock and read metadata off disk */
800 lock_extent(io_tree, start, start + len - 1, GFP_NOFS);
801 em = btrfs_get_extent(inode, NULL, 0, start, len, 0);
802 unlock_extent(io_tree, start, start + len - 1, GFP_NOFS);
808 /* this will cover holes, and inline extents */
809 if (em->block_start >= EXTENT_MAP_LAST_BYTE)
813 * we hit a real extent, if it is big don't bother defragging it again
815 if ((*last_len == 0 || *last_len >= thresh) && em->len >= thresh)
819 * last_len ends up being a counter of how many bytes we've defragged.
820 * every time we choose not to defrag an extent, we reset *last_len
821 * so that the next tiny extent will force a defrag.
823 * The end result of this is that tiny extents before a single big
824 * extent will force at least part of that big extent to be defragged.
827 *defrag_end = extent_map_end(em);
830 *skip = extent_map_end(em);
839 * it doesn't do much good to defrag one or two pages
840 * at a time. This pulls in a nice chunk of pages
843 * It also makes sure the delalloc code has enough
844 * dirty data to avoid making new small extents as part
847 * It's a good idea to start RA on this range
848 * before calling this.
850 static int cluster_pages_for_defrag(struct inode *inode,
852 unsigned long start_index,
855 unsigned long file_end;
856 u64 isize = i_size_read(inode);
862 struct btrfs_ordered_extent *ordered;
863 struct extent_state *cached_state = NULL;
864 gfp_t mask = btrfs_alloc_write_mask(inode->i_mapping);
868 file_end = (isize - 1) >> PAGE_CACHE_SHIFT;
870 ret = btrfs_delalloc_reserve_space(inode,
871 num_pages << PAGE_CACHE_SHIFT);
878 /* step one, lock all the pages */
879 for (i = 0; i < num_pages; i++) {
881 page = find_or_create_page(inode->i_mapping,
882 start_index + i, mask);
886 if (!PageUptodate(page)) {
887 btrfs_readpage(NULL, page);
889 if (!PageUptodate(page)) {
891 page_cache_release(page);
896 isize = i_size_read(inode);
897 file_end = (isize - 1) >> PAGE_CACHE_SHIFT;
898 if (!isize || page->index > file_end ||
899 page->mapping != inode->i_mapping) {
900 /* whoops, we blew past eof, skip this page */
902 page_cache_release(page);
911 if (!(inode->i_sb->s_flags & MS_ACTIVE))
915 * so now we have a nice long stream of locked
916 * and up to date pages, lets wait on them
918 for (i = 0; i < i_done; i++)
919 wait_on_page_writeback(pages[i]);
921 page_start = page_offset(pages[0]);
922 page_end = page_offset(pages[i_done - 1]) + PAGE_CACHE_SIZE;
924 lock_extent_bits(&BTRFS_I(inode)->io_tree,
925 page_start, page_end - 1, 0, &cached_state,
927 ordered = btrfs_lookup_first_ordered_extent(inode, page_end - 1);
929 ordered->file_offset + ordered->len > page_start &&
930 ordered->file_offset < page_end) {
931 btrfs_put_ordered_extent(ordered);
932 unlock_extent_cached(&BTRFS_I(inode)->io_tree,
933 page_start, page_end - 1,
934 &cached_state, GFP_NOFS);
935 for (i = 0; i < i_done; i++) {
936 unlock_page(pages[i]);
937 page_cache_release(pages[i]);
939 btrfs_wait_ordered_range(inode, page_start,
940 page_end - page_start);
944 btrfs_put_ordered_extent(ordered);
946 clear_extent_bit(&BTRFS_I(inode)->io_tree, page_start,
947 page_end - 1, EXTENT_DIRTY | EXTENT_DELALLOC |
948 EXTENT_DO_ACCOUNTING, 0, 0, &cached_state,
951 if (i_done != num_pages) {
952 spin_lock(&BTRFS_I(inode)->lock);
953 BTRFS_I(inode)->outstanding_extents++;
954 spin_unlock(&BTRFS_I(inode)->lock);
955 btrfs_delalloc_release_space(inode,
956 (num_pages - i_done) << PAGE_CACHE_SHIFT);
960 btrfs_set_extent_delalloc(inode, page_start, page_end - 1,
963 unlock_extent_cached(&BTRFS_I(inode)->io_tree,
964 page_start, page_end - 1, &cached_state,
967 for (i = 0; i < i_done; i++) {
968 clear_page_dirty_for_io(pages[i]);
969 ClearPageChecked(pages[i]);
970 set_page_extent_mapped(pages[i]);
971 set_page_dirty(pages[i]);
972 unlock_page(pages[i]);
973 page_cache_release(pages[i]);
977 for (i = 0; i < i_done; i++) {
978 unlock_page(pages[i]);
979 page_cache_release(pages[i]);
981 btrfs_delalloc_release_space(inode, num_pages << PAGE_CACHE_SHIFT);
986 int btrfs_defrag_file(struct inode *inode, struct file *file,
987 struct btrfs_ioctl_defrag_range_args *range,
988 u64 newer_than, unsigned long max_to_defrag)
990 struct btrfs_root *root = BTRFS_I(inode)->root;
991 struct btrfs_super_block *disk_super;
992 struct file_ra_state *ra = NULL;
993 unsigned long last_index;
994 u64 isize = i_size_read(inode);
999 u64 newer_off = range->start;
1001 unsigned long ra_index = 0;
1003 int defrag_count = 0;
1004 int compress_type = BTRFS_COMPRESS_ZLIB;
1005 int extent_thresh = range->extent_thresh;
1006 int max_cluster = (256 * 1024) >> PAGE_CACHE_SHIFT;
1007 int cluster = max_cluster;
1008 u64 new_align = ~((u64)128 * 1024 - 1);
1009 struct page **pages = NULL;
1011 if (extent_thresh == 0)
1012 extent_thresh = 256 * 1024;
1014 if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS) {
1015 if (range->compress_type > BTRFS_COMPRESS_TYPES)
1017 if (range->compress_type)
1018 compress_type = range->compress_type;
1025 * if we were not given a file, allocate a readahead
1029 ra = kzalloc(sizeof(*ra), GFP_NOFS);
1032 file_ra_state_init(ra, inode->i_mapping);
1037 pages = kmalloc(sizeof(struct page *) * max_cluster,
1044 /* find the last page to defrag */
1045 if (range->start + range->len > range->start) {
1046 last_index = min_t(u64, isize - 1,
1047 range->start + range->len - 1) >> PAGE_CACHE_SHIFT;
1049 last_index = (isize - 1) >> PAGE_CACHE_SHIFT;
1053 ret = find_new_extents(root, inode, newer_than,
1054 &newer_off, 64 * 1024);
1056 range->start = newer_off;
1058 * we always align our defrag to help keep
1059 * the extents in the file evenly spaced
1061 i = (newer_off & new_align) >> PAGE_CACHE_SHIFT;
1065 i = range->start >> PAGE_CACHE_SHIFT;
1068 max_to_defrag = last_index + 1;
1071 * make writeback starts from i, so the defrag range can be
1072 * written sequentially.
1074 if (i < inode->i_mapping->writeback_index)
1075 inode->i_mapping->writeback_index = i;
1077 while (i <= last_index && defrag_count < max_to_defrag &&
1078 (i < (i_size_read(inode) + PAGE_CACHE_SIZE - 1) >>
1079 PAGE_CACHE_SHIFT)) {
1081 * make sure we stop running if someone unmounts
1084 if (!(inode->i_sb->s_flags & MS_ACTIVE))
1088 !should_defrag_range(inode, (u64)i << PAGE_CACHE_SHIFT,
1095 * the should_defrag function tells us how much to skip
1096 * bump our counter by the suggested amount
1098 next = (skip + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT;
1099 i = max(i + 1, next);
1104 cluster = (PAGE_CACHE_ALIGN(defrag_end) >>
1105 PAGE_CACHE_SHIFT) - i;
1106 cluster = min(cluster, max_cluster);
1108 cluster = max_cluster;
1111 if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)
1112 BTRFS_I(inode)->force_compress = compress_type;
1114 if (i + cluster > ra_index) {
1115 ra_index = max(i, ra_index);
1116 btrfs_force_ra(inode->i_mapping, ra, file, ra_index,
1118 ra_index += max_cluster;
1121 ret = cluster_pages_for_defrag(inode, pages, i, cluster);
1125 defrag_count += ret;
1126 balance_dirty_pages_ratelimited_nr(inode->i_mapping, ret);
1129 if (newer_off == (u64)-1)
1132 newer_off = max(newer_off + 1,
1133 (u64)i << PAGE_CACHE_SHIFT);
1135 ret = find_new_extents(root, inode,
1136 newer_than, &newer_off,
1139 range->start = newer_off;
1140 i = (newer_off & new_align) >> PAGE_CACHE_SHIFT;
1147 last_len += ret << PAGE_CACHE_SHIFT;
1155 if ((range->flags & BTRFS_DEFRAG_RANGE_START_IO))
1156 filemap_flush(inode->i_mapping);
1158 if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
1159 /* the filemap_flush will queue IO into the worker threads, but
1160 * we have to make sure the IO is actually started and that
1161 * ordered extents get created before we return
1163 atomic_inc(&root->fs_info->async_submit_draining);
1164 while (atomic_read(&root->fs_info->nr_async_submits) ||
1165 atomic_read(&root->fs_info->async_delalloc_pages)) {
1166 wait_event(root->fs_info->async_submit_wait,
1167 (atomic_read(&root->fs_info->nr_async_submits) == 0 &&
1168 atomic_read(&root->fs_info->async_delalloc_pages) == 0));
1170 atomic_dec(&root->fs_info->async_submit_draining);
1172 mutex_lock(&inode->i_mutex);
1173 BTRFS_I(inode)->force_compress = BTRFS_COMPRESS_NONE;
1174 mutex_unlock(&inode->i_mutex);
1177 disk_super = root->fs_info->super_copy;
1178 features = btrfs_super_incompat_flags(disk_super);
1179 if (range->compress_type == BTRFS_COMPRESS_LZO) {
1180 features |= BTRFS_FEATURE_INCOMPAT_COMPRESS_LZO;
1181 btrfs_set_super_incompat_flags(disk_super, features);
1193 static noinline int btrfs_ioctl_resize(struct btrfs_root *root,
1199 struct btrfs_ioctl_vol_args *vol_args;
1200 struct btrfs_trans_handle *trans;
1201 struct btrfs_device *device = NULL;
1203 char *devstr = NULL;
1207 if (root->fs_info->sb->s_flags & MS_RDONLY)
1210 if (!capable(CAP_SYS_ADMIN))
1213 mutex_lock(&root->fs_info->volume_mutex);
1214 if (root->fs_info->balance_ctl) {
1215 printk(KERN_INFO "btrfs: balance in progress\n");
1220 vol_args = memdup_user(arg, sizeof(*vol_args));
1221 if (IS_ERR(vol_args)) {
1222 ret = PTR_ERR(vol_args);
1226 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1228 sizestr = vol_args->name;
1229 devstr = strchr(sizestr, ':');
1232 sizestr = devstr + 1;
1234 devstr = vol_args->name;
1235 devid = simple_strtoull(devstr, &end, 10);
1236 printk(KERN_INFO "btrfs: resizing devid %llu\n",
1237 (unsigned long long)devid);
1239 device = btrfs_find_device(root, devid, NULL, NULL);
1241 printk(KERN_INFO "btrfs: resizer unable to find device %llu\n",
1242 (unsigned long long)devid);
1246 if (!strcmp(sizestr, "max"))
1247 new_size = device->bdev->bd_inode->i_size;
1249 if (sizestr[0] == '-') {
1252 } else if (sizestr[0] == '+') {
1256 new_size = memparse(sizestr, NULL);
1257 if (new_size == 0) {
1263 old_size = device->total_bytes;
1266 if (new_size > old_size) {
1270 new_size = old_size - new_size;
1271 } else if (mod > 0) {
1272 new_size = old_size + new_size;
1275 if (new_size < 256 * 1024 * 1024) {
1279 if (new_size > device->bdev->bd_inode->i_size) {
1284 do_div(new_size, root->sectorsize);
1285 new_size *= root->sectorsize;
1287 printk(KERN_INFO "btrfs: new size for %s is %llu\n",
1288 device->name, (unsigned long long)new_size);
1290 if (new_size > old_size) {
1291 trans = btrfs_start_transaction(root, 0);
1292 if (IS_ERR(trans)) {
1293 ret = PTR_ERR(trans);
1296 ret = btrfs_grow_device(trans, device, new_size);
1297 btrfs_commit_transaction(trans, root);
1298 } else if (new_size < old_size) {
1299 ret = btrfs_shrink_device(device, new_size);
1305 mutex_unlock(&root->fs_info->volume_mutex);
1309 static noinline int btrfs_ioctl_snap_create_transid(struct file *file,
1316 struct btrfs_root *root = BTRFS_I(fdentry(file)->d_inode)->root;
1317 struct file *src_file;
1321 if (root->fs_info->sb->s_flags & MS_RDONLY)
1324 namelen = strlen(name);
1325 if (strchr(name, '/')) {
1331 ret = btrfs_mksubvol(&file->f_path, name, namelen,
1332 NULL, transid, readonly);
1334 struct inode *src_inode;
1335 src_file = fget(fd);
1341 src_inode = src_file->f_path.dentry->d_inode;
1342 if (src_inode->i_sb != file->f_path.dentry->d_inode->i_sb) {
1343 printk(KERN_INFO "btrfs: Snapshot src from "
1349 ret = btrfs_mksubvol(&file->f_path, name, namelen,
1350 BTRFS_I(src_inode)->root,
1358 static noinline int btrfs_ioctl_snap_create(struct file *file,
1359 void __user *arg, int subvol)
1361 struct btrfs_ioctl_vol_args *vol_args;
1364 vol_args = memdup_user(arg, sizeof(*vol_args));
1365 if (IS_ERR(vol_args))
1366 return PTR_ERR(vol_args);
1367 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1369 ret = btrfs_ioctl_snap_create_transid(file, vol_args->name,
1370 vol_args->fd, subvol,
1377 static noinline int btrfs_ioctl_snap_create_v2(struct file *file,
1378 void __user *arg, int subvol)
1380 struct btrfs_ioctl_vol_args_v2 *vol_args;
1384 bool readonly = false;
1386 vol_args = memdup_user(arg, sizeof(*vol_args));
1387 if (IS_ERR(vol_args))
1388 return PTR_ERR(vol_args);
1389 vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0';
1391 if (vol_args->flags &
1392 ~(BTRFS_SUBVOL_CREATE_ASYNC | BTRFS_SUBVOL_RDONLY)) {
1397 if (vol_args->flags & BTRFS_SUBVOL_CREATE_ASYNC)
1399 if (vol_args->flags & BTRFS_SUBVOL_RDONLY)
1402 ret = btrfs_ioctl_snap_create_transid(file, vol_args->name,
1403 vol_args->fd, subvol,
1406 if (ret == 0 && ptr &&
1408 offsetof(struct btrfs_ioctl_vol_args_v2,
1409 transid), ptr, sizeof(*ptr)))
1416 static noinline int btrfs_ioctl_subvol_getflags(struct file *file,
1419 struct inode *inode = fdentry(file)->d_inode;
1420 struct btrfs_root *root = BTRFS_I(inode)->root;
1424 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID)
1427 down_read(&root->fs_info->subvol_sem);
1428 if (btrfs_root_readonly(root))
1429 flags |= BTRFS_SUBVOL_RDONLY;
1430 up_read(&root->fs_info->subvol_sem);
1432 if (copy_to_user(arg, &flags, sizeof(flags)))
1438 static noinline int btrfs_ioctl_subvol_setflags(struct file *file,
1441 struct inode *inode = fdentry(file)->d_inode;
1442 struct btrfs_root *root = BTRFS_I(inode)->root;
1443 struct btrfs_trans_handle *trans;
1448 if (root->fs_info->sb->s_flags & MS_RDONLY)
1451 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID)
1454 if (copy_from_user(&flags, arg, sizeof(flags)))
1457 if (flags & BTRFS_SUBVOL_CREATE_ASYNC)
1460 if (flags & ~BTRFS_SUBVOL_RDONLY)
1463 if (!inode_owner_or_capable(inode))
1466 down_write(&root->fs_info->subvol_sem);
1469 if (!!(flags & BTRFS_SUBVOL_RDONLY) == btrfs_root_readonly(root))
1472 root_flags = btrfs_root_flags(&root->root_item);
1473 if (flags & BTRFS_SUBVOL_RDONLY)
1474 btrfs_set_root_flags(&root->root_item,
1475 root_flags | BTRFS_ROOT_SUBVOL_RDONLY);
1477 btrfs_set_root_flags(&root->root_item,
1478 root_flags & ~BTRFS_ROOT_SUBVOL_RDONLY);
1480 trans = btrfs_start_transaction(root, 1);
1481 if (IS_ERR(trans)) {
1482 ret = PTR_ERR(trans);
1486 ret = btrfs_update_root(trans, root->fs_info->tree_root,
1487 &root->root_key, &root->root_item);
1489 btrfs_commit_transaction(trans, root);
1492 btrfs_set_root_flags(&root->root_item, root_flags);
1494 up_write(&root->fs_info->subvol_sem);
1499 * helper to check if the subvolume references other subvolumes
1501 static noinline int may_destroy_subvol(struct btrfs_root *root)
1503 struct btrfs_path *path;
1504 struct btrfs_key key;
1507 path = btrfs_alloc_path();
1511 key.objectid = root->root_key.objectid;
1512 key.type = BTRFS_ROOT_REF_KEY;
1513 key.offset = (u64)-1;
1515 ret = btrfs_search_slot(NULL, root->fs_info->tree_root,
1522 if (path->slots[0] > 0) {
1524 btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
1525 if (key.objectid == root->root_key.objectid &&
1526 key.type == BTRFS_ROOT_REF_KEY)
1530 btrfs_free_path(path);
1534 static noinline int key_in_sk(struct btrfs_key *key,
1535 struct btrfs_ioctl_search_key *sk)
1537 struct btrfs_key test;
1540 test.objectid = sk->min_objectid;
1541 test.type = sk->min_type;
1542 test.offset = sk->min_offset;
1544 ret = btrfs_comp_cpu_keys(key, &test);
1548 test.objectid = sk->max_objectid;
1549 test.type = sk->max_type;
1550 test.offset = sk->max_offset;
1552 ret = btrfs_comp_cpu_keys(key, &test);
1558 static noinline int copy_to_sk(struct btrfs_root *root,
1559 struct btrfs_path *path,
1560 struct btrfs_key *key,
1561 struct btrfs_ioctl_search_key *sk,
1563 unsigned long *sk_offset,
1567 struct extent_buffer *leaf;
1568 struct btrfs_ioctl_search_header sh;
1569 unsigned long item_off;
1570 unsigned long item_len;
1576 leaf = path->nodes[0];
1577 slot = path->slots[0];
1578 nritems = btrfs_header_nritems(leaf);
1580 if (btrfs_header_generation(leaf) > sk->max_transid) {
1584 found_transid = btrfs_header_generation(leaf);
1586 for (i = slot; i < nritems; i++) {
1587 item_off = btrfs_item_ptr_offset(leaf, i);
1588 item_len = btrfs_item_size_nr(leaf, i);
1590 if (item_len > BTRFS_SEARCH_ARGS_BUFSIZE)
1593 if (sizeof(sh) + item_len + *sk_offset >
1594 BTRFS_SEARCH_ARGS_BUFSIZE) {
1599 btrfs_item_key_to_cpu(leaf, key, i);
1600 if (!key_in_sk(key, sk))
1603 sh.objectid = key->objectid;
1604 sh.offset = key->offset;
1605 sh.type = key->type;
1607 sh.transid = found_transid;
1609 /* copy search result header */
1610 memcpy(buf + *sk_offset, &sh, sizeof(sh));
1611 *sk_offset += sizeof(sh);
1614 char *p = buf + *sk_offset;
1616 read_extent_buffer(leaf, p,
1617 item_off, item_len);
1618 *sk_offset += item_len;
1622 if (*num_found >= sk->nr_items)
1627 if (key->offset < (u64)-1 && key->offset < sk->max_offset)
1629 else if (key->type < (u8)-1 && key->type < sk->max_type) {
1632 } else if (key->objectid < (u64)-1 && key->objectid < sk->max_objectid) {
1642 static noinline int search_ioctl(struct inode *inode,
1643 struct btrfs_ioctl_search_args *args)
1645 struct btrfs_root *root;
1646 struct btrfs_key key;
1647 struct btrfs_key max_key;
1648 struct btrfs_path *path;
1649 struct btrfs_ioctl_search_key *sk = &args->key;
1650 struct btrfs_fs_info *info = BTRFS_I(inode)->root->fs_info;
1653 unsigned long sk_offset = 0;
1655 path = btrfs_alloc_path();
1659 if (sk->tree_id == 0) {
1660 /* search the root of the inode that was passed */
1661 root = BTRFS_I(inode)->root;
1663 key.objectid = sk->tree_id;
1664 key.type = BTRFS_ROOT_ITEM_KEY;
1665 key.offset = (u64)-1;
1666 root = btrfs_read_fs_root_no_name(info, &key);
1668 printk(KERN_ERR "could not find root %llu\n",
1670 btrfs_free_path(path);
1675 key.objectid = sk->min_objectid;
1676 key.type = sk->min_type;
1677 key.offset = sk->min_offset;
1679 max_key.objectid = sk->max_objectid;
1680 max_key.type = sk->max_type;
1681 max_key.offset = sk->max_offset;
1683 path->keep_locks = 1;
1686 ret = btrfs_search_forward(root, &key, &max_key, path, 0,
1693 ret = copy_to_sk(root, path, &key, sk, args->buf,
1694 &sk_offset, &num_found);
1695 btrfs_release_path(path);
1696 if (ret || num_found >= sk->nr_items)
1702 sk->nr_items = num_found;
1703 btrfs_free_path(path);
1707 static noinline int btrfs_ioctl_tree_search(struct file *file,
1710 struct btrfs_ioctl_search_args *args;
1711 struct inode *inode;
1714 if (!capable(CAP_SYS_ADMIN))
1717 args = memdup_user(argp, sizeof(*args));
1719 return PTR_ERR(args);
1721 inode = fdentry(file)->d_inode;
1722 ret = search_ioctl(inode, args);
1723 if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
1730 * Search INODE_REFs to identify path name of 'dirid' directory
1731 * in a 'tree_id' tree. and sets path name to 'name'.
1733 static noinline int btrfs_search_path_in_tree(struct btrfs_fs_info *info,
1734 u64 tree_id, u64 dirid, char *name)
1736 struct btrfs_root *root;
1737 struct btrfs_key key;
1743 struct btrfs_inode_ref *iref;
1744 struct extent_buffer *l;
1745 struct btrfs_path *path;
1747 if (dirid == BTRFS_FIRST_FREE_OBJECTID) {
1752 path = btrfs_alloc_path();
1756 ptr = &name[BTRFS_INO_LOOKUP_PATH_MAX];
1758 key.objectid = tree_id;
1759 key.type = BTRFS_ROOT_ITEM_KEY;
1760 key.offset = (u64)-1;
1761 root = btrfs_read_fs_root_no_name(info, &key);
1763 printk(KERN_ERR "could not find root %llu\n", tree_id);
1768 key.objectid = dirid;
1769 key.type = BTRFS_INODE_REF_KEY;
1770 key.offset = (u64)-1;
1773 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
1778 slot = path->slots[0];
1779 if (ret > 0 && slot > 0)
1781 btrfs_item_key_to_cpu(l, &key, slot);
1783 if (ret > 0 && (key.objectid != dirid ||
1784 key.type != BTRFS_INODE_REF_KEY)) {
1789 iref = btrfs_item_ptr(l, slot, struct btrfs_inode_ref);
1790 len = btrfs_inode_ref_name_len(l, iref);
1792 total_len += len + 1;
1797 read_extent_buffer(l, ptr,(unsigned long)(iref + 1), len);
1799 if (key.offset == BTRFS_FIRST_FREE_OBJECTID)
1802 btrfs_release_path(path);
1803 key.objectid = key.offset;
1804 key.offset = (u64)-1;
1805 dirid = key.objectid;
1809 memmove(name, ptr, total_len);
1810 name[total_len]='\0';
1813 btrfs_free_path(path);
1817 static noinline int btrfs_ioctl_ino_lookup(struct file *file,
1820 struct btrfs_ioctl_ino_lookup_args *args;
1821 struct inode *inode;
1824 if (!capable(CAP_SYS_ADMIN))
1827 args = memdup_user(argp, sizeof(*args));
1829 return PTR_ERR(args);
1831 inode = fdentry(file)->d_inode;
1833 if (args->treeid == 0)
1834 args->treeid = BTRFS_I(inode)->root->root_key.objectid;
1836 ret = btrfs_search_path_in_tree(BTRFS_I(inode)->root->fs_info,
1837 args->treeid, args->objectid,
1840 if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
1847 static noinline int btrfs_ioctl_snap_destroy(struct file *file,
1850 struct dentry *parent = fdentry(file);
1851 struct dentry *dentry;
1852 struct inode *dir = parent->d_inode;
1853 struct inode *inode;
1854 struct btrfs_root *root = BTRFS_I(dir)->root;
1855 struct btrfs_root *dest = NULL;
1856 struct btrfs_ioctl_vol_args *vol_args;
1857 struct btrfs_trans_handle *trans;
1862 vol_args = memdup_user(arg, sizeof(*vol_args));
1863 if (IS_ERR(vol_args))
1864 return PTR_ERR(vol_args);
1866 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1867 namelen = strlen(vol_args->name);
1868 if (strchr(vol_args->name, '/') ||
1869 strncmp(vol_args->name, "..", namelen) == 0) {
1874 err = mnt_want_write_file(file);
1878 mutex_lock_nested(&dir->i_mutex, I_MUTEX_PARENT);
1879 dentry = lookup_one_len(vol_args->name, parent, namelen);
1880 if (IS_ERR(dentry)) {
1881 err = PTR_ERR(dentry);
1882 goto out_unlock_dir;
1885 if (!dentry->d_inode) {
1890 inode = dentry->d_inode;
1891 dest = BTRFS_I(inode)->root;
1892 if (!capable(CAP_SYS_ADMIN)){
1894 * Regular user. Only allow this with a special mount
1895 * option, when the user has write+exec access to the
1896 * subvol root, and when rmdir(2) would have been
1899 * Note that this is _not_ check that the subvol is
1900 * empty or doesn't contain data that we wouldn't
1901 * otherwise be able to delete.
1903 * Users who want to delete empty subvols should try
1907 if (!btrfs_test_opt(root, USER_SUBVOL_RM_ALLOWED))
1911 * Do not allow deletion if the parent dir is the same
1912 * as the dir to be deleted. That means the ioctl
1913 * must be called on the dentry referencing the root
1914 * of the subvol, not a random directory contained
1921 err = inode_permission(inode, MAY_WRITE | MAY_EXEC);
1925 /* check if subvolume may be deleted by a non-root user */
1926 err = btrfs_may_delete(dir, dentry, 1);
1931 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) {
1936 mutex_lock(&inode->i_mutex);
1937 err = d_invalidate(dentry);
1941 down_write(&root->fs_info->subvol_sem);
1943 err = may_destroy_subvol(dest);
1947 trans = btrfs_start_transaction(root, 0);
1948 if (IS_ERR(trans)) {
1949 err = PTR_ERR(trans);
1952 trans->block_rsv = &root->fs_info->global_block_rsv;
1954 ret = btrfs_unlink_subvol(trans, root, dir,
1955 dest->root_key.objectid,
1956 dentry->d_name.name,
1957 dentry->d_name.len);
1960 btrfs_record_root_in_trans(trans, dest);
1962 memset(&dest->root_item.drop_progress, 0,
1963 sizeof(dest->root_item.drop_progress));
1964 dest->root_item.drop_level = 0;
1965 btrfs_set_root_refs(&dest->root_item, 0);
1967 if (!xchg(&dest->orphan_item_inserted, 1)) {
1968 ret = btrfs_insert_orphan_item(trans,
1969 root->fs_info->tree_root,
1970 dest->root_key.objectid);
1974 ret = btrfs_end_transaction(trans, root);
1976 inode->i_flags |= S_DEAD;
1978 up_write(&root->fs_info->subvol_sem);
1980 mutex_unlock(&inode->i_mutex);
1982 shrink_dcache_sb(root->fs_info->sb);
1983 btrfs_invalidate_inodes(dest);
1989 mutex_unlock(&dir->i_mutex);
1990 mnt_drop_write_file(file);
1996 static int btrfs_ioctl_defrag(struct file *file, void __user *argp)
1998 struct inode *inode = fdentry(file)->d_inode;
1999 struct btrfs_root *root = BTRFS_I(inode)->root;
2000 struct btrfs_ioctl_defrag_range_args *range;
2003 if (btrfs_root_readonly(root))
2006 ret = mnt_want_write_file(file);
2010 switch (inode->i_mode & S_IFMT) {
2012 if (!capable(CAP_SYS_ADMIN)) {
2016 ret = btrfs_defrag_root(root, 0);
2019 ret = btrfs_defrag_root(root->fs_info->extent_root, 0);
2022 if (!(file->f_mode & FMODE_WRITE)) {
2027 range = kzalloc(sizeof(*range), GFP_KERNEL);
2034 if (copy_from_user(range, argp,
2040 /* compression requires us to start the IO */
2041 if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
2042 range->flags |= BTRFS_DEFRAG_RANGE_START_IO;
2043 range->extent_thresh = (u32)-1;
2046 /* the rest are all set to zero by kzalloc */
2047 range->len = (u64)-1;
2049 ret = btrfs_defrag_file(fdentry(file)->d_inode, file,
2059 mnt_drop_write_file(file);
2063 static long btrfs_ioctl_add_dev(struct btrfs_root *root, void __user *arg)
2065 struct btrfs_ioctl_vol_args *vol_args;
2068 if (!capable(CAP_SYS_ADMIN))
2071 mutex_lock(&root->fs_info->volume_mutex);
2072 if (root->fs_info->balance_ctl) {
2073 printk(KERN_INFO "btrfs: balance in progress\n");
2078 vol_args = memdup_user(arg, sizeof(*vol_args));
2079 if (IS_ERR(vol_args)) {
2080 ret = PTR_ERR(vol_args);
2084 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2085 ret = btrfs_init_new_device(root, vol_args->name);
2089 mutex_unlock(&root->fs_info->volume_mutex);
2093 static long btrfs_ioctl_rm_dev(struct btrfs_root *root, void __user *arg)
2095 struct btrfs_ioctl_vol_args *vol_args;
2098 if (!capable(CAP_SYS_ADMIN))
2101 if (root->fs_info->sb->s_flags & MS_RDONLY)
2104 mutex_lock(&root->fs_info->volume_mutex);
2105 if (root->fs_info->balance_ctl) {
2106 printk(KERN_INFO "btrfs: balance in progress\n");
2111 vol_args = memdup_user(arg, sizeof(*vol_args));
2112 if (IS_ERR(vol_args)) {
2113 ret = PTR_ERR(vol_args);
2117 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2118 ret = btrfs_rm_device(root, vol_args->name);
2122 mutex_unlock(&root->fs_info->volume_mutex);
2126 static long btrfs_ioctl_fs_info(struct btrfs_root *root, void __user *arg)
2128 struct btrfs_ioctl_fs_info_args *fi_args;
2129 struct btrfs_device *device;
2130 struct btrfs_device *next;
2131 struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
2134 if (!capable(CAP_SYS_ADMIN))
2137 fi_args = kzalloc(sizeof(*fi_args), GFP_KERNEL);
2141 fi_args->num_devices = fs_devices->num_devices;
2142 memcpy(&fi_args->fsid, root->fs_info->fsid, sizeof(fi_args->fsid));
2144 mutex_lock(&fs_devices->device_list_mutex);
2145 list_for_each_entry_safe(device, next, &fs_devices->devices, dev_list) {
2146 if (device->devid > fi_args->max_id)
2147 fi_args->max_id = device->devid;
2149 mutex_unlock(&fs_devices->device_list_mutex);
2151 if (copy_to_user(arg, fi_args, sizeof(*fi_args)))
2158 static long btrfs_ioctl_dev_info(struct btrfs_root *root, void __user *arg)
2160 struct btrfs_ioctl_dev_info_args *di_args;
2161 struct btrfs_device *dev;
2162 struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
2164 char *s_uuid = NULL;
2165 char empty_uuid[BTRFS_UUID_SIZE] = {0};
2167 if (!capable(CAP_SYS_ADMIN))
2170 di_args = memdup_user(arg, sizeof(*di_args));
2171 if (IS_ERR(di_args))
2172 return PTR_ERR(di_args);
2174 if (memcmp(empty_uuid, di_args->uuid, BTRFS_UUID_SIZE) != 0)
2175 s_uuid = di_args->uuid;
2177 mutex_lock(&fs_devices->device_list_mutex);
2178 dev = btrfs_find_device(root, di_args->devid, s_uuid, NULL);
2179 mutex_unlock(&fs_devices->device_list_mutex);
2186 di_args->devid = dev->devid;
2187 di_args->bytes_used = dev->bytes_used;
2188 di_args->total_bytes = dev->total_bytes;
2189 memcpy(di_args->uuid, dev->uuid, sizeof(di_args->uuid));
2190 strncpy(di_args->path, dev->name, sizeof(di_args->path));
2193 if (ret == 0 && copy_to_user(arg, di_args, sizeof(*di_args)))
2200 static noinline long btrfs_ioctl_clone(struct file *file, unsigned long srcfd,
2201 u64 off, u64 olen, u64 destoff)
2203 struct inode *inode = fdentry(file)->d_inode;
2204 struct btrfs_root *root = BTRFS_I(inode)->root;
2205 struct file *src_file;
2207 struct btrfs_trans_handle *trans;
2208 struct btrfs_path *path;
2209 struct extent_buffer *leaf;
2211 struct btrfs_key key;
2216 u64 bs = root->fs_info->sb->s_blocksize;
2221 * - split compressed inline extents. annoying: we need to
2222 * decompress into destination's address_space (the file offset
2223 * may change, so source mapping won't do), then recompress (or
2224 * otherwise reinsert) a subrange.
2225 * - allow ranges within the same file to be cloned (provided
2226 * they don't overlap)?
2229 /* the destination must be opened for writing */
2230 if (!(file->f_mode & FMODE_WRITE) || (file->f_flags & O_APPEND))
2233 if (btrfs_root_readonly(root))
2236 ret = mnt_want_write_file(file);
2240 src_file = fget(srcfd);
2243 goto out_drop_write;
2246 src = src_file->f_dentry->d_inode;
2252 /* the src must be open for reading */
2253 if (!(src_file->f_mode & FMODE_READ))
2256 /* don't make the dst file partly checksummed */
2257 if ((BTRFS_I(src)->flags & BTRFS_INODE_NODATASUM) !=
2258 (BTRFS_I(inode)->flags & BTRFS_INODE_NODATASUM))
2262 if (S_ISDIR(src->i_mode) || S_ISDIR(inode->i_mode))
2266 if (src->i_sb != inode->i_sb || BTRFS_I(src)->root != root)
2270 buf = vmalloc(btrfs_level_size(root, 0));
2274 path = btrfs_alloc_path();
2282 mutex_lock_nested(&inode->i_mutex, I_MUTEX_PARENT);
2283 mutex_lock_nested(&src->i_mutex, I_MUTEX_CHILD);
2285 mutex_lock_nested(&src->i_mutex, I_MUTEX_PARENT);
2286 mutex_lock_nested(&inode->i_mutex, I_MUTEX_CHILD);
2289 /* determine range to clone */
2291 if (off + len > src->i_size || off + len < off)
2294 olen = len = src->i_size - off;
2295 /* if we extend to eof, continue to block boundary */
2296 if (off + len == src->i_size)
2297 len = ALIGN(src->i_size, bs) - off;
2299 /* verify the end result is block aligned */
2300 if (!IS_ALIGNED(off, bs) || !IS_ALIGNED(off + len, bs) ||
2301 !IS_ALIGNED(destoff, bs))
2304 if (destoff > inode->i_size) {
2305 ret = btrfs_cont_expand(inode, inode->i_size, destoff);
2310 /* truncate page cache pages from target inode range */
2311 truncate_inode_pages_range(&inode->i_data, destoff,
2312 PAGE_CACHE_ALIGN(destoff + len) - 1);
2314 /* do any pending delalloc/csum calc on src, one way or
2315 another, and lock file content */
2317 struct btrfs_ordered_extent *ordered;
2318 lock_extent(&BTRFS_I(src)->io_tree, off, off+len, GFP_NOFS);
2319 ordered = btrfs_lookup_first_ordered_extent(src, off+len);
2321 !test_range_bit(&BTRFS_I(src)->io_tree, off, off+len,
2322 EXTENT_DELALLOC, 0, NULL))
2324 unlock_extent(&BTRFS_I(src)->io_tree, off, off+len, GFP_NOFS);
2326 btrfs_put_ordered_extent(ordered);
2327 btrfs_wait_ordered_range(src, off, len);
2331 key.objectid = btrfs_ino(src);
2332 key.type = BTRFS_EXTENT_DATA_KEY;
2337 * note the key will change type as we walk through the
2340 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
2344 nritems = btrfs_header_nritems(path->nodes[0]);
2345 if (path->slots[0] >= nritems) {
2346 ret = btrfs_next_leaf(root, path);
2351 nritems = btrfs_header_nritems(path->nodes[0]);
2353 leaf = path->nodes[0];
2354 slot = path->slots[0];
2356 btrfs_item_key_to_cpu(leaf, &key, slot);
2357 if (btrfs_key_type(&key) > BTRFS_EXTENT_DATA_KEY ||
2358 key.objectid != btrfs_ino(src))
2361 if (btrfs_key_type(&key) == BTRFS_EXTENT_DATA_KEY) {
2362 struct btrfs_file_extent_item *extent;
2365 struct btrfs_key new_key;
2366 u64 disko = 0, diskl = 0;
2367 u64 datao = 0, datal = 0;
2371 size = btrfs_item_size_nr(leaf, slot);
2372 read_extent_buffer(leaf, buf,
2373 btrfs_item_ptr_offset(leaf, slot),
2376 extent = btrfs_item_ptr(leaf, slot,
2377 struct btrfs_file_extent_item);
2378 comp = btrfs_file_extent_compression(leaf, extent);
2379 type = btrfs_file_extent_type(leaf, extent);
2380 if (type == BTRFS_FILE_EXTENT_REG ||
2381 type == BTRFS_FILE_EXTENT_PREALLOC) {
2382 disko = btrfs_file_extent_disk_bytenr(leaf,
2384 diskl = btrfs_file_extent_disk_num_bytes(leaf,
2386 datao = btrfs_file_extent_offset(leaf, extent);
2387 datal = btrfs_file_extent_num_bytes(leaf,
2389 } else if (type == BTRFS_FILE_EXTENT_INLINE) {
2390 /* take upper bound, may be compressed */
2391 datal = btrfs_file_extent_ram_bytes(leaf,
2394 btrfs_release_path(path);
2396 if (key.offset + datal <= off ||
2397 key.offset >= off+len)
2400 memcpy(&new_key, &key, sizeof(new_key));
2401 new_key.objectid = btrfs_ino(inode);
2402 if (off <= key.offset)
2403 new_key.offset = key.offset + destoff - off;
2405 new_key.offset = destoff;
2408 * 1 - adjusting old extent (we may have to split it)
2409 * 1 - add new extent
2412 trans = btrfs_start_transaction(root, 3);
2413 if (IS_ERR(trans)) {
2414 ret = PTR_ERR(trans);
2418 if (type == BTRFS_FILE_EXTENT_REG ||
2419 type == BTRFS_FILE_EXTENT_PREALLOC) {
2421 * a | --- range to clone ---| b
2422 * | ------------- extent ------------- |
2425 /* substract range b */
2426 if (key.offset + datal > off + len)
2427 datal = off + len - key.offset;
2429 /* substract range a */
2430 if (off > key.offset) {
2431 datao += off - key.offset;
2432 datal -= off - key.offset;
2435 ret = btrfs_drop_extents(trans, inode,
2437 new_key.offset + datal,
2441 ret = btrfs_insert_empty_item(trans, root, path,
2445 leaf = path->nodes[0];
2446 slot = path->slots[0];
2447 write_extent_buffer(leaf, buf,
2448 btrfs_item_ptr_offset(leaf, slot),
2451 extent = btrfs_item_ptr(leaf, slot,
2452 struct btrfs_file_extent_item);
2454 /* disko == 0 means it's a hole */
2458 btrfs_set_file_extent_offset(leaf, extent,
2460 btrfs_set_file_extent_num_bytes(leaf, extent,
2463 inode_add_bytes(inode, datal);
2464 ret = btrfs_inc_extent_ref(trans, root,
2466 root->root_key.objectid,
2468 new_key.offset - datao,
2472 } else if (type == BTRFS_FILE_EXTENT_INLINE) {
2475 if (off > key.offset) {
2476 skip = off - key.offset;
2477 new_key.offset += skip;
2480 if (key.offset + datal > off+len)
2481 trim = key.offset + datal - (off+len);
2483 if (comp && (skip || trim)) {
2485 btrfs_end_transaction(trans, root);
2488 size -= skip + trim;
2489 datal -= skip + trim;
2491 ret = btrfs_drop_extents(trans, inode,
2493 new_key.offset + datal,
2497 ret = btrfs_insert_empty_item(trans, root, path,
2503 btrfs_file_extent_calc_inline_size(0);
2504 memmove(buf+start, buf+start+skip,
2508 leaf = path->nodes[0];
2509 slot = path->slots[0];
2510 write_extent_buffer(leaf, buf,
2511 btrfs_item_ptr_offset(leaf, slot),
2513 inode_add_bytes(inode, datal);
2516 btrfs_mark_buffer_dirty(leaf);
2517 btrfs_release_path(path);
2519 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
2522 * we round up to the block size at eof when
2523 * determining which extents to clone above,
2524 * but shouldn't round up the file size
2526 endoff = new_key.offset + datal;
2527 if (endoff > destoff+olen)
2528 endoff = destoff+olen;
2529 if (endoff > inode->i_size)
2530 btrfs_i_size_write(inode, endoff);
2532 ret = btrfs_update_inode(trans, root, inode);
2534 btrfs_end_transaction(trans, root);
2537 btrfs_release_path(path);
2542 btrfs_release_path(path);
2543 unlock_extent(&BTRFS_I(src)->io_tree, off, off+len, GFP_NOFS);
2545 mutex_unlock(&src->i_mutex);
2546 mutex_unlock(&inode->i_mutex);
2548 btrfs_free_path(path);
2552 mnt_drop_write_file(file);
2556 static long btrfs_ioctl_clone_range(struct file *file, void __user *argp)
2558 struct btrfs_ioctl_clone_range_args args;
2560 if (copy_from_user(&args, argp, sizeof(args)))
2562 return btrfs_ioctl_clone(file, args.src_fd, args.src_offset,
2563 args.src_length, args.dest_offset);
2567 * there are many ways the trans_start and trans_end ioctls can lead
2568 * to deadlocks. They should only be used by applications that
2569 * basically own the machine, and have a very in depth understanding
2570 * of all the possible deadlocks and enospc problems.
2572 static long btrfs_ioctl_trans_start(struct file *file)
2574 struct inode *inode = fdentry(file)->d_inode;
2575 struct btrfs_root *root = BTRFS_I(inode)->root;
2576 struct btrfs_trans_handle *trans;
2580 if (!capable(CAP_SYS_ADMIN))
2584 if (file->private_data)
2588 if (btrfs_root_readonly(root))
2591 ret = mnt_want_write_file(file);
2595 atomic_inc(&root->fs_info->open_ioctl_trans);
2598 trans = btrfs_start_ioctl_transaction(root);
2602 file->private_data = trans;
2606 atomic_dec(&root->fs_info->open_ioctl_trans);
2607 mnt_drop_write_file(file);
2612 static long btrfs_ioctl_default_subvol(struct file *file, void __user *argp)
2614 struct inode *inode = fdentry(file)->d_inode;
2615 struct btrfs_root *root = BTRFS_I(inode)->root;
2616 struct btrfs_root *new_root;
2617 struct btrfs_dir_item *di;
2618 struct btrfs_trans_handle *trans;
2619 struct btrfs_path *path;
2620 struct btrfs_key location;
2621 struct btrfs_disk_key disk_key;
2622 struct btrfs_super_block *disk_super;
2627 if (!capable(CAP_SYS_ADMIN))
2630 if (copy_from_user(&objectid, argp, sizeof(objectid)))
2634 objectid = root->root_key.objectid;
2636 location.objectid = objectid;
2637 location.type = BTRFS_ROOT_ITEM_KEY;
2638 location.offset = (u64)-1;
2640 new_root = btrfs_read_fs_root_no_name(root->fs_info, &location);
2641 if (IS_ERR(new_root))
2642 return PTR_ERR(new_root);
2644 if (btrfs_root_refs(&new_root->root_item) == 0)
2647 path = btrfs_alloc_path();
2650 path->leave_spinning = 1;
2652 trans = btrfs_start_transaction(root, 1);
2653 if (IS_ERR(trans)) {
2654 btrfs_free_path(path);
2655 return PTR_ERR(trans);
2658 dir_id = btrfs_super_root_dir(root->fs_info->super_copy);
2659 di = btrfs_lookup_dir_item(trans, root->fs_info->tree_root, path,
2660 dir_id, "default", 7, 1);
2661 if (IS_ERR_OR_NULL(di)) {
2662 btrfs_free_path(path);
2663 btrfs_end_transaction(trans, root);
2664 printk(KERN_ERR "Umm, you don't have the default dir item, "
2665 "this isn't going to work\n");
2669 btrfs_cpu_key_to_disk(&disk_key, &new_root->root_key);
2670 btrfs_set_dir_item_key(path->nodes[0], di, &disk_key);
2671 btrfs_mark_buffer_dirty(path->nodes[0]);
2672 btrfs_free_path(path);
2674 disk_super = root->fs_info->super_copy;
2675 features = btrfs_super_incompat_flags(disk_super);
2676 if (!(features & BTRFS_FEATURE_INCOMPAT_DEFAULT_SUBVOL)) {
2677 features |= BTRFS_FEATURE_INCOMPAT_DEFAULT_SUBVOL;
2678 btrfs_set_super_incompat_flags(disk_super, features);
2680 btrfs_end_transaction(trans, root);
2685 static void get_block_group_info(struct list_head *groups_list,
2686 struct btrfs_ioctl_space_info *space)
2688 struct btrfs_block_group_cache *block_group;
2690 space->total_bytes = 0;
2691 space->used_bytes = 0;
2693 list_for_each_entry(block_group, groups_list, list) {
2694 space->flags = block_group->flags;
2695 space->total_bytes += block_group->key.offset;
2696 space->used_bytes +=
2697 btrfs_block_group_used(&block_group->item);
2701 long btrfs_ioctl_space_info(struct btrfs_root *root, void __user *arg)
2703 struct btrfs_ioctl_space_args space_args;
2704 struct btrfs_ioctl_space_info space;
2705 struct btrfs_ioctl_space_info *dest;
2706 struct btrfs_ioctl_space_info *dest_orig;
2707 struct btrfs_ioctl_space_info __user *user_dest;
2708 struct btrfs_space_info *info;
2709 u64 types[] = {BTRFS_BLOCK_GROUP_DATA,
2710 BTRFS_BLOCK_GROUP_SYSTEM,
2711 BTRFS_BLOCK_GROUP_METADATA,
2712 BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA};
2719 if (copy_from_user(&space_args,
2720 (struct btrfs_ioctl_space_args __user *)arg,
2721 sizeof(space_args)))
2724 for (i = 0; i < num_types; i++) {
2725 struct btrfs_space_info *tmp;
2729 list_for_each_entry_rcu(tmp, &root->fs_info->space_info,
2731 if (tmp->flags == types[i]) {
2741 down_read(&info->groups_sem);
2742 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
2743 if (!list_empty(&info->block_groups[c]))
2746 up_read(&info->groups_sem);
2749 /* space_slots == 0 means they are asking for a count */
2750 if (space_args.space_slots == 0) {
2751 space_args.total_spaces = slot_count;
2755 slot_count = min_t(u64, space_args.space_slots, slot_count);
2757 alloc_size = sizeof(*dest) * slot_count;
2759 /* we generally have at most 6 or so space infos, one for each raid
2760 * level. So, a whole page should be more than enough for everyone
2762 if (alloc_size > PAGE_CACHE_SIZE)
2765 space_args.total_spaces = 0;
2766 dest = kmalloc(alloc_size, GFP_NOFS);
2771 /* now we have a buffer to copy into */
2772 for (i = 0; i < num_types; i++) {
2773 struct btrfs_space_info *tmp;
2780 list_for_each_entry_rcu(tmp, &root->fs_info->space_info,
2782 if (tmp->flags == types[i]) {
2791 down_read(&info->groups_sem);
2792 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
2793 if (!list_empty(&info->block_groups[c])) {
2794 get_block_group_info(&info->block_groups[c],
2796 memcpy(dest, &space, sizeof(space));
2798 space_args.total_spaces++;
2804 up_read(&info->groups_sem);
2807 user_dest = (struct btrfs_ioctl_space_info *)
2808 (arg + sizeof(struct btrfs_ioctl_space_args));
2810 if (copy_to_user(user_dest, dest_orig, alloc_size))
2815 if (ret == 0 && copy_to_user(arg, &space_args, sizeof(space_args)))
2822 * there are many ways the trans_start and trans_end ioctls can lead
2823 * to deadlocks. They should only be used by applications that
2824 * basically own the machine, and have a very in depth understanding
2825 * of all the possible deadlocks and enospc problems.
2827 long btrfs_ioctl_trans_end(struct file *file)
2829 struct inode *inode = fdentry(file)->d_inode;
2830 struct btrfs_root *root = BTRFS_I(inode)->root;
2831 struct btrfs_trans_handle *trans;
2833 trans = file->private_data;
2836 file->private_data = NULL;
2838 btrfs_end_transaction(trans, root);
2840 atomic_dec(&root->fs_info->open_ioctl_trans);
2842 mnt_drop_write_file(file);
2846 static noinline long btrfs_ioctl_start_sync(struct file *file, void __user *argp)
2848 struct btrfs_root *root = BTRFS_I(file->f_dentry->d_inode)->root;
2849 struct btrfs_trans_handle *trans;
2853 trans = btrfs_start_transaction(root, 0);
2855 return PTR_ERR(trans);
2856 transid = trans->transid;
2857 ret = btrfs_commit_transaction_async(trans, root, 0);
2859 btrfs_end_transaction(trans, root);
2864 if (copy_to_user(argp, &transid, sizeof(transid)))
2869 static noinline long btrfs_ioctl_wait_sync(struct file *file, void __user *argp)
2871 struct btrfs_root *root = BTRFS_I(file->f_dentry->d_inode)->root;
2875 if (copy_from_user(&transid, argp, sizeof(transid)))
2878 transid = 0; /* current trans */
2880 return btrfs_wait_for_commit(root, transid);
2883 static long btrfs_ioctl_scrub(struct btrfs_root *root, void __user *arg)
2886 struct btrfs_ioctl_scrub_args *sa;
2888 if (!capable(CAP_SYS_ADMIN))
2891 sa = memdup_user(arg, sizeof(*sa));
2895 ret = btrfs_scrub_dev(root, sa->devid, sa->start, sa->end,
2896 &sa->progress, sa->flags & BTRFS_SCRUB_READONLY);
2898 if (copy_to_user(arg, sa, sizeof(*sa)))
2905 static long btrfs_ioctl_scrub_cancel(struct btrfs_root *root, void __user *arg)
2907 if (!capable(CAP_SYS_ADMIN))
2910 return btrfs_scrub_cancel(root);
2913 static long btrfs_ioctl_scrub_progress(struct btrfs_root *root,
2916 struct btrfs_ioctl_scrub_args *sa;
2919 if (!capable(CAP_SYS_ADMIN))
2922 sa = memdup_user(arg, sizeof(*sa));
2926 ret = btrfs_scrub_progress(root, sa->devid, &sa->progress);
2928 if (copy_to_user(arg, sa, sizeof(*sa)))
2935 static long btrfs_ioctl_ino_to_path(struct btrfs_root *root, void __user *arg)
2941 struct btrfs_ioctl_ino_path_args *ipa = NULL;
2942 struct inode_fs_paths *ipath = NULL;
2943 struct btrfs_path *path;
2945 if (!capable(CAP_SYS_ADMIN))
2948 path = btrfs_alloc_path();
2954 ipa = memdup_user(arg, sizeof(*ipa));
2961 size = min_t(u32, ipa->size, 4096);
2962 ipath = init_ipath(size, root, path);
2963 if (IS_ERR(ipath)) {
2964 ret = PTR_ERR(ipath);
2969 ret = paths_from_inode(ipa->inum, ipath);
2973 for (i = 0; i < ipath->fspath->elem_cnt; ++i) {
2974 rel_ptr = ipath->fspath->val[i] -
2975 (u64)(unsigned long)ipath->fspath->val;
2976 ipath->fspath->val[i] = rel_ptr;
2979 ret = copy_to_user((void *)(unsigned long)ipa->fspath,
2980 (void *)(unsigned long)ipath->fspath, size);
2987 btrfs_free_path(path);
2994 static int build_ino_list(u64 inum, u64 offset, u64 root, void *ctx)
2996 struct btrfs_data_container *inodes = ctx;
2997 const size_t c = 3 * sizeof(u64);
2999 if (inodes->bytes_left >= c) {
3000 inodes->bytes_left -= c;
3001 inodes->val[inodes->elem_cnt] = inum;
3002 inodes->val[inodes->elem_cnt + 1] = offset;
3003 inodes->val[inodes->elem_cnt + 2] = root;
3004 inodes->elem_cnt += 3;
3006 inodes->bytes_missing += c - inodes->bytes_left;
3007 inodes->bytes_left = 0;
3008 inodes->elem_missed += 3;
3014 static long btrfs_ioctl_logical_to_ino(struct btrfs_root *root,
3019 u64 extent_item_pos;
3020 struct btrfs_ioctl_logical_ino_args *loi;
3021 struct btrfs_data_container *inodes = NULL;
3022 struct btrfs_path *path = NULL;
3023 struct btrfs_key key;
3025 if (!capable(CAP_SYS_ADMIN))
3028 loi = memdup_user(arg, sizeof(*loi));
3035 path = btrfs_alloc_path();
3041 size = min_t(u32, loi->size, 4096);
3042 inodes = init_data_container(size);
3043 if (IS_ERR(inodes)) {
3044 ret = PTR_ERR(inodes);
3049 ret = extent_from_logical(root->fs_info, loi->logical, path, &key);
3050 btrfs_release_path(path);
3052 if (ret & BTRFS_EXTENT_FLAG_TREE_BLOCK)
3057 extent_item_pos = loi->logical - key.objectid;
3058 ret = iterate_extent_inodes(root->fs_info, path, key.objectid,
3059 extent_item_pos, build_ino_list,
3065 ret = copy_to_user((void *)(unsigned long)loi->inodes,
3066 (void *)(unsigned long)inodes, size);
3071 btrfs_free_path(path);
3078 void update_ioctl_balance_args(struct btrfs_fs_info *fs_info, int lock,
3079 struct btrfs_ioctl_balance_args *bargs)
3081 struct btrfs_balance_control *bctl = fs_info->balance_ctl;
3083 bargs->flags = bctl->flags;
3085 if (atomic_read(&fs_info->balance_running))
3086 bargs->state |= BTRFS_BALANCE_STATE_RUNNING;
3087 if (atomic_read(&fs_info->balance_pause_req))
3088 bargs->state |= BTRFS_BALANCE_STATE_PAUSE_REQ;
3089 if (atomic_read(&fs_info->balance_cancel_req))
3090 bargs->state |= BTRFS_BALANCE_STATE_CANCEL_REQ;
3092 memcpy(&bargs->data, &bctl->data, sizeof(bargs->data));
3093 memcpy(&bargs->meta, &bctl->meta, sizeof(bargs->meta));
3094 memcpy(&bargs->sys, &bctl->sys, sizeof(bargs->sys));
3097 spin_lock(&fs_info->balance_lock);
3098 memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat));
3099 spin_unlock(&fs_info->balance_lock);
3101 memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat));
3105 static long btrfs_ioctl_balance(struct btrfs_root *root, void __user *arg)
3107 struct btrfs_fs_info *fs_info = root->fs_info;
3108 struct btrfs_ioctl_balance_args *bargs;
3109 struct btrfs_balance_control *bctl;
3112 if (!capable(CAP_SYS_ADMIN))
3115 if (fs_info->sb->s_flags & MS_RDONLY)
3118 mutex_lock(&fs_info->volume_mutex);
3119 mutex_lock(&fs_info->balance_mutex);
3122 bargs = memdup_user(arg, sizeof(*bargs));
3123 if (IS_ERR(bargs)) {
3124 ret = PTR_ERR(bargs);
3128 if (bargs->flags & BTRFS_BALANCE_RESUME) {
3129 if (!fs_info->balance_ctl) {
3134 bctl = fs_info->balance_ctl;
3135 spin_lock(&fs_info->balance_lock);
3136 bctl->flags |= BTRFS_BALANCE_RESUME;
3137 spin_unlock(&fs_info->balance_lock);
3145 if (fs_info->balance_ctl) {
3150 bctl = kzalloc(sizeof(*bctl), GFP_NOFS);
3156 bctl->fs_info = fs_info;
3158 memcpy(&bctl->data, &bargs->data, sizeof(bctl->data));
3159 memcpy(&bctl->meta, &bargs->meta, sizeof(bctl->meta));
3160 memcpy(&bctl->sys, &bargs->sys, sizeof(bctl->sys));
3162 bctl->flags = bargs->flags;
3164 /* balance everything - no filters */
3165 bctl->flags |= BTRFS_BALANCE_TYPE_MASK;
3169 ret = btrfs_balance(bctl, bargs);
3171 * bctl is freed in __cancel_balance or in free_fs_info if
3172 * restriper was paused all the way until unmount
3175 if (copy_to_user(arg, bargs, sizeof(*bargs)))
3182 mutex_unlock(&fs_info->balance_mutex);
3183 mutex_unlock(&fs_info->volume_mutex);
3187 static long btrfs_ioctl_balance_ctl(struct btrfs_root *root, int cmd)
3189 if (!capable(CAP_SYS_ADMIN))
3193 case BTRFS_BALANCE_CTL_PAUSE:
3194 return btrfs_pause_balance(root->fs_info);
3195 case BTRFS_BALANCE_CTL_CANCEL:
3196 return btrfs_cancel_balance(root->fs_info);
3202 static long btrfs_ioctl_balance_progress(struct btrfs_root *root,
3205 struct btrfs_fs_info *fs_info = root->fs_info;
3206 struct btrfs_ioctl_balance_args *bargs;
3209 if (!capable(CAP_SYS_ADMIN))
3212 mutex_lock(&fs_info->balance_mutex);
3213 if (!fs_info->balance_ctl) {
3218 bargs = kzalloc(sizeof(*bargs), GFP_NOFS);
3224 update_ioctl_balance_args(fs_info, 1, bargs);
3226 if (copy_to_user(arg, bargs, sizeof(*bargs)))
3231 mutex_unlock(&fs_info->balance_mutex);
3235 long btrfs_ioctl(struct file *file, unsigned int
3236 cmd, unsigned long arg)
3238 struct btrfs_root *root = BTRFS_I(fdentry(file)->d_inode)->root;
3239 void __user *argp = (void __user *)arg;
3242 case FS_IOC_GETFLAGS:
3243 return btrfs_ioctl_getflags(file, argp);
3244 case FS_IOC_SETFLAGS:
3245 return btrfs_ioctl_setflags(file, argp);
3246 case FS_IOC_GETVERSION:
3247 return btrfs_ioctl_getversion(file, argp);
3249 return btrfs_ioctl_fitrim(file, argp);
3250 case BTRFS_IOC_SNAP_CREATE:
3251 return btrfs_ioctl_snap_create(file, argp, 0);
3252 case BTRFS_IOC_SNAP_CREATE_V2:
3253 return btrfs_ioctl_snap_create_v2(file, argp, 0);
3254 case BTRFS_IOC_SUBVOL_CREATE:
3255 return btrfs_ioctl_snap_create(file, argp, 1);
3256 case BTRFS_IOC_SNAP_DESTROY:
3257 return btrfs_ioctl_snap_destroy(file, argp);
3258 case BTRFS_IOC_SUBVOL_GETFLAGS:
3259 return btrfs_ioctl_subvol_getflags(file, argp);
3260 case BTRFS_IOC_SUBVOL_SETFLAGS:
3261 return btrfs_ioctl_subvol_setflags(file, argp);
3262 case BTRFS_IOC_DEFAULT_SUBVOL:
3263 return btrfs_ioctl_default_subvol(file, argp);
3264 case BTRFS_IOC_DEFRAG:
3265 return btrfs_ioctl_defrag(file, NULL);
3266 case BTRFS_IOC_DEFRAG_RANGE:
3267 return btrfs_ioctl_defrag(file, argp);
3268 case BTRFS_IOC_RESIZE:
3269 return btrfs_ioctl_resize(root, argp);
3270 case BTRFS_IOC_ADD_DEV:
3271 return btrfs_ioctl_add_dev(root, argp);
3272 case BTRFS_IOC_RM_DEV:
3273 return btrfs_ioctl_rm_dev(root, argp);
3274 case BTRFS_IOC_FS_INFO:
3275 return btrfs_ioctl_fs_info(root, argp);
3276 case BTRFS_IOC_DEV_INFO:
3277 return btrfs_ioctl_dev_info(root, argp);
3278 case BTRFS_IOC_BALANCE:
3279 return btrfs_ioctl_balance(root, NULL);
3280 case BTRFS_IOC_CLONE:
3281 return btrfs_ioctl_clone(file, arg, 0, 0, 0);
3282 case BTRFS_IOC_CLONE_RANGE:
3283 return btrfs_ioctl_clone_range(file, argp);
3284 case BTRFS_IOC_TRANS_START:
3285 return btrfs_ioctl_trans_start(file);
3286 case BTRFS_IOC_TRANS_END:
3287 return btrfs_ioctl_trans_end(file);
3288 case BTRFS_IOC_TREE_SEARCH:
3289 return btrfs_ioctl_tree_search(file, argp);
3290 case BTRFS_IOC_INO_LOOKUP:
3291 return btrfs_ioctl_ino_lookup(file, argp);
3292 case BTRFS_IOC_INO_PATHS:
3293 return btrfs_ioctl_ino_to_path(root, argp);
3294 case BTRFS_IOC_LOGICAL_INO:
3295 return btrfs_ioctl_logical_to_ino(root, argp);
3296 case BTRFS_IOC_SPACE_INFO:
3297 return btrfs_ioctl_space_info(root, argp);
3298 case BTRFS_IOC_SYNC:
3299 btrfs_sync_fs(file->f_dentry->d_sb, 1);
3301 case BTRFS_IOC_START_SYNC:
3302 return btrfs_ioctl_start_sync(file, argp);
3303 case BTRFS_IOC_WAIT_SYNC:
3304 return btrfs_ioctl_wait_sync(file, argp);
3305 case BTRFS_IOC_SCRUB:
3306 return btrfs_ioctl_scrub(root, argp);
3307 case BTRFS_IOC_SCRUB_CANCEL:
3308 return btrfs_ioctl_scrub_cancel(root, argp);
3309 case BTRFS_IOC_SCRUB_PROGRESS:
3310 return btrfs_ioctl_scrub_progress(root, argp);
3311 case BTRFS_IOC_BALANCE_V2:
3312 return btrfs_ioctl_balance(root, argp);
3313 case BTRFS_IOC_BALANCE_CTL:
3314 return btrfs_ioctl_balance_ctl(root, arg);
3315 case BTRFS_IOC_BALANCE_PROGRESS:
3316 return btrfs_ioctl_balance_progress(root, argp);