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/pagemap.h>
25 #include <linux/highmem.h>
26 #include <linux/time.h>
27 #include <linux/init.h>
28 #include <linux/string.h>
29 #include <linux/smp_lock.h>
30 #include <linux/backing-dev.h>
31 #include <linux/mpage.h>
32 #include <linux/swap.h>
33 #include <linux/writeback.h>
34 #include <linux/statfs.h>
35 #include <linux/compat.h>
36 #include <linux/bit_spinlock.h>
37 #include <linux/version.h>
38 #include <linux/xattr.h>
41 #include "transaction.h"
42 #include "btrfs_inode.h"
44 #include "print-tree.h"
50 static noinline int create_subvol(struct btrfs_root *root, char *name,
53 struct btrfs_trans_handle *trans;
55 struct btrfs_root_item root_item;
56 struct btrfs_inode_item *inode_item;
57 struct extent_buffer *leaf;
58 struct btrfs_root *new_root = root;
63 u64 new_dirid = BTRFS_FIRST_FREE_OBJECTID;
66 ret = btrfs_check_free_space(root, 1, 0);
70 trans = btrfs_start_transaction(root, 1);
73 ret = btrfs_find_free_objectid(trans, root->fs_info->tree_root,
78 leaf = btrfs_alloc_free_block(trans, root, root->leafsize,
79 objectid, trans->transid, 0, 0,
84 btrfs_set_header_nritems(leaf, 0);
85 btrfs_set_header_level(leaf, 0);
86 btrfs_set_header_bytenr(leaf, leaf->start);
87 btrfs_set_header_generation(leaf, trans->transid);
88 btrfs_set_header_owner(leaf, objectid);
90 write_extent_buffer(leaf, root->fs_info->fsid,
91 (unsigned long)btrfs_header_fsid(leaf),
93 btrfs_mark_buffer_dirty(leaf);
95 inode_item = &root_item.inode;
96 memset(inode_item, 0, sizeof(*inode_item));
97 inode_item->generation = cpu_to_le64(1);
98 inode_item->size = cpu_to_le64(3);
99 inode_item->nlink = cpu_to_le32(1);
100 inode_item->nblocks = cpu_to_le64(1);
101 inode_item->mode = cpu_to_le32(S_IFDIR | 0755);
103 btrfs_set_root_bytenr(&root_item, leaf->start);
104 btrfs_set_root_level(&root_item, 0);
105 btrfs_set_root_refs(&root_item, 1);
106 btrfs_set_root_used(&root_item, 0);
108 memset(&root_item.drop_progress, 0, sizeof(root_item.drop_progress));
109 root_item.drop_level = 0;
111 btrfs_tree_unlock(leaf);
112 free_extent_buffer(leaf);
115 btrfs_set_root_dirid(&root_item, new_dirid);
117 key.objectid = objectid;
119 btrfs_set_key_type(&key, BTRFS_ROOT_ITEM_KEY);
120 ret = btrfs_insert_root(trans, root->fs_info->tree_root, &key,
126 * insert the directory item
128 key.offset = (u64)-1;
129 dir = root->fs_info->sb->s_root->d_inode;
130 ret = btrfs_insert_dir_item(trans, root->fs_info->tree_root,
131 name, namelen, dir->i_ino, &key,
136 ret = btrfs_insert_inode_ref(trans, root->fs_info->tree_root,
137 name, namelen, objectid,
138 root->fs_info->sb->s_root->d_inode->i_ino);
142 ret = btrfs_commit_transaction(trans, root);
146 new_root = btrfs_read_fs_root(root->fs_info, &key, name, namelen);
149 trans = btrfs_start_transaction(new_root, 1);
152 ret = btrfs_create_subvol_root(new_root, trans, new_dirid,
153 BTRFS_I(dir)->block_group);
157 /* Invalidate existing dcache entry for new subvolume. */
158 btrfs_invalidate_dcache_root(root, name, namelen);
161 nr = trans->blocks_used;
162 err = btrfs_commit_transaction(trans, new_root);
166 btrfs_btree_balance_dirty(root, nr);
167 btrfs_throttle(root);
171 static int create_snapshot(struct btrfs_root *root, char *name, int namelen)
173 struct btrfs_pending_snapshot *pending_snapshot;
174 struct btrfs_trans_handle *trans;
177 unsigned long nr = 0;
182 ret = btrfs_check_free_space(root, 1, 0);
186 pending_snapshot = kmalloc(sizeof(*pending_snapshot), GFP_NOFS);
187 if (!pending_snapshot) {
191 pending_snapshot->name = kmalloc(namelen + 1, GFP_NOFS);
192 if (!pending_snapshot->name) {
194 kfree(pending_snapshot);
197 memcpy(pending_snapshot->name, name, namelen);
198 pending_snapshot->name[namelen] = '\0';
199 trans = btrfs_start_transaction(root, 1);
201 pending_snapshot->root = root;
202 list_add(&pending_snapshot->list,
203 &trans->transaction->pending_snapshots);
204 ret = btrfs_update_inode(trans, root, root->inode);
205 err = btrfs_commit_transaction(trans, root);
208 btrfs_btree_balance_dirty(root, nr);
209 btrfs_throttle(root);
213 int btrfs_defrag_file(struct file *file)
215 struct inode *inode = fdentry(file)->d_inode;
216 struct btrfs_root *root = BTRFS_I(inode)->root;
217 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
219 unsigned long last_index;
220 unsigned long ra_pages = root->fs_info->bdi.ra_pages;
221 unsigned long total_read = 0;
227 ret = btrfs_check_free_space(root, inode->i_size, 0);
231 mutex_lock(&inode->i_mutex);
232 last_index = inode->i_size >> PAGE_CACHE_SHIFT;
233 for (i = 0; i <= last_index; i++) {
234 if (total_read % ra_pages == 0) {
235 btrfs_force_ra(inode->i_mapping, &file->f_ra, file, i,
236 min(last_index, i + ra_pages - 1));
239 page = grab_cache_page(inode->i_mapping, i);
242 if (!PageUptodate(page)) {
243 btrfs_readpage(NULL, page);
245 if (!PageUptodate(page)) {
247 page_cache_release(page);
252 #if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,18)
253 ClearPageDirty(page);
255 cancel_dirty_page(page, PAGE_CACHE_SIZE);
257 wait_on_page_writeback(page);
258 set_page_extent_mapped(page);
260 page_start = (u64)page->index << PAGE_CACHE_SHIFT;
261 page_end = page_start + PAGE_CACHE_SIZE - 1;
263 lock_extent(io_tree, page_start, page_end, GFP_NOFS);
264 set_extent_delalloc(io_tree, page_start,
267 unlock_extent(io_tree, page_start, page_end, GFP_NOFS);
268 set_page_dirty(page);
270 page_cache_release(page);
271 balance_dirty_pages_ratelimited_nr(inode->i_mapping, 1);
275 mutex_unlock(&inode->i_mutex);
280 * Called inside transaction, so use GFP_NOFS
283 static int btrfs_ioctl_resize(struct btrfs_root *root, void __user *arg)
288 struct btrfs_ioctl_vol_args *vol_args;
289 struct btrfs_trans_handle *trans;
290 struct btrfs_device *device = NULL;
297 vol_args = kmalloc(sizeof(*vol_args), GFP_NOFS);
302 if (copy_from_user(vol_args, arg, sizeof(*vol_args))) {
306 namelen = strlen(vol_args->name);
307 if (namelen > BTRFS_VOL_NAME_MAX) {
312 mutex_lock(&root->fs_info->alloc_mutex);
313 mutex_lock(&root->fs_info->chunk_mutex);
314 sizestr = vol_args->name;
315 devstr = strchr(sizestr, ':');
318 sizestr = devstr + 1;
320 devstr = vol_args->name;
321 devid = simple_strtoull(devstr, &end, 10);
322 printk(KERN_INFO "resizing devid %llu\n", devid);
324 device = btrfs_find_device(root, devid, NULL);
326 printk(KERN_INFO "resizer unable to find device %llu\n", devid);
330 if (!strcmp(sizestr, "max"))
331 new_size = device->bdev->bd_inode->i_size;
333 if (sizestr[0] == '-') {
336 } else if (sizestr[0] == '+') {
340 new_size = btrfs_parse_size(sizestr);
347 old_size = device->total_bytes;
350 if (new_size > old_size) {
354 new_size = old_size - new_size;
355 } else if (mod > 0) {
356 new_size = old_size + new_size;
359 if (new_size < 256 * 1024 * 1024) {
363 if (new_size > device->bdev->bd_inode->i_size) {
368 do_div(new_size, root->sectorsize);
369 new_size *= root->sectorsize;
371 printk(KERN_INFO "new size for %s is %llu\n",
372 device->name, (unsigned long long)new_size);
374 if (new_size > old_size) {
375 trans = btrfs_start_transaction(root, 1);
376 ret = btrfs_grow_device(trans, device, new_size);
377 btrfs_commit_transaction(trans, root);
379 ret = btrfs_shrink_device(device, new_size);
383 mutex_lock(&root->fs_info->alloc_mutex);
384 mutex_lock(&root->fs_info->chunk_mutex);
390 static noinline int btrfs_ioctl_snap_create(struct btrfs_root *root,
393 struct btrfs_ioctl_vol_args *vol_args;
394 struct btrfs_dir_item *di;
395 struct btrfs_path *path;
400 vol_args = kmalloc(sizeof(*vol_args), GFP_NOFS);
405 if (copy_from_user(vol_args, arg, sizeof(*vol_args))) {
410 namelen = strlen(vol_args->name);
411 if (namelen > BTRFS_VOL_NAME_MAX) {
415 if (strchr(vol_args->name, '/')) {
420 path = btrfs_alloc_path();
426 root_dirid = root->fs_info->sb->s_root->d_inode->i_ino,
427 di = btrfs_lookup_dir_item(NULL, root->fs_info->tree_root,
429 vol_args->name, namelen, 0);
430 btrfs_free_path(path);
432 if (di && !IS_ERR(di)) {
442 mutex_lock(&root->fs_info->drop_mutex);
443 if (root == root->fs_info->tree_root)
444 ret = create_subvol(root, vol_args->name, namelen);
446 ret = create_snapshot(root, vol_args->name, namelen);
447 mutex_unlock(&root->fs_info->drop_mutex);
453 static int btrfs_ioctl_defrag(struct file *file)
455 struct inode *inode = fdentry(file)->d_inode;
456 struct btrfs_root *root = BTRFS_I(inode)->root;
458 switch (inode->i_mode & S_IFMT) {
460 btrfs_defrag_root(root, 0);
461 btrfs_defrag_root(root->fs_info->extent_root, 0);
464 btrfs_defrag_file(file);
471 long btrfs_ioctl_add_dev(struct btrfs_root *root, void __user *arg)
473 struct btrfs_ioctl_vol_args *vol_args;
476 vol_args = kmalloc(sizeof(*vol_args), GFP_NOFS);
481 if (copy_from_user(vol_args, arg, sizeof(*vol_args))) {
485 ret = btrfs_init_new_device(root, vol_args->name);
492 long btrfs_ioctl_rm_dev(struct btrfs_root *root, void __user *arg)
494 struct btrfs_ioctl_vol_args *vol_args;
497 vol_args = kmalloc(sizeof(*vol_args), GFP_NOFS);
502 if (copy_from_user(vol_args, arg, sizeof(*vol_args))) {
506 ret = btrfs_rm_device(root, vol_args->name);
513 int dup_item_to_inode(struct btrfs_trans_handle *trans,
514 struct btrfs_root *root,
515 struct btrfs_path *path,
516 struct extent_buffer *leaf,
518 struct btrfs_key *key,
522 int len = btrfs_item_size_nr(leaf, slot);
523 struct btrfs_key ckey = *key;
526 dup = kmalloc(len, GFP_NOFS);
530 read_extent_buffer(leaf, dup, btrfs_item_ptr_offset(leaf, slot), len);
531 btrfs_release_path(root, path);
533 ckey.objectid = destino;
534 ret = btrfs_insert_item(trans, root, &ckey, dup, len);
539 long btrfs_ioctl_clone(struct file *file, unsigned long src_fd)
541 struct inode *inode = fdentry(file)->d_inode;
542 struct btrfs_root *root = BTRFS_I(inode)->root;
543 struct file *src_file;
545 struct btrfs_trans_handle *trans;
548 struct btrfs_path *path;
549 struct btrfs_key key;
550 struct extent_buffer *leaf;
554 src_file = fget(src_fd);
557 src = src_file->f_dentry->d_inode;
560 if (src->i_sb != inode->i_sb)
564 mutex_lock(&inode->i_mutex);
565 mutex_lock(&src->i_mutex);
567 mutex_lock(&src->i_mutex);
568 mutex_lock(&inode->i_mutex);
575 /* do any pending delalloc/csum calc on src, one way or
576 another, and lock file content */
578 filemap_write_and_wait(src->i_mapping);
579 lock_extent(&BTRFS_I(src)->io_tree, 0, (u64)-1, GFP_NOFS);
580 if (BTRFS_I(src)->delalloc_bytes == 0)
582 unlock_extent(&BTRFS_I(src)->io_tree, 0, (u64)-1, GFP_NOFS);
585 trans = btrfs_start_transaction(root, 0);
586 path = btrfs_alloc_path();
592 key.type = BTRFS_EXTENT_DATA_KEY;
593 key.objectid = src->i_ino;
599 * note the key will change type as we walk through the
602 ret = btrfs_search_slot(trans, root, &key, path, 0, 0);
606 if (path->slots[0] >= btrfs_header_nritems(path->nodes[0])) {
607 ret = btrfs_next_leaf(root, path);
613 leaf = path->nodes[0];
614 slot = path->slots[0];
615 btrfs_item_key_to_cpu(leaf, &key, slot);
616 nritems = btrfs_header_nritems(leaf);
618 if (btrfs_key_type(&key) > BTRFS_CSUM_ITEM_KEY ||
619 key.objectid != src->i_ino)
622 if (btrfs_key_type(&key) == BTRFS_EXTENT_DATA_KEY) {
623 struct btrfs_file_extent_item *extent;
626 extent = btrfs_item_ptr(leaf, slot,
627 struct btrfs_file_extent_item);
628 found_type = btrfs_file_extent_type(leaf, extent);
629 if (found_type == BTRFS_FILE_EXTENT_REG) {
630 u64 len = btrfs_file_extent_num_bytes(leaf,
632 u64 ds = btrfs_file_extent_disk_bytenr(leaf,
634 u64 dl = btrfs_file_extent_disk_num_bytes(leaf,
636 u64 off = btrfs_file_extent_offset(leaf,
638 btrfs_insert_file_extent(trans, root,
641 /* ds == 0 means there's a hole */
643 btrfs_inc_extent_ref(trans, root,
645 root->root_key.objectid,
649 pos = key.offset + len;
650 } else if (found_type == BTRFS_FILE_EXTENT_INLINE) {
651 ret = dup_item_to_inode(trans, root, path,
656 pos = key.offset + btrfs_item_size_nr(leaf,
659 } else if (btrfs_key_type(&key) == BTRFS_CSUM_ITEM_KEY) {
660 ret = dup_item_to_inode(trans, root, path, leaf,
661 slot, &key, inode->i_ino);
667 btrfs_release_path(root, path);
672 btrfs_free_path(path);
674 inode->i_blocks = src->i_blocks;
675 i_size_write(inode, src->i_size);
676 btrfs_update_inode(trans, root, inode);
678 unlock_extent(&BTRFS_I(src)->io_tree, 0, (u64)-1, GFP_NOFS);
680 btrfs_end_transaction(trans, root);
683 mutex_unlock(&src->i_mutex);
684 mutex_unlock(&inode->i_mutex);
691 * there are many ways the trans_start and trans_end ioctls can lead
692 * to deadlocks. They should only be used by applications that
693 * basically own the machine, and have a very in depth understanding
694 * of all the possible deadlocks and enospc problems.
696 long btrfs_ioctl_trans_start(struct file *file)
698 struct inode *inode = fdentry(file)->d_inode;
699 struct btrfs_root *root = BTRFS_I(inode)->root;
700 struct btrfs_trans_handle *trans;
703 if (!capable(CAP_SYS_ADMIN))
706 if (file->private_data) {
710 trans = btrfs_start_transaction(root, 0);
712 file->private_data = trans;
715 /*printk(KERN_INFO "btrfs_ioctl_trans_start on %p\n", file);*/
721 * there are many ways the trans_start and trans_end ioctls can lead
722 * to deadlocks. They should only be used by applications that
723 * basically own the machine, and have a very in depth understanding
724 * of all the possible deadlocks and enospc problems.
726 long btrfs_ioctl_trans_end(struct file *file)
728 struct inode *inode = fdentry(file)->d_inode;
729 struct btrfs_root *root = BTRFS_I(inode)->root;
730 struct btrfs_trans_handle *trans;
733 trans = file->private_data;
738 btrfs_end_transaction(trans, root);
739 file->private_data = 0;
744 long btrfs_ioctl(struct file *file, unsigned int
745 cmd, unsigned long arg)
747 struct btrfs_root *root = BTRFS_I(fdentry(file)->d_inode)->root;
750 case BTRFS_IOC_SNAP_CREATE:
751 return btrfs_ioctl_snap_create(root, (void __user *)arg);
752 case BTRFS_IOC_DEFRAG:
753 return btrfs_ioctl_defrag(file);
754 case BTRFS_IOC_RESIZE:
755 return btrfs_ioctl_resize(root, (void __user *)arg);
756 case BTRFS_IOC_ADD_DEV:
757 return btrfs_ioctl_add_dev(root, (void __user *)arg);
758 case BTRFS_IOC_RM_DEV:
759 return btrfs_ioctl_rm_dev(root, (void __user *)arg);
760 case BTRFS_IOC_BALANCE:
761 return btrfs_balance(root->fs_info->dev_root);
762 case BTRFS_IOC_CLONE:
763 return btrfs_ioctl_clone(file, arg);
764 case BTRFS_IOC_TRANS_START:
765 return btrfs_ioctl_trans_start(file);
766 case BTRFS_IOC_TRANS_END:
767 return btrfs_ioctl_trans_end(file);
769 btrfs_sync_fs(file->f_dentry->d_sb, 1);