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);
170 static int create_snapshot(struct btrfs_root *root, char *name, int namelen)
172 struct btrfs_pending_snapshot *pending_snapshot;
173 struct btrfs_trans_handle *trans;
176 unsigned long nr = 0;
181 ret = btrfs_check_free_space(root, 1, 0);
185 pending_snapshot = kmalloc(sizeof(*pending_snapshot), GFP_NOFS);
186 if (!pending_snapshot) {
190 pending_snapshot->name = kmalloc(namelen + 1, GFP_NOFS);
191 if (!pending_snapshot->name) {
193 kfree(pending_snapshot);
196 memcpy(pending_snapshot->name, name, namelen);
197 pending_snapshot->name[namelen] = '\0';
198 trans = btrfs_start_transaction(root, 1);
200 pending_snapshot->root = root;
201 list_add(&pending_snapshot->list,
202 &trans->transaction->pending_snapshots);
203 ret = btrfs_update_inode(trans, root, root->inode);
204 err = btrfs_commit_transaction(trans, root);
207 btrfs_btree_balance_dirty(root, nr);
211 int btrfs_defrag_file(struct file *file)
213 struct inode *inode = fdentry(file)->d_inode;
214 struct btrfs_root *root = BTRFS_I(inode)->root;
215 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
217 unsigned long last_index;
218 unsigned long ra_pages = root->fs_info->bdi.ra_pages;
219 unsigned long total_read = 0;
225 ret = btrfs_check_free_space(root, inode->i_size, 0);
229 mutex_lock(&inode->i_mutex);
230 last_index = inode->i_size >> PAGE_CACHE_SHIFT;
231 for (i = 0; i <= last_index; i++) {
232 if (total_read % ra_pages == 0) {
233 btrfs_force_ra(inode->i_mapping, &file->f_ra, file, i,
234 min(last_index, i + ra_pages - 1));
237 page = grab_cache_page(inode->i_mapping, i);
240 if (!PageUptodate(page)) {
241 btrfs_readpage(NULL, page);
243 if (!PageUptodate(page)) {
245 page_cache_release(page);
250 #if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,18)
251 ClearPageDirty(page);
253 cancel_dirty_page(page, PAGE_CACHE_SIZE);
255 wait_on_page_writeback(page);
256 set_page_extent_mapped(page);
258 page_start = (u64)page->index << PAGE_CACHE_SHIFT;
259 page_end = page_start + PAGE_CACHE_SIZE - 1;
261 lock_extent(io_tree, page_start, page_end, GFP_NOFS);
262 set_extent_delalloc(io_tree, page_start,
265 unlock_extent(io_tree, page_start, page_end, GFP_NOFS);
266 set_page_dirty(page);
268 page_cache_release(page);
269 balance_dirty_pages_ratelimited_nr(inode->i_mapping, 1);
273 mutex_unlock(&inode->i_mutex);
278 * Called inside transaction, so use GFP_NOFS
281 static int btrfs_ioctl_resize(struct btrfs_root *root, void __user *arg)
286 struct btrfs_ioctl_vol_args *vol_args;
287 struct btrfs_trans_handle *trans;
288 struct btrfs_device *device = NULL;
295 vol_args = kmalloc(sizeof(*vol_args), GFP_NOFS);
300 if (copy_from_user(vol_args, arg, sizeof(*vol_args))) {
304 namelen = strlen(vol_args->name);
305 if (namelen > BTRFS_VOL_NAME_MAX) {
310 mutex_lock(&root->fs_info->volume_mutex);
311 sizestr = vol_args->name;
312 devstr = strchr(sizestr, ':');
315 sizestr = devstr + 1;
317 devstr = vol_args->name;
318 devid = simple_strtoull(devstr, &end, 10);
319 printk(KERN_INFO "resizing devid %llu\n", devid);
321 device = btrfs_find_device(root, devid, NULL);
323 printk(KERN_INFO "resizer unable to find device %llu\n", devid);
327 if (!strcmp(sizestr, "max"))
328 new_size = device->bdev->bd_inode->i_size;
330 if (sizestr[0] == '-') {
333 } else if (sizestr[0] == '+') {
337 new_size = btrfs_parse_size(sizestr);
344 old_size = device->total_bytes;
347 if (new_size > old_size) {
351 new_size = old_size - new_size;
352 } else if (mod > 0) {
353 new_size = old_size + new_size;
356 if (new_size < 256 * 1024 * 1024) {
360 if (new_size > device->bdev->bd_inode->i_size) {
365 do_div(new_size, root->sectorsize);
366 new_size *= root->sectorsize;
368 printk(KERN_INFO "new size for %s is %llu\n",
369 device->name, (unsigned long long)new_size);
371 if (new_size > old_size) {
372 trans = btrfs_start_transaction(root, 1);
373 ret = btrfs_grow_device(trans, device, new_size);
374 btrfs_commit_transaction(trans, root);
376 ret = btrfs_shrink_device(device, new_size);
380 mutex_unlock(&root->fs_info->volume_mutex);
386 static noinline int btrfs_ioctl_snap_create(struct btrfs_root *root,
389 struct btrfs_ioctl_vol_args *vol_args;
390 struct btrfs_dir_item *di;
391 struct btrfs_path *path;
396 vol_args = kmalloc(sizeof(*vol_args), GFP_NOFS);
401 if (copy_from_user(vol_args, arg, sizeof(*vol_args))) {
406 namelen = strlen(vol_args->name);
407 if (namelen > BTRFS_VOL_NAME_MAX) {
411 if (strchr(vol_args->name, '/')) {
416 path = btrfs_alloc_path();
422 root_dirid = root->fs_info->sb->s_root->d_inode->i_ino,
423 di = btrfs_lookup_dir_item(NULL, root->fs_info->tree_root,
425 vol_args->name, namelen, 0);
426 btrfs_free_path(path);
428 if (di && !IS_ERR(di)) {
438 mutex_lock(&root->fs_info->drop_mutex);
439 if (root == root->fs_info->tree_root)
440 ret = create_subvol(root, vol_args->name, namelen);
442 ret = create_snapshot(root, vol_args->name, namelen);
443 mutex_unlock(&root->fs_info->drop_mutex);
449 static int btrfs_ioctl_defrag(struct file *file)
451 struct inode *inode = fdentry(file)->d_inode;
452 struct btrfs_root *root = BTRFS_I(inode)->root;
454 switch (inode->i_mode & S_IFMT) {
456 btrfs_defrag_root(root, 0);
457 btrfs_defrag_root(root->fs_info->extent_root, 0);
460 btrfs_defrag_file(file);
467 long btrfs_ioctl_add_dev(struct btrfs_root *root, void __user *arg)
469 struct btrfs_ioctl_vol_args *vol_args;
472 vol_args = kmalloc(sizeof(*vol_args), GFP_NOFS);
477 if (copy_from_user(vol_args, arg, sizeof(*vol_args))) {
481 ret = btrfs_init_new_device(root, vol_args->name);
488 long btrfs_ioctl_rm_dev(struct btrfs_root *root, void __user *arg)
490 struct btrfs_ioctl_vol_args *vol_args;
493 vol_args = kmalloc(sizeof(*vol_args), GFP_NOFS);
498 if (copy_from_user(vol_args, arg, sizeof(*vol_args))) {
502 ret = btrfs_rm_device(root, vol_args->name);
509 int dup_item_to_inode(struct btrfs_trans_handle *trans,
510 struct btrfs_root *root,
511 struct btrfs_path *path,
512 struct extent_buffer *leaf,
514 struct btrfs_key *key,
518 int len = btrfs_item_size_nr(leaf, slot);
519 struct btrfs_key ckey = *key;
522 dup = kmalloc(len, GFP_NOFS);
526 read_extent_buffer(leaf, dup, btrfs_item_ptr_offset(leaf, slot), len);
527 btrfs_release_path(root, path);
529 ckey.objectid = destino;
530 ret = btrfs_insert_item(trans, root, &ckey, dup, len);
535 long btrfs_ioctl_clone(struct file *file, unsigned long src_fd)
537 struct inode *inode = fdentry(file)->d_inode;
538 struct btrfs_root *root = BTRFS_I(inode)->root;
539 struct file *src_file;
541 struct btrfs_trans_handle *trans;
544 struct btrfs_path *path;
545 struct btrfs_key key;
546 struct extent_buffer *leaf;
550 src_file = fget(src_fd);
553 src = src_file->f_dentry->d_inode;
556 if (src->i_sb != inode->i_sb)
560 mutex_lock(&inode->i_mutex);
561 mutex_lock(&src->i_mutex);
563 mutex_lock(&src->i_mutex);
564 mutex_lock(&inode->i_mutex);
571 /* do any pending delalloc/csum calc on src, one way or
572 another, and lock file content */
574 filemap_write_and_wait(src->i_mapping);
575 lock_extent(&BTRFS_I(src)->io_tree, 0, (u64)-1, GFP_NOFS);
576 if (BTRFS_I(src)->delalloc_bytes == 0)
578 unlock_extent(&BTRFS_I(src)->io_tree, 0, (u64)-1, GFP_NOFS);
581 trans = btrfs_start_transaction(root, 0);
582 path = btrfs_alloc_path();
588 key.type = BTRFS_EXTENT_DATA_KEY;
589 key.objectid = src->i_ino;
595 * note the key will change type as we walk through the
598 ret = btrfs_search_slot(trans, root, &key, path, 0, 0);
602 if (path->slots[0] >= btrfs_header_nritems(path->nodes[0])) {
603 ret = btrfs_next_leaf(root, path);
609 leaf = path->nodes[0];
610 slot = path->slots[0];
611 btrfs_item_key_to_cpu(leaf, &key, slot);
612 nritems = btrfs_header_nritems(leaf);
614 if (btrfs_key_type(&key) > BTRFS_CSUM_ITEM_KEY ||
615 key.objectid != src->i_ino)
618 if (btrfs_key_type(&key) == BTRFS_EXTENT_DATA_KEY) {
619 struct btrfs_file_extent_item *extent;
622 extent = btrfs_item_ptr(leaf, slot,
623 struct btrfs_file_extent_item);
624 found_type = btrfs_file_extent_type(leaf, extent);
625 if (found_type == BTRFS_FILE_EXTENT_REG) {
626 u64 len = btrfs_file_extent_num_bytes(leaf,
628 u64 ds = btrfs_file_extent_disk_bytenr(leaf,
630 u64 dl = btrfs_file_extent_disk_num_bytes(leaf,
632 u64 off = btrfs_file_extent_offset(leaf,
634 btrfs_insert_file_extent(trans, root,
637 /* ds == 0 means there's a hole */
639 btrfs_inc_extent_ref(trans, root,
641 root->root_key.objectid,
645 pos = key.offset + len;
646 } else if (found_type == BTRFS_FILE_EXTENT_INLINE) {
647 ret = dup_item_to_inode(trans, root, path,
652 pos = key.offset + btrfs_item_size_nr(leaf,
655 } else if (btrfs_key_type(&key) == BTRFS_CSUM_ITEM_KEY) {
656 ret = dup_item_to_inode(trans, root, path, leaf,
657 slot, &key, inode->i_ino);
663 btrfs_release_path(root, path);
668 btrfs_free_path(path);
670 inode->i_blocks = src->i_blocks;
671 i_size_write(inode, src->i_size);
672 btrfs_update_inode(trans, root, inode);
674 unlock_extent(&BTRFS_I(src)->io_tree, 0, (u64)-1, GFP_NOFS);
676 btrfs_end_transaction(trans, root);
679 mutex_unlock(&src->i_mutex);
680 mutex_unlock(&inode->i_mutex);
687 * there are many ways the trans_start and trans_end ioctls can lead
688 * to deadlocks. They should only be used by applications that
689 * basically own the machine, and have a very in depth understanding
690 * of all the possible deadlocks and enospc problems.
692 long btrfs_ioctl_trans_start(struct file *file)
694 struct inode *inode = fdentry(file)->d_inode;
695 struct btrfs_root *root = BTRFS_I(inode)->root;
696 struct btrfs_trans_handle *trans;
699 if (!capable(CAP_SYS_ADMIN))
702 if (file->private_data) {
706 trans = btrfs_start_transaction(root, 0);
708 file->private_data = trans;
711 /*printk(KERN_INFO "btrfs_ioctl_trans_start on %p\n", file);*/
717 * there are many ways the trans_start and trans_end ioctls can lead
718 * to deadlocks. They should only be used by applications that
719 * basically own the machine, and have a very in depth understanding
720 * of all the possible deadlocks and enospc problems.
722 long btrfs_ioctl_trans_end(struct file *file)
724 struct inode *inode = fdentry(file)->d_inode;
725 struct btrfs_root *root = BTRFS_I(inode)->root;
726 struct btrfs_trans_handle *trans;
729 trans = file->private_data;
734 btrfs_end_transaction(trans, root);
735 file->private_data = 0;
740 long btrfs_ioctl(struct file *file, unsigned int
741 cmd, unsigned long arg)
743 struct btrfs_root *root = BTRFS_I(fdentry(file)->d_inode)->root;
746 case BTRFS_IOC_SNAP_CREATE:
747 return btrfs_ioctl_snap_create(root, (void __user *)arg);
748 case BTRFS_IOC_DEFRAG:
749 return btrfs_ioctl_defrag(file);
750 case BTRFS_IOC_RESIZE:
751 return btrfs_ioctl_resize(root, (void __user *)arg);
752 case BTRFS_IOC_ADD_DEV:
753 return btrfs_ioctl_add_dev(root, (void __user *)arg);
754 case BTRFS_IOC_RM_DEV:
755 return btrfs_ioctl_rm_dev(root, (void __user *)arg);
756 case BTRFS_IOC_BALANCE:
757 return btrfs_balance(root->fs_info->dev_root);
758 case BTRFS_IOC_CLONE:
759 return btrfs_ioctl_clone(file, arg);
760 case BTRFS_IOC_TRANS_START:
761 return btrfs_ioctl_trans_start(file);
762 case BTRFS_IOC_TRANS_END:
763 return btrfs_ioctl_trans_end(file);
765 btrfs_sync_fs(file->f_dentry->d_sb, 1);