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"
49 static noinline int create_subvol(struct btrfs_root *root, char *name,
52 struct btrfs_trans_handle *trans;
54 struct btrfs_root_item root_item;
55 struct btrfs_inode_item *inode_item;
56 struct extent_buffer *leaf;
57 struct btrfs_root *new_root = root;
62 u64 new_dirid = BTRFS_FIRST_FREE_OBJECTID;
65 mutex_lock(&root->fs_info->fs_mutex);
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 free_extent_buffer(leaf);
114 btrfs_set_root_dirid(&root_item, new_dirid);
116 key.objectid = objectid;
118 btrfs_set_key_type(&key, BTRFS_ROOT_ITEM_KEY);
119 ret = btrfs_insert_root(trans, root->fs_info->tree_root, &key,
125 * insert the directory item
127 key.offset = (u64)-1;
128 dir = root->fs_info->sb->s_root->d_inode;
129 ret = btrfs_insert_dir_item(trans, root->fs_info->tree_root,
130 name, namelen, dir->i_ino, &key,
135 ret = btrfs_insert_inode_ref(trans, root->fs_info->tree_root,
136 name, namelen, objectid,
137 root->fs_info->sb->s_root->d_inode->i_ino);
141 ret = btrfs_commit_transaction(trans, root);
145 new_root = btrfs_read_fs_root(root->fs_info, &key, name, namelen);
148 trans = btrfs_start_transaction(new_root, 1);
151 ret = btrfs_create_subvol_root(new_root, trans, new_dirid,
152 BTRFS_I(dir)->block_group);
156 /* Invalidate existing dcache entry for new subvolume. */
157 btrfs_invalidate_dcache_root(root, name, namelen);
160 nr = trans->blocks_used;
161 err = btrfs_commit_transaction(trans, new_root);
165 mutex_unlock(&root->fs_info->fs_mutex);
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 mutex_lock(&root->fs_info->fs_mutex);
183 ret = btrfs_check_free_space(root, 1, 0);
187 pending_snapshot = kmalloc(sizeof(*pending_snapshot), GFP_NOFS);
188 if (!pending_snapshot) {
192 pending_snapshot->name = kmalloc(namelen + 1, GFP_NOFS);
193 if (!pending_snapshot->name) {
195 kfree(pending_snapshot);
198 memcpy(pending_snapshot->name, name, namelen);
199 pending_snapshot->name[namelen] = '\0';
200 trans = btrfs_start_transaction(root, 1);
202 pending_snapshot->root = root;
203 list_add(&pending_snapshot->list,
204 &trans->transaction->pending_snapshots);
205 ret = btrfs_update_inode(trans, root, root->inode);
206 err = btrfs_commit_transaction(trans, root);
209 mutex_unlock(&root->fs_info->fs_mutex);
210 btrfs_btree_balance_dirty(root, nr);
211 btrfs_throttle(root);
215 int btrfs_defrag_file(struct file *file)
217 struct inode *inode = fdentry(file)->d_inode;
218 struct btrfs_root *root = BTRFS_I(inode)->root;
219 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
221 unsigned long last_index;
222 unsigned long ra_pages = root->fs_info->bdi.ra_pages;
223 unsigned long total_read = 0;
229 mutex_lock(&root->fs_info->fs_mutex);
230 ret = btrfs_check_free_space(root, inode->i_size, 0);
231 mutex_unlock(&root->fs_info->fs_mutex);
235 mutex_lock(&inode->i_mutex);
236 last_index = inode->i_size >> PAGE_CACHE_SHIFT;
237 for (i = 0; i <= last_index; i++) {
238 if (total_read % ra_pages == 0) {
239 btrfs_force_ra(inode->i_mapping, &file->f_ra, file, i,
240 min(last_index, i + ra_pages - 1));
243 page = grab_cache_page(inode->i_mapping, i);
246 if (!PageUptodate(page)) {
247 btrfs_readpage(NULL, page);
249 if (!PageUptodate(page)) {
251 page_cache_release(page);
256 #if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,18)
257 ClearPageDirty(page);
259 cancel_dirty_page(page, PAGE_CACHE_SIZE);
261 wait_on_page_writeback(page);
262 set_page_extent_mapped(page);
264 page_start = (u64)page->index << PAGE_CACHE_SHIFT;
265 page_end = page_start + PAGE_CACHE_SIZE - 1;
267 lock_extent(io_tree, page_start, page_end, GFP_NOFS);
268 set_extent_delalloc(io_tree, page_start,
271 unlock_extent(io_tree, page_start, page_end, GFP_NOFS);
272 set_page_dirty(page);
274 page_cache_release(page);
275 balance_dirty_pages_ratelimited_nr(inode->i_mapping, 1);
279 mutex_unlock(&inode->i_mutex);
284 * Called inside transaction, so use GFP_NOFS
287 static int btrfs_ioctl_resize(struct btrfs_root *root, void __user *arg)
292 struct btrfs_ioctl_vol_args *vol_args;
293 struct btrfs_trans_handle *trans;
294 struct btrfs_device *device = NULL;
301 vol_args = kmalloc(sizeof(*vol_args), GFP_NOFS);
306 if (copy_from_user(vol_args, arg, sizeof(*vol_args))) {
310 namelen = strlen(vol_args->name);
311 if (namelen > BTRFS_VOL_NAME_MAX) {
316 mutex_lock(&root->fs_info->fs_mutex);
317 sizestr = vol_args->name;
318 devstr = strchr(sizestr, ':');
321 sizestr = devstr + 1;
323 devstr = vol_args->name;
324 devid = simple_strtoull(devstr, &end, 10);
325 printk(KERN_INFO "resizing devid %llu\n", devid);
327 device = btrfs_find_device(root, devid, NULL);
329 printk(KERN_INFO "resizer unable to find device %llu\n", devid);
333 if (!strcmp(sizestr, "max"))
334 new_size = device->bdev->bd_inode->i_size;
336 if (sizestr[0] == '-') {
339 } else if (sizestr[0] == '+') {
343 new_size = btrfs_parse_size(sizestr);
350 old_size = device->total_bytes;
353 if (new_size > old_size) {
357 new_size = old_size - new_size;
358 } else if (mod > 0) {
359 new_size = old_size + new_size;
362 if (new_size < 256 * 1024 * 1024) {
366 if (new_size > device->bdev->bd_inode->i_size) {
371 do_div(new_size, root->sectorsize);
372 new_size *= root->sectorsize;
374 printk(KERN_INFO "new size for %s is %llu\n",
375 device->name, (unsigned long long)new_size);
377 if (new_size > old_size) {
378 trans = btrfs_start_transaction(root, 1);
379 ret = btrfs_grow_device(trans, device, new_size);
380 btrfs_commit_transaction(trans, root);
382 ret = btrfs_shrink_device(device, new_size);
386 mutex_unlock(&root->fs_info->fs_mutex);
392 static noinline int btrfs_ioctl_snap_create(struct btrfs_root *root,
395 struct btrfs_ioctl_vol_args *vol_args;
396 struct btrfs_dir_item *di;
397 struct btrfs_path *path;
402 vol_args = kmalloc(sizeof(*vol_args), GFP_NOFS);
407 if (copy_from_user(vol_args, arg, sizeof(*vol_args))) {
412 namelen = strlen(vol_args->name);
413 if (namelen > BTRFS_VOL_NAME_MAX) {
417 if (strchr(vol_args->name, '/')) {
422 path = btrfs_alloc_path();
428 root_dirid = root->fs_info->sb->s_root->d_inode->i_ino,
429 mutex_lock(&root->fs_info->fs_mutex);
430 di = btrfs_lookup_dir_item(NULL, root->fs_info->tree_root,
432 vol_args->name, namelen, 0);
433 mutex_unlock(&root->fs_info->fs_mutex);
434 btrfs_free_path(path);
436 if (di && !IS_ERR(di)) {
446 if (root == root->fs_info->tree_root)
447 ret = create_subvol(root, vol_args->name, namelen);
449 ret = create_snapshot(root, vol_args->name, namelen);
455 static int btrfs_ioctl_defrag(struct file *file)
457 struct inode *inode = fdentry(file)->d_inode;
458 struct btrfs_root *root = BTRFS_I(inode)->root;
460 switch (inode->i_mode & S_IFMT) {
462 mutex_lock(&root->fs_info->fs_mutex);
463 btrfs_defrag_root(root, 0);
464 btrfs_defrag_root(root->fs_info->extent_root, 0);
465 mutex_unlock(&root->fs_info->fs_mutex);
468 btrfs_defrag_file(file);
475 long btrfs_ioctl_add_dev(struct btrfs_root *root, void __user *arg)
477 struct btrfs_ioctl_vol_args *vol_args;
480 vol_args = kmalloc(sizeof(*vol_args), GFP_NOFS);
485 if (copy_from_user(vol_args, arg, sizeof(*vol_args))) {
489 ret = btrfs_init_new_device(root, vol_args->name);
496 long btrfs_ioctl_rm_dev(struct btrfs_root *root, void __user *arg)
498 struct btrfs_ioctl_vol_args *vol_args;
501 vol_args = kmalloc(sizeof(*vol_args), GFP_NOFS);
506 if (copy_from_user(vol_args, arg, sizeof(*vol_args))) {
510 ret = btrfs_rm_device(root, vol_args->name);
517 int dup_item_to_inode(struct btrfs_trans_handle *trans,
518 struct btrfs_root *root,
519 struct btrfs_path *path,
520 struct extent_buffer *leaf,
522 struct btrfs_key *key,
526 int len = btrfs_item_size_nr(leaf, slot);
527 struct btrfs_key ckey = *key;
530 dup = kmalloc(len, GFP_NOFS);
534 read_extent_buffer(leaf, dup, btrfs_item_ptr_offset(leaf, slot), len);
535 btrfs_release_path(root, path);
537 ckey.objectid = destino;
538 ret = btrfs_insert_item(trans, root, &ckey, dup, len);
543 long btrfs_ioctl_clone(struct file *file, unsigned long src_fd)
545 struct inode *inode = fdentry(file)->d_inode;
546 struct btrfs_root *root = BTRFS_I(inode)->root;
547 struct file *src_file;
549 struct btrfs_trans_handle *trans;
552 struct btrfs_path *path;
553 struct btrfs_key key;
554 struct extent_buffer *leaf;
558 src_file = fget(src_fd);
561 src = src_file->f_dentry->d_inode;
564 if (src->i_sb != inode->i_sb)
568 mutex_lock(&inode->i_mutex);
569 mutex_lock(&src->i_mutex);
571 mutex_lock(&src->i_mutex);
572 mutex_lock(&inode->i_mutex);
579 /* do any pending delalloc/csum calc on src, one way or
580 another, and lock file content */
582 filemap_write_and_wait(src->i_mapping);
583 lock_extent(&BTRFS_I(src)->io_tree, 0, (u64)-1, GFP_NOFS);
584 if (BTRFS_I(src)->delalloc_bytes == 0)
586 unlock_extent(&BTRFS_I(src)->io_tree, 0, (u64)-1, GFP_NOFS);
589 mutex_lock(&root->fs_info->fs_mutex);
590 trans = btrfs_start_transaction(root, 0);
591 path = btrfs_alloc_path();
597 key.type = BTRFS_EXTENT_DATA_KEY;
598 key.objectid = src->i_ino;
604 * note the key will change type as we walk through the
607 ret = btrfs_search_slot(trans, root, &key, path, 0, 0);
611 if (path->slots[0] >= btrfs_header_nritems(path->nodes[0])) {
612 ret = btrfs_next_leaf(root, path);
618 leaf = path->nodes[0];
619 slot = path->slots[0];
620 btrfs_item_key_to_cpu(leaf, &key, slot);
621 nritems = btrfs_header_nritems(leaf);
623 if (btrfs_key_type(&key) > BTRFS_CSUM_ITEM_KEY ||
624 key.objectid != src->i_ino)
627 if (btrfs_key_type(&key) == BTRFS_EXTENT_DATA_KEY) {
628 struct btrfs_file_extent_item *extent;
631 extent = btrfs_item_ptr(leaf, slot,
632 struct btrfs_file_extent_item);
633 found_type = btrfs_file_extent_type(leaf, extent);
634 if (found_type == BTRFS_FILE_EXTENT_REG) {
635 u64 len = btrfs_file_extent_num_bytes(leaf,
637 u64 ds = btrfs_file_extent_disk_bytenr(leaf,
639 u64 dl = btrfs_file_extent_disk_num_bytes(leaf,
641 u64 off = btrfs_file_extent_offset(leaf,
643 btrfs_insert_file_extent(trans, root,
646 /* ds == 0 means there's a hole */
648 btrfs_inc_extent_ref(trans, root,
650 root->root_key.objectid,
654 pos = key.offset + len;
655 } else if (found_type == BTRFS_FILE_EXTENT_INLINE) {
656 ret = dup_item_to_inode(trans, root, path,
661 pos = key.offset + btrfs_item_size_nr(leaf,
664 } else if (btrfs_key_type(&key) == BTRFS_CSUM_ITEM_KEY) {
665 ret = dup_item_to_inode(trans, root, path, leaf,
666 slot, &key, inode->i_ino);
672 btrfs_release_path(root, path);
677 btrfs_free_path(path);
679 inode->i_blocks = src->i_blocks;
680 i_size_write(inode, src->i_size);
681 btrfs_update_inode(trans, root, inode);
683 unlock_extent(&BTRFS_I(src)->io_tree, 0, (u64)-1, GFP_NOFS);
685 btrfs_end_transaction(trans, root);
686 mutex_unlock(&root->fs_info->fs_mutex);
689 mutex_unlock(&src->i_mutex);
690 mutex_unlock(&inode->i_mutex);
697 * there are many ways the trans_start and trans_end ioctls can lead
698 * to deadlocks. They should only be used by applications that
699 * basically own the machine, and have a very in depth understanding
700 * of all the possible deadlocks and enospc problems.
702 long btrfs_ioctl_trans_start(struct file *file)
704 struct inode *inode = fdentry(file)->d_inode;
705 struct btrfs_root *root = BTRFS_I(inode)->root;
706 struct btrfs_trans_handle *trans;
709 mutex_lock(&root->fs_info->fs_mutex);
710 if (file->private_data) {
714 trans = btrfs_start_transaction(root, 0);
716 file->private_data = trans;
719 /*printk(KERN_INFO "btrfs_ioctl_trans_start on %p\n", file);*/
721 mutex_unlock(&root->fs_info->fs_mutex);
726 * there are many ways the trans_start and trans_end ioctls can lead
727 * to deadlocks. They should only be used by applications that
728 * basically own the machine, and have a very in depth understanding
729 * of all the possible deadlocks and enospc problems.
731 long btrfs_ioctl_trans_end(struct file *file)
733 struct inode *inode = fdentry(file)->d_inode;
734 struct btrfs_root *root = BTRFS_I(inode)->root;
735 struct btrfs_trans_handle *trans;
738 mutex_lock(&root->fs_info->fs_mutex);
739 trans = file->private_data;
744 btrfs_end_transaction(trans, root);
745 file->private_data = 0;
747 mutex_unlock(&root->fs_info->fs_mutex);
751 long btrfs_ioctl(struct file *file, unsigned int
752 cmd, unsigned long arg)
754 struct btrfs_root *root = BTRFS_I(fdentry(file)->d_inode)->root;
757 case BTRFS_IOC_SNAP_CREATE:
758 return btrfs_ioctl_snap_create(root, (void __user *)arg);
759 case BTRFS_IOC_DEFRAG:
760 return btrfs_ioctl_defrag(file);
761 case BTRFS_IOC_RESIZE:
762 return btrfs_ioctl_resize(root, (void __user *)arg);
763 case BTRFS_IOC_ADD_DEV:
764 return btrfs_ioctl_add_dev(root, (void __user *)arg);
765 case BTRFS_IOC_RM_DEV:
766 return btrfs_ioctl_rm_dev(root, (void __user *)arg);
767 case BTRFS_IOC_BALANCE:
768 return btrfs_balance(root->fs_info->dev_root);
769 case BTRFS_IOC_CLONE:
770 return btrfs_ioctl_clone(file, arg);
771 case BTRFS_IOC_TRANS_START:
772 return btrfs_ioctl_trans_start(file);
773 case BTRFS_IOC_TRANS_END:
774 return btrfs_ioctl_trans_end(file);
776 btrfs_sync_fs(file->f_dentry->d_sb, 1);