Merge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/jmorris...
[sfrench/cifs-2.6.git] / fs / btrfs / ioctl.c
1 /*
2  * Copyright (C) 2007 Oracle.  All rights reserved.
3  *
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.
7  *
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.
12  *
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.
17  */
18
19 #include <linux/kernel.h>
20 #include <linux/bio.h>
21 #include <linux/buffer_head.h>
22 #include <linux/file.h>
23 #include <linux/fs.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 "compat.h"
43 #include "ctree.h"
44 #include "disk-io.h"
45 #include "transaction.h"
46 #include "btrfs_inode.h"
47 #include "ioctl.h"
48 #include "print-tree.h"
49 #include "volumes.h"
50 #include "locking.h"
51
52 /* Mask out flags that are inappropriate for the given type of inode. */
53 static inline __u32 btrfs_mask_flags(umode_t mode, __u32 flags)
54 {
55         if (S_ISDIR(mode))
56                 return flags;
57         else if (S_ISREG(mode))
58                 return flags & ~FS_DIRSYNC_FL;
59         else
60                 return flags & (FS_NODUMP_FL | FS_NOATIME_FL);
61 }
62
63 /*
64  * Export inode flags to the format expected by the FS_IOC_GETFLAGS ioctl.
65  */
66 static unsigned int btrfs_flags_to_ioctl(unsigned int flags)
67 {
68         unsigned int iflags = 0;
69
70         if (flags & BTRFS_INODE_SYNC)
71                 iflags |= FS_SYNC_FL;
72         if (flags & BTRFS_INODE_IMMUTABLE)
73                 iflags |= FS_IMMUTABLE_FL;
74         if (flags & BTRFS_INODE_APPEND)
75                 iflags |= FS_APPEND_FL;
76         if (flags & BTRFS_INODE_NODUMP)
77                 iflags |= FS_NODUMP_FL;
78         if (flags & BTRFS_INODE_NOATIME)
79                 iflags |= FS_NOATIME_FL;
80         if (flags & BTRFS_INODE_DIRSYNC)
81                 iflags |= FS_DIRSYNC_FL;
82
83         return iflags;
84 }
85
86 /*
87  * Update inode->i_flags based on the btrfs internal flags.
88  */
89 void btrfs_update_iflags(struct inode *inode)
90 {
91         struct btrfs_inode *ip = BTRFS_I(inode);
92
93         inode->i_flags &= ~(S_SYNC|S_APPEND|S_IMMUTABLE|S_NOATIME|S_DIRSYNC);
94
95         if (ip->flags & BTRFS_INODE_SYNC)
96                 inode->i_flags |= S_SYNC;
97         if (ip->flags & BTRFS_INODE_IMMUTABLE)
98                 inode->i_flags |= S_IMMUTABLE;
99         if (ip->flags & BTRFS_INODE_APPEND)
100                 inode->i_flags |= S_APPEND;
101         if (ip->flags & BTRFS_INODE_NOATIME)
102                 inode->i_flags |= S_NOATIME;
103         if (ip->flags & BTRFS_INODE_DIRSYNC)
104                 inode->i_flags |= S_DIRSYNC;
105 }
106
107 /*
108  * Inherit flags from the parent inode.
109  *
110  * Unlike extN we don't have any flags we don't want to inherit currently.
111  */
112 void btrfs_inherit_iflags(struct inode *inode, struct inode *dir)
113 {
114         unsigned int flags;
115
116         if (!dir)
117                 return;
118
119         flags = BTRFS_I(dir)->flags;
120
121         if (S_ISREG(inode->i_mode))
122                 flags &= ~BTRFS_INODE_DIRSYNC;
123         else if (!S_ISDIR(inode->i_mode))
124                 flags &= (BTRFS_INODE_NODUMP | BTRFS_INODE_NOATIME);
125
126         BTRFS_I(inode)->flags = flags;
127         btrfs_update_iflags(inode);
128 }
129
130 static int btrfs_ioctl_getflags(struct file *file, void __user *arg)
131 {
132         struct btrfs_inode *ip = BTRFS_I(file->f_path.dentry->d_inode);
133         unsigned int flags = btrfs_flags_to_ioctl(ip->flags);
134
135         if (copy_to_user(arg, &flags, sizeof(flags)))
136                 return -EFAULT;
137         return 0;
138 }
139
140 static int btrfs_ioctl_setflags(struct file *file, void __user *arg)
141 {
142         struct inode *inode = file->f_path.dentry->d_inode;
143         struct btrfs_inode *ip = BTRFS_I(inode);
144         struct btrfs_root *root = ip->root;
145         struct btrfs_trans_handle *trans;
146         unsigned int flags, oldflags;
147         int ret;
148
149         if (copy_from_user(&flags, arg, sizeof(flags)))
150                 return -EFAULT;
151
152         if (flags & ~(FS_IMMUTABLE_FL | FS_APPEND_FL | \
153                       FS_NOATIME_FL | FS_NODUMP_FL | \
154                       FS_SYNC_FL | FS_DIRSYNC_FL))
155                 return -EOPNOTSUPP;
156
157         if (!is_owner_or_cap(inode))
158                 return -EACCES;
159
160         mutex_lock(&inode->i_mutex);
161
162         flags = btrfs_mask_flags(inode->i_mode, flags);
163         oldflags = btrfs_flags_to_ioctl(ip->flags);
164         if ((flags ^ oldflags) & (FS_APPEND_FL | FS_IMMUTABLE_FL)) {
165                 if (!capable(CAP_LINUX_IMMUTABLE)) {
166                         ret = -EPERM;
167                         goto out_unlock;
168                 }
169         }
170
171         ret = mnt_want_write(file->f_path.mnt);
172         if (ret)
173                 goto out_unlock;
174
175         if (flags & FS_SYNC_FL)
176                 ip->flags |= BTRFS_INODE_SYNC;
177         else
178                 ip->flags &= ~BTRFS_INODE_SYNC;
179         if (flags & FS_IMMUTABLE_FL)
180                 ip->flags |= BTRFS_INODE_IMMUTABLE;
181         else
182                 ip->flags &= ~BTRFS_INODE_IMMUTABLE;
183         if (flags & FS_APPEND_FL)
184                 ip->flags |= BTRFS_INODE_APPEND;
185         else
186                 ip->flags &= ~BTRFS_INODE_APPEND;
187         if (flags & FS_NODUMP_FL)
188                 ip->flags |= BTRFS_INODE_NODUMP;
189         else
190                 ip->flags &= ~BTRFS_INODE_NODUMP;
191         if (flags & FS_NOATIME_FL)
192                 ip->flags |= BTRFS_INODE_NOATIME;
193         else
194                 ip->flags &= ~BTRFS_INODE_NOATIME;
195         if (flags & FS_DIRSYNC_FL)
196                 ip->flags |= BTRFS_INODE_DIRSYNC;
197         else
198                 ip->flags &= ~BTRFS_INODE_DIRSYNC;
199
200
201         trans = btrfs_join_transaction(root, 1);
202         BUG_ON(!trans);
203
204         ret = btrfs_update_inode(trans, root, inode);
205         BUG_ON(ret);
206
207         btrfs_update_iflags(inode);
208         inode->i_ctime = CURRENT_TIME;
209         btrfs_end_transaction(trans, root);
210
211         mnt_drop_write(file->f_path.mnt);
212  out_unlock:
213         mutex_unlock(&inode->i_mutex);
214         return 0;
215 }
216
217 static int btrfs_ioctl_getversion(struct file *file, int __user *arg)
218 {
219         struct inode *inode = file->f_path.dentry->d_inode;
220
221         return put_user(inode->i_generation, arg);
222 }
223
224 static noinline int create_subvol(struct btrfs_root *root,
225                                   struct dentry *dentry,
226                                   char *name, int namelen)
227 {
228         struct btrfs_trans_handle *trans;
229         struct btrfs_key key;
230         struct btrfs_root_item root_item;
231         struct btrfs_inode_item *inode_item;
232         struct extent_buffer *leaf;
233         struct btrfs_root *new_root;
234         struct inode *dir = dentry->d_parent->d_inode;
235         int ret;
236         int err;
237         u64 objectid;
238         u64 new_dirid = BTRFS_FIRST_FREE_OBJECTID;
239         u64 index = 0;
240
241         /*
242          * 1 - inode item
243          * 2 - refs
244          * 1 - root item
245          * 2 - dir items
246          */
247         ret = btrfs_reserve_metadata_space(root, 6);
248         if (ret)
249                 return ret;
250
251         trans = btrfs_start_transaction(root, 1);
252         BUG_ON(!trans);
253
254         ret = btrfs_find_free_objectid(trans, root->fs_info->tree_root,
255                                        0, &objectid);
256         if (ret)
257                 goto fail;
258
259         leaf = btrfs_alloc_free_block(trans, root, root->leafsize,
260                                       0, objectid, NULL, 0, 0, 0);
261         if (IS_ERR(leaf)) {
262                 ret = PTR_ERR(leaf);
263                 goto fail;
264         }
265
266         memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header));
267         btrfs_set_header_bytenr(leaf, leaf->start);
268         btrfs_set_header_generation(leaf, trans->transid);
269         btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV);
270         btrfs_set_header_owner(leaf, objectid);
271
272         write_extent_buffer(leaf, root->fs_info->fsid,
273                             (unsigned long)btrfs_header_fsid(leaf),
274                             BTRFS_FSID_SIZE);
275         write_extent_buffer(leaf, root->fs_info->chunk_tree_uuid,
276                             (unsigned long)btrfs_header_chunk_tree_uuid(leaf),
277                             BTRFS_UUID_SIZE);
278         btrfs_mark_buffer_dirty(leaf);
279
280         inode_item = &root_item.inode;
281         memset(inode_item, 0, sizeof(*inode_item));
282         inode_item->generation = cpu_to_le64(1);
283         inode_item->size = cpu_to_le64(3);
284         inode_item->nlink = cpu_to_le32(1);
285         inode_item->nbytes = cpu_to_le64(root->leafsize);
286         inode_item->mode = cpu_to_le32(S_IFDIR | 0755);
287
288         btrfs_set_root_bytenr(&root_item, leaf->start);
289         btrfs_set_root_generation(&root_item, trans->transid);
290         btrfs_set_root_level(&root_item, 0);
291         btrfs_set_root_refs(&root_item, 1);
292         btrfs_set_root_used(&root_item, leaf->len);
293         btrfs_set_root_last_snapshot(&root_item, 0);
294
295         memset(&root_item.drop_progress, 0, sizeof(root_item.drop_progress));
296         root_item.drop_level = 0;
297
298         btrfs_tree_unlock(leaf);
299         free_extent_buffer(leaf);
300         leaf = NULL;
301
302         btrfs_set_root_dirid(&root_item, new_dirid);
303
304         key.objectid = objectid;
305         key.offset = 0;
306         btrfs_set_key_type(&key, BTRFS_ROOT_ITEM_KEY);
307         ret = btrfs_insert_root(trans, root->fs_info->tree_root, &key,
308                                 &root_item);
309         if (ret)
310                 goto fail;
311
312         key.offset = (u64)-1;
313         new_root = btrfs_read_fs_root_no_name(root->fs_info, &key);
314         BUG_ON(IS_ERR(new_root));
315
316         btrfs_record_root_in_trans(trans, new_root);
317
318         ret = btrfs_create_subvol_root(trans, new_root, new_dirid,
319                                        BTRFS_I(dir)->block_group);
320         /*
321          * insert the directory item
322          */
323         ret = btrfs_set_inode_index(dir, &index);
324         BUG_ON(ret);
325
326         ret = btrfs_insert_dir_item(trans, root,
327                                     name, namelen, dir->i_ino, &key,
328                                     BTRFS_FT_DIR, index);
329         if (ret)
330                 goto fail;
331
332         btrfs_i_size_write(dir, dir->i_size + namelen * 2);
333         ret = btrfs_update_inode(trans, root, dir);
334         BUG_ON(ret);
335
336         ret = btrfs_add_root_ref(trans, root->fs_info->tree_root,
337                                  objectid, root->root_key.objectid,
338                                  dir->i_ino, index, name, namelen);
339
340         BUG_ON(ret);
341
342         d_instantiate(dentry, btrfs_lookup_dentry(dir, dentry));
343 fail:
344         err = btrfs_commit_transaction(trans, root);
345         if (err && !ret)
346                 ret = err;
347
348         btrfs_unreserve_metadata_space(root, 6);
349         return ret;
350 }
351
352 static int create_snapshot(struct btrfs_root *root, struct dentry *dentry,
353                            char *name, int namelen)
354 {
355         struct inode *inode;
356         struct btrfs_pending_snapshot *pending_snapshot;
357         struct btrfs_trans_handle *trans;
358         int ret;
359
360         if (!root->ref_cows)
361                 return -EINVAL;
362
363         /*
364          * 1 - inode item
365          * 2 - refs
366          * 1 - root item
367          * 2 - dir items
368          */
369         ret = btrfs_reserve_metadata_space(root, 6);
370         if (ret)
371                 goto fail;
372
373         pending_snapshot = kzalloc(sizeof(*pending_snapshot), GFP_NOFS);
374         if (!pending_snapshot) {
375                 ret = -ENOMEM;
376                 btrfs_unreserve_metadata_space(root, 6);
377                 goto fail;
378         }
379         pending_snapshot->name = kmalloc(namelen + 1, GFP_NOFS);
380         if (!pending_snapshot->name) {
381                 ret = -ENOMEM;
382                 kfree(pending_snapshot);
383                 btrfs_unreserve_metadata_space(root, 6);
384                 goto fail;
385         }
386         memcpy(pending_snapshot->name, name, namelen);
387         pending_snapshot->name[namelen] = '\0';
388         pending_snapshot->dentry = dentry;
389         trans = btrfs_start_transaction(root, 1);
390         BUG_ON(!trans);
391         pending_snapshot->root = root;
392         list_add(&pending_snapshot->list,
393                  &trans->transaction->pending_snapshots);
394         ret = btrfs_commit_transaction(trans, root);
395         BUG_ON(ret);
396         btrfs_unreserve_metadata_space(root, 6);
397
398         inode = btrfs_lookup_dentry(dentry->d_parent->d_inode, dentry);
399         if (IS_ERR(inode)) {
400                 ret = PTR_ERR(inode);
401                 goto fail;
402         }
403         BUG_ON(!inode);
404         d_instantiate(dentry, inode);
405         ret = 0;
406 fail:
407         return ret;
408 }
409
410 /* copy of may_create in fs/namei.c() */
411 static inline int btrfs_may_create(struct inode *dir, struct dentry *child)
412 {
413         if (child->d_inode)
414                 return -EEXIST;
415         if (IS_DEADDIR(dir))
416                 return -ENOENT;
417         return inode_permission(dir, MAY_WRITE | MAY_EXEC);
418 }
419
420 /*
421  * Create a new subvolume below @parent.  This is largely modeled after
422  * sys_mkdirat and vfs_mkdir, but we only do a single component lookup
423  * inside this filesystem so it's quite a bit simpler.
424  */
425 static noinline int btrfs_mksubvol(struct path *parent,
426                                    char *name, int namelen,
427                                    struct btrfs_root *snap_src)
428 {
429         struct inode *dir  = parent->dentry->d_inode;
430         struct dentry *dentry;
431         int error;
432
433         mutex_lock_nested(&dir->i_mutex, I_MUTEX_PARENT);
434
435         dentry = lookup_one_len(name, parent->dentry, namelen);
436         error = PTR_ERR(dentry);
437         if (IS_ERR(dentry))
438                 goto out_unlock;
439
440         error = -EEXIST;
441         if (dentry->d_inode)
442                 goto out_dput;
443
444         error = mnt_want_write(parent->mnt);
445         if (error)
446                 goto out_dput;
447
448         error = btrfs_may_create(dir, dentry);
449         if (error)
450                 goto out_drop_write;
451
452         down_read(&BTRFS_I(dir)->root->fs_info->subvol_sem);
453
454         if (btrfs_root_refs(&BTRFS_I(dir)->root->root_item) == 0)
455                 goto out_up_read;
456
457         if (snap_src) {
458                 error = create_snapshot(snap_src, dentry,
459                                         name, namelen);
460         } else {
461                 error = create_subvol(BTRFS_I(dir)->root, dentry,
462                                       name, namelen);
463         }
464         if (!error)
465                 fsnotify_mkdir(dir, dentry);
466 out_up_read:
467         up_read(&BTRFS_I(dir)->root->fs_info->subvol_sem);
468 out_drop_write:
469         mnt_drop_write(parent->mnt);
470 out_dput:
471         dput(dentry);
472 out_unlock:
473         mutex_unlock(&dir->i_mutex);
474         return error;
475 }
476
477 static int btrfs_defrag_file(struct file *file)
478 {
479         struct inode *inode = fdentry(file)->d_inode;
480         struct btrfs_root *root = BTRFS_I(inode)->root;
481         struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
482         struct btrfs_ordered_extent *ordered;
483         struct page *page;
484         unsigned long last_index;
485         unsigned long ra_pages = root->fs_info->bdi.ra_pages;
486         unsigned long total_read = 0;
487         u64 page_start;
488         u64 page_end;
489         unsigned long i;
490         int ret;
491
492         ret = btrfs_check_data_free_space(root, inode, inode->i_size);
493         if (ret)
494                 return -ENOSPC;
495
496         mutex_lock(&inode->i_mutex);
497         last_index = inode->i_size >> PAGE_CACHE_SHIFT;
498         for (i = 0; i <= last_index; i++) {
499                 if (total_read % ra_pages == 0) {
500                         btrfs_force_ra(inode->i_mapping, &file->f_ra, file, i,
501                                        min(last_index, i + ra_pages - 1));
502                 }
503                 total_read++;
504 again:
505                 page = grab_cache_page(inode->i_mapping, i);
506                 if (!page)
507                         goto out_unlock;
508                 if (!PageUptodate(page)) {
509                         btrfs_readpage(NULL, page);
510                         lock_page(page);
511                         if (!PageUptodate(page)) {
512                                 unlock_page(page);
513                                 page_cache_release(page);
514                                 goto out_unlock;
515                         }
516                 }
517
518                 wait_on_page_writeback(page);
519
520                 page_start = (u64)page->index << PAGE_CACHE_SHIFT;
521                 page_end = page_start + PAGE_CACHE_SIZE - 1;
522                 lock_extent(io_tree, page_start, page_end, GFP_NOFS);
523
524                 ordered = btrfs_lookup_ordered_extent(inode, page_start);
525                 if (ordered) {
526                         unlock_extent(io_tree, page_start, page_end, GFP_NOFS);
527                         unlock_page(page);
528                         page_cache_release(page);
529                         btrfs_start_ordered_extent(inode, ordered, 1);
530                         btrfs_put_ordered_extent(ordered);
531                         goto again;
532                 }
533                 set_page_extent_mapped(page);
534
535                 /*
536                  * this makes sure page_mkwrite is called on the
537                  * page if it is dirtied again later
538                  */
539                 clear_page_dirty_for_io(page);
540
541                 btrfs_set_extent_delalloc(inode, page_start, page_end);
542                 set_page_dirty(page);
543                 unlock_extent(io_tree, page_start, page_end, GFP_NOFS);
544                 unlock_page(page);
545                 page_cache_release(page);
546                 balance_dirty_pages_ratelimited_nr(inode->i_mapping, 1);
547         }
548
549 out_unlock:
550         mutex_unlock(&inode->i_mutex);
551         return 0;
552 }
553
554 static noinline int btrfs_ioctl_resize(struct btrfs_root *root,
555                                         void __user *arg)
556 {
557         u64 new_size;
558         u64 old_size;
559         u64 devid = 1;
560         struct btrfs_ioctl_vol_args *vol_args;
561         struct btrfs_trans_handle *trans;
562         struct btrfs_device *device = NULL;
563         char *sizestr;
564         char *devstr = NULL;
565         int ret = 0;
566         int namelen;
567         int mod = 0;
568
569         if (root->fs_info->sb->s_flags & MS_RDONLY)
570                 return -EROFS;
571
572         if (!capable(CAP_SYS_ADMIN))
573                 return -EPERM;
574
575         vol_args = memdup_user(arg, sizeof(*vol_args));
576         if (IS_ERR(vol_args))
577                 return PTR_ERR(vol_args);
578
579         vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
580         namelen = strlen(vol_args->name);
581
582         mutex_lock(&root->fs_info->volume_mutex);
583         sizestr = vol_args->name;
584         devstr = strchr(sizestr, ':');
585         if (devstr) {
586                 char *end;
587                 sizestr = devstr + 1;
588                 *devstr = '\0';
589                 devstr = vol_args->name;
590                 devid = simple_strtoull(devstr, &end, 10);
591                 printk(KERN_INFO "resizing devid %llu\n",
592                        (unsigned long long)devid);
593         }
594         device = btrfs_find_device(root, devid, NULL, NULL);
595         if (!device) {
596                 printk(KERN_INFO "resizer unable to find device %llu\n",
597                        (unsigned long long)devid);
598                 ret = -EINVAL;
599                 goto out_unlock;
600         }
601         if (!strcmp(sizestr, "max"))
602                 new_size = device->bdev->bd_inode->i_size;
603         else {
604                 if (sizestr[0] == '-') {
605                         mod = -1;
606                         sizestr++;
607                 } else if (sizestr[0] == '+') {
608                         mod = 1;
609                         sizestr++;
610                 }
611                 new_size = btrfs_parse_size(sizestr);
612                 if (new_size == 0) {
613                         ret = -EINVAL;
614                         goto out_unlock;
615                 }
616         }
617
618         old_size = device->total_bytes;
619
620         if (mod < 0) {
621                 if (new_size > old_size) {
622                         ret = -EINVAL;
623                         goto out_unlock;
624                 }
625                 new_size = old_size - new_size;
626         } else if (mod > 0) {
627                 new_size = old_size + new_size;
628         }
629
630         if (new_size < 256 * 1024 * 1024) {
631                 ret = -EINVAL;
632                 goto out_unlock;
633         }
634         if (new_size > device->bdev->bd_inode->i_size) {
635                 ret = -EFBIG;
636                 goto out_unlock;
637         }
638
639         do_div(new_size, root->sectorsize);
640         new_size *= root->sectorsize;
641
642         printk(KERN_INFO "new size for %s is %llu\n",
643                 device->name, (unsigned long long)new_size);
644
645         if (new_size > old_size) {
646                 trans = btrfs_start_transaction(root, 1);
647                 ret = btrfs_grow_device(trans, device, new_size);
648                 btrfs_commit_transaction(trans, root);
649         } else {
650                 ret = btrfs_shrink_device(device, new_size);
651         }
652
653 out_unlock:
654         mutex_unlock(&root->fs_info->volume_mutex);
655         kfree(vol_args);
656         return ret;
657 }
658
659 static noinline int btrfs_ioctl_snap_create(struct file *file,
660                                             void __user *arg, int subvol)
661 {
662         struct btrfs_root *root = BTRFS_I(fdentry(file)->d_inode)->root;
663         struct btrfs_ioctl_vol_args *vol_args;
664         struct file *src_file;
665         int namelen;
666         int ret = 0;
667
668         if (root->fs_info->sb->s_flags & MS_RDONLY)
669                 return -EROFS;
670
671         vol_args = memdup_user(arg, sizeof(*vol_args));
672         if (IS_ERR(vol_args))
673                 return PTR_ERR(vol_args);
674
675         vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
676         namelen = strlen(vol_args->name);
677         if (strchr(vol_args->name, '/')) {
678                 ret = -EINVAL;
679                 goto out;
680         }
681
682         if (subvol) {
683                 ret = btrfs_mksubvol(&file->f_path, vol_args->name, namelen,
684                                      NULL);
685         } else {
686                 struct inode *src_inode;
687                 src_file = fget(vol_args->fd);
688                 if (!src_file) {
689                         ret = -EINVAL;
690                         goto out;
691                 }
692
693                 src_inode = src_file->f_path.dentry->d_inode;
694                 if (src_inode->i_sb != file->f_path.dentry->d_inode->i_sb) {
695                         printk(KERN_INFO "btrfs: Snapshot src from "
696                                "another FS\n");
697                         ret = -EINVAL;
698                         fput(src_file);
699                         goto out;
700                 }
701                 ret = btrfs_mksubvol(&file->f_path, vol_args->name, namelen,
702                                      BTRFS_I(src_inode)->root);
703                 fput(src_file);
704         }
705 out:
706         kfree(vol_args);
707         return ret;
708 }
709
710 /*
711  * helper to check if the subvolume references other subvolumes
712  */
713 static noinline int may_destroy_subvol(struct btrfs_root *root)
714 {
715         struct btrfs_path *path;
716         struct btrfs_key key;
717         int ret;
718
719         path = btrfs_alloc_path();
720         if (!path)
721                 return -ENOMEM;
722
723         key.objectid = root->root_key.objectid;
724         key.type = BTRFS_ROOT_REF_KEY;
725         key.offset = (u64)-1;
726
727         ret = btrfs_search_slot(NULL, root->fs_info->tree_root,
728                                 &key, path, 0, 0);
729         if (ret < 0)
730                 goto out;
731         BUG_ON(ret == 0);
732
733         ret = 0;
734         if (path->slots[0] > 0) {
735                 path->slots[0]--;
736                 btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
737                 if (key.objectid == root->root_key.objectid &&
738                     key.type == BTRFS_ROOT_REF_KEY)
739                         ret = -ENOTEMPTY;
740         }
741 out:
742         btrfs_free_path(path);
743         return ret;
744 }
745
746 static noinline int btrfs_ioctl_snap_destroy(struct file *file,
747                                              void __user *arg)
748 {
749         struct dentry *parent = fdentry(file);
750         struct dentry *dentry;
751         struct inode *dir = parent->d_inode;
752         struct inode *inode;
753         struct btrfs_root *root = BTRFS_I(dir)->root;
754         struct btrfs_root *dest = NULL;
755         struct btrfs_ioctl_vol_args *vol_args;
756         struct btrfs_trans_handle *trans;
757         int namelen;
758         int ret;
759         int err = 0;
760
761         if (!capable(CAP_SYS_ADMIN))
762                 return -EPERM;
763
764         vol_args = memdup_user(arg, sizeof(*vol_args));
765         if (IS_ERR(vol_args))
766                 return PTR_ERR(vol_args);
767
768         vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
769         namelen = strlen(vol_args->name);
770         if (strchr(vol_args->name, '/') ||
771             strncmp(vol_args->name, "..", namelen) == 0) {
772                 err = -EINVAL;
773                 goto out;
774         }
775
776         err = mnt_want_write(file->f_path.mnt);
777         if (err)
778                 goto out;
779
780         mutex_lock_nested(&dir->i_mutex, I_MUTEX_PARENT);
781         dentry = lookup_one_len(vol_args->name, parent, namelen);
782         if (IS_ERR(dentry)) {
783                 err = PTR_ERR(dentry);
784                 goto out_unlock_dir;
785         }
786
787         if (!dentry->d_inode) {
788                 err = -ENOENT;
789                 goto out_dput;
790         }
791
792         inode = dentry->d_inode;
793         if (inode->i_ino != BTRFS_FIRST_FREE_OBJECTID) {
794                 err = -EINVAL;
795                 goto out_dput;
796         }
797
798         dest = BTRFS_I(inode)->root;
799
800         mutex_lock(&inode->i_mutex);
801         err = d_invalidate(dentry);
802         if (err)
803                 goto out_unlock;
804
805         down_write(&root->fs_info->subvol_sem);
806
807         err = may_destroy_subvol(dest);
808         if (err)
809                 goto out_up_write;
810
811         trans = btrfs_start_transaction(root, 1);
812         ret = btrfs_unlink_subvol(trans, root, dir,
813                                 dest->root_key.objectid,
814                                 dentry->d_name.name,
815                                 dentry->d_name.len);
816         BUG_ON(ret);
817
818         btrfs_record_root_in_trans(trans, dest);
819
820         memset(&dest->root_item.drop_progress, 0,
821                 sizeof(dest->root_item.drop_progress));
822         dest->root_item.drop_level = 0;
823         btrfs_set_root_refs(&dest->root_item, 0);
824
825         ret = btrfs_insert_orphan_item(trans,
826                                 root->fs_info->tree_root,
827                                 dest->root_key.objectid);
828         BUG_ON(ret);
829
830         ret = btrfs_commit_transaction(trans, root);
831         BUG_ON(ret);
832         inode->i_flags |= S_DEAD;
833 out_up_write:
834         up_write(&root->fs_info->subvol_sem);
835 out_unlock:
836         mutex_unlock(&inode->i_mutex);
837         if (!err) {
838                 shrink_dcache_sb(root->fs_info->sb);
839                 btrfs_invalidate_inodes(dest);
840                 d_delete(dentry);
841         }
842 out_dput:
843         dput(dentry);
844 out_unlock_dir:
845         mutex_unlock(&dir->i_mutex);
846         mnt_drop_write(file->f_path.mnt);
847 out:
848         kfree(vol_args);
849         return err;
850 }
851
852 static int btrfs_ioctl_defrag(struct file *file)
853 {
854         struct inode *inode = fdentry(file)->d_inode;
855         struct btrfs_root *root = BTRFS_I(inode)->root;
856         int ret;
857
858         ret = mnt_want_write(file->f_path.mnt);
859         if (ret)
860                 return ret;
861
862         switch (inode->i_mode & S_IFMT) {
863         case S_IFDIR:
864                 if (!capable(CAP_SYS_ADMIN)) {
865                         ret = -EPERM;
866                         goto out;
867                 }
868                 btrfs_defrag_root(root, 0);
869                 btrfs_defrag_root(root->fs_info->extent_root, 0);
870                 break;
871         case S_IFREG:
872                 if (!(file->f_mode & FMODE_WRITE)) {
873                         ret = -EINVAL;
874                         goto out;
875                 }
876                 btrfs_defrag_file(file);
877                 break;
878         }
879 out:
880         mnt_drop_write(file->f_path.mnt);
881         return ret;
882 }
883
884 static long btrfs_ioctl_add_dev(struct btrfs_root *root, void __user *arg)
885 {
886         struct btrfs_ioctl_vol_args *vol_args;
887         int ret;
888
889         if (!capable(CAP_SYS_ADMIN))
890                 return -EPERM;
891
892         vol_args = memdup_user(arg, sizeof(*vol_args));
893         if (IS_ERR(vol_args))
894                 return PTR_ERR(vol_args);
895
896         vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
897         ret = btrfs_init_new_device(root, vol_args->name);
898
899         kfree(vol_args);
900         return ret;
901 }
902
903 static long btrfs_ioctl_rm_dev(struct btrfs_root *root, void __user *arg)
904 {
905         struct btrfs_ioctl_vol_args *vol_args;
906         int ret;
907
908         if (!capable(CAP_SYS_ADMIN))
909                 return -EPERM;
910
911         if (root->fs_info->sb->s_flags & MS_RDONLY)
912                 return -EROFS;
913
914         vol_args = memdup_user(arg, sizeof(*vol_args));
915         if (IS_ERR(vol_args))
916                 return PTR_ERR(vol_args);
917
918         vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
919         ret = btrfs_rm_device(root, vol_args->name);
920
921         kfree(vol_args);
922         return ret;
923 }
924
925 static noinline long btrfs_ioctl_clone(struct file *file, unsigned long srcfd,
926                                        u64 off, u64 olen, u64 destoff)
927 {
928         struct inode *inode = fdentry(file)->d_inode;
929         struct btrfs_root *root = BTRFS_I(inode)->root;
930         struct file *src_file;
931         struct inode *src;
932         struct btrfs_trans_handle *trans;
933         struct btrfs_path *path;
934         struct extent_buffer *leaf;
935         char *buf;
936         struct btrfs_key key;
937         u32 nritems;
938         int slot;
939         int ret;
940         u64 len = olen;
941         u64 bs = root->fs_info->sb->s_blocksize;
942         u64 hint_byte;
943
944         /*
945          * TODO:
946          * - split compressed inline extents.  annoying: we need to
947          *   decompress into destination's address_space (the file offset
948          *   may change, so source mapping won't do), then recompress (or
949          *   otherwise reinsert) a subrange.
950          * - allow ranges within the same file to be cloned (provided
951          *   they don't overlap)?
952          */
953
954         /* the destination must be opened for writing */
955         if (!(file->f_mode & FMODE_WRITE))
956                 return -EINVAL;
957
958         ret = mnt_want_write(file->f_path.mnt);
959         if (ret)
960                 return ret;
961
962         src_file = fget(srcfd);
963         if (!src_file) {
964                 ret = -EBADF;
965                 goto out_drop_write;
966         }
967         src = src_file->f_dentry->d_inode;
968
969         ret = -EINVAL;
970         if (src == inode)
971                 goto out_fput;
972
973         ret = -EISDIR;
974         if (S_ISDIR(src->i_mode) || S_ISDIR(inode->i_mode))
975                 goto out_fput;
976
977         ret = -EXDEV;
978         if (src->i_sb != inode->i_sb || BTRFS_I(src)->root != root)
979                 goto out_fput;
980
981         ret = -ENOMEM;
982         buf = vmalloc(btrfs_level_size(root, 0));
983         if (!buf)
984                 goto out_fput;
985
986         path = btrfs_alloc_path();
987         if (!path) {
988                 vfree(buf);
989                 goto out_fput;
990         }
991         path->reada = 2;
992
993         if (inode < src) {
994                 mutex_lock(&inode->i_mutex);
995                 mutex_lock(&src->i_mutex);
996         } else {
997                 mutex_lock(&src->i_mutex);
998                 mutex_lock(&inode->i_mutex);
999         }
1000
1001         /* determine range to clone */
1002         ret = -EINVAL;
1003         if (off >= src->i_size || off + len > src->i_size)
1004                 goto out_unlock;
1005         if (len == 0)
1006                 olen = len = src->i_size - off;
1007         /* if we extend to eof, continue to block boundary */
1008         if (off + len == src->i_size)
1009                 len = ((src->i_size + bs-1) & ~(bs-1))
1010                         - off;
1011
1012         /* verify the end result is block aligned */
1013         if ((off & (bs-1)) ||
1014             ((off + len) & (bs-1)))
1015                 goto out_unlock;
1016
1017         /* do any pending delalloc/csum calc on src, one way or
1018            another, and lock file content */
1019         while (1) {
1020                 struct btrfs_ordered_extent *ordered;
1021                 lock_extent(&BTRFS_I(src)->io_tree, off, off+len, GFP_NOFS);
1022                 ordered = btrfs_lookup_first_ordered_extent(inode, off+len);
1023                 if (BTRFS_I(src)->delalloc_bytes == 0 && !ordered)
1024                         break;
1025                 unlock_extent(&BTRFS_I(src)->io_tree, off, off+len, GFP_NOFS);
1026                 if (ordered)
1027                         btrfs_put_ordered_extent(ordered);
1028                 btrfs_wait_ordered_range(src, off, off+len);
1029         }
1030
1031         trans = btrfs_start_transaction(root, 1);
1032         BUG_ON(!trans);
1033
1034         /* punch hole in destination first */
1035         btrfs_drop_extents(trans, inode, off, off + len, &hint_byte, 1);
1036
1037         /* clone data */
1038         key.objectid = src->i_ino;
1039         key.type = BTRFS_EXTENT_DATA_KEY;
1040         key.offset = 0;
1041
1042         while (1) {
1043                 /*
1044                  * note the key will change type as we walk through the
1045                  * tree.
1046                  */
1047                 ret = btrfs_search_slot(trans, root, &key, path, 0, 0);
1048                 if (ret < 0)
1049                         goto out;
1050
1051                 nritems = btrfs_header_nritems(path->nodes[0]);
1052                 if (path->slots[0] >= nritems) {
1053                         ret = btrfs_next_leaf(root, path);
1054                         if (ret < 0)
1055                                 goto out;
1056                         if (ret > 0)
1057                                 break;
1058                         nritems = btrfs_header_nritems(path->nodes[0]);
1059                 }
1060                 leaf = path->nodes[0];
1061                 slot = path->slots[0];
1062
1063                 btrfs_item_key_to_cpu(leaf, &key, slot);
1064                 if (btrfs_key_type(&key) > BTRFS_EXTENT_DATA_KEY ||
1065                     key.objectid != src->i_ino)
1066                         break;
1067
1068                 if (btrfs_key_type(&key) == BTRFS_EXTENT_DATA_KEY) {
1069                         struct btrfs_file_extent_item *extent;
1070                         int type;
1071                         u32 size;
1072                         struct btrfs_key new_key;
1073                         u64 disko = 0, diskl = 0;
1074                         u64 datao = 0, datal = 0;
1075                         u8 comp;
1076
1077                         size = btrfs_item_size_nr(leaf, slot);
1078                         read_extent_buffer(leaf, buf,
1079                                            btrfs_item_ptr_offset(leaf, slot),
1080                                            size);
1081
1082                         extent = btrfs_item_ptr(leaf, slot,
1083                                                 struct btrfs_file_extent_item);
1084                         comp = btrfs_file_extent_compression(leaf, extent);
1085                         type = btrfs_file_extent_type(leaf, extent);
1086                         if (type == BTRFS_FILE_EXTENT_REG ||
1087                             type == BTRFS_FILE_EXTENT_PREALLOC) {
1088                                 disko = btrfs_file_extent_disk_bytenr(leaf,
1089                                                                       extent);
1090                                 diskl = btrfs_file_extent_disk_num_bytes(leaf,
1091                                                                  extent);
1092                                 datao = btrfs_file_extent_offset(leaf, extent);
1093                                 datal = btrfs_file_extent_num_bytes(leaf,
1094                                                                     extent);
1095                         } else if (type == BTRFS_FILE_EXTENT_INLINE) {
1096                                 /* take upper bound, may be compressed */
1097                                 datal = btrfs_file_extent_ram_bytes(leaf,
1098                                                                     extent);
1099                         }
1100                         btrfs_release_path(root, path);
1101
1102                         if (key.offset + datal < off ||
1103                             key.offset >= off+len)
1104                                 goto next;
1105
1106                         memcpy(&new_key, &key, sizeof(new_key));
1107                         new_key.objectid = inode->i_ino;
1108                         new_key.offset = key.offset + destoff - off;
1109
1110                         if (type == BTRFS_FILE_EXTENT_REG ||
1111                             type == BTRFS_FILE_EXTENT_PREALLOC) {
1112                                 ret = btrfs_insert_empty_item(trans, root, path,
1113                                                               &new_key, size);
1114                                 if (ret)
1115                                         goto out;
1116
1117                                 leaf = path->nodes[0];
1118                                 slot = path->slots[0];
1119                                 write_extent_buffer(leaf, buf,
1120                                             btrfs_item_ptr_offset(leaf, slot),
1121                                             size);
1122
1123                                 extent = btrfs_item_ptr(leaf, slot,
1124                                                 struct btrfs_file_extent_item);
1125
1126                                 if (off > key.offset) {
1127                                         datao += off - key.offset;
1128                                         datal -= off - key.offset;
1129                                 }
1130
1131                                 if (key.offset + datal > off + len)
1132                                         datal = off + len - key.offset;
1133
1134                                 /* disko == 0 means it's a hole */
1135                                 if (!disko)
1136                                         datao = 0;
1137
1138                                 btrfs_set_file_extent_offset(leaf, extent,
1139                                                              datao);
1140                                 btrfs_set_file_extent_num_bytes(leaf, extent,
1141                                                                 datal);
1142                                 if (disko) {
1143                                         inode_add_bytes(inode, datal);
1144                                         ret = btrfs_inc_extent_ref(trans, root,
1145                                                         disko, diskl, 0,
1146                                                         root->root_key.objectid,
1147                                                         inode->i_ino,
1148                                                         new_key.offset - datao);
1149                                         BUG_ON(ret);
1150                                 }
1151                         } else if (type == BTRFS_FILE_EXTENT_INLINE) {
1152                                 u64 skip = 0;
1153                                 u64 trim = 0;
1154                                 if (off > key.offset) {
1155                                         skip = off - key.offset;
1156                                         new_key.offset += skip;
1157                                 }
1158
1159                                 if (key.offset + datal > off+len)
1160                                         trim = key.offset + datal - (off+len);
1161
1162                                 if (comp && (skip || trim)) {
1163                                         ret = -EINVAL;
1164                                         goto out;
1165                                 }
1166                                 size -= skip + trim;
1167                                 datal -= skip + trim;
1168                                 ret = btrfs_insert_empty_item(trans, root, path,
1169                                                               &new_key, size);
1170                                 if (ret)
1171                                         goto out;
1172
1173                                 if (skip) {
1174                                         u32 start =
1175                                           btrfs_file_extent_calc_inline_size(0);
1176                                         memmove(buf+start, buf+start+skip,
1177                                                 datal);
1178                                 }
1179
1180                                 leaf = path->nodes[0];
1181                                 slot = path->slots[0];
1182                                 write_extent_buffer(leaf, buf,
1183                                             btrfs_item_ptr_offset(leaf, slot),
1184                                             size);
1185                                 inode_add_bytes(inode, datal);
1186                         }
1187
1188                         btrfs_mark_buffer_dirty(leaf);
1189                 }
1190
1191 next:
1192                 btrfs_release_path(root, path);
1193                 key.offset++;
1194         }
1195         ret = 0;
1196 out:
1197         btrfs_release_path(root, path);
1198         if (ret == 0) {
1199                 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
1200                 if (destoff + olen > inode->i_size)
1201                         btrfs_i_size_write(inode, destoff + olen);
1202                 BTRFS_I(inode)->flags = BTRFS_I(src)->flags;
1203                 ret = btrfs_update_inode(trans, root, inode);
1204         }
1205         btrfs_end_transaction(trans, root);
1206         unlock_extent(&BTRFS_I(src)->io_tree, off, off+len, GFP_NOFS);
1207         if (ret)
1208                 vmtruncate(inode, 0);
1209 out_unlock:
1210         mutex_unlock(&src->i_mutex);
1211         mutex_unlock(&inode->i_mutex);
1212         vfree(buf);
1213         btrfs_free_path(path);
1214 out_fput:
1215         fput(src_file);
1216 out_drop_write:
1217         mnt_drop_write(file->f_path.mnt);
1218         return ret;
1219 }
1220
1221 static long btrfs_ioctl_clone_range(struct file *file, void __user *argp)
1222 {
1223         struct btrfs_ioctl_clone_range_args args;
1224
1225         if (copy_from_user(&args, argp, sizeof(args)))
1226                 return -EFAULT;
1227         return btrfs_ioctl_clone(file, args.src_fd, args.src_offset,
1228                                  args.src_length, args.dest_offset);
1229 }
1230
1231 /*
1232  * there are many ways the trans_start and trans_end ioctls can lead
1233  * to deadlocks.  They should only be used by applications that
1234  * basically own the machine, and have a very in depth understanding
1235  * of all the possible deadlocks and enospc problems.
1236  */
1237 static long btrfs_ioctl_trans_start(struct file *file)
1238 {
1239         struct inode *inode = fdentry(file)->d_inode;
1240         struct btrfs_root *root = BTRFS_I(inode)->root;
1241         struct btrfs_trans_handle *trans;
1242         int ret;
1243
1244         ret = -EPERM;
1245         if (!capable(CAP_SYS_ADMIN))
1246                 goto out;
1247
1248         ret = -EINPROGRESS;
1249         if (file->private_data)
1250                 goto out;
1251
1252         ret = mnt_want_write(file->f_path.mnt);
1253         if (ret)
1254                 goto out;
1255
1256         mutex_lock(&root->fs_info->trans_mutex);
1257         root->fs_info->open_ioctl_trans++;
1258         mutex_unlock(&root->fs_info->trans_mutex);
1259
1260         ret = -ENOMEM;
1261         trans = btrfs_start_ioctl_transaction(root, 0);
1262         if (!trans)
1263                 goto out_drop;
1264
1265         file->private_data = trans;
1266         return 0;
1267
1268 out_drop:
1269         mutex_lock(&root->fs_info->trans_mutex);
1270         root->fs_info->open_ioctl_trans--;
1271         mutex_unlock(&root->fs_info->trans_mutex);
1272         mnt_drop_write(file->f_path.mnt);
1273 out:
1274         return ret;
1275 }
1276
1277 /*
1278  * there are many ways the trans_start and trans_end ioctls can lead
1279  * to deadlocks.  They should only be used by applications that
1280  * basically own the machine, and have a very in depth understanding
1281  * of all the possible deadlocks and enospc problems.
1282  */
1283 long btrfs_ioctl_trans_end(struct file *file)
1284 {
1285         struct inode *inode = fdentry(file)->d_inode;
1286         struct btrfs_root *root = BTRFS_I(inode)->root;
1287         struct btrfs_trans_handle *trans;
1288
1289         trans = file->private_data;
1290         if (!trans)
1291                 return -EINVAL;
1292         file->private_data = NULL;
1293
1294         btrfs_end_transaction(trans, root);
1295
1296         mutex_lock(&root->fs_info->trans_mutex);
1297         root->fs_info->open_ioctl_trans--;
1298         mutex_unlock(&root->fs_info->trans_mutex);
1299
1300         mnt_drop_write(file->f_path.mnt);
1301         return 0;
1302 }
1303
1304 long btrfs_ioctl(struct file *file, unsigned int
1305                 cmd, unsigned long arg)
1306 {
1307         struct btrfs_root *root = BTRFS_I(fdentry(file)->d_inode)->root;
1308         void __user *argp = (void __user *)arg;
1309
1310         switch (cmd) {
1311         case FS_IOC_GETFLAGS:
1312                 return btrfs_ioctl_getflags(file, argp);
1313         case FS_IOC_SETFLAGS:
1314                 return btrfs_ioctl_setflags(file, argp);
1315         case FS_IOC_GETVERSION:
1316                 return btrfs_ioctl_getversion(file, argp);
1317         case BTRFS_IOC_SNAP_CREATE:
1318                 return btrfs_ioctl_snap_create(file, argp, 0);
1319         case BTRFS_IOC_SUBVOL_CREATE:
1320                 return btrfs_ioctl_snap_create(file, argp, 1);
1321         case BTRFS_IOC_SNAP_DESTROY:
1322                 return btrfs_ioctl_snap_destroy(file, argp);
1323         case BTRFS_IOC_DEFRAG:
1324                 return btrfs_ioctl_defrag(file);
1325         case BTRFS_IOC_RESIZE:
1326                 return btrfs_ioctl_resize(root, argp);
1327         case BTRFS_IOC_ADD_DEV:
1328                 return btrfs_ioctl_add_dev(root, argp);
1329         case BTRFS_IOC_RM_DEV:
1330                 return btrfs_ioctl_rm_dev(root, argp);
1331         case BTRFS_IOC_BALANCE:
1332                 return btrfs_balance(root->fs_info->dev_root);
1333         case BTRFS_IOC_CLONE:
1334                 return btrfs_ioctl_clone(file, arg, 0, 0, 0);
1335         case BTRFS_IOC_CLONE_RANGE:
1336                 return btrfs_ioctl_clone_range(file, argp);
1337         case BTRFS_IOC_TRANS_START:
1338                 return btrfs_ioctl_trans_start(file);
1339         case BTRFS_IOC_TRANS_END:
1340                 return btrfs_ioctl_trans_end(file);
1341         case BTRFS_IOC_SYNC:
1342                 btrfs_sync_fs(file->f_dentry->d_sb, 1);
1343                 return 0;
1344         }
1345
1346         return -ENOTTY;
1347 }