~ [ source navigation ] ~ [ diff markup ] ~ [ identifier search ] ~

TOMOYO Linux Cross Reference
Linux/fs/btrfs/ioctl.c

Version: ~ [ linux-5.5-rc1 ] ~ [ linux-5.4.2 ] ~ [ linux-5.3.15 ] ~ [ linux-5.2.21 ] ~ [ linux-5.1.21 ] ~ [ linux-5.0.21 ] ~ [ linux-4.20.17 ] ~ [ linux-4.19.88 ] ~ [ linux-4.18.20 ] ~ [ linux-4.17.19 ] ~ [ linux-4.16.18 ] ~ [ linux-4.15.18 ] ~ [ linux-4.14.158 ] ~ [ linux-4.13.16 ] ~ [ linux-4.12.14 ] ~ [ linux-4.11.12 ] ~ [ linux-4.10.17 ] ~ [ linux-4.9.206 ] ~ [ linux-4.8.17 ] ~ [ linux-4.7.10 ] ~ [ linux-4.6.7 ] ~ [ linux-4.5.7 ] ~ [ linux-4.4.206 ] ~ [ linux-4.3.6 ] ~ [ linux-4.2.8 ] ~ [ linux-4.1.52 ] ~ [ linux-4.0.9 ] ~ [ linux-3.19.8 ] ~ [ linux-3.18.140 ] ~ [ linux-3.17.8 ] ~ [ linux-3.16.78 ] ~ [ linux-3.15.10 ] ~ [ linux-3.14.79 ] ~ [ linux-3.13.11 ] ~ [ linux-3.12.74 ] ~ [ linux-3.11.10 ] ~ [ linux-3.10.108 ] ~ [ linux-3.9.11 ] ~ [ linux-3.8.13 ] ~ [ linux-3.7.10 ] ~ [ linux-3.6.11 ] ~ [ linux-3.5.7 ] ~ [ linux-3.4.113 ] ~ [ linux-3.3.8 ] ~ [ linux-3.2.102 ] ~ [ linux-3.1.10 ] ~ [ linux-3.0.101 ] ~ [ linux-2.6.32.71 ] ~ [ linux-2.6.0 ] ~ [ linux-2.4.37.11 ] ~ [ unix-v6-master ] ~ [ ccs-tools-1.8.5 ] ~ [ policy-sample ] ~
Architecture: ~ [ i386 ] ~ [ alpha ] ~ [ m68k ] ~ [ mips ] ~ [ ppc ] ~ [ sparc ] ~ [ sparc64 ] ~

  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 <linux/slab.h>
 43 #include <linux/blkdev.h>
 44 #include <linux/uuid.h>
 45 #include <linux/btrfs.h>
 46 #include "compat.h"
 47 #include "ctree.h"
 48 #include "disk-io.h"
 49 #include "transaction.h"
 50 #include "btrfs_inode.h"
 51 #include "print-tree.h"
 52 #include "volumes.h"
 53 #include "locking.h"
 54 #include "inode-map.h"
 55 #include "backref.h"
 56 #include "rcu-string.h"
 57 #include "send.h"
 58 #include "dev-replace.h"
 59 
 60 /* Mask out flags that are inappropriate for the given type of inode. */
 61 static inline __u32 btrfs_mask_flags(umode_t mode, __u32 flags)
 62 {
 63         if (S_ISDIR(mode))
 64                 return flags;
 65         else if (S_ISREG(mode))
 66                 return flags & ~FS_DIRSYNC_FL;
 67         else
 68                 return flags & (FS_NODUMP_FL | FS_NOATIME_FL);
 69 }
 70 
 71 /*
 72  * Export inode flags to the format expected by the FS_IOC_GETFLAGS ioctl.
 73  */
 74 static unsigned int btrfs_flags_to_ioctl(unsigned int flags)
 75 {
 76         unsigned int iflags = 0;
 77 
 78         if (flags & BTRFS_INODE_SYNC)
 79                 iflags |= FS_SYNC_FL;
 80         if (flags & BTRFS_INODE_IMMUTABLE)
 81                 iflags |= FS_IMMUTABLE_FL;
 82         if (flags & BTRFS_INODE_APPEND)
 83                 iflags |= FS_APPEND_FL;
 84         if (flags & BTRFS_INODE_NODUMP)
 85                 iflags |= FS_NODUMP_FL;
 86         if (flags & BTRFS_INODE_NOATIME)
 87                 iflags |= FS_NOATIME_FL;
 88         if (flags & BTRFS_INODE_DIRSYNC)
 89                 iflags |= FS_DIRSYNC_FL;
 90         if (flags & BTRFS_INODE_NODATACOW)
 91                 iflags |= FS_NOCOW_FL;
 92 
 93         if ((flags & BTRFS_INODE_COMPRESS) && !(flags & BTRFS_INODE_NOCOMPRESS))
 94                 iflags |= FS_COMPR_FL;
 95         else if (flags & BTRFS_INODE_NOCOMPRESS)
 96                 iflags |= FS_NOCOMP_FL;
 97 
 98         return iflags;
 99 }
100 
101 /*
102  * Update inode->i_flags based on the btrfs internal flags.
103  */
104 void btrfs_update_iflags(struct inode *inode)
105 {
106         struct btrfs_inode *ip = BTRFS_I(inode);
107 
108         inode->i_flags &= ~(S_SYNC|S_APPEND|S_IMMUTABLE|S_NOATIME|S_DIRSYNC);
109 
110         if (ip->flags & BTRFS_INODE_SYNC)
111                 inode->i_flags |= S_SYNC;
112         if (ip->flags & BTRFS_INODE_IMMUTABLE)
113                 inode->i_flags |= S_IMMUTABLE;
114         if (ip->flags & BTRFS_INODE_APPEND)
115                 inode->i_flags |= S_APPEND;
116         if (ip->flags & BTRFS_INODE_NOATIME)
117                 inode->i_flags |= S_NOATIME;
118         if (ip->flags & BTRFS_INODE_DIRSYNC)
119                 inode->i_flags |= S_DIRSYNC;
120 }
121 
122 /*
123  * Inherit flags from the parent inode.
124  *
125  * Currently only the compression flags and the cow flags are inherited.
126  */
127 void btrfs_inherit_iflags(struct inode *inode, struct inode *dir)
128 {
129         unsigned int flags;
130 
131         if (!dir)
132                 return;
133 
134         flags = BTRFS_I(dir)->flags;
135 
136         if (flags & BTRFS_INODE_NOCOMPRESS) {
137                 BTRFS_I(inode)->flags &= ~BTRFS_INODE_COMPRESS;
138                 BTRFS_I(inode)->flags |= BTRFS_INODE_NOCOMPRESS;
139         } else if (flags & BTRFS_INODE_COMPRESS) {
140                 BTRFS_I(inode)->flags &= ~BTRFS_INODE_NOCOMPRESS;
141                 BTRFS_I(inode)->flags |= BTRFS_INODE_COMPRESS;
142         }
143 
144         if (flags & BTRFS_INODE_NODATACOW) {
145                 BTRFS_I(inode)->flags |= BTRFS_INODE_NODATACOW;
146                 if (S_ISREG(inode->i_mode))
147                         BTRFS_I(inode)->flags |= BTRFS_INODE_NODATASUM;
148         }
149 
150         btrfs_update_iflags(inode);
151 }
152 
153 static int btrfs_ioctl_getflags(struct file *file, void __user *arg)
154 {
155         struct btrfs_inode *ip = BTRFS_I(file_inode(file));
156         unsigned int flags = btrfs_flags_to_ioctl(ip->flags);
157 
158         if (copy_to_user(arg, &flags, sizeof(flags)))
159                 return -EFAULT;
160         return 0;
161 }
162 
163 static int check_flags(unsigned int flags)
164 {
165         if (flags & ~(FS_IMMUTABLE_FL | FS_APPEND_FL | \
166                       FS_NOATIME_FL | FS_NODUMP_FL | \
167                       FS_SYNC_FL | FS_DIRSYNC_FL | \
168                       FS_NOCOMP_FL | FS_COMPR_FL |
169                       FS_NOCOW_FL))
170                 return -EOPNOTSUPP;
171 
172         if ((flags & FS_NOCOMP_FL) && (flags & FS_COMPR_FL))
173                 return -EINVAL;
174 
175         return 0;
176 }
177 
178 static int btrfs_ioctl_setflags(struct file *file, void __user *arg)
179 {
180         struct inode *inode = file_inode(file);
181         struct btrfs_inode *ip = BTRFS_I(inode);
182         struct btrfs_root *root = ip->root;
183         struct btrfs_trans_handle *trans;
184         unsigned int flags, oldflags;
185         int ret;
186         u64 ip_oldflags;
187         unsigned int i_oldflags;
188         umode_t mode;
189 
190         if (btrfs_root_readonly(root))
191                 return -EROFS;
192 
193         if (copy_from_user(&flags, arg, sizeof(flags)))
194                 return -EFAULT;
195 
196         ret = check_flags(flags);
197         if (ret)
198                 return ret;
199 
200         if (!inode_owner_or_capable(inode))
201                 return -EACCES;
202 
203         ret = mnt_want_write_file(file);
204         if (ret)
205                 return ret;
206 
207         mutex_lock(&inode->i_mutex);
208 
209         ip_oldflags = ip->flags;
210         i_oldflags = inode->i_flags;
211         mode = inode->i_mode;
212 
213         flags = btrfs_mask_flags(inode->i_mode, flags);
214         oldflags = btrfs_flags_to_ioctl(ip->flags);
215         if ((flags ^ oldflags) & (FS_APPEND_FL | FS_IMMUTABLE_FL)) {
216                 if (!capable(CAP_LINUX_IMMUTABLE)) {
217                         ret = -EPERM;
218                         goto out_unlock;
219                 }
220         }
221 
222         if (flags & FS_SYNC_FL)
223                 ip->flags |= BTRFS_INODE_SYNC;
224         else
225                 ip->flags &= ~BTRFS_INODE_SYNC;
226         if (flags & FS_IMMUTABLE_FL)
227                 ip->flags |= BTRFS_INODE_IMMUTABLE;
228         else
229                 ip->flags &= ~BTRFS_INODE_IMMUTABLE;
230         if (flags & FS_APPEND_FL)
231                 ip->flags |= BTRFS_INODE_APPEND;
232         else
233                 ip->flags &= ~BTRFS_INODE_APPEND;
234         if (flags & FS_NODUMP_FL)
235                 ip->flags |= BTRFS_INODE_NODUMP;
236         else
237                 ip->flags &= ~BTRFS_INODE_NODUMP;
238         if (flags & FS_NOATIME_FL)
239                 ip->flags |= BTRFS_INODE_NOATIME;
240         else
241                 ip->flags &= ~BTRFS_INODE_NOATIME;
242         if (flags & FS_DIRSYNC_FL)
243                 ip->flags |= BTRFS_INODE_DIRSYNC;
244         else
245                 ip->flags &= ~BTRFS_INODE_DIRSYNC;
246         if (flags & FS_NOCOW_FL) {
247                 if (S_ISREG(mode)) {
248                         /*
249                          * It's safe to turn csums off here, no extents exist.
250                          * Otherwise we want the flag to reflect the real COW
251                          * status of the file and will not set it.
252                          */
253                         if (inode->i_size == 0)
254                                 ip->flags |= BTRFS_INODE_NODATACOW
255                                            | BTRFS_INODE_NODATASUM;
256                 } else {
257                         ip->flags |= BTRFS_INODE_NODATACOW;
258                 }
259         } else {
260                 /*
261                  * Revert back under same assuptions as above
262                  */
263                 if (S_ISREG(mode)) {
264                         if (inode->i_size == 0)
265                                 ip->flags &= ~(BTRFS_INODE_NODATACOW
266                                              | BTRFS_INODE_NODATASUM);
267                 } else {
268                         ip->flags &= ~BTRFS_INODE_NODATACOW;
269                 }
270         }
271 
272         /*
273          * The COMPRESS flag can only be changed by users, while the NOCOMPRESS
274          * flag may be changed automatically if compression code won't make
275          * things smaller.
276          */
277         if (flags & FS_NOCOMP_FL) {
278                 ip->flags &= ~BTRFS_INODE_COMPRESS;
279                 ip->flags |= BTRFS_INODE_NOCOMPRESS;
280         } else if (flags & FS_COMPR_FL) {
281                 ip->flags |= BTRFS_INODE_COMPRESS;
282                 ip->flags &= ~BTRFS_INODE_NOCOMPRESS;
283         } else {
284                 ip->flags &= ~(BTRFS_INODE_COMPRESS | BTRFS_INODE_NOCOMPRESS);
285         }
286 
287         trans = btrfs_start_transaction(root, 1);
288         if (IS_ERR(trans)) {
289                 ret = PTR_ERR(trans);
290                 goto out_drop;
291         }
292 
293         btrfs_update_iflags(inode);
294         inode_inc_iversion(inode);
295         inode->i_ctime = CURRENT_TIME;
296         ret = btrfs_update_inode(trans, root, inode);
297 
298         btrfs_end_transaction(trans, root);
299  out_drop:
300         if (ret) {
301                 ip->flags = ip_oldflags;
302                 inode->i_flags = i_oldflags;
303         }
304 
305  out_unlock:
306         mutex_unlock(&inode->i_mutex);
307         mnt_drop_write_file(file);
308         return ret;
309 }
310 
311 static int btrfs_ioctl_getversion(struct file *file, int __user *arg)
312 {
313         struct inode *inode = file_inode(file);
314 
315         return put_user(inode->i_generation, arg);
316 }
317 
318 static noinline int btrfs_ioctl_fitrim(struct file *file, void __user *arg)
319 {
320         struct btrfs_fs_info *fs_info = btrfs_sb(fdentry(file)->d_sb);
321         struct btrfs_device *device;
322         struct request_queue *q;
323         struct fstrim_range range;
324         u64 minlen = ULLONG_MAX;
325         u64 num_devices = 0;
326         u64 total_bytes = btrfs_super_total_bytes(fs_info->super_copy);
327         int ret;
328 
329         if (!capable(CAP_SYS_ADMIN))
330                 return -EPERM;
331 
332         rcu_read_lock();
333         list_for_each_entry_rcu(device, &fs_info->fs_devices->devices,
334                                 dev_list) {
335                 if (!device->bdev)
336                         continue;
337                 q = bdev_get_queue(device->bdev);
338                 if (blk_queue_discard(q)) {
339                         num_devices++;
340                         minlen = min((u64)q->limits.discard_granularity,
341                                      minlen);
342                 }
343         }
344         rcu_read_unlock();
345 
346         if (!num_devices)
347                 return -EOPNOTSUPP;
348         if (copy_from_user(&range, arg, sizeof(range)))
349                 return -EFAULT;
350         if (range.start > total_bytes ||
351             range.len < fs_info->sb->s_blocksize)
352                 return -EINVAL;
353 
354         range.len = min(range.len, total_bytes - range.start);
355         range.minlen = max(range.minlen, minlen);
356         ret = btrfs_trim_fs(fs_info->tree_root, &range);
357         if (ret < 0)
358                 return ret;
359 
360         if (copy_to_user(arg, &range, sizeof(range)))
361                 return -EFAULT;
362 
363         return 0;
364 }
365 
366 static noinline int create_subvol(struct inode *dir,
367                                   struct dentry *dentry,
368                                   char *name, int namelen,
369                                   u64 *async_transid,
370                                   struct btrfs_qgroup_inherit *inherit)
371 {
372         struct btrfs_trans_handle *trans;
373         struct btrfs_key key;
374         struct btrfs_root_item root_item;
375         struct btrfs_inode_item *inode_item;
376         struct extent_buffer *leaf;
377         struct btrfs_root *root = BTRFS_I(dir)->root;
378         struct btrfs_root *new_root;
379         struct btrfs_block_rsv block_rsv;
380         struct timespec cur_time = CURRENT_TIME;
381         int ret;
382         int err;
383         u64 objectid;
384         u64 new_dirid = BTRFS_FIRST_FREE_OBJECTID;
385         u64 index = 0;
386         u64 qgroup_reserved;
387         uuid_le new_uuid;
388 
389         ret = btrfs_find_free_objectid(root->fs_info->tree_root, &objectid);
390         if (ret)
391                 return ret;
392 
393         btrfs_init_block_rsv(&block_rsv, BTRFS_BLOCK_RSV_TEMP);
394         /*
395          * The same as the snapshot creation, please see the comment
396          * of create_snapshot().
397          */
398         ret = btrfs_subvolume_reserve_metadata(root, &block_rsv,
399                                                7, &qgroup_reserved);
400         if (ret)
401                 return ret;
402 
403         trans = btrfs_start_transaction(root, 0);
404         if (IS_ERR(trans)) {
405                 ret = PTR_ERR(trans);
406                 goto out;
407         }
408         trans->block_rsv = &block_rsv;
409         trans->bytes_reserved = block_rsv.size;
410 
411         ret = btrfs_qgroup_inherit(trans, root->fs_info, 0, objectid, inherit);
412         if (ret)
413                 goto fail;
414 
415         leaf = btrfs_alloc_free_block(trans, root, root->leafsize,
416                                       0, objectid, NULL, 0, 0, 0);
417         if (IS_ERR(leaf)) {
418                 ret = PTR_ERR(leaf);
419                 goto fail;
420         }
421 
422         memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header));
423         btrfs_set_header_bytenr(leaf, leaf->start);
424         btrfs_set_header_generation(leaf, trans->transid);
425         btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV);
426         btrfs_set_header_owner(leaf, objectid);
427 
428         write_extent_buffer(leaf, root->fs_info->fsid,
429                             (unsigned long)btrfs_header_fsid(leaf),
430                             BTRFS_FSID_SIZE);
431         write_extent_buffer(leaf, root->fs_info->chunk_tree_uuid,
432                             (unsigned long)btrfs_header_chunk_tree_uuid(leaf),
433                             BTRFS_UUID_SIZE);
434         btrfs_mark_buffer_dirty(leaf);
435 
436         memset(&root_item, 0, sizeof(root_item));
437 
438         inode_item = &root_item.inode;
439         inode_item->generation = cpu_to_le64(1);
440         inode_item->size = cpu_to_le64(3);
441         inode_item->nlink = cpu_to_le32(1);
442         inode_item->nbytes = cpu_to_le64(root->leafsize);
443         inode_item->mode = cpu_to_le32(S_IFDIR | 0755);
444 
445         root_item.flags = 0;
446         root_item.byte_limit = 0;
447         inode_item->flags = cpu_to_le64(BTRFS_INODE_ROOT_ITEM_INIT);
448 
449         btrfs_set_root_bytenr(&root_item, leaf->start);
450         btrfs_set_root_generation(&root_item, trans->transid);
451         btrfs_set_root_level(&root_item, 0);
452         btrfs_set_root_refs(&root_item, 1);
453         btrfs_set_root_used(&root_item, leaf->len);
454         btrfs_set_root_last_snapshot(&root_item, 0);
455 
456         btrfs_set_root_generation_v2(&root_item,
457                         btrfs_root_generation(&root_item));
458         uuid_le_gen(&new_uuid);
459         memcpy(root_item.uuid, new_uuid.b, BTRFS_UUID_SIZE);
460         root_item.otime.sec = cpu_to_le64(cur_time.tv_sec);
461         root_item.otime.nsec = cpu_to_le32(cur_time.tv_nsec);
462         root_item.ctime = root_item.otime;
463         btrfs_set_root_ctransid(&root_item, trans->transid);
464         btrfs_set_root_otransid(&root_item, trans->transid);
465 
466         btrfs_tree_unlock(leaf);
467         free_extent_buffer(leaf);
468         leaf = NULL;
469 
470         btrfs_set_root_dirid(&root_item, new_dirid);
471 
472         key.objectid = objectid;
473         key.offset = 0;
474         btrfs_set_key_type(&key, BTRFS_ROOT_ITEM_KEY);
475         ret = btrfs_insert_root(trans, root->fs_info->tree_root, &key,
476                                 &root_item);
477         if (ret)
478                 goto fail;
479 
480         key.offset = (u64)-1;
481         new_root = btrfs_read_fs_root_no_name(root->fs_info, &key);
482         if (IS_ERR(new_root)) {
483                 btrfs_abort_transaction(trans, root, PTR_ERR(new_root));
484                 ret = PTR_ERR(new_root);
485                 goto fail;
486         }
487 
488         btrfs_record_root_in_trans(trans, new_root);
489 
490         ret = btrfs_create_subvol_root(trans, new_root, new_dirid);
491         if (ret) {
492                 /* We potentially lose an unused inode item here */
493                 btrfs_abort_transaction(trans, root, ret);
494                 goto fail;
495         }
496 
497         /*
498          * insert the directory item
499          */
500         ret = btrfs_set_inode_index(dir, &index);
501         if (ret) {
502                 btrfs_abort_transaction(trans, root, ret);
503                 goto fail;
504         }
505 
506         ret = btrfs_insert_dir_item(trans, root,
507                                     name, namelen, dir, &key,
508                                     BTRFS_FT_DIR, index);
509         if (ret) {
510                 btrfs_abort_transaction(trans, root, ret);
511                 goto fail;
512         }
513 
514         btrfs_i_size_write(dir, dir->i_size + namelen * 2);
515         ret = btrfs_update_inode(trans, root, dir);
516         BUG_ON(ret);
517 
518         ret = btrfs_add_root_ref(trans, root->fs_info->tree_root,
519                                  objectid, root->root_key.objectid,
520                                  btrfs_ino(dir), index, name, namelen);
521 
522         BUG_ON(ret);
523 
524 fail:
525         trans->block_rsv = NULL;
526         trans->bytes_reserved = 0;
527         if (async_transid) {
528                 *async_transid = trans->transid;
529                 err = btrfs_commit_transaction_async(trans, root, 1);
530                 if (err)
531                         err = btrfs_commit_transaction(trans, root);
532         } else {
533                 err = btrfs_commit_transaction(trans, root);
534         }
535         if (err && !ret)
536                 ret = err;
537 
538         if (!ret)
539                 d_instantiate(dentry, btrfs_lookup_dentry(dir, dentry));
540 out:
541         btrfs_subvolume_release_metadata(root, &block_rsv, qgroup_reserved);
542         return ret;
543 }
544 
545 static int create_snapshot(struct btrfs_root *root, struct inode *dir,
546                            struct dentry *dentry, char *name, int namelen,
547                            u64 *async_transid, bool readonly,
548                            struct btrfs_qgroup_inherit *inherit)
549 {
550         struct inode *inode;
551         struct btrfs_pending_snapshot *pending_snapshot;
552         struct btrfs_trans_handle *trans;
553         int ret;
554 
555         if (!root->ref_cows)
556                 return -EINVAL;
557 
558         ret = btrfs_start_delalloc_inodes(root, 0);
559         if (ret)
560                 return ret;
561 
562         btrfs_wait_ordered_extents(root, 0);
563 
564         pending_snapshot = kzalloc(sizeof(*pending_snapshot), GFP_NOFS);
565         if (!pending_snapshot)
566                 return -ENOMEM;
567 
568         btrfs_init_block_rsv(&pending_snapshot->block_rsv,
569                              BTRFS_BLOCK_RSV_TEMP);
570         /*
571          * 1 - parent dir inode
572          * 2 - dir entries
573          * 1 - root item
574          * 2 - root ref/backref
575          * 1 - root of snapshot
576          */
577         ret = btrfs_subvolume_reserve_metadata(BTRFS_I(dir)->root,
578                                         &pending_snapshot->block_rsv, 7,
579                                         &pending_snapshot->qgroup_reserved);
580         if (ret)
581                 goto out;
582 
583         pending_snapshot->dentry = dentry;
584         pending_snapshot->root = root;
585         pending_snapshot->readonly = readonly;
586         pending_snapshot->dir = dir;
587         pending_snapshot->inherit = inherit;
588 
589         trans = btrfs_start_transaction(root, 0);
590         if (IS_ERR(trans)) {
591                 ret = PTR_ERR(trans);
592                 goto fail;
593         }
594 
595         spin_lock(&root->fs_info->trans_lock);
596         list_add(&pending_snapshot->list,
597                  &trans->transaction->pending_snapshots);
598         spin_unlock(&root->fs_info->trans_lock);
599         if (async_transid) {
600                 *async_transid = trans->transid;
601                 ret = btrfs_commit_transaction_async(trans,
602                                      root->fs_info->extent_root, 1);
603                 if (ret)
604                         ret = btrfs_commit_transaction(trans, root);
605         } else {
606                 ret = btrfs_commit_transaction(trans,
607                                                root->fs_info->extent_root);
608         }
609         if (ret)
610                 goto fail;
611 
612         ret = pending_snapshot->error;
613         if (ret)
614                 goto fail;
615 
616         ret = btrfs_orphan_cleanup(pending_snapshot->snap);
617         if (ret)
618                 goto fail;
619 
620         inode = btrfs_lookup_dentry(dentry->d_parent->d_inode, dentry);
621         if (IS_ERR(inode)) {
622                 ret = PTR_ERR(inode);
623                 goto fail;
624         }
625         BUG_ON(!inode);
626         d_instantiate(dentry, inode);
627         ret = 0;
628 fail:
629         btrfs_subvolume_release_metadata(BTRFS_I(dir)->root,
630                                          &pending_snapshot->block_rsv,
631                                          pending_snapshot->qgroup_reserved);
632 out:
633         kfree(pending_snapshot);
634         return ret;
635 }
636 
637 /*  copy of check_sticky in fs/namei.c()
638 * It's inline, so penalty for filesystems that don't use sticky bit is
639 * minimal.
640 */
641 static inline int btrfs_check_sticky(struct inode *dir, struct inode *inode)
642 {
643         kuid_t fsuid = current_fsuid();
644 
645         if (!(dir->i_mode & S_ISVTX))
646                 return 0;
647         if (uid_eq(inode->i_uid, fsuid))
648                 return 0;
649         if (uid_eq(dir->i_uid, fsuid))
650                 return 0;
651         return !capable(CAP_FOWNER);
652 }
653 
654 /*  copy of may_delete in fs/namei.c()
655  *      Check whether we can remove a link victim from directory dir, check
656  *  whether the type of victim is right.
657  *  1. We can't do it if dir is read-only (done in permission())
658  *  2. We should have write and exec permissions on dir
659  *  3. We can't remove anything from append-only dir
660  *  4. We can't do anything with immutable dir (done in permission())
661  *  5. If the sticky bit on dir is set we should either
662  *      a. be owner of dir, or
663  *      b. be owner of victim, or
664  *      c. have CAP_FOWNER capability
665  *  6. If the victim is append-only or immutable we can't do antyhing with
666  *     links pointing to it.
667  *  7. If we were asked to remove a directory and victim isn't one - ENOTDIR.
668  *  8. If we were asked to remove a non-directory and victim isn't one - EISDIR.
669  *  9. We can't remove a root or mountpoint.
670  * 10. We don't allow removal of NFS sillyrenamed files; it's handled by
671  *     nfs_async_unlink().
672  */
673 
674 static int btrfs_may_delete(struct inode *dir,struct dentry *victim,int isdir)
675 {
676         int error;
677 
678         if (!victim->d_inode)
679                 return -ENOENT;
680 
681         BUG_ON(victim->d_parent->d_inode != dir);
682         audit_inode_child(dir, victim, AUDIT_TYPE_CHILD_DELETE);
683 
684         error = inode_permission(dir, MAY_WRITE | MAY_EXEC);
685         if (error)
686                 return error;
687         if (IS_APPEND(dir))
688                 return -EPERM;
689         if (btrfs_check_sticky(dir, victim->d_inode)||
690                 IS_APPEND(victim->d_inode)||
691             IS_IMMUTABLE(victim->d_inode) || IS_SWAPFILE(victim->d_inode))
692                 return -EPERM;
693         if (isdir) {
694                 if (!S_ISDIR(victim->d_inode->i_mode))
695                         return -ENOTDIR;
696                 if (IS_ROOT(victim))
697                         return -EBUSY;
698         } else if (S_ISDIR(victim->d_inode->i_mode))
699                 return -EISDIR;
700         if (IS_DEADDIR(dir))
701                 return -ENOENT;
702         if (victim->d_flags & DCACHE_NFSFS_RENAMED)
703                 return -EBUSY;
704         return 0;
705 }
706 
707 /* copy of may_create in fs/namei.c() */
708 static inline int btrfs_may_create(struct inode *dir, struct dentry *child)
709 {
710         if (child->d_inode)
711                 return -EEXIST;
712         if (IS_DEADDIR(dir))
713                 return -ENOENT;
714         return inode_permission(dir, MAY_WRITE | MAY_EXEC);
715 }
716 
717 /*
718  * Create a new subvolume below @parent.  This is largely modeled after
719  * sys_mkdirat and vfs_mkdir, but we only do a single component lookup
720  * inside this filesystem so it's quite a bit simpler.
721  */
722 static noinline int btrfs_mksubvol(struct path *parent,
723                                    char *name, int namelen,
724                                    struct btrfs_root *snap_src,
725                                    u64 *async_transid, bool readonly,
726                                    struct btrfs_qgroup_inherit *inherit)
727 {
728         struct inode *dir  = parent->dentry->d_inode;
729         struct dentry *dentry;
730         int error;
731 
732         error = mutex_lock_killable_nested(&dir->i_mutex, I_MUTEX_PARENT);
733         if (error == -EINTR)
734                 return error;
735 
736         dentry = lookup_one_len(name, parent->dentry, namelen);
737         error = PTR_ERR(dentry);
738         if (IS_ERR(dentry))
739                 goto out_unlock;
740 
741         error = -EEXIST;
742         if (dentry->d_inode)
743                 goto out_dput;
744 
745         error = btrfs_may_create(dir, dentry);
746         if (error)
747                 goto out_dput;
748 
749         /*
750          * even if this name doesn't exist, we may get hash collisions.
751          * check for them now when we can safely fail
752          */
753         error = btrfs_check_dir_item_collision(BTRFS_I(dir)->root,
754                                                dir->i_ino, name,
755                                                namelen);
756         if (error)
757                 goto out_dput;
758 
759         down_read(&BTRFS_I(dir)->root->fs_info->subvol_sem);
760 
761         if (btrfs_root_refs(&BTRFS_I(dir)->root->root_item) == 0)
762                 goto out_up_read;
763 
764         if (snap_src) {
765                 error = create_snapshot(snap_src, dir, dentry, name, namelen,
766                                         async_transid, readonly, inherit);
767         } else {
768                 error = create_subvol(dir, dentry, name, namelen,
769                                       async_transid, inherit);
770         }
771         if (!error)
772                 fsnotify_mkdir(dir, dentry);
773 out_up_read:
774         up_read(&BTRFS_I(dir)->root->fs_info->subvol_sem);
775 out_dput:
776         dput(dentry);
777 out_unlock:
778         mutex_unlock(&dir->i_mutex);
779         return error;
780 }
781 
782 /*
783  * When we're defragging a range, we don't want to kick it off again
784  * if it is really just waiting for delalloc to send it down.
785  * If we find a nice big extent or delalloc range for the bytes in the
786  * file you want to defrag, we return 0 to let you know to skip this
787  * part of the file
788  */
789 static int check_defrag_in_cache(struct inode *inode, u64 offset, int thresh)
790 {
791         struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
792         struct extent_map *em = NULL;
793         struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
794         u64 end;
795 
796         read_lock(&em_tree->lock);
797         em = lookup_extent_mapping(em_tree, offset, PAGE_CACHE_SIZE);
798         read_unlock(&em_tree->lock);
799 
800         if (em) {
801                 end = extent_map_end(em);
802                 free_extent_map(em);
803                 if (end - offset > thresh)
804                         return 0;
805         }
806         /* if we already have a nice delalloc here, just stop */
807         thresh /= 2;
808         end = count_range_bits(io_tree, &offset, offset + thresh,
809                                thresh, EXTENT_DELALLOC, 1);
810         if (end >= thresh)
811                 return 0;
812         return 1;
813 }
814 
815 /*
816  * helper function to walk through a file and find extents
817  * newer than a specific transid, and smaller than thresh.
818  *
819  * This is used by the defragging code to find new and small
820  * extents
821  */
822 static int find_new_extents(struct btrfs_root *root,
823                             struct inode *inode, u64 newer_than,
824                             u64 *off, int thresh)
825 {
826         struct btrfs_path *path;
827         struct btrfs_key min_key;
828         struct btrfs_key max_key;
829         struct extent_buffer *leaf;
830         struct btrfs_file_extent_item *extent;
831         int type;
832         int ret;
833         u64 ino = btrfs_ino(inode);
834 
835         path = btrfs_alloc_path();
836         if (!path)
837                 return -ENOMEM;
838 
839         min_key.objectid = ino;
840         min_key.type = BTRFS_EXTENT_DATA_KEY;
841         min_key.offset = *off;
842 
843         max_key.objectid = ino;
844         max_key.type = (u8)-1;
845         max_key.offset = (u64)-1;
846 
847         path->keep_locks = 1;
848 
849         while(1) {
850                 ret = btrfs_search_forward(root, &min_key, &max_key,
851                                            path, newer_than);
852                 if (ret != 0)
853                         goto none;
854                 if (min_key.objectid != ino)
855                         goto none;
856                 if (min_key.type != BTRFS_EXTENT_DATA_KEY)
857                         goto none;
858 
859                 leaf = path->nodes[0];
860                 extent = btrfs_item_ptr(leaf, path->slots[0],
861                                         struct btrfs_file_extent_item);
862 
863                 type = btrfs_file_extent_type(leaf, extent);
864                 if (type == BTRFS_FILE_EXTENT_REG &&
865                     btrfs_file_extent_num_bytes(leaf, extent) < thresh &&
866                     check_defrag_in_cache(inode, min_key.offset, thresh)) {
867                         *off = min_key.offset;
868                         btrfs_free_path(path);
869                         return 0;
870                 }
871 
872                 if (min_key.offset == (u64)-1)
873                         goto none;
874 
875                 min_key.offset++;
876                 btrfs_release_path(path);
877         }
878 none:
879         btrfs_free_path(path);
880         return -ENOENT;
881 }
882 
883 static struct extent_map *defrag_lookup_extent(struct inode *inode, u64 start)
884 {
885         struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
886         struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
887         struct extent_map *em;
888         u64 len = PAGE_CACHE_SIZE;
889 
890         /*
891          * hopefully we have this extent in the tree already, try without
892          * the full extent lock
893          */
894         read_lock(&em_tree->lock);
895         em = lookup_extent_mapping(em_tree, start, len);
896         read_unlock(&em_tree->lock);
897 
898         if (!em) {
899                 /* get the big lock and read metadata off disk */
900                 lock_extent(io_tree, start, start + len - 1);
901                 em = btrfs_get_extent(inode, NULL, 0, start, len, 0);
902                 unlock_extent(io_tree, start, start + len - 1);
903 
904                 if (IS_ERR(em))
905                         return NULL;
906         }
907 
908         return em;
909 }
910 
911 static bool defrag_check_next_extent(struct inode *inode, struct extent_map *em)
912 {
913         struct extent_map *next;
914         bool ret = true;
915 
916         /* this is the last extent */
917         if (em->start + em->len >= i_size_read(inode))
918                 return false;
919 
920         next = defrag_lookup_extent(inode, em->start + em->len);
921         if (!next || next->block_start >= EXTENT_MAP_LAST_BYTE)
922                 ret = false;
923 
924         free_extent_map(next);
925         return ret;
926 }
927 
928 static int should_defrag_range(struct inode *inode, u64 start, int thresh,
929                                u64 *last_len, u64 *skip, u64 *defrag_end,
930                                int compress)
931 {
932         struct extent_map *em;
933         int ret = 1;
934         bool next_mergeable = true;
935 
936         /*
937          * make sure that once we start defragging an extent, we keep on
938          * defragging it
939          */
940         if (start < *defrag_end)
941                 return 1;
942 
943         *skip = 0;
944 
945         em = defrag_lookup_extent(inode, start);
946         if (!em)
947                 return 0;
948 
949         /* this will cover holes, and inline extents */
950         if (em->block_start >= EXTENT_MAP_LAST_BYTE) {
951                 ret = 0;
952                 goto out;
953         }
954 
955         next_mergeable = defrag_check_next_extent(inode, em);
956 
957         /*
958          * we hit a real extent, if it is big or the next extent is not a
959          * real extent, don't bother defragging it
960          */
961         if (!compress && (*last_len == 0 || *last_len >= thresh) &&
962             (em->len >= thresh || !next_mergeable))
963                 ret = 0;
964 out:
965         /*
966          * last_len ends up being a counter of how many bytes we've defragged.
967          * every time we choose not to defrag an extent, we reset *last_len
968          * so that the next tiny extent will force a defrag.
969          *
970          * The end result of this is that tiny extents before a single big
971          * extent will force at least part of that big extent to be defragged.
972          */
973         if (ret) {
974                 *defrag_end = extent_map_end(em);
975         } else {
976                 *last_len = 0;
977                 *skip = extent_map_end(em);
978                 *defrag_end = 0;
979         }
980 
981         free_extent_map(em);
982         return ret;
983 }
984 
985 /*
986  * it doesn't do much good to defrag one or two pages
987  * at a time.  This pulls in a nice chunk of pages
988  * to COW and defrag.
989  *
990  * It also makes sure the delalloc code has enough
991  * dirty data to avoid making new small extents as part
992  * of the defrag
993  *
994  * It's a good idea to start RA on this range
995  * before calling this.
996  */
997 static int cluster_pages_for_defrag(struct inode *inode,
998                                     struct page **pages,
999                                     unsigned long start_index,
1000                                     int num_pages)
1001 {
1002         unsigned long file_end;
1003         u64 isize = i_size_read(inode);
1004         u64 page_start;
1005         u64 page_end;
1006         u64 page_cnt;
1007         int ret;
1008         int i;
1009         int i_done;
1010         struct btrfs_ordered_extent *ordered;
1011         struct extent_state *cached_state = NULL;
1012         struct extent_io_tree *tree;
1013         gfp_t mask = btrfs_alloc_write_mask(inode->i_mapping);
1014 
1015         file_end = (isize - 1) >> PAGE_CACHE_SHIFT;
1016         if (!isize || start_index > file_end)
1017                 return 0;
1018 
1019         page_cnt = min_t(u64, (u64)num_pages, (u64)file_end - start_index + 1);
1020 
1021         ret = btrfs_delalloc_reserve_space(inode,
1022                                            page_cnt << PAGE_CACHE_SHIFT);
1023         if (ret)
1024                 return ret;
1025         i_done = 0;
1026         tree = &BTRFS_I(inode)->io_tree;
1027 
1028         /* step one, lock all the pages */
1029         for (i = 0; i < page_cnt; i++) {
1030                 struct page *page;
1031 again:
1032                 page = find_or_create_page(inode->i_mapping,
1033                                            start_index + i, mask);
1034                 if (!page)
1035                         break;
1036 
1037                 page_start = page_offset(page);
1038                 page_end = page_start + PAGE_CACHE_SIZE - 1;
1039                 while (1) {
1040                         lock_extent(tree, page_start, page_end);
1041                         ordered = btrfs_lookup_ordered_extent(inode,
1042                                                               page_start);
1043                         unlock_extent(tree, page_start, page_end);
1044                         if (!ordered)
1045                                 break;
1046 
1047                         unlock_page(page);
1048                         btrfs_start_ordered_extent(inode, ordered, 1);
1049                         btrfs_put_ordered_extent(ordered);
1050                         lock_page(page);
1051                         /*
1052                          * we unlocked the page above, so we need check if
1053                          * it was released or not.
1054                          */
1055                         if (page->mapping != inode->i_mapping) {
1056                                 unlock_page(page);
1057                                 page_cache_release(page);
1058                                 goto again;
1059                         }
1060                 }
1061 
1062                 if (!PageUptodate(page)) {
1063                         btrfs_readpage(NULL, page);
1064                         lock_page(page);
1065                         if (!PageUptodate(page)) {
1066                                 unlock_page(page);
1067                                 page_cache_release(page);
1068                                 ret = -EIO;
1069                                 break;
1070                         }
1071                 }
1072 
1073                 if (page->mapping != inode->i_mapping) {
1074                         unlock_page(page);
1075                         page_cache_release(page);
1076                         goto again;
1077                 }
1078 
1079                 pages[i] = page;
1080                 i_done++;
1081         }
1082         if (!i_done || ret)
1083                 goto out;
1084 
1085         if (!(inode->i_sb->s_flags & MS_ACTIVE))
1086                 goto out;
1087 
1088         /*
1089          * so now we have a nice long stream of locked
1090          * and up to date pages, lets wait on them
1091          */
1092         for (i = 0; i < i_done; i++)
1093                 wait_on_page_writeback(pages[i]);
1094 
1095         page_start = page_offset(pages[0]);
1096         page_end = page_offset(pages[i_done - 1]) + PAGE_CACHE_SIZE;
1097 
1098         lock_extent_bits(&BTRFS_I(inode)->io_tree,
1099                          page_start, page_end - 1, 0, &cached_state);
1100         clear_extent_bit(&BTRFS_I(inode)->io_tree, page_start,
1101                           page_end - 1, EXTENT_DIRTY | EXTENT_DELALLOC |
1102                           EXTENT_DO_ACCOUNTING | EXTENT_DEFRAG, 0, 0,
1103                           &cached_state, GFP_NOFS);
1104 
1105         if (i_done != page_cnt) {
1106                 spin_lock(&BTRFS_I(inode)->lock);
1107                 BTRFS_I(inode)->outstanding_extents++;
1108                 spin_unlock(&BTRFS_I(inode)->lock);
1109                 btrfs_delalloc_release_space(inode,
1110                                      (page_cnt - i_done) << PAGE_CACHE_SHIFT);
1111         }
1112 
1113 
1114         set_extent_defrag(&BTRFS_I(inode)->io_tree, page_start, page_end - 1,
1115                           &cached_state, GFP_NOFS);
1116 
1117         unlock_extent_cached(&BTRFS_I(inode)->io_tree,
1118                              page_start, page_end - 1, &cached_state,
1119                              GFP_NOFS);
1120 
1121         for (i = 0; i < i_done; i++) {
1122                 clear_page_dirty_for_io(pages[i]);
1123                 ClearPageChecked(pages[i]);
1124                 set_page_extent_mapped(pages[i]);
1125                 set_page_dirty(pages[i]);
1126                 unlock_page(pages[i]);
1127                 page_cache_release(pages[i]);
1128         }
1129         return i_done;
1130 out:
1131         for (i = 0; i < i_done; i++) {
1132                 unlock_page(pages[i]);
1133                 page_cache_release(pages[i]);
1134         }
1135         btrfs_delalloc_release_space(inode, page_cnt << PAGE_CACHE_SHIFT);
1136         return ret;
1137 
1138 }
1139 
1140 int btrfs_defrag_file(struct inode *inode, struct file *file,
1141                       struct btrfs_ioctl_defrag_range_args *range,
1142                       u64 newer_than, unsigned long max_to_defrag)
1143 {
1144         struct btrfs_root *root = BTRFS_I(inode)->root;
1145         struct file_ra_state *ra = NULL;
1146         unsigned long last_index;
1147         u64 isize = i_size_read(inode);
1148         u64 last_len = 0;
1149         u64 skip = 0;
1150         u64 defrag_end = 0;
1151         u64 newer_off = range->start;
1152         unsigned long i;
1153         unsigned long ra_index = 0;
1154         int ret;
1155         int defrag_count = 0;
1156         int compress_type = BTRFS_COMPRESS_ZLIB;
1157         int extent_thresh = range->extent_thresh;
1158         int max_cluster = (256 * 1024) >> PAGE_CACHE_SHIFT;
1159         int cluster = max_cluster;
1160         u64 new_align = ~((u64)128 * 1024 - 1);
1161         struct page **pages = NULL;
1162 
1163         if (isize == 0)
1164                 return 0;
1165 
1166         if (range->start >= isize)
1167                 return -EINVAL;
1168 
1169         if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS) {
1170                 if (range->compress_type > BTRFS_COMPRESS_TYPES)
1171                         return -EINVAL;
1172                 if (range->compress_type)
1173                         compress_type = range->compress_type;
1174         }
1175 
1176         if (extent_thresh == 0)
1177                 extent_thresh = 256 * 1024;
1178 
1179         /*
1180          * if we were not given a file, allocate a readahead
1181          * context
1182          */
1183         if (!file) {
1184                 ra = kzalloc(sizeof(*ra), GFP_NOFS);
1185                 if (!ra)
1186                         return -ENOMEM;
1187                 file_ra_state_init(ra, inode->i_mapping);
1188         } else {
1189                 ra = &file->f_ra;
1190         }
1191 
1192         pages = kmalloc(sizeof(struct page *) * max_cluster,
1193                         GFP_NOFS);
1194         if (!pages) {
1195                 ret = -ENOMEM;
1196                 goto out_ra;
1197         }
1198 
1199         /* find the last page to defrag */
1200         if (range->start + range->len > range->start) {
1201                 last_index = min_t(u64, isize - 1,
1202                          range->start + range->len - 1) >> PAGE_CACHE_SHIFT;
1203         } else {
1204                 last_index = (isize - 1) >> PAGE_CACHE_SHIFT;
1205         }
1206 
1207         if (newer_than) {
1208                 ret = find_new_extents(root, inode, newer_than,
1209                                        &newer_off, 64 * 1024);
1210                 if (!ret) {
1211                         range->start = newer_off;
1212                         /*
1213                          * we always align our defrag to help keep
1214                          * the extents in the file evenly spaced
1215                          */
1216                         i = (newer_off & new_align) >> PAGE_CACHE_SHIFT;
1217                 } else
1218                         goto out_ra;
1219         } else {
1220                 i = range->start >> PAGE_CACHE_SHIFT;
1221         }
1222         if (!max_to_defrag)
1223                 max_to_defrag = last_index + 1;
1224 
1225         /*
1226          * make writeback starts from i, so the defrag range can be
1227          * written sequentially.
1228          */
1229         if (i < inode->i_mapping->writeback_index)
1230                 inode->i_mapping->writeback_index = i;
1231 
1232         while (i <= last_index && defrag_count < max_to_defrag &&
1233                (i < (i_size_read(inode) + PAGE_CACHE_SIZE - 1) >>
1234                 PAGE_CACHE_SHIFT)) {
1235                 /*
1236                  * make sure we stop running if someone unmounts
1237                  * the FS
1238                  */
1239                 if (!(inode->i_sb->s_flags & MS_ACTIVE))
1240                         break;
1241 
1242                 if (btrfs_defrag_cancelled(root->fs_info)) {
1243                         printk(KERN_DEBUG "btrfs: defrag_file cancelled\n");
1244                         ret = -EAGAIN;
1245                         break;
1246                 }
1247 
1248                 if (!should_defrag_range(inode, (u64)i << PAGE_CACHE_SHIFT,
1249                                          extent_thresh, &last_len, &skip,
1250                                          &defrag_end, range->flags &
1251                                          BTRFS_DEFRAG_RANGE_COMPRESS)) {
1252                         unsigned long next;
1253                         /*
1254                          * the should_defrag function tells us how much to skip
1255                          * bump our counter by the suggested amount
1256                          */
1257                         next = (skip + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT;
1258                         i = max(i + 1, next);
1259                         continue;
1260                 }
1261 
1262                 if (!newer_than) {
1263                         cluster = (PAGE_CACHE_ALIGN(defrag_end) >>
1264                                    PAGE_CACHE_SHIFT) - i;
1265                         cluster = min(cluster, max_cluster);
1266                 } else {
1267                         cluster = max_cluster;
1268                 }
1269 
1270                 if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)
1271                         BTRFS_I(inode)->force_compress = compress_type;
1272 
1273                 if (i + cluster > ra_index) {
1274                         ra_index = max(i, ra_index);
1275                         btrfs_force_ra(inode->i_mapping, ra, file, ra_index,
1276                                        cluster);
1277                         ra_index += max_cluster;
1278                 }
1279 
1280                 mutex_lock(&inode->i_mutex);
1281                 ret = cluster_pages_for_defrag(inode, pages, i, cluster);
1282                 if (ret < 0) {
1283                         mutex_unlock(&inode->i_mutex);
1284                         goto out_ra;
1285                 }
1286 
1287                 defrag_count += ret;
1288                 balance_dirty_pages_ratelimited(inode->i_mapping);
1289                 mutex_unlock(&inode->i_mutex);
1290 
1291                 if (newer_than) {
1292                         if (newer_off == (u64)-1)
1293                                 break;
1294 
1295                         if (ret > 0)
1296                                 i += ret;
1297 
1298                         newer_off = max(newer_off + 1,
1299                                         (u64)i << PAGE_CACHE_SHIFT);
1300 
1301                         ret = find_new_extents(root, inode,
1302                                                newer_than, &newer_off,
1303                                                64 * 1024);
1304                         if (!ret) {
1305                                 range->start = newer_off;
1306                                 i = (newer_off & new_align) >> PAGE_CACHE_SHIFT;
1307                         } else {
1308                                 break;
1309                         }
1310                 } else {
1311                         if (ret > 0) {
1312                                 i += ret;
1313                                 last_len += ret << PAGE_CACHE_SHIFT;
1314                         } else {
1315                                 i++;
1316                                 last_len = 0;
1317                         }
1318                 }
1319         }
1320 
1321         if ((range->flags & BTRFS_DEFRAG_RANGE_START_IO))
1322                 filemap_flush(inode->i_mapping);
1323 
1324         if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
1325                 /* the filemap_flush will queue IO into the worker threads, but
1326                  * we have to make sure the IO is actually started and that
1327                  * ordered extents get created before we return
1328                  */
1329                 atomic_inc(&root->fs_info->async_submit_draining);
1330                 while (atomic_read(&root->fs_info->nr_async_submits) ||
1331                       atomic_read(&root->fs_info->async_delalloc_pages)) {
1332                         wait_event(root->fs_info->async_submit_wait,
1333                            (atomic_read(&root->fs_info->nr_async_submits) == 0 &&
1334                             atomic_read(&root->fs_info->async_delalloc_pages) == 0));
1335                 }
1336                 atomic_dec(&root->fs_info->async_submit_draining);
1337 
1338                 mutex_lock(&inode->i_mutex);
1339                 BTRFS_I(inode)->force_compress = BTRFS_COMPRESS_NONE;
1340                 mutex_unlock(&inode->i_mutex);
1341         }
1342 
1343         if (range->compress_type == BTRFS_COMPRESS_LZO) {
1344                 btrfs_set_fs_incompat(root->fs_info, COMPRESS_LZO);
1345         }
1346 
1347         ret = defrag_count;
1348 
1349 out_ra:
1350         if (!file)
1351                 kfree(ra);
1352         kfree(pages);
1353         return ret;
1354 }
1355 
1356 static noinline int btrfs_ioctl_resize(struct file *file,
1357                                         void __user *arg)
1358 {
1359         u64 new_size;
1360         u64 old_size;
1361         u64 devid = 1;
1362         struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
1363         struct btrfs_ioctl_vol_args *vol_args;
1364         struct btrfs_trans_handle *trans;
1365         struct btrfs_device *device = NULL;
1366         char *sizestr;
1367         char *devstr = NULL;
1368         int ret = 0;
1369         int mod = 0;
1370 
1371         if (!capable(CAP_SYS_ADMIN))
1372                 return -EPERM;
1373 
1374         ret = mnt_want_write_file(file);
1375         if (ret)
1376                 return ret;
1377 
1378         if (atomic_xchg(&root->fs_info->mutually_exclusive_operation_running,
1379                         1)) {
1380                 pr_info("btrfs: dev add/delete/balance/replace/resize operation in progress\n");
1381                 mnt_drop_write_file(file);
1382                 return -EINVAL;
1383         }
1384 
1385         mutex_lock(&root->fs_info->volume_mutex);
1386         vol_args = memdup_user(arg, sizeof(*vol_args));
1387         if (IS_ERR(vol_args)) {
1388                 ret = PTR_ERR(vol_args);
1389                 goto out;
1390         }
1391 
1392         vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1393 
1394         sizestr = vol_args->name;
1395         devstr = strchr(sizestr, ':');
1396         if (devstr) {
1397                 char *end;
1398                 sizestr = devstr + 1;
1399                 *devstr = '\0';
1400                 devstr = vol_args->name;
1401                 devid = simple_strtoull(devstr, &end, 10);
1402                 if (!devid) {
1403                         ret = -EINVAL;
1404                         goto out_free;
1405                 }
1406                 printk(KERN_INFO "btrfs: resizing devid %llu\n",
1407                        (unsigned long long)devid);
1408         }
1409 
1410         device = btrfs_find_device(root->fs_info, devid, NULL, NULL);
1411         if (!device) {
1412                 printk(KERN_INFO "btrfs: resizer unable to find device %llu\n",
1413                        (unsigned long long)devid);
1414                 ret = -ENODEV;
1415                 goto out_free;
1416         }
1417 
1418         if (!device->writeable) {
1419                 printk(KERN_INFO "btrfs: resizer unable to apply on "
1420                        "readonly device %llu\n",
1421                        (unsigned long long)devid);
1422                 ret = -EPERM;
1423                 goto out_free;
1424         }
1425 
1426         if (!strcmp(sizestr, "max"))
1427                 new_size = device->bdev->bd_inode->i_size;
1428         else {
1429                 if (sizestr[0] == '-') {
1430                         mod = -1;
1431                         sizestr++;
1432                 } else if (sizestr[0] == '+') {
1433                         mod = 1;
1434                         sizestr++;
1435                 }
1436                 new_size = memparse(sizestr, NULL);
1437                 if (new_size == 0) {
1438                         ret = -EINVAL;
1439                         goto out_free;
1440                 }
1441         }
1442 
1443         if (device->is_tgtdev_for_dev_replace) {
1444                 ret = -EPERM;
1445                 goto out_free;
1446         }
1447 
1448         old_size = device->total_bytes;
1449 
1450         if (mod < 0) {
1451                 if (new_size > old_size) {
1452                         ret = -EINVAL;
1453                         goto out_free;
1454                 }
1455                 new_size = old_size - new_size;
1456         } else if (mod > 0) {
1457                 new_size = old_size + new_size;
1458         }
1459 
1460         if (new_size < 256 * 1024 * 1024) {
1461                 ret = -EINVAL;
1462                 goto out_free;
1463         }
1464         if (new_size > device->bdev->bd_inode->i_size) {
1465                 ret = -EFBIG;
1466                 goto out_free;
1467         }
1468 
1469         do_div(new_size, root->sectorsize);
1470         new_size *= root->sectorsize;
1471 
1472         printk_in_rcu(KERN_INFO "btrfs: new size for %s is %llu\n",
1473                       rcu_str_deref(device->name),
1474                       (unsigned long long)new_size);
1475 
1476         if (new_size > old_size) {
1477                 trans = btrfs_start_transaction(root, 0);
1478                 if (IS_ERR(trans)) {
1479                         ret = PTR_ERR(trans);
1480                         goto out_free;
1481                 }
1482                 ret = btrfs_grow_device(trans, device, new_size);
1483                 btrfs_commit_transaction(trans, root);
1484         } else if (new_size < old_size) {
1485                 ret = btrfs_shrink_device(device, new_size);
1486         } /* equal, nothing need to do */
1487 
1488 out_free:
1489         kfree(vol_args);
1490 out:
1491         mutex_unlock(&root->fs_info->volume_mutex);
1492         atomic_set(&root->fs_info->mutually_exclusive_operation_running, 0);
1493         mnt_drop_write_file(file);
1494         return ret;
1495 }
1496 
1497 static noinline int btrfs_ioctl_snap_create_transid(struct file *file,
1498                                 char *name, unsigned long fd, int subvol,
1499                                 u64 *transid, bool readonly,
1500                                 struct btrfs_qgroup_inherit *inherit)
1501 {
1502         int namelen;
1503         int ret = 0;
1504 
1505         ret = mnt_want_write_file(file);
1506         if (ret)
1507                 goto out;
1508 
1509         namelen = strlen(name);
1510         if (strchr(name, '/')) {
1511                 ret = -EINVAL;
1512                 goto out_drop_write;
1513         }
1514 
1515         if (name[0] == '.' &&
1516            (namelen == 1 || (name[1] == '.' && namelen == 2))) {
1517                 ret = -EEXIST;
1518                 goto out_drop_write;
1519         }
1520 
1521         if (subvol) {
1522                 ret = btrfs_mksubvol(&file->f_path, name, namelen,
1523                                      NULL, transid, readonly, inherit);
1524         } else {
1525                 struct fd src = fdget(fd);
1526                 struct inode *src_inode;
1527                 if (!src.file) {
1528                         ret = -EINVAL;
1529                         goto out_drop_write;
1530                 }
1531 
1532                 src_inode = file_inode(src.file);
1533                 if (src_inode->i_sb != file_inode(file)->i_sb) {
1534                         printk(KERN_INFO "btrfs: Snapshot src from "
1535                                "another FS\n");
1536                         ret = -EINVAL;
1537                 } else {
1538                         ret = btrfs_mksubvol(&file->f_path, name, namelen,
1539                                              BTRFS_I(src_inode)->root,
1540                                              transid, readonly, inherit);
1541                 }
1542                 fdput(src);
1543         }
1544 out_drop_write:
1545         mnt_drop_write_file(file);
1546 out:
1547         return ret;
1548 }
1549 
1550 static noinline int btrfs_ioctl_snap_create(struct file *file,
1551                                             void __user *arg, int subvol)
1552 {
1553         struct btrfs_ioctl_vol_args *vol_args;
1554         int ret;
1555 
1556         vol_args = memdup_user(arg, sizeof(*vol_args));
1557         if (IS_ERR(vol_args))
1558                 return PTR_ERR(vol_args);
1559         vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1560 
1561         ret = btrfs_ioctl_snap_create_transid(file, vol_args->name,
1562                                               vol_args->fd, subvol,
1563                                               NULL, false, NULL);
1564 
1565         kfree(vol_args);
1566         return ret;
1567 }
1568 
1569 static noinline int btrfs_ioctl_snap_create_v2(struct file *file,
1570                                                void __user *arg, int subvol)
1571 {
1572         struct btrfs_ioctl_vol_args_v2 *vol_args;
1573         int ret;
1574         u64 transid = 0;
1575         u64 *ptr = NULL;
1576         bool readonly = false;
1577         struct btrfs_qgroup_inherit *inherit = NULL;
1578 
1579         vol_args = memdup_user(arg, sizeof(*vol_args));
1580         if (IS_ERR(vol_args))
1581                 return PTR_ERR(vol_args);
1582         vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0';
1583 
1584         if (vol_args->flags &
1585             ~(BTRFS_SUBVOL_CREATE_ASYNC | BTRFS_SUBVOL_RDONLY |
1586               BTRFS_SUBVOL_QGROUP_INHERIT)) {
1587                 ret = -EOPNOTSUPP;
1588                 goto out;
1589         }
1590 
1591         if (vol_args->flags & BTRFS_SUBVOL_CREATE_ASYNC)
1592                 ptr = &transid;
1593         if (vol_args->flags & BTRFS_SUBVOL_RDONLY)
1594                 readonly = true;
1595         if (vol_args->flags & BTRFS_SUBVOL_QGROUP_INHERIT) {
1596                 if (vol_args->size > PAGE_CACHE_SIZE) {
1597                         ret = -EINVAL;
1598                         goto out;
1599                 }
1600                 inherit = memdup_user(vol_args->qgroup_inherit, vol_args->size);
1601                 if (IS_ERR(inherit)) {
1602                         ret = PTR_ERR(inherit);
1603                         goto out;
1604                 }
1605         }
1606 
1607         ret = btrfs_ioctl_snap_create_transid(file, vol_args->name,
1608                                               vol_args->fd, subvol, ptr,
1609                                               readonly, inherit);
1610 
1611         if (ret == 0 && ptr &&
1612             copy_to_user(arg +
1613                          offsetof(struct btrfs_ioctl_vol_args_v2,
1614                                   transid), ptr, sizeof(*ptr)))
1615                 ret = -EFAULT;
1616 out:
1617         kfree(vol_args);
1618         kfree(inherit);
1619         return ret;
1620 }
1621 
1622 static noinline int btrfs_ioctl_subvol_getflags(struct file *file,
1623                                                 void __user *arg)
1624 {
1625         struct inode *inode = file_inode(file);
1626         struct btrfs_root *root = BTRFS_I(inode)->root;
1627         int ret = 0;
1628         u64 flags = 0;
1629 
1630         if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID)
1631                 return -EINVAL;
1632 
1633         down_read(&root->fs_info->subvol_sem);
1634         if (btrfs_root_readonly(root))
1635                 flags |= BTRFS_SUBVOL_RDONLY;
1636         up_read(&root->fs_info->subvol_sem);
1637 
1638         if (copy_to_user(arg, &flags, sizeof(flags)))
1639                 ret = -EFAULT;
1640 
1641         return ret;
1642 }
1643 
1644 static noinline int btrfs_ioctl_subvol_setflags(struct file *file,
1645                                               void __user *arg)
1646 {
1647         struct inode *inode = file_inode(file);
1648         struct btrfs_root *root = BTRFS_I(inode)->root;
1649         struct btrfs_trans_handle *trans;
1650         u64 root_flags;
1651         u64 flags;
1652         int ret = 0;
1653 
1654         ret = mnt_want_write_file(file);
1655         if (ret)
1656                 goto out;
1657 
1658         if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) {
1659                 ret = -EINVAL;
1660                 goto out_drop_write;
1661         }
1662 
1663         if (copy_from_user(&flags, arg, sizeof(flags))) {
1664                 ret = -EFAULT;
1665                 goto out_drop_write;
1666         }
1667 
1668         if (flags & BTRFS_SUBVOL_CREATE_ASYNC) {
1669                 ret = -EINVAL;
1670                 goto out_drop_write;
1671         }
1672 
1673         if (flags & ~BTRFS_SUBVOL_RDONLY) {
1674                 ret = -EOPNOTSUPP;
1675                 goto out_drop_write;
1676         }
1677 
1678         if (!inode_owner_or_capable(inode)) {
1679                 ret = -EACCES;
1680                 goto out_drop_write;
1681         }
1682 
1683         down_write(&root->fs_info->subvol_sem);
1684 
1685         /* nothing to do */
1686         if (!!(flags & BTRFS_SUBVOL_RDONLY) == btrfs_root_readonly(root))
1687                 goto out_drop_sem;
1688 
1689         root_flags = btrfs_root_flags(&root->root_item);
1690         if (flags & BTRFS_SUBVOL_RDONLY)
1691                 btrfs_set_root_flags(&root->root_item,
1692                                      root_flags | BTRFS_ROOT_SUBVOL_RDONLY);
1693         else
1694                 btrfs_set_root_flags(&root->root_item,
1695                                      root_flags & ~BTRFS_ROOT_SUBVOL_RDONLY);
1696 
1697         trans = btrfs_start_transaction(root, 1);
1698         if (IS_ERR(trans)) {
1699                 ret = PTR_ERR(trans);
1700                 goto out_reset;
1701         }
1702 
1703         ret = btrfs_update_root(trans, root->fs_info->tree_root,
1704                                 &root->root_key, &root->root_item);
1705 
1706         btrfs_commit_transaction(trans, root);
1707 out_reset:
1708         if (ret)
1709                 btrfs_set_root_flags(&root->root_item, root_flags);
1710 out_drop_sem:
1711         up_write(&root->fs_info->subvol_sem);
1712 out_drop_write:
1713         mnt_drop_write_file(file);
1714 out:
1715         return ret;
1716 }
1717 
1718 /*
1719  * helper to check if the subvolume references other subvolumes
1720  */
1721 static noinline int may_destroy_subvol(struct btrfs_root *root)
1722 {
1723         struct btrfs_path *path;
1724         struct btrfs_key key;
1725         int ret;
1726 
1727         path = btrfs_alloc_path();
1728         if (!path)
1729                 return -ENOMEM;
1730 
1731         key.objectid = root->root_key.objectid;
1732         key.type = BTRFS_ROOT_REF_KEY;
1733         key.offset = (u64)-1;
1734 
1735         ret = btrfs_search_slot(NULL, root->fs_info->tree_root,
1736                                 &key, path, 0, 0);
1737         if (ret < 0)
1738                 goto out;
1739         BUG_ON(ret == 0);
1740 
1741         ret = 0;
1742         if (path->slots[0] > 0) {
1743                 path->slots[0]--;
1744                 btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
1745                 if (key.objectid == root->root_key.objectid &&
1746                     key.type == BTRFS_ROOT_REF_KEY)
1747                         ret = -ENOTEMPTY;
1748         }
1749 out:
1750         btrfs_free_path(path);
1751         return ret;
1752 }
1753 
1754 static noinline int key_in_sk(struct btrfs_key *key,
1755                               struct btrfs_ioctl_search_key *sk)
1756 {
1757         struct btrfs_key test;
1758         int ret;
1759 
1760         test.objectid = sk->min_objectid;
1761         test.type = sk->min_type;
1762         test.offset = sk->min_offset;
1763 
1764         ret = btrfs_comp_cpu_keys(key, &test);
1765         if (ret < 0)
1766                 return 0;
1767 
1768         test.objectid = sk->max_objectid;
1769         test.type = sk->max_type;
1770         test.offset = sk->max_offset;
1771 
1772         ret = btrfs_comp_cpu_keys(key, &test);
1773         if (ret > 0)
1774                 return 0;
1775         return 1;
1776 }
1777 
1778 static noinline int copy_to_sk(struct btrfs_root *root,
1779                                struct btrfs_path *path,
1780                                struct btrfs_key *key,
1781                                struct btrfs_ioctl_search_key *sk,
1782                                char *buf,
1783                                unsigned long *sk_offset,
1784                                int *num_found)
1785 {
1786         u64 found_transid;
1787         struct extent_buffer *leaf;
1788         struct btrfs_ioctl_search_header sh;
1789         unsigned long item_off;
1790         unsigned long item_len;
1791         int nritems;
1792         int i;
1793         int slot;
1794         int ret = 0;
1795 
1796         leaf = path->nodes[0];
1797         slot = path->slots[0];
1798         nritems = btrfs_header_nritems(leaf);
1799 
1800         if (btrfs_header_generation(leaf) > sk->max_transid) {
1801                 i = nritems;
1802                 goto advance_key;
1803         }
1804         found_transid = btrfs_header_generation(leaf);
1805 
1806         for (i = slot; i < nritems; i++) {
1807                 item_off = btrfs_item_ptr_offset(leaf, i);
1808                 item_len = btrfs_item_size_nr(leaf, i);
1809 
1810                 btrfs_item_key_to_cpu(leaf, key, i);
1811                 if (!key_in_sk(key, sk))
1812                         continue;
1813 
1814                 if (sizeof(sh) + item_len > BTRFS_SEARCH_ARGS_BUFSIZE)
1815                         item_len = 0;
1816 
1817                 if (sizeof(sh) + item_len + *sk_offset >
1818                     BTRFS_SEARCH_ARGS_BUFSIZE) {
1819                         ret = 1;
1820                         goto overflow;
1821                 }
1822 
1823                 sh.objectid = key->objectid;
1824                 sh.offset = key->offset;
1825                 sh.type = key->type;
1826                 sh.len = item_len;
1827                 sh.transid = found_transid;
1828 
1829                 /* copy search result header */
1830                 memcpy(buf + *sk_offset, &sh, sizeof(sh));
1831                 *sk_offset += sizeof(sh);
1832 
1833                 if (item_len) {
1834                         char *p = buf + *sk_offset;
1835                         /* copy the item */
1836                         read_extent_buffer(leaf, p,
1837                                            item_off, item_len);
1838                         *sk_offset += item_len;
1839                 }
1840                 (*num_found)++;
1841 
1842                 if (*num_found >= sk->nr_items)
1843                         break;
1844         }
1845 advance_key:
1846         ret = 0;
1847         if (key->offset < (u64)-1 && key->offset < sk->max_offset)
1848                 key->offset++;
1849         else if (key->type < (u8)-1 && key->type < sk->max_type) {
1850                 key->offset = 0;
1851                 key->type++;
1852         } else if (key->objectid < (u64)-1 && key->objectid < sk->max_objectid) {
1853                 key->offset = 0;
1854                 key->type = 0;
1855                 key->objectid++;
1856         } else
1857                 ret = 1;
1858 overflow:
1859         return ret;
1860 }
1861 
1862 static noinline int search_ioctl(struct inode *inode,
1863                                  struct btrfs_ioctl_search_args *args)
1864 {
1865         struct btrfs_root *root;
1866         struct btrfs_key key;
1867         struct btrfs_key max_key;
1868         struct btrfs_path *path;
1869         struct btrfs_ioctl_search_key *sk = &args->key;
1870         struct btrfs_fs_info *info = BTRFS_I(inode)->root->fs_info;
1871         int ret;
1872         int num_found = 0;
1873         unsigned long sk_offset = 0;
1874 
1875         path = btrfs_alloc_path();
1876         if (!path)
1877                 return -ENOMEM;
1878 
1879         if (sk->tree_id == 0) {
1880                 /* search the root of the inode that was passed */
1881                 root = BTRFS_I(inode)->root;
1882         } else {
1883                 key.objectid = sk->tree_id;
1884                 key.type = BTRFS_ROOT_ITEM_KEY;
1885                 key.offset = (u64)-1;
1886                 root = btrfs_read_fs_root_no_name(info, &key);
1887                 if (IS_ERR(root)) {
1888                         printk(KERN_ERR "could not find root %llu\n",
1889                                sk->tree_id);
1890                         btrfs_free_path(path);
1891                         return -ENOENT;
1892                 }
1893         }
1894 
1895         key.objectid = sk->min_objectid;
1896         key.type = sk->min_type;
1897         key.offset = sk->min_offset;
1898 
1899         max_key.objectid = sk->max_objectid;
1900         max_key.type = sk->max_type;
1901         max_key.offset = sk->max_offset;
1902 
1903         path->keep_locks = 1;
1904 
1905         while(1) {
1906                 ret = btrfs_search_forward(root, &key, &max_key, path,
1907                                            sk->min_transid);
1908                 if (ret != 0) {
1909                         if (ret > 0)
1910                                 ret = 0;
1911                         goto err;
1912                 }
1913                 ret = copy_to_sk(root, path, &key, sk, args->buf,
1914                                  &sk_offset, &num_found);
1915                 btrfs_release_path(path);
1916                 if (ret || num_found >= sk->nr_items)
1917                         break;
1918 
1919         }
1920         ret = 0;
1921 err:
1922         sk->nr_items = num_found;
1923         btrfs_free_path(path);
1924         return ret;
1925 }
1926 
1927 static noinline int btrfs_ioctl_tree_search(struct file *file,
1928                                            void __user *argp)
1929 {
1930          struct btrfs_ioctl_search_args *args;
1931          struct inode *inode;
1932          int ret;
1933 
1934         if (!capable(CAP_SYS_ADMIN))
1935                 return -EPERM;
1936 
1937         args = memdup_user(argp, sizeof(*args));
1938         if (IS_ERR(args))
1939                 return PTR_ERR(args);
1940 
1941         inode = file_inode(file);
1942         ret = search_ioctl(inode, args);
1943         if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
1944                 ret = -EFAULT;
1945         kfree(args);
1946         return ret;
1947 }
1948 
1949 /*
1950  * Search INODE_REFs to identify path name of 'dirid' directory
1951  * in a 'tree_id' tree. and sets path name to 'name'.
1952  */
1953 static noinline int btrfs_search_path_in_tree(struct btrfs_fs_info *info,
1954                                 u64 tree_id, u64 dirid, char *name)
1955 {
1956         struct btrfs_root *root;
1957         struct btrfs_key key;
1958         char *ptr;
1959         int ret = -1;
1960         int slot;
1961         int len;
1962         int total_len = 0;
1963         struct btrfs_inode_ref *iref;
1964         struct extent_buffer *l;
1965         struct btrfs_path *path;
1966 
1967         if (dirid == BTRFS_FIRST_FREE_OBJECTID) {
1968                 name[0]='\0';
1969                 return 0;
1970         }
1971 
1972         path = btrfs_alloc_path();
1973         if (!path)
1974                 return -ENOMEM;
1975 
1976         ptr = &name[BTRFS_INO_LOOKUP_PATH_MAX];
1977 
1978         key.objectid = tree_id;
1979         key.type = BTRFS_ROOT_ITEM_KEY;
1980         key.offset = (u64)-1;
1981         root = btrfs_read_fs_root_no_name(info, &key);
1982         if (IS_ERR(root)) {
1983                 printk(KERN_ERR "could not find root %llu\n", tree_id);
1984                 ret = -ENOENT;
1985                 goto out;
1986         }
1987 
1988         key.objectid = dirid;
1989         key.type = BTRFS_INODE_REF_KEY;
1990         key.offset = (u64)-1;
1991 
1992         while(1) {
1993                 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
1994                 if (ret < 0)
1995                         goto out;
1996 
1997                 l = path->nodes[0];
1998                 slot = path->slots[0];
1999                 if (ret > 0 && slot > 0)
2000                         slot--;
2001                 btrfs_item_key_to_cpu(l, &key, slot);
2002 
2003                 if (ret > 0 && (key.objectid != dirid ||
2004                                 key.type != BTRFS_INODE_REF_KEY)) {
2005                         ret = -ENOENT;
2006                         goto out;
2007                 }
2008 
2009                 iref = btrfs_item_ptr(l, slot, struct btrfs_inode_ref);
2010                 len = btrfs_inode_ref_name_len(l, iref);
2011                 ptr -= len + 1;
2012                 total_len += len + 1;
2013                 if (ptr < name)
2014                         goto out;
2015 
2016                 *(ptr + len) = '/';
2017                 read_extent_buffer(l, ptr,(unsigned long)(iref + 1), len);
2018 
2019                 if (key.offset == BTRFS_FIRST_FREE_OBJECTID)
2020                         break;
2021 
2022                 btrfs_release_path(path);
2023                 key.objectid = key.offset;
2024                 key.offset = (u64)-1;
2025                 dirid = key.objectid;
2026         }
2027         if (ptr < name)
2028                 goto out;
2029         memmove(name, ptr, total_len);
2030         name[total_len]='\0';
2031         ret = 0;
2032 out:
2033         btrfs_free_path(path);
2034         return ret;
2035 }
2036 
2037 static noinline int btrfs_ioctl_ino_lookup(struct file *file,
2038                                            void __user *argp)
2039 {
2040          struct btrfs_ioctl_ino_lookup_args *args;
2041          struct inode *inode;
2042          int ret;
2043 
2044         if (!capable(CAP_SYS_ADMIN))
2045                 return -EPERM;
2046 
2047         args = memdup_user(argp, sizeof(*args));
2048         if (IS_ERR(args))
2049                 return PTR_ERR(args);
2050 
2051         inode = file_inode(file);
2052 
2053         if (args->treeid == 0)
2054                 args->treeid = BTRFS_I(inode)->root->root_key.objectid;
2055 
2056         ret = btrfs_search_path_in_tree(BTRFS_I(inode)->root->fs_info,
2057                                         args->treeid, args->objectid,
2058                                         args->name);
2059 
2060         if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
2061                 ret = -EFAULT;
2062 
2063         kfree(args);
2064         return ret;
2065 }
2066 
2067 static noinline int btrfs_ioctl_snap_destroy(struct file *file,
2068                                              void __user *arg)
2069 {
2070         struct dentry *parent = fdentry(file);
2071         struct dentry *dentry;
2072         struct inode *dir = parent->d_inode;
2073         struct inode *inode;
2074         struct btrfs_root *root = BTRFS_I(dir)->root;
2075         struct btrfs_root *dest = NULL;
2076         struct btrfs_ioctl_vol_args *vol_args;
2077         struct btrfs_trans_handle *trans;
2078         struct btrfs_block_rsv block_rsv;
2079         u64 qgroup_reserved;
2080         int namelen;
2081         int ret;
2082         int err = 0;
2083 
2084         vol_args = memdup_user(arg, sizeof(*vol_args));
2085         if (IS_ERR(vol_args))
2086                 return PTR_ERR(vol_args);
2087 
2088         vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2089         namelen = strlen(vol_args->name);
2090         if (strchr(vol_args->name, '/') ||
2091             strncmp(vol_args->name, "..", namelen) == 0) {
2092                 err = -EINVAL;
2093                 goto out;
2094         }
2095 
2096         err = mnt_want_write_file(file);
2097         if (err)
2098                 goto out;
2099 
2100         err = mutex_lock_killable_nested(&dir->i_mutex, I_MUTEX_PARENT);
2101         if (err == -EINTR)
2102                 goto out;
2103         dentry = lookup_one_len(vol_args->name, parent, namelen);
2104         if (IS_ERR(dentry)) {
2105                 err = PTR_ERR(dentry);
2106                 goto out_unlock_dir;
2107         }
2108 
2109         if (!dentry->d_inode) {
2110                 err = -ENOENT;
2111                 goto out_dput;
2112         }
2113 
2114         inode = dentry->d_inode;
2115         dest = BTRFS_I(inode)->root;
2116         if (!capable(CAP_SYS_ADMIN)){
2117                 /*
2118                  * Regular user.  Only allow this with a special mount
2119                  * option, when the user has write+exec access to the
2120                  * subvol root, and when rmdir(2) would have been
2121                  * allowed.
2122                  *
2123                  * Note that this is _not_ check that the subvol is
2124                  * empty or doesn't contain data that we wouldn't
2125                  * otherwise be able to delete.
2126                  *
2127                  * Users who want to delete empty subvols should try
2128                  * rmdir(2).
2129                  */
2130                 err = -EPERM;
2131                 if (!btrfs_test_opt(root, USER_SUBVOL_RM_ALLOWED))
2132                         goto out_dput;
2133 
2134                 /*
2135                  * Do not allow deletion if the parent dir is the same
2136                  * as the dir to be deleted.  That means the ioctl
2137                  * must be called on the dentry referencing the root
2138                  * of the subvol, not a random directory contained
2139                  * within it.
2140                  */
2141                 err = -EINVAL;
2142                 if (root == dest)
2143                         goto out_dput;
2144 
2145                 err = inode_permission(inode, MAY_WRITE | MAY_EXEC);
2146                 if (err)
2147                         goto out_dput;
2148         }
2149 
2150         /* check if subvolume may be deleted by a user */
2151         err = btrfs_may_delete(dir, dentry, 1);
2152         if (err)
2153                 goto out_dput;
2154 
2155         if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) {
2156                 err = -EINVAL;
2157                 goto out_dput;
2158         }
2159 
2160         mutex_lock(&inode->i_mutex);
2161         err = d_invalidate(dentry);
2162         if (err)
2163                 goto out_unlock;
2164 
2165         down_write(&root->fs_info->subvol_sem);
2166 
2167         err = may_destroy_subvol(dest);
2168         if (err)
2169                 goto out_up_write;
2170 
2171         btrfs_init_block_rsv(&block_rsv, BTRFS_BLOCK_RSV_TEMP);
2172         /*
2173          * One for dir inode, two for dir entries, two for root
2174          * ref/backref.
2175          */
2176         err = btrfs_subvolume_reserve_metadata(root, &block_rsv,
2177                                                5, &qgroup_reserved);
2178         if (err)
2179                 goto out_up_write;
2180 
2181         trans = btrfs_start_transaction(root, 0);
2182         if (IS_ERR(trans)) {
2183                 err = PTR_ERR(trans);
2184                 goto out_release;
2185         }
2186         trans->block_rsv = &block_rsv;
2187         trans->bytes_reserved = block_rsv.size;
2188 
2189         ret = btrfs_unlink_subvol(trans, root, dir,
2190                                 dest->root_key.objectid,
2191                                 dentry->d_name.name,
2192                                 dentry->d_name.len);
2193         if (ret) {
2194                 err = ret;
2195                 btrfs_abort_transaction(trans, root, ret);
2196                 goto out_end_trans;
2197         }
2198 
2199         btrfs_record_root_in_trans(trans, dest);
2200 
2201         memset(&dest->root_item.drop_progress, 0,
2202                 sizeof(dest->root_item.drop_progress));
2203         dest->root_item.drop_level = 0;
2204         btrfs_set_root_refs(&dest->root_item, 0);
2205 
2206         if (!xchg(&dest->orphan_item_inserted, 1)) {
2207                 ret = btrfs_insert_orphan_item(trans,
2208                                         root->fs_info->tree_root,
2209                                         dest->root_key.objectid);
2210                 if (ret) {
2211                         btrfs_abort_transaction(trans, root, ret);
2212                         err = ret;
2213                         goto out_end_trans;
2214                 }
2215         }
2216 out_end_trans:
2217         trans->block_rsv = NULL;
2218         trans->bytes_reserved = 0;
2219         ret = btrfs_end_transaction(trans, root);
2220         if (ret && !err)
2221                 err = ret;
2222         inode->i_flags |= S_DEAD;
2223 out_release:
2224         btrfs_subvolume_release_metadata(root, &block_rsv, qgroup_reserved);
2225 out_up_write:
2226         up_write(&root->fs_info->subvol_sem);
2227 out_unlock:
2228         mutex_unlock(&inode->i_mutex);
2229         if (!err) {
2230                 shrink_dcache_sb(root->fs_info->sb);
2231                 btrfs_invalidate_inodes(dest);
2232                 d_delete(dentry);
2233 
2234                 /* the last ref */
2235                 if (dest->cache_inode) {
2236                         iput(dest->cache_inode);
2237                         dest->cache_inode = NULL;
2238                 }
2239         }
2240 out_dput:
2241         dput(dentry);
2242 out_unlock_dir:
2243         mutex_unlock(&dir->i_mutex);
2244         mnt_drop_write_file(file);
2245 out:
2246         kfree(vol_args);
2247         return err;
2248 }
2249 
2250 static int btrfs_ioctl_defrag(struct file *file, void __user *argp)
2251 {
2252         struct inode *inode = file_inode(file);
2253         struct btrfs_root *root = BTRFS_I(inode)->root;
2254         struct btrfs_ioctl_defrag_range_args *range;
2255         int ret;
2256 
2257         ret = mnt_want_write_file(file);
2258         if (ret)
2259                 return ret;
2260 
2261         if (btrfs_root_readonly(root)) {
2262                 ret = -EROFS;
2263                 goto out;
2264         }
2265 
2266         switch (inode->i_mode & S_IFMT) {
2267         case S_IFDIR:
2268                 if (!capable(CAP_SYS_ADMIN)) {
2269                         ret = -EPERM;
2270                         goto out;
2271                 }
2272                 ret = btrfs_defrag_root(root);
2273                 if (ret)
2274                         goto out;
2275                 ret = btrfs_defrag_root(root->fs_info->extent_root);
2276                 break;
2277         case S_IFREG:
2278                 if (!(file->f_mode & FMODE_WRITE)) {
2279                         ret = -EINVAL;
2280                         goto out;
2281                 }
2282 
2283                 range = kzalloc(sizeof(*range), GFP_KERNEL);
2284                 if (!range) {
2285                         ret = -ENOMEM;
2286                         goto out;
2287                 }
2288 
2289                 if (argp) {
2290                         if (copy_from_user(range, argp,
2291                                            sizeof(*range))) {
2292                                 ret = -EFAULT;
2293                                 kfree(range);
2294                                 goto out;
2295                         }
2296                         /* compression requires us to start the IO */
2297                         if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
2298                                 range->flags |= BTRFS_DEFRAG_RANGE_START_IO;
2299                                 range->extent_thresh = (u32)-1;
2300                         }
2301                 } else {
2302                         /* the rest are all set to zero by kzalloc */
2303                         range->len = (u64)-1;
2304                 }
2305                 ret = btrfs_defrag_file(file_inode(file), file,
2306                                         range, 0, 0);
2307                 if (ret > 0)
2308                         ret = 0;
2309                 kfree(range);
2310                 break;
2311         default:
2312                 ret = -EINVAL;
2313         }
2314 out:
2315         mnt_drop_write_file(file);
2316         return ret;
2317 }
2318 
2319 static long btrfs_ioctl_add_dev(struct btrfs_root *root, void __user *arg)
2320 {
2321         struct btrfs_ioctl_vol_args *vol_args;
2322         int ret;
2323 
2324         if (!capable(CAP_SYS_ADMIN))
2325                 return -EPERM;
2326 
2327         if (atomic_xchg(&root->fs_info->mutually_exclusive_operation_running,
2328                         1)) {
2329                 pr_info("btrfs: dev add/delete/balance/replace/resize operation in progress\n");
2330                 return -EINVAL;
2331         }
2332 
2333         mutex_lock(&root->fs_info->volume_mutex);
2334         vol_args = memdup_user(arg, sizeof(*vol_args));
2335         if (IS_ERR(vol_args)) {
2336                 ret = PTR_ERR(vol_args);
2337                 goto out;
2338         }
2339 
2340         vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2341         ret = btrfs_init_new_device(root, vol_args->name);
2342 
2343         kfree(vol_args);
2344 out:
2345         mutex_unlock(&root->fs_info->volume_mutex);
2346         atomic_set(&root->fs_info->mutually_exclusive_operation_running, 0);
2347         return ret;
2348 }
2349 
2350 static long btrfs_ioctl_rm_dev(struct file *file, void __user *arg)
2351 {
2352         struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
2353         struct btrfs_ioctl_vol_args *vol_args;
2354         int ret;
2355 
2356         if (!capable(CAP_SYS_ADMIN))
2357                 return -EPERM;
2358 
2359         ret = mnt_want_write_file(file);
2360         if (ret)
2361                 return ret;
2362 
2363         vol_args = memdup_user(arg, sizeof(*vol_args));
2364         if (IS_ERR(vol_args)) {
2365                 ret = PTR_ERR(vol_args);
2366                 goto out;
2367         }
2368 
2369         vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2370 
2371         if (atomic_xchg(&root->fs_info->mutually_exclusive_operation_running,
2372                         1)) {
2373                 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
2374                 goto out;
2375         }
2376 
2377         mutex_lock(&root->fs_info->volume_mutex);
2378         ret = btrfs_rm_device(root, vol_args->name);
2379         mutex_unlock(&root->fs_info->volume_mutex);
2380         atomic_set(&root->fs_info->mutually_exclusive_operation_running, 0);
2381 
2382 out:
2383         kfree(vol_args);
2384         mnt_drop_write_file(file);
2385         return ret;
2386 }
2387 
2388 static long btrfs_ioctl_fs_info(struct btrfs_root *root, void __user *arg)
2389 {
2390         struct btrfs_ioctl_fs_info_args *fi_args;
2391         struct btrfs_device *device;
2392         struct btrfs_device *next;
2393         struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
2394         int ret = 0;
2395 
2396         if (!capable(CAP_SYS_ADMIN))
2397                 return -EPERM;
2398 
2399         fi_args = kzalloc(sizeof(*fi_args), GFP_KERNEL);
2400         if (!fi_args)
2401                 return -ENOMEM;
2402 
2403         fi_args->num_devices = fs_devices->num_devices;
2404         memcpy(&fi_args->fsid, root->fs_info->fsid, sizeof(fi_args->fsid));
2405 
2406         mutex_lock(&fs_devices->device_list_mutex);
2407         list_for_each_entry_safe(device, next, &fs_devices->devices, dev_list) {
2408                 if (device->devid > fi_args->max_id)
2409                         fi_args->max_id = device->devid;
2410         }
2411         mutex_unlock(&fs_devices->device_list_mutex);
2412 
2413         if (copy_to_user(arg, fi_args, sizeof(*fi_args)))
2414                 ret = -EFAULT;
2415 
2416         kfree(fi_args);
2417         return ret;
2418 }
2419 
2420 static long btrfs_ioctl_dev_info(struct btrfs_root *root, void __user *arg)
2421 {
2422         struct btrfs_ioctl_dev_info_args *di_args;
2423         struct btrfs_device *dev;
2424         struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
2425         int ret = 0;
2426         char *s_uuid = NULL;
2427         char empty_uuid[BTRFS_UUID_SIZE] = {0};
2428 
2429         if (!capable(CAP_SYS_ADMIN))
2430                 return -EPERM;
2431 
2432         di_args = memdup_user(arg, sizeof(*di_args));
2433         if (IS_ERR(di_args))
2434                 return PTR_ERR(di_args);
2435 
2436         if (memcmp(empty_uuid, di_args->uuid, BTRFS_UUID_SIZE) != 0)
2437                 s_uuid = di_args->uuid;
2438 
2439         mutex_lock(&fs_devices->device_list_mutex);
2440         dev = btrfs_find_device(root->fs_info, di_args->devid, s_uuid, NULL);
2441 
2442         if (!dev) {
2443                 ret = -ENODEV;
2444                 goto out;
2445         }
2446 
2447         di_args->devid = dev->devid;
2448         di_args->bytes_used = dev->bytes_used;
2449         di_args->total_bytes = dev->total_bytes;
2450         memcpy(di_args->uuid, dev->uuid, sizeof(di_args->uuid));
2451         if (dev->name) {
2452                 struct rcu_string *name;
2453 
2454                 rcu_read_lock();
2455                 name = rcu_dereference(dev->name);
2456                 strncpy(di_args->path, name->str, sizeof(di_args->path));
2457                 rcu_read_unlock();
2458                 di_args->path[sizeof(di_args->path) - 1] = 0;
2459         } else {
2460                 di_args->path[0] = '\0';
2461         }
2462 
2463 out:
2464         mutex_unlock(&fs_devices->device_list_mutex);
2465         if (ret == 0 && copy_to_user(arg, di_args, sizeof(*di_args)))
2466                 ret = -EFAULT;
2467 
2468         kfree(di_args);
2469         return ret;
2470 }
2471 
2472 static noinline long btrfs_ioctl_clone(struct file *file, unsigned long srcfd,
2473                                        u64 off, u64 olen, u64 destoff)
2474 {
2475         struct inode *inode = file_inode(file);
2476         struct btrfs_root *root = BTRFS_I(inode)->root;
2477         struct fd src_file;
2478         struct inode *src;
2479         struct btrfs_trans_handle *trans;
2480         struct btrfs_path *path;
2481         struct extent_buffer *leaf;
2482         char *buf;
2483         struct btrfs_key key;
2484         u32 nritems;
2485         int slot;
2486         int ret;
2487         u64 len = olen;
2488         u64 bs = root->fs_info->sb->s_blocksize;
2489         int same_inode = 0;
2490 
2491         /*
2492          * TODO:
2493          * - split compressed inline extents.  annoying: we need to
2494          *   decompress into destination's address_space (the file offset
2495          *   may change, so source mapping won't do), then recompress (or
2496          *   otherwise reinsert) a subrange.
2497          * - allow ranges within the same file to be cloned (provided
2498          *   they don't overlap)?
2499          */
2500 
2501         /* the destination must be opened for writing */
2502         if (!(file->f_mode & FMODE_WRITE) || (file->f_flags & O_APPEND))
2503                 return -EINVAL;
2504 
2505         if (btrfs_root_readonly(root))
2506                 return -EROFS;
2507 
2508         ret = mnt_want_write_file(file);
2509         if (ret)
2510                 return ret;
2511 
2512         src_file = fdget(srcfd);
2513         if (!src_file.file) {
2514                 ret = -EBADF;
2515                 goto out_drop_write;
2516         }
2517 
2518         ret = -EXDEV;
2519         if (src_file.file->f_path.mnt != file->f_path.mnt)
2520                 goto out_fput;
2521 
2522         src = file_inode(src_file.file);
2523 
2524         ret = -EINVAL;
2525         if (src == inode)
2526                 same_inode = 1;
2527 
2528         /* the src must be open for reading */
2529         if (!(src_file.file->f_mode & FMODE_READ))
2530                 goto out_fput;
2531 
2532         /* don't make the dst file partly checksummed */
2533         if ((BTRFS_I(src)->flags & BTRFS_INODE_NODATASUM) !=
2534             (BTRFS_I(inode)->flags & BTRFS_INODE_NODATASUM))
2535                 goto out_fput;
2536 
2537         ret = -EISDIR;
2538         if (S_ISDIR(src->i_mode) || S_ISDIR(inode->i_mode))
2539                 goto out_fput;
2540 
2541         ret = -EXDEV;
2542         if (src->i_sb != inode->i_sb)
2543                 goto out_fput;
2544 
2545         ret = -ENOMEM;
2546         buf = vmalloc(btrfs_level_size(root, 0));
2547         if (!buf)
2548                 goto out_fput;
2549 
2550         path = btrfs_alloc_path();
2551         if (!path) {
2552                 vfree(buf);
2553                 goto out_fput;
2554         }
2555         path->reada = 2;
2556 
2557         if (!same_inode) {
2558                 if (inode < src) {
2559                         mutex_lock_nested(&inode->i_mutex, I_MUTEX_PARENT);
2560                         mutex_lock_nested(&src->i_mutex, I_MUTEX_CHILD);
2561                 } else {
2562                         mutex_lock_nested(&src->i_mutex, I_MUTEX_PARENT);
2563                         mutex_lock_nested(&inode->i_mutex, I_MUTEX_CHILD);
2564                 }
2565         } else {
2566                 mutex_lock(&src->i_mutex);
2567         }
2568 
2569         /* determine range to clone */
2570         ret = -EINVAL;
2571         if (off + len > src->i_size || off + len < off)
2572                 goto out_unlock;
2573         if (len == 0)
2574                 olen = len = src->i_size - off;
2575         /* if we extend to eof, continue to block boundary */
2576         if (off + len == src->i_size)
2577                 len = ALIGN(src->i_size, bs) - off;
2578 
2579         /* verify the end result is block aligned */
2580         if (!IS_ALIGNED(off, bs) || !IS_ALIGNED(off + len, bs) ||
2581             !IS_ALIGNED(destoff, bs))
2582                 goto out_unlock;
2583 
2584         /* verify if ranges are overlapped within the same file */
2585         if (same_inode) {
2586                 if (destoff + len > off && destoff < off + len)
2587                         goto out_unlock;
2588         }
2589 
2590         if (destoff > inode->i_size) {
2591                 ret = btrfs_cont_expand(inode, inode->i_size, destoff);
2592                 if (ret)
2593                         goto out_unlock;
2594         }
2595 
2596         /* truncate page cache pages from target inode range */
2597         truncate_inode_pages_range(&inode->i_data, destoff,
2598                                    PAGE_CACHE_ALIGN(destoff + len) - 1);
2599 
2600         /* do any pending delalloc/csum calc on src, one way or
2601            another, and lock file content */
2602         while (1) {
2603                 struct btrfs_ordered_extent *ordered;
2604                 lock_extent(&BTRFS_I(src)->io_tree, off, off + len - 1);
2605                 ordered = btrfs_lookup_first_ordered_extent(src, off + len - 1);
2606                 if (!ordered &&
2607                     !test_range_bit(&BTRFS_I(src)->io_tree, off, off + len - 1,
2608                                     EXTENT_DELALLOC, 0, NULL))
2609                         break;
2610                 unlock_extent(&BTRFS_I(src)->io_tree, off, off + len - 1);
2611                 if (ordered)
2612                         btrfs_put_ordered_extent(ordered);
2613                 btrfs_wait_ordered_range(src, off, len);
2614         }
2615 
2616         /* clone data */
2617         key.objectid = btrfs_ino(src);
2618         key.type = BTRFS_EXTENT_DATA_KEY;
2619         key.offset = 0;
2620 
2621         while (1) {
2622                 /*
2623                  * note the key will change type as we walk through the
2624                  * tree.
2625                  */
2626                 ret = btrfs_search_slot(NULL, BTRFS_I(src)->root, &key, path,
2627                                 0, 0);
2628                 if (ret < 0)
2629                         goto out;
2630 
2631                 nritems = btrfs_header_nritems(path->nodes[0]);
2632                 if (path->slots[0] >= nritems) {
2633                         ret = btrfs_next_leaf(BTRFS_I(src)->root, path);
2634                         if (ret < 0)
2635                                 goto out;
2636                         if (ret > 0)
2637                                 break;
2638                         nritems = btrfs_header_nritems(path->nodes[0]);
2639                 }
2640                 leaf = path->nodes[0];
2641                 slot = path->slots[0];
2642 
2643                 btrfs_item_key_to_cpu(leaf, &key, slot);
2644                 if (btrfs_key_type(&key) > BTRFS_EXTENT_DATA_KEY ||
2645                     key.objectid != btrfs_ino(src))
2646                         break;
2647 
2648                 if (btrfs_key_type(&key) == BTRFS_EXTENT_DATA_KEY) {
2649                         struct btrfs_file_extent_item *extent;
2650                         int type;
2651                         u32 size;
2652                         struct btrfs_key new_key;
2653                         u64 disko = 0, diskl = 0;
2654                         u64 datao = 0, datal = 0;
2655                         u8 comp;
2656                         u64 endoff;
2657 
2658                         size = btrfs_item_size_nr(leaf, slot);
2659                         read_extent_buffer(leaf, buf,
2660                                            btrfs_item_ptr_offset(leaf, slot),
2661                                            size);
2662 
2663                         extent = btrfs_item_ptr(leaf, slot,
2664                                                 struct btrfs_file_extent_item);
2665                         comp = btrfs_file_extent_compression(leaf, extent);
2666                         type = btrfs_file_extent_type(leaf, extent);
2667                         if (type == BTRFS_FILE_EXTENT_REG ||
2668                             type == BTRFS_FILE_EXTENT_PREALLOC) {
2669                                 disko = btrfs_file_extent_disk_bytenr(leaf,
2670                                                                       extent);
2671                                 diskl = btrfs_file_extent_disk_num_bytes(leaf,
2672                                                                  extent);
2673                                 datao = btrfs_file_extent_offset(leaf, extent);
2674                                 datal = btrfs_file_extent_num_bytes(leaf,
2675                                                                     extent);
2676                         } else if (type == BTRFS_FILE_EXTENT_INLINE) {
2677                                 /* take upper bound, may be compressed */
2678                                 datal = btrfs_file_extent_ram_bytes(leaf,
2679                                                                     extent);
2680                         }
2681                         btrfs_release_path(path);
2682 
2683                         if (key.offset + datal <= off ||
2684                             key.offset >= off + len - 1)
2685                                 goto next;
2686 
2687                         memcpy(&new_key, &key, sizeof(new_key));
2688                         new_key.objectid = btrfs_ino(inode);
2689                         if (off <= key.offset)
2690                                 new_key.offset = key.offset + destoff - off;
2691                         else
2692                                 new_key.offset = destoff;
2693 
2694                         /*
2695                          * 1 - adjusting old extent (we may have to split it)
2696                          * 1 - add new extent
2697                          * 1 - inode update
2698                          */
2699                         trans = btrfs_start_transaction(root, 3);
2700                         if (IS_ERR(trans)) {
2701                                 ret = PTR_ERR(trans);
2702                                 goto out;
2703                         }
2704 
2705                         if (type == BTRFS_FILE_EXTENT_REG ||
2706                             type == BTRFS_FILE_EXTENT_PREALLOC) {
2707                                 /*
2708                                  *    a  | --- range to clone ---|  b
2709                                  * | ------------- extent ------------- |
2710                                  */
2711 
2712                                 /* substract range b */
2713                                 if (key.offset + datal > off + len)
2714                                         datal = off + len - key.offset;
2715 
2716                                 /* substract range a */
2717                                 if (off > key.offset) {
2718                                         datao += off - key.offset;
2719                                         datal -= off - key.offset;
2720                                 }
2721 
2722                                 ret = btrfs_drop_extents(trans, root, inode,
2723                                                          new_key.offset,
2724                                                          new_key.offset + datal,
2725                                                          1);
2726                                 if (ret) {
2727                                         btrfs_abort_transaction(trans, root,
2728                                                                 ret);
2729                                         btrfs_end_transaction(trans, root);
2730                                         goto out;
2731                                 }
2732 
2733                                 ret = btrfs_insert_empty_item(trans, root, path,
2734                                                               &new_key, size);
2735                                 if (ret) {
2736                                         btrfs_abort_transaction(trans, root,
2737                                                                 ret);
2738                                         btrfs_end_transaction(trans, root);
2739                                         goto out;
2740                                 }
2741 
2742                                 leaf = path->nodes[0];
2743                                 slot = path->slots[0];
2744                                 write_extent_buffer(leaf, buf,
2745                                             btrfs_item_ptr_offset(leaf, slot),
2746                                             size);
2747 
2748                                 extent = btrfs_item_ptr(leaf, slot,
2749                                                 struct btrfs_file_extent_item);
2750 
2751                                 /* disko == 0 means it's a hole */
2752                                 if (!disko)
2753                                         datao = 0;
2754 
2755                                 btrfs_set_file_extent_offset(leaf, extent,
2756                                                              datao);
2757                                 btrfs_set_file_extent_num_bytes(leaf, extent,
2758                                                                 datal);
2759                                 if (disko) {
2760                                         inode_add_bytes(inode, datal);
2761                                         ret = btrfs_inc_extent_ref(trans, root,
2762                                                         disko, diskl, 0,
2763                                                         root->root_key.objectid,
2764                                                         btrfs_ino(inode),
2765                                                         new_key.offset - datao,
2766                                                         0);
2767                                         if (ret) {
2768                                                 btrfs_abort_transaction(trans,
2769                                                                         root,
2770                                                                         ret);
2771                                                 btrfs_end_transaction(trans,
2772                                                                       root);
2773                                                 goto out;
2774 
2775                                         }
2776                                 }
2777                         } else if (type == BTRFS_FILE_EXTENT_INLINE) {
2778                                 u64 skip = 0;
2779                                 u64 trim = 0;
2780                                 if (off > key.offset) {
2781                                         skip = off - key.offset;
2782                                         new_key.offset += skip;
2783                                 }
2784 
2785                                 if (key.offset + datal > off + len)
2786                                         trim = key.offset + datal - (off + len);
2787 
2788                                 if (comp && (skip || trim)) {
2789                                         ret = -EINVAL;
2790                                         btrfs_end_transaction(trans, root);
2791                                         goto out;
2792                                 }
2793                                 size -= skip + trim;
2794                                 datal -= skip + trim;
2795 
2796                                 ret = btrfs_drop_extents(trans, root, inode,
2797                                                          new_key.offset,
2798                                                          new_key.offset + datal,
2799                                                          1);
2800                                 if (ret) {
2801                                         btrfs_abort_transaction(trans, root,
2802                                                                 ret);
2803                                         btrfs_end_transaction(trans, root);
2804                                         goto out;
2805                                 }
2806 
2807                                 ret = btrfs_insert_empty_item(trans, root, path,
2808                                                               &new_key, size);
2809                                 if (ret) {
2810                                         btrfs_abort_transaction(trans, root,
2811                                                                 ret);
2812                                         btrfs_end_transaction(trans, root);
2813                                         goto out;
2814                                 }
2815 
2816                                 if (skip) {
2817                                         u32 start =
2818                                           btrfs_file_extent_calc_inline_size(0);
2819                                         memmove(buf+start, buf+start+skip,
2820                                                 datal);
2821                                 }
2822 
2823                                 leaf = path->nodes[0];
2824                                 slot = path->slots[0];
2825                                 write_extent_buffer(leaf, buf,
2826                                             btrfs_item_ptr_offset(leaf, slot),
2827                                             size);
2828                                 inode_add_bytes(inode, datal);
2829                         }
2830 
2831                         btrfs_mark_buffer_dirty(leaf);
2832                         btrfs_release_path(path);
2833 
2834                         inode_inc_iversion(inode);
2835                         inode->i_mtime = inode->i_ctime = CURRENT_TIME;
2836 
2837                         /*
2838                          * we round up to the block size at eof when
2839                          * determining which extents to clone above,
2840                          * but shouldn't round up the file size
2841                          */
2842                         endoff = new_key.offset + datal;
2843                         if (endoff > destoff+olen)
2844                                 endoff = destoff+olen;
2845                         if (endoff > inode->i_size)
2846                                 btrfs_i_size_write(inode, endoff);
2847 
2848                         ret = btrfs_update_inode(trans, root, inode);
2849                         if (ret) {
2850                                 btrfs_abort_transaction(trans, root, ret);
2851                                 btrfs_end_transaction(trans, root);
2852                                 goto out;
2853                         }
2854                         ret = btrfs_end_transaction(trans, root);
2855                 }
2856 next:
2857                 btrfs_release_path(path);
2858                 key.offset++;
2859         }
2860         ret = 0;
2861 out:
2862         btrfs_release_path(path);
2863         unlock_extent(&BTRFS_I(src)->io_tree, off, off + len - 1);
2864 out_unlock:
2865         mutex_unlock(&src->i_mutex);
2866         if (!same_inode)
2867                 mutex_unlock(&inode->i_mutex);
2868         vfree(buf);
2869         btrfs_free_path(path);
2870 out_fput:
2871         fdput(src_file);
2872 out_drop_write:
2873         mnt_drop_write_file(file);
2874         return ret;
2875 }
2876 
2877 static long btrfs_ioctl_clone_range(struct file *file, void __user *argp)
2878 {
2879         struct btrfs_ioctl_clone_range_args args;
2880 
2881         if (copy_from_user(&args, argp, sizeof(args)))
2882                 return -EFAULT;
2883         return btrfs_ioctl_clone(file, args.src_fd, args.src_offset,
2884                                  args.src_length, args.dest_offset);
2885 }
2886 
2887 /*
2888  * there are many ways the trans_start and trans_end ioctls can lead
2889  * to deadlocks.  They should only be used by applications that
2890  * basically own the machine, and have a very in depth understanding
2891  * of all the possible deadlocks and enospc problems.
2892  */
2893 static long btrfs_ioctl_trans_start(struct file *file)
2894 {
2895         struct inode *inode = file_inode(file);
2896         struct btrfs_root *root = BTRFS_I(inode)->root;
2897         struct btrfs_trans_handle *trans;
2898         int ret;
2899 
2900         ret = -EPERM;
2901         if (!capable(CAP_SYS_ADMIN))
2902                 goto out;
2903 
2904         ret = -EINPROGRESS;
2905         if (file->private_data)
2906                 goto out;
2907 
2908         ret = -EROFS;
2909         if (btrfs_root_readonly(root))
2910                 goto out;
2911 
2912         ret = mnt_want_write_file(file);
2913         if (ret)
2914                 goto out;
2915 
2916         atomic_inc(&root->fs_info->open_ioctl_trans);
2917 
2918         ret = -ENOMEM;
2919         trans = btrfs_start_ioctl_transaction(root);
2920         if (IS_ERR(trans))
2921                 goto out_drop;
2922 
2923         file->private_data = trans;
2924         return 0;
2925 
2926 out_drop:
2927         atomic_dec(&root->fs_info->open_ioctl_trans);
2928         mnt_drop_write_file(file);
2929 out:
2930         return ret;
2931 }
2932 
2933 static long btrfs_ioctl_default_subvol(struct file *file, void __user *argp)
2934 {
2935         struct inode *inode = file_inode(file);
2936         struct btrfs_root *root = BTRFS_I(inode)->root;
2937         struct btrfs_root *new_root;
2938         struct btrfs_dir_item *di;
2939         struct btrfs_trans_handle *trans;
2940         struct btrfs_path *path;
2941         struct btrfs_key location;
2942         struct btrfs_disk_key disk_key;
2943         u64 objectid = 0;
2944         u64 dir_id;
2945         int ret;
2946 
2947         if (!capable(CAP_SYS_ADMIN))
2948                 return -EPERM;
2949 
2950         ret = mnt_want_write_file(file);
2951         if (ret)
2952                 return ret;
2953 
2954         if (copy_from_user(&objectid, argp, sizeof(objectid))) {
2955                 ret = -EFAULT;
2956                 goto out;
2957         }
2958 
2959         if (!objectid)
2960                 objectid = root->root_key.objectid;
2961 
2962         location.objectid = objectid;
2963         location.type = BTRFS_ROOT_ITEM_KEY;
2964         location.offset = (u64)-1;
2965 
2966         new_root = btrfs_read_fs_root_no_name(root->fs_info, &location);
2967         if (IS_ERR(new_root)) {
2968                 ret = PTR_ERR(new_root);
2969                 goto out;
2970         }
2971 
2972         path = btrfs_alloc_path();
2973         if (!path) {
2974                 ret = -ENOMEM;
2975                 goto out;
2976         }
2977         path->leave_spinning = 1;
2978 
2979         trans = btrfs_start_transaction(root, 1);
2980         if (IS_ERR(trans)) {
2981                 btrfs_free_path(path);
2982                 ret = PTR_ERR(trans);
2983                 goto out;
2984         }
2985 
2986         dir_id = btrfs_super_root_dir(root->fs_info->super_copy);
2987         di = btrfs_lookup_dir_item(trans, root->fs_info->tree_root, path,
2988                                    dir_id, "default", 7, 1);
2989         if (IS_ERR_OR_NULL(di)) {
2990                 btrfs_free_path(path);
2991                 btrfs_end_transaction(trans, root);
2992                 printk(KERN_ERR "Umm, you don't have the default dir item, "
2993                        "this isn't going to work\n");
2994                 ret = -ENOENT;
2995                 goto out;
2996         }
2997 
2998         btrfs_cpu_key_to_disk(&disk_key, &new_root->root_key);
2999         btrfs_set_dir_item_key(path->nodes[0], di, &disk_key);
3000         btrfs_mark_buffer_dirty(path->nodes[0]);
3001         btrfs_free_path(path);
3002 
3003         btrfs_set_fs_incompat(root->fs_info, DEFAULT_SUBVOL);
3004         btrfs_end_transaction(trans, root);
3005 out:
3006         mnt_drop_write_file(file);
3007         return ret;
3008 }
3009 
3010 void btrfs_get_block_group_info(struct list_head *groups_list,
3011                                 struct btrfs_ioctl_space_info *space)
3012 {
3013         struct btrfs_block_group_cache *block_group;
3014 
3015         space->total_bytes = 0;
3016         space->used_bytes = 0;
3017         space->flags = 0;
3018         list_for_each_entry(block_group, groups_list, list) {
3019                 space->flags = block_group->flags;
3020                 space->total_bytes += block_group->key.offset;
3021                 space->used_bytes +=
3022                         btrfs_block_group_used(&block_group->item);
3023         }
3024 }
3025 
3026 static long btrfs_ioctl_space_info(struct btrfs_root *root, void __user *arg)
3027 {
3028         struct btrfs_ioctl_space_args space_args;
3029         struct btrfs_ioctl_space_info space;
3030         struct btrfs_ioctl_space_info *dest;
3031         struct btrfs_ioctl_space_info *dest_orig;
3032         struct btrfs_ioctl_space_info __user *user_dest;
3033         struct btrfs_space_info *info;
3034         u64 types[] = {BTRFS_BLOCK_GROUP_DATA,
3035                        BTRFS_BLOCK_GROUP_SYSTEM,
3036                        BTRFS_BLOCK_GROUP_METADATA,
3037                        BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA};
3038         int num_types = 4;
3039         int alloc_size;
3040         int ret = 0;
3041         u64 slot_count = 0;
3042         int i, c;
3043 
3044         if (copy_from_user(&space_args,
3045                            (struct btrfs_ioctl_space_args __user *)arg,
3046                            sizeof(space_args)))
3047                 return -EFAULT;
3048 
3049         for (i = 0; i < num_types; i++) {
3050                 struct btrfs_space_info *tmp;
3051 
3052                 info = NULL;
3053                 rcu_read_lock();
3054                 list_for_each_entry_rcu(tmp, &root->fs_info->space_info,
3055                                         list) {
3056                         if (tmp->flags == types[i]) {
3057                                 info = tmp;
3058                                 break;
3059                         }
3060                 }
3061                 rcu_read_unlock();
3062 
3063                 if (!info)
3064                         continue;
3065 
3066                 down_read(&info->groups_sem);
3067                 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
3068                         if (!list_empty(&info->block_groups[c]))
3069                                 slot_count++;
3070                 }
3071                 up_read(&info->groups_sem);
3072         }
3073 
3074         /* space_slots == 0 means they are asking for a count */
3075         if (space_args.space_slots == 0) {
3076                 space_args.total_spaces = slot_count;
3077                 goto out;
3078         }
3079 
3080         slot_count = min_t(u64, space_args.space_slots, slot_count);
3081 
3082         alloc_size = sizeof(*dest) * slot_count;
3083 
3084         /* we generally have at most 6 or so space infos, one for each raid
3085          * level.  So, a whole page should be more than enough for everyone
3086          */
3087         if (alloc_size > PAGE_CACHE_SIZE)
3088                 return -ENOMEM;
3089 
3090         space_args.total_spaces = 0;
3091         dest = kmalloc(alloc_size, GFP_NOFS);
3092         if (!dest)
3093                 return -ENOMEM;
3094         dest_orig = dest;
3095 
3096         /* now we have a buffer to copy into */
3097         for (i = 0; i < num_types; i++) {
3098                 struct btrfs_space_info *tmp;
3099 
3100                 if (!slot_count)
3101                         break;
3102 
3103                 info = NULL;
3104                 rcu_read_lock();
3105                 list_for_each_entry_rcu(tmp, &root->fs_info->space_info,
3106                                         list) {
3107                         if (tmp->flags == types[i]) {
3108                                 info = tmp;
3109                                 break;
3110                         }
3111                 }
3112                 rcu_read_unlock();
3113 
3114                 if (!info)
3115                         continue;
3116                 down_read(&info->groups_sem);
3117                 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
3118                         if (!list_empty(&info->block_groups[c])) {
3119                                 btrfs_get_block_group_info(
3120                                         &info->block_groups[c], &space);
3121                                 memcpy(dest, &space, sizeof(space));
3122                                 dest++;
3123                                 space_args.total_spaces++;
3124                                 slot_count--;
3125                         }
3126                         if (!slot_count)
3127                                 break;
3128                 }
3129                 up_read(&info->groups_sem);
3130         }
3131 
3132         user_dest = (struct btrfs_ioctl_space_info __user *)
3133                 (arg + sizeof(struct btrfs_ioctl_space_args));
3134 
3135         if (copy_to_user(user_dest, dest_orig, alloc_size))
3136                 ret = -EFAULT;
3137 
3138         kfree(dest_orig);
3139 out:
3140         if (ret == 0 && copy_to_user(arg, &space_args, sizeof(space_args)))
3141                 ret = -EFAULT;
3142 
3143         return ret;
3144 }
3145 
3146 /*
3147  * there are many ways the trans_start and trans_end ioctls can lead
3148  * to deadlocks.  They should only be used by applications that
3149  * basically own the machine, and have a very in depth understanding
3150  * of all the possible deadlocks and enospc problems.
3151  */
3152 long btrfs_ioctl_trans_end(struct file *file)
3153 {
3154         struct inode *inode = file_inode(file);
3155         struct btrfs_root *root = BTRFS_I(inode)->root;
3156         struct btrfs_trans_handle *trans;
3157 
3158         trans = file->private_data;
3159         if (!trans)
3160                 return -EINVAL;
3161         file->private_data = NULL;
3162 
3163         btrfs_end_transaction(trans, root);
3164 
3165         atomic_dec(&root->fs_info->open_ioctl_trans);
3166 
3167         mnt_drop_write_file(file);
3168         return 0;
3169 }
3170 
3171 static noinline long btrfs_ioctl_start_sync(struct btrfs_root *root,
3172                                             void __user *argp)
3173 {
3174         struct btrfs_trans_handle *trans;
3175         u64 transid;
3176         int ret;
3177 
3178         trans = btrfs_attach_transaction_barrier(root);
3179         if (IS_ERR(trans)) {
3180                 if (PTR_ERR(trans) != -ENOENT)
3181                         return PTR_ERR(trans);
3182 
3183                 /* No running transaction, don't bother */
3184                 transid = root->fs_info->last_trans_committed;
3185                 goto out;
3186         }
3187         transid = trans->transid;
3188         ret = btrfs_commit_transaction_async(trans, root, 0);
3189         if (ret) {
3190                 btrfs_end_transaction(trans, root);
3191                 return ret;
3192         }
3193 out:
3194         if (argp)
3195                 if (copy_to_user(argp, &transid, sizeof(transid)))
3196                         return -EFAULT;
3197         return 0;
3198 }
3199 
3200 static noinline long btrfs_ioctl_wait_sync(struct btrfs_root *root,
3201                                            void __user *argp)
3202 {
3203         u64 transid;
3204 
3205         if (argp) {
3206                 if (copy_from_user(&transid, argp, sizeof(transid)))
3207                         return -EFAULT;
3208         } else {
3209                 transid = 0;  /* current trans */
3210         }
3211         return btrfs_wait_for_commit(root, transid);
3212 }
3213 
3214 static long btrfs_ioctl_scrub(struct file *file, void __user *arg)
3215 {
3216         struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
3217         struct btrfs_ioctl_scrub_args *sa;
3218         int ret;
3219 
3220         if (!capable(CAP_SYS_ADMIN))
3221                 return -EPERM;
3222 
3223         sa = memdup_user(arg, sizeof(*sa));
3224         if (IS_ERR(sa))
3225                 return PTR_ERR(sa);
3226 
3227         if (!(sa->flags & BTRFS_SCRUB_READONLY)) {
3228                 ret = mnt_want_write_file(file);
3229                 if (ret)
3230                         goto out;
3231         }
3232 
3233         ret = btrfs_scrub_dev(root->fs_info, sa->devid, sa->start, sa->end,
3234                               &sa->progress, sa->flags & BTRFS_SCRUB_READONLY,
3235                               0);
3236 
3237         if (copy_to_user(arg, sa, sizeof(*sa)))
3238                 ret = -EFAULT;
3239 
3240         if (!(sa->flags & BTRFS_SCRUB_READONLY))
3241                 mnt_drop_write_file(file);
3242 out:
3243         kfree(sa);
3244         return ret;
3245 }
3246 
3247 static long btrfs_ioctl_scrub_cancel(struct btrfs_root *root, void __user *arg)
3248 {
3249         if (!capable(CAP_SYS_ADMIN))
3250                 return -EPERM;
3251 
3252         return btrfs_scrub_cancel(root->fs_info);
3253 }
3254 
3255 static long btrfs_ioctl_scrub_progress(struct btrfs_root *root,
3256                                        void __user *arg)
3257 {
3258         struct btrfs_ioctl_scrub_args *sa;
3259         int ret;
3260 
3261         if (!capable(CAP_SYS_ADMIN))
3262                 return -EPERM;
3263 
3264         sa = memdup_user(arg, sizeof(*sa));
3265         if (IS_ERR(sa))
3266                 return PTR_ERR(sa);
3267 
3268         ret = btrfs_scrub_progress(root, sa->devid, &sa->progress);
3269 
3270         if (copy_to_user(arg, sa, sizeof(*sa)))
3271                 ret = -EFAULT;
3272 
3273         kfree(sa);
3274         return ret;
3275 }
3276 
3277 static long btrfs_ioctl_get_dev_stats(struct btrfs_root *root,
3278                                       void __user *arg)
3279 {
3280         struct btrfs_ioctl_get_dev_stats *sa;
3281         int ret;
3282 
3283         sa = memdup_user(arg, sizeof(*sa));
3284         if (IS_ERR(sa))
3285                 return PTR_ERR(sa);
3286 
3287         if ((sa->flags & BTRFS_DEV_STATS_RESET) && !capable(CAP_SYS_ADMIN)) {
3288                 kfree(sa);
3289                 return -EPERM;
3290         }
3291 
3292         ret = btrfs_get_dev_stats(root, sa);
3293 
3294         if (copy_to_user(arg, sa, sizeof(*sa)))
3295                 ret = -EFAULT;
3296 
3297         kfree(sa);
3298         return ret;
3299 }
3300 
3301 static long btrfs_ioctl_dev_replace(struct btrfs_root *root, void __user *arg)
3302 {
3303         struct btrfs_ioctl_dev_replace_args *p;
3304         int ret;
3305 
3306         if (!capable(CAP_SYS_ADMIN))
3307                 return -EPERM;
3308 
3309         p = memdup_user(arg, sizeof(*p));
3310         if (IS_ERR(p))
3311                 return PTR_ERR(p);
3312 
3313         switch (p->cmd) {
3314         case BTRFS_IOCTL_DEV_REPLACE_CMD_START:
3315                 if (root->fs_info->sb->s_flags & MS_RDONLY)
3316                         return -EROFS;
3317 
3318                 if (atomic_xchg(
3319                         &root->fs_info->mutually_exclusive_operation_running,
3320                         1)) {
3321                         pr_info("btrfs: dev add/delete/balance/replace/resize operation in progress\n");
3322                         ret = -EINPROGRESS;
3323                 } else {
3324                         ret = btrfs_dev_replace_start(root, p);
3325                         atomic_set(
3326                          &root->fs_info->mutually_exclusive_operation_running,
3327                          0);
3328                 }
3329                 break;
3330         case BTRFS_IOCTL_DEV_REPLACE_CMD_STATUS:
3331                 btrfs_dev_replace_status(root->fs_info, p);
3332                 ret = 0;
3333                 break;
3334         case BTRFS_IOCTL_DEV_REPLACE_CMD_CANCEL:
3335                 ret = btrfs_dev_replace_cancel(root->fs_info, p);
3336                 break;
3337         default:
3338                 ret = -EINVAL;
3339                 break;
3340         }
3341 
3342         if (copy_to_user(arg, p, sizeof(*p)))
3343                 ret = -EFAULT;
3344 
3345         kfree(p);
3346         return ret;
3347 }
3348 
3349 static long btrfs_ioctl_ino_to_path(struct btrfs_root *root, void __user *arg)
3350 {
3351         int ret = 0;
3352         int i;
3353         u64 rel_ptr;
3354         int size;
3355         struct btrfs_ioctl_ino_path_args *ipa = NULL;
3356         struct inode_fs_paths *ipath = NULL;
3357         struct btrfs_path *path;
3358 
3359         if (!capable(CAP_DAC_READ_SEARCH))
3360                 return -EPERM;
3361 
3362         path = btrfs_alloc_path();
3363         if (!path) {
3364                 ret = -ENOMEM;
3365                 goto out;
3366         }
3367 
3368         ipa = memdup_user(arg, sizeof(*ipa));
3369         if (IS_ERR(ipa)) {
3370                 ret = PTR_ERR(ipa);
3371                 ipa = NULL;
3372                 goto out;
3373         }
3374 
3375         size = min_t(u32, ipa->size, 4096);
3376         ipath = init_ipath(size, root, path);
3377         if (IS_ERR(ipath)) {
3378                 ret = PTR_ERR(ipath);
3379                 ipath = NULL;
3380                 goto out;
3381         }
3382 
3383         ret = paths_from_inode(ipa->inum, ipath);
3384         if (ret < 0)
3385                 goto out;
3386 
3387         for (i = 0; i < ipath->fspath->elem_cnt; ++i) {
3388                 rel_ptr = ipath->fspath->val[i] -
3389                           (u64)(unsigned long)ipath->fspath->val;
3390                 ipath->fspath->val[i] = rel_ptr;
3391         }
3392 
3393         ret = copy_to_user((void *)(unsigned long)ipa->fspath,
3394                            (void *)(unsigned long)ipath->fspath, size);
3395         if (ret) {
3396                 ret = -EFAULT;
3397                 goto out;
3398         }
3399 
3400 out:
3401         btrfs_free_path(path);
3402         free_ipath(ipath);
3403         kfree(ipa);
3404 
3405         return ret;
3406 }
3407 
3408 static int build_ino_list(u64 inum, u64 offset, u64 root, void *ctx)
3409 {
3410         struct btrfs_data_container *inodes = ctx;
3411         const size_t c = 3 * sizeof(u64);
3412 
3413         if (inodes->bytes_left >= c) {
3414                 inodes->bytes_left -= c;
3415                 inodes->val[inodes->elem_cnt] = inum;
3416                 inodes->val[inodes->elem_cnt + 1] = offset;
3417                 inodes->val[inodes->elem_cnt + 2] = root;
3418                 inodes->elem_cnt += 3;
3419         } else {
3420                 inodes->bytes_missing += c - inodes->bytes_left;
3421                 inodes->bytes_left = 0;
3422                 inodes->elem_missed += 3;
3423         }
3424 
3425         return 0;
3426 }
3427 
3428 static long btrfs_ioctl_logical_to_ino(struct btrfs_root *root,
3429                                         void __user *arg)
3430 {
3431         int ret = 0;
3432         int size;
3433         struct btrfs_ioctl_logical_ino_args *loi;
3434         struct btrfs_data_container *inodes = NULL;
3435         struct btrfs_path *path = NULL;
3436 
3437         if (!capable(CAP_SYS_ADMIN))
3438                 return -EPERM;
3439 
3440         loi = memdup_user(arg, sizeof(*loi));
3441         if (IS_ERR(loi)) {
3442                 ret = PTR_ERR(loi);
3443                 loi = NULL;
3444                 goto out;
3445         }
3446 
3447         path = btrfs_alloc_path();
3448         if (!path) {
3449                 ret = -ENOMEM;
3450                 goto out;
3451         }
3452 
3453         size = min_t(u32, loi->size, 64 * 1024);
3454         inodes = init_data_container(size);
3455         if (IS_ERR(inodes)) {
3456                 ret = PTR_ERR(inodes);
3457                 inodes = NULL;
3458                 goto out;
3459         }
3460 
3461         ret = iterate_inodes_from_logical(loi->logical, root->fs_info, path,
3462                                           build_ino_list, inodes);
3463         if (ret == -EINVAL)
3464                 ret = -ENOENT;
3465         if (ret < 0)
3466                 goto out;
3467 
3468         ret = copy_to_user((void *)(unsigned long)loi->inodes,
3469                            (void *)(unsigned long)inodes, size);
3470         if (ret)
3471                 ret = -EFAULT;
3472 
3473 out:
3474         btrfs_free_path(path);
3475         vfree(inodes);
3476         kfree(loi);
3477 
3478         return ret;
3479 }
3480 
3481 void update_ioctl_balance_args(struct btrfs_fs_info *fs_info, int lock,
3482                                struct btrfs_ioctl_balance_args *bargs)
3483 {
3484         struct btrfs_balance_control *bctl = fs_info->balance_ctl;
3485 
3486         bargs->flags = bctl->flags;
3487 
3488         if (atomic_read(&fs_info->balance_running))
3489                 bargs->state |= BTRFS_BALANCE_STATE_RUNNING;
3490         if (atomic_read(&fs_info->balance_pause_req))
3491                 bargs->state |= BTRFS_BALANCE_STATE_PAUSE_REQ;
3492         if (atomic_read(&fs_info->balance_cancel_req))
3493                 bargs->state |= BTRFS_BALANCE_STATE_CANCEL_REQ;
3494 
3495         memcpy(&bargs->data, &bctl->data, sizeof(bargs->data));
3496         memcpy(&bargs->meta, &bctl->meta, sizeof(bargs->meta));
3497         memcpy(&bargs->sys, &bctl->sys, sizeof(bargs->sys));
3498 
3499         if (lock) {
3500                 spin_lock(&fs_info->balance_lock);
3501                 memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat));
3502                 spin_unlock(&fs_info->balance_lock);
3503         } else {
3504                 memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat));
3505         }
3506 }
3507 
3508 static long btrfs_ioctl_balance(struct file *file, void __user *arg)
3509 {
3510         struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
3511         struct btrfs_fs_info *fs_info = root->fs_info;
3512         struct btrfs_ioctl_balance_args *bargs;
3513         struct btrfs_balance_control *bctl;
3514         bool need_unlock; /* for mut. excl. ops lock */
3515         int ret;
3516 
3517         if (!capable(CAP_SYS_ADMIN))
3518                 return -EPERM;
3519 
3520         ret = mnt_want_write_file(file);
3521         if (ret)
3522                 return ret;
3523 
3524 again:
3525         if (!atomic_xchg(&fs_info->mutually_exclusive_operation_running, 1)) {
3526                 mutex_lock(&fs_info->volume_mutex);
3527                 mutex_lock(&fs_info->balance_mutex);
3528                 need_unlock = true;
3529                 goto locked;
3530         }
3531 
3532         /*
3533          * mut. excl. ops lock is locked.  Three possibilites:
3534          *   (1) some other op is running
3535          *   (2) balance is running
3536          *   (3) balance is paused -- special case (think resume)
3537          */
3538         mutex_lock(&fs_info->balance_mutex);
3539         if (fs_info->balance_ctl) {
3540                 /* this is either (2) or (3) */
3541                 if (!atomic_read(&fs_info->balance_running)) {
3542                         mutex_unlock(&fs_info->balance_mutex);
3543                         if (!mutex_trylock(&fs_info->volume_mutex))
3544                                 goto again;
3545                         mutex_lock(&fs_info->balance_mutex);
3546 
3547                         if (fs_info->balance_ctl &&
3548                             !atomic_read(&fs_info->balance_running)) {
3549                                 /* this is (3) */
3550                                 need_unlock = false;
3551                                 goto locked;
3552                         }
3553 
3554                         mutex_unlock(&fs_info->balance_mutex);
3555                         mutex_unlock(&fs_info->volume_mutex);
3556                         goto again;
3557                 } else {
3558                         /* this is (2) */
3559                         mutex_unlock(&fs_info->balance_mutex);
3560                         ret = -EINPROGRESS;
3561                         goto out;
3562                 }
3563         } else {
3564                 /* this is (1) */
3565                 mutex_unlock(&fs_info->balance_mutex);
3566                 pr_info("btrfs: dev add/delete/balance/replace/resize operation in progress\n");
3567                 ret = -EINVAL;
3568                 goto out;
3569         }
3570 
3571 locked:
3572         BUG_ON(!atomic_read(&fs_info->mutually_exclusive_operation_running));
3573 
3574         if (arg) {
3575                 bargs = memdup_user(arg, sizeof(*bargs));
3576                 if (IS_ERR(bargs)) {
3577                         ret = PTR_ERR(bargs);
3578                         goto out_unlock;
3579                 }
3580 
3581                 if (bargs->flags & BTRFS_BALANCE_RESUME) {
3582                         if (!fs_info->balance_ctl) {
3583                                 ret = -ENOTCONN;
3584                                 goto out_bargs;
3585                         }
3586 
3587                         bctl = fs_info->balance_ctl;
3588                         spin_lock(&fs_info->balance_lock);
3589                         bctl->flags |= BTRFS_BALANCE_RESUME;
3590                         spin_unlock(&fs_info->balance_lock);
3591 
3592                         goto do_balance;
3593                 }
3594         } else {
3595                 bargs = NULL;
3596         }
3597 
3598         if (fs_info->balance_ctl) {
3599                 ret = -EINPROGRESS;
3600                 goto out_bargs;
3601         }
3602 
3603         bctl = kzalloc(sizeof(*bctl), GFP_NOFS);
3604         if (!bctl) {
3605                 ret = -ENOMEM;
3606                 goto out_bargs;
3607         }
3608 
3609         bctl->fs_info = fs_info;
3610         if (arg) {
3611                 memcpy(&bctl->data, &bargs->data, sizeof(bctl->data));
3612                 memcpy(&bctl->meta, &bargs->meta, sizeof(bctl->meta));
3613                 memcpy(&bctl->sys, &bargs->sys, sizeof(bctl->sys));
3614 
3615                 bctl->flags = bargs->flags;
3616         } else {
3617                 /* balance everything - no filters */
3618                 bctl->flags |= BTRFS_BALANCE_TYPE_MASK;
3619         }
3620 
3621 do_balance:
3622         /*
3623          * Ownership of bctl and mutually_exclusive_operation_running
3624          * goes to to btrfs_balance.  bctl is freed in __cancel_balance,
3625          * or, if restriper was paused all the way until unmount, in
3626          * free_fs_info.  mutually_exclusive_operation_running is
3627          * cleared in __cancel_balance.
3628          */
3629         need_unlock = false;
3630 
3631         ret = btrfs_balance(bctl, bargs);
3632 
3633         if (arg) {
3634                 if (copy_to_user(arg, bargs, sizeof(*bargs)))
3635                         ret = -EFAULT;
3636         }
3637 
3638 out_bargs:
3639         kfree(bargs);
3640 out_unlock:
3641         mutex_unlock(&fs_info->balance_mutex);
3642         mutex_unlock(&fs_info->volume_mutex);
3643         if (need_unlock)
3644                 atomic_set(&fs_info->mutually_exclusive_operation_running, 0);
3645 out:
3646         mnt_drop_write_file(file);
3647         return ret;
3648 }
3649 
3650 static long btrfs_ioctl_balance_ctl(struct btrfs_root *root, int cmd)
3651 {
3652         if (!capable(CAP_SYS_ADMIN))
3653                 return -EPERM;
3654 
3655         switch (cmd) {
3656         case BTRFS_BALANCE_CTL_PAUSE:
3657                 return btrfs_pause_balance(root->fs_info);
3658         case BTRFS_BALANCE_CTL_CANCEL:
3659                 return btrfs_cancel_balance(root->fs_info);
3660         }
3661 
3662         return -EINVAL;
3663 }
3664 
3665 static long btrfs_ioctl_balance_progress(struct btrfs_root *root,
3666                                          void __user *arg)
3667 {
3668         struct btrfs_fs_info *fs_info = root->fs_info;
3669         struct btrfs_ioctl_balance_args *bargs;
3670         int ret = 0;
3671 
3672         if (!capable(CAP_SYS_ADMIN))
3673                 return -EPERM;
3674 
3675         mutex_lock(&fs_info->balance_mutex);
3676         if (!fs_info->balance_ctl) {
3677                 ret = -ENOTCONN;
3678                 goto out;
3679         }
3680 
3681         bargs = kzalloc(sizeof(*bargs), GFP_NOFS);
3682         if (!bargs) {
3683                 ret = -ENOMEM;
3684                 goto out;
3685         }
3686 
3687         update_ioctl_balance_args(fs_info, 1, bargs);
3688 
3689         if (copy_to_user(arg, bargs, sizeof(*bargs)))
3690                 ret = -EFAULT;
3691 
3692         kfree(bargs);
3693 out:
3694         mutex_unlock(&fs_info->balance_mutex);
3695         return ret;
3696 }
3697 
3698 static long btrfs_ioctl_quota_ctl(struct file *file, void __user *arg)
3699 {
3700         struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
3701         struct btrfs_ioctl_quota_ctl_args *sa;
3702         struct btrfs_trans_handle *trans = NULL;
3703         int ret;
3704         int err;
3705 
3706         if (!capable(CAP_SYS_ADMIN))
3707                 return -EPERM;
3708 
3709         ret = mnt_want_write_file(file);
3710         if (ret)
3711                 return ret;
3712 
3713         sa = memdup_user(arg, sizeof(*sa));
3714         if (IS_ERR(sa)) {
3715                 ret = PTR_ERR(sa);
3716                 goto drop_write;
3717         }
3718 
3719         down_write(&root->fs_info->subvol_sem);
3720         trans = btrfs_start_transaction(root->fs_info->tree_root, 2);
3721         if (IS_ERR(trans)) {
3722                 ret = PTR_ERR(trans);
3723                 goto out;
3724         }
3725 
3726         switch (sa->cmd) {
3727         case BTRFS_QUOTA_CTL_ENABLE:
3728                 ret = btrfs_quota_enable(trans, root->fs_info);
3729                 break;
3730         case BTRFS_QUOTA_CTL_DISABLE:
3731                 ret = btrfs_quota_disable(trans, root->fs_info);
3732                 break;
3733         default:
3734                 ret = -EINVAL;
3735                 break;
3736         }
3737 
3738         err = btrfs_commit_transaction(trans, root->fs_info->tree_root);
3739         if (err && !ret)
3740                 ret = err;
3741 out:
3742         kfree(sa);
3743         up_write(&root->fs_info->subvol_sem);
3744 drop_write:
3745         mnt_drop_write_file(file);
3746         return ret;
3747 }
3748 
3749 static long btrfs_ioctl_qgroup_assign(struct file *file, void __user *arg)
3750 {
3751         struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
3752         struct btrfs_ioctl_qgroup_assign_args *sa;
3753         struct btrfs_trans_handle *trans;
3754         int ret;
3755         int err;
3756 
3757         if (!capable(CAP_SYS_ADMIN))
3758                 return -EPERM;
3759 
3760         ret = mnt_want_write_file(file);
3761         if (ret)
3762                 return ret;
3763 
3764         sa = memdup_user(arg, sizeof(*sa));
3765         if (IS_ERR(sa)) {
3766                 ret = PTR_ERR(sa);
3767                 goto drop_write;
3768         }
3769 
3770         trans = btrfs_join_transaction(root);
3771         if (IS_ERR(trans)) {
3772                 ret = PTR_ERR(trans);
3773                 goto out;
3774         }
3775 
3776         /* FIXME: check if the IDs really exist */
3777         if (sa->assign) {
3778                 ret = btrfs_add_qgroup_relation(trans, root->fs_info,
3779                                                 sa->src, sa->dst);
3780         } else {
3781                 ret = btrfs_del_qgroup_relation(trans, root->fs_info,
3782                                                 sa->src, sa->dst);
3783         }
3784 
3785         err = btrfs_end_transaction(trans, root);
3786         if (err && !ret)
3787                 ret = err;
3788 
3789 out:
3790         kfree(sa);
3791 drop_write:
3792         mnt_drop_write_file(file);
3793         return ret;
3794 }
3795 
3796 static long btrfs_ioctl_qgroup_create(struct file *file, void __user *arg)
3797 {
3798         struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
3799         struct btrfs_ioctl_qgroup_create_args *sa;
3800         struct btrfs_trans_handle *trans;
3801         int ret;
3802         int err;
3803 
3804         if (!capable(CAP_SYS_ADMIN))
3805                 return -EPERM;
3806 
3807         ret = mnt_want_write_file(file);
3808         if (ret)
3809                 return ret;
3810 
3811         sa = memdup_user(arg, sizeof(*sa));
3812         if (IS_ERR(sa)) {
3813                 ret = PTR_ERR(sa);
3814                 goto drop_write;
3815         }
3816 
3817         if (!sa->qgroupid) {
3818                 ret = -EINVAL;
3819                 goto out;
3820         }
3821 
3822         trans = btrfs_join_transaction(root);
3823         if (IS_ERR(trans)) {
3824                 ret = PTR_ERR(trans);
3825                 goto out;
3826         }
3827 
3828         /* FIXME: check if the IDs really exist */
3829         if (sa->create) {
3830                 ret = btrfs_create_qgroup(trans, root->fs_info, sa->qgroupid,
3831                                           NULL);
3832         } else {
3833                 ret = btrfs_remove_qgroup(trans, root->fs_info, sa->qgroupid);
3834         }
3835 
3836         err = btrfs_end_transaction(trans, root);
3837         if (err && !ret)
3838                 ret = err;
3839 
3840 out:
3841         kfree(sa);
3842 drop_write:
3843         mnt_drop_write_file(file);
3844         return ret;
3845 }
3846 
3847 static long btrfs_ioctl_qgroup_limit(struct file *file, void __user *arg)
3848 {
3849         struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
3850         struct btrfs_ioctl_qgroup_limit_args *sa;
3851         struct btrfs_trans_handle *trans;
3852         int ret;
3853         int err;
3854         u64 qgroupid;
3855 
3856         if (!capable(CAP_SYS_ADMIN))
3857                 return -EPERM;
3858 
3859         ret = mnt_want_write_file(file);
3860         if (ret)
3861                 return ret;
3862 
3863         sa = memdup_user(arg, sizeof(*sa));
3864         if (IS_ERR(sa)) {
3865                 ret = PTR_ERR(sa);
3866                 goto drop_write;
3867         }
3868 
3869         trans = btrfs_join_transaction(root);
3870         if (IS_ERR(trans)) {
3871                 ret = PTR_ERR(trans);
3872                 goto out;
3873         }
3874 
3875         qgroupid = sa->qgroupid;
3876         if (!qgroupid) {
3877                 /* take the current subvol as qgroup */
3878                 qgroupid = root->root_key.objectid;
3879         }
3880 
3881         /* FIXME: check if the IDs really exist */
3882         ret = btrfs_limit_qgroup(trans, root->fs_info, qgroupid, &sa->lim);
3883 
3884         err = btrfs_end_transaction(trans, root);
3885         if (err && !ret)
3886                 ret = err;
3887 
3888 out:
3889         kfree(sa);
3890 drop_write:
3891         mnt_drop_write_file(file);
3892         return ret;
3893 }
3894 
3895 static long btrfs_ioctl_quota_rescan(struct file *file, void __user *arg)
3896 {
3897         struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
3898         struct btrfs_ioctl_quota_rescan_args *qsa;
3899         int ret;
3900 
3901         if (!capable(CAP_SYS_ADMIN))
3902                 return -EPERM;
3903 
3904         ret = mnt_want_write_file(file);
3905         if (ret)
3906                 return ret;
3907 
3908         qsa = memdup_user(arg, sizeof(*qsa));
3909         if (IS_ERR(qsa)) {
3910                 ret = PTR_ERR(qsa);
3911                 goto drop_write;
3912         }
3913 
3914         if (qsa->flags) {
3915                 ret = -EINVAL;
3916                 goto out;
3917         }
3918 
3919         ret = btrfs_qgroup_rescan(root->fs_info);
3920 
3921 out:
3922         kfree(qsa);
3923 drop_write:
3924         mnt_drop_write_file(file);
3925         return ret;
3926 }
3927 
3928 static long btrfs_ioctl_quota_rescan_status(struct file *file, void __user *arg)
3929 {
3930         struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
3931         struct btrfs_ioctl_quota_rescan_args *qsa;
3932         int ret = 0;
3933 
3934         if (!capable(CAP_SYS_ADMIN))
3935                 return -EPERM;
3936 
3937         qsa = kzalloc(sizeof(*qsa), GFP_NOFS);
3938         if (!qsa)
3939                 return -ENOMEM;
3940 
3941         if (root->fs_info->qgroup_flags & BTRFS_QGROUP_STATUS_FLAG_RESCAN) {
3942                 qsa->flags = 1;
3943                 qsa->progress = root->fs_info->qgroup_rescan_progress.objectid;
3944         }
3945 
3946         if (copy_to_user(arg, qsa, sizeof(*qsa)))
3947                 ret = -EFAULT;
3948 
3949         kfree(qsa);
3950         return ret;
3951 }
3952 
3953 static long btrfs_ioctl_quota_rescan_wait(struct file *file, void __user *arg)
3954 {
3955         struct btrfs_root *root = BTRFS_I(fdentry(file)->d_inode)->root;
3956 
3957         if (!capable(CAP_SYS_ADMIN))
3958                 return -EPERM;
3959 
3960         return btrfs_qgroup_wait_for_completion(root->fs_info);
3961 }
3962 
3963 static long btrfs_ioctl_set_received_subvol(struct file *file,
3964                                             void __user *arg)
3965 {
3966         struct btrfs_ioctl_received_subvol_args *sa = NULL;
3967         struct inode *inode = file_inode(file);
3968         struct btrfs_root *root = BTRFS_I(inode)->root;
3969         struct btrfs_root_item *root_item = &root->root_item;
3970         struct btrfs_trans_handle *trans;
3971         struct timespec ct = CURRENT_TIME;
3972         int ret = 0;
3973 
3974         ret = mnt_want_write_file(file);
3975         if (ret < 0)
3976                 return ret;
3977 
3978         down_write(&root->fs_info->subvol_sem);
3979 
3980         if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) {
3981                 ret = -EINVAL;
3982                 goto out;
3983         }
3984 
3985         if (btrfs_root_readonly(root)) {
3986                 ret = -EROFS;
3987                 goto out;
3988         }
3989 
3990         if (!inode_owner_or_capable(inode)) {
3991                 ret = -EACCES;
3992                 goto out;
3993         }
3994 
3995         sa = memdup_user(arg, sizeof(*sa));
3996         if (IS_ERR(sa)) {
3997                 ret = PTR_ERR(sa);
3998                 sa = NULL;
3999                 goto out;
4000         }
4001 
4002         trans = btrfs_start_transaction(root, 1);
4003         if (IS_ERR(trans)) {
4004                 ret = PTR_ERR(trans);
4005                 trans = NULL;
4006                 goto out;
4007         }
4008 
4009         sa->rtransid = trans->transid;
4010         sa->rtime.sec = ct.tv_sec;
4011         sa->rtime.nsec = ct.tv_nsec;
4012 
4013         memcpy(root_item->received_uuid, sa->uuid, BTRFS_UUID_SIZE);
4014         btrfs_set_root_stransid(root_item, sa->stransid);
4015         btrfs_set_root_rtransid(root_item, sa->rtransid);
4016         root_item->stime.sec = cpu_to_le64(sa->stime.sec);
4017         root_item->stime.nsec = cpu_to_le32(sa->stime.nsec);
4018         root_item->rtime.sec = cpu_to_le64(sa->rtime.sec);
4019         root_item->rtime.nsec = cpu_to_le32(sa->rtime.nsec);
4020 
4021         ret = btrfs_update_root(trans, root->fs_info->tree_root,
4022                                 &root->root_key, &root->root_item);
4023         if (ret < 0) {
4024                 btrfs_end_transaction(trans, root);
4025                 trans = NULL;
4026                 goto out;
4027         } else {
4028                 ret = btrfs_commit_transaction(trans, root);
4029                 if (ret < 0)
4030                         goto out;
4031         }
4032 
4033         ret = copy_to_user(arg, sa, sizeof(*sa));
4034         if (ret)
4035                 ret = -EFAULT;
4036 
4037 out:
4038         kfree(sa);
4039         up_write(&root->fs_info->subvol_sem);
4040         mnt_drop_write_file(file);
4041         return ret;
4042 }
4043 
4044 static int btrfs_ioctl_get_fslabel(struct file *file, void __user *arg)
4045 {
4046         struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4047         const char *label = root->fs_info->super_copy->label;
4048         size_t len = strnlen(label, BTRFS_LABEL_SIZE);
4049         int ret;
4050 
4051         if (len == BTRFS_LABEL_SIZE) {
4052                 pr_warn("btrfs: label is too long, return the first %zu bytes\n",
4053                         --len);
4054         }
4055 
4056         mutex_lock(&root->fs_info->volume_mutex);
4057         ret = copy_to_user(arg, label, len);
4058         mutex_unlock(&root->fs_info->volume_mutex);
4059 
4060         return ret ? -EFAULT : 0;
4061 }
4062 
4063 static int btrfs_ioctl_set_fslabel(struct file *file, void __user *arg)
4064 {
4065         struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4066         struct btrfs_super_block *super_block = root->fs_info->super_copy;
4067         struct btrfs_trans_handle *trans;
4068         char label[BTRFS_LABEL_SIZE];
4069         int ret;
4070 
4071         if (!capable(CAP_SYS_ADMIN))
4072                 return -EPERM;
4073 
4074         if (copy_from_user(label, arg, sizeof(label)))
4075                 return -EFAULT;
4076 
4077         if (strnlen(label, BTRFS_LABEL_SIZE) == BTRFS_LABEL_SIZE) {
4078                 pr_err("btrfs: unable to set label with more than %d bytes\n",
4079                        BTRFS_LABEL_SIZE - 1);
4080                 return -EINVAL;
4081         }
4082 
4083         ret = mnt_want_write_file(file);
4084         if (ret)
4085                 return ret;
4086 
4087         mutex_lock(&root->fs_info->volume_mutex);
4088         trans = btrfs_start_transaction(root, 0);
4089         if (IS_ERR(trans)) {
4090                 ret = PTR_ERR(trans);
4091                 goto out_unlock;
4092         }
4093 
4094         strcpy(super_block->label, label);
4095         ret = btrfs_end_transaction(trans, root);
4096 
4097 out_unlock:
4098         mutex_unlock(&root->fs_info->volume_mutex);
4099         mnt_drop_write_file(file);
4100         return ret;
4101 }
4102 
4103 long btrfs_ioctl(struct file *file, unsigned int
4104                 cmd, unsigned long arg)
4105 {
4106         struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4107         void __user *argp = (void __user *)arg;
4108 
4109         switch (cmd) {
4110         case FS_IOC_GETFLAGS:
4111                 return btrfs_ioctl_getflags(file, argp);
4112         case FS_IOC_SETFLAGS:
4113                 return btrfs_ioctl_setflags(file, argp);
4114         case FS_IOC_GETVERSION:
4115                 return btrfs_ioctl_getversion(file, argp);
4116         case FITRIM:
4117                 return btrfs_ioctl_fitrim(file, argp);
4118         case BTRFS_IOC_SNAP_CREATE:
4119                 return btrfs_ioctl_snap_create(file, argp, 0);
4120         case BTRFS_IOC_SNAP_CREATE_V2:
4121                 return btrfs_ioctl_snap_create_v2(file, argp, 0);
4122         case BTRFS_IOC_SUBVOL_CREATE:
4123                 return btrfs_ioctl_snap_create(file, argp, 1);
4124         case BTRFS_IOC_SUBVOL_CREATE_V2:
4125                 return btrfs_ioctl_snap_create_v2(file, argp, 1);
4126         case BTRFS_IOC_SNAP_DESTROY:
4127                 return btrfs_ioctl_snap_destroy(file, argp);
4128         case BTRFS_IOC_SUBVOL_GETFLAGS:
4129                 return btrfs_ioctl_subvol_getflags(file, argp);
4130         case BTRFS_IOC_SUBVOL_SETFLAGS:
4131                 return btrfs_ioctl_subvol_setflags(file, argp);
4132         case BTRFS_IOC_DEFAULT_SUBVOL:
4133                 return btrfs_ioctl_default_subvol(file, argp);
4134         case BTRFS_IOC_DEFRAG:
4135                 return btrfs_ioctl_defrag(file, NULL);
4136         case BTRFS_IOC_DEFRAG_RANGE:
4137                 return btrfs_ioctl_defrag(file, argp);
4138         case BTRFS_IOC_RESIZE:
4139                 return btrfs_ioctl_resize(file, argp);
4140         case BTRFS_IOC_ADD_DEV:
4141                 return btrfs_ioctl_add_dev(root, argp);
4142         case BTRFS_IOC_RM_DEV:
4143                 return btrfs_ioctl_rm_dev(file, argp);
4144         case BTRFS_IOC_FS_INFO:
4145                 return btrfs_ioctl_fs_info(root, argp);
4146         case BTRFS_IOC_DEV_INFO:
4147                 return btrfs_ioctl_dev_info(root, argp);
4148         case BTRFS_IOC_BALANCE:
4149                 return btrfs_ioctl_balance(file, NULL);
4150         case BTRFS_IOC_CLONE:
4151                 return btrfs_ioctl_clone(file, arg, 0, 0, 0);
4152         case BTRFS_IOC_CLONE_RANGE:
4153                 return btrfs_ioctl_clone_range(file, argp);
4154         case BTRFS_IOC_TRANS_START:
4155                 return btrfs_ioctl_trans_start(file);
4156         case BTRFS_IOC_TRANS_END:
4157                 return btrfs_ioctl_trans_end(file);
4158         case BTRFS_IOC_TREE_SEARCH:
4159                 return btrfs_ioctl_tree_search(file, argp);
4160         case BTRFS_IOC_INO_LOOKUP:
4161                 return btrfs_ioctl_ino_lookup(file, argp);
4162         case BTRFS_IOC_INO_PATHS:
4163                 return btrfs_ioctl_ino_to_path(root, argp);
4164         case BTRFS_IOC_LOGICAL_INO:
4165                 return btrfs_ioctl_logical_to_ino(root, argp);
4166         case BTRFS_IOC_SPACE_INFO:
4167                 return btrfs_ioctl_space_info(root, argp);
4168         case BTRFS_IOC_SYNC:
4169                 btrfs_sync_fs(file->f_dentry->d_sb, 1);
4170                 return 0;
4171         case BTRFS_IOC_START_SYNC:
4172                 return btrfs_ioctl_start_sync(root, argp);
4173         case BTRFS_IOC_WAIT_SYNC:
4174                 return btrfs_ioctl_wait_sync(root, argp);
4175         case BTRFS_IOC_SCRUB:
4176                 return btrfs_ioctl_scrub(file, argp);
4177         case BTRFS_IOC_SCRUB_CANCEL:
4178                 return btrfs_ioctl_scrub_cancel(root, argp);
4179         case BTRFS_IOC_SCRUB_PROGRESS:
4180                 return btrfs_ioctl_scrub_progress(root, argp);
4181         case BTRFS_IOC_BALANCE_V2:
4182                 return btrfs_ioctl_balance(file, argp);
4183         case BTRFS_IOC_BALANCE_CTL:
4184                 return btrfs_ioctl_balance_ctl(root, arg);
4185         case BTRFS_IOC_BALANCE_PROGRESS:
4186                 return btrfs_ioctl_balance_progress(root, argp);
4187         case BTRFS_IOC_SET_RECEIVED_SUBVOL:
4188                 return btrfs_ioctl_set_received_subvol(file, argp);
4189         case BTRFS_IOC_SEND:
4190                 return btrfs_ioctl_send(file, argp);
4191         case BTRFS_IOC_GET_DEV_STATS:
4192                 return btrfs_ioctl_get_dev_stats(root, argp);
4193         case BTRFS_IOC_QUOTA_CTL:
4194                 return btrfs_ioctl_quota_ctl(file, argp);
4195         case BTRFS_IOC_QGROUP_ASSIGN:
4196                 return btrfs_ioctl_qgroup_assign(file, argp);
4197         case BTRFS_IOC_QGROUP_CREATE:
4198                 return btrfs_ioctl_qgroup_create(file, argp);
4199         case BTRFS_IOC_QGROUP_LIMIT:
4200                 return btrfs_ioctl_qgroup_limit(file, argp);
4201         case BTRFS_IOC_QUOTA_RESCAN:
4202                 return btrfs_ioctl_quota_rescan(file, argp);
4203         case BTRFS_IOC_QUOTA_RESCAN_STATUS:
4204                 return btrfs_ioctl_quota_rescan_status(file, argp);
4205         case BTRFS_IOC_QUOTA_RESCAN_WAIT:
4206                 return btrfs_ioctl_quota_rescan_wait(file, argp);
4207         case BTRFS_IOC_DEV_REPLACE:
4208                 return btrfs_ioctl_dev_replace(root, argp);
4209         case BTRFS_IOC_GET_FSLABEL:
4210                 return btrfs_ioctl_get_fslabel(file, argp);
4211         case BTRFS_IOC_SET_FSLABEL:
4212                 return btrfs_ioctl_set_fslabel(file, argp);
4213         }
4214 
4215         return -ENOTTY;
4216 }
4217 

~ [ source navigation ] ~ [ diff markup ] ~ [ identifier search ] ~

kernel.org | git.kernel.org | LWN.net | Project Home | Wiki (Japanese) | Wiki (English) | SVN repository | Mail admin

Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.

osdn.jp