fs/btrfs/ioctl.c

   1 /*
   2  * Copyright (C) 2007 Oracle.  All rights reserved.
   3  *
   4  * This program is free software; you can redistribute it and/or
   5  * modify it under the terms of the GNU General Public
   6  * License v2 as published by the Free Software Foundation.
   7  *
   8  * This program is distributed in the hope that it will be useful,
   9  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  10  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  11  * General Public License for more details.
  12  *
  13  * You should have received a copy of the GNU General Public
  14  * License along with this program; if not, write to the
  15  * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
  16  * Boston, MA 021110-1307, USA.
  17  */
  18
  19 #include <linux/kernel.h>
  20 #include <linux/bio.h>
  21 #include <linux/buffer_head.h>
  22 #include <linux/file.h>
  23 #include <linux/fs.h>
  24 #include <linux/fsnotify.h>
  25 #include <linux/pagemap.h>
  26 #include <linux/highmem.h>
  27 #include <linux/time.h>
  28 #include <linux/init.h>
  29 #include <linux/string.h>
  30 #include <linux/backing-dev.h>
  31 #include <linux/mount.h>
  32 #include <linux/mpage.h>
  33 #include <linux/namei.h>
  34 #include <linux/swap.h>
  35 #include <linux/writeback.h>
  36 #include <linux/statfs.h>
  37 #include <linux/compat.h>
  38 #include <linux/bit_spinlock.h>
  39 #include <linux/security.h>
  40 #include <linux/xattr.h>
  41 #include <linux/vmalloc.h>
  42 #include <linux/slab.h>
  43 #include <linux/blkdev.h>
  44 #include "compat.h"
  45 #include "ctree.h"
  46 #include "disk-io.h"
  47 #include "transaction.h"
  48 #include "btrfs_inode.h"
  49 #include "ioctl.h"
  50 #include "print-tree.h"
  51 #include "volumes.h"
  52 #include "locking.h"
  53 #include "inode-map.h"
  54 #include "backref.h"
  55
  56 /* Mask out flags that are inappropriate for the given type of inode. */
  57 static inline __u32 btrfs_mask_flags(umode_t mode, __u32 flags)
  58 {
  59         if (S_ISDIR(mode))
  60                 return flags;
  61         else if (S_ISREG(mode))
  62                 return flags & ~FS_DIRSYNC_FL;
  63         else
  64                 return flags & (FS_NODUMP_FL | FS_NOATIME_FL);
  65 }
  66
  67 /*
  68  * Export inode flags to the format expected by the FS_IOC_GETFLAGS ioctl.
  69  */
  70 static unsigned int btrfs_flags_to_ioctl(unsigned int flags)
  71 {
  72         unsigned int iflags = 0;
  73
  74         if (flags & BTRFS_INODE_SYNC)
  75                 iflags |= FS_SYNC_FL;
  76         if (flags & BTRFS_INODE_IMMUTABLE)
  77                 iflags |= FS_IMMUTABLE_FL;
  78         if (flags & BTRFS_INODE_APPEND)
  79                 iflags |= FS_APPEND_FL;
  80         if (flags & BTRFS_INODE_NODUMP)
  81                 iflags |= FS_NODUMP_FL;
  82         if (flags & BTRFS_INODE_NOATIME)
  83                 iflags |= FS_NOATIME_FL;
  84         if (flags & BTRFS_INODE_DIRSYNC)
  85                 iflags |= FS_DIRSYNC_FL;
  86         if (flags & BTRFS_INODE_NODATACOW)
  87                 iflags |= FS_NOCOW_FL;
  88
  89         if ((flags & BTRFS_INODE_COMPRESS) && !(flags & BTRFS_INODE_NOCOMPRESS))
  90                 iflags |= FS_COMPR_FL;
  91         else if (flags & BTRFS_INODE_NOCOMPRESS)
  92                 iflags |= FS_NOCOMP_FL;
  93
  94         return iflags;
  95 }
  96
  97 /*
  98  * Update inode->i_flags based on the btrfs internal flags.
  99  */
 100 void btrfs_update_iflags(struct inode *inode)
 101 {
 102         struct btrfs_inode *ip = BTRFS_I(inode);
 103
 104         inode->i_flags &= ~(S_SYNC|S_APPEND|S_IMMUTABLE|S_NOATIME|S_DIRSYNC);
 105
 106         if (ip->flags & BTRFS_INODE_SYNC)
 107                 inode->i_flags |= S_SYNC;
 108         if (ip->flags & BTRFS_INODE_IMMUTABLE)
 109                 inode->i_flags |= S_IMMUTABLE;
 110         if (ip->flags & BTRFS_INODE_APPEND)
 111                 inode->i_flags |= S_APPEND;
 112         if (ip->flags & BTRFS_INODE_NOATIME)
 113                 inode->i_flags |= S_NOATIME;
 114         if (ip->flags & BTRFS_INODE_DIRSYNC)
 115                 inode->i_flags |= S_DIRSYNC;
 116 }
 117
 118 /*
 119  * Inherit flags from the parent inode.
 120  *
 121  * Currently only the compression flags and the cow flags are inherited.
 122  */
 123 void btrfs_inherit_iflags(struct inode *inode, struct inode *dir)
 124 {
 125         unsigned int flags;
 126
 127         if (!dir)
 128                 return;
 129
 130         flags = BTRFS_I(dir)->flags;
 131
 132         if (flags & BTRFS_INODE_NOCOMPRESS) {
 133                 BTRFS_I(inode)->flags &= ~BTRFS_INODE_COMPRESS;
 134                 BTRFS_I(inode)->flags |= BTRFS_INODE_NOCOMPRESS;
 135         } else if (flags & BTRFS_INODE_COMPRESS) {
 136                 BTRFS_I(inode)->flags &= ~BTRFS_INODE_NOCOMPRESS;
 137                 BTRFS_I(inode)->flags |= BTRFS_INODE_COMPRESS;
 138         }
 139
 140         if (flags & BTRFS_INODE_NODATACOW)
 141                 BTRFS_I(inode)->flags |= BTRFS_INODE_NODATACOW;
 142
 143         btrfs_update_iflags(inode);
 144 }
 145
 146 static int btrfs_ioctl_getflags(struct file *file, void __user *arg)
 147 {
 148         struct btrfs_inode *ip = BTRFS_I(file->f_path.dentry->d_inode);
 149         unsigned int flags = btrfs_flags_to_ioctl(ip->flags);
 150
 151         if (copy_to_user(arg, &flags, sizeof(flags)))
 152                 return -EFAULT;
 153         return 0;
 154 }
 155
 156 static int check_flags(unsigned int flags)
 157 {
 158         if (flags & ~(FS_IMMUTABLE_FL | FS_APPEND_FL | \
 159                       FS_NOATIME_FL | FS_NODUMP_FL | \
 160                       FS_SYNC_FL | FS_DIRSYNC_FL | \
 161                       FS_NOCOMP_FL | FS_COMPR_FL |
 162                       FS_NOCOW_FL))
 163                 return -EOPNOTSUPP;
 164
 165         if ((flags & FS_NOCOMP_FL) && (flags & FS_COMPR_FL))
 166                 return -EINVAL;
 167
 168         return 0;
 169 }
 170
 171 static int btrfs_ioctl_setflags(struct file *file, void __user *arg)
 172 {
 173         struct inode *inode = file->f_path.dentry->d_inode;
 174         struct btrfs_inode *ip = BTRFS_I(inode);
 175         struct btrfs_root *root = ip->root;
 176         struct btrfs_trans_handle *trans;
 177         unsigned int flags, oldflags;
 178         int ret;
 179         u64 ip_oldflags;
 180         unsigned int i_oldflags;
 181
 182         if (btrfs_root_readonly(root))
 183                 return -EROFS;
 184
 185         if (copy_from_user(&flags, arg, sizeof(flags)))
 186                 return -EFAULT;
 187
 188         ret = check_flags(flags);
 189         if (ret)
 190                 return ret;
 191
 192         if (!inode_owner_or_capable(inode))
 193                 return -EACCES;
 194
 195         mutex_lock(&inode->i_mutex);
 196
 197         ip_oldflags = ip->flags;
 198         i_oldflags = inode->i_flags;
 199
 200         flags = btrfs_mask_flags(inode->i_mode, flags);
 201         oldflags = btrfs_flags_to_ioctl(ip->flags);
 202         if ((flags ^ oldflags) & (FS_APPEND_FL | FS_IMMUTABLE_FL)) {
 203                 if (!capable(CAP_LINUX_IMMUTABLE)) {
 204                         ret = -EPERM;
 205                         goto out_unlock;
 206                 }
 207         }
 208
 209         ret = mnt_want_write_file(file);
 210         if (ret)
 211                 goto out_unlock;
 212
 213         if (flags & FS_SYNC_FL)
 214                 ip->flags |= BTRFS_INODE_SYNC;
 215         else
 216                 ip->flags &= ~BTRFS_INODE_SYNC;
 217         if (flags & FS_IMMUTABLE_FL)
 218                 ip->flags |= BTRFS_INODE_IMMUTABLE;
 219         else
 220                 ip->flags &= ~BTRFS_INODE_IMMUTABLE;
 221         if (flags & FS_APPEND_FL)
 222                 ip->flags |= BTRFS_INODE_APPEND;
 223         else
 224                 ip->flags &= ~BTRFS_INODE_APPEND;
 225         if (flags & FS_NODUMP_FL)
 226                 ip->flags |= BTRFS_INODE_NODUMP;
 227         else
 228                 ip->flags &= ~BTRFS_INODE_NODUMP;
 229         if (flags & FS_NOATIME_FL)
 230                 ip->flags |= BTRFS_INODE_NOATIME;
 231         else
 232                 ip->flags &= ~BTRFS_INODE_NOATIME;
 233         if (flags & FS_DIRSYNC_FL)
 234                 ip->flags |= BTRFS_INODE_DIRSYNC;
 235         else
 236                 ip->flags &= ~BTRFS_INODE_DIRSYNC;
 237         if (flags & FS_NOCOW_FL)
 238                 ip->flags |= BTRFS_INODE_NODATACOW;
 239         else
 240                 ip->flags &= ~BTRFS_INODE_NODATACOW;
 241
 242         /*
 243          * The COMPRESS flag can only be changed by users, while the NOCOMPRESS
 244          * flag may be changed automatically if compression code won't make
 245          * things smaller.
 246          */
 247         if (flags & FS_NOCOMP_FL) {
 248                 ip->flags &= ~BTRFS_INODE_COMPRESS;
 249                 ip->flags |= BTRFS_INODE_NOCOMPRESS;
 250         } else if (flags & FS_COMPR_FL) {
 251                 ip->flags |= BTRFS_INODE_COMPRESS;
 252                 ip->flags &= ~BTRFS_INODE_NOCOMPRESS;
 253         } else {
 254                 ip->flags &= ~(BTRFS_INODE_COMPRESS | BTRFS_INODE_NOCOMPRESS);
 255         }
 256
 257         trans = btrfs_start_transaction(root, 1);
 258         if (IS_ERR(trans)) {
 259                 ret = PTR_ERR(trans);
 260                 goto out_drop;
 261         }
 262
 263         btrfs_update_iflags(inode);
 264         inode->i_ctime = CURRENT_TIME;
 265         ret = btrfs_update_inode(trans, root, inode);
 266
 267         btrfs_end_transaction(trans, root);
 268  out_drop:
 269         if (ret) {
 270                 ip->flags = ip_oldflags;
 271                 inode->i_flags = i_oldflags;
 272         }
 273
 274         mnt_drop_write_file(file);
 275  out_unlock:
 276         mutex_unlock(&inode->i_mutex);
 277         return ret;
 278 }
 279
 280 static int btrfs_ioctl_getversion(struct file *file, int __user *arg)
 281 {
 282         struct inode *inode = file->f_path.dentry->d_inode;
 283
 284         return put_user(inode->i_generation, arg);
 285 }
 286
 287 static noinline int btrfs_ioctl_fitrim(struct file *file, void __user *arg)
 288 {
 289         struct btrfs_fs_info *fs_info = btrfs_sb(fdentry(file)->d_sb);
 290         struct btrfs_device *device;
 291         struct request_queue *q;
 292         struct fstrim_range range;
 293         u64 minlen = ULLONG_MAX;
 294         u64 num_devices = 0;
 295         u64 total_bytes = btrfs_super_total_bytes(fs_info->super_copy);
 296         int ret;
 297
 298         if (!capable(CAP_SYS_ADMIN))
 299                 return -EPERM;
 300
 301         rcu_read_lock();
 302         list_for_each_entry_rcu(device, &fs_info->fs_devices->devices,
 303                                 dev_list) {
 304                 if (!device->bdev)
 305                         continue;
 306                 q = bdev_get_queue(device->bdev);
 307                 if (blk_queue_discard(q)) {
 308                         num_devices++;
 309                         minlen = min((u64)q->limits.discard_granularity,
 310                                      minlen);
 311                 }
 312         }
 313         rcu_read_unlock();
 314
 315         if (!num_devices)
 316                 return -EOPNOTSUPP;
 317         if (copy_from_user(&range, arg, sizeof(range)))
 318                 return -EFAULT;
 319         if (range.start > total_bytes)
 320                 return -EINVAL;
 321
 322         range.len = min(range.len, total_bytes - range.start);
 323         range.minlen = max(range.minlen, minlen);
 324         ret = btrfs_trim_fs(fs_info->tree_root, &range);
 325         if (ret < 0)
 326                 return ret;
 327
 328         if (copy_to_user(arg, &range, sizeof(range)))
 329                 return -EFAULT;
 330
 331         return 0;
 332 }
 333
 334 static noinline int create_subvol(struct btrfs_root *root,
 335                                   struct dentry *dentry,
 336                                   char *name, int namelen,
 337                                   u64 *async_transid)
 338 {
 339         struct btrfs_trans_handle *trans;
 340         struct btrfs_key key;
 341         struct btrfs_root_item root_item;
 342         struct btrfs_inode_item *inode_item;
 343         struct extent_buffer *leaf;
 344         struct btrfs_root *new_root;
 345         struct dentry *parent = dentry->d_parent;
 346         struct inode *dir;
 347         int ret;
 348         int err;
 349         u64 objectid;
 350         u64 new_dirid = BTRFS_FIRST_FREE_OBJECTID;
 351         u64 index = 0;
 352
 353         ret = btrfs_find_free_objectid(root->fs_info->tree_root, &objectid);
 354         if (ret)
 355                 return ret;
 356
 357         dir = parent->d_inode;
 358
 359         /*
 360          * 1 - inode item
 361          * 2 - refs
 362          * 1 - root item
 363          * 2 - dir items
 364          */
 365         trans = btrfs_start_transaction(root, 6);
 366         if (IS_ERR(trans))
 367                 return PTR_ERR(trans);
 368
 369         leaf = btrfs_alloc_free_block(trans, root, root->leafsize,
 370                                       0, objectid, NULL, 0, 0, 0, 0);
 371         if (IS_ERR(leaf)) {
 372                 ret = PTR_ERR(leaf);
 373                 goto fail;
 374         }
 375
 376         memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header));
 377         btrfs_set_header_bytenr(leaf, leaf->start);
 378         btrfs_set_header_generation(leaf, trans->transid);
 379         btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV);
 380         btrfs_set_header_owner(leaf, objectid);
 381
 382         write_extent_buffer(leaf, root->fs_info->fsid,
 383                             (unsigned long)btrfs_header_fsid(leaf),
 384                             BTRFS_FSID_SIZE);
 385         write_extent_buffer(leaf, root->fs_info->chunk_tree_uuid,
 386                             (unsigned long)btrfs_header_chunk_tree_uuid(leaf),
 387                             BTRFS_UUID_SIZE);
 388         btrfs_mark_buffer_dirty(leaf);
 389
 390         inode_item = &root_item.inode;
 391         memset(inode_item, 0, sizeof(*inode_item));
 392         inode_item->generation = cpu_to_le64(1);
 393         inode_item->size = cpu_to_le64(3);
 394         inode_item->nlink = cpu_to_le32(1);
 395         inode_item->nbytes = cpu_to_le64(root->leafsize);
 396         inode_item->mode = cpu_to_le32(S_IFDIR | 0755);
 397
 398         root_item.flags = 0;
 399         root_item.byte_limit = 0;
 400         inode_item->flags = cpu_to_le64(BTRFS_INODE_ROOT_ITEM_INIT);
 401
 402         btrfs_set_root_bytenr(&root_item, leaf->start);
 403         btrfs_set_root_generation(&root_item, trans->transid);
 404         btrfs_set_root_level(&root_item, 0);
 405         btrfs_set_root_refs(&root_item, 1);
 406         btrfs_set_root_used(&root_item, leaf->len);
 407         btrfs_set_root_last_snapshot(&root_item, 0);
 408
 409         memset(&root_item.drop_progress, 0, sizeof(root_item.drop_progress));
 410         root_item.drop_level = 0;
 411
 412         btrfs_tree_unlock(leaf);
 413         free_extent_buffer(leaf);
 414         leaf = NULL;
 415
 416         btrfs_set_root_dirid(&root_item, new_dirid);
 417
 418         key.objectid = objectid;
 419         key.offset = 0;
 420         btrfs_set_key_type(&key, BTRFS_ROOT_ITEM_KEY);
 421         ret = btrfs_insert_root(trans, root->fs_info->tree_root, &key,
 422                                 &root_item);
 423         if (ret)
 424                 goto fail;
 425
 426         key.offset = (u64)-1;
 427         new_root = btrfs_read_fs_root_no_name(root->fs_info, &key);
 428         if (IS_ERR(new_root)) {
 429                 btrfs_abort_transaction(trans, root, PTR_ERR(new_root));
 430                 ret = PTR_ERR(new_root);
 431                 goto fail;
 432         }
 433
 434         btrfs_record_root_in_trans(trans, new_root);
 435
 436         ret = btrfs_create_subvol_root(trans, new_root, new_dirid);
 437         if (ret) {
 438                 /* We potentially lose an unused inode item here */
 439                 btrfs_abort_transaction(trans, root, ret);
 440                 goto fail;
 441         }
 442
 443         /*
 444          * insert the directory item
 445          */
 446         ret = btrfs_set_inode_index(dir, &index);
 447         if (ret) {
 448                 btrfs_abort_transaction(trans, root, ret);
 449                 goto fail;
 450         }
 451
 452         ret = btrfs_insert_dir_item(trans, root,
 453                                     name, namelen, dir, &key,
 454                                     BTRFS_FT_DIR, index);
 455         if (ret) {
 456                 btrfs_abort_transaction(trans, root, ret);
 457                 goto fail;
 458         }
 459
 460         btrfs_i_size_write(dir, dir->i_size + namelen * 2);
 461         ret = btrfs_update_inode(trans, root, dir);
 462         BUG_ON(ret);
 463
 464         ret = btrfs_add_root_ref(trans, root->fs_info->tree_root,
 465                                  objectid, root->root_key.objectid,
 466                                  btrfs_ino(dir), index, name, namelen);
 467
 468         BUG_ON(ret);
 469
 470         d_instantiate(dentry, btrfs_lookup_dentry(dir, dentry));
 471 fail:
 472         if (async_transid) {
 473                 *async_transid = trans->transid;
 474                 err = btrfs_commit_transaction_async(trans, root, 1);
 475         } else {
 476                 err = btrfs_commit_transaction(trans, root);
 477         }
 478         if (err && !ret)
 479                 ret = err;
 480         return ret;
 481 }
 482
 483 static int create_snapshot(struct btrfs_root *root, struct dentry *dentry,
 484                            char *name, int namelen, u64 *async_transid,
 485                            bool readonly)
 486 {
 487         struct inode *inode;
 488         struct btrfs_pending_snapshot *pending_snapshot;
 489         struct btrfs_trans_handle *trans;
 490         int ret;
 491
 492         if (!root->ref_cows)
 493                 return -EINVAL;
 494
 495         pending_snapshot = kzalloc(sizeof(*pending_snapshot), GFP_NOFS);
 496         if (!pending_snapshot)
 497                 return -ENOMEM;
 498
 499         btrfs_init_block_rsv(&pending_snapshot->block_rsv);
 500         pending_snapshot->dentry = dentry;
 501         pending_snapshot->root = root;
 502         pending_snapshot->readonly = readonly;
 503
 504         trans = btrfs_start_transaction(root->fs_info->extent_root, 5);
 505         if (IS_ERR(trans)) {
 506                 ret = PTR_ERR(trans);
 507                 goto fail;
 508         }
 509
 510         ret = btrfs_snap_reserve_metadata(trans, pending_snapshot);
 511         BUG_ON(ret);
 512
 513         spin_lock(&root->fs_info->trans_lock);
 514         list_add(&pending_snapshot->list,
 515                  &trans->transaction->pending_snapshots);
 516         spin_unlock(&root->fs_info->trans_lock);
 517         if (async_transid) {
 518                 *async_transid = trans->transid;
 519                 ret = btrfs_commit_transaction_async(trans,
 520                                      root->fs_info->extent_root, 1);
 521         } else {
 522                 ret = btrfs_commit_transaction(trans,
 523                                                root->fs_info->extent_root);
 524         }
 525         BUG_ON(ret);
 526
 527         ret = pending_snapshot->error;
 528         if (ret)
 529                 goto fail;
 530
 531         ret = btrfs_orphan_cleanup(pending_snapshot->snap);
 532         if (ret)
 533                 goto fail;
 534
 535         inode = btrfs_lookup_dentry(dentry->d_parent->d_inode, dentry);
 536         if (IS_ERR(inode)) {
 537                 ret = PTR_ERR(inode);
 538                 goto fail;
 539         }
 540         BUG_ON(!inode);
 541         d_instantiate(dentry, inode);
 542         ret = 0;
 543 fail:
 544         kfree(pending_snapshot);
 545         return ret;
 546 }
 547
 548 /*  copy of check_sticky in fs/namei.c()
 549 * It's inline, so penalty for filesystems that don't use sticky bit is
 550 * minimal.
 551 */
 552 static inline int btrfs_check_sticky(struct inode *dir, struct inode *inode)
 553 {
 554         uid_t fsuid = current_fsuid();
 555
 556         if (!(dir->i_mode & S_ISVTX))
 557                 return 0;
 558         if (inode->i_uid == fsuid)
 559                 return 0;
 560         if (dir->i_uid == fsuid)
 561                 return 0;
 562         return !capable(CAP_FOWNER);
 563 }
 564
 565 /*  copy of may_delete in fs/namei.c()
 566  *      Check whether we can remove a link victim from directory dir, check
 567  *  whether the type of victim is right.
 568  *  1. We can't do it if dir is read-only (done in permission())
 569  *  2. We should have write and exec permissions on dir
 570  *  3. We can't remove anything from append-only dir
 571  *  4. We can't do anything with immutable dir (done in permission())
 572  *  5. If the sticky bit on dir is set we should either
 573  *      a. be owner of dir, or
 574  *      b. be owner of victim, or
 575  *      c. have CAP_FOWNER capability
 576  *  6. If the victim is append-only or immutable we can't do antyhing with
 577  *     links pointing to it.
 578  *  7. If we were asked to remove a directory and victim isn't one - ENOTDIR.
 579  *  8. If we were asked to remove a non-directory and victim isn't one - EISDIR.
 580  *  9. We can't remove a root or mountpoint.
 581  * 10. We don't allow removal of NFS sillyrenamed files; it's handled by
 582  *     nfs_async_unlink().
 583  */
 584
 585 static int btrfs_may_delete(struct inode *dir,struct dentry *victim,int isdir)
 586 {
 587         int error;
 588
 589         if (!victim->d_inode)
 590                 return -ENOENT;
 591
 592         BUG_ON(victim->d_parent->d_inode != dir);
 593         audit_inode_child(victim, dir);
 594
 595         error = inode_permission(dir, MAY_WRITE | MAY_EXEC);
 596         if (error)
 597                 return error;
 598         if (IS_APPEND(dir))
 599                 return -EPERM;
 600         if (btrfs_check_sticky(dir, victim->d_inode)||
 601                 IS_APPEND(victim->d_inode)||
 602             IS_IMMUTABLE(victim->d_inode) || IS_SWAPFILE(victim->d_inode))
 603                 return -EPERM;
 604         if (isdir) {
 605                 if (!S_ISDIR(victim->d_inode->i_mode))
 606                         return -ENOTDIR;
 607                 if (IS_ROOT(victim))
 608                         return -EBUSY;
 609         } else if (S_ISDIR(victim->d_inode->i_mode))
 610                 return -EISDIR;
 611         if (IS_DEADDIR(dir))
 612                 return -ENOENT;
 613         if (victim->d_flags & DCACHE_NFSFS_RENAMED)
 614                 return -EBUSY;
 615         return 0;
 616 }
 617
 618 /* copy of may_create in fs/namei.c() */
 619 static inline int btrfs_may_create(struct inode *dir, struct dentry *child)
 620 {
 621         if (child->d_inode)
 622                 return -EEXIST;
 623         if (IS_DEADDIR(dir))
 624                 return -ENOENT;
 625         return inode_permission(dir, MAY_WRITE | MAY_EXEC);
 626 }
 627
 628 /*
 629  * Create a new subvolume below @parent.  This is largely modeled after
 630  * sys_mkdirat and vfs_mkdir, but we only do a single component lookup
 631  * inside this filesystem so it's quite a bit simpler.
 632  */
 633 static noinline int btrfs_mksubvol(struct path *parent,
 634                                    char *name, int namelen,
 635                                    struct btrfs_root *snap_src,
 636                                    u64 *async_transid, bool readonly)
 637 {
 638         struct inode *dir  = parent->dentry->d_inode;
 639         struct dentry *dentry;
 640         int error;
 641
 642         mutex_lock_nested(&dir->i_mutex, I_MUTEX_PARENT);
 643
 644         dentry = lookup_one_len(name, parent->dentry, namelen);
 645         error = PTR_ERR(dentry);
 646         if (IS_ERR(dentry))
 647                 goto out_unlock;
 648
 649         error = -EEXIST;
 650         if (dentry->d_inode)
 651                 goto out_dput;
 652
 653         error = mnt_want_write(parent->mnt);
 654         if (error)
 655                 goto out_dput;
 656
 657         error = btrfs_may_create(dir, dentry);
 658         if (error)
 659                 goto out_drop_write;
 660
 661         down_read(&BTRFS_I(dir)->root->fs_info->subvol_sem);
 662
 663         if (btrfs_root_refs(&BTRFS_I(dir)->root->root_item) == 0)
 664                 goto out_up_read;
 665
 666         if (snap_src) {
 667                 error = create_snapshot(snap_src, dentry,
 668                                         name, namelen, async_transid, readonly);
 669         } else {
 670                 error = create_subvol(BTRFS_I(dir)->root, dentry,
 671                                       name, namelen, async_transid);
 672         }
 673         if (!error)
 674                 fsnotify_mkdir(dir, dentry);
 675 out_up_read:
 676         up_read(&BTRFS_I(dir)->root->fs_info->subvol_sem);
 677 out_drop_write:
 678         mnt_drop_write(parent->mnt);
 679 out_dput:
 680         dput(dentry);
 681 out_unlock:
 682         mutex_unlock(&dir->i_mutex);
 683         return error;
 684 }
 685
 686 /*
 687  * When we're defragging a range, we don't want to kick it off again
 688  * if it is really just waiting for delalloc to send it down.
 689  * If we find a nice big extent or delalloc range for the bytes in the
 690  * file you want to defrag, we return 0 to let you know to skip this
 691  * part of the file
 692  */
 693 static int check_defrag_in_cache(struct inode *inode, u64 offset, int thresh)
 694 {
 695         struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
 696         struct extent_map *em = NULL;
 697         struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
 698         u64 end;
 699
 700         read_lock(&em_tree->lock);
 701         em = lookup_extent_mapping(em_tree, offset, PAGE_CACHE_SIZE);
 702         read_unlock(&em_tree->lock);
 703
 704         if (em) {
 705                 end = extent_map_end(em);
 706                 free_extent_map(em);
 707                 if (end - offset > thresh)
 708                         return 0;
 709         }
 710         /* if we already have a nice delalloc here, just stop */
 711         thresh /= 2;
 712         end = count_range_bits(io_tree, &offset, offset + thresh,
 713                                thresh, EXTENT_DELALLOC, 1);
 714         if (end >= thresh)
 715                 return 0;
 716         return 1;
 717 }
 718
 719 /*
 720  * helper function to walk through a file and find extents
 721  * newer than a specific transid, and smaller than thresh.
 722  *
 723  * This is used by the defragging code to find new and small
 724  * extents
 725  */
 726 static int find_new_extents(struct btrfs_root *root,
 727                             struct inode *inode, u64 newer_than,
 728                             u64 *off, int thresh)
 729 {
 730         struct btrfs_path *path;
 731         struct btrfs_key min_key;
 732         struct btrfs_key max_key;
 733         struct extent_buffer *leaf;
 734         struct btrfs_file_extent_item *extent;
 735         int type;
 736         int ret;
 737         u64 ino = btrfs_ino(inode);
 738
 739         path = btrfs_alloc_path();
 740         if (!path)
 741                 return -ENOMEM;
 742
 743         min_key.objectid = ino;
 744         min_key.type = BTRFS_EXTENT_DATA_KEY;
 745         min_key.offset = *off;
 746
 747         max_key.objectid = ino;
 748         max_key.type = (u8)-1;
 749         max_key.offset = (u64)-1;
 750
 751         path->keep_locks = 1;
 752
 753         while(1) {
 754                 ret = btrfs_search_forward(root, &min_key, &max_key,
 755                                            path, 0, newer_than);
 756                 if (ret != 0)
 757                         goto none;
 758                 if (min_key.objectid != ino)
 759                         goto none;
 760                 if (min_key.type != BTRFS_EXTENT_DATA_KEY)
 761                         goto none;
 762
 763                 leaf = path->nodes[0];
 764                 extent = btrfs_item_ptr(leaf, path->slots[0],
 765                                         struct btrfs_file_extent_item);
 766
 767                 type = btrfs_file_extent_type(leaf, extent);
 768                 if (type == BTRFS_FILE_EXTENT_REG &&
 769                     btrfs_file_extent_num_bytes(leaf, extent) < thresh &&
 770                     check_defrag_in_cache(inode, min_key.offset, thresh)) {
 771                         *off = min_key.offset;
 772                         btrfs_free_path(path);
 773                         return 0;
 774                 }
 775
 776                 if (min_key.offset == (u64)-1)
 777                         goto none;
 778
 779                 min_key.offset++;
 780                 btrfs_release_path(path);
 781         }
 782 none:
 783         btrfs_free_path(path);
 784         return -ENOENT;
 785 }
 786
 787 /*
 788  * Validaty check of prev em and next em:
 789  * 1) no prev/next em
 790  * 2) prev/next em is an hole/inline extent
 791  */
 792 static int check_adjacent_extents(struct inode *inode, struct extent_map *em)
 793 {
 794         struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
 795         struct extent_map *prev = NULL, *next = NULL;
 796         int ret = 0;
 797
 798         read_lock(&em_tree->lock);
 799         prev = lookup_extent_mapping(em_tree, em->start - 1, (u64)-1);
 800         next = lookup_extent_mapping(em_tree, em->start + em->len, (u64)-1);
 801         read_unlock(&em_tree->lock);
 802
 803         if ((!prev || prev->block_start >= EXTENT_MAP_LAST_BYTE) &&
 804             (!next || next->block_start >= EXTENT_MAP_LAST_BYTE))
 805                 ret = 1;
 806         free_extent_map(prev);
 807         free_extent_map(next);
 808
 809         return ret;
 810 }
 811
 812 static int should_defrag_range(struct inode *inode, u64 start, u64 len,
 813                                int thresh, u64 *last_len, u64 *skip,
 814                                u64 *defrag_end)
 815 {
 816         struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
 817         struct extent_map *em = NULL;
 818         struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
 819         int ret = 1;
 820
 821         /*
 822          * make sure that once we start defragging an extent, we keep on
 823          * defragging it
 824          */
 825         if (start < *defrag_end)
 826                 return 1;
 827
 828         *skip = 0;
 829
 830         /*
 831          * hopefully we have this extent in the tree already, try without
 832          * the full extent lock
 833          */
 834         read_lock(&em_tree->lock);
 835         em = lookup_extent_mapping(em_tree, start, len);
 836         read_unlock(&em_tree->lock);
 837
 838         if (!em) {
 839                 /* get the big lock and read metadata off disk */
 840                 lock_extent(io_tree, start, start + len - 1);
 841                 em = btrfs_get_extent(inode, NULL, 0, start, len, 0);
 842                 unlock_extent(io_tree, start, start + len - 1);
 843
 844                 if (IS_ERR(em))
 845                         return 0;
 846         }
 847
 848         /* this will cover holes, and inline extents */
 849         if (em->block_start >= EXTENT_MAP_LAST_BYTE) {
 850                 ret = 0;
 851                 goto out;
 852         }
 853
 854         /* If we have nothing to merge with us, just skip. */
 855         if (check_adjacent_extents(inode, em)) {
 856                 ret = 0;
 857                 goto out;
 858         }
 859
 860         /*
 861          * we hit a real extent, if it is big don't bother defragging it again
 862          */
 863         if ((*last_len == 0 || *last_len >= thresh) && em->len >= thresh)
 864                 ret = 0;
 865
 866 out:
 867         /*
 868          * last_len ends up being a counter of how many bytes we've defragged.
 869          * every time we choose not to defrag an extent, we reset *last_len
 870          * so that the next tiny extent will force a defrag.
 871          *
 872          * The end result of this is that tiny extents before a single big
 873          * extent will force at least part of that big extent to be defragged.
 874          */
 875         if (ret) {
 876                 *defrag_end = extent_map_end(em);
 877         } else {
 878                 *last_len = 0;
 879                 *skip = extent_map_end(em);
 880                 *defrag_end = 0;
 881         }
 882
 883         free_extent_map(em);
 884         return ret;
 885 }
 886
 887 /*
 888  * it doesn't do much good to defrag one or two pages
 889  * at a time.  This pulls in a nice chunk of pages
 890  * to COW and defrag.
 891  *
 892  * It also makes sure the delalloc code has enough
 893  * dirty data to avoid making new small extents as part
 894  * of the defrag
 895  *
 896  * It's a good idea to start RA on this range
 897  * before calling this.
 898  */
 899 static int cluster_pages_for_defrag(struct inode *inode,
 900                                     struct page **pages,
 901                                     unsigned long start_index,
 902                                     int num_pages)
 903 {
 904         unsigned long file_end;
 905         u64 isize = i_size_read(inode);
 906         u64 page_start;
 907         u64 page_end;
 908         u64 page_cnt;
 909         int ret;
 910         int i;
 911         int i_done;
 912         struct btrfs_ordered_extent *ordered;
 913         struct extent_state *cached_state = NULL;
 914         struct extent_io_tree *tree;
 915         gfp_t mask = btrfs_alloc_write_mask(inode->i_mapping);
 916
 917         file_end = (isize - 1) >> PAGE_CACHE_SHIFT;
 918         if (!isize || start_index > file_end)
 919                 return 0;
 920
 921         page_cnt = min_t(u64, (u64)num_pages, (u64)file_end - start_index + 1);
 922
 923         ret = btrfs_delalloc_reserve_space(inode,
 924                                            page_cnt << PAGE_CACHE_SHIFT);
 925         if (ret)
 926                 return ret;
 927         i_done = 0;
 928         tree = &BTRFS_I(inode)->io_tree;
 929
 930         /* step one, lock all the pages */
 931         for (i = 0; i < page_cnt; i++) {
 932                 struct page *page;
 933 again:
 934                 page = find_or_create_page(inode->i_mapping,
 935                                            start_index + i, mask);
 936                 if (!page)
 937                         break;
 938
 939                 page_start = page_offset(page);
 940                 page_end = page_start + PAGE_CACHE_SIZE - 1;
 941                 while (1) {
 942                         lock_extent(tree, page_start, page_end);
 943                         ordered = btrfs_lookup_ordered_extent(inode,
 944                                                               page_start);
 945                         unlock_extent(tree, page_start, page_end);
 946                         if (!ordered)
 947                                 break;
 948
 949                         unlock_page(page);
 950                         btrfs_start_ordered_extent(inode, ordered, 1);
 951                         btrfs_put_ordered_extent(ordered);
 952                         lock_page(page);
 953                         /*
 954                          * we unlocked the page above, so we need check if
 955                          * it was released or not.
 956                          */
 957                         if (page->mapping != inode->i_mapping) {
 958                                 unlock_page(page);
 959                                 page_cache_release(page);
 960                                 goto again;
 961                         }
 962                 }
 963
 964                 if (!PageUptodate(page)) {
 965                         btrfs_readpage(NULL, page);
 966                         lock_page(page);
 967                         if (!PageUptodate(page)) {
 968                                 unlock_page(page);
 969                                 page_cache_release(page);
 970                                 ret = -EIO;
 971                                 break;
 972                         }
 973                 }
 974
 975                 if (page->mapping != inode->i_mapping) {
 976                         unlock_page(page);
 977                         page_cache_release(page);
 978                         goto again;
 979                 }
 980
 981                 pages[i] = page;
 982                 i_done++;
 983         }
 984         if (!i_done || ret)
 985                 goto out;
 986
 987         if (!(inode->i_sb->s_flags & MS_ACTIVE))
 988                 goto out;
 989
 990         /*
 991          * so now we have a nice long stream of locked
 992          * and up to date pages, lets wait on them
 993          */
 994         for (i = 0; i < i_done; i++)
 995                 wait_on_page_writeback(pages[i]);
 996
 997         page_start = page_offset(pages[0]);
 998         page_end = page_offset(pages[i_done - 1]) + PAGE_CACHE_SIZE;
 999
1000         lock_extent_bits(&BTRFS_I(inode)->io_tree,
1001                          page_start, page_end - 1, 0, &cached_state);
1002         clear_extent_bit(&BTRFS_I(inode)->io_tree, page_start,
1003                           page_end - 1, EXTENT_DIRTY | EXTENT_DELALLOC |
1004                           EXTENT_DO_ACCOUNTING, 0, 0, &cached_state,
1005                           GFP_NOFS);
1006
1007         if (i_done != page_cnt) {
1008                 spin_lock(&BTRFS_I(inode)->lock);
1009                 BTRFS_I(inode)->outstanding_extents++;
1010                 spin_unlock(&BTRFS_I(inode)->lock);
1011                 btrfs_delalloc_release_space(inode,
1012                                      (page_cnt - i_done) << PAGE_CACHE_SHIFT);
1013         }
1014
1015
1016         btrfs_set_extent_delalloc(inode, page_start, page_end - 1,
1017                                   &cached_state);
1018
1019         unlock_extent_cached(&BTRFS_I(inode)->io_tree,
1020                              page_start, page_end - 1, &cached_state,
1021                              GFP_NOFS);
1022
1023         for (i = 0; i < i_done; i++) {
1024                 clear_page_dirty_for_io(pages[i]);
1025                 ClearPageChecked(pages[i]);
1026                 set_page_extent_mapped(pages[i]);
1027                 set_page_dirty(pages[i]);
1028                 unlock_page(pages[i]);
1029                 page_cache_release(pages[i]);
1030         }
1031         return i_done;
1032 out:
1033         for (i = 0; i < i_done; i++) {
1034                 unlock_page(pages[i]);
1035                 page_cache_release(pages[i]);
1036         }
1037         btrfs_delalloc_release_space(inode, page_cnt << PAGE_CACHE_SHIFT);
1038         return ret;
1039
1040 }
1041
1042 int btrfs_defrag_file(struct inode *inode, struct file *file,
1043                       struct btrfs_ioctl_defrag_range_args *range,
1044                       u64 newer_than, unsigned long max_to_defrag)
1045 {
1046         struct btrfs_root *root = BTRFS_I(inode)->root;
1047         struct btrfs_super_block *disk_super;
1048         struct file_ra_state *ra = NULL;
1049         unsigned long last_index;
1050         u64 isize = i_size_read(inode);
1051         u64 features;
1052         u64 last_len = 0;
1053         u64 skip = 0;
1054         u64 defrag_end = 0;
1055         u64 newer_off = range->start;
1056         unsigned long i;
1057         unsigned long ra_index = 0;
1058         int ret;
1059         int defrag_count = 0;
1060         int compress_type = BTRFS_COMPRESS_ZLIB;
1061         int extent_thresh = range->extent_thresh;
1062         int max_cluster = (256 * 1024) >> PAGE_CACHE_SHIFT;
1063         int cluster = max_cluster;
1064         u64 new_align = ~((u64)128 * 1024 - 1);
1065         struct page **pages = NULL;
1066
1067         if (extent_thresh == 0)
1068                 extent_thresh = 256 * 1024;
1069
1070         if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS) {
1071                 if (range->compress_type > BTRFS_COMPRESS_TYPES)
1072                         return -EINVAL;
1073                 if (range->compress_type)
1074                         compress_type = range->compress_type;
1075         }
1076
1077         if (isize == 0)
1078                 return 0;
1079
1080         /*
1081          * if we were not given a file, allocate a readahead
1082          * context
1083          */
1084         if (!file) {
1085                 ra = kzalloc(sizeof(*ra), GFP_NOFS);
1086                 if (!ra)
1087                         return -ENOMEM;
1088                 file_ra_state_init(ra, inode->i_mapping);
1089         } else {
1090                 ra = &file->f_ra;
1091         }
1092
1093         pages = kmalloc(sizeof(struct page *) * max_cluster,
1094                         GFP_NOFS);
1095         if (!pages) {
1096                 ret = -ENOMEM;
1097                 goto out_ra;
1098         }
1099
1100         /* find the last page to defrag */
1101         if (range->start + range->len > range->start) {
1102                 last_index = min_t(u64, isize - 1,
1103                          range->start + range->len - 1) >> PAGE_CACHE_SHIFT;
1104         } else {
1105                 last_index = (isize - 1) >> PAGE_CACHE_SHIFT;
1106         }
1107
1108         if (newer_than) {
1109                 ret = find_new_extents(root, inode, newer_than,
1110                                        &newer_off, 64 * 1024);
1111                 if (!ret) {
1112                         range->start = newer_off;
1113                         /*
1114                          * we always align our defrag to help keep
1115                          * the extents in the file evenly spaced
1116                          */
1117                         i = (newer_off & new_align) >> PAGE_CACHE_SHIFT;
1118                 } else
1119                         goto out_ra;
1120         } else {
1121                 i = range->start >> PAGE_CACHE_SHIFT;
1122         }
1123         if (!max_to_defrag)
1124                 max_to_defrag = last_index + 1;
1125
1126         /*
1127          * make writeback starts from i, so the defrag range can be
1128          * written sequentially.
1129          */
1130         if (i < inode->i_mapping->writeback_index)
1131                 inode->i_mapping->writeback_index = i;
1132
1133         while (i <= last_index && defrag_count < max_to_defrag &&
1134                (i < (i_size_read(inode) + PAGE_CACHE_SIZE - 1) >>
1135                 PAGE_CACHE_SHIFT)) {
1136                 /*
1137                  * make sure we stop running if someone unmounts
1138                  * the FS
1139                  */
1140                 if (!(inode->i_sb->s_flags & MS_ACTIVE))
1141                         break;
1142
1143                 if (!should_defrag_range(inode, (u64)i << PAGE_CACHE_SHIFT,
1144                                          PAGE_CACHE_SIZE, extent_thresh,
1145                                          &last_len, &skip, &defrag_end)) {
1146                         unsigned long next;
1147                         /*
1148                          * the should_defrag function tells us how much to skip
1149                          * bump our counter by the suggested amount
1150                          */
1151                         next = (skip + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT;
1152                         i = max(i + 1, next);
1153                         continue;
1154                 }
1155
1156                 if (!newer_than) {
1157                         cluster = (PAGE_CACHE_ALIGN(defrag_end) >>
1158                                    PAGE_CACHE_SHIFT) - i;
1159                         cluster = min(cluster, max_cluster);
1160                 } else {
1161                         cluster = max_cluster;
1162                 }
1163
1164                 if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)
1165                         BTRFS_I(inode)->force_compress = compress_type;
1166
1167                 if (i + cluster > ra_index) {
1168                         ra_index = max(i, ra_index);
1169                         btrfs_force_ra(inode->i_mapping, ra, file, ra_index,
1170                                        cluster);
1171                         ra_index += max_cluster;
1172                 }
1173
1174                 mutex_lock(&inode->i_mutex);
1175                 ret = cluster_pages_for_defrag(inode, pages, i, cluster);
1176                 if (ret < 0) {
1177                         mutex_unlock(&inode->i_mutex);
1178                         goto out_ra;
1179                 }
1180
1181                 defrag_count += ret;
1182                 balance_dirty_pages_ratelimited_nr(inode->i_mapping, ret);
1183                 mutex_unlock(&inode->i_mutex);
1184
1185                 if (newer_than) {
1186                         if (newer_off == (u64)-1)
1187                                 break;
1188
1189                         newer_off = max(newer_off + 1,
1190                                         (u64)i << PAGE_CACHE_SHIFT);
1191
1192                         ret = find_new_extents(root, inode,
1193                                                newer_than, &newer_off,
1194                                                64 * 1024);
1195                         if (!ret) {
1196                                 range->start = newer_off;
1197                                 i = (newer_off & new_align) >> PAGE_CACHE_SHIFT;
1198                         } else {
1199                                 break;
1200                         }
1201                 } else {
1202                         if (ret > 0) {
1203                                 i += ret;
1204                                 last_len += ret << PAGE_CACHE_SHIFT;
1205                         } else {
1206                                 i++;
1207                                 last_len = 0;
1208                         }
1209                 }
1210         }
1211
1212         if ((range->flags & BTRFS_DEFRAG_RANGE_START_IO))
1213                 filemap_flush(inode->i_mapping);
1214
1215         if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
1216                 /* the filemap_flush will queue IO into the worker threads, but
1217                  * we have to make sure the IO is actually started and that
1218                  * ordered extents get created before we return
1219                  */
1220                 atomic_inc(&root->fs_info->async_submit_draining);
1221                 while (atomic_read(&root->fs_info->nr_async_submits) ||
1222                       atomic_read(&root->fs_info->async_delalloc_pages)) {
1223                         wait_event(root->fs_info->async_submit_wait,
1224                            (atomic_read(&root->fs_info->nr_async_submits) == 0 &&
1225                             atomic_read(&root->fs_info->async_delalloc_pages) == 0));
1226                 }
1227                 atomic_dec(&root->fs_info->async_submit_draining);
1228
1229                 mutex_lock(&inode->i_mutex);
1230                 BTRFS_I(inode)->force_compress = BTRFS_COMPRESS_NONE;
1231                 mutex_unlock(&inode->i_mutex);
1232         }
1233
1234         disk_super = root->fs_info->super_copy;
1235         features = btrfs_super_incompat_flags(disk_super);
1236         if (range->compress_type == BTRFS_COMPRESS_LZO) {
1237                 features |= BTRFS_FEATURE_INCOMPAT_COMPRESS_LZO;
1238                 btrfs_set_super_incompat_flags(disk_super, features);
1239         }
1240
1241         ret = defrag_count;
1242
1243 out_ra:
1244         if (!file)
1245                 kfree(ra);
1246         kfree(pages);
1247         return ret;
1248 }
1249
1250 static noinline int btrfs_ioctl_resize(struct btrfs_root *root,
1251                                         void __user *arg)
1252 {
1253         u64 new_size;
1254         u64 old_size;
1255         u64 devid = 1;
1256         struct btrfs_ioctl_vol_args *vol_args;
1257         struct btrfs_trans_handle *trans;
1258         struct btrfs_device *device = NULL;
1259         char *sizestr;
1260         char *devstr = NULL;
1261         int ret = 0;
1262         int mod = 0;
1263
1264         if (root->fs_info->sb->s_flags & MS_RDONLY)
1265                 return -EROFS;
1266
1267         if (!capable(CAP_SYS_ADMIN))
1268                 return -EPERM;
1269
1270         mutex_lock(&root->fs_info->volume_mutex);
1271         if (root->fs_info->balance_ctl) {
1272                 printk(KERN_INFO "btrfs: balance in progress\n");
1273                 ret = -EINVAL;
1274                 goto out;
1275         }
1276
1277         vol_args = memdup_user(arg, sizeof(*vol_args));
1278         if (IS_ERR(vol_args)) {
1279                 ret = PTR_ERR(vol_args);
1280                 goto out;
1281         }
1282
1283         vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1284
1285         sizestr = vol_args->name;
1286         devstr = strchr(sizestr, ':');
1287         if (devstr) {
1288                 char *end;
1289                 sizestr = devstr + 1;
1290                 *devstr = '\0';
1291                 devstr = vol_args->name;
1292                 devid = simple_strtoull(devstr, &end, 10);
1293                 printk(KERN_INFO "btrfs: resizing devid %llu\n",
1294                        (unsigned long long)devid);
1295         }
1296         device = btrfs_find_device(root, devid, NULL, NULL);
1297         if (!device) {
1298                 printk(KERN_INFO "btrfs: resizer unable to find device %llu\n",
1299                        (unsigned long long)devid);
1300                 ret = -EINVAL;
1301                 goto out_free;
1302         }
1303         if (!strcmp(sizestr, "max"))
1304                 new_size = device->bdev->bd_inode->i_size;
1305         else {
1306                 if (sizestr[0] == '-') {
1307                         mod = -1;
1308                         sizestr++;
1309                 } else if (sizestr[0] == '+') {
1310                         mod = 1;
1311                         sizestr++;
1312                 }
1313                 new_size = memparse(sizestr, NULL);
1314                 if (new_size == 0) {
1315                         ret = -EINVAL;
1316                         goto out_free;
1317                 }
1318         }
1319
1320         old_size = device->total_bytes;
1321
1322         if (mod < 0) {
1323                 if (new_size > old_size) {
1324                         ret = -EINVAL;
1325                         goto out_free;
1326                 }
1327                 new_size = old_size - new_size;
1328         } else if (mod > 0) {
1329                 new_size = old_size + new_size;
1330         }
1331
1332         if (new_size < 256 * 1024 * 1024) {
1333                 ret = -EINVAL;
1334                 goto out_free;
1335         }
1336         if (new_size > device->bdev->bd_inode->i_size) {
1337                 ret = -EFBIG;
1338                 goto out_free;
1339         }
1340
1341         do_div(new_size, root->sectorsize);
1342         new_size *= root->sectorsize;
1343
1344         printk(KERN_INFO "btrfs: new size for %s is %llu\n",
1345                 device->name, (unsigned long long)new_size);
1346
1347         if (new_size > old_size) {
1348                 trans = btrfs_start_transaction(root, 0);
1349                 if (IS_ERR(trans)) {
1350                         ret = PTR_ERR(trans);
1351                         goto out_free;
1352                 }
1353                 ret = btrfs_grow_device(trans, device, new_size);
1354                 btrfs_commit_transaction(trans, root);
1355         } else if (new_size < old_size) {
1356                 ret = btrfs_shrink_device(device, new_size);
1357         }
1358
1359 out_free:
1360         kfree(vol_args);
1361 out:
1362         mutex_unlock(&root->fs_info->volume_mutex);
1363         return ret;
1364 }
1365
1366 static noinline int btrfs_ioctl_snap_create_transid(struct file *file,
1367                                                     char *name,
1368                                                     unsigned long fd,
1369                                                     int subvol,
1370                                                     u64 *transid,
1371                                                     bool readonly)
1372 {
1373         struct btrfs_root *root = BTRFS_I(fdentry(file)->d_inode)->root;
1374         struct file *src_file;
1375         int namelen;
1376         int ret = 0;
1377
1378         if (root->fs_info->sb->s_flags & MS_RDONLY)
1379                 return -EROFS;
1380
1381         namelen = strlen(name);
1382         if (strchr(name, '/')) {
1383                 ret = -EINVAL;
1384                 goto out;
1385         }
1386
1387         if (name[0] == '.' &&
1388            (namelen == 1 || (name[1] == '.' && namelen == 2))) {
1389                 ret = -EEXIST;
1390                 goto out;
1391         }
1392
1393         if (subvol) {
1394                 ret = btrfs_mksubvol(&file->f_path, name, namelen,
1395                                      NULL, transid, readonly);
1396         } else {
1397                 struct inode *src_inode;
1398                 src_file = fget(fd);
1399                 if (!src_file) {
1400                         ret = -EINVAL;
1401                         goto out;
1402                 }
1403
1404                 src_inode = src_file->f_path.dentry->d_inode;
1405                 if (src_inode->i_sb != file->f_path.dentry->d_inode->i_sb) {
1406                         printk(KERN_INFO "btrfs: Snapshot src from "
1407                                "another FS\n");
1408                         ret = -EINVAL;
1409                         fput(src_file);
1410                         goto out;
1411                 }
1412                 ret = btrfs_mksubvol(&file->f_path, name, namelen,
1413                                      BTRFS_I(src_inode)->root,
1414                                      transid, readonly);
1415                 fput(src_file);
1416         }
1417 out:
1418         return ret;
1419 }
1420
1421 static noinline int btrfs_ioctl_snap_create(struct file *file,
1422                                             void __user *arg, int subvol)
1423 {
1424         struct btrfs_ioctl_vol_args *vol_args;
1425         int ret;
1426
1427         vol_args = memdup_user(arg, sizeof(*vol_args));
1428         if (IS_ERR(vol_args))
1429                 return PTR_ERR(vol_args);
1430         vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1431
1432         ret = btrfs_ioctl_snap_create_transid(file, vol_args->name,
1433                                               vol_args->fd, subvol,
1434                                               NULL, false);
1435
1436         kfree(vol_args);
1437         return ret;
1438 }
1439
1440 static noinline int btrfs_ioctl_snap_create_v2(struct file *file,
1441                                                void __user *arg, int subvol)
1442 {
1443         struct btrfs_ioctl_vol_args_v2 *vol_args;
1444         int ret;
1445         u64 transid = 0;
1446         u64 *ptr = NULL;
1447         bool readonly = false;
1448
1449         vol_args = memdup_user(arg, sizeof(*vol_args));
1450         if (IS_ERR(vol_args))
1451                 return PTR_ERR(vol_args);
1452         vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0';
1453
1454         if (vol_args->flags &
1455             ~(BTRFS_SUBVOL_CREATE_ASYNC | BTRFS_SUBVOL_RDONLY)) {
1456                 ret = -EOPNOTSUPP;
1457                 goto out;
1458         }
1459
1460         if (vol_args->flags & BTRFS_SUBVOL_CREATE_ASYNC)
1461                 ptr = &transid;
1462         if (vol_args->flags & BTRFS_SUBVOL_RDONLY)
1463                 readonly = true;
1464
1465         ret = btrfs_ioctl_snap_create_transid(file, vol_args->name,
1466                                               vol_args->fd, subvol,
1467                                               ptr, readonly);
1468
1469         if (ret == 0 && ptr &&
1470             copy_to_user(arg +
1471                          offsetof(struct btrfs_ioctl_vol_args_v2,
1472                                   transid), ptr, sizeof(*ptr)))
1473                 ret = -EFAULT;
1474 out:
1475         kfree(vol_args);
1476         return ret;
1477 }
1478
1479 static noinline int btrfs_ioctl_subvol_getflags(struct file *file,
1480                                                 void __user *arg)
1481 {
1482         struct inode *inode = fdentry(file)->d_inode;
1483         struct btrfs_root *root = BTRFS_I(inode)->root;
1484         int ret = 0;
1485         u64 flags = 0;
1486
1487         if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID)
1488                 return -EINVAL;
1489
1490         down_read(&root->fs_info->subvol_sem);
1491         if (btrfs_root_readonly(root))
1492                 flags |= BTRFS_SUBVOL_RDONLY;
1493         up_read(&root->fs_info->subvol_sem);
1494
1495         if (copy_to_user(arg, &flags, sizeof(flags)))
1496                 ret = -EFAULT;
1497
1498         return ret;
1499 }
1500
1501 static noinline int btrfs_ioctl_subvol_setflags(struct file *file,
1502                                               void __user *arg)
1503 {
1504         struct inode *inode = fdentry(file)->d_inode;
1505         struct btrfs_root *root = BTRFS_I(inode)->root;
1506         struct btrfs_trans_handle *trans;
1507         u64 root_flags;
1508         u64 flags;
1509         int ret = 0;
1510
1511         if (root->fs_info->sb->s_flags & MS_RDONLY)
1512                 return -EROFS;
1513
1514         if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID)
1515                 return -EINVAL;
1516
1517         if (copy_from_user(&flags, arg, sizeof(flags)))
1518                 return -EFAULT;
1519
1520         if (flags & BTRFS_SUBVOL_CREATE_ASYNC)
1521                 return -EINVAL;
1522
1523         if (flags & ~BTRFS_SUBVOL_RDONLY)
1524                 return -EOPNOTSUPP;
1525
1526         if (!inode_owner_or_capable(inode))
1527                 return -EACCES;
1528
1529         down_write(&root->fs_info->subvol_sem);
1530
1531         /* nothing to do */
1532         if (!!(flags & BTRFS_SUBVOL_RDONLY) == btrfs_root_readonly(root))
1533                 goto out;
1534
1535         root_flags = btrfs_root_flags(&root->root_item);
1536         if (flags & BTRFS_SUBVOL_RDONLY)
1537                 btrfs_set_root_flags(&root->root_item,
1538                                      root_flags | BTRFS_ROOT_SUBVOL_RDONLY);
1539         else
1540                 btrfs_set_root_flags(&root->root_item,
1541                                      root_flags & ~BTRFS_ROOT_SUBVOL_RDONLY);
1542
1543         trans = btrfs_start_transaction(root, 1);
1544         if (IS_ERR(trans)) {
1545                 ret = PTR_ERR(trans);
1546                 goto out_reset;
1547         }
1548
1549         ret = btrfs_update_root(trans, root->fs_info->tree_root,
1550                                 &root->root_key, &root->root_item);
1551
1552         btrfs_commit_transaction(trans, root);
1553 out_reset:
1554         if (ret)
1555                 btrfs_set_root_flags(&root->root_item, root_flags);
1556 out:
1557         up_write(&root->fs_info->subvol_sem);
1558         return ret;
1559 }
1560
1561 /*
1562  * helper to check if the subvolume references other subvolumes
1563  */
1564 static noinline int may_destroy_subvol(struct btrfs_root *root)
1565 {
1566         struct btrfs_path *path;
1567         struct btrfs_key key;
1568         int ret;
1569
1570         path = btrfs_alloc_path();
1571         if (!path)
1572                 return -ENOMEM;
1573
1574         key.objectid = root->root_key.objectid;
1575         key.type = BTRFS_ROOT_REF_KEY;
1576         key.offset = (u64)-1;
1577
1578         ret = btrfs_search_slot(NULL, root->fs_info->tree_root,
1579                                 &key, path, 0, 0);
1580         if (ret < 0)
1581                 goto out;
1582         BUG_ON(ret == 0);
1583
1584         ret = 0;
1585         if (path->slots[0] > 0) {
1586                 path->slots[0]--;
1587                 btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
1588                 if (key.objectid == root->root_key.objectid &&
1589                     key.type == BTRFS_ROOT_REF_KEY)
1590                         ret = -ENOTEMPTY;
1591         }
1592 out:
1593         btrfs_free_path(path);
1594         return ret;
1595 }
1596
1597 static noinline int key_in_sk(struct btrfs_key *key,
1598                               struct btrfs_ioctl_search_key *sk)
1599 {
1600         struct btrfs_key test;
1601         int ret;
1602
1603         test.objectid = sk->min_objectid;
1604         test.type = sk->min_type;
1605         test.offset = sk->min_offset;
1606
1607         ret = btrfs_comp_cpu_keys(key, &test);
1608         if (ret < 0)
1609                 return 0;
1610
1611         test.objectid = sk->max_objectid;
1612         test.type = sk->max_type;
1613         test.offset = sk->max_offset;
1614
1615         ret = btrfs_comp_cpu_keys(key, &test);
1616         if (ret > 0)
1617                 return 0;
1618         return 1;
1619 }
1620
1621 static noinline int copy_to_sk(struct btrfs_root *root,
1622                                struct btrfs_path *path,
1623                                struct btrfs_key *key,
1624                                struct btrfs_ioctl_search_key *sk,
1625                                char *buf,
1626                                unsigned long *sk_offset,
1627                                int *num_found)
1628 {
1629         u64 found_transid;
1630         struct extent_buffer *leaf;
1631         struct btrfs_ioctl_search_header sh;
1632         unsigned long item_off;
1633         unsigned long item_len;
1634         int nritems;
1635         int i;
1636         int slot;
1637         int ret = 0;
1638
1639         leaf = path->nodes[0];
1640         slot = path->slots[0];
1641         nritems = btrfs_header_nritems(leaf);
1642
1643         if (btrfs_header_generation(leaf) > sk->max_transid) {
1644                 i = nritems;
1645                 goto advance_key;
1646         }
1647         found_transid = btrfs_header_generation(leaf);
1648
1649         for (i = slot; i < nritems; i++) {
1650                 item_off = btrfs_item_ptr_offset(leaf, i);
1651                 item_len = btrfs_item_size_nr(leaf, i);
1652
1653                 if (item_len > BTRFS_SEARCH_ARGS_BUFSIZE)
1654                         item_len = 0;
1655
1656                 if (sizeof(sh) + item_len + *sk_offset >
1657                     BTRFS_SEARCH_ARGS_BUFSIZE) {
1658                         ret = 1;
1659                         goto overflow;
1660                 }
1661
1662                 btrfs_item_key_to_cpu(leaf, key, i);
1663                 if (!key_in_sk(key, sk))
1664                         continue;
1665
1666                 sh.objectid = key->objectid;
1667                 sh.offset = key->offset;
1668                 sh.type = key->type;
1669                 sh.len = item_len;
1670                 sh.transid = found_transid;
1671
1672                 /* copy search result header */
1673                 memcpy(buf + *sk_offset, &sh, sizeof(sh));
1674                 *sk_offset += sizeof(sh);
1675
1676                 if (item_len) {
1677                         char *p = buf + *sk_offset;
1678                         /* copy the item */
1679                         read_extent_buffer(leaf, p,
1680                                            item_off, item_len);
1681                         *sk_offset += item_len;
1682                 }
1683                 (*num_found)++;
1684
1685                 if (*num_found >= sk->nr_items)
1686                         break;
1687         }
1688 advance_key:
1689         ret = 0;
1690         if (key->offset < (u64)-1 && key->offset < sk->max_offset)
1691                 key->offset++;
1692         else if (key->type < (u8)-1 && key->type < sk->max_type) {
1693                 key->offset = 0;
1694                 key->type++;
1695         } else if (key->objectid < (u64)-1 && key->objectid < sk->max_objectid) {
1696                 key->offset = 0;
1697                 key->type = 0;
1698                 key->objectid++;
1699         } else
1700                 ret = 1;
1701 overflow:
1702         return ret;
1703 }
1704
1705 static noinline int search_ioctl(struct inode *inode,
1706                                  struct btrfs_ioctl_search_args *args)
1707 {
1708         struct btrfs_root *root;
1709         struct btrfs_key key;
1710         struct btrfs_key max_key;
1711         struct btrfs_path *path;
1712         struct btrfs_ioctl_search_key *sk = &args->key;
1713         struct btrfs_fs_info *info = BTRFS_I(inode)->root->fs_info;
1714         int ret;
1715         int num_found = 0;
1716         unsigned long sk_offset = 0;
1717
1718         path = btrfs_alloc_path();
1719         if (!path)
1720                 return -ENOMEM;
1721
1722         if (sk->tree_id == 0) {
1723                 /* search the root of the inode that was passed */
1724                 root = BTRFS_I(inode)->root;
1725         } else {
1726                 key.objectid = sk->tree_id;
1727                 key.type = BTRFS_ROOT_ITEM_KEY;
1728                 key.offset = (u64)-1;
1729                 root = btrfs_read_fs_root_no_name(info, &key);
1730                 if (IS_ERR(root)) {
1731                         printk(KERN_ERR "could not find root %llu\n",
1732                                sk->tree_id);
1733                         btrfs_free_path(path);
1734                         return -ENOENT;
1735                 }
1736         }
1737
1738         key.objectid = sk->min_objectid;
1739         key.type = sk->min_type;
1740         key.offset = sk->min_offset;
1741
1742         max_key.objectid = sk->max_objectid;
1743         max_key.type = sk->max_type;
1744         max_key.offset = sk->max_offset;
1745
1746         path->keep_locks = 1;
1747
1748         while(1) {
1749                 ret = btrfs_search_forward(root, &key, &max_key, path, 0,
1750                                            sk->min_transid);
1751                 if (ret != 0) {
1752                         if (ret > 0)
1753                                 ret = 0;
1754                         goto err;
1755                 }
1756                 ret = copy_to_sk(root, path, &key, sk, args->buf,
1757                                  &sk_offset, &num_found);
1758                 btrfs_release_path(path);
1759                 if (ret || num_found >= sk->nr_items)
1760                         break;
1761
1762         }
1763         ret = 0;
1764 err:
1765         sk->nr_items = num_found;
1766         btrfs_free_path(path);
1767         return ret;
1768 }
1769
1770 static noinline int btrfs_ioctl_tree_search(struct file *file,
1771                                            void __user *argp)
1772 {
1773          struct btrfs_ioctl_search_args *args;
1774          struct inode *inode;
1775          int ret;
1776
1777         if (!capable(CAP_SYS_ADMIN))
1778                 return -EPERM;
1779
1780         args = memdup_user(argp, sizeof(*args));
1781         if (IS_ERR(args))
1782                 return PTR_ERR(args);
1783
1784         inode = fdentry(file)->d_inode;
1785         ret = search_ioctl(inode, args);
1786         if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
1787                 ret = -EFAULT;
1788         kfree(args);
1789         return ret;
1790 }
1791
1792 /*
1793  * Search INODE_REFs to identify path name of 'dirid' directory
1794  * in a 'tree_id' tree. and sets path name to 'name'.
1795  */
1796 static noinline int btrfs_search_path_in_tree(struct btrfs_fs_info *info,
1797                                 u64 tree_id, u64 dirid, char *name)
1798 {
1799         struct btrfs_root *root;
1800         struct btrfs_key key;
1801         char *ptr;
1802         int ret = -1;
1803         int slot;
1804         int len;
1805         int total_len = 0;
1806         struct btrfs_inode_ref *iref;
1807         struct extent_buffer *l;
1808         struct btrfs_path *path;
1809
1810         if (dirid == BTRFS_FIRST_FREE_OBJECTID) {
1811                 name[0]='\0';
1812                 return 0;
1813         }
1814
1815         path = btrfs_alloc_path();
1816         if (!path)
1817                 return -ENOMEM;
1818
1819         ptr = &name[BTRFS_INO_LOOKUP_PATH_MAX];
1820
1821         key.objectid = tree_id;
1822         key.type = BTRFS_ROOT_ITEM_KEY;
1823         key.offset = (u64)-1;
1824         root = btrfs_read_fs_root_no_name(info, &key);
1825         if (IS_ERR(root)) {
1826                 printk(KERN_ERR "could not find root %llu\n", tree_id);
1827                 ret = -ENOENT;
1828                 goto out;
1829         }
1830
1831         key.objectid = dirid;
1832         key.type = BTRFS_INODE_REF_KEY;
1833         key.offset = (u64)-1;
1834
1835         while(1) {
1836                 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
1837                 if (ret < 0)
1838                         goto out;
1839
1840                 l = path->nodes[0];
1841                 slot = path->slots[0];
1842                 if (ret > 0 && slot > 0)
1843                         slot--;
1844                 btrfs_item_key_to_cpu(l, &key, slot);
1845
1846                 if (ret > 0 && (key.objectid != dirid ||
1847                                 key.type != BTRFS_INODE_REF_KEY)) {
1848                         ret = -ENOENT;
1849                         goto out;
1850                 }
1851
1852                 iref = btrfs_item_ptr(l, slot, struct btrfs_inode_ref);
1853                 len = btrfs_inode_ref_name_len(l, iref);
1854                 ptr -= len + 1;
1855                 total_len += len + 1;
1856                 if (ptr < name)
1857                         goto out;
1858
1859                 *(ptr + len) = '/';
1860                 read_extent_buffer(l, ptr,(unsigned long)(iref + 1), len);
1861
1862                 if (key.offset == BTRFS_FIRST_FREE_OBJECTID)
1863                         break;
1864
1865                 btrfs_release_path(path);
1866                 key.objectid = key.offset;
1867                 key.offset = (u64)-1;
1868                 dirid = key.objectid;
1869         }
1870         if (ptr < name)
1871                 goto out;
1872         memmove(name, ptr, total_len);
1873         name[total_len]='\0';
1874         ret = 0;
1875 out:
1876         btrfs_free_path(path);
1877         return ret;
1878 }
1879
1880 static noinline int btrfs_ioctl_ino_lookup(struct file *file,
1881                                            void __user *argp)
1882 {
1883          struct btrfs_ioctl_ino_lookup_args *args;
1884          struct inode *inode;
1885          int ret;
1886
1887         if (!capable(CAP_SYS_ADMIN))
1888                 return -EPERM;
1889
1890         args = memdup_user(argp, sizeof(*args));
1891         if (IS_ERR(args))
1892                 return PTR_ERR(args);
1893
1894         inode = fdentry(file)->d_inode;
1895
1896         if (args->treeid == 0)
1897                 args->treeid = BTRFS_I(inode)->root->root_key.objectid;
1898
1899         ret = btrfs_search_path_in_tree(BTRFS_I(inode)->root->fs_info,
1900                                         args->treeid, args->objectid,
1901                                         args->name);
1902
1903         if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
1904                 ret = -EFAULT;
1905
1906         kfree(args);
1907         return ret;
1908 }
1909
1910 static noinline int btrfs_ioctl_snap_destroy(struct file *file,
1911                                              void __user *arg)
1912 {
1913         struct dentry *parent = fdentry(file);
1914         struct dentry *dentry;
1915         struct inode *dir = parent->d_inode;
1916         struct inode *inode;
1917         struct btrfs_root *root = BTRFS_I(dir)->root;
1918         struct btrfs_root *dest = NULL;
1919         struct btrfs_ioctl_vol_args *vol_args;
1920         struct btrfs_trans_handle *trans;
1921         int namelen;
1922         int ret;
1923         int err = 0;
1924
1925         vol_args = memdup_user(arg, sizeof(*vol_args));
1926         if (IS_ERR(vol_args))
1927                 return PTR_ERR(vol_args);
1928
1929         vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1930         namelen = strlen(vol_args->name);
1931         if (strchr(vol_args->name, '/') ||
1932             strncmp(vol_args->name, "..", namelen) == 0) {
1933                 err = -EINVAL;
1934                 goto out;
1935         }
1936
1937         err = mnt_want_write_file(file);
1938         if (err)
1939                 goto out;
1940
1941         mutex_lock_nested(&dir->i_mutex, I_MUTEX_PARENT);
1942         dentry = lookup_one_len(vol_args->name, parent, namelen);
1943         if (IS_ERR(dentry)) {
1944                 err = PTR_ERR(dentry);
1945                 goto out_unlock_dir;
1946         }
1947
1948         if (!dentry->d_inode) {
1949                 err = -ENOENT;
1950                 goto out_dput;
1951         }
1952
1953         inode = dentry->d_inode;
1954         dest = BTRFS_I(inode)->root;
1955         if (!capable(CAP_SYS_ADMIN)){
1956                 /*
1957                  * Regular user.  Only allow this with a special mount
1958                  * option, when the user has write+exec access to the
1959                  * subvol root, and when rmdir(2) would have been
1960                  * allowed.
1961                  *
1962                  * Note that this is _not_ check that the subvol is
1963                  * empty or doesn't contain data that we wouldn't
1964                  * otherwise be able to delete.
1965                  *
1966                  * Users who want to delete empty subvols should try
1967                  * rmdir(2).
1968                  */
1969                 err = -EPERM;
1970                 if (!btrfs_test_opt(root, USER_SUBVOL_RM_ALLOWED))
1971                         goto out_dput;
1972
1973                 /*
1974                  * Do not allow deletion if the parent dir is the same
1975                  * as the dir to be deleted.  That means the ioctl
1976                  * must be called on the dentry referencing the root
1977                  * of the subvol, not a random directory contained
1978                  * within it.
1979                  */
1980                 err = -EINVAL;
1981                 if (root == dest)
1982                         goto out_dput;
1983
1984                 err = inode_permission(inode, MAY_WRITE | MAY_EXEC);
1985                 if (err)
1986                         goto out_dput;
1987
1988                 /* check if subvolume may be deleted by a non-root user */
1989                 err = btrfs_may_delete(dir, dentry, 1);
1990                 if (err)
1991                         goto out_dput;
1992         }
1993
1994         if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) {
1995                 err = -EINVAL;
1996                 goto out_dput;
1997         }
1998
1999         mutex_lock(&inode->i_mutex);
2000         err = d_invalidate(dentry);
2001         if (err)
2002                 goto out_unlock;
2003
2004         down_write(&root->fs_info->subvol_sem);
2005
2006         err = may_destroy_subvol(dest);
2007         if (err)
2008                 goto out_up_write;
2009
2010         trans = btrfs_start_transaction(root, 0);
2011         if (IS_ERR(trans)) {
2012                 err = PTR_ERR(trans);
2013                 goto out_up_write;
2014         }
2015         trans->block_rsv = &root->fs_info->global_block_rsv;
2016
2017         ret = btrfs_unlink_subvol(trans, root, dir,
2018                                 dest->root_key.objectid,
2019                                 dentry->d_name.name,
2020                                 dentry->d_name.len);
2021         if (ret) {
2022                 err = ret;
2023                 btrfs_abort_transaction(trans, root, ret);
2024                 goto out_end_trans;
2025         }
2026
2027         btrfs_record_root_in_trans(trans, dest);
2028
2029         memset(&dest->root_item.drop_progress, 0,
2030                 sizeof(dest->root_item.drop_progress));
2031         dest->root_item.drop_level = 0;
2032         btrfs_set_root_refs(&dest->root_item, 0);
2033
2034         if (!xchg(&dest->orphan_item_inserted, 1)) {
2035                 ret = btrfs_insert_orphan_item(trans,
2036                                         root->fs_info->tree_root,
2037                                         dest->root_key.objectid);
2038                 if (ret) {
2039                         btrfs_abort_transaction(trans, root, ret);
2040                         err = ret;
2041                         goto out_end_trans;
2042                 }
2043         }
2044 out_end_trans:
2045         ret = btrfs_end_transaction(trans, root);
2046         if (ret && !err)
2047                 err = ret;
2048         inode->i_flags |= S_DEAD;
2049 out_up_write:
2050         up_write(&root->fs_info->subvol_sem);
2051 out_unlock:
2052         mutex_unlock(&inode->i_mutex);
2053         if (!err) {
2054                 shrink_dcache_sb(root->fs_info->sb);
2055                 btrfs_invalidate_inodes(dest);
2056                 d_delete(dentry);
2057         }
2058 out_dput:
2059         dput(dentry);
2060 out_unlock_dir:
2061         mutex_unlock(&dir->i_mutex);
2062         mnt_drop_write_file(file);
2063 out:
2064         kfree(vol_args);
2065         return err;
2066 }
2067
2068 static int btrfs_ioctl_defrag(struct file *file, void __user *argp)
2069 {
2070         struct inode *inode = fdentry(file)->d_inode;
2071         struct btrfs_root *root = BTRFS_I(inode)->root;
2072         struct btrfs_ioctl_defrag_range_args *range;
2073         int ret;
2074
2075         if (btrfs_root_readonly(root))
2076                 return -EROFS;
2077
2078         ret = mnt_want_write_file(file);
2079         if (ret)
2080                 return ret;
2081
2082         switch (inode->i_mode & S_IFMT) {
2083         case S_IFDIR:
2084                 if (!capable(CAP_SYS_ADMIN)) {
2085                         ret = -EPERM;
2086                         goto out;
2087                 }
2088                 ret = btrfs_defrag_root(root, 0);
2089                 if (ret)
2090                         goto out;
2091                 ret = btrfs_defrag_root(root->fs_info->extent_root, 0);
2092                 break;
2093         case S_IFREG:
2094                 if (!(file->f_mode & FMODE_WRITE)) {
2095                         ret = -EINVAL;
2096                         goto out;
2097                 }
2098
2099                 range = kzalloc(sizeof(*range), GFP_KERNEL);
2100                 if (!range) {
2101                         ret = -ENOMEM;
2102                         goto out;
2103                 }
2104
2105                 if (argp) {
2106                         if (copy_from_user(range, argp,
2107                                            sizeof(*range))) {
2108                                 ret = -EFAULT;
2109                                 kfree(range);
2110                                 goto out;
2111                         }
2112                         /* compression requires us to start the IO */
2113                         if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
2114                                 range->flags |= BTRFS_DEFRAG_RANGE_START_IO;
2115                                 range->extent_thresh = (u32)-1;
2116                         }
2117                 } else {
2118                         /* the rest are all set to zero by kzalloc */
2119                         range->len = (u64)-1;
2120                 }
2121                 ret = btrfs_defrag_file(fdentry(file)->d_inode, file,
2122                                         range, 0, 0);
2123                 if (ret > 0)
2124                         ret = 0;
2125                 kfree(range);
2126                 break;
2127         default:
2128                 ret = -EINVAL;
2129         }
2130 out:
2131         mnt_drop_write_file(file);
2132         return ret;
2133 }
2134
2135 static long btrfs_ioctl_add_dev(struct btrfs_root *root, void __user *arg)
2136 {
2137         struct btrfs_ioctl_vol_args *vol_args;
2138         int ret;
2139
2140         if (!capable(CAP_SYS_ADMIN))
2141                 return -EPERM;
2142
2143         mutex_lock(&root->fs_info->volume_mutex);
2144         if (root->fs_info->balance_ctl) {
2145                 printk(KERN_INFO "btrfs: balance in progress\n");
2146                 ret = -EINVAL;
2147                 goto out;
2148         }
2149
2150         vol_args = memdup_user(arg, sizeof(*vol_args));
2151         if (IS_ERR(vol_args)) {
2152                 ret = PTR_ERR(vol_args);
2153                 goto out;
2154         }
2155
2156         vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2157         ret = btrfs_init_new_device(root, vol_args->name);
2158
2159         kfree(vol_args);
2160 out:
2161         mutex_unlock(&root->fs_info->volume_mutex);
2162         return ret;
2163 }
2164
2165 static long btrfs_ioctl_rm_dev(struct btrfs_root *root, void __user *arg)
2166 {
2167         struct btrfs_ioctl_vol_args *vol_args;
2168         int ret;
2169
2170         if (!capable(CAP_SYS_ADMIN))
2171                 return -EPERM;
2172
2173         if (root->fs_info->sb->s_flags & MS_RDONLY)
2174                 return -EROFS;
2175
2176         mutex_lock(&root->fs_info->volume_mutex);
2177         if (root->fs_info->balance_ctl) {
2178                 printk(KERN_INFO "btrfs: balance in progress\n");
2179                 ret = -EINVAL;
2180                 goto out;
2181         }
2182
2183         vol_args = memdup_user(arg, sizeof(*vol_args));
2184         if (IS_ERR(vol_args)) {
2185                 ret = PTR_ERR(vol_args);
2186                 goto out;
2187         }
2188
2189         vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2190         ret = btrfs_rm_device(root, vol_args->name);
2191
2192         kfree(vol_args);
2193 out:
2194         mutex_unlock(&root->fs_info->volume_mutex);
2195         return ret;
2196 }
2197
2198 static long btrfs_ioctl_fs_info(struct btrfs_root *root, void __user *arg)
2199 {
2200         struct btrfs_ioctl_fs_info_args *fi_args;
2201         struct btrfs_device *device;
2202         struct btrfs_device *next;
2203         struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
2204         int ret = 0;
2205
2206         if (!capable(CAP_SYS_ADMIN))
2207                 return -EPERM;
2208
2209         fi_args = kzalloc(sizeof(*fi_args), GFP_KERNEL);
2210         if (!fi_args)
2211                 return -ENOMEM;
2212
2213         fi_args->num_devices = fs_devices->num_devices;
2214         memcpy(&fi_args->fsid, root->fs_info->fsid, sizeof(fi_args->fsid));
2215
2216         mutex_lock(&fs_devices->device_list_mutex);
2217         list_for_each_entry_safe(device, next, &fs_devices->devices, dev_list) {
2218                 if (device->devid > fi_args->max_id)
2219                         fi_args->max_id = device->devid;
2220         }
2221         mutex_unlock(&fs_devices->device_list_mutex);
2222
2223         if (copy_to_user(arg, fi_args, sizeof(*fi_args)))
2224                 ret = -EFAULT;
2225
2226         kfree(fi_args);
2227         return ret;
2228 }
2229
2230 static long btrfs_ioctl_dev_info(struct btrfs_root *root, void __user *arg)
2231 {
2232         struct btrfs_ioctl_dev_info_args *di_args;
2233         struct btrfs_device *dev;
2234         struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
2235         int ret = 0;
2236         char *s_uuid = NULL;
2237         char empty_uuid[BTRFS_UUID_SIZE] = {0};
2238
2239         if (!capable(CAP_SYS_ADMIN))
2240                 return -EPERM;
2241
2242         di_args = memdup_user(arg, sizeof(*di_args));
2243         if (IS_ERR(di_args))
2244                 return PTR_ERR(di_args);
2245
2246         if (memcmp(empty_uuid, di_args->uuid, BTRFS_UUID_SIZE) != 0)
2247                 s_uuid = di_args->uuid;
2248
2249         mutex_lock(&fs_devices->device_list_mutex);
2250         dev = btrfs_find_device(root, di_args->devid, s_uuid, NULL);
2251         mutex_unlock(&fs_devices->device_list_mutex);
2252
2253         if (!dev) {
2254                 ret = -ENODEV;
2255                 goto out;
2256         }
2257
2258         di_args->devid = dev->devid;
2259         di_args->bytes_used = dev->bytes_used;
2260         di_args->total_bytes = dev->total_bytes;
2261         memcpy(di_args->uuid, dev->uuid, sizeof(di_args->uuid));
2262         strncpy(di_args->path, dev->name, sizeof(di_args->path));
2263
2264 out:
2265         if (ret == 0 && copy_to_user(arg, di_args, sizeof(*di_args)))
2266                 ret = -EFAULT;
2267
2268         kfree(di_args);
2269         return ret;
2270 }
2271
2272 static noinline long btrfs_ioctl_clone(struct file *file, unsigned long srcfd,
2273                                        u64 off, u64 olen, u64 destoff)
2274 {
2275         struct inode *inode = fdentry(file)->d_inode;
2276         struct btrfs_root *root = BTRFS_I(inode)->root;
2277         struct file *src_file;
2278         struct inode *src;
2279         struct btrfs_trans_handle *trans;
2280         struct btrfs_path *path;
2281         struct extent_buffer *leaf;
2282         char *buf;
2283         struct btrfs_key key;
2284         u32 nritems;
2285         int slot;
2286         int ret;
2287         u64 len = olen;
2288         u64 bs = root->fs_info->sb->s_blocksize;
2289         u64 hint_byte;
2290
2291         /*
2292          * TODO:
2293          * - split compressed inline extents.  annoying: we need to
2294          *   decompress into destination's address_space (the file offset
2295          *   may change, so source mapping won't do), then recompress (or
2296          *   otherwise reinsert) a subrange.
2297          * - allow ranges within the same file to be cloned (provided
2298          *   they don't overlap)?
2299          */
2300
2301         /* the destination must be opened for writing */
2302         if (!(file->f_mode & FMODE_WRITE) || (file->f_flags & O_APPEND))
2303                 return -EINVAL;
2304
2305         if (btrfs_root_readonly(root))
2306                 return -EROFS;
2307
2308         ret = mnt_want_write_file(file);
2309         if (ret)
2310                 return ret;
2311
2312         src_file = fget(srcfd);
2313         if (!src_file) {
2314                 ret = -EBADF;
2315                 goto out_drop_write;
2316         }
2317
2318         src = src_file->f_dentry->d_inode;
2319
2320         ret = -EINVAL;
2321         if (src == inode)
2322                 goto out_fput;
2323
2324         /* the src must be open for reading */
2325         if (!(src_file->f_mode & FMODE_READ))
2326                 goto out_fput;
2327
2328         /* don't make the dst file partly checksummed */
2329         if ((BTRFS_I(src)->flags & BTRFS_INODE_NODATASUM) !=
2330             (BTRFS_I(inode)->flags & BTRFS_INODE_NODATASUM))
2331                 goto out_fput;
2332
2333         ret = -EISDIR;
2334         if (S_ISDIR(src->i_mode) || S_ISDIR(inode->i_mode))
2335                 goto out_fput;
2336
2337         ret = -EXDEV;
2338         if (src->i_sb != inode->i_sb || BTRFS_I(src)->root != root)
2339                 goto out_fput;
2340
2341         ret = -ENOMEM;
2342         buf = vmalloc(btrfs_level_size(root, 0));
2343         if (!buf)
2344                 goto out_fput;
2345
2346         path = btrfs_alloc_path();
2347         if (!path) {
2348                 vfree(buf);
2349                 goto out_fput;
2350         }
2351         path->reada = 2;
2352
2353         if (inode < src) {
2354                 mutex_lock_nested(&inode->i_mutex, I_MUTEX_PARENT);
2355                 mutex_lock_nested(&src->i_mutex, I_MUTEX_CHILD);
2356         } else {
2357                 mutex_lock_nested(&src->i_mutex, I_MUTEX_PARENT);
2358                 mutex_lock_nested(&inode->i_mutex, I_MUTEX_CHILD);
2359         }
2360
2361         /* determine range to clone */
2362         ret = -EINVAL;
2363         if (off + len > src->i_size || off + len < off)
2364                 goto out_unlock;
2365         if (len == 0)
2366                 olen = len = src->i_size - off;
2367         /* if we extend to eof, continue to block boundary */
2368         if (off + len == src->i_size)
2369                 len = ALIGN(src->i_size, bs) - off;
2370
2371         /* verify the end result is block aligned */
2372         if (!IS_ALIGNED(off, bs) || !IS_ALIGNED(off + len, bs) ||
2373             !IS_ALIGNED(destoff, bs))
2374                 goto out_unlock;
2375
2376         if (destoff > inode->i_size) {
2377                 ret = btrfs_cont_expand(inode, inode->i_size, destoff);
2378                 if (ret)
2379                         goto out_unlock;
2380         }
2381
2382         /* truncate page cache pages from target inode range */
2383         truncate_inode_pages_range(&inode->i_data, destoff,
2384                                    PAGE_CACHE_ALIGN(destoff + len) - 1);
2385
2386         /* do any pending delalloc/csum calc on src, one way or
2387            another, and lock file content */
2388         while (1) {
2389                 struct btrfs_ordered_extent *ordered;
2390                 lock_extent(&BTRFS_I(src)->io_tree, off, off+len);
2391                 ordered = btrfs_lookup_first_ordered_extent(src, off+len);
2392                 if (!ordered &&
2393                     !test_range_bit(&BTRFS_I(src)->io_tree, off, off+len,
2394                                    EXTENT_DELALLOC, 0, NULL))
2395                         break;
2396                 unlock_extent(&BTRFS_I(src)->io_tree, off, off+len);
2397                 if (ordered)
2398                         btrfs_put_ordered_extent(ordered);
2399                 btrfs_wait_ordered_range(src, off, len);
2400         }
2401
2402         /* clone data */
2403         key.objectid = btrfs_ino(src);
2404         key.type = BTRFS_EXTENT_DATA_KEY;
2405         key.offset = 0;
2406
2407         while (1) {
2408                 /*
2409                  * note the key will change type as we walk through the
2410                  * tree.
2411                  */
2412                 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
2413                 if (ret < 0)
2414                         goto out;
2415
2416                 nritems = btrfs_header_nritems(path->nodes[0]);
2417                 if (path->slots[0] >= nritems) {
2418                         ret = btrfs_next_leaf(root, path);
2419                         if (ret < 0)
2420                                 goto out;
2421                         if (ret > 0)
2422                                 break;
2423                         nritems = btrfs_header_nritems(path->nodes[0]);
2424                 }
2425                 leaf = path->nodes[0];
2426                 slot = path->slots[0];
2427
2428                 btrfs_item_key_to_cpu(leaf, &key, slot);
2429                 if (btrfs_key_type(&key) > BTRFS_EXTENT_DATA_KEY ||
2430                     key.objectid != btrfs_ino(src))
2431                         break;
2432
2433                 if (btrfs_key_type(&key) == BTRFS_EXTENT_DATA_KEY) {
2434                         struct btrfs_file_extent_item *extent;
2435                         int type;
2436                         u32 size;
2437                         struct btrfs_key new_key;
2438                         u64 disko = 0, diskl = 0;
2439                         u64 datao = 0, datal = 0;
2440                         u8 comp;
2441                         u64 endoff;
2442
2443                         size = btrfs_item_size_nr(leaf, slot);
2444                         read_extent_buffer(leaf, buf,
2445                                            btrfs_item_ptr_offset(leaf, slot),
2446                                            size);
2447
2448                         extent = btrfs_item_ptr(leaf, slot,
2449                                                 struct btrfs_file_extent_item);
2450                         comp = btrfs_file_extent_compression(leaf, extent);
2451                         type = btrfs_file_extent_type(leaf, extent);
2452                         if (type == BTRFS_FILE_EXTENT_REG ||
2453                             type == BTRFS_FILE_EXTENT_PREALLOC) {
2454                                 disko = btrfs_file_extent_disk_bytenr(leaf,
2455                                                                       extent);
2456                                 diskl = btrfs_file_extent_disk_num_bytes(leaf,
2457                                                                  extent);
2458                                 datao = btrfs_file_extent_offset(leaf, extent);
2459                                 datal = btrfs_file_extent_num_bytes(leaf,
2460                                                                     extent);
2461                         } else if (type == BTRFS_FILE_EXTENT_INLINE) {
2462                                 /* take upper bound, may be compressed */
2463                                 datal = btrfs_file_extent_ram_bytes(leaf,
2464                                                                     extent);
2465                         }
2466                         btrfs_release_path(path);
2467
2468                         if (key.offset + datal <= off ||
2469                             key.offset >= off+len)
2470                                 goto next;
2471
2472                         memcpy(&new_key, &key, sizeof(new_key));
2473                         new_key.objectid = btrfs_ino(inode);
2474                         if (off <= key.offset)
2475                                 new_key.offset = key.offset + destoff - off;
2476                         else
2477                                 new_key.offset = destoff;
2478
2479                         /*
2480                          * 1 - adjusting old extent (we may have to split it)
2481                          * 1 - add new extent
2482                          * 1 - inode update
2483                          */
2484                         trans = btrfs_start_transaction(root, 3);
2485                         if (IS_ERR(trans)) {
2486                                 ret = PTR_ERR(trans);
2487                                 goto out;
2488                         }
2489
2490                         if (type == BTRFS_FILE_EXTENT_REG ||
2491                             type == BTRFS_FILE_EXTENT_PREALLOC) {
2492                                 /*
2493                                  *    a  | --- range to clone ---|  b
2494                                  * | ------------- extent ------------- |
2495                                  */
2496
2497                                 /* substract range b */
2498                                 if (key.offset + datal > off + len)
2499                                         datal = off + len - key.offset;
2500
2501                                 /* substract range a */
2502                                 if (off > key.offset) {
2503                                         datao += off - key.offset;
2504                                         datal -= off - key.offset;
2505                                 }
2506
2507                                 ret = btrfs_drop_extents(trans, inode,
2508                                                          new_key.offset,
2509                                                          new_key.offset + datal,
2510                                                          &hint_byte, 1);
2511                                 if (ret) {
2512                                         btrfs_abort_transaction(trans, root,
2513                                                                 ret);
2514                                         btrfs_end_transaction(trans, root);
2515                                         goto out;
2516                                 }
2517
2518                                 ret = btrfs_insert_empty_item(trans, root, path,
2519                                                               &new_key, size);
2520                                 if (ret) {
2521                                         btrfs_abort_transaction(trans, root,
2522                                                                 ret);
2523                                         btrfs_end_transaction(trans, root);
2524                                         goto out;
2525                                 }
2526
2527                                 leaf = path->nodes[0];
2528                                 slot = path->slots[0];
2529                                 write_extent_buffer(leaf, buf,
2530                                             btrfs_item_ptr_offset(leaf, slot),
2531                                             size);
2532
2533                                 extent = btrfs_item_ptr(leaf, slot,
2534                                                 struct btrfs_file_extent_item);
2535
2536                                 /* disko == 0 means it's a hole */
2537                                 if (!disko)
2538                                         datao = 0;
2539
2540                                 btrfs_set_file_extent_offset(leaf, extent,
2541                                                              datao);
2542                                 btrfs_set_file_extent_num_bytes(leaf, extent,
2543                                                                 datal);
2544                                 if (disko) {
2545                                         inode_add_bytes(inode, datal);
2546                                         ret = btrfs_inc_extent_ref(trans, root,
2547                                                         disko, diskl, 0,
2548                                                         root->root_key.objectid,
2549                                                         btrfs_ino(inode),
2550                                                         new_key.offset - datao,
2551                                                         0);
2552                                         if (ret) {
2553                                                 btrfs_abort_transaction(trans,
2554                                                                         root,
2555                                                                         ret);
2556                                                 btrfs_end_transaction(trans,
2557                                                                       root);
2558                                                 goto out;
2559
2560                                         }
2561                                 }
2562                         } else if (type == BTRFS_FILE_EXTENT_INLINE) {
2563                                 u64 skip = 0;
2564                                 u64 trim = 0;
2565                                 if (off > key.offset) {
2566                                         skip = off - key.offset;
2567                                         new_key.offset += skip;
2568                                 }
2569
2570                                 if (key.offset + datal > off+len)
2571                                         trim = key.offset + datal - (off+len);
2572
2573                                 if (comp && (skip || trim)) {
2574                                         ret = -EINVAL;
2575                                         btrfs_end_transaction(trans, root);
2576                                         goto out;
2577                                 }
2578                                 size -= skip + trim;
2579                                 datal -= skip + trim;
2580
2581                                 ret = btrfs_drop_extents(trans, inode,
2582                                                          new_key.offset,
2583                                                          new_key.offset + datal,
2584                                                          &hint_byte, 1);
2585                                 if (ret) {
2586                                         btrfs_abort_transaction(trans, root,
2587                                                                 ret);
2588                                         btrfs_end_transaction(trans, root);
2589                                         goto out;
2590                                 }
2591
2592                                 ret = btrfs_insert_empty_item(trans, root, path,
2593                                                               &new_key, size);
2594                                 if (ret) {
2595                                         btrfs_abort_transaction(trans, root,
2596                                                                 ret);
2597                                         btrfs_end_transaction(trans, root);
2598                                         goto out;
2599                                 }
2600
2601                                 if (skip) {
2602                                         u32 start =
2603                                           btrfs_file_extent_calc_inline_size(0);
2604                                         memmove(buf+start, buf+start+skip,
2605                                                 datal);
2606                                 }
2607
2608                                 leaf = path->nodes[0];
2609                                 slot = path->slots[0];
2610                                 write_extent_buffer(leaf, buf,
2611                                             btrfs_item_ptr_offset(leaf, slot),
2612                                             size);
2613                                 inode_add_bytes(inode, datal);
2614                         }
2615
2616                         btrfs_mark_buffer_dirty(leaf);
2617                         btrfs_release_path(path);
2618
2619                         inode->i_mtime = inode->i_ctime = CURRENT_TIME;
2620
2621                         /*
2622                          * we round up to the block size at eof when
2623                          * determining which extents to clone above,
2624                          * but shouldn't round up the file size
2625                          */
2626                         endoff = new_key.offset + datal;
2627                         if (endoff > destoff+olen)
2628                                 endoff = destoff+olen;
2629                         if (endoff > inode->i_size)
2630                                 btrfs_i_size_write(inode, endoff);
2631
2632                         ret = btrfs_update_inode(trans, root, inode);
2633                         if (ret) {
2634                                 btrfs_abort_transaction(trans, root, ret);
2635                                 btrfs_end_transaction(trans, root);
2636                                 goto out;
2637                         }
2638                         ret = btrfs_end_transaction(trans, root);
2639                 }
2640 next:
2641                 btrfs_release_path(path);
2642                 key.offset++;
2643         }
2644         ret = 0;
2645 out:
2646         btrfs_release_path(path);
2647         unlock_extent(&BTRFS_I(src)->io_tree, off, off+len);
2648 out_unlock:
2649         mutex_unlock(&src->i_mutex);
2650         mutex_unlock(&inode->i_mutex);
2651         vfree(buf);
2652         btrfs_free_path(path);
2653 out_fput:
2654         fput(src_file);
2655 out_drop_write:
2656         mnt_drop_write_file(file);
2657         return ret;
2658 }
2659
2660 static long btrfs_ioctl_clone_range(struct file *file, void __user *argp)
2661 {
2662         struct btrfs_ioctl_clone_range_args args;
2663
2664         if (copy_from_user(&args, argp, sizeof(args)))
2665                 return -EFAULT;
2666         return btrfs_ioctl_clone(file, args.src_fd, args.src_offset,
2667                                  args.src_length, args.dest_offset);
2668 }
2669
2670 /*
2671  * there are many ways the trans_start and trans_end ioctls can lead
2672  * to deadlocks.  They should only be used by applications that
2673  * basically own the machine, and have a very in depth understanding
2674  * of all the possible deadlocks and enospc problems.
2675  */
2676 static long btrfs_ioctl_trans_start(struct file *file)
2677 {
2678         struct inode *inode = fdentry(file)->d_inode;
2679         struct btrfs_root *root = BTRFS_I(inode)->root;
2680         struct btrfs_trans_handle *trans;
2681         int ret;
2682
2683         ret = -EPERM;
2684         if (!capable(CAP_SYS_ADMIN))
2685                 goto out;
2686
2687         ret = -EINPROGRESS;
2688         if (file->private_data)
2689                 goto out;
2690
2691         ret = -EROFS;
2692         if (btrfs_root_readonly(root))
2693                 goto out;
2694
2695         ret = mnt_want_write_file(file);
2696         if (ret)
2697                 goto out;
2698
2699         atomic_inc(&root->fs_info->open_ioctl_trans);
2700
2701         ret = -ENOMEM;
2702         trans = btrfs_start_ioctl_transaction(root);
2703         if (IS_ERR(trans))
2704                 goto out_drop;
2705
2706         file->private_data = trans;
2707         return 0;
2708
2709 out_drop:
2710         atomic_dec(&root->fs_info->open_ioctl_trans);
2711         mnt_drop_write_file(file);
2712 out:
2713         return ret;
2714 }
2715
2716 static long btrfs_ioctl_default_subvol(struct file *file, void __user *argp)
2717 {
2718         struct inode *inode = fdentry(file)->d_inode;
2719         struct btrfs_root *root = BTRFS_I(inode)->root;
2720         struct btrfs_root *new_root;
2721         struct btrfs_dir_item *di;
2722         struct btrfs_trans_handle *trans;
2723         struct btrfs_path *path;
2724         struct btrfs_key location;
2725         struct btrfs_disk_key disk_key;
2726         struct btrfs_super_block *disk_super;
2727         u64 features;
2728         u64 objectid = 0;
2729         u64 dir_id;
2730
2731         if (!capable(CAP_SYS_ADMIN))
2732                 return -EPERM;
2733
2734         if (copy_from_user(&objectid, argp, sizeof(objectid)))
2735                 return -EFAULT;
2736
2737         if (!objectid)
2738                 objectid = root->root_key.objectid;
2739
2740         location.objectid = objectid;
2741         location.type = BTRFS_ROOT_ITEM_KEY;
2742         location.offset = (u64)-1;
2743
2744         new_root = btrfs_read_fs_root_no_name(root->fs_info, &location);
2745         if (IS_ERR(new_root))
2746                 return PTR_ERR(new_root);
2747
2748         if (btrfs_root_refs(&new_root->root_item) == 0)
2749                 return -ENOENT;
2750
2751         path = btrfs_alloc_path();
2752         if (!path)
2753                 return -ENOMEM;
2754         path->leave_spinning = 1;
2755
2756         trans = btrfs_start_transaction(root, 1);
2757         if (IS_ERR(trans)) {
2758                 btrfs_free_path(path);
2759                 return PTR_ERR(trans);
2760         }
2761
2762         dir_id = btrfs_super_root_dir(root->fs_info->super_copy);
2763         di = btrfs_lookup_dir_item(trans, root->fs_info->tree_root, path,
2764                                    dir_id, "default", 7, 1);
2765         if (IS_ERR_OR_NULL(di)) {
2766                 btrfs_free_path(path);
2767                 btrfs_end_transaction(trans, root);
2768                 printk(KERN_ERR "Umm, you don't have the default dir item, "
2769                        "this isn't going to work\n");
2770                 return -ENOENT;
2771         }
2772
2773         btrfs_cpu_key_to_disk(&disk_key, &new_root->root_key);
2774         btrfs_set_dir_item_key(path->nodes[0], di, &disk_key);
2775         btrfs_mark_buffer_dirty(path->nodes[0]);
2776         btrfs_free_path(path);
2777
2778         disk_super = root->fs_info->super_copy;
2779         features = btrfs_super_incompat_flags(disk_super);
2780         if (!(features & BTRFS_FEATURE_INCOMPAT_DEFAULT_SUBVOL)) {
2781                 features |= BTRFS_FEATURE_INCOMPAT_DEFAULT_SUBVOL;
2782                 btrfs_set_super_incompat_flags(disk_super, features);
2783         }
2784         btrfs_end_transaction(trans, root);
2785
2786         return 0;
2787 }
2788
2789 static void get_block_group_info(struct list_head *groups_list,
2790                                  struct btrfs_ioctl_space_info *space)
2791 {
2792         struct btrfs_block_group_cache *block_group;
2793
2794         space->total_bytes = 0;
2795         space->used_bytes = 0;
2796         space->flags = 0;
2797         list_for_each_entry(block_group, groups_list, list) {
2798                 space->flags = block_group->flags;
2799                 space->total_bytes += block_group->key.offset;
2800                 space->used_bytes +=
2801                         btrfs_block_group_used(&block_group->item);
2802         }
2803 }
2804
2805 long btrfs_ioctl_space_info(struct btrfs_root *root, void __user *arg)
2806 {
2807         struct btrfs_ioctl_space_args space_args;
2808         struct btrfs_ioctl_space_info space;
2809         struct btrfs_ioctl_space_info *dest;
2810         struct btrfs_ioctl_space_info *dest_orig;
2811         struct btrfs_ioctl_space_info __user *user_dest;
2812         struct btrfs_space_info *info;
2813         u64 types[] = {BTRFS_BLOCK_GROUP_DATA,
2814                        BTRFS_BLOCK_GROUP_SYSTEM,
2815                        BTRFS_BLOCK_GROUP_METADATA,
2816                        BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA};
2817         int num_types = 4;
2818         int alloc_size;
2819         int ret = 0;
2820         u64 slot_count = 0;
2821         int i, c;
2822
2823         if (copy_from_user(&space_args,
2824                            (struct btrfs_ioctl_space_args __user *)arg,
2825                            sizeof(space_args)))
2826                 return -EFAULT;
2827
2828         for (i = 0; i < num_types; i++) {
2829                 struct btrfs_space_info *tmp;
2830
2831                 info = NULL;
2832                 rcu_read_lock();
2833                 list_for_each_entry_rcu(tmp, &root->fs_info->space_info,
2834                                         list) {
2835                         if (tmp->flags == types[i]) {
2836                                 info = tmp;
2837                                 break;
2838                         }
2839                 }
2840                 rcu_read_unlock();
2841
2842                 if (!info)
2843                         continue;
2844
2845                 down_read(&info->groups_sem);
2846                 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
2847                         if (!list_empty(&info->block_groups[c]))
2848                                 slot_count++;
2849                 }
2850                 up_read(&info->groups_sem);
2851         }
2852
2853         /* space_slots == 0 means they are asking for a count */
2854         if (space_args.space_slots == 0) {
2855                 space_args.total_spaces = slot_count;
2856                 goto out;
2857         }
2858
2859         slot_count = min_t(u64, space_args.space_slots, slot_count);
2860
2861         alloc_size = sizeof(*dest) * slot_count;
2862
2863         /* we generally have at most 6 or so space infos, one for each raid
2864          * level.  So, a whole page should be more than enough for everyone
2865          */
2866         if (alloc_size > PAGE_CACHE_SIZE)
2867                 return -ENOMEM;
2868
2869         space_args.total_spaces = 0;
2870         dest = kmalloc(alloc_size, GFP_NOFS);
2871         if (!dest)
2872                 return -ENOMEM;
2873         dest_orig = dest;
2874
2875         /* now we have a buffer to copy into */
2876         for (i = 0; i < num_types; i++) {
2877                 struct btrfs_space_info *tmp;
2878
2879                 if (!slot_count)
2880                         break;
2881
2882                 info = NULL;
2883                 rcu_read_lock();
2884                 list_for_each_entry_rcu(tmp, &root->fs_info->space_info,
2885                                         list) {
2886                         if (tmp->flags == types[i]) {
2887                                 info = tmp;
2888                                 break;
2889                         }
2890                 }
2891                 rcu_read_unlock();
2892
2893                 if (!info)
2894                         continue;
2895                 down_read(&info->groups_sem);
2896                 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
2897                         if (!list_empty(&info->block_groups[c])) {
2898                                 get_block_group_info(&info->block_groups[c],
2899                                                      &space);
2900                                 memcpy(dest, &space, sizeof(space));
2901                                 dest++;
2902                                 space_args.total_spaces++;
2903                                 slot_count--;
2904                         }
2905                         if (!slot_count)
2906                                 break;
2907                 }
2908                 up_read(&info->groups_sem);
2909         }
2910
2911         user_dest = (struct btrfs_ioctl_space_info *)
2912                 (arg + sizeof(struct btrfs_ioctl_space_args));
2913
2914         if (copy_to_user(user_dest, dest_orig, alloc_size))
2915                 ret = -EFAULT;
2916
2917         kfree(dest_orig);
2918 out:
2919         if (ret == 0 && copy_to_user(arg, &space_args, sizeof(space_args)))
2920                 ret = -EFAULT;
2921
2922         return ret;
2923 }
2924
2925 /*
2926  * there are many ways the trans_start and trans_end ioctls can lead
2927  * to deadlocks.  They should only be used by applications that
2928  * basically own the machine, and have a very in depth understanding
2929  * of all the possible deadlocks and enospc problems.
2930  */
2931 long btrfs_ioctl_trans_end(struct file *file)
2932 {
2933         struct inode *inode = fdentry(file)->d_inode;
2934         struct btrfs_root *root = BTRFS_I(inode)->root;
2935         struct btrfs_trans_handle *trans;
2936
2937         trans = file->private_data;
2938         if (!trans)
2939                 return -EINVAL;
2940         file->private_data = NULL;
2941
2942         btrfs_end_transaction(trans, root);
2943
2944         atomic_dec(&root->fs_info->open_ioctl_trans);
2945
2946         mnt_drop_write_file(file);
2947         return 0;
2948 }
2949
2950 static noinline long btrfs_ioctl_start_sync(struct file *file, void __user *argp)
2951 {
2952         struct btrfs_root *root = BTRFS_I(file->f_dentry->d_inode)->root;
2953         struct btrfs_trans_handle *trans;
2954         u64 transid;
2955         int ret;
2956
2957         trans = btrfs_start_transaction(root, 0);
2958         if (IS_ERR(trans))
2959                 return PTR_ERR(trans);
2960         transid = trans->transid;
2961         ret = btrfs_commit_transaction_async(trans, root, 0);
2962         if (ret) {
2963                 btrfs_end_transaction(trans, root);
2964                 return ret;
2965         }
2966
2967         if (argp)
2968                 if (copy_to_user(argp, &transid, sizeof(transid)))
2969                         return -EFAULT;
2970         return 0;
2971 }
2972
2973 static noinline long btrfs_ioctl_wait_sync(struct file *file, void __user *argp)
2974 {
2975         struct btrfs_root *root = BTRFS_I(file->f_dentry->d_inode)->root;
2976         u64 transid;
2977
2978         if (argp) {
2979                 if (copy_from_user(&transid, argp, sizeof(transid)))
2980                         return -EFAULT;
2981         } else {
2982                 transid = 0;  /* current trans */
2983         }
2984         return btrfs_wait_for_commit(root, transid);
2985 }
2986
2987 static long btrfs_ioctl_scrub(struct btrfs_root *root, void __user *arg)
2988 {
2989         int ret;
2990         struct btrfs_ioctl_scrub_args *sa;
2991
2992         if (!capable(CAP_SYS_ADMIN))
2993                 return -EPERM;
2994
2995         sa = memdup_user(arg, sizeof(*sa));
2996         if (IS_ERR(sa))
2997                 return PTR_ERR(sa);
2998
2999         ret = btrfs_scrub_dev(root, sa->devid, sa->start, sa->end,
3000                               &sa->progress, sa->flags & BTRFS_SCRUB_READONLY);
3001
3002         if (copy_to_user(arg, sa, sizeof(*sa)))
3003                 ret = -EFAULT;
3004
3005         kfree(sa);
3006         return ret;
3007 }
3008
3009 static long btrfs_ioctl_scrub_cancel(struct btrfs_root *root, void __user *arg)
3010 {
3011         if (!capable(CAP_SYS_ADMIN))
3012                 return -EPERM;
3013
3014         return btrfs_scrub_cancel(root);
3015 }
3016
3017 static long btrfs_ioctl_scrub_progress(struct btrfs_root *root,
3018                                        void __user *arg)
3019 {
3020         struct btrfs_ioctl_scrub_args *sa;
3021         int ret;
3022
3023         if (!capable(CAP_SYS_ADMIN))
3024                 return -EPERM;
3025
3026         sa = memdup_user(arg, sizeof(*sa));
3027         if (IS_ERR(sa))
3028                 return PTR_ERR(sa);
3029
3030         ret = btrfs_scrub_progress(root, sa->devid, &sa->progress);
3031
3032         if (copy_to_user(arg, sa, sizeof(*sa)))
3033                 ret = -EFAULT;
3034
3035         kfree(sa);
3036         return ret;
3037 }
3038
3039 static long btrfs_ioctl_ino_to_path(struct btrfs_root *root, void __user *arg)
3040 {
3041         int ret = 0;
3042         int i;
3043         u64 rel_ptr;
3044         int size;
3045         struct btrfs_ioctl_ino_path_args *ipa = NULL;
3046         struct inode_fs_paths *ipath = NULL;
3047         struct btrfs_path *path;
3048
3049         if (!capable(CAP_SYS_ADMIN))
3050                 return -EPERM;
3051
3052         path = btrfs_alloc_path();
3053         if (!path) {
3054                 ret = -ENOMEM;
3055                 goto out;
3056         }
3057
3058         ipa = memdup_user(arg, sizeof(*ipa));
3059         if (IS_ERR(ipa)) {
3060                 ret = PTR_ERR(ipa);
3061                 ipa = NULL;
3062                 goto out;
3063         }
3064
3065         size = min_t(u32, ipa->size, 4096);
3066         ipath = init_ipath(size, root, path);
3067         if (IS_ERR(ipath)) {
3068                 ret = PTR_ERR(ipath);
3069                 ipath = NULL;
3070                 goto out;
3071         }
3072
3073         ret = paths_from_inode(ipa->inum, ipath);
3074         if (ret < 0)
3075                 goto out;
3076
3077         for (i = 0; i < ipath->fspath->elem_cnt; ++i) {
3078                 rel_ptr = ipath->fspath->val[i] -
3079                           (u64)(unsigned long)ipath->fspath->val;
3080                 ipath->fspath->val[i] = rel_ptr;
3081         }
3082
3083         ret = copy_to_user((void *)(unsigned long)ipa->fspath,
3084                            (void *)(unsigned long)ipath->fspath, size);
3085         if (ret) {
3086                 ret = -EFAULT;
3087                 goto out;
3088         }
3089
3090 out:
3091         btrfs_free_path(path);
3092         free_ipath(ipath);
3093         kfree(ipa);
3094
3095         return ret;
3096 }
3097
3098 static int build_ino_list(u64 inum, u64 offset, u64 root, void *ctx)
3099 {
3100         struct btrfs_data_container *inodes = ctx;
3101         const size_t c = 3 * sizeof(u64);
3102
3103         if (inodes->bytes_left >= c) {
3104                 inodes->bytes_left -= c;
3105                 inodes->val[inodes->elem_cnt] = inum;
3106                 inodes->val[inodes->elem_cnt + 1] = offset;
3107                 inodes->val[inodes->elem_cnt + 2] = root;
3108                 inodes->elem_cnt += 3;
3109         } else {
3110                 inodes->bytes_missing += c - inodes->bytes_left;
3111                 inodes->bytes_left = 0;
3112                 inodes->elem_missed += 3;
3113         }
3114
3115         return 0;
3116 }
3117
3118 static long btrfs_ioctl_logical_to_ino(struct btrfs_root *root,
3119                                         void __user *arg)
3120 {
3121         int ret = 0;
3122         int size;
3123         u64 extent_item_pos;
3124         struct btrfs_ioctl_logical_ino_args *loi;
3125         struct btrfs_data_container *inodes = NULL;
3126         struct btrfs_path *path = NULL;
3127         struct btrfs_key key;
3128
3129         if (!capable(CAP_SYS_ADMIN))
3130                 return -EPERM;
3131
3132         loi = memdup_user(arg, sizeof(*loi));
3133         if (IS_ERR(loi)) {
3134                 ret = PTR_ERR(loi);
3135                 loi = NULL;
3136                 goto out;
3137         }
3138
3139         path = btrfs_alloc_path();
3140         if (!path) {
3141                 ret = -ENOMEM;
3142                 goto out;
3143         }
3144
3145         size = min_t(u32, loi->size, 4096);
3146         inodes = init_data_container(size);
3147         if (IS_ERR(inodes)) {
3148                 ret = PTR_ERR(inodes);
3149                 inodes = NULL;
3150                 goto out;
3151         }
3152
3153         ret = extent_from_logical(root->fs_info, loi->logical, path, &key);
3154         btrfs_release_path(path);
3155
3156         if (ret & BTRFS_EXTENT_FLAG_TREE_BLOCK)
3157                 ret = -ENOENT;
3158         if (ret < 0)
3159                 goto out;
3160
3161         extent_item_pos = loi->logical - key.objectid;
3162         ret = iterate_extent_inodes(root->fs_info, key.objectid,
3163                                         extent_item_pos, 0, build_ino_list,
3164                                         inodes);
3165
3166         if (ret < 0)
3167                 goto out;
3168
3169         ret = copy_to_user((void *)(unsigned long)loi->inodes,
3170                            (void *)(unsigned long)inodes, size);
3171         if (ret)
3172                 ret = -EFAULT;
3173
3174 out:
3175         btrfs_free_path(path);
3176         kfree(inodes);
3177         kfree(loi);
3178
3179         return ret;
3180 }
3181
3182 void update_ioctl_balance_args(struct btrfs_fs_info *fs_info, int lock,
3183                                struct btrfs_ioctl_balance_args *bargs)
3184 {
3185         struct btrfs_balance_control *bctl = fs_info->balance_ctl;
3186
3187         bargs->flags = bctl->flags;
3188
3189         if (atomic_read(&fs_info->balance_running))
3190                 bargs->state |= BTRFS_BALANCE_STATE_RUNNING;
3191         if (atomic_read(&fs_info->balance_pause_req))
3192                 bargs->state |= BTRFS_BALANCE_STATE_PAUSE_REQ;
3193         if (atomic_read(&fs_info->balance_cancel_req))
3194                 bargs->state |= BTRFS_BALANCE_STATE_CANCEL_REQ;
3195
3196         memcpy(&bargs->data, &bctl->data, sizeof(bargs->data));
3197         memcpy(&bargs->meta, &bctl->meta, sizeof(bargs->meta));
3198         memcpy(&bargs->sys, &bctl->sys, sizeof(bargs->sys));
3199
3200         if (lock) {
3201                 spin_lock(&fs_info->balance_lock);
3202                 memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat));
3203                 spin_unlock(&fs_info->balance_lock);
3204         } else {
3205                 memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat));
3206         }
3207 }
3208
3209 static long btrfs_ioctl_balance(struct btrfs_root *root, void __user *arg)
3210 {
3211         struct btrfs_fs_info *fs_info = root->fs_info;
3212         struct btrfs_ioctl_balance_args *bargs;
3213         struct btrfs_balance_control *bctl;
3214         int ret;
3215
3216         if (!capable(CAP_SYS_ADMIN))
3217                 return -EPERM;
3218
3219         if (fs_info->sb->s_flags & MS_RDONLY)
3220                 return -EROFS;
3221
3222         mutex_lock(&fs_info->volume_mutex);
3223         mutex_lock(&fs_info->balance_mutex);
3224
3225         if (arg) {
3226                 bargs = memdup_user(arg, sizeof(*bargs));
3227                 if (IS_ERR(bargs)) {
3228                         ret = PTR_ERR(bargs);
3229                         goto out;
3230                 }
3231
3232                 if (bargs->flags & BTRFS_BALANCE_RESUME) {
3233                         if (!fs_info->balance_ctl) {
3234                                 ret = -ENOTCONN;
3235                                 goto out_bargs;
3236                         }
3237
3238                         bctl = fs_info->balance_ctl;
3239                         spin_lock(&fs_info->balance_lock);
3240                         bctl->flags |= BTRFS_BALANCE_RESUME;
3241                         spin_unlock(&fs_info->balance_lock);
3242
3243                         goto do_balance;
3244                 }
3245         } else {
3246                 bargs = NULL;
3247         }
3248
3249         if (fs_info->balance_ctl) {
3250                 ret = -EINPROGRESS;
3251                 goto out_bargs;
3252         }
3253
3254         bctl = kzalloc(sizeof(*bctl), GFP_NOFS);
3255         if (!bctl) {
3256                 ret = -ENOMEM;
3257                 goto out_bargs;
3258         }
3259
3260         bctl->fs_info = fs_info;
3261         if (arg) {
3262                 memcpy(&bctl->data, &bargs->data, sizeof(bctl->data));
3263                 memcpy(&bctl->meta, &bargs->meta, sizeof(bctl->meta));
3264                 memcpy(&bctl->sys, &bargs->sys, sizeof(bctl->sys));
3265
3266                 bctl->flags = bargs->flags;
3267         } else {
3268                 /* balance everything - no filters */
3269                 bctl->flags |= BTRFS_BALANCE_TYPE_MASK;
3270         }
3271
3272 do_balance:
3273         ret = btrfs_balance(bctl, bargs);
3274         /*
3275          * bctl is freed in __cancel_balance or in free_fs_info if
3276          * restriper was paused all the way until unmount
3277          */
3278         if (arg) {
3279                 if (copy_to_user(arg, bargs, sizeof(*bargs)))
3280                         ret = -EFAULT;
3281         }
3282
3283 out_bargs:
3284         kfree(bargs);
3285 out:
3286         mutex_unlock(&fs_info->balance_mutex);
3287         mutex_unlock(&fs_info->volume_mutex);
3288         return ret;
3289 }
3290
3291 static long btrfs_ioctl_balance_ctl(struct btrfs_root *root, int cmd)
3292 {
3293         if (!capable(CAP_SYS_ADMIN))
3294                 return -EPERM;
3295
3296         switch (cmd) {
3297         case BTRFS_BALANCE_CTL_PAUSE:
3298                 return btrfs_pause_balance(root->fs_info);
3299         case BTRFS_BALANCE_CTL_CANCEL:
3300                 return btrfs_cancel_balance(root->fs_info);
3301         }
3302
3303         return -EINVAL;
3304 }
3305
3306 static long btrfs_ioctl_balance_progress(struct btrfs_root *root,
3307                                          void __user *arg)
3308 {
3309         struct btrfs_fs_info *fs_info = root->fs_info;
3310         struct btrfs_ioctl_balance_args *bargs;
3311         int ret = 0;
3312
3313         if (!capable(CAP_SYS_ADMIN))
3314                 return -EPERM;
3315
3316         mutex_lock(&fs_info->balance_mutex);
3317         if (!fs_info->balance_ctl) {
3318                 ret = -ENOTCONN;
3319                 goto out;
3320         }
3321
3322         bargs = kzalloc(sizeof(*bargs), GFP_NOFS);
3323         if (!bargs) {
3324                 ret = -ENOMEM;
3325                 goto out;
3326         }
3327
3328         update_ioctl_balance_args(fs_info, 1, bargs);
3329
3330         if (copy_to_user(arg, bargs, sizeof(*bargs)))
3331                 ret = -EFAULT;
3332
3333         kfree(bargs);
3334 out:
3335         mutex_unlock(&fs_info->balance_mutex);
3336         return ret;
3337 }
3338
3339 long btrfs_ioctl(struct file *file, unsigned int
3340                 cmd, unsigned long arg)
3341 {
3342         struct btrfs_root *root = BTRFS_I(fdentry(file)->d_inode)->root;
3343         void __user *argp = (void __user *)arg;
3344
3345         switch (cmd) {
3346         case FS_IOC_GETFLAGS:
3347                 return btrfs_ioctl_getflags(file, argp);
3348         case FS_IOC_SETFLAGS:
3349                 return btrfs_ioctl_setflags(file, argp);
3350         case FS_IOC_GETVERSION:
3351                 return btrfs_ioctl_getversion(file, argp);
3352         case FITRIM:
3353                 return btrfs_ioctl_fitrim(file, argp);
3354         case BTRFS_IOC_SNAP_CREATE:
3355                 return btrfs_ioctl_snap_create(file, argp, 0);
3356         case BTRFS_IOC_SNAP_CREATE_V2:
3357                 return btrfs_ioctl_snap_create_v2(file, argp, 0);
3358         case BTRFS_IOC_SUBVOL_CREATE:
3359                 return btrfs_ioctl_snap_create(file, argp, 1);
3360         case BTRFS_IOC_SNAP_DESTROY:
3361                 return btrfs_ioctl_snap_destroy(file, argp);
3362         case BTRFS_IOC_SUBVOL_GETFLAGS:
3363                 return btrfs_ioctl_subvol_getflags(file, argp);
3364         case BTRFS_IOC_SUBVOL_SETFLAGS:
3365                 return btrfs_ioctl_subvol_setflags(file, argp);
3366         case BTRFS_IOC_DEFAULT_SUBVOL:
3367                 return btrfs_ioctl_default_subvol(file, argp);
3368         case BTRFS_IOC_DEFRAG:
3369                 return btrfs_ioctl_defrag(file, NULL);
3370         case BTRFS_IOC_DEFRAG_RANGE:
3371                 return btrfs_ioctl_defrag(file, argp);
3372         case BTRFS_IOC_RESIZE:
3373                 return btrfs_ioctl_resize(root, argp);
3374         case BTRFS_IOC_ADD_DEV:
3375                 return btrfs_ioctl_add_dev(root, argp);
3376         case BTRFS_IOC_RM_DEV:
3377                 return btrfs_ioctl_rm_dev(root, argp);
3378         case BTRFS_IOC_FS_INFO:
3379                 return btrfs_ioctl_fs_info(root, argp);
3380         case BTRFS_IOC_DEV_INFO:
3381                 return btrfs_ioctl_dev_info(root, argp);
3382         case BTRFS_IOC_BALANCE:
3383                 return btrfs_ioctl_balance(root, NULL);
3384         case BTRFS_IOC_CLONE:
3385                 return btrfs_ioctl_clone(file, arg, 0, 0, 0);
3386         case BTRFS_IOC_CLONE_RANGE:
3387                 return btrfs_ioctl_clone_range(file, argp);
3388         case BTRFS_IOC_TRANS_START:
3389                 return btrfs_ioctl_trans_start(file);
3390         case BTRFS_IOC_TRANS_END:
3391                 return btrfs_ioctl_trans_end(file);
3392         case BTRFS_IOC_TREE_SEARCH:
3393                 return btrfs_ioctl_tree_search(file, argp);
3394         case BTRFS_IOC_INO_LOOKUP:
3395                 return btrfs_ioctl_ino_lookup(file, argp);
3396         case BTRFS_IOC_INO_PATHS:
3397                 return btrfs_ioctl_ino_to_path(root, argp);
3398         case BTRFS_IOC_LOGICAL_INO:
3399                 return btrfs_ioctl_logical_to_ino(root, argp);
3400         case BTRFS_IOC_SPACE_INFO:
3401                 return btrfs_ioctl_space_info(root, argp);
3402         case BTRFS_IOC_SYNC:
3403                 btrfs_sync_fs(file->f_dentry->d_sb, 1);
3404                 return 0;
3405         case BTRFS_IOC_START_SYNC:
3406                 return btrfs_ioctl_start_sync(file, argp);
3407         case BTRFS_IOC_WAIT_SYNC:
3408                 return btrfs_ioctl_wait_sync(file, argp);
3409         case BTRFS_IOC_SCRUB:
3410                 return btrfs_ioctl_scrub(root, argp);
3411         case BTRFS_IOC_SCRUB_CANCEL:
3412                 return btrfs_ioctl_scrub_cancel(root, argp);
3413         case BTRFS_IOC_SCRUB_PROGRESS:
3414                 return btrfs_ioctl_scrub_progress(root, argp);
3415         case BTRFS_IOC_BALANCE_V2:
3416                 return btrfs_ioctl_balance(root, argp);
3417         case BTRFS_IOC_BALANCE_CTL:
3418                 return btrfs_ioctl_balance_ctl(root, arg);
3419         case BTRFS_IOC_BALANCE_PROGRESS:
3420                 return btrfs_ioctl_balance_progress(root, argp);
3421         }
3422
3423         return -ENOTTY;
3424 }