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