fs/btrfs/ioctl.c

   1 /*
   2  * Copyright (C) 2007 Oracle.  All rights reserved.
   3  *
   4  * This program is free software; you can redistribute it and/or
   5  * modify it under the terms of the GNU General Public
   6  * License v2 as published by the Free Software Foundation.
   7  *
   8  * This program is distributed in the hope that it will be useful,
   9  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  10  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  11  * General Public License for more details.
  12  *
  13  * You should have received a copy of the GNU General Public
  14  * License along with this program; if not, write to the
  15  * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
  16  * Boston, MA 021110-1307, USA.
  17  */
  18
  19 #include <linux/kernel.h>
  20 #include <linux/bio.h>
  21 #include <linux/buffer_head.h>
  22 #include <linux/file.h>
  23 #include <linux/fs.h>
  24 #include <linux/fsnotify.h>
  25 #include <linux/pagemap.h>
  26 #include <linux/highmem.h>
  27 #include <linux/time.h>
  28 #include <linux/init.h>
  29 #include <linux/string.h>
  30 #include <linux/backing-dev.h>
  31 #include <linux/mount.h>
  32 #include <linux/mpage.h>
  33 #include <linux/namei.h>
  34 #include <linux/swap.h>
  35 #include <linux/writeback.h>
  36 #include <linux/statfs.h>
  37 #include <linux/compat.h>
  38 #include <linux/bit_spinlock.h>
  39 #include <linux/security.h>
  40 #include <linux/xattr.h>
  41 #include <linux/vmalloc.h>
  42 #include <linux/slab.h>
  43 #include <linux/blkdev.h>
  44 #include "compat.h"
  45 #include "ctree.h"
  46 #include "disk-io.h"
  47 #include "transaction.h"
  48 #include "btrfs_inode.h"
  49 #include "ioctl.h"
  50 #include "print-tree.h"
  51 #include "volumes.h"
  52 #include "locking.h"
  53 #include "inode-map.h"
  54 #include "backref.h"
  55 #include "rcu-string.h"
  56
  57 /* Mask out flags that are inappropriate for the given type of inode. */
  58 static inline __u32 btrfs_mask_flags(umode_t mode, __u32 flags)
  59 {
  60         if (S_ISDIR(mode))
  61                 return flags;
  62         else if (S_ISREG(mode))
  63                 return flags & ~FS_DIRSYNC_FL;
  64         else
  65                 return flags & (FS_NODUMP_FL | FS_NOATIME_FL);
  66 }
  67
  68 /*
  69  * Export inode flags to the format expected by the FS_IOC_GETFLAGS ioctl.
  70  */
  71 static unsigned int btrfs_flags_to_ioctl(unsigned int flags)
  72 {
  73         unsigned int iflags = 0;
  74
  75         if (flags & BTRFS_INODE_SYNC)
  76                 iflags |= FS_SYNC_FL;
  77         if (flags & BTRFS_INODE_IMMUTABLE)
  78                 iflags |= FS_IMMUTABLE_FL;
  79         if (flags & BTRFS_INODE_APPEND)
  80                 iflags |= FS_APPEND_FL;
  81         if (flags & BTRFS_INODE_NODUMP)
  82                 iflags |= FS_NODUMP_FL;
  83         if (flags & BTRFS_INODE_NOATIME)
  84                 iflags |= FS_NOATIME_FL;
  85         if (flags & BTRFS_INODE_DIRSYNC)
  86                 iflags |= FS_DIRSYNC_FL;
  87         if (flags & BTRFS_INODE_NODATACOW)
  88                 iflags |= FS_NOCOW_FL;
  89
  90         if ((flags & BTRFS_INODE_COMPRESS) && !(flags & BTRFS_INODE_NOCOMPRESS))
  91                 iflags |= FS_COMPR_FL;
  92         else if (flags & BTRFS_INODE_NOCOMPRESS)
  93                 iflags |= FS_NOCOMP_FL;
  94
  95         return iflags;
  96 }
  97
  98 /*
  99  * Update inode->i_flags based on the btrfs internal flags.
 100  */
 101 void btrfs_update_iflags(struct inode *inode)
 102 {
 103         struct btrfs_inode *ip = BTRFS_I(inode);
 104
 105         inode->i_flags &= ~(S_SYNC|S_APPEND|S_IMMUTABLE|S_NOATIME|S_DIRSYNC);
 106
 107         if (ip->flags & BTRFS_INODE_SYNC)
 108                 inode->i_flags |= S_SYNC;
 109         if (ip->flags & BTRFS_INODE_IMMUTABLE)
 110                 inode->i_flags |= S_IMMUTABLE;
 111         if (ip->flags & BTRFS_INODE_APPEND)
 112                 inode->i_flags |= S_APPEND;
 113         if (ip->flags & BTRFS_INODE_NOATIME)
 114                 inode->i_flags |= S_NOATIME;
 115         if (ip->flags & BTRFS_INODE_DIRSYNC)
 116                 inode->i_flags |= S_DIRSYNC;
 117 }
 118
 119 /*
 120  * Inherit flags from the parent inode.
 121  *
 122  * Currently only the compression flags and the cow flags are inherited.
 123  */
 124 void btrfs_inherit_iflags(struct inode *inode, struct inode *dir)
 125 {
 126         unsigned int flags;
 127
 128         if (!dir)
 129                 return;
 130
 131         flags = BTRFS_I(dir)->flags;
 132
 133         if (flags & BTRFS_INODE_NOCOMPRESS) {
 134                 BTRFS_I(inode)->flags &= ~BTRFS_INODE_COMPRESS;
 135                 BTRFS_I(inode)->flags |= BTRFS_INODE_NOCOMPRESS;
 136         } else if (flags & BTRFS_INODE_COMPRESS) {
 137                 BTRFS_I(inode)->flags &= ~BTRFS_INODE_NOCOMPRESS;
 138                 BTRFS_I(inode)->flags |= BTRFS_INODE_COMPRESS;
 139         }
 140
 141         if (flags & BTRFS_INODE_NODATACOW)
 142                 BTRFS_I(inode)->flags |= BTRFS_INODE_NODATACOW;
 143
 144         btrfs_update_iflags(inode);
 145 }
 146
 147 static int btrfs_ioctl_getflags(struct file *file, void __user *arg)
 148 {
 149         struct btrfs_inode *ip = BTRFS_I(file->f_path.dentry->d_inode);
 150         unsigned int flags = btrfs_flags_to_ioctl(ip->flags);
 151
 152         if (copy_to_user(arg, &flags, sizeof(flags)))
 153                 return -EFAULT;
 154         return 0;
 155 }
 156
 157 static int check_flags(unsigned int flags)
 158 {
 159         if (flags & ~(FS_IMMUTABLE_FL | FS_APPEND_FL | \
 160                       FS_NOATIME_FL | FS_NODUMP_FL | \
 161                       FS_SYNC_FL | FS_DIRSYNC_FL | \
 162                       FS_NOCOMP_FL | FS_COMPR_FL |
 163                       FS_NOCOW_FL))
 164                 return -EOPNOTSUPP;
 165
 166         if ((flags & FS_NOCOMP_FL) && (flags & FS_COMPR_FL))
 167                 return -EINVAL;
 168
 169         return 0;
 170 }
 171
 172 static int btrfs_ioctl_setflags(struct file *file, void __user *arg)
 173 {
 174         struct inode *inode = file->f_path.dentry->d_inode;
 175         struct btrfs_inode *ip = BTRFS_I(inode);
 176         struct btrfs_root *root = ip->root;
 177         struct btrfs_trans_handle *trans;
 178         unsigned int flags, oldflags;
 179         int ret;
 180         u64 ip_oldflags;
 181         unsigned int i_oldflags;
 182
 183         if (btrfs_root_readonly(root))
 184                 return -EROFS;
 185
 186         if (copy_from_user(&flags, arg, sizeof(flags)))
 187                 return -EFAULT;
 188
 189         ret = check_flags(flags);
 190         if (ret)
 191                 return ret;
 192
 193         if (!inode_owner_or_capable(inode))
 194                 return -EACCES;
 195
 196         mutex_lock(&inode->i_mutex);
 197
 198         ip_oldflags = ip->flags;
 199         i_oldflags = inode->i_flags;
 200
 201         flags = btrfs_mask_flags(inode->i_mode, flags);
 202         oldflags = btrfs_flags_to_ioctl(ip->flags);
 203         if ((flags ^ oldflags) & (FS_APPEND_FL | FS_IMMUTABLE_FL)) {
 204                 if (!capable(CAP_LINUX_IMMUTABLE)) {
 205                         ret = -EPERM;
 206                         goto out_unlock;
 207                 }
 208         }
 209
 210         ret = mnt_want_write_file(file);
 211         if (ret)
 212                 goto out_unlock;
 213
 214         if (flags & FS_SYNC_FL)
 215                 ip->flags |= BTRFS_INODE_SYNC;
 216         else
 217                 ip->flags &= ~BTRFS_INODE_SYNC;
 218         if (flags & FS_IMMUTABLE_FL)
 219                 ip->flags |= BTRFS_INODE_IMMUTABLE;
 220         else
 221                 ip->flags &= ~BTRFS_INODE_IMMUTABLE;
 222         if (flags & FS_APPEND_FL)
 223                 ip->flags |= BTRFS_INODE_APPEND;
 224         else
 225                 ip->flags &= ~BTRFS_INODE_APPEND;
 226         if (flags & FS_NODUMP_FL)
 227                 ip->flags |= BTRFS_INODE_NODUMP;
 228         else
 229                 ip->flags &= ~BTRFS_INODE_NODUMP;
 230         if (flags & FS_NOATIME_FL)
 231                 ip->flags |= BTRFS_INODE_NOATIME;
 232         else
 233                 ip->flags &= ~BTRFS_INODE_NOATIME;
 234         if (flags & FS_DIRSYNC_FL)
 235                 ip->flags |= BTRFS_INODE_DIRSYNC;
 236         else
 237                 ip->flags &= ~BTRFS_INODE_DIRSYNC;
 238         if (flags & FS_NOCOW_FL)
 239                 ip->flags |= BTRFS_INODE_NODATACOW;
 240         else
 241                 ip->flags &= ~BTRFS_INODE_NODATACOW;
 242
 243         /*
 244          * The COMPRESS flag can only be changed by users, while the NOCOMPRESS
 245          * flag may be changed automatically if compression code won't make
 246          * things smaller.
 247          */
 248         if (flags & FS_NOCOMP_FL) {
 249                 ip->flags &= ~BTRFS_INODE_COMPRESS;
 250                 ip->flags |= BTRFS_INODE_NOCOMPRESS;
 251         } else if (flags & FS_COMPR_FL) {
 252                 ip->flags |= BTRFS_INODE_COMPRESS;
 253                 ip->flags &= ~BTRFS_INODE_NOCOMPRESS;
 254         } else {
 255                 ip->flags &= ~(BTRFS_INODE_COMPRESS | BTRFS_INODE_NOCOMPRESS);
 256         }
 257
 258         trans = btrfs_start_transaction(root, 1);
 259         if (IS_ERR(trans)) {
 260                 ret = PTR_ERR(trans);
 261                 goto out_drop;
 262         }
 263
 264         btrfs_update_iflags(inode);
 265         inode_inc_iversion(inode);
 266         inode->i_ctime = CURRENT_TIME;
 267         ret = btrfs_update_inode(trans, root, inode);
 268
 269         btrfs_end_transaction(trans, root);
 270  out_drop:
 271         if (ret) {
 272                 ip->flags = ip_oldflags;
 273                 inode->i_flags = i_oldflags;
 274         }
 275
 276         mnt_drop_write_file(file);
 277  out_unlock:
 278         mutex_unlock(&inode->i_mutex);
 279         return ret;
 280 }
 281
 282 static int btrfs_ioctl_getversion(struct file *file, int __user *arg)
 283 {
 284         struct inode *inode = file->f_path.dentry->d_inode;
 285
 286         return put_user(inode->i_generation, arg);
 287 }
 288
 289 static noinline int btrfs_ioctl_fitrim(struct file *file, void __user *arg)
 290 {
 291         struct btrfs_fs_info *fs_info = btrfs_sb(fdentry(file)->d_sb);
 292         struct btrfs_device *device;
 293         struct request_queue *q;
 294         struct fstrim_range range;
 295         u64 minlen = ULLONG_MAX;
 296         u64 num_devices = 0;
 297         u64 total_bytes = btrfs_super_total_bytes(fs_info->super_copy);
 298         int ret;
 299
 300         if (!capable(CAP_SYS_ADMIN))
 301                 return -EPERM;
 302
 303         rcu_read_lock();
 304         list_for_each_entry_rcu(device, &fs_info->fs_devices->devices,
 305                                 dev_list) {
 306                 if (!device->bdev)
 307                         continue;
 308                 q = bdev_get_queue(device->bdev);
 309                 if (blk_queue_discard(q)) {
 310                         num_devices++;
 311                         minlen = min((u64)q->limits.discard_granularity,
 312                                      minlen);
 313                 }
 314         }
 315         rcu_read_unlock();
 316
 317         if (!num_devices)
 318                 return -EOPNOTSUPP;
 319         if (copy_from_user(&range, arg, sizeof(range)))
 320                 return -EFAULT;
 321         if (range.start > total_bytes)
 322                 return -EINVAL;
 323
 324         range.len = min(range.len, total_bytes - range.start);
 325         range.minlen = max(range.minlen, minlen);
 326         ret = btrfs_trim_fs(fs_info->tree_root, &range);
 327         if (ret < 0)
 328                 return ret;
 329
 330         if (copy_to_user(arg, &range, sizeof(range)))
 331                 return -EFAULT;
 332
 333         return 0;
 334 }
 335
 336 static noinline int create_subvol(struct btrfs_root *root,
 337                                   struct dentry *dentry,
 338                                   char *name, int namelen,
 339                                   u64 *async_transid)
 340 {
 341         struct btrfs_trans_handle *trans;
 342         struct btrfs_key key;
 343         struct btrfs_root_item root_item;
 344         struct btrfs_inode_item *inode_item;
 345         struct extent_buffer *leaf;
 346         struct btrfs_root *new_root;
 347         struct dentry *parent = dentry->d_parent;
 348         struct inode *dir;
 349         int ret;
 350         int err;
 351         u64 objectid;
 352         u64 new_dirid = BTRFS_FIRST_FREE_OBJECTID;
 353         u64 index = 0;
 354
 355         ret = btrfs_find_free_objectid(root->fs_info->tree_root, &objectid);
 356         if (ret)
 357                 return ret;
 358
 359         dir = parent->d_inode;
 360
 361         /*
 362          * 1 - inode item
 363          * 2 - refs
 364          * 1 - root item
 365          * 2 - dir items
 366          */
 367         trans = btrfs_start_transaction(root, 6);
 368         if (IS_ERR(trans))
 369                 return PTR_ERR(trans);
 370
 371         leaf = btrfs_alloc_free_block(trans, root, root->leafsize,
 372                                       0, objectid, NULL, 0, 0, 0);
 373         if (IS_ERR(leaf)) {
 374                 ret = PTR_ERR(leaf);
 375                 goto fail;
 376         }
 377
 378         memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header));
 379         btrfs_set_header_bytenr(leaf, leaf->start);
 380         btrfs_set_header_generation(leaf, trans->transid);
 381         btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV);
 382         btrfs_set_header_owner(leaf, objectid);
 383
 384         write_extent_buffer(leaf, root->fs_info->fsid,
 385                             (unsigned long)btrfs_header_fsid(leaf),
 386                             BTRFS_FSID_SIZE);
 387         write_extent_buffer(leaf, root->fs_info->chunk_tree_uuid,
 388                             (unsigned long)btrfs_header_chunk_tree_uuid(leaf),
 389                             BTRFS_UUID_SIZE);
 390         btrfs_mark_buffer_dirty(leaf);
 391
 392         inode_item = &root_item.inode;
 393         memset(inode_item, 0, sizeof(*inode_item));
 394         inode_item->generation = cpu_to_le64(1);
 395         inode_item->size = cpu_to_le64(3);
 396         inode_item->nlink = cpu_to_le32(1);
 397         inode_item->nbytes = cpu_to_le64(root->leafsize);
 398         inode_item->mode = cpu_to_le32(S_IFDIR | 0755);
 399
 400         root_item.flags = 0;
 401         root_item.byte_limit = 0;
 402         inode_item->flags = cpu_to_le64(BTRFS_INODE_ROOT_ITEM_INIT);
 403
 404         btrfs_set_root_bytenr(&root_item, leaf->start);
 405         btrfs_set_root_generation(&root_item, trans->transid);
 406         btrfs_set_root_level(&root_item, 0);
 407         btrfs_set_root_refs(&root_item, 1);
 408         btrfs_set_root_used(&root_item, leaf->len);
 409         btrfs_set_root_last_snapshot(&root_item, 0);
 410
 411         memset(&root_item.drop_progress, 0, sizeof(root_item.drop_progress));
 412         root_item.drop_level = 0;
 413
 414         btrfs_tree_unlock(leaf);
 415         free_extent_buffer(leaf);
 416         leaf = NULL;
 417
 418         btrfs_set_root_dirid(&root_item, new_dirid);
 419
 420         key.objectid = objectid;
 421         key.offset = 0;
 422         btrfs_set_key_type(&key, BTRFS_ROOT_ITEM_KEY);
 423         ret = btrfs_insert_root(trans, root->fs_info->tree_root, &key,
 424                                 &root_item);
 425         if (ret)
 426                 goto fail;
 427
 428         key.offset = (u64)-1;
 429         new_root = btrfs_read_fs_root_no_name(root->fs_info, &key);
 430         if (IS_ERR(new_root)) {
 431                 btrfs_abort_transaction(trans, root, PTR_ERR(new_root));
 432                 ret = PTR_ERR(new_root);
 433                 goto fail;
 434         }
 435
 436         btrfs_record_root_in_trans(trans, new_root);
 437
 438         ret = btrfs_create_subvol_root(trans, new_root, new_dirid);
 439         if (ret) {
 440                 /* We potentially lose an unused inode item here */
 441                 btrfs_abort_transaction(trans, root, ret);
 442                 goto fail;
 443         }
 444
 445         /*
 446          * insert the directory item
 447          */
 448         ret = btrfs_set_inode_index(dir, &index);
 449         if (ret) {
 450                 btrfs_abort_transaction(trans, root, ret);
 451                 goto fail;
 452         }
 453
 454         ret = btrfs_insert_dir_item(trans, root,
 455                                     name, namelen, dir, &key,
 456                                     BTRFS_FT_DIR, index);
 457         if (ret) {
 458                 btrfs_abort_transaction(trans, root, ret);
 459                 goto fail;
 460         }
 461
 462         btrfs_i_size_write(dir, dir->i_size + namelen * 2);
 463         ret = btrfs_update_inode(trans, root, dir);
 464         BUG_ON(ret);
 465
 466         ret = btrfs_add_root_ref(trans, root->fs_info->tree_root,
 467                                  objectid, root->root_key.objectid,
 468                                  btrfs_ino(dir), index, name, namelen);
 469
 470         BUG_ON(ret);
 471
 472         d_instantiate(dentry, btrfs_lookup_dentry(dir, dentry));
 473 fail:
 474         if (async_transid) {
 475                 *async_transid = trans->transid;
 476                 err = btrfs_commit_transaction_async(trans, root, 1);
 477         } else {
 478                 err = btrfs_commit_transaction(trans, root);
 479         }
 480         if (err && !ret)
 481                 ret = err;
 482         return ret;
 483 }
 484
 485 static int create_snapshot(struct btrfs_root *root, struct dentry *dentry,
 486                            char *name, int namelen, u64 *async_transid,
 487                            bool readonly)
 488 {
 489         struct inode *inode;
 490         struct btrfs_pending_snapshot *pending_snapshot;
 491         struct btrfs_trans_handle *trans;
 492         int ret;
 493
 494         if (!root->ref_cows)
 495                 return -EINVAL;
 496
 497         pending_snapshot = kzalloc(sizeof(*pending_snapshot), GFP_NOFS);
 498         if (!pending_snapshot)
 499                 return -ENOMEM;
 500
 501         btrfs_init_block_rsv(&pending_snapshot->block_rsv);
 502         pending_snapshot->dentry = dentry;
 503         pending_snapshot->root = root;
 504         pending_snapshot->readonly = readonly;
 505
 506         trans = btrfs_start_transaction(root->fs_info->extent_root, 5);
 507         if (IS_ERR(trans)) {
 508                 ret = PTR_ERR(trans);
 509                 goto fail;
 510         }
 511
 512         ret = btrfs_snap_reserve_metadata(trans, pending_snapshot);
 513         BUG_ON(ret);
 514
 515         spin_lock(&root->fs_info->trans_lock);
 516         list_add(&pending_snapshot->list,
 517                  &trans->transaction->pending_snapshots);
 518         spin_unlock(&root->fs_info->trans_lock);
 519         if (async_transid) {
 520                 *async_transid = trans->transid;
 521                 ret = btrfs_commit_transaction_async(trans,
 522                                      root->fs_info->extent_root, 1);
 523         } else {
 524                 ret = btrfs_commit_transaction(trans,
 525                                                root->fs_info->extent_root);
 526         }
 527         BUG_ON(ret);
 528
 529         ret = pending_snapshot->error;
 530         if (ret)
 531                 goto fail;
 532
 533         ret = btrfs_orphan_cleanup(pending_snapshot->snap);
 534         if (ret)
 535                 goto fail;
 536
 537         inode = btrfs_lookup_dentry(dentry->d_parent->d_inode, dentry);
 538         if (IS_ERR(inode)) {
 539                 ret = PTR_ERR(inode);
 540                 goto fail;
 541         }
 542         BUG_ON(!inode);
 543         d_instantiate(dentry, inode);
 544         ret = 0;
 545 fail:
 546         kfree(pending_snapshot);
 547         return ret;
 548 }
 549
 550 /*  copy of check_sticky in fs/namei.c()
 551 * It's inline, so penalty for filesystems that don't use sticky bit is
 552 * minimal.
 553 */
 554 static inline int btrfs_check_sticky(struct inode *dir, struct inode *inode)
 555 {
 556         uid_t fsuid = current_fsuid();
 557
 558         if (!(dir->i_mode & S_ISVTX))
 559                 return 0;
 560         if (inode->i_uid == fsuid)
 561                 return 0;
 562         if (dir->i_uid == fsuid)
 563                 return 0;
 564         return !capable(CAP_FOWNER);
 565 }
 566
 567 /*  copy of may_delete in fs/namei.c()
 568  *      Check whether we can remove a link victim from directory dir, check
 569  *  whether the type of victim is right.
 570  *  1. We can't do it if dir is read-only (done in permission())
 571  *  2. We should have write and exec permissions on dir
 572  *  3. We can't remove anything from append-only dir
 573  *  4. We can't do anything with immutable dir (done in permission())
 574  *  5. If the sticky bit on dir is set we should either
 575  *      a. be owner of dir, or
 576  *      b. be owner of victim, or
 577  *      c. have CAP_FOWNER capability
 578  *  6. If the victim is append-only or immutable we can't do antyhing with
 579  *     links pointing to it.
 580  *  7. If we were asked to remove a directory and victim isn't one - ENOTDIR.
 581  *  8. If we were asked to remove a non-directory and victim isn't one - EISDIR.
 582  *  9. We can't remove a root or mountpoint.
 583  * 10. We don't allow removal of NFS sillyrenamed files; it's handled by
 584  *     nfs_async_unlink().
 585  */
 586
 587 static int btrfs_may_delete(struct inode *dir,struct dentry *victim,int isdir)
 588 {
 589         int error;
 590
 591         if (!victim->d_inode)
 592                 return -ENOENT;
 593
 594         BUG_ON(victim->d_parent->d_inode != dir);
 595         audit_inode_child(victim, dir);
 596
 597         error = inode_permission(dir, MAY_WRITE | MAY_EXEC);
 598         if (error)
 599                 return error;
 600         if (IS_APPEND(dir))
 601                 return -EPERM;
 602         if (btrfs_check_sticky(dir, victim->d_inode)||
 603                 IS_APPEND(victim->d_inode)||
 604             IS_IMMUTABLE(victim->d_inode) || IS_SWAPFILE(victim->d_inode))
 605                 return -EPERM;
 606         if (isdir) {
 607                 if (!S_ISDIR(victim->d_inode->i_mode))
 608                         return -ENOTDIR;
 609                 if (IS_ROOT(victim))
 610                         return -EBUSY;
 611         } else if (S_ISDIR(victim->d_inode->i_mode))
 612                 return -EISDIR;
 613         if (IS_DEADDIR(dir))
 614                 return -ENOENT;
 615         if (victim->d_flags & DCACHE_NFSFS_RENAMED)
 616                 return -EBUSY;
 617         return 0;
 618 }
 619
 620 /* copy of may_create in fs/namei.c() */
 621 static inline int btrfs_may_create(struct inode *dir, struct dentry *child)
 622 {
 623         if (child->d_inode)
 624                 return -EEXIST;
 625         if (IS_DEADDIR(dir))
 626                 return -ENOENT;
 627         return inode_permission(dir, MAY_WRITE | MAY_EXEC);
 628 }
 629
 630 /*
 631  * Create a new subvolume below @parent.  This is largely modeled after
 632  * sys_mkdirat and vfs_mkdir, but we only do a single component lookup
 633  * inside this filesystem so it's quite a bit simpler.
 634  */
 635 static noinline int btrfs_mksubvol(struct path *parent,
 636                                    char *name, int namelen,
 637                                    struct btrfs_root *snap_src,
 638                                    u64 *async_transid, bool readonly)
 639 {
 640         struct inode *dir  = parent->dentry->d_inode;
 641         struct dentry *dentry;
 642         int error;
 643
 644         mutex_lock_nested(&dir->i_mutex, I_MUTEX_PARENT);
 645
 646         dentry = lookup_one_len(name, parent->dentry, namelen);
 647         error = PTR_ERR(dentry);
 648         if (IS_ERR(dentry))
 649                 goto out_unlock;
 650
 651         error = -EEXIST;
 652         if (dentry->d_inode)
 653                 goto out_dput;
 654
 655         error = mnt_want_write(parent->mnt);
 656         if (error)
 657                 goto out_dput;
 658
 659         error = btrfs_may_create(dir, dentry);
 660         if (error)
 661                 goto out_drop_write;
 662
 663         down_read(&BTRFS_I(dir)->root->fs_info->subvol_sem);
 664
 665         if (btrfs_root_refs(&BTRFS_I(dir)->root->root_item) == 0)
 666                 goto out_up_read;
 667
 668         if (snap_src) {
 669                 error = create_snapshot(snap_src, dentry,
 670                                         name, namelen, async_transid, readonly);
 671         } else {
 672                 error = create_subvol(BTRFS_I(dir)->root, dentry,
 673                                       name, namelen, async_transid);
 674         }
 675         if (!error)
 676                 fsnotify_mkdir(dir, dentry);
 677 out_up_read:
 678         up_read(&BTRFS_I(dir)->root->fs_info->subvol_sem);
 679 out_drop_write:
 680         mnt_drop_write(parent->mnt);
 681 out_dput:
 682         dput(dentry);
 683 out_unlock:
 684         mutex_unlock(&dir->i_mutex);
 685         return error;
 686 }
 687
 688 /*
 689  * When we're defragging a range, we don't want to kick it off again
 690  * if it is really just waiting for delalloc to send it down.
 691  * If we find a nice big extent or delalloc range for the bytes in the
 692  * file you want to defrag, we return 0 to let you know to skip this
 693  * part of the file
 694  */
 695 static int check_defrag_in_cache(struct inode *inode, u64 offset, int thresh)
 696 {
 697         struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
 698         struct extent_map *em = NULL;
 699         struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
 700         u64 end;
 701
 702         read_lock(&em_tree->lock);
 703         em = lookup_extent_mapping(em_tree, offset, PAGE_CACHE_SIZE);
 704         read_unlock(&em_tree->lock);
 705
 706         if (em) {
 707                 end = extent_map_end(em);
 708                 free_extent_map(em);
 709                 if (end - offset > thresh)
 710                         return 0;
 711         }
 712         /* if we already have a nice delalloc here, just stop */
 713         thresh /= 2;
 714         end = count_range_bits(io_tree, &offset, offset + thresh,
 715                                thresh, EXTENT_DELALLOC, 1);
 716         if (end >= thresh)
 717                 return 0;
 718         return 1;
 719 }
 720
 721 /*
 722  * helper function to walk through a file and find extents
 723  * newer than a specific transid, and smaller than thresh.
 724  *
 725  * This is used by the defragging code to find new and small
 726  * extents
 727  */
 728 static int find_new_extents(struct btrfs_root *root,
 729                             struct inode *inode, u64 newer_than,
 730                             u64 *off, int thresh)
 731 {
 732         struct btrfs_path *path;
 733         struct btrfs_key min_key;
 734         struct btrfs_key max_key;
 735         struct extent_buffer *leaf;
 736         struct btrfs_file_extent_item *extent;
 737         int type;
 738         int ret;
 739         u64 ino = btrfs_ino(inode);
 740
 741         path = btrfs_alloc_path();
 742         if (!path)
 743                 return -ENOMEM;
 744
 745         min_key.objectid = ino;
 746         min_key.type = BTRFS_EXTENT_DATA_KEY;
 747         min_key.offset = *off;
 748
 749         max_key.objectid = ino;
 750         max_key.type = (u8)-1;
 751         max_key.offset = (u64)-1;
 752
 753         path->keep_locks = 1;
 754
 755         while(1) {
 756                 ret = btrfs_search_forward(root, &min_key, &max_key,
 757                                            path, 0, newer_than);
 758                 if (ret != 0)
 759                         goto none;
 760                 if (min_key.objectid != ino)
 761                         goto none;
 762                 if (min_key.type != BTRFS_EXTENT_DATA_KEY)
 763                         goto none;
 764
 765                 leaf = path->nodes[0];
 766                 extent = btrfs_item_ptr(leaf, path->slots[0],
 767                                         struct btrfs_file_extent_item);
 768
 769                 type = btrfs_file_extent_type(leaf, extent);
 770                 if (type == BTRFS_FILE_EXTENT_REG &&
 771                     btrfs_file_extent_num_bytes(leaf, extent) < thresh &&
 772                     check_defrag_in_cache(inode, min_key.offset, thresh)) {
 773                         *off = min_key.offset;
 774                         btrfs_free_path(path);
 775                         return 0;
 776                 }
 777
 778                 if (min_key.offset == (u64)-1)
 779                         goto none;
 780
 781                 min_key.offset++;
 782                 btrfs_release_path(path);
 783         }
 784 none:
 785         btrfs_free_path(path);
 786         return -ENOENT;
 787 }
 788
 789 /*
 790  * Validaty check of prev em and next em:
 791  * 1) no prev/next em
 792  * 2) prev/next em is an hole/inline extent
 793  */
 794 static int check_adjacent_extents(struct inode *inode, struct extent_map *em)
 795 {
 796         struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
 797         struct extent_map *prev = NULL, *next = NULL;
 798         int ret = 0;
 799
 800         read_lock(&em_tree->lock);
 801         prev = lookup_extent_mapping(em_tree, em->start - 1, (u64)-1);
 802         next = lookup_extent_mapping(em_tree, em->start + em->len, (u64)-1);
 803         read_unlock(&em_tree->lock);
 804
 805         if ((!prev || prev->block_start >= EXTENT_MAP_LAST_BYTE) &&
 806             (!next || next->block_start >= EXTENT_MAP_LAST_BYTE))
 807                 ret = 1;
 808         free_extent_map(prev);
 809         free_extent_map(next);
 810
 811         return ret;
 812 }
 813
 814 static int should_defrag_range(struct inode *inode, u64 start, u64 len,
 815                                int thresh, u64 *last_len, u64 *skip,
 816                                u64 *defrag_end)
 817 {
 818         struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
 819         struct extent_map *em = NULL;
 820         struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
 821         int ret = 1;
 822
 823         /*
 824          * make sure that once we start defragging an extent, we keep on
 825          * defragging it
 826          */
 827         if (start < *defrag_end)
 828                 return 1;
 829
 830         *skip = 0;
 831
 832         /*
 833          * hopefully we have this extent in the tree already, try without
 834          * the full extent lock
 835          */
 836         read_lock(&em_tree->lock);
 837         em = lookup_extent_mapping(em_tree, start, len);
 838         read_unlock(&em_tree->lock);
 839
 840         if (!em) {
 841                 /* get the big lock and read metadata off disk */
 842                 lock_extent(io_tree, start, start + len - 1);
 843                 em = btrfs_get_extent(inode, NULL, 0, start, len, 0);
 844                 unlock_extent(io_tree, start, start + len - 1);
 845
 846                 if (IS_ERR(em))
 847                         return 0;
 848         }
 849
 850         /* this will cover holes, and inline extents */
 851         if (em->block_start >= EXTENT_MAP_LAST_BYTE) {
 852                 ret = 0;
 853                 goto out;
 854         }
 855
 856         /* If we have nothing to merge with us, just skip. */
 857         if (check_adjacent_extents(inode, em)) {
 858                 ret = 0;
 859                 goto out;
 860         }
 861
 862         /*
 863          * we hit a real extent, if it is big don't bother defragging it again
 864          */
 865         if ((*last_len == 0 || *last_len >= thresh) && em->len >= thresh)
 866                 ret = 0;
 867
 868 out:
 869         /*
 870          * last_len ends up being a counter of how many bytes we've defragged.
 871          * every time we choose not to defrag an extent, we reset *last_len
 872          * so that the next tiny extent will force a defrag.
 873          *
 874          * The end result of this is that tiny extents before a single big
 875          * extent will force at least part of that big extent to be defragged.
 876          */
 877         if (ret) {
 878                 *defrag_end = extent_map_end(em);
 879         } else {
 880                 *last_len = 0;
 881                 *skip = extent_map_end(em);
 882                 *defrag_end = 0;
 883         }
 884
 885         free_extent_map(em);
 886         return ret;
 887 }
 888
 889 /*
 890  * it doesn't do much good to defrag one or two pages
 891  * at a time.  This pulls in a nice chunk of pages
 892  * to COW and defrag.
 893  *
 894  * It also makes sure the delalloc code has enough
 895  * dirty data to avoid making new small extents as part
 896  * of the defrag
 897  *
 898  * It's a good idea to start RA on this range
 899  * before calling this.
 900  */
 901 static int cluster_pages_for_defrag(struct inode *inode,
 902                                     struct page **pages,
 903                                     unsigned long start_index,
 904                                     int num_pages)
 905 {
 906         unsigned long file_end;
 907         u64 isize = i_size_read(inode);
 908         u64 page_start;
 909         u64 page_end;
 910         u64 page_cnt;
 911         int ret;
 912         int i;
 913         int i_done;
 914         struct btrfs_ordered_extent *ordered;
 915         struct extent_state *cached_state = NULL;
 916         struct extent_io_tree *tree;
 917         gfp_t mask = btrfs_alloc_write_mask(inode->i_mapping);
 918
 919         file_end = (isize - 1) >> PAGE_CACHE_SHIFT;
 920         if (!isize || start_index > file_end)
 921                 return 0;
 922
 923         page_cnt = min_t(u64, (u64)num_pages, (u64)file_end - start_index + 1);
 924
 925         ret = btrfs_delalloc_reserve_space(inode,
 926                                            page_cnt << PAGE_CACHE_SHIFT);
 927         if (ret)
 928                 return ret;
 929         i_done = 0;
 930         tree = &BTRFS_I(inode)->io_tree;
 931
 932         /* step one, lock all the pages */
 933         for (i = 0; i < page_cnt; i++) {
 934                 struct page *page;
 935 again:
 936                 page = find_or_create_page(inode->i_mapping,
 937                                            start_index + i, mask);
 938                 if (!page)
 939                         break;
 940
 941                 page_start = page_offset(page);
 942                 page_end = page_start + PAGE_CACHE_SIZE - 1;
 943                 while (1) {
 944                         lock_extent(tree, page_start, page_end);
 945                         ordered = btrfs_lookup_ordered_extent(inode,
 946                                                               page_start);
 947                         unlock_extent(tree, page_start, page_end);
 948                         if (!ordered)
 949                                 break;
 950
 951                         unlock_page(page);
 952                         btrfs_start_ordered_extent(inode, ordered, 1);
 953                         btrfs_put_ordered_extent(ordered);
 954                         lock_page(page);
 955                         /*
 956                          * we unlocked the page above, so we need check if
 957                          * it was released or not.
 958                          */
 959                         if (page->mapping != inode->i_mapping) {
 960                                 unlock_page(page);
 961                                 page_cache_release(page);
 962                                 goto again;
 963                         }
 964                 }
 965
 966                 if (!PageUptodate(page)) {
 967                         btrfs_readpage(NULL, page);
 968                         lock_page(page);
 969                         if (!PageUptodate(page)) {
 970                                 unlock_page(page);
 971                                 page_cache_release(page);
 972                                 ret = -EIO;
 973                                 break;
 974                         }
 975                 }
 976
 977                 if (page->mapping != inode->i_mapping) {
 978                         unlock_page(page);
 979                         page_cache_release(page);
 980                         goto again;
 981                 }
 982
 983                 pages[i] = page;
 984                 i_done++;
 985         }
 986         if (!i_done || ret)
 987                 goto out;
 988
 989         if (!(inode->i_sb->s_flags & MS_ACTIVE))
 990                 goto out;
 991
 992         /*
 993          * so now we have a nice long stream of locked
 994          * and up to date pages, lets wait on them
 995          */
 996         for (i = 0; i < i_done; i++)
 997                 wait_on_page_writeback(pages[i]);
 998
 999         page_start = page_offset(pages[0]);
1000         page_end = page_offset(pages[i_done - 1]) + PAGE_CACHE_SIZE;
1001
1002         lock_extent_bits(&BTRFS_I(inode)->io_tree,
1003                          page_start, page_end - 1, 0, &cached_state);
1004         clear_extent_bit(&BTRFS_I(inode)->io_tree, page_start,
1005                           page_end - 1, EXTENT_DIRTY | EXTENT_DELALLOC |
1006                           EXTENT_DO_ACCOUNTING, 0, 0, &cached_state,
1007                           GFP_NOFS);
1008
1009         if (i_done != page_cnt) {
1010                 spin_lock(&BTRFS_I(inode)->lock);
1011                 BTRFS_I(inode)->outstanding_extents++;
1012                 spin_unlock(&BTRFS_I(inode)->lock);
1013                 btrfs_delalloc_release_space(inode,
1014                                      (page_cnt - i_done) << PAGE_CACHE_SHIFT);
1015         }
1016
1017
1018         btrfs_set_extent_delalloc(inode, page_start, page_end - 1,
1019                                   &cached_state);
1020
1021         unlock_extent_cached(&BTRFS_I(inode)->io_tree,
1022                              page_start, page_end - 1, &cached_state,
1023                              GFP_NOFS);
1024
1025         for (i = 0; i < i_done; i++) {
1026                 clear_page_dirty_for_io(pages[i]);
1027                 ClearPageChecked(pages[i]);
1028                 set_page_extent_mapped(pages[i]);
1029                 set_page_dirty(pages[i]);
1030                 unlock_page(pages[i]);
1031                 page_cache_release(pages[i]);
1032         }
1033         return i_done;
1034 out:
1035         for (i = 0; i < i_done; i++) {
1036                 unlock_page(pages[i]);
1037                 page_cache_release(pages[i]);
1038         }
1039         btrfs_delalloc_release_space(inode, page_cnt << PAGE_CACHE_SHIFT);
1040         return ret;
1041
1042 }
1043
1044 int btrfs_defrag_file(struct inode *inode, struct file *file,
1045                       struct btrfs_ioctl_defrag_range_args *range,
1046                       u64 newer_than, unsigned long max_to_defrag)
1047 {
1048         struct btrfs_root *root = BTRFS_I(inode)->root;
1049         struct btrfs_super_block *disk_super;
1050         struct file_ra_state *ra = NULL;
1051         unsigned long last_index;
1052         u64 isize = i_size_read(inode);
1053         u64 features;
1054         u64 last_len = 0;
1055         u64 skip = 0;
1056         u64 defrag_end = 0;
1057         u64 newer_off = range->start;
1058         unsigned long i;
1059         unsigned long ra_index = 0;
1060         int ret;
1061         int defrag_count = 0;
1062         int compress_type = BTRFS_COMPRESS_ZLIB;
1063         int extent_thresh = range->extent_thresh;
1064         int max_cluster = (256 * 1024) >> PAGE_CACHE_SHIFT;
1065         int cluster = max_cluster;
1066         u64 new_align = ~((u64)128 * 1024 - 1);
1067         struct page **pages = NULL;
1068
1069         if (extent_thresh == 0)
1070                 extent_thresh = 256 * 1024;
1071
1072         if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS) {
1073                 if (range->compress_type > BTRFS_COMPRESS_TYPES)
1074                         return -EINVAL;
1075                 if (range->compress_type)
1076                         compress_type = range->compress_type;
1077         }
1078
1079         if (isize == 0)
1080                 return 0;
1081
1082         /*
1083          * if we were not given a file, allocate a readahead
1084          * context
1085          */
1086         if (!file) {
1087                 ra = kzalloc(sizeof(*ra), GFP_NOFS);
1088                 if (!ra)
1089                         return -ENOMEM;
1090                 file_ra_state_init(ra, inode->i_mapping);
1091         } else {
1092                 ra = &file->f_ra;
1093         }
1094
1095         pages = kmalloc(sizeof(struct page *) * max_cluster,
1096                         GFP_NOFS);
1097         if (!pages) {
1098                 ret = -ENOMEM;
1099                 goto out_ra;
1100         }
1101
1102         /* find the last page to defrag */
1103         if (range->start + range->len > range->start) {
1104                 last_index = min_t(u64, isize - 1,
1105                          range->start + range->len - 1) >> PAGE_CACHE_SHIFT;
1106         } else {
1107                 last_index = (isize - 1) >> PAGE_CACHE_SHIFT;
1108         }
1109
1110         if (newer_than) {
1111                 ret = find_new_extents(root, inode, newer_than,
1112                                        &newer_off, 64 * 1024);
1113                 if (!ret) {
1114                         range->start = newer_off;
1115                         /*
1116                          * we always align our defrag to help keep
1117                          * the extents in the file evenly spaced
1118                          */
1119                         i = (newer_off & new_align) >> PAGE_CACHE_SHIFT;
1120                 } else
1121                         goto out_ra;
1122         } else {
1123                 i = range->start >> PAGE_CACHE_SHIFT;
1124         }
1125         if (!max_to_defrag)
1126                 max_to_defrag = last_index + 1;
1127
1128         /*
1129          * make writeback starts from i, so the defrag range can be
1130          * written sequentially.
1131          */
1132         if (i < inode->i_mapping->writeback_index)
1133                 inode->i_mapping->writeback_index = i;
1134
1135         while (i <= last_index && defrag_count < max_to_defrag &&
1136                (i < (i_size_read(inode) + PAGE_CACHE_SIZE - 1) >>
1137                 PAGE_CACHE_SHIFT)) {
1138                 /*
1139                  * make sure we stop running if someone unmounts
1140                  * the FS
1141                  */
1142                 if (!(inode->i_sb->s_flags & MS_ACTIVE))
1143                         break;
1144
1145                 if (!should_defrag_range(inode, (u64)i << PAGE_CACHE_SHIFT,
1146                                          PAGE_CACHE_SIZE, extent_thresh,
1147                                          &last_len, &skip, &defrag_end)) {
1148                         unsigned long next;
1149                         /*
1150                          * the should_defrag function tells us how much to skip
1151                          * bump our counter by the suggested amount
1152                          */
1153                         next = (skip + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT;
1154                         i = max(i + 1, next);
1155                         continue;
1156                 }
1157
1158                 if (!newer_than) {
1159                         cluster = (PAGE_CACHE_ALIGN(defrag_end) >>
1160                                    PAGE_CACHE_SHIFT) - i;
1161                         cluster = min(cluster, max_cluster);
1162                 } else {
1163                         cluster = max_cluster;
1164                 }
1165
1166                 if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)
1167                         BTRFS_I(inode)->force_compress = compress_type;
1168
1169                 if (i + cluster > ra_index) {
1170                         ra_index = max(i, ra_index);
1171                         btrfs_force_ra(inode->i_mapping, ra, file, ra_index,
1172                                        cluster);
1173                         ra_index += max_cluster;
1174                 }
1175
1176                 mutex_lock(&inode->i_mutex);
1177                 ret = cluster_pages_for_defrag(inode, pages, i, cluster);
1178                 if (ret < 0) {
1179                         mutex_unlock(&inode->i_mutex);
1180                         goto out_ra;
1181                 }
1182
1183                 defrag_count += ret;
1184                 balance_dirty_pages_ratelimited_nr(inode->i_mapping, ret);
1185                 mutex_unlock(&inode->i_mutex);
1186
1187                 if (newer_than) {
1188                         if (newer_off == (u64)-1)
1189                                 break;
1190
1191                         if (ret > 0)
1192                                 i += ret;
1193
1194                         newer_off = max(newer_off + 1,
1195                                         (u64)i << PAGE_CACHE_SHIFT);
1196
1197                         ret = find_new_extents(root, inode,
1198                                                newer_than, &newer_off,
1199                                                64 * 1024);
1200                         if (!ret) {
1201                                 range->start = newer_off;
1202                                 i = (newer_off & new_align) >> PAGE_CACHE_SHIFT;
1203                         } else {
1204                                 break;
1205                         }
1206                 } else {
1207                         if (ret > 0) {
1208                                 i += ret;
1209                                 last_len += ret << PAGE_CACHE_SHIFT;
1210                         } else {
1211                                 i++;
1212                                 last_len = 0;
1213                         }
1214                 }
1215         }
1216
1217         if ((range->flags & BTRFS_DEFRAG_RANGE_START_IO))
1218                 filemap_flush(inode->i_mapping);
1219
1220         if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
1221                 /* the filemap_flush will queue IO into the worker threads, but
1222                  * we have to make sure the IO is actually started and that
1223                  * ordered extents get created before we return
1224                  */
1225                 atomic_inc(&root->fs_info->async_submit_draining);
1226                 while (atomic_read(&root->fs_info->nr_async_submits) ||
1227                       atomic_read(&root->fs_info->async_delalloc_pages)) {
1228                         wait_event(root->fs_info->async_submit_wait,
1229                            (atomic_read(&root->fs_info->nr_async_submits) == 0 &&
1230                             atomic_read(&root->fs_info->async_delalloc_pages) == 0));
1231                 }
1232                 atomic_dec(&root->fs_info->async_submit_draining);
1233
1234                 mutex_lock(&inode->i_mutex);
1235                 BTRFS_I(inode)->force_compress = BTRFS_COMPRESS_NONE;
1236                 mutex_unlock(&inode->i_mutex);
1237         }
1238
1239         disk_super = root->fs_info->super_copy;
1240         features = btrfs_super_incompat_flags(disk_super);
1241         if (range->compress_type == BTRFS_COMPRESS_LZO) {
1242                 features |= BTRFS_FEATURE_INCOMPAT_COMPRESS_LZO;
1243                 btrfs_set_super_incompat_flags(disk_super, features);
1244         }
1245
1246         ret = defrag_count;
1247
1248 out_ra:
1249         if (!file)
1250                 kfree(ra);
1251         kfree(pages);
1252         return ret;
1253 }
1254
1255 static noinline int btrfs_ioctl_resize(struct btrfs_root *root,
1256                                         void __user *arg)
1257 {
1258         u64 new_size;
1259         u64 old_size;
1260         u64 devid = 1;
1261         struct btrfs_ioctl_vol_args *vol_args;
1262         struct btrfs_trans_handle *trans;
1263         struct btrfs_device *device = NULL;
1264         char *sizestr;
1265         char *devstr = NULL;
1266         int ret = 0;
1267         int mod = 0;
1268
1269         if (root->fs_info->sb->s_flags & MS_RDONLY)
1270                 return -EROFS;
1271
1272         if (!capable(CAP_SYS_ADMIN))
1273                 return -EPERM;
1274
1275         mutex_lock(&root->fs_info->volume_mutex);
1276         if (root->fs_info->balance_ctl) {
1277                 printk(KERN_INFO "btrfs: balance in progress\n");
1278                 ret = -EINVAL;
1279                 goto out;
1280         }
1281
1282         vol_args = memdup_user(arg, sizeof(*vol_args));
1283         if (IS_ERR(vol_args)) {
1284                 ret = PTR_ERR(vol_args);
1285                 goto out;
1286         }
1287
1288         vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1289
1290         sizestr = vol_args->name;
1291         devstr = strchr(sizestr, ':');
1292         if (devstr) {
1293                 char *end;
1294                 sizestr = devstr + 1;
1295                 *devstr = '\0';
1296                 devstr = vol_args->name;
1297                 devid = simple_strtoull(devstr, &end, 10);
1298                 printk(KERN_INFO "btrfs: resizing devid %llu\n",
1299                        (unsigned long long)devid);
1300         }
1301         device = btrfs_find_device(root, devid, NULL, NULL);
1302         if (!device) {
1303                 printk(KERN_INFO "btrfs: resizer unable to find device %llu\n",
1304                        (unsigned long long)devid);
1305                 ret = -EINVAL;
1306                 goto out_free;
1307         }
1308         if (!strcmp(sizestr, "max"))
1309                 new_size = device->bdev->bd_inode->i_size;
1310         else {
1311                 if (sizestr[0] == '-') {
1312                         mod = -1;
1313                         sizestr++;
1314                 } else if (sizestr[0] == '+') {
1315                         mod = 1;
1316                         sizestr++;
1317                 }
1318                 new_size = memparse(sizestr, NULL);
1319                 if (new_size == 0) {
1320                         ret = -EINVAL;
1321                         goto out_free;
1322                 }
1323         }
1324
1325         old_size = device->total_bytes;
1326
1327         if (mod < 0) {
1328                 if (new_size > old_size) {
1329                         ret = -EINVAL;
1330                         goto out_free;
1331                 }
1332                 new_size = old_size - new_size;
1333         } else if (mod > 0) {
1334                 new_size = old_size + new_size;
1335         }
1336
1337         if (new_size < 256 * 1024 * 1024) {
1338                 ret = -EINVAL;
1339                 goto out_free;
1340         }
1341         if (new_size > device->bdev->bd_inode->i_size) {
1342                 ret = -EFBIG;
1343                 goto out_free;
1344         }
1345
1346         do_div(new_size, root->sectorsize);
1347         new_size *= root->sectorsize;
1348
1349         printk_in_rcu(KERN_INFO "btrfs: new size for %s is %llu\n",
1350                       rcu_str_deref(device->name),
1351                       (unsigned long long)new_size);
1352
1353         if (new_size > old_size) {
1354                 trans = btrfs_start_transaction(root, 0);
1355                 if (IS_ERR(trans)) {
1356                         ret = PTR_ERR(trans);
1357                         goto out_free;
1358                 }
1359                 ret = btrfs_grow_device(trans, device, new_size);
1360                 btrfs_commit_transaction(trans, root);
1361         } else if (new_size < old_size) {
1362                 ret = btrfs_shrink_device(device, new_size);
1363         }
1364
1365 out_free:
1366         kfree(vol_args);
1367 out:
1368         mutex_unlock(&root->fs_info->volume_mutex);
1369         return ret;
1370 }
1371
1372 static noinline int btrfs_ioctl_snap_create_transid(struct file *file,
1373                                                     char *name,
1374                                                     unsigned long fd,
1375                                                     int subvol,
1376                                                     u64 *transid,
1377                                                     bool readonly)
1378 {
1379         struct btrfs_root *root = BTRFS_I(fdentry(file)->d_inode)->root;
1380         struct file *src_file;
1381         int namelen;
1382         int ret = 0;
1383
1384         if (root->fs_info->sb->s_flags & MS_RDONLY)
1385                 return -EROFS;
1386
1387         namelen = strlen(name);
1388         if (strchr(name, '/')) {
1389                 ret = -EINVAL;
1390                 goto out;
1391         }
1392
1393         if (name[0] == '.' &&
1394            (namelen == 1 || (name[1] == '.' && namelen == 2))) {
1395                 ret = -EEXIST;
1396                 goto out;
1397         }
1398
1399         if (subvol) {
1400                 ret = btrfs_mksubvol(&file->f_path, name, namelen,
1401                                      NULL, transid, readonly);
1402         } else {
1403                 struct inode *src_inode;
1404                 src_file = fget(fd);
1405                 if (!src_file) {
1406                         ret = -EINVAL;
1407                         goto out;
1408                 }
1409
1410                 src_inode = src_file->f_path.dentry->d_inode;
1411                 if (src_inode->i_sb != file->f_path.dentry->d_inode->i_sb) {
1412                         printk(KERN_INFO "btrfs: Snapshot src from "
1413                                "another FS\n");
1414                         ret = -EINVAL;
1415                         fput(src_file);
1416                         goto out;
1417                 }
1418                 ret = btrfs_mksubvol(&file->f_path, name, namelen,
1419                                      BTRFS_I(src_inode)->root,
1420                                      transid, readonly);
1421                 fput(src_file);
1422         }
1423 out:
1424         return ret;
1425 }
1426
1427 static noinline int btrfs_ioctl_snap_create(struct file *file,
1428                                             void __user *arg, int subvol)
1429 {
1430         struct btrfs_ioctl_vol_args *vol_args;
1431         int ret;
1432
1433         vol_args = memdup_user(arg, sizeof(*vol_args));
1434         if (IS_ERR(vol_args))
1435                 return PTR_ERR(vol_args);
1436         vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1437
1438         ret = btrfs_ioctl_snap_create_transid(file, vol_args->name,
1439                                               vol_args->fd, subvol,
1440                                               NULL, false);
1441
1442         kfree(vol_args);
1443         return ret;
1444 }
1445
1446 static noinline int btrfs_ioctl_snap_create_v2(struct file *file,
1447                                                void __user *arg, int subvol)
1448 {
1449         struct btrfs_ioctl_vol_args_v2 *vol_args;
1450         int ret;
1451         u64 transid = 0;
1452         u64 *ptr = NULL;
1453         bool readonly = false;
1454
1455         vol_args = memdup_user(arg, sizeof(*vol_args));
1456         if (IS_ERR(vol_args))
1457                 return PTR_ERR(vol_args);
1458         vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0';
1459
1460         if (vol_args->flags &
1461             ~(BTRFS_SUBVOL_CREATE_ASYNC | BTRFS_SUBVOL_RDONLY)) {
1462                 ret = -EOPNOTSUPP;
1463                 goto out;
1464         }
1465
1466         if (vol_args->flags & BTRFS_SUBVOL_CREATE_ASYNC)
1467                 ptr = &transid;
1468         if (vol_args->flags & BTRFS_SUBVOL_RDONLY)
1469                 readonly = true;
1470
1471         ret = btrfs_ioctl_snap_create_transid(file, vol_args->name,
1472                                               vol_args->fd, subvol,
1473                                               ptr, readonly);
1474
1475         if (ret == 0 && ptr &&
1476             copy_to_user(arg +
1477                          offsetof(struct btrfs_ioctl_vol_args_v2,
1478                                   transid), ptr, sizeof(*ptr)))
1479                 ret = -EFAULT;
1480 out:
1481         kfree(vol_args);
1482         return ret;
1483 }
1484
1485 static noinline int btrfs_ioctl_subvol_getflags(struct file *file,
1486                                                 void __user *arg)
1487 {
1488         struct inode *inode = fdentry(file)->d_inode;
1489         struct btrfs_root *root = BTRFS_I(inode)->root;
1490         int ret = 0;
1491         u64 flags = 0;
1492
1493         if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID)
1494                 return -EINVAL;
1495
1496         down_read(&root->fs_info->subvol_sem);
1497         if (btrfs_root_readonly(root))
1498                 flags |= BTRFS_SUBVOL_RDONLY;
1499         up_read(&root->fs_info->subvol_sem);
1500
1501         if (copy_to_user(arg, &flags, sizeof(flags)))
1502                 ret = -EFAULT;
1503
1504         return ret;
1505 }
1506
1507 static noinline int btrfs_ioctl_subvol_setflags(struct file *file,
1508                                               void __user *arg)
1509 {
1510         struct inode *inode = fdentry(file)->d_inode;
1511         struct btrfs_root *root = BTRFS_I(inode)->root;
1512         struct btrfs_trans_handle *trans;
1513         u64 root_flags;
1514         u64 flags;
1515         int ret = 0;
1516
1517         if (root->fs_info->sb->s_flags & MS_RDONLY)
1518                 return -EROFS;
1519
1520         if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID)
1521                 return -EINVAL;
1522
1523         if (copy_from_user(&flags, arg, sizeof(flags)))
1524                 return -EFAULT;
1525
1526         if (flags & BTRFS_SUBVOL_CREATE_ASYNC)
1527                 return -EINVAL;
1528
1529         if (flags & ~BTRFS_SUBVOL_RDONLY)
1530                 return -EOPNOTSUPP;
1531
1532         if (!inode_owner_or_capable(inode))
1533                 return -EACCES;
1534
1535         down_write(&root->fs_info->subvol_sem);
1536
1537         /* nothing to do */
1538         if (!!(flags & BTRFS_SUBVOL_RDONLY) == btrfs_root_readonly(root))
1539                 goto out;
1540
1541         root_flags = btrfs_root_flags(&root->root_item);
1542         if (flags & BTRFS_SUBVOL_RDONLY)
1543                 btrfs_set_root_flags(&root->root_item,
1544                                      root_flags | BTRFS_ROOT_SUBVOL_RDONLY);
1545         else
1546                 btrfs_set_root_flags(&root->root_item,
1547                                      root_flags & ~BTRFS_ROOT_SUBVOL_RDONLY);
1548
1549         trans = btrfs_start_transaction(root, 1);
1550         if (IS_ERR(trans)) {
1551                 ret = PTR_ERR(trans);
1552                 goto out_reset;
1553         }
1554
1555         ret = btrfs_update_root(trans, root->fs_info->tree_root,
1556                                 &root->root_key, &root->root_item);
1557
1558         btrfs_commit_transaction(trans, root);
1559 out_reset:
1560         if (ret)
1561                 btrfs_set_root_flags(&root->root_item, root_flags);
1562 out:
1563         up_write(&root->fs_info->subvol_sem);
1564         return ret;
1565 }
1566
1567 /*
1568  * helper to check if the subvolume references other subvolumes
1569  */
1570 static noinline int may_destroy_subvol(struct btrfs_root *root)
1571 {
1572         struct btrfs_path *path;
1573         struct btrfs_key key;
1574         int ret;
1575
1576         path = btrfs_alloc_path();
1577         if (!path)
1578                 return -ENOMEM;
1579
1580         key.objectid = root->root_key.objectid;
1581         key.type = BTRFS_ROOT_REF_KEY;
1582         key.offset = (u64)-1;
1583
1584         ret = btrfs_search_slot(NULL, root->fs_info->tree_root,
1585                                 &key, path, 0, 0);
1586         if (ret < 0)
1587                 goto out;
1588         BUG_ON(ret == 0);
1589
1590         ret = 0;
1591         if (path->slots[0] > 0) {
1592                 path->slots[0]--;
1593                 btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
1594                 if (key.objectid == root->root_key.objectid &&
1595                     key.type == BTRFS_ROOT_REF_KEY)
1596                         ret = -ENOTEMPTY;
1597         }
1598 out:
1599         btrfs_free_path(path);
1600         return ret;
1601 }
1602
1603 static noinline int key_in_sk(struct btrfs_key *key,
1604                               struct btrfs_ioctl_search_key *sk)
1605 {
1606         struct btrfs_key test;
1607         int ret;
1608
1609         test.objectid = sk->min_objectid;
1610         test.type = sk->min_type;
1611         test.offset = sk->min_offset;
1612
1613         ret = btrfs_comp_cpu_keys(key, &test);
1614         if (ret < 0)
1615                 return 0;
1616
1617         test.objectid = sk->max_objectid;
1618         test.type = sk->max_type;
1619         test.offset = sk->max_offset;
1620
1621         ret = btrfs_comp_cpu_keys(key, &test);
1622         if (ret > 0)
1623                 return 0;
1624         return 1;
1625 }
1626
1627 static noinline int copy_to_sk(struct btrfs_root *root,
1628                                struct btrfs_path *path,
1629                                struct btrfs_key *key,
1630                                struct btrfs_ioctl_search_key *sk,
1631                                char *buf,
1632                                unsigned long *sk_offset,
1633                                int *num_found)
1634 {
1635         u64 found_transid;
1636         struct extent_buffer *leaf;
1637         struct btrfs_ioctl_search_header sh;
1638         unsigned long item_off;
1639         unsigned long item_len;
1640         int nritems;
1641         int i;
1642         int slot;
1643         int ret = 0;
1644
1645         leaf = path->nodes[0];
1646         slot = path->slots[0];
1647         nritems = btrfs_header_nritems(leaf);
1648
1649         if (btrfs_header_generation(leaf) > sk->max_transid) {
1650                 i = nritems;
1651                 goto advance_key;
1652         }
1653         found_transid = btrfs_header_generation(leaf);
1654
1655         for (i = slot; i < nritems; i++) {
1656                 item_off = btrfs_item_ptr_offset(leaf, i);
1657                 item_len = btrfs_item_size_nr(leaf, i);
1658
1659                 if (item_len > BTRFS_SEARCH_ARGS_BUFSIZE)
1660                         item_len = 0;
1661
1662                 if (sizeof(sh) + item_len + *sk_offset >
1663                     BTRFS_SEARCH_ARGS_BUFSIZE) {
1664                         ret = 1;
1665                         goto overflow;
1666                 }
1667
1668                 btrfs_item_key_to_cpu(leaf, key, i);
1669                 if (!key_in_sk(key, sk))
1670                         continue;
1671
1672                 sh.objectid = key->objectid;
1673                 sh.offset = key->offset;
1674                 sh.type = key->type;
1675                 sh.len = item_len;
1676                 sh.transid = found_transid;
1677
1678                 /* copy search result header */
1679                 memcpy(buf + *sk_offset, &sh, sizeof(sh));
1680                 *sk_offset += sizeof(sh);
1681
1682                 if (item_len) {
1683                         char *p = buf + *sk_offset;
1684                         /* copy the item */
1685                         read_extent_buffer(leaf, p,
1686                                            item_off, item_len);
1687                         *sk_offset += item_len;
1688                 }
1689                 (*num_found)++;
1690
1691                 if (*num_found >= sk->nr_items)
1692                         break;
1693         }
1694 advance_key:
1695         ret = 0;
1696         if (key->offset < (u64)-1 && key->offset < sk->max_offset)
1697                 key->offset++;
1698         else if (key->type < (u8)-1 && key->type < sk->max_type) {
1699                 key->offset = 0;
1700                 key->type++;
1701         } else if (key->objectid < (u64)-1 && key->objectid < sk->max_objectid) {
1702                 key->offset = 0;
1703                 key->type = 0;
1704                 key->objectid++;
1705         } else
1706                 ret = 1;
1707 overflow:
1708         return ret;
1709 }
1710
1711 static noinline int search_ioctl(struct inode *inode,
1712                                  struct btrfs_ioctl_search_args *args)
1713 {
1714         struct btrfs_root *root;
1715         struct btrfs_key key;
1716         struct btrfs_key max_key;
1717         struct btrfs_path *path;
1718         struct btrfs_ioctl_search_key *sk = &args->key;
1719         struct btrfs_fs_info *info = BTRFS_I(inode)->root->fs_info;
1720         int ret;
1721         int num_found = 0;
1722         unsigned long sk_offset = 0;
1723
1724         path = btrfs_alloc_path();
1725         if (!path)
1726                 return -ENOMEM;
1727
1728         if (sk->tree_id == 0) {
1729                 /* search the root of the inode that was passed */
1730                 root = BTRFS_I(inode)->root;
1731         } else {
1732                 key.objectid = sk->tree_id;
1733                 key.type = BTRFS_ROOT_ITEM_KEY;
1734                 key.offset = (u64)-1;
1735                 root = btrfs_read_fs_root_no_name(info, &key);
1736                 if (IS_ERR(root)) {
1737                         printk(KERN_ERR "could not find root %llu\n",
1738                                sk->tree_id);
1739                         btrfs_free_path(path);
1740                         return -ENOENT;
1741                 }
1742         }
1743
1744         key.objectid = sk->min_objectid;
1745         key.type = sk->min_type;
1746         key.offset = sk->min_offset;
1747
1748         max_key.objectid = sk->max_objectid;
1749         max_key.type = sk->max_type;
1750         max_key.offset = sk->max_offset;
1751
1752         path->keep_locks = 1;
1753
1754         while(1) {
1755                 ret = btrfs_search_forward(root, &key, &max_key, path, 0,
1756                                            sk->min_transid);
1757                 if (ret != 0) {
1758                         if (ret > 0)
1759                                 ret = 0;
1760                         goto err;
1761                 }
1762                 ret = copy_to_sk(root, path, &key, sk, args->buf,
1763                                  &sk_offset, &num_found);
1764                 btrfs_release_path(path);
1765                 if (ret || num_found >= sk->nr_items)
1766                         break;
1767
1768         }
1769         ret = 0;
1770 err:
1771         sk->nr_items = num_found;
1772         btrfs_free_path(path);
1773         return ret;
1774 }
1775
1776 static noinline int btrfs_ioctl_tree_search(struct file *file,
1777                                            void __user *argp)
1778 {
1779          struct btrfs_ioctl_search_args *args;
1780          struct inode *inode;
1781          int ret;
1782
1783         if (!capable(CAP_SYS_ADMIN))
1784                 return -EPERM;
1785
1786         args = memdup_user(argp, sizeof(*args));
1787         if (IS_ERR(args))
1788                 return PTR_ERR(args);
1789
1790         inode = fdentry(file)->d_inode;
1791         ret = search_ioctl(inode, args);
1792         if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
1793                 ret = -EFAULT;
1794         kfree(args);
1795         return ret;
1796 }
1797
1798 /*
1799  * Search INODE_REFs to identify path name of 'dirid' directory
1800  * in a 'tree_id' tree. and sets path name to 'name'.
1801  */
1802 static noinline int btrfs_search_path_in_tree(struct btrfs_fs_info *info,
1803                                 u64 tree_id, u64 dirid, char *name)
1804 {
1805         struct btrfs_root *root;
1806         struct btrfs_key key;
1807         char *ptr;
1808         int ret = -1;
1809         int slot;
1810         int len;
1811         int total_len = 0;
1812         struct btrfs_inode_ref *iref;
1813         struct extent_buffer *l;
1814         struct btrfs_path *path;
1815
1816         if (dirid == BTRFS_FIRST_FREE_OBJECTID) {
1817                 name[0]='\0';
1818                 return 0;
1819         }
1820
1821         path = btrfs_alloc_path();
1822         if (!path)
1823                 return -ENOMEM;
1824
1825         ptr = &name[BTRFS_INO_LOOKUP_PATH_MAX];
1826
1827         key.objectid = tree_id;
1828         key.type = BTRFS_ROOT_ITEM_KEY;
1829         key.offset = (u64)-1;
1830         root = btrfs_read_fs_root_no_name(info, &key);
1831         if (IS_ERR(root)) {
1832                 printk(KERN_ERR "could not find root %llu\n", tree_id);
1833                 ret = -ENOENT;
1834                 goto out;
1835         }
1836
1837         key.objectid = dirid;
1838         key.type = BTRFS_INODE_REF_KEY;
1839         key.offset = (u64)-1;
1840
1841         while(1) {
1842                 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
1843                 if (ret < 0)
1844                         goto out;
1845
1846                 l = path->nodes[0];
1847                 slot = path->slots[0];
1848                 if (ret > 0 && slot > 0)
1849                         slot--;
1850                 btrfs_item_key_to_cpu(l, &key, slot);
1851
1852                 if (ret > 0 && (key.objectid != dirid ||
1853                                 key.type != BTRFS_INODE_REF_KEY)) {
1854                         ret = -ENOENT;
1855                         goto out;
1856                 }
1857
1858                 iref = btrfs_item_ptr(l, slot, struct btrfs_inode_ref);
1859                 len = btrfs_inode_ref_name_len(l, iref);
1860                 ptr -= len + 1;
1861                 total_len += len + 1;
1862                 if (ptr < name)
1863                         goto out;
1864
1865                 *(ptr + len) = '/';
1866                 read_extent_buffer(l, ptr,(unsigned long)(iref + 1), len);
1867
1868                 if (key.offset == BTRFS_FIRST_FREE_OBJECTID)
1869                         break;
1870
1871                 btrfs_release_path(path);
1872                 key.objectid = key.offset;
1873                 key.offset = (u64)-1;
1874                 dirid = key.objectid;
1875         }
1876         if (ptr < name)
1877                 goto out;
1878         memmove(name, ptr, total_len);
1879         name[total_len]='\0';
1880         ret = 0;
1881 out:
1882         btrfs_free_path(path);
1883         return ret;
1884 }
1885
1886 static noinline int btrfs_ioctl_ino_lookup(struct file *file,
1887                                            void __user *argp)
1888 {
1889          struct btrfs_ioctl_ino_lookup_args *args;
1890          struct inode *inode;
1891          int ret;
1892
1893         if (!capable(CAP_SYS_ADMIN))
1894                 return -EPERM;
1895
1896         args = memdup_user(argp, sizeof(*args));
1897         if (IS_ERR(args))
1898                 return PTR_ERR(args);
1899
1900         inode = fdentry(file)->d_inode;
1901
1902         if (args->treeid == 0)
1903                 args->treeid = BTRFS_I(inode)->root->root_key.objectid;
1904
1905         ret = btrfs_search_path_in_tree(BTRFS_I(inode)->root->fs_info,
1906                                         args->treeid, args->objectid,
1907                                         args->name);
1908
1909         if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
1910                 ret = -EFAULT;
1911
1912         kfree(args);
1913         return ret;
1914 }
1915
1916 static noinline int btrfs_ioctl_snap_destroy(struct file *file,
1917                                              void __user *arg)
1918 {
1919         struct dentry *parent = fdentry(file);
1920         struct dentry *dentry;
1921         struct inode *dir = parent->d_inode;
1922         struct inode *inode;
1923         struct btrfs_root *root = BTRFS_I(dir)->root;
1924         struct btrfs_root *dest = NULL;
1925         struct btrfs_ioctl_vol_args *vol_args;
1926         struct btrfs_trans_handle *trans;
1927         int namelen;
1928         int ret;
1929         int err = 0;
1930
1931         vol_args = memdup_user(arg, sizeof(*vol_args));
1932         if (IS_ERR(vol_args))
1933                 return PTR_ERR(vol_args);
1934
1935         vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1936         namelen = strlen(vol_args->name);
1937         if (strchr(vol_args->name, '/') ||
1938             strncmp(vol_args->name, "..", namelen) == 0) {
1939                 err = -EINVAL;
1940                 goto out;
1941         }
1942
1943         err = mnt_want_write_file(file);
1944         if (err)
1945                 goto out;
1946
1947         mutex_lock_nested(&dir->i_mutex, I_MUTEX_PARENT);
1948         dentry = lookup_one_len(vol_args->name, parent, namelen);
1949         if (IS_ERR(dentry)) {
1950                 err = PTR_ERR(dentry);
1951                 goto out_unlock_dir;
1952         }
1953
1954         if (!dentry->d_inode) {
1955                 err = -ENOENT;
1956                 goto out_dput;
1957         }
1958
1959         inode = dentry->d_inode;
1960         dest = BTRFS_I(inode)->root;
1961         if (!capable(CAP_SYS_ADMIN)){
1962                 /*
1963                  * Regular user.  Only allow this with a special mount
1964                  * option, when the user has write+exec access to the
1965                  * subvol root, and when rmdir(2) would have been
1966                  * allowed.
1967                  *
1968                  * Note that this is _not_ check that the subvol is
1969                  * empty or doesn't contain data that we wouldn't
1970                  * otherwise be able to delete.
1971                  *
1972                  * Users who want to delete empty subvols should try
1973                  * rmdir(2).
1974                  */
1975                 err = -EPERM;
1976                 if (!btrfs_test_opt(root, USER_SUBVOL_RM_ALLOWED))
1977                         goto out_dput;
1978
1979                 /*
1980                  * Do not allow deletion if the parent dir is the same
1981                  * as the dir to be deleted.  That means the ioctl
1982                  * must be called on the dentry referencing the root
1983                  * of the subvol, not a random directory contained
1984                  * within it.
1985                  */
1986                 err = -EINVAL;
1987                 if (root == dest)
1988                         goto out_dput;
1989
1990                 err = inode_permission(inode, MAY_WRITE | MAY_EXEC);
1991                 if (err)
1992                         goto out_dput;
1993
1994                 /* check if subvolume may be deleted by a non-root user */
1995                 err = btrfs_may_delete(dir, dentry, 1);
1996                 if (err)
1997                         goto out_dput;
1998         }
1999
2000         if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) {
2001                 err = -EINVAL;
2002                 goto out_dput;
2003         }
2004
2005         mutex_lock(&inode->i_mutex);
2006         err = d_invalidate(dentry);
2007         if (err)
2008                 goto out_unlock;
2009
2010         down_write(&root->fs_info->subvol_sem);
2011
2012         err = may_destroy_subvol(dest);
2013         if (err)
2014                 goto out_up_write;
2015
2016         trans = btrfs_start_transaction(root, 0);
2017         if (IS_ERR(trans)) {
2018                 err = PTR_ERR(trans);
2019                 goto out_up_write;
2020         }
2021         trans->block_rsv = &root->fs_info->global_block_rsv;
2022
2023         ret = btrfs_unlink_subvol(trans, root, dir,
2024                                 dest->root_key.objectid,
2025                                 dentry->d_name.name,
2026                                 dentry->d_name.len);
2027         if (ret) {
2028                 err = ret;
2029                 btrfs_abort_transaction(trans, root, ret);
2030                 goto out_end_trans;
2031         }
2032
2033         btrfs_record_root_in_trans(trans, dest);
2034
2035         memset(&dest->root_item.drop_progress, 0,
2036                 sizeof(dest->root_item.drop_progress));
2037         dest->root_item.drop_level = 0;
2038         btrfs_set_root_refs(&dest->root_item, 0);
2039
2040         if (!xchg(&dest->orphan_item_inserted, 1)) {
2041                 ret = btrfs_insert_orphan_item(trans,
2042                                         root->fs_info->tree_root,
2043                                         dest->root_key.objectid);
2044                 if (ret) {
2045                         btrfs_abort_transaction(trans, root, ret);
2046                         err = ret;
2047                         goto out_end_trans;
2048                 }
2049         }
2050 out_end_trans:
2051         ret = btrfs_end_transaction(trans, root);
2052         if (ret && !err)
2053                 err = ret;
2054         inode->i_flags |= S_DEAD;
2055 out_up_write:
2056         up_write(&root->fs_info->subvol_sem);
2057 out_unlock:
2058         mutex_unlock(&inode->i_mutex);
2059         if (!err) {
2060                 shrink_dcache_sb(root->fs_info->sb);
2061                 btrfs_invalidate_inodes(dest);
2062                 d_delete(dentry);
2063         }
2064 out_dput:
2065         dput(dentry);
2066 out_unlock_dir:
2067         mutex_unlock(&dir->i_mutex);
2068         mnt_drop_write_file(file);
2069 out:
2070         kfree(vol_args);
2071         return err;
2072 }
2073
2074 static int btrfs_ioctl_defrag(struct file *file, void __user *argp)
2075 {
2076         struct inode *inode = fdentry(file)->d_inode;
2077         struct btrfs_root *root = BTRFS_I(inode)->root;
2078         struct btrfs_ioctl_defrag_range_args *range;
2079         int ret;
2080
2081         if (btrfs_root_readonly(root))
2082                 return -EROFS;
2083
2084         ret = mnt_want_write_file(file);
2085         if (ret)
2086                 return ret;
2087
2088         switch (inode->i_mode & S_IFMT) {
2089         case S_IFDIR:
2090                 if (!capable(CAP_SYS_ADMIN)) {
2091                         ret = -EPERM;
2092                         goto out;
2093                 }
2094                 ret = btrfs_defrag_root(root, 0);
2095                 if (ret)
2096                         goto out;
2097                 ret = btrfs_defrag_root(root->fs_info->extent_root, 0);
2098                 break;
2099         case S_IFREG:
2100                 if (!(file->f_mode & FMODE_WRITE)) {
2101                         ret = -EINVAL;
2102                         goto out;
2103                 }
2104
2105                 range = kzalloc(sizeof(*range), GFP_KERNEL);
2106                 if (!range) {
2107                         ret = -ENOMEM;
2108                         goto out;
2109                 }
2110
2111                 if (argp) {
2112                         if (copy_from_user(range, argp,
2113                                            sizeof(*range))) {
2114                                 ret = -EFAULT;
2115                                 kfree(range);
2116                                 goto out;
2117                         }
2118                         /* compression requires us to start the IO */
2119                         if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
2120                                 range->flags |= BTRFS_DEFRAG_RANGE_START_IO;
2121                                 range->extent_thresh = (u32)-1;
2122                         }
2123                 } else {
2124                         /* the rest are all set to zero by kzalloc */
2125                         range->len = (u64)-1;
2126                 }
2127                 ret = btrfs_defrag_file(fdentry(file)->d_inode, file,
2128                                         range, 0, 0);
2129                 if (ret > 0)
2130                         ret = 0;
2131                 kfree(range);
2132                 break;
2133         default:
2134                 ret = -EINVAL;
2135         }
2136 out:
2137         mnt_drop_write_file(file);
2138         return ret;
2139 }
2140
2141 static long btrfs_ioctl_add_dev(struct btrfs_root *root, void __user *arg)
2142 {
2143         struct btrfs_ioctl_vol_args *vol_args;
2144         int ret;
2145
2146         if (!capable(CAP_SYS_ADMIN))
2147                 return -EPERM;
2148
2149         mutex_lock(&root->fs_info->volume_mutex);
2150         if (root->fs_info->balance_ctl) {
2151                 printk(KERN_INFO "btrfs: balance in progress\n");
2152                 ret = -EINVAL;
2153                 goto out;
2154         }
2155
2156         vol_args = memdup_user(arg, sizeof(*vol_args));
2157         if (IS_ERR(vol_args)) {
2158                 ret = PTR_ERR(vol_args);
2159                 goto out;
2160         }
2161
2162         vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2163         ret = btrfs_init_new_device(root, vol_args->name);
2164
2165         kfree(vol_args);
2166 out:
2167         mutex_unlock(&root->fs_info->volume_mutex);
2168         return ret;
2169 }
2170
2171 static long btrfs_ioctl_rm_dev(struct btrfs_root *root, void __user *arg)
2172 {
2173         struct btrfs_ioctl_vol_args *vol_args;
2174         int ret;
2175
2176         if (!capable(CAP_SYS_ADMIN))
2177                 return -EPERM;
2178
2179         if (root->fs_info->sb->s_flags & MS_RDONLY)
2180                 return -EROFS;
2181
2182         mutex_lock(&root->fs_info->volume_mutex);
2183         if (root->fs_info->balance_ctl) {
2184                 printk(KERN_INFO "btrfs: balance in progress\n");
2185                 ret = -EINVAL;
2186                 goto out;
2187         }
2188
2189         vol_args = memdup_user(arg, sizeof(*vol_args));
2190         if (IS_ERR(vol_args)) {
2191                 ret = PTR_ERR(vol_args);
2192                 goto out;
2193         }
2194
2195         vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2196         ret = btrfs_rm_device(root, vol_args->name);
2197
2198         kfree(vol_args);
2199 out:
2200         mutex_unlock(&root->fs_info->volume_mutex);
2201         return ret;
2202 }
2203
2204 static long btrfs_ioctl_fs_info(struct btrfs_root *root, void __user *arg)
2205 {
2206         struct btrfs_ioctl_fs_info_args *fi_args;
2207         struct btrfs_device *device;
2208         struct btrfs_device *next;
2209         struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
2210         int ret = 0;
2211
2212         if (!capable(CAP_SYS_ADMIN))
2213                 return -EPERM;
2214
2215         fi_args = kzalloc(sizeof(*fi_args), GFP_KERNEL);
2216         if (!fi_args)
2217                 return -ENOMEM;
2218
2219         fi_args->num_devices = fs_devices->num_devices;
2220         memcpy(&fi_args->fsid, root->fs_info->fsid, sizeof(fi_args->fsid));
2221
2222         mutex_lock(&fs_devices->device_list_mutex);
2223         list_for_each_entry_safe(device, next, &fs_devices->devices, dev_list) {
2224                 if (device->devid > fi_args->max_id)
2225                         fi_args->max_id = device->devid;
2226         }
2227         mutex_unlock(&fs_devices->device_list_mutex);
2228
2229         if (copy_to_user(arg, fi_args, sizeof(*fi_args)))
2230                 ret = -EFAULT;
2231
2232         kfree(fi_args);
2233         return ret;
2234 }
2235
2236 static long btrfs_ioctl_dev_info(struct btrfs_root *root, void __user *arg)
2237 {
2238         struct btrfs_ioctl_dev_info_args *di_args;
2239         struct btrfs_device *dev;
2240         struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
2241         int ret = 0;
2242         char *s_uuid = NULL;
2243         char empty_uuid[BTRFS_UUID_SIZE] = {0};
2244
2245         if (!capable(CAP_SYS_ADMIN))
2246                 return -EPERM;
2247
2248         di_args = memdup_user(arg, sizeof(*di_args));
2249         if (IS_ERR(di_args))
2250                 return PTR_ERR(di_args);
2251
2252         if (memcmp(empty_uuid, di_args->uuid, BTRFS_UUID_SIZE) != 0)
2253                 s_uuid = di_args->uuid;
2254
2255         mutex_lock(&fs_devices->device_list_mutex);
2256         dev = btrfs_find_device(root, di_args->devid, s_uuid, NULL);
2257         mutex_unlock(&fs_devices->device_list_mutex);
2258
2259         if (!dev) {
2260                 ret = -ENODEV;
2261                 goto out;
2262         }
2263
2264         di_args->devid = dev->devid;
2265         di_args->bytes_used = dev->bytes_used;
2266         di_args->total_bytes = dev->total_bytes;
2267         memcpy(di_args->uuid, dev->uuid, sizeof(di_args->uuid));
2268         if (dev->name) {
2269                 struct rcu_string *name;
2270
2271                 rcu_read_lock();
2272                 name = rcu_dereference(dev->name);
2273                 strncpy(di_args->path, name->str, sizeof(di_args->path));
2274                 rcu_read_unlock();
2275                 di_args->path[sizeof(di_args->path) - 1] = 0;
2276         } else {
2277                 di_args->path[0] = '\0';
2278         }
2279
2280 out:
2281         if (ret == 0 && copy_to_user(arg, di_args, sizeof(*di_args)))
2282                 ret = -EFAULT;
2283
2284         kfree(di_args);
2285         return ret;
2286 }
2287
2288 static noinline long btrfs_ioctl_clone(struct file *file, unsigned long srcfd,
2289                                        u64 off, u64 olen, u64 destoff)
2290 {
2291         struct inode *inode = fdentry(file)->d_inode;
2292         struct btrfs_root *root = BTRFS_I(inode)->root;
2293         struct file *src_file;
2294         struct inode *src;
2295         struct btrfs_trans_handle *trans;
2296         struct btrfs_path *path;
2297         struct extent_buffer *leaf;
2298         char *buf;
2299         struct btrfs_key key;
2300         u32 nritems;
2301         int slot;
2302         int ret;
2303         u64 len = olen;
2304         u64 bs = root->fs_info->sb->s_blocksize;
2305         u64 hint_byte;
2306
2307         /*
2308          * TODO:
2309          * - split compressed inline extents.  annoying: we need to
2310          *   decompress into destination's address_space (the file offset
2311          *   may change, so source mapping won't do), then recompress (or
2312          *   otherwise reinsert) a subrange.
2313          * - allow ranges within the same file to be cloned (provided
2314          *   they don't overlap)?
2315          */
2316
2317         /* the destination must be opened for writing */
2318         if (!(file->f_mode & FMODE_WRITE) || (file->f_flags & O_APPEND))
2319                 return -EINVAL;
2320
2321         if (btrfs_root_readonly(root))
2322                 return -EROFS;
2323
2324         ret = mnt_want_write_file(file);
2325         if (ret)
2326                 return ret;
2327
2328         src_file = fget(srcfd);
2329         if (!src_file) {
2330                 ret = -EBADF;
2331                 goto out_drop_write;
2332         }
2333
2334         src = src_file->f_dentry->d_inode;
2335
2336         ret = -EINVAL;
2337         if (src == inode)
2338                 goto out_fput;
2339
2340         /* the src must be open for reading */
2341         if (!(src_file->f_mode & FMODE_READ))
2342                 goto out_fput;
2343
2344         /* don't make the dst file partly checksummed */
2345         if ((BTRFS_I(src)->flags & BTRFS_INODE_NODATASUM) !=
2346             (BTRFS_I(inode)->flags & BTRFS_INODE_NODATASUM))
2347                 goto out_fput;
2348
2349         ret = -EISDIR;
2350         if (S_ISDIR(src->i_mode) || S_ISDIR(inode->i_mode))
2351                 goto out_fput;
2352
2353         ret = -EXDEV;
2354         if (src->i_sb != inode->i_sb || BTRFS_I(src)->root != root)
2355                 goto out_fput;
2356
2357         ret = -ENOMEM;
2358         buf = vmalloc(btrfs_level_size(root, 0));
2359         if (!buf)
2360                 goto out_fput;
2361
2362         path = btrfs_alloc_path();
2363         if (!path) {
2364                 vfree(buf);
2365                 goto out_fput;
2366         }
2367         path->reada = 2;
2368
2369         if (inode < src) {
2370                 mutex_lock_nested(&inode->i_mutex, I_MUTEX_PARENT);
2371                 mutex_lock_nested(&src->i_mutex, I_MUTEX_CHILD);
2372         } else {
2373                 mutex_lock_nested(&src->i_mutex, I_MUTEX_PARENT);
2374                 mutex_lock_nested(&inode->i_mutex, I_MUTEX_CHILD);
2375         }
2376
2377         /* determine range to clone */
2378         ret = -EINVAL;
2379         if (off + len > src->i_size || off + len < off)
2380                 goto out_unlock;
2381         if (len == 0)
2382                 olen = len = src->i_size - off;
2383         /* if we extend to eof, continue to block boundary */
2384         if (off + len == src->i_size)
2385                 len = ALIGN(src->i_size, bs) - off;
2386
2387         /* verify the end result is block aligned */
2388         if (!IS_ALIGNED(off, bs) || !IS_ALIGNED(off + len, bs) ||
2389             !IS_ALIGNED(destoff, bs))
2390                 goto out_unlock;
2391
2392         if (destoff > inode->i_size) {
2393                 ret = btrfs_cont_expand(inode, inode->i_size, destoff);
2394                 if (ret)
2395                         goto out_unlock;
2396         }
2397
2398         /* truncate page cache pages from target inode range */
2399         truncate_inode_pages_range(&inode->i_data, destoff,
2400                                    PAGE_CACHE_ALIGN(destoff + len) - 1);
2401
2402         /* do any pending delalloc/csum calc on src, one way or
2403            another, and lock file content */
2404         while (1) {
2405                 struct btrfs_ordered_extent *ordered;
2406                 lock_extent(&BTRFS_I(src)->io_tree, off, off+len);
2407                 ordered = btrfs_lookup_first_ordered_extent(src, off+len);
2408                 if (!ordered &&
2409                     !test_range_bit(&BTRFS_I(src)->io_tree, off, off+len,
2410                                    EXTENT_DELALLOC, 0, NULL))
2411                         break;
2412                 unlock_extent(&BTRFS_I(src)->io_tree, off, off+len);
2413                 if (ordered)
2414                         btrfs_put_ordered_extent(ordered);
2415                 btrfs_wait_ordered_range(src, off, len);
2416         }
2417
2418         /* clone data */
2419         key.objectid = btrfs_ino(src);
2420         key.type = BTRFS_EXTENT_DATA_KEY;
2421         key.offset = 0;
2422
2423         while (1) {
2424                 /*
2425                  * note the key will change type as we walk through the
2426                  * tree.
2427                  */
2428                 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
2429                 if (ret < 0)
2430                         goto out;
2431
2432                 nritems = btrfs_header_nritems(path->nodes[0]);
2433                 if (path->slots[0] >= nritems) {
2434                         ret = btrfs_next_leaf(root, path);
2435                         if (ret < 0)
2436                                 goto out;
2437                         if (ret > 0)
2438                                 break;
2439                         nritems = btrfs_header_nritems(path->nodes[0]);
2440                 }
2441                 leaf = path->nodes[0];
2442                 slot = path->slots[0];
2443
2444                 btrfs_item_key_to_cpu(leaf, &key, slot);
2445                 if (btrfs_key_type(&key) > BTRFS_EXTENT_DATA_KEY ||
2446                     key.objectid != btrfs_ino(src))
2447                         break;
2448
2449                 if (btrfs_key_type(&key) == BTRFS_EXTENT_DATA_KEY) {
2450                         struct btrfs_file_extent_item *extent;
2451                         int type;
2452                         u32 size;
2453                         struct btrfs_key new_key;
2454                         u64 disko = 0, diskl = 0;
2455                         u64 datao = 0, datal = 0;
2456                         u8 comp;
2457                         u64 endoff;
2458
2459                         size = btrfs_item_size_nr(leaf, slot);
2460                         read_extent_buffer(leaf, buf,
2461                                            btrfs_item_ptr_offset(leaf, slot),
2462                                            size);
2463
2464                         extent = btrfs_item_ptr(leaf, slot,
2465                                                 struct btrfs_file_extent_item);
2466                         comp = btrfs_file_extent_compression(leaf, extent);
2467                         type = btrfs_file_extent_type(leaf, extent);
2468                         if (type == BTRFS_FILE_EXTENT_REG ||
2469                             type == BTRFS_FILE_EXTENT_PREALLOC) {
2470                                 disko = btrfs_file_extent_disk_bytenr(leaf,
2471                                                                       extent);
2472                                 diskl = btrfs_file_extent_disk_num_bytes(leaf,
2473                                                                  extent);
2474                                 datao = btrfs_file_extent_offset(leaf, extent);
2475                                 datal = btrfs_file_extent_num_bytes(leaf,
2476                                                                     extent);
2477                         } else if (type == BTRFS_FILE_EXTENT_INLINE) {
2478                                 /* take upper bound, may be compressed */
2479                                 datal = btrfs_file_extent_ram_bytes(leaf,
2480                                                                     extent);
2481                         }
2482                         btrfs_release_path(path);
2483
2484                         if (key.offset + datal <= off ||
2485                             key.offset >= off+len)
2486                                 goto next;
2487
2488                         memcpy(&new_key, &key, sizeof(new_key));
2489                         new_key.objectid = btrfs_ino(inode);
2490                         if (off <= key.offset)
2491                                 new_key.offset = key.offset + destoff - off;
2492                         else
2493                                 new_key.offset = destoff;
2494
2495                         /*
2496                          * 1 - adjusting old extent (we may have to split it)
2497                          * 1 - add new extent
2498                          * 1 - inode update
2499                          */
2500                         trans = btrfs_start_transaction(root, 3);
2501                         if (IS_ERR(trans)) {
2502                                 ret = PTR_ERR(trans);
2503                                 goto out;
2504                         }
2505
2506                         if (type == BTRFS_FILE_EXTENT_REG ||
2507                             type == BTRFS_FILE_EXTENT_PREALLOC) {
2508                                 /*
2509                                  *    a  | --- range to clone ---|  b
2510                                  * | ------------- extent ------------- |
2511                                  */
2512
2513                                 /* substract range b */
2514                                 if (key.offset + datal > off + len)
2515                                         datal = off + len - key.offset;
2516
2517                                 /* substract range a */
2518                                 if (off > key.offset) {
2519                                         datao += off - key.offset;
2520                                         datal -= off - key.offset;
2521                                 }
2522
2523                                 ret = btrfs_drop_extents(trans, inode,
2524                                                          new_key.offset,
2525                                                          new_key.offset + datal,
2526                                                          &hint_byte, 1);
2527                                 if (ret) {
2528                                         btrfs_abort_transaction(trans, root,
2529                                                                 ret);
2530                                         btrfs_end_transaction(trans, root);
2531                                         goto out;
2532                                 }
2533
2534                                 ret = btrfs_insert_empty_item(trans, root, path,
2535                                                               &new_key, size);
2536                                 if (ret) {
2537                                         btrfs_abort_transaction(trans, root,
2538                                                                 ret);
2539                                         btrfs_end_transaction(trans, root);
2540                                         goto out;
2541                                 }
2542
2543                                 leaf = path->nodes[0];
2544                                 slot = path->slots[0];
2545                                 write_extent_buffer(leaf, buf,
2546                                             btrfs_item_ptr_offset(leaf, slot),
2547                                             size);
2548
2549                                 extent = btrfs_item_ptr(leaf, slot,
2550                                                 struct btrfs_file_extent_item);
2551
2552                                 /* disko == 0 means it's a hole */
2553                                 if (!disko)
2554                                         datao = 0;
2555
2556                                 btrfs_set_file_extent_offset(leaf, extent,
2557                                                              datao);
2558                                 btrfs_set_file_extent_num_bytes(leaf, extent,
2559                                                                 datal);
2560                                 if (disko) {
2561                                         inode_add_bytes(inode, datal);
2562                                         ret = btrfs_inc_extent_ref(trans, root,
2563                                                         disko, diskl, 0,
2564                                                         root->root_key.objectid,
2565                                                         btrfs_ino(inode),
2566                                                         new_key.offset - datao,
2567                                                         0);
2568                                         if (ret) {
2569                                                 btrfs_abort_transaction(trans,
2570                                                                         root,
2571                                                                         ret);
2572                                                 btrfs_end_transaction(trans,
2573                                                                       root);
2574                                                 goto out;
2575
2576                                         }
2577                                 }
2578                         } else if (type == BTRFS_FILE_EXTENT_INLINE) {
2579                                 u64 skip = 0;
2580                                 u64 trim = 0;
2581                                 if (off > key.offset) {
2582                                         skip = off - key.offset;
2583                                         new_key.offset += skip;
2584                                 }
2585
2586                                 if (key.offset + datal > off+len)
2587                                         trim = key.offset + datal - (off+len);
2588
2589                                 if (comp && (skip || trim)) {
2590                                         ret = -EINVAL;
2591                                         btrfs_end_transaction(trans, root);
2592                                         goto out;
2593                                 }
2594                                 size -= skip + trim;
2595                                 datal -= skip + trim;
2596
2597                                 ret = btrfs_drop_extents(trans, inode,
2598                                                          new_key.offset,
2599                                                          new_key.offset + datal,
2600                                                          &hint_byte, 1);
2601                                 if (ret) {
2602                                         btrfs_abort_transaction(trans, root,
2603                                                                 ret);
2604                                         btrfs_end_transaction(trans, root);
2605                                         goto out;
2606                                 }
2607
2608                                 ret = btrfs_insert_empty_item(trans, root, path,
2609                                                               &new_key, size);
2610                                 if (ret) {
2611                                         btrfs_abort_transaction(trans, root,
2612                                                                 ret);
2613                                         btrfs_end_transaction(trans, root);
2614                                         goto out;
2615                                 }
2616
2617                                 if (skip) {
2618                                         u32 start =
2619                                           btrfs_file_extent_calc_inline_size(0);
2620                                         memmove(buf+start, buf+start+skip,
2621                                                 datal);
2622                                 }
2623
2624                                 leaf = path->nodes[0];
2625                                 slot = path->slots[0];
2626                                 write_extent_buffer(leaf, buf,
2627                                             btrfs_item_ptr_offset(leaf, slot),
2628                                             size);
2629                                 inode_add_bytes(inode, datal);
2630                         }
2631
2632                         btrfs_mark_buffer_dirty(leaf);
2633                         btrfs_release_path(path);
2634
2635                         inode_inc_iversion(inode);
2636                         inode->i_mtime = inode->i_ctime = CURRENT_TIME;
2637
2638                         /*
2639                          * we round up to the block size at eof when
2640                          * determining which extents to clone above,
2641                          * but shouldn't round up the file size
2642                          */
2643                         endoff = new_key.offset + datal;
2644                         if (endoff > destoff+olen)
2645                                 endoff = destoff+olen;
2646                         if (endoff > inode->i_size)
2647                                 btrfs_i_size_write(inode, endoff);
2648
2649                         ret = btrfs_update_inode(trans, root, inode);
2650                         if (ret) {
2651                                 btrfs_abort_transaction(trans, root, ret);
2652                                 btrfs_end_transaction(trans, root);
2653                                 goto out;
2654                         }
2655                         ret = btrfs_end_transaction(trans, root);
2656                 }
2657 next:
2658                 btrfs_release_path(path);
2659                 key.offset++;
2660         }
2661         ret = 0;
2662 out:
2663         btrfs_release_path(path);
2664         unlock_extent(&BTRFS_I(src)->io_tree, off, off+len);
2665 out_unlock:
2666         mutex_unlock(&src->i_mutex);
2667         mutex_unlock(&inode->i_mutex);
2668         vfree(buf);
2669         btrfs_free_path(path);
2670 out_fput:
2671         fput(src_file);
2672 out_drop_write:
2673         mnt_drop_write_file(file);
2674         return ret;
2675 }
2676
2677 static long btrfs_ioctl_clone_range(struct file *file, void __user *argp)
2678 {
2679         struct btrfs_ioctl_clone_range_args args;
2680
2681         if (copy_from_user(&args, argp, sizeof(args)))
2682                 return -EFAULT;
2683         return btrfs_ioctl_clone(file, args.src_fd, args.src_offset,
2684                                  args.src_length, args.dest_offset);
2685 }
2686
2687 /*
2688  * there are many ways the trans_start and trans_end ioctls can lead
2689  * to deadlocks.  They should only be used by applications that
2690  * basically own the machine, and have a very in depth understanding
2691  * of all the possible deadlocks and enospc problems.
2692  */
2693 static long btrfs_ioctl_trans_start(struct file *file)
2694 {
2695         struct inode *inode = fdentry(file)->d_inode;
2696         struct btrfs_root *root = BTRFS_I(inode)->root;
2697         struct btrfs_trans_handle *trans;
2698         int ret;
2699
2700         ret = -EPERM;
2701         if (!capable(CAP_SYS_ADMIN))
2702                 goto out;
2703
2704         ret = -EINPROGRESS;
2705         if (file->private_data)
2706                 goto out;
2707
2708         ret = -EROFS;
2709         if (btrfs_root_readonly(root))
2710                 goto out;
2711
2712         ret = mnt_want_write_file(file);
2713         if (ret)
2714                 goto out;
2715
2716         atomic_inc(&root->fs_info->open_ioctl_trans);
2717
2718         ret = -ENOMEM;
2719         trans = btrfs_start_ioctl_transaction(root);
2720         if (IS_ERR(trans))
2721                 goto out_drop;
2722
2723         file->private_data = trans;
2724         return 0;
2725
2726 out_drop:
2727         atomic_dec(&root->fs_info->open_ioctl_trans);
2728         mnt_drop_write_file(file);
2729 out:
2730         return ret;
2731 }
2732
2733 static long btrfs_ioctl_default_subvol(struct file *file, void __user *argp)
2734 {
2735         struct inode *inode = fdentry(file)->d_inode;
2736         struct btrfs_root *root = BTRFS_I(inode)->root;
2737         struct btrfs_root *new_root;
2738         struct btrfs_dir_item *di;
2739         struct btrfs_trans_handle *trans;
2740         struct btrfs_path *path;
2741         struct btrfs_key location;
2742         struct btrfs_disk_key disk_key;
2743         struct btrfs_super_block *disk_super;
2744         u64 features;
2745         u64 objectid = 0;
2746         u64 dir_id;
2747
2748         if (!capable(CAP_SYS_ADMIN))
2749                 return -EPERM;
2750
2751         if (copy_from_user(&objectid, argp, sizeof(objectid)))
2752                 return -EFAULT;
2753
2754         if (!objectid)
2755                 objectid = root->root_key.objectid;
2756
2757         location.objectid = objectid;
2758         location.type = BTRFS_ROOT_ITEM_KEY;
2759         location.offset = (u64)-1;
2760
2761         new_root = btrfs_read_fs_root_no_name(root->fs_info, &location);
2762         if (IS_ERR(new_root))
2763                 return PTR_ERR(new_root);
2764
2765         if (btrfs_root_refs(&new_root->root_item) == 0)
2766                 return -ENOENT;
2767
2768         path = btrfs_alloc_path();
2769         if (!path)
2770                 return -ENOMEM;
2771         path->leave_spinning = 1;
2772
2773         trans = btrfs_start_transaction(root, 1);
2774         if (IS_ERR(trans)) {
2775                 btrfs_free_path(path);
2776                 return PTR_ERR(trans);
2777         }
2778
2779         dir_id = btrfs_super_root_dir(root->fs_info->super_copy);
2780         di = btrfs_lookup_dir_item(trans, root->fs_info->tree_root, path,
2781                                    dir_id, "default", 7, 1);
2782         if (IS_ERR_OR_NULL(di)) {
2783                 btrfs_free_path(path);
2784                 btrfs_end_transaction(trans, root);
2785                 printk(KERN_ERR "Umm, you don't have the default dir item, "
2786                        "this isn't going to work\n");
2787                 return -ENOENT;
2788         }
2789
2790         btrfs_cpu_key_to_disk(&disk_key, &new_root->root_key);
2791         btrfs_set_dir_item_key(path->nodes[0], di, &disk_key);
2792         btrfs_mark_buffer_dirty(path->nodes[0]);
2793         btrfs_free_path(path);
2794
2795         disk_super = root->fs_info->super_copy;
2796         features = btrfs_super_incompat_flags(disk_super);
2797         if (!(features & BTRFS_FEATURE_INCOMPAT_DEFAULT_SUBVOL)) {
2798                 features |= BTRFS_FEATURE_INCOMPAT_DEFAULT_SUBVOL;
2799                 btrfs_set_super_incompat_flags(disk_super, features);
2800         }
2801         btrfs_end_transaction(trans, root);
2802
2803         return 0;
2804 }
2805
2806 static void get_block_group_info(struct list_head *groups_list,
2807                                  struct btrfs_ioctl_space_info *space)
2808 {
2809         struct btrfs_block_group_cache *block_group;
2810
2811         space->total_bytes = 0;
2812         space->used_bytes = 0;
2813         space->flags = 0;
2814         list_for_each_entry(block_group, groups_list, list) {
2815                 space->flags = block_group->flags;
2816                 space->total_bytes += block_group->key.offset;
2817                 space->used_bytes +=
2818                         btrfs_block_group_used(&block_group->item);
2819         }
2820 }
2821
2822 long btrfs_ioctl_space_info(struct btrfs_root *root, void __user *arg)
2823 {
2824         struct btrfs_ioctl_space_args space_args;
2825         struct btrfs_ioctl_space_info space;
2826         struct btrfs_ioctl_space_info *dest;
2827         struct btrfs_ioctl_space_info *dest_orig;
2828         struct btrfs_ioctl_space_info __user *user_dest;
2829         struct btrfs_space_info *info;
2830         u64 types[] = {BTRFS_BLOCK_GROUP_DATA,
2831                        BTRFS_BLOCK_GROUP_SYSTEM,
2832                        BTRFS_BLOCK_GROUP_METADATA,
2833                        BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA};
2834         int num_types = 4;
2835         int alloc_size;
2836         int ret = 0;
2837         u64 slot_count = 0;
2838         int i, c;
2839
2840         if (copy_from_user(&space_args,
2841                            (struct btrfs_ioctl_space_args __user *)arg,
2842                            sizeof(space_args)))
2843                 return -EFAULT;
2844
2845         for (i = 0; i < num_types; i++) {
2846                 struct btrfs_space_info *tmp;
2847
2848                 info = NULL;
2849                 rcu_read_lock();
2850                 list_for_each_entry_rcu(tmp, &root->fs_info->space_info,
2851                                         list) {
2852                         if (tmp->flags == types[i]) {
2853                                 info = tmp;
2854                                 break;
2855                         }
2856                 }
2857                 rcu_read_unlock();
2858
2859                 if (!info)
2860                         continue;
2861
2862                 down_read(&info->groups_sem);
2863                 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
2864                         if (!list_empty(&info->block_groups[c]))
2865                                 slot_count++;
2866                 }
2867                 up_read(&info->groups_sem);
2868         }
2869
2870         /* space_slots == 0 means they are asking for a count */
2871         if (space_args.space_slots == 0) {
2872                 space_args.total_spaces = slot_count;
2873                 goto out;
2874         }
2875
2876         slot_count = min_t(u64, space_args.space_slots, slot_count);
2877
2878         alloc_size = sizeof(*dest) * slot_count;
2879
2880         /* we generally have at most 6 or so space infos, one for each raid
2881          * level.  So, a whole page should be more than enough for everyone
2882          */
2883         if (alloc_size > PAGE_CACHE_SIZE)
2884                 return -ENOMEM;
2885
2886         space_args.total_spaces = 0;
2887         dest = kmalloc(alloc_size, GFP_NOFS);
2888         if (!dest)
2889                 return -ENOMEM;
2890         dest_orig = dest;
2891
2892         /* now we have a buffer to copy into */
2893         for (i = 0; i < num_types; i++) {
2894                 struct btrfs_space_info *tmp;
2895
2896                 if (!slot_count)
2897                         break;
2898
2899                 info = NULL;
2900                 rcu_read_lock();
2901                 list_for_each_entry_rcu(tmp, &root->fs_info->space_info,
2902                                         list) {
2903                         if (tmp->flags == types[i]) {
2904                                 info = tmp;
2905                                 break;
2906                         }
2907                 }
2908                 rcu_read_unlock();
2909
2910                 if (!info)
2911                         continue;
2912                 down_read(&info->groups_sem);
2913                 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
2914                         if (!list_empty(&info->block_groups[c])) {
2915                                 get_block_group_info(&info->block_groups[c],
2916                                                      &space);
2917                                 memcpy(dest, &space, sizeof(space));
2918                                 dest++;
2919                                 space_args.total_spaces++;
2920                                 slot_count--;
2921                         }
2922                         if (!slot_count)
2923                                 break;
2924                 }
2925                 up_read(&info->groups_sem);
2926         }
2927
2928         user_dest = (struct btrfs_ioctl_space_info __user *)
2929                 (arg + sizeof(struct btrfs_ioctl_space_args));
2930
2931         if (copy_to_user(user_dest, dest_orig, alloc_size))
2932                 ret = -EFAULT;
2933
2934         kfree(dest_orig);
2935 out:
2936         if (ret == 0 && copy_to_user(arg, &space_args, sizeof(space_args)))
2937                 ret = -EFAULT;
2938
2939         return ret;
2940 }
2941
2942 /*
2943  * there are many ways the trans_start and trans_end ioctls can lead
2944  * to deadlocks.  They should only be used by applications that
2945  * basically own the machine, and have a very in depth understanding
2946  * of all the possible deadlocks and enospc problems.
2947  */
2948 long btrfs_ioctl_trans_end(struct file *file)
2949 {
2950         struct inode *inode = fdentry(file)->d_inode;
2951         struct btrfs_root *root = BTRFS_I(inode)->root;
2952         struct btrfs_trans_handle *trans;
2953
2954         trans = file->private_data;
2955         if (!trans)
2956                 return -EINVAL;
2957         file->private_data = NULL;
2958
2959         btrfs_end_transaction(trans, root);
2960
2961         atomic_dec(&root->fs_info->open_ioctl_trans);
2962
2963         mnt_drop_write_file(file);
2964         return 0;
2965 }
2966
2967 static noinline long btrfs_ioctl_start_sync(struct file *file, void __user *argp)
2968 {
2969         struct btrfs_root *root = BTRFS_I(file->f_dentry->d_inode)->root;
2970         struct btrfs_trans_handle *trans;
2971         u64 transid;
2972         int ret;
2973
2974         trans = btrfs_start_transaction(root, 0);
2975         if (IS_ERR(trans))
2976                 return PTR_ERR(trans);
2977         transid = trans->transid;
2978         ret = btrfs_commit_transaction_async(trans, root, 0);
2979         if (ret) {
2980                 btrfs_end_transaction(trans, root);
2981                 return ret;
2982         }
2983
2984         if (argp)
2985                 if (copy_to_user(argp, &transid, sizeof(transid)))
2986                         return -EFAULT;
2987         return 0;
2988 }
2989
2990 static noinline long btrfs_ioctl_wait_sync(struct file *file, void __user *argp)
2991 {
2992         struct btrfs_root *root = BTRFS_I(file->f_dentry->d_inode)->root;
2993         u64 transid;
2994
2995         if (argp) {
2996                 if (copy_from_user(&transid, argp, sizeof(transid)))
2997                         return -EFAULT;
2998         } else {
2999                 transid = 0;  /* current trans */
3000         }
3001         return btrfs_wait_for_commit(root, transid);
3002 }
3003
3004 static long btrfs_ioctl_scrub(struct btrfs_root *root, void __user *arg)
3005 {
3006         int ret;
3007         struct btrfs_ioctl_scrub_args *sa;
3008
3009         if (!capable(CAP_SYS_ADMIN))
3010                 return -EPERM;
3011
3012         sa = memdup_user(arg, sizeof(*sa));
3013         if (IS_ERR(sa))
3014                 return PTR_ERR(sa);
3015
3016         ret = btrfs_scrub_dev(root, sa->devid, sa->start, sa->end,
3017                               &sa->progress, sa->flags & BTRFS_SCRUB_READONLY);
3018
3019         if (copy_to_user(arg, sa, sizeof(*sa)))
3020                 ret = -EFAULT;
3021
3022         kfree(sa);
3023         return ret;
3024 }
3025
3026 static long btrfs_ioctl_scrub_cancel(struct btrfs_root *root, void __user *arg)
3027 {
3028         if (!capable(CAP_SYS_ADMIN))
3029                 return -EPERM;
3030
3031         return btrfs_scrub_cancel(root);
3032 }
3033
3034 static long btrfs_ioctl_scrub_progress(struct btrfs_root *root,
3035                                        void __user *arg)
3036 {
3037         struct btrfs_ioctl_scrub_args *sa;
3038         int ret;
3039
3040         if (!capable(CAP_SYS_ADMIN))
3041                 return -EPERM;
3042
3043         sa = memdup_user(arg, sizeof(*sa));
3044         if (IS_ERR(sa))
3045                 return PTR_ERR(sa);
3046
3047         ret = btrfs_scrub_progress(root, sa->devid, &sa->progress);
3048
3049         if (copy_to_user(arg, sa, sizeof(*sa)))
3050                 ret = -EFAULT;
3051
3052         kfree(sa);
3053         return ret;
3054 }
3055
3056 static long btrfs_ioctl_get_dev_stats(struct btrfs_root *root,
3057                                       void __user *arg, int reset_after_read)
3058 {
3059         struct btrfs_ioctl_get_dev_stats *sa;
3060         int ret;
3061
3062         if (reset_after_read && !capable(CAP_SYS_ADMIN))
3063                 return -EPERM;
3064
3065         sa = memdup_user(arg, sizeof(*sa));
3066         if (IS_ERR(sa))
3067                 return PTR_ERR(sa);
3068
3069         ret = btrfs_get_dev_stats(root, sa, reset_after_read);
3070
3071         if (copy_to_user(arg, sa, sizeof(*sa)))
3072                 ret = -EFAULT;
3073
3074         kfree(sa);
3075         return ret;
3076 }
3077
3078 static long btrfs_ioctl_ino_to_path(struct btrfs_root *root, void __user *arg)
3079 {
3080         int ret = 0;
3081         int i;
3082         u64 rel_ptr;
3083         int size;
3084         struct btrfs_ioctl_ino_path_args *ipa = NULL;
3085         struct inode_fs_paths *ipath = NULL;
3086         struct btrfs_path *path;
3087
3088         if (!capable(CAP_SYS_ADMIN))
3089                 return -EPERM;
3090
3091         path = btrfs_alloc_path();
3092         if (!path) {
3093                 ret = -ENOMEM;
3094                 goto out;
3095         }
3096
3097         ipa = memdup_user(arg, sizeof(*ipa));
3098         if (IS_ERR(ipa)) {
3099                 ret = PTR_ERR(ipa);
3100                 ipa = NULL;
3101                 goto out;
3102         }
3103
3104         size = min_t(u32, ipa->size, 4096);
3105         ipath = init_ipath(size, root, path);
3106         if (IS_ERR(ipath)) {
3107                 ret = PTR_ERR(ipath);
3108                 ipath = NULL;
3109                 goto out;
3110         }
3111
3112         ret = paths_from_inode(ipa->inum, ipath);
3113         if (ret < 0)
3114                 goto out;
3115
3116         for (i = 0; i < ipath->fspath->elem_cnt; ++i) {
3117                 rel_ptr = ipath->fspath->val[i] -
3118                           (u64)(unsigned long)ipath->fspath->val;
3119                 ipath->fspath->val[i] = rel_ptr;
3120         }
3121
3122         ret = copy_to_user((void *)(unsigned long)ipa->fspath,
3123                            (void *)(unsigned long)ipath->fspath, size);
3124         if (ret) {
3125                 ret = -EFAULT;
3126                 goto out;
3127         }
3128
3129 out:
3130         btrfs_free_path(path);
3131         free_ipath(ipath);
3132         kfree(ipa);
3133
3134         return ret;
3135 }
3136
3137 static int build_ino_list(u64 inum, u64 offset, u64 root, void *ctx)
3138 {
3139         struct btrfs_data_container *inodes = ctx;
3140         const size_t c = 3 * sizeof(u64);
3141
3142         if (inodes->bytes_left >= c) {
3143                 inodes->bytes_left -= c;
3144                 inodes->val[inodes->elem_cnt] = inum;
3145                 inodes->val[inodes->elem_cnt + 1] = offset;
3146                 inodes->val[inodes->elem_cnt + 2] = root;
3147                 inodes->elem_cnt += 3;
3148         } else {
3149                 inodes->bytes_missing += c - inodes->bytes_left;
3150                 inodes->bytes_left = 0;
3151                 inodes->elem_missed += 3;
3152         }
3153
3154         return 0;
3155 }
3156
3157 static long btrfs_ioctl_logical_to_ino(struct btrfs_root *root,
3158                                         void __user *arg)
3159 {
3160         int ret = 0;
3161         int size;
3162         u64 extent_item_pos;
3163         struct btrfs_ioctl_logical_ino_args *loi;
3164         struct btrfs_data_container *inodes = NULL;
3165         struct btrfs_path *path = NULL;
3166         struct btrfs_key key;
3167
3168         if (!capable(CAP_SYS_ADMIN))
3169                 return -EPERM;
3170
3171         loi = memdup_user(arg, sizeof(*loi));
3172         if (IS_ERR(loi)) {
3173                 ret = PTR_ERR(loi);
3174                 loi = NULL;
3175                 goto out;
3176         }
3177
3178         path = btrfs_alloc_path();
3179         if (!path) {
3180                 ret = -ENOMEM;
3181                 goto out;
3182         }
3183
3184         size = min_t(u32, loi->size, 4096);
3185         inodes = init_data_container(size);
3186         if (IS_ERR(inodes)) {
3187                 ret = PTR_ERR(inodes);
3188                 inodes = NULL;
3189                 goto out;
3190         }
3191
3192         ret = extent_from_logical(root->fs_info, loi->logical, path, &key);
3193         btrfs_release_path(path);
3194
3195         if (ret & BTRFS_EXTENT_FLAG_TREE_BLOCK)
3196                 ret = -ENOENT;
3197         if (ret < 0)
3198                 goto out;
3199
3200         extent_item_pos = loi->logical - key.objectid;
3201         ret = iterate_extent_inodes(root->fs_info, key.objectid,
3202                                         extent_item_pos, 0, build_ino_list,
3203                                         inodes);
3204
3205         if (ret < 0)
3206                 goto out;
3207
3208         ret = copy_to_user((void *)(unsigned long)loi->inodes,
3209                            (void *)(unsigned long)inodes, size);
3210         if (ret)
3211                 ret = -EFAULT;
3212
3213 out:
3214         btrfs_free_path(path);
3215         kfree(inodes);
3216         kfree(loi);
3217
3218         return ret;
3219 }
3220
3221 void update_ioctl_balance_args(struct btrfs_fs_info *fs_info, int lock,
3222                                struct btrfs_ioctl_balance_args *bargs)
3223 {
3224         struct btrfs_balance_control *bctl = fs_info->balance_ctl;
3225
3226         bargs->flags = bctl->flags;
3227
3228         if (atomic_read(&fs_info->balance_running))
3229                 bargs->state |= BTRFS_BALANCE_STATE_RUNNING;
3230         if (atomic_read(&fs_info->balance_pause_req))
3231                 bargs->state |= BTRFS_BALANCE_STATE_PAUSE_REQ;
3232         if (atomic_read(&fs_info->balance_cancel_req))
3233                 bargs->state |= BTRFS_BALANCE_STATE_CANCEL_REQ;
3234
3235         memcpy(&bargs->data, &bctl->data, sizeof(bargs->data));
3236         memcpy(&bargs->meta, &bctl->meta, sizeof(bargs->meta));
3237         memcpy(&bargs->sys, &bctl->sys, sizeof(bargs->sys));
3238
3239         if (lock) {
3240                 spin_lock(&fs_info->balance_lock);
3241                 memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat));
3242                 spin_unlock(&fs_info->balance_lock);
3243         } else {
3244                 memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat));
3245         }
3246 }
3247
3248 static long btrfs_ioctl_balance(struct file *file, void __user *arg)
3249 {
3250         struct btrfs_root *root = BTRFS_I(fdentry(file)->d_inode)->root;
3251         struct btrfs_fs_info *fs_info = root->fs_info;
3252         struct btrfs_ioctl_balance_args *bargs;
3253         struct btrfs_balance_control *bctl;
3254         int ret;
3255
3256         if (!capable(CAP_SYS_ADMIN))
3257                 return -EPERM;
3258
3259         if (fs_info->sb->s_flags & MS_RDONLY)
3260                 return -EROFS;
3261
3262         ret = mnt_want_write(file->f_path.mnt);
3263         if (ret)
3264                 return ret;
3265
3266         mutex_lock(&fs_info->volume_mutex);
3267         mutex_lock(&fs_info->balance_mutex);
3268
3269         if (arg) {
3270                 bargs = memdup_user(arg, sizeof(*bargs));
3271                 if (IS_ERR(bargs)) {
3272                         ret = PTR_ERR(bargs);
3273                         goto out;
3274                 }
3275
3276                 if (bargs->flags & BTRFS_BALANCE_RESUME) {
3277                         if (!fs_info->balance_ctl) {
3278                                 ret = -ENOTCONN;
3279                                 goto out_bargs;
3280                         }
3281
3282                         bctl = fs_info->balance_ctl;
3283                         spin_lock(&fs_info->balance_lock);
3284                         bctl->flags |= BTRFS_BALANCE_RESUME;
3285                         spin_unlock(&fs_info->balance_lock);
3286
3287                         goto do_balance;
3288                 }
3289         } else {
3290                 bargs = NULL;
3291         }
3292
3293         if (fs_info->balance_ctl) {
3294                 ret = -EINPROGRESS;
3295                 goto out_bargs;
3296         }
3297
3298         bctl = kzalloc(sizeof(*bctl), GFP_NOFS);
3299         if (!bctl) {
3300                 ret = -ENOMEM;
3301                 goto out_bargs;
3302         }
3303
3304         bctl->fs_info = fs_info;
3305         if (arg) {
3306                 memcpy(&bctl->data, &bargs->data, sizeof(bctl->data));
3307                 memcpy(&bctl->meta, &bargs->meta, sizeof(bctl->meta));
3308                 memcpy(&bctl->sys, &bargs->sys, sizeof(bctl->sys));
3309
3310                 bctl->flags = bargs->flags;
3311         } else {
3312                 /* balance everything - no filters */
3313                 bctl->flags |= BTRFS_BALANCE_TYPE_MASK;
3314         }
3315
3316 do_balance:
3317         ret = btrfs_balance(bctl, bargs);
3318         /*
3319          * bctl is freed in __cancel_balance or in free_fs_info if
3320          * restriper was paused all the way until unmount
3321          */
3322         if (arg) {
3323                 if (copy_to_user(arg, bargs, sizeof(*bargs)))
3324                         ret = -EFAULT;
3325         }
3326
3327 out_bargs:
3328         kfree(bargs);
3329 out:
3330         mutex_unlock(&fs_info->balance_mutex);
3331         mutex_unlock(&fs_info->volume_mutex);
3332         mnt_drop_write(file->f_path.mnt);
3333         return ret;
3334 }
3335
3336 static long btrfs_ioctl_balance_ctl(struct btrfs_root *root, int cmd)
3337 {
3338         if (!capable(CAP_SYS_ADMIN))
3339                 return -EPERM;
3340
3341         switch (cmd) {
3342         case BTRFS_BALANCE_CTL_PAUSE:
3343                 return btrfs_pause_balance(root->fs_info);
3344         case BTRFS_BALANCE_CTL_CANCEL:
3345                 return btrfs_cancel_balance(root->fs_info);
3346         }
3347
3348         return -EINVAL;
3349 }
3350
3351 static long btrfs_ioctl_balance_progress(struct btrfs_root *root,
3352                                          void __user *arg)
3353 {
3354         struct btrfs_fs_info *fs_info = root->fs_info;
3355         struct btrfs_ioctl_balance_args *bargs;
3356         int ret = 0;
3357
3358         if (!capable(CAP_SYS_ADMIN))
3359                 return -EPERM;
3360
3361         mutex_lock(&fs_info->balance_mutex);
3362         if (!fs_info->balance_ctl) {
3363                 ret = -ENOTCONN;
3364                 goto out;
3365         }
3366
3367         bargs = kzalloc(sizeof(*bargs), GFP_NOFS);
3368         if (!bargs) {
3369                 ret = -ENOMEM;
3370                 goto out;
3371         }
3372
3373         update_ioctl_balance_args(fs_info, 1, bargs);
3374
3375         if (copy_to_user(arg, bargs, sizeof(*bargs)))
3376                 ret = -EFAULT;
3377
3378         kfree(bargs);
3379 out:
3380         mutex_unlock(&fs_info->balance_mutex);
3381         return ret;
3382 }
3383
3384 long btrfs_ioctl(struct file *file, unsigned int
3385                 cmd, unsigned long arg)
3386 {
3387         struct btrfs_root *root = BTRFS_I(fdentry(file)->d_inode)->root;
3388         void __user *argp = (void __user *)arg;
3389
3390         switch (cmd) {
3391         case FS_IOC_GETFLAGS:
3392                 return btrfs_ioctl_getflags(file, argp);
3393         case FS_IOC_SETFLAGS:
3394                 return btrfs_ioctl_setflags(file, argp);
3395         case FS_IOC_GETVERSION:
3396                 return btrfs_ioctl_getversion(file, argp);
3397         case FITRIM:
3398                 return btrfs_ioctl_fitrim(file, argp);
3399         case BTRFS_IOC_SNAP_CREATE:
3400                 return btrfs_ioctl_snap_create(file, argp, 0);
3401         case BTRFS_IOC_SNAP_CREATE_V2:
3402                 return btrfs_ioctl_snap_create_v2(file, argp, 0);
3403         case BTRFS_IOC_SUBVOL_CREATE:
3404                 return btrfs_ioctl_snap_create(file, argp, 1);
3405         case BTRFS_IOC_SNAP_DESTROY:
3406                 return btrfs_ioctl_snap_destroy(file, argp);
3407         case BTRFS_IOC_SUBVOL_GETFLAGS:
3408                 return btrfs_ioctl_subvol_getflags(file, argp);
3409         case BTRFS_IOC_SUBVOL_SETFLAGS:
3410                 return btrfs_ioctl_subvol_setflags(file, argp);
3411         case BTRFS_IOC_DEFAULT_SUBVOL:
3412                 return btrfs_ioctl_default_subvol(file, argp);
3413         case BTRFS_IOC_DEFRAG:
3414                 return btrfs_ioctl_defrag(file, NULL);
3415         case BTRFS_IOC_DEFRAG_RANGE:
3416                 return btrfs_ioctl_defrag(file, argp);
3417         case BTRFS_IOC_RESIZE:
3418                 return btrfs_ioctl_resize(root, argp);
3419         case BTRFS_IOC_ADD_DEV:
3420                 return btrfs_ioctl_add_dev(root, argp);
3421         case BTRFS_IOC_RM_DEV:
3422                 return btrfs_ioctl_rm_dev(root, argp);
3423         case BTRFS_IOC_FS_INFO:
3424                 return btrfs_ioctl_fs_info(root, argp);
3425         case BTRFS_IOC_DEV_INFO:
3426                 return btrfs_ioctl_dev_info(root, argp);
3427         case BTRFS_IOC_BALANCE:
3428                 return btrfs_ioctl_balance(file, NULL);
3429         case BTRFS_IOC_CLONE:
3430                 return btrfs_ioctl_clone(file, arg, 0, 0, 0);
3431         case BTRFS_IOC_CLONE_RANGE:
3432                 return btrfs_ioctl_clone_range(file, argp);
3433         case BTRFS_IOC_TRANS_START:
3434                 return btrfs_ioctl_trans_start(file);
3435         case BTRFS_IOC_TRANS_END:
3436                 return btrfs_ioctl_trans_end(file);
3437         case BTRFS_IOC_TREE_SEARCH:
3438                 return btrfs_ioctl_tree_search(file, argp);
3439         case BTRFS_IOC_INO_LOOKUP:
3440                 return btrfs_ioctl_ino_lookup(file, argp);
3441         case BTRFS_IOC_INO_PATHS:
3442                 return btrfs_ioctl_ino_to_path(root, argp);
3443         case BTRFS_IOC_LOGICAL_INO:
3444                 return btrfs_ioctl_logical_to_ino(root, argp);
3445         case BTRFS_IOC_SPACE_INFO:
3446                 return btrfs_ioctl_space_info(root, argp);
3447         case BTRFS_IOC_SYNC:
3448                 btrfs_sync_fs(file->f_dentry->d_sb, 1);
3449                 return 0;
3450         case BTRFS_IOC_START_SYNC:
3451                 return btrfs_ioctl_start_sync(file, argp);
3452         case BTRFS_IOC_WAIT_SYNC:
3453                 return btrfs_ioctl_wait_sync(file, argp);
3454         case BTRFS_IOC_SCRUB:
3455                 return btrfs_ioctl_scrub(root, argp);
3456         case BTRFS_IOC_SCRUB_CANCEL:
3457                 return btrfs_ioctl_scrub_cancel(root, argp);
3458         case BTRFS_IOC_SCRUB_PROGRESS:
3459                 return btrfs_ioctl_scrub_progress(root, argp);
3460         case BTRFS_IOC_BALANCE_V2:
3461                 return btrfs_ioctl_balance(file, argp);
3462         case BTRFS_IOC_BALANCE_CTL:
3463                 return btrfs_ioctl_balance_ctl(root, arg);
3464         case BTRFS_IOC_BALANCE_PROGRESS:
3465                 return btrfs_ioctl_balance_progress(root, argp);
3466         case BTRFS_IOC_GET_DEV_STATS:
3467                 return btrfs_ioctl_get_dev_stats(root, argp, 0);
3468         case BTRFS_IOC_GET_AND_RESET_DEV_STATS:
3469                 return btrfs_ioctl_get_dev_stats(root, argp, 1);
3470         }
3471
3472         return -ENOTTY;
3473 }