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 static struct extent_map *defrag_lookup_extent(struct inode *inode, u64 start)
 790 {
 791         struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
 792         struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
 793         struct extent_map *em;
 794         u64 len = PAGE_CACHE_SIZE;
 795
 796         /*
 797          * hopefully we have this extent in the tree already, try without
 798          * the full extent lock
 799          */
 800         read_lock(&em_tree->lock);
 801         em = lookup_extent_mapping(em_tree, start, len);
 802         read_unlock(&em_tree->lock);
 803
 804         if (!em) {
 805                 /* get the big lock and read metadata off disk */
 806                 lock_extent(io_tree, start, start + len - 1);
 807                 em = btrfs_get_extent(inode, NULL, 0, start, len, 0);
 808                 unlock_extent(io_tree, start, start + len - 1);
 809
 810                 if (IS_ERR(em))
 811                         return NULL;
 812         }
 813
 814         return em;
 815 }
 816
 817 static bool defrag_check_next_extent(struct inode *inode, struct extent_map *em)
 818 {
 819         struct extent_map *next;
 820         bool ret = true;
 821
 822         /* this is the last extent */
 823         if (em->start + em->len >= i_size_read(inode))
 824                 return false;
 825
 826         next = defrag_lookup_extent(inode, em->start + em->len);
 827         if (!next || next->block_start >= EXTENT_MAP_LAST_BYTE)
 828                 ret = false;
 829
 830         free_extent_map(next);
 831         return ret;
 832 }
 833
 834 static int should_defrag_range(struct inode *inode, u64 start, int thresh,
 835                                u64 *last_len, u64 *skip, u64 *defrag_end)
 836 {
 837         struct extent_map *em;
 838         int ret = 1;
 839         bool next_mergeable = true;
 840
 841         /*
 842          * make sure that once we start defragging an extent, we keep on
 843          * defragging it
 844          */
 845         if (start < *defrag_end)
 846                 return 1;
 847
 848         *skip = 0;
 849
 850         em = defrag_lookup_extent(inode, start);
 851         if (!em)
 852                 return 0;
 853
 854         /* this will cover holes, and inline extents */
 855         if (em->block_start >= EXTENT_MAP_LAST_BYTE) {
 856                 ret = 0;
 857                 goto out;
 858         }
 859
 860         next_mergeable = defrag_check_next_extent(inode, em);
 861
 862         /*
 863          * we hit a real extent, if it is big or the next extent is not a
 864          * real extent, don't bother defragging it
 865          */
 866         if ((*last_len == 0 || *last_len >= thresh) &&
 867             (em->len >= thresh || !next_mergeable))
 868                 ret = 0;
 869 out:
 870         /*
 871          * last_len ends up being a counter of how many bytes we've defragged.
 872          * every time we choose not to defrag an extent, we reset *last_len
 873          * so that the next tiny extent will force a defrag.
 874          *
 875          * The end result of this is that tiny extents before a single big
 876          * extent will force at least part of that big extent to be defragged.
 877          */
 878         if (ret) {
 879                 *defrag_end = extent_map_end(em);
 880         } else {
 881                 *last_len = 0;
 882                 *skip = extent_map_end(em);
 883                 *defrag_end = 0;
 884         }
 885
 886         free_extent_map(em);
 887         return ret;
 888 }
 889
 890 /*
 891  * it doesn't do much good to defrag one or two pages
 892  * at a time.  This pulls in a nice chunk of pages
 893  * to COW and defrag.
 894  *
 895  * It also makes sure the delalloc code has enough
 896  * dirty data to avoid making new small extents as part
 897  * of the defrag
 898  *
 899  * It's a good idea to start RA on this range
 900  * before calling this.
 901  */
 902 static int cluster_pages_for_defrag(struct inode *inode,
 903                                     struct page **pages,
 904                                     unsigned long start_index,
 905                                     int num_pages)
 906 {
 907         unsigned long file_end;
 908         u64 isize = i_size_read(inode);
 909         u64 page_start;
 910         u64 page_end;
 911         u64 page_cnt;
 912         int ret;
 913         int i;
 914         int i_done;
 915         struct btrfs_ordered_extent *ordered;
 916         struct extent_state *cached_state = NULL;
 917         struct extent_io_tree *tree;
 918         gfp_t mask = btrfs_alloc_write_mask(inode->i_mapping);
 919
 920         file_end = (isize - 1) >> PAGE_CACHE_SHIFT;
 921         if (!isize || start_index > file_end)
 922                 return 0;
 923
 924         page_cnt = min_t(u64, (u64)num_pages, (u64)file_end - start_index + 1);
 925
 926         ret = btrfs_delalloc_reserve_space(inode,
 927                                            page_cnt << PAGE_CACHE_SHIFT);
 928         if (ret)
 929                 return ret;
 930         i_done = 0;
 931         tree = &BTRFS_I(inode)->io_tree;
 932
 933         /* step one, lock all the pages */
 934         for (i = 0; i < page_cnt; i++) {
 935                 struct page *page;
 936 again:
 937                 page = find_or_create_page(inode->i_mapping,
 938                                            start_index + i, mask);
 939                 if (!page)
 940                         break;
 941
 942                 page_start = page_offset(page);
 943                 page_end = page_start + PAGE_CACHE_SIZE - 1;
 944                 while (1) {
 945                         lock_extent(tree, page_start, page_end);
 946                         ordered = btrfs_lookup_ordered_extent(inode,
 947                                                               page_start);
 948                         unlock_extent(tree, page_start, page_end);
 949                         if (!ordered)
 950                                 break;
 951
 952                         unlock_page(page);
 953                         btrfs_start_ordered_extent(inode, ordered, 1);
 954                         btrfs_put_ordered_extent(ordered);
 955                         lock_page(page);
 956                         /*
 957                          * we unlocked the page above, so we need check if
 958                          * it was released or not.
 959                          */
 960                         if (page->mapping != inode->i_mapping) {
 961                                 unlock_page(page);
 962                                 page_cache_release(page);
 963                                 goto again;
 964                         }
 965                 }
 966
 967                 if (!PageUptodate(page)) {
 968                         btrfs_readpage(NULL, page);
 969                         lock_page(page);
 970                         if (!PageUptodate(page)) {
 971                                 unlock_page(page);
 972                                 page_cache_release(page);
 973                                 ret = -EIO;
 974                                 break;
 975                         }
 976                 }
 977
 978                 if (page->mapping != inode->i_mapping) {
 979                         unlock_page(page);
 980                         page_cache_release(page);
 981                         goto again;
 982                 }
 983
 984                 pages[i] = page;
 985                 i_done++;
 986         }
 987         if (!i_done || ret)
 988                 goto out;
 989
 990         if (!(inode->i_sb->s_flags & MS_ACTIVE))
 991                 goto out;
 992
 993         /*
 994          * so now we have a nice long stream of locked
 995          * and up to date pages, lets wait on them
 996          */
 997         for (i = 0; i < i_done; i++)
 998                 wait_on_page_writeback(pages[i]);
 999
1000         page_start = page_offset(pages[0]);
1001         page_end = page_offset(pages[i_done - 1]) + PAGE_CACHE_SIZE;
1002
1003         lock_extent_bits(&BTRFS_I(inode)->io_tree,
1004                          page_start, page_end - 1, 0, &cached_state);
1005         clear_extent_bit(&BTRFS_I(inode)->io_tree, page_start,
1006                           page_end - 1, EXTENT_DIRTY | EXTENT_DELALLOC |
1007                           EXTENT_DO_ACCOUNTING, 0, 0, &cached_state,
1008                           GFP_NOFS);
1009
1010         if (i_done != page_cnt) {
1011                 spin_lock(&BTRFS_I(inode)->lock);
1012                 BTRFS_I(inode)->outstanding_extents++;
1013                 spin_unlock(&BTRFS_I(inode)->lock);
1014                 btrfs_delalloc_release_space(inode,
1015                                      (page_cnt - i_done) << PAGE_CACHE_SHIFT);
1016         }
1017
1018
1019         btrfs_set_extent_delalloc(inode, page_start, page_end - 1,
1020                                   &cached_state);
1021
1022         unlock_extent_cached(&BTRFS_I(inode)->io_tree,
1023                              page_start, page_end - 1, &cached_state,
1024                              GFP_NOFS);
1025
1026         for (i = 0; i < i_done; i++) {
1027                 clear_page_dirty_for_io(pages[i]);
1028                 ClearPageChecked(pages[i]);
1029                 set_page_extent_mapped(pages[i]);
1030                 set_page_dirty(pages[i]);
1031                 unlock_page(pages[i]);
1032                 page_cache_release(pages[i]);
1033         }
1034         return i_done;
1035 out:
1036         for (i = 0; i < i_done; i++) {
1037                 unlock_page(pages[i]);
1038                 page_cache_release(pages[i]);
1039         }
1040         btrfs_delalloc_release_space(inode, page_cnt << PAGE_CACHE_SHIFT);
1041         return ret;
1042
1043 }
1044
1045 int btrfs_defrag_file(struct inode *inode, struct file *file,
1046                       struct btrfs_ioctl_defrag_range_args *range,
1047                       u64 newer_than, unsigned long max_to_defrag)
1048 {
1049         struct btrfs_root *root = BTRFS_I(inode)->root;
1050         struct btrfs_super_block *disk_super;
1051         struct file_ra_state *ra = NULL;
1052         unsigned long last_index;
1053         u64 isize = i_size_read(inode);
1054         u64 features;
1055         u64 last_len = 0;
1056         u64 skip = 0;
1057         u64 defrag_end = 0;
1058         u64 newer_off = range->start;
1059         unsigned long i;
1060         unsigned long ra_index = 0;
1061         int ret;
1062         int defrag_count = 0;
1063         int compress_type = BTRFS_COMPRESS_ZLIB;
1064         int extent_thresh = range->extent_thresh;
1065         int max_cluster = (256 * 1024) >> PAGE_CACHE_SHIFT;
1066         int cluster = max_cluster;
1067         u64 new_align = ~((u64)128 * 1024 - 1);
1068         struct page **pages = NULL;
1069
1070         if (extent_thresh == 0)
1071                 extent_thresh = 256 * 1024;
1072
1073         if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS) {
1074                 if (range->compress_type > BTRFS_COMPRESS_TYPES)
1075                         return -EINVAL;
1076                 if (range->compress_type)
1077                         compress_type = range->compress_type;
1078         }
1079
1080         if (isize == 0)
1081                 return 0;
1082
1083         /*
1084          * if we were not given a file, allocate a readahead
1085          * context
1086          */
1087         if (!file) {
1088                 ra = kzalloc(sizeof(*ra), GFP_NOFS);
1089                 if (!ra)
1090                         return -ENOMEM;
1091                 file_ra_state_init(ra, inode->i_mapping);
1092         } else {
1093                 ra = &file->f_ra;
1094         }
1095
1096         pages = kmalloc(sizeof(struct page *) * max_cluster,
1097                         GFP_NOFS);
1098         if (!pages) {
1099                 ret = -ENOMEM;
1100                 goto out_ra;
1101         }
1102
1103         /* find the last page to defrag */
1104         if (range->start + range->len > range->start) {
1105                 last_index = min_t(u64, isize - 1,
1106                          range->start + range->len - 1) >> PAGE_CACHE_SHIFT;
1107         } else {
1108                 last_index = (isize - 1) >> PAGE_CACHE_SHIFT;
1109         }
1110
1111         if (newer_than) {
1112                 ret = find_new_extents(root, inode, newer_than,
1113                                        &newer_off, 64 * 1024);
1114                 if (!ret) {
1115                         range->start = newer_off;
1116                         /*
1117                          * we always align our defrag to help keep
1118                          * the extents in the file evenly spaced
1119                          */
1120                         i = (newer_off & new_align) >> PAGE_CACHE_SHIFT;
1121                 } else
1122                         goto out_ra;
1123         } else {
1124                 i = range->start >> PAGE_CACHE_SHIFT;
1125         }
1126         if (!max_to_defrag)
1127                 max_to_defrag = last_index + 1;
1128
1129         /*
1130          * make writeback starts from i, so the defrag range can be
1131          * written sequentially.
1132          */
1133         if (i < inode->i_mapping->writeback_index)
1134                 inode->i_mapping->writeback_index = i;
1135
1136         while (i <= last_index && defrag_count < max_to_defrag &&
1137                (i < (i_size_read(inode) + PAGE_CACHE_SIZE - 1) >>
1138                 PAGE_CACHE_SHIFT)) {
1139                 /*
1140                  * make sure we stop running if someone unmounts
1141                  * the FS
1142                  */
1143                 if (!(inode->i_sb->s_flags & MS_ACTIVE))
1144                         break;
1145
1146                 if (!should_defrag_range(inode, (u64)i << PAGE_CACHE_SHIFT,
1147                                          extent_thresh, &last_len, &skip,
1148                                          &defrag_end)) {
1149                         unsigned long next;
1150                         /*
1151                          * the should_defrag function tells us how much to skip
1152                          * bump our counter by the suggested amount
1153                          */
1154                         next = (skip + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT;
1155                         i = max(i + 1, next);
1156                         continue;
1157                 }
1158
1159                 if (!newer_than) {
1160                         cluster = (PAGE_CACHE_ALIGN(defrag_end) >>
1161                                    PAGE_CACHE_SHIFT) - i;
1162                         cluster = min(cluster, max_cluster);
1163                 } else {
1164                         cluster = max_cluster;
1165                 }
1166
1167                 if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)
1168                         BTRFS_I(inode)->force_compress = compress_type;
1169
1170                 if (i + cluster > ra_index) {
1171                         ra_index = max(i, ra_index);
1172                         btrfs_force_ra(inode->i_mapping, ra, file, ra_index,
1173                                        cluster);
1174                         ra_index += max_cluster;
1175                 }
1176
1177                 mutex_lock(&inode->i_mutex);
1178                 ret = cluster_pages_for_defrag(inode, pages, i, cluster);
1179                 if (ret < 0) {
1180                         mutex_unlock(&inode->i_mutex);
1181                         goto out_ra;
1182                 }
1183
1184                 defrag_count += ret;
1185                 balance_dirty_pages_ratelimited_nr(inode->i_mapping, ret);
1186                 mutex_unlock(&inode->i_mutex);
1187
1188                 if (newer_than) {
1189                         if (newer_off == (u64)-1)
1190                                 break;
1191
1192                         if (ret > 0)
1193                                 i += ret;
1194
1195                         newer_off = max(newer_off + 1,
1196                                         (u64)i << PAGE_CACHE_SHIFT);
1197
1198                         ret = find_new_extents(root, inode,
1199                                                newer_than, &newer_off,
1200                                                64 * 1024);
1201                         if (!ret) {
1202                                 range->start = newer_off;
1203                                 i = (newer_off & new_align) >> PAGE_CACHE_SHIFT;
1204                         } else {
1205                                 break;
1206                         }
1207                 } else {
1208                         if (ret > 0) {
1209                                 i += ret;
1210                                 last_len += ret << PAGE_CACHE_SHIFT;
1211                         } else {
1212                                 i++;
1213                                 last_len = 0;
1214                         }
1215                 }
1216         }
1217
1218         if ((range->flags & BTRFS_DEFRAG_RANGE_START_IO))
1219                 filemap_flush(inode->i_mapping);
1220
1221         if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
1222                 /* the filemap_flush will queue IO into the worker threads, but
1223                  * we have to make sure the IO is actually started and that
1224                  * ordered extents get created before we return
1225                  */
1226                 atomic_inc(&root->fs_info->async_submit_draining);
1227                 while (atomic_read(&root->fs_info->nr_async_submits) ||
1228                       atomic_read(&root->fs_info->async_delalloc_pages)) {
1229                         wait_event(root->fs_info->async_submit_wait,
1230                            (atomic_read(&root->fs_info->nr_async_submits) == 0 &&
1231                             atomic_read(&root->fs_info->async_delalloc_pages) == 0));
1232                 }
1233                 atomic_dec(&root->fs_info->async_submit_draining);
1234
1235                 mutex_lock(&inode->i_mutex);
1236                 BTRFS_I(inode)->force_compress = BTRFS_COMPRESS_NONE;
1237                 mutex_unlock(&inode->i_mutex);
1238         }
1239
1240         disk_super = root->fs_info->super_copy;
1241         features = btrfs_super_incompat_flags(disk_super);
1242         if (range->compress_type == BTRFS_COMPRESS_LZO) {
1243                 features |= BTRFS_FEATURE_INCOMPAT_COMPRESS_LZO;
1244                 btrfs_set_super_incompat_flags(disk_super, features);
1245         }
1246
1247         ret = defrag_count;
1248
1249 out_ra:
1250         if (!file)
1251                 kfree(ra);
1252         kfree(pages);
1253         return ret;
1254 }
1255
1256 static noinline int btrfs_ioctl_resize(struct btrfs_root *root,
1257                                         void __user *arg)
1258 {
1259         u64 new_size;
1260         u64 old_size;
1261         u64 devid = 1;
1262         struct btrfs_ioctl_vol_args *vol_args;
1263         struct btrfs_trans_handle *trans;
1264         struct btrfs_device *device = NULL;
1265         char *sizestr;
1266         char *devstr = NULL;
1267         int ret = 0;
1268         int mod = 0;
1269
1270         if (root->fs_info->sb->s_flags & MS_RDONLY)
1271                 return -EROFS;
1272
1273         if (!capable(CAP_SYS_ADMIN))
1274                 return -EPERM;
1275
1276         mutex_lock(&root->fs_info->volume_mutex);
1277         if (root->fs_info->balance_ctl) {
1278                 printk(KERN_INFO "btrfs: balance in progress\n");
1279                 ret = -EINVAL;
1280                 goto out;
1281         }
1282
1283         vol_args = memdup_user(arg, sizeof(*vol_args));
1284         if (IS_ERR(vol_args)) {
1285                 ret = PTR_ERR(vol_args);
1286                 goto out;
1287         }
1288
1289         vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1290
1291         sizestr = vol_args->name;
1292         devstr = strchr(sizestr, ':');
1293         if (devstr) {
1294                 char *end;
1295                 sizestr = devstr + 1;
1296                 *devstr = '\0';
1297                 devstr = vol_args->name;
1298                 devid = simple_strtoull(devstr, &end, 10);
1299                 printk(KERN_INFO "btrfs: resizing devid %llu\n",
1300                        (unsigned long long)devid);
1301         }
1302         device = btrfs_find_device(root, devid, NULL, NULL);
1303         if (!device) {
1304                 printk(KERN_INFO "btrfs: resizer unable to find device %llu\n",
1305                        (unsigned long long)devid);
1306                 ret = -EINVAL;
1307                 goto out_free;
1308         }
1309         if (!strcmp(sizestr, "max"))
1310                 new_size = device->bdev->bd_inode->i_size;
1311         else {
1312                 if (sizestr[0] == '-') {
1313                         mod = -1;
1314                         sizestr++;
1315                 } else if (sizestr[0] == '+') {
1316                         mod = 1;
1317                         sizestr++;
1318                 }
1319                 new_size = memparse(sizestr, NULL);
1320                 if (new_size == 0) {
1321                         ret = -EINVAL;
1322                         goto out_free;
1323                 }
1324         }
1325
1326         old_size = device->total_bytes;
1327
1328         if (mod < 0) {
1329                 if (new_size > old_size) {
1330                         ret = -EINVAL;
1331                         goto out_free;
1332                 }
1333                 new_size = old_size - new_size;
1334         } else if (mod > 0) {
1335                 new_size = old_size + new_size;
1336         }
1337
1338         if (new_size < 256 * 1024 * 1024) {
1339                 ret = -EINVAL;
1340                 goto out_free;
1341         }
1342         if (new_size > device->bdev->bd_inode->i_size) {
1343                 ret = -EFBIG;
1344                 goto out_free;
1345         }
1346
1347         do_div(new_size, root->sectorsize);
1348         new_size *= root->sectorsize;
1349
1350         printk_in_rcu(KERN_INFO "btrfs: new size for %s is %llu\n",
1351                       rcu_str_deref(device->name),
1352                       (unsigned long long)new_size);
1353
1354         if (new_size > old_size) {
1355                 trans = btrfs_start_transaction(root, 0);
1356                 if (IS_ERR(trans)) {
1357                         ret = PTR_ERR(trans);
1358                         goto out_free;
1359                 }
1360                 ret = btrfs_grow_device(trans, device, new_size);
1361                 btrfs_commit_transaction(trans, root);
1362         } else if (new_size < old_size) {
1363                 ret = btrfs_shrink_device(device, new_size);
1364         }
1365
1366 out_free:
1367         kfree(vol_args);
1368 out:
1369         mutex_unlock(&root->fs_info->volume_mutex);
1370         return ret;
1371 }
1372
1373 static noinline int btrfs_ioctl_snap_create_transid(struct file *file,
1374                                                     char *name,
1375                                                     unsigned long fd,
1376                                                     int subvol,
1377                                                     u64 *transid,
1378                                                     bool readonly)
1379 {
1380         struct btrfs_root *root = BTRFS_I(fdentry(file)->d_inode)->root;
1381         struct file *src_file;
1382         int namelen;
1383         int ret = 0;
1384
1385         if (root->fs_info->sb->s_flags & MS_RDONLY)
1386                 return -EROFS;
1387
1388         namelen = strlen(name);
1389         if (strchr(name, '/')) {
1390                 ret = -EINVAL;
1391                 goto out;
1392         }
1393
1394         if (name[0] == '.' &&
1395            (namelen == 1 || (name[1] == '.' && namelen == 2))) {
1396                 ret = -EEXIST;
1397                 goto out;
1398         }
1399
1400         if (subvol) {
1401                 ret = btrfs_mksubvol(&file->f_path, name, namelen,
1402                                      NULL, transid, readonly);
1403         } else {
1404                 struct inode *src_inode;
1405                 src_file = fget(fd);
1406                 if (!src_file) {
1407                         ret = -EINVAL;
1408                         goto out;
1409                 }
1410
1411                 src_inode = src_file->f_path.dentry->d_inode;
1412                 if (src_inode->i_sb != file->f_path.dentry->d_inode->i_sb) {
1413                         printk(KERN_INFO "btrfs: Snapshot src from "
1414                                "another FS\n");
1415                         ret = -EINVAL;
1416                         fput(src_file);
1417                         goto out;
1418                 }
1419                 ret = btrfs_mksubvol(&file->f_path, name, namelen,
1420                                      BTRFS_I(src_inode)->root,
1421                                      transid, readonly);
1422                 fput(src_file);
1423         }
1424 out:
1425         return ret;
1426 }
1427
1428 static noinline int btrfs_ioctl_snap_create(struct file *file,
1429                                             void __user *arg, int subvol)
1430 {
1431         struct btrfs_ioctl_vol_args *vol_args;
1432         int ret;
1433
1434         vol_args = memdup_user(arg, sizeof(*vol_args));
1435         if (IS_ERR(vol_args))
1436                 return PTR_ERR(vol_args);
1437         vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1438
1439         ret = btrfs_ioctl_snap_create_transid(file, vol_args->name,
1440                                               vol_args->fd, subvol,
1441                                               NULL, false);
1442
1443         kfree(vol_args);
1444         return ret;
1445 }
1446
1447 static noinline int btrfs_ioctl_snap_create_v2(struct file *file,
1448                                                void __user *arg, int subvol)
1449 {
1450         struct btrfs_ioctl_vol_args_v2 *vol_args;
1451         int ret;
1452         u64 transid = 0;
1453         u64 *ptr = NULL;
1454         bool readonly = false;
1455
1456         vol_args = memdup_user(arg, sizeof(*vol_args));
1457         if (IS_ERR(vol_args))
1458                 return PTR_ERR(vol_args);
1459         vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0';
1460
1461         if (vol_args->flags &
1462             ~(BTRFS_SUBVOL_CREATE_ASYNC | BTRFS_SUBVOL_RDONLY)) {
1463                 ret = -EOPNOTSUPP;
1464                 goto out;
1465         }
1466
1467         if (vol_args->flags & BTRFS_SUBVOL_CREATE_ASYNC)
1468                 ptr = &transid;
1469         if (vol_args->flags & BTRFS_SUBVOL_RDONLY)
1470                 readonly = true;
1471
1472         ret = btrfs_ioctl_snap_create_transid(file, vol_args->name,
1473                                               vol_args->fd, subvol,
1474                                               ptr, readonly);
1475
1476         if (ret == 0 && ptr &&
1477             copy_to_user(arg +
1478                          offsetof(struct btrfs_ioctl_vol_args_v2,
1479                                   transid), ptr, sizeof(*ptr)))
1480                 ret = -EFAULT;
1481 out:
1482         kfree(vol_args);
1483         return ret;
1484 }
1485
1486 static noinline int btrfs_ioctl_subvol_getflags(struct file *file,
1487                                                 void __user *arg)
1488 {
1489         struct inode *inode = fdentry(file)->d_inode;
1490         struct btrfs_root *root = BTRFS_I(inode)->root;
1491         int ret = 0;
1492         u64 flags = 0;
1493
1494         if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID)
1495                 return -EINVAL;
1496
1497         down_read(&root->fs_info->subvol_sem);
1498         if (btrfs_root_readonly(root))
1499                 flags |= BTRFS_SUBVOL_RDONLY;
1500         up_read(&root->fs_info->subvol_sem);
1501
1502         if (copy_to_user(arg, &flags, sizeof(flags)))
1503                 ret = -EFAULT;
1504
1505         return ret;
1506 }
1507
1508 static noinline int btrfs_ioctl_subvol_setflags(struct file *file,
1509                                               void __user *arg)
1510 {
1511         struct inode *inode = fdentry(file)->d_inode;
1512         struct btrfs_root *root = BTRFS_I(inode)->root;
1513         struct btrfs_trans_handle *trans;
1514         u64 root_flags;
1515         u64 flags;
1516         int ret = 0;
1517
1518         if (root->fs_info->sb->s_flags & MS_RDONLY)
1519                 return -EROFS;
1520
1521         if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID)
1522                 return -EINVAL;
1523
1524         if (copy_from_user(&flags, arg, sizeof(flags)))
1525                 return -EFAULT;
1526
1527         if (flags & BTRFS_SUBVOL_CREATE_ASYNC)
1528                 return -EINVAL;
1529
1530         if (flags & ~BTRFS_SUBVOL_RDONLY)
1531                 return -EOPNOTSUPP;
1532
1533         if (!inode_owner_or_capable(inode))
1534                 return -EACCES;
1535
1536         down_write(&root->fs_info->subvol_sem);
1537
1538         /* nothing to do */
1539         if (!!(flags & BTRFS_SUBVOL_RDONLY) == btrfs_root_readonly(root))
1540                 goto out;
1541
1542         root_flags = btrfs_root_flags(&root->root_item);
1543         if (flags & BTRFS_SUBVOL_RDONLY)
1544                 btrfs_set_root_flags(&root->root_item,
1545                                      root_flags | BTRFS_ROOT_SUBVOL_RDONLY);
1546         else
1547                 btrfs_set_root_flags(&root->root_item,
1548                                      root_flags & ~BTRFS_ROOT_SUBVOL_RDONLY);
1549
1550         trans = btrfs_start_transaction(root, 1);
1551         if (IS_ERR(trans)) {
1552                 ret = PTR_ERR(trans);
1553                 goto out_reset;
1554         }
1555
1556         ret = btrfs_update_root(trans, root->fs_info->tree_root,
1557                                 &root->root_key, &root->root_item);
1558
1559         btrfs_commit_transaction(trans, root);
1560 out_reset:
1561         if (ret)
1562                 btrfs_set_root_flags(&root->root_item, root_flags);
1563 out:
1564         up_write(&root->fs_info->subvol_sem);
1565         return ret;
1566 }
1567
1568 /*
1569  * helper to check if the subvolume references other subvolumes
1570  */
1571 static noinline int may_destroy_subvol(struct btrfs_root *root)
1572 {
1573         struct btrfs_path *path;
1574         struct btrfs_key key;
1575         int ret;
1576
1577         path = btrfs_alloc_path();
1578         if (!path)
1579                 return -ENOMEM;
1580
1581         key.objectid = root->root_key.objectid;
1582         key.type = BTRFS_ROOT_REF_KEY;
1583         key.offset = (u64)-1;
1584
1585         ret = btrfs_search_slot(NULL, root->fs_info->tree_root,
1586                                 &key, path, 0, 0);
1587         if (ret < 0)
1588                 goto out;
1589         BUG_ON(ret == 0);
1590
1591         ret = 0;
1592         if (path->slots[0] > 0) {
1593                 path->slots[0]--;
1594                 btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
1595                 if (key.objectid == root->root_key.objectid &&
1596                     key.type == BTRFS_ROOT_REF_KEY)
1597                         ret = -ENOTEMPTY;
1598         }
1599 out:
1600         btrfs_free_path(path);
1601         return ret;
1602 }
1603
1604 static noinline int key_in_sk(struct btrfs_key *key,
1605                               struct btrfs_ioctl_search_key *sk)
1606 {
1607         struct btrfs_key test;
1608         int ret;
1609
1610         test.objectid = sk->min_objectid;
1611         test.type = sk->min_type;
1612         test.offset = sk->min_offset;
1613
1614         ret = btrfs_comp_cpu_keys(key, &test);
1615         if (ret < 0)
1616                 return 0;
1617
1618         test.objectid = sk->max_objectid;
1619         test.type = sk->max_type;
1620         test.offset = sk->max_offset;
1621
1622         ret = btrfs_comp_cpu_keys(key, &test);
1623         if (ret > 0)
1624                 return 0;
1625         return 1;
1626 }
1627
1628 static noinline int copy_to_sk(struct btrfs_root *root,
1629                                struct btrfs_path *path,
1630                                struct btrfs_key *key,
1631                                struct btrfs_ioctl_search_key *sk,
1632                                char *buf,
1633                                unsigned long *sk_offset,
1634                                int *num_found)
1635 {
1636         u64 found_transid;
1637         struct extent_buffer *leaf;
1638         struct btrfs_ioctl_search_header sh;
1639         unsigned long item_off;
1640         unsigned long item_len;
1641         int nritems;
1642         int i;
1643         int slot;
1644         int ret = 0;
1645
1646         leaf = path->nodes[0];
1647         slot = path->slots[0];
1648         nritems = btrfs_header_nritems(leaf);
1649
1650         if (btrfs_header_generation(leaf) > sk->max_transid) {
1651                 i = nritems;
1652                 goto advance_key;
1653         }
1654         found_transid = btrfs_header_generation(leaf);
1655
1656         for (i = slot; i < nritems; i++) {
1657                 item_off = btrfs_item_ptr_offset(leaf, i);
1658                 item_len = btrfs_item_size_nr(leaf, i);
1659
1660                 if (item_len > BTRFS_SEARCH_ARGS_BUFSIZE)
1661                         item_len = 0;
1662
1663                 if (sizeof(sh) + item_len + *sk_offset >
1664                     BTRFS_SEARCH_ARGS_BUFSIZE) {
1665                         ret = 1;
1666                         goto overflow;
1667                 }
1668
1669                 btrfs_item_key_to_cpu(leaf, key, i);
1670                 if (!key_in_sk(key, sk))
1671                         continue;
1672
1673                 sh.objectid = key->objectid;
1674                 sh.offset = key->offset;
1675                 sh.type = key->type;
1676                 sh.len = item_len;
1677                 sh.transid = found_transid;
1678
1679                 /* copy search result header */
1680                 memcpy(buf + *sk_offset, &sh, sizeof(sh));
1681                 *sk_offset += sizeof(sh);
1682
1683                 if (item_len) {
1684                         char *p = buf + *sk_offset;
1685                         /* copy the item */
1686                         read_extent_buffer(leaf, p,
1687                                            item_off, item_len);
1688                         *sk_offset += item_len;
1689                 }
1690                 (*num_found)++;
1691
1692                 if (*num_found >= sk->nr_items)
1693                         break;
1694         }
1695 advance_key:
1696         ret = 0;
1697         if (key->offset < (u64)-1 && key->offset < sk->max_offset)
1698                 key->offset++;
1699         else if (key->type < (u8)-1 && key->type < sk->max_type) {
1700                 key->offset = 0;
1701                 key->type++;
1702         } else if (key->objectid < (u64)-1 && key->objectid < sk->max_objectid) {
1703                 key->offset = 0;
1704                 key->type = 0;
1705                 key->objectid++;
1706         } else
1707                 ret = 1;
1708 overflow:
1709         return ret;
1710 }
1711
1712 static noinline int search_ioctl(struct inode *inode,
1713                                  struct btrfs_ioctl_search_args *args)
1714 {
1715         struct btrfs_root *root;
1716         struct btrfs_key key;
1717         struct btrfs_key max_key;
1718         struct btrfs_path *path;
1719         struct btrfs_ioctl_search_key *sk = &args->key;
1720         struct btrfs_fs_info *info = BTRFS_I(inode)->root->fs_info;
1721         int ret;
1722         int num_found = 0;
1723         unsigned long sk_offset = 0;
1724
1725         path = btrfs_alloc_path();
1726         if (!path)
1727                 return -ENOMEM;
1728
1729         if (sk->tree_id == 0) {
1730                 /* search the root of the inode that was passed */
1731                 root = BTRFS_I(inode)->root;
1732         } else {
1733                 key.objectid = sk->tree_id;
1734                 key.type = BTRFS_ROOT_ITEM_KEY;
1735                 key.offset = (u64)-1;
1736                 root = btrfs_read_fs_root_no_name(info, &key);
1737                 if (IS_ERR(root)) {
1738                         printk(KERN_ERR "could not find root %llu\n",
1739                                sk->tree_id);
1740                         btrfs_free_path(path);
1741                         return -ENOENT;
1742                 }
1743         }
1744
1745         key.objectid = sk->min_objectid;
1746         key.type = sk->min_type;
1747         key.offset = sk->min_offset;
1748
1749         max_key.objectid = sk->max_objectid;
1750         max_key.type = sk->max_type;
1751         max_key.offset = sk->max_offset;
1752
1753         path->keep_locks = 1;
1754
1755         while(1) {
1756                 ret = btrfs_search_forward(root, &key, &max_key, path, 0,
1757                                            sk->min_transid);
1758                 if (ret != 0) {
1759                         if (ret > 0)
1760                                 ret = 0;
1761                         goto err;
1762                 }
1763                 ret = copy_to_sk(root, path, &key, sk, args->buf,
1764                                  &sk_offset, &num_found);
1765                 btrfs_release_path(path);
1766                 if (ret || num_found >= sk->nr_items)
1767                         break;
1768
1769         }
1770         ret = 0;
1771 err:
1772         sk->nr_items = num_found;
1773         btrfs_free_path(path);
1774         return ret;
1775 }
1776
1777 static noinline int btrfs_ioctl_tree_search(struct file *file,
1778                                            void __user *argp)
1779 {
1780          struct btrfs_ioctl_search_args *args;
1781          struct inode *inode;
1782          int ret;
1783
1784         if (!capable(CAP_SYS_ADMIN))
1785                 return -EPERM;
1786
1787         args = memdup_user(argp, sizeof(*args));
1788         if (IS_ERR(args))
1789                 return PTR_ERR(args);
1790
1791         inode = fdentry(file)->d_inode;
1792         ret = search_ioctl(inode, args);
1793         if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
1794                 ret = -EFAULT;
1795         kfree(args);
1796         return ret;
1797 }
1798
1799 /*
1800  * Search INODE_REFs to identify path name of 'dirid' directory
1801  * in a 'tree_id' tree. and sets path name to 'name'.
1802  */
1803 static noinline int btrfs_search_path_in_tree(struct btrfs_fs_info *info,
1804                                 u64 tree_id, u64 dirid, char *name)
1805 {
1806         struct btrfs_root *root;
1807         struct btrfs_key key;
1808         char *ptr;
1809         int ret = -1;
1810         int slot;
1811         int len;
1812         int total_len = 0;
1813         struct btrfs_inode_ref *iref;
1814         struct extent_buffer *l;
1815         struct btrfs_path *path;
1816
1817         if (dirid == BTRFS_FIRST_FREE_OBJECTID) {
1818                 name[0]='\0';
1819                 return 0;
1820         }
1821
1822         path = btrfs_alloc_path();
1823         if (!path)
1824                 return -ENOMEM;
1825
1826         ptr = &name[BTRFS_INO_LOOKUP_PATH_MAX];
1827
1828         key.objectid = tree_id;
1829         key.type = BTRFS_ROOT_ITEM_KEY;
1830         key.offset = (u64)-1;
1831         root = btrfs_read_fs_root_no_name(info, &key);
1832         if (IS_ERR(root)) {
1833                 printk(KERN_ERR "could not find root %llu\n", tree_id);
1834                 ret = -ENOENT;
1835                 goto out;
1836         }
1837
1838         key.objectid = dirid;
1839         key.type = BTRFS_INODE_REF_KEY;
1840         key.offset = (u64)-1;
1841
1842         while(1) {
1843                 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
1844                 if (ret < 0)
1845                         goto out;
1846
1847                 l = path->nodes[0];
1848                 slot = path->slots[0];
1849                 if (ret > 0 && slot > 0)
1850                         slot--;
1851                 btrfs_item_key_to_cpu(l, &key, slot);
1852
1853                 if (ret > 0 && (key.objectid != dirid ||
1854                                 key.type != BTRFS_INODE_REF_KEY)) {
1855                         ret = -ENOENT;
1856                         goto out;
1857                 }
1858
1859                 iref = btrfs_item_ptr(l, slot, struct btrfs_inode_ref);
1860                 len = btrfs_inode_ref_name_len(l, iref);
1861                 ptr -= len + 1;
1862                 total_len += len + 1;
1863                 if (ptr < name)
1864                         goto out;
1865
1866                 *(ptr + len) = '/';
1867                 read_extent_buffer(l, ptr,(unsigned long)(iref + 1), len);
1868
1869                 if (key.offset == BTRFS_FIRST_FREE_OBJECTID)
1870                         break;
1871
1872                 btrfs_release_path(path);
1873                 key.objectid = key.offset;
1874                 key.offset = (u64)-1;
1875                 dirid = key.objectid;
1876         }
1877         if (ptr < name)
1878                 goto out;
1879         memmove(name, ptr, total_len);
1880         name[total_len]='\0';
1881         ret = 0;
1882 out:
1883         btrfs_free_path(path);
1884         return ret;
1885 }
1886
1887 static noinline int btrfs_ioctl_ino_lookup(struct file *file,
1888                                            void __user *argp)
1889 {
1890          struct btrfs_ioctl_ino_lookup_args *args;
1891          struct inode *inode;
1892          int ret;
1893
1894         if (!capable(CAP_SYS_ADMIN))
1895                 return -EPERM;
1896
1897         args = memdup_user(argp, sizeof(*args));
1898         if (IS_ERR(args))
1899                 return PTR_ERR(args);
1900
1901         inode = fdentry(file)->d_inode;
1902
1903         if (args->treeid == 0)
1904                 args->treeid = BTRFS_I(inode)->root->root_key.objectid;
1905
1906         ret = btrfs_search_path_in_tree(BTRFS_I(inode)->root->fs_info,
1907                                         args->treeid, args->objectid,
1908                                         args->name);
1909
1910         if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
1911                 ret = -EFAULT;
1912
1913         kfree(args);
1914         return ret;
1915 }
1916
1917 static noinline int btrfs_ioctl_snap_destroy(struct file *file,
1918                                              void __user *arg)
1919 {
1920         struct dentry *parent = fdentry(file);
1921         struct dentry *dentry;
1922         struct inode *dir = parent->d_inode;
1923         struct inode *inode;
1924         struct btrfs_root *root = BTRFS_I(dir)->root;
1925         struct btrfs_root *dest = NULL;
1926         struct btrfs_ioctl_vol_args *vol_args;
1927         struct btrfs_trans_handle *trans;
1928         int namelen;
1929         int ret;
1930         int err = 0;
1931
1932         vol_args = memdup_user(arg, sizeof(*vol_args));
1933         if (IS_ERR(vol_args))
1934                 return PTR_ERR(vol_args);
1935
1936         vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1937         namelen = strlen(vol_args->name);
1938         if (strchr(vol_args->name, '/') ||
1939             strncmp(vol_args->name, "..", namelen) == 0) {
1940                 err = -EINVAL;
1941                 goto out;
1942         }
1943
1944         err = mnt_want_write_file(file);
1945         if (err)
1946                 goto out;
1947
1948         mutex_lock_nested(&dir->i_mutex, I_MUTEX_PARENT);
1949         dentry = lookup_one_len(vol_args->name, parent, namelen);
1950         if (IS_ERR(dentry)) {
1951                 err = PTR_ERR(dentry);
1952                 goto out_unlock_dir;
1953         }
1954
1955         if (!dentry->d_inode) {
1956                 err = -ENOENT;
1957                 goto out_dput;
1958         }
1959
1960         inode = dentry->d_inode;
1961         dest = BTRFS_I(inode)->root;
1962         if (!capable(CAP_SYS_ADMIN)){
1963                 /*
1964                  * Regular user.  Only allow this with a special mount
1965                  * option, when the user has write+exec access to the
1966                  * subvol root, and when rmdir(2) would have been
1967                  * allowed.
1968                  *
1969                  * Note that this is _not_ check that the subvol is
1970                  * empty or doesn't contain data that we wouldn't
1971                  * otherwise be able to delete.
1972                  *
1973                  * Users who want to delete empty subvols should try
1974                  * rmdir(2).
1975                  */
1976                 err = -EPERM;
1977                 if (!btrfs_test_opt(root, USER_SUBVOL_RM_ALLOWED))
1978                         goto out_dput;
1979
1980                 /*
1981                  * Do not allow deletion if the parent dir is the same
1982                  * as the dir to be deleted.  That means the ioctl
1983                  * must be called on the dentry referencing the root
1984                  * of the subvol, not a random directory contained
1985                  * within it.
1986                  */
1987                 err = -EINVAL;
1988                 if (root == dest)
1989                         goto out_dput;
1990
1991                 err = inode_permission(inode, MAY_WRITE | MAY_EXEC);
1992                 if (err)
1993                         goto out_dput;
1994
1995                 /* check if subvolume may be deleted by a non-root user */
1996                 err = btrfs_may_delete(dir, dentry, 1);
1997                 if (err)
1998                         goto out_dput;
1999         }
2000
2001         if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) {
2002                 err = -EINVAL;
2003                 goto out_dput;
2004         }
2005
2006         mutex_lock(&inode->i_mutex);
2007         err = d_invalidate(dentry);
2008         if (err)
2009                 goto out_unlock;
2010
2011         down_write(&root->fs_info->subvol_sem);
2012
2013         err = may_destroy_subvol(dest);
2014         if (err)
2015                 goto out_up_write;
2016
2017         trans = btrfs_start_transaction(root, 0);
2018         if (IS_ERR(trans)) {
2019                 err = PTR_ERR(trans);
2020                 goto out_up_write;
2021         }
2022         trans->block_rsv = &root->fs_info->global_block_rsv;
2023
2024         ret = btrfs_unlink_subvol(trans, root, dir,
2025                                 dest->root_key.objectid,
2026                                 dentry->d_name.name,
2027                                 dentry->d_name.len);
2028         if (ret) {
2029                 err = ret;
2030                 btrfs_abort_transaction(trans, root, ret);
2031                 goto out_end_trans;
2032         }
2033
2034         btrfs_record_root_in_trans(trans, dest);
2035
2036         memset(&dest->root_item.drop_progress, 0,
2037                 sizeof(dest->root_item.drop_progress));
2038         dest->root_item.drop_level = 0;
2039         btrfs_set_root_refs(&dest->root_item, 0);
2040
2041         if (!xchg(&dest->orphan_item_inserted, 1)) {
2042                 ret = btrfs_insert_orphan_item(trans,
2043                                         root->fs_info->tree_root,
2044                                         dest->root_key.objectid);
2045                 if (ret) {
2046                         btrfs_abort_transaction(trans, root, ret);
2047                         err = ret;
2048                         goto out_end_trans;
2049                 }
2050         }
2051 out_end_trans:
2052         ret = btrfs_end_transaction(trans, root);
2053         if (ret && !err)
2054                 err = ret;
2055         inode->i_flags |= S_DEAD;
2056 out_up_write:
2057         up_write(&root->fs_info->subvol_sem);
2058 out_unlock:
2059         mutex_unlock(&inode->i_mutex);
2060         if (!err) {
2061                 shrink_dcache_sb(root->fs_info->sb);
2062                 btrfs_invalidate_inodes(dest);
2063                 d_delete(dentry);
2064         }
2065 out_dput:
2066         dput(dentry);
2067 out_unlock_dir:
2068         mutex_unlock(&dir->i_mutex);
2069         mnt_drop_write_file(file);
2070 out:
2071         kfree(vol_args);
2072         return err;
2073 }
2074
2075 static int btrfs_ioctl_defrag(struct file *file, void __user *argp)
2076 {
2077         struct inode *inode = fdentry(file)->d_inode;
2078         struct btrfs_root *root = BTRFS_I(inode)->root;
2079         struct btrfs_ioctl_defrag_range_args *range;
2080         int ret;
2081
2082         if (btrfs_root_readonly(root))
2083                 return -EROFS;
2084
2085         ret = mnt_want_write_file(file);
2086         if (ret)
2087                 return ret;
2088
2089         switch (inode->i_mode & S_IFMT) {
2090         case S_IFDIR:
2091                 if (!capable(CAP_SYS_ADMIN)) {
2092                         ret = -EPERM;
2093                         goto out;
2094                 }
2095                 ret = btrfs_defrag_root(root, 0);
2096                 if (ret)
2097                         goto out;
2098                 ret = btrfs_defrag_root(root->fs_info->extent_root, 0);
2099                 break;
2100         case S_IFREG:
2101                 if (!(file->f_mode & FMODE_WRITE)) {
2102                         ret = -EINVAL;
2103                         goto out;
2104                 }
2105
2106                 range = kzalloc(sizeof(*range), GFP_KERNEL);
2107                 if (!range) {
2108                         ret = -ENOMEM;
2109                         goto out;
2110                 }
2111
2112                 if (argp) {
2113                         if (copy_from_user(range, argp,
2114                                            sizeof(*range))) {
2115                                 ret = -EFAULT;
2116                                 kfree(range);
2117                                 goto out;
2118                         }
2119                         /* compression requires us to start the IO */
2120                         if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
2121                                 range->flags |= BTRFS_DEFRAG_RANGE_START_IO;
2122                                 range->extent_thresh = (u32)-1;
2123                         }
2124                 } else {
2125                         /* the rest are all set to zero by kzalloc */
2126                         range->len = (u64)-1;
2127                 }
2128                 ret = btrfs_defrag_file(fdentry(file)->d_inode, file,
2129                                         range, 0, 0);
2130                 if (ret > 0)
2131                         ret = 0;
2132                 kfree(range);
2133                 break;
2134         default:
2135                 ret = -EINVAL;
2136         }
2137 out:
2138         mnt_drop_write_file(file);
2139         return ret;
2140 }
2141
2142 static long btrfs_ioctl_add_dev(struct btrfs_root *root, void __user *arg)
2143 {
2144         struct btrfs_ioctl_vol_args *vol_args;
2145         int ret;
2146
2147         if (!capable(CAP_SYS_ADMIN))
2148                 return -EPERM;
2149
2150         mutex_lock(&root->fs_info->volume_mutex);
2151         if (root->fs_info->balance_ctl) {
2152                 printk(KERN_INFO "btrfs: balance in progress\n");
2153                 ret = -EINVAL;
2154                 goto out;
2155         }
2156
2157         vol_args = memdup_user(arg, sizeof(*vol_args));
2158         if (IS_ERR(vol_args)) {
2159                 ret = PTR_ERR(vol_args);
2160                 goto out;
2161         }
2162
2163         vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2164         ret = btrfs_init_new_device(root, vol_args->name);
2165
2166         kfree(vol_args);
2167 out:
2168         mutex_unlock(&root->fs_info->volume_mutex);
2169         return ret;
2170 }
2171
2172 static long btrfs_ioctl_rm_dev(struct btrfs_root *root, void __user *arg)
2173 {
2174         struct btrfs_ioctl_vol_args *vol_args;
2175         int ret;
2176
2177         if (!capable(CAP_SYS_ADMIN))
2178                 return -EPERM;
2179
2180         if (root->fs_info->sb->s_flags & MS_RDONLY)
2181                 return -EROFS;
2182
2183         mutex_lock(&root->fs_info->volume_mutex);
2184         if (root->fs_info->balance_ctl) {
2185                 printk(KERN_INFO "btrfs: balance in progress\n");
2186                 ret = -EINVAL;
2187                 goto out;
2188         }
2189
2190         vol_args = memdup_user(arg, sizeof(*vol_args));
2191         if (IS_ERR(vol_args)) {
2192                 ret = PTR_ERR(vol_args);
2193                 goto out;
2194         }
2195
2196         vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2197         ret = btrfs_rm_device(root, vol_args->name);
2198
2199         kfree(vol_args);
2200 out:
2201         mutex_unlock(&root->fs_info->volume_mutex);
2202         return ret;
2203 }
2204
2205 static long btrfs_ioctl_fs_info(struct btrfs_root *root, void __user *arg)
2206 {
2207         struct btrfs_ioctl_fs_info_args *fi_args;
2208         struct btrfs_device *device;
2209         struct btrfs_device *next;
2210         struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
2211         int ret = 0;
2212
2213         if (!capable(CAP_SYS_ADMIN))
2214                 return -EPERM;
2215
2216         fi_args = kzalloc(sizeof(*fi_args), GFP_KERNEL);
2217         if (!fi_args)
2218                 return -ENOMEM;
2219
2220         fi_args->num_devices = fs_devices->num_devices;
2221         memcpy(&fi_args->fsid, root->fs_info->fsid, sizeof(fi_args->fsid));
2222
2223         mutex_lock(&fs_devices->device_list_mutex);
2224         list_for_each_entry_safe(device, next, &fs_devices->devices, dev_list) {
2225                 if (device->devid > fi_args->max_id)
2226                         fi_args->max_id = device->devid;
2227         }
2228         mutex_unlock(&fs_devices->device_list_mutex);
2229
2230         if (copy_to_user(arg, fi_args, sizeof(*fi_args)))
2231                 ret = -EFAULT;
2232
2233         kfree(fi_args);
2234         return ret;
2235 }
2236
2237 static long btrfs_ioctl_dev_info(struct btrfs_root *root, void __user *arg)
2238 {
2239         struct btrfs_ioctl_dev_info_args *di_args;
2240         struct btrfs_device *dev;
2241         struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
2242         int ret = 0;
2243         char *s_uuid = NULL;
2244         char empty_uuid[BTRFS_UUID_SIZE] = {0};
2245
2246         if (!capable(CAP_SYS_ADMIN))
2247                 return -EPERM;
2248
2249         di_args = memdup_user(arg, sizeof(*di_args));
2250         if (IS_ERR(di_args))
2251                 return PTR_ERR(di_args);
2252
2253         if (memcmp(empty_uuid, di_args->uuid, BTRFS_UUID_SIZE) != 0)
2254                 s_uuid = di_args->uuid;
2255
2256         mutex_lock(&fs_devices->device_list_mutex);
2257         dev = btrfs_find_device(root, di_args->devid, s_uuid, NULL);
2258         mutex_unlock(&fs_devices->device_list_mutex);
2259
2260         if (!dev) {
2261                 ret = -ENODEV;
2262                 goto out;
2263         }
2264
2265         di_args->devid = dev->devid;
2266         di_args->bytes_used = dev->bytes_used;
2267         di_args->total_bytes = dev->total_bytes;
2268         memcpy(di_args->uuid, dev->uuid, sizeof(di_args->uuid));
2269         if (dev->name) {
2270                 struct rcu_string *name;
2271
2272                 rcu_read_lock();
2273                 name = rcu_dereference(dev->name);
2274                 strncpy(di_args->path, name->str, sizeof(di_args->path));
2275                 rcu_read_unlock();
2276                 di_args->path[sizeof(di_args->path) - 1] = 0;
2277         } else {
2278                 di_args->path[0] = '\0';
2279         }
2280
2281 out:
2282         if (ret == 0 && copy_to_user(arg, di_args, sizeof(*di_args)))
2283                 ret = -EFAULT;
2284
2285         kfree(di_args);
2286         return ret;
2287 }
2288
2289 static noinline long btrfs_ioctl_clone(struct file *file, unsigned long srcfd,
2290                                        u64 off, u64 olen, u64 destoff)
2291 {
2292         struct inode *inode = fdentry(file)->d_inode;
2293         struct btrfs_root *root = BTRFS_I(inode)->root;
2294         struct file *src_file;
2295         struct inode *src;
2296         struct btrfs_trans_handle *trans;
2297         struct btrfs_path *path;
2298         struct extent_buffer *leaf;
2299         char *buf;
2300         struct btrfs_key key;
2301         u32 nritems;
2302         int slot;
2303         int ret;
2304         u64 len = olen;
2305         u64 bs = root->fs_info->sb->s_blocksize;
2306         u64 hint_byte;
2307
2308         /*
2309          * TODO:
2310          * - split compressed inline extents.  annoying: we need to
2311          *   decompress into destination's address_space (the file offset
2312          *   may change, so source mapping won't do), then recompress (or
2313          *   otherwise reinsert) a subrange.
2314          * - allow ranges within the same file to be cloned (provided
2315          *   they don't overlap)?
2316          */
2317
2318         /* the destination must be opened for writing */
2319         if (!(file->f_mode & FMODE_WRITE) || (file->f_flags & O_APPEND))
2320                 return -EINVAL;
2321
2322         if (btrfs_root_readonly(root))
2323                 return -EROFS;
2324
2325         ret = mnt_want_write_file(file);
2326         if (ret)
2327                 return ret;
2328
2329         src_file = fget(srcfd);
2330         if (!src_file) {
2331                 ret = -EBADF;
2332                 goto out_drop_write;
2333         }
2334
2335         src = src_file->f_dentry->d_inode;
2336
2337         ret = -EINVAL;
2338         if (src == inode)
2339                 goto out_fput;
2340
2341         /* the src must be open for reading */
2342         if (!(src_file->f_mode & FMODE_READ))
2343                 goto out_fput;
2344
2345         /* don't make the dst file partly checksummed */
2346         if ((BTRFS_I(src)->flags & BTRFS_INODE_NODATASUM) !=
2347             (BTRFS_I(inode)->flags & BTRFS_INODE_NODATASUM))
2348                 goto out_fput;
2349
2350         ret = -EISDIR;
2351         if (S_ISDIR(src->i_mode) || S_ISDIR(inode->i_mode))
2352                 goto out_fput;
2353
2354         ret = -EXDEV;
2355         if (src->i_sb != inode->i_sb || BTRFS_I(src)->root != root)
2356                 goto out_fput;
2357
2358         ret = -ENOMEM;
2359         buf = vmalloc(btrfs_level_size(root, 0));
2360         if (!buf)
2361                 goto out_fput;
2362
2363         path = btrfs_alloc_path();
2364         if (!path) {
2365                 vfree(buf);
2366                 goto out_fput;
2367         }
2368         path->reada = 2;
2369
2370         if (inode < src) {
2371                 mutex_lock_nested(&inode->i_mutex, I_MUTEX_PARENT);
2372                 mutex_lock_nested(&src->i_mutex, I_MUTEX_CHILD);
2373         } else {
2374                 mutex_lock_nested(&src->i_mutex, I_MUTEX_PARENT);
2375                 mutex_lock_nested(&inode->i_mutex, I_MUTEX_CHILD);
2376         }
2377
2378         /* determine range to clone */
2379         ret = -EINVAL;
2380         if (off + len > src->i_size || off + len < off)
2381                 goto out_unlock;
2382         if (len == 0)
2383                 olen = len = src->i_size - off;
2384         /* if we extend to eof, continue to block boundary */
2385         if (off + len == src->i_size)
2386                 len = ALIGN(src->i_size, bs) - off;
2387
2388         /* verify the end result is block aligned */
2389         if (!IS_ALIGNED(off, bs) || !IS_ALIGNED(off + len, bs) ||
2390             !IS_ALIGNED(destoff, bs))
2391                 goto out_unlock;
2392
2393         if (destoff > inode->i_size) {
2394                 ret = btrfs_cont_expand(inode, inode->i_size, destoff);
2395                 if (ret)
2396                         goto out_unlock;
2397         }
2398
2399         /* truncate page cache pages from target inode range */
2400         truncate_inode_pages_range(&inode->i_data, destoff,
2401                                    PAGE_CACHE_ALIGN(destoff + len) - 1);
2402
2403         /* do any pending delalloc/csum calc on src, one way or
2404            another, and lock file content */
2405         while (1) {
2406                 struct btrfs_ordered_extent *ordered;
2407                 lock_extent(&BTRFS_I(src)->io_tree, off, off+len);
2408                 ordered = btrfs_lookup_first_ordered_extent(src, off+len);
2409                 if (!ordered &&
2410                     !test_range_bit(&BTRFS_I(src)->io_tree, off, off+len,
2411                                    EXTENT_DELALLOC, 0, NULL))
2412                         break;
2413                 unlock_extent(&BTRFS_I(src)->io_tree, off, off+len);
2414                 if (ordered)
2415                         btrfs_put_ordered_extent(ordered);
2416                 btrfs_wait_ordered_range(src, off, len);
2417         }
2418
2419         /* clone data */
2420         key.objectid = btrfs_ino(src);
2421         key.type = BTRFS_EXTENT_DATA_KEY;
2422         key.offset = 0;
2423
2424         while (1) {
2425                 /*
2426                  * note the key will change type as we walk through the
2427                  * tree.
2428                  */
2429                 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
2430                 if (ret < 0)
2431                         goto out;
2432
2433                 nritems = btrfs_header_nritems(path->nodes[0]);
2434                 if (path->slots[0] >= nritems) {
2435                         ret = btrfs_next_leaf(root, path);
2436                         if (ret < 0)
2437                                 goto out;
2438                         if (ret > 0)
2439                                 break;
2440                         nritems = btrfs_header_nritems(path->nodes[0]);
2441                 }
2442                 leaf = path->nodes[0];
2443                 slot = path->slots[0];
2444
2445                 btrfs_item_key_to_cpu(leaf, &key, slot);
2446                 if (btrfs_key_type(&key) > BTRFS_EXTENT_DATA_KEY ||
2447                     key.objectid != btrfs_ino(src))
2448                         break;
2449
2450                 if (btrfs_key_type(&key) == BTRFS_EXTENT_DATA_KEY) {
2451                         struct btrfs_file_extent_item *extent;
2452                         int type;
2453                         u32 size;
2454                         struct btrfs_key new_key;
2455                         u64 disko = 0, diskl = 0;
2456                         u64 datao = 0, datal = 0;
2457                         u8 comp;
2458                         u64 endoff;
2459
2460                         size = btrfs_item_size_nr(leaf, slot);
2461                         read_extent_buffer(leaf, buf,
2462                                            btrfs_item_ptr_offset(leaf, slot),
2463                                            size);
2464
2465                         extent = btrfs_item_ptr(leaf, slot,
2466                                                 struct btrfs_file_extent_item);
2467                         comp = btrfs_file_extent_compression(leaf, extent);
2468                         type = btrfs_file_extent_type(leaf, extent);
2469                         if (type == BTRFS_FILE_EXTENT_REG ||
2470                             type == BTRFS_FILE_EXTENT_PREALLOC) {
2471                                 disko = btrfs_file_extent_disk_bytenr(leaf,
2472                                                                       extent);
2473                                 diskl = btrfs_file_extent_disk_num_bytes(leaf,
2474                                                                  extent);
2475                                 datao = btrfs_file_extent_offset(leaf, extent);
2476                                 datal = btrfs_file_extent_num_bytes(leaf,
2477                                                                     extent);
2478                         } else if (type == BTRFS_FILE_EXTENT_INLINE) {
2479                                 /* take upper bound, may be compressed */
2480                                 datal = btrfs_file_extent_ram_bytes(leaf,
2481                                                                     extent);
2482                         }
2483                         btrfs_release_path(path);
2484
2485                         if (key.offset + datal <= off ||
2486                             key.offset >= off+len)
2487                                 goto next;
2488
2489                         memcpy(&new_key, &key, sizeof(new_key));
2490                         new_key.objectid = btrfs_ino(inode);
2491                         if (off <= key.offset)
2492                                 new_key.offset = key.offset + destoff - off;
2493                         else
2494                                 new_key.offset = destoff;
2495
2496                         /*
2497                          * 1 - adjusting old extent (we may have to split it)
2498                          * 1 - add new extent
2499                          * 1 - inode update
2500                          */
2501                         trans = btrfs_start_transaction(root, 3);
2502                         if (IS_ERR(trans)) {
2503                                 ret = PTR_ERR(trans);
2504                                 goto out;
2505                         }
2506
2507                         if (type == BTRFS_FILE_EXTENT_REG ||
2508                             type == BTRFS_FILE_EXTENT_PREALLOC) {
2509                                 /*
2510                                  *    a  | --- range to clone ---|  b
2511                                  * | ------------- extent ------------- |
2512                                  */
2513
2514                                 /* substract range b */
2515                                 if (key.offset + datal > off + len)
2516                                         datal = off + len - key.offset;
2517
2518                                 /* substract range a */
2519                                 if (off > key.offset) {
2520                                         datao += off - key.offset;
2521                                         datal -= off - key.offset;
2522                                 }
2523
2524                                 ret = btrfs_drop_extents(trans, inode,
2525                                                          new_key.offset,
2526                                                          new_key.offset + datal,
2527                                                          &hint_byte, 1);
2528                                 if (ret) {
2529                                         btrfs_abort_transaction(trans, root,
2530                                                                 ret);
2531                                         btrfs_end_transaction(trans, root);
2532                                         goto out;
2533                                 }
2534
2535                                 ret = btrfs_insert_empty_item(trans, root, path,
2536                                                               &new_key, size);
2537                                 if (ret) {
2538                                         btrfs_abort_transaction(trans, root,
2539                                                                 ret);
2540                                         btrfs_end_transaction(trans, root);
2541                                         goto out;
2542                                 }
2543
2544                                 leaf = path->nodes[0];
2545                                 slot = path->slots[0];
2546                                 write_extent_buffer(leaf, buf,
2547                                             btrfs_item_ptr_offset(leaf, slot),
2548                                             size);
2549
2550                                 extent = btrfs_item_ptr(leaf, slot,
2551                                                 struct btrfs_file_extent_item);
2552
2553                                 /* disko == 0 means it's a hole */
2554                                 if (!disko)
2555                                         datao = 0;
2556
2557                                 btrfs_set_file_extent_offset(leaf, extent,
2558                                                              datao);
2559                                 btrfs_set_file_extent_num_bytes(leaf, extent,
2560                                                                 datal);
2561                                 if (disko) {
2562                                         inode_add_bytes(inode, datal);
2563                                         ret = btrfs_inc_extent_ref(trans, root,
2564                                                         disko, diskl, 0,
2565                                                         root->root_key.objectid,
2566                                                         btrfs_ino(inode),
2567                                                         new_key.offset - datao,
2568                                                         0);
2569                                         if (ret) {
2570                                                 btrfs_abort_transaction(trans,
2571                                                                         root,
2572                                                                         ret);
2573                                                 btrfs_end_transaction(trans,
2574                                                                       root);
2575                                                 goto out;
2576
2577                                         }
2578                                 }
2579                         } else if (type == BTRFS_FILE_EXTENT_INLINE) {
2580                                 u64 skip = 0;
2581                                 u64 trim = 0;
2582                                 if (off > key.offset) {
2583                                         skip = off - key.offset;
2584                                         new_key.offset += skip;
2585                                 }
2586
2587                                 if (key.offset + datal > off+len)
2588                                         trim = key.offset + datal - (off+len);
2589
2590                                 if (comp && (skip || trim)) {
2591                                         ret = -EINVAL;
2592                                         btrfs_end_transaction(trans, root);
2593                                         goto out;
2594                                 }
2595                                 size -= skip + trim;
2596                                 datal -= skip + trim;
2597
2598                                 ret = btrfs_drop_extents(trans, inode,
2599                                                          new_key.offset,
2600                                                          new_key.offset + datal,
2601                                                          &hint_byte, 1);
2602                                 if (ret) {
2603                                         btrfs_abort_transaction(trans, root,
2604                                                                 ret);
2605                                         btrfs_end_transaction(trans, root);
2606                                         goto out;
2607                                 }
2608
2609                                 ret = btrfs_insert_empty_item(trans, root, path,
2610                                                               &new_key, size);
2611                                 if (ret) {
2612                                         btrfs_abort_transaction(trans, root,
2613                                                                 ret);
2614                                         btrfs_end_transaction(trans, root);
2615                                         goto out;
2616                                 }
2617
2618                                 if (skip) {
2619                                         u32 start =
2620                                           btrfs_file_extent_calc_inline_size(0);
2621                                         memmove(buf+start, buf+start+skip,
2622                                                 datal);
2623                                 }
2624
2625                                 leaf = path->nodes[0];
2626                                 slot = path->slots[0];
2627                                 write_extent_buffer(leaf, buf,
2628                                             btrfs_item_ptr_offset(leaf, slot),
2629                                             size);
2630                                 inode_add_bytes(inode, datal);
2631                         }
2632
2633                         btrfs_mark_buffer_dirty(leaf);
2634                         btrfs_release_path(path);
2635
2636                         inode_inc_iversion(inode);
2637                         inode->i_mtime = inode->i_ctime = CURRENT_TIME;
2638
2639                         /*
2640                          * we round up to the block size at eof when
2641                          * determining which extents to clone above,
2642                          * but shouldn't round up the file size
2643                          */
2644                         endoff = new_key.offset + datal;
2645                         if (endoff > destoff+olen)
2646                                 endoff = destoff+olen;
2647                         if (endoff > inode->i_size)
2648                                 btrfs_i_size_write(inode, endoff);
2649
2650                         ret = btrfs_update_inode(trans, root, inode);
2651                         if (ret) {
2652                                 btrfs_abort_transaction(trans, root, ret);
2653                                 btrfs_end_transaction(trans, root);
2654                                 goto out;
2655                         }
2656                         ret = btrfs_end_transaction(trans, root);
2657                 }
2658 next:
2659                 btrfs_release_path(path);
2660                 key.offset++;
2661         }
2662         ret = 0;
2663 out:
2664         btrfs_release_path(path);
2665         unlock_extent(&BTRFS_I(src)->io_tree, off, off+len);
2666 out_unlock:
2667         mutex_unlock(&src->i_mutex);
2668         mutex_unlock(&inode->i_mutex);
2669         vfree(buf);
2670         btrfs_free_path(path);
2671 out_fput:
2672         fput(src_file);
2673 out_drop_write:
2674         mnt_drop_write_file(file);
2675         return ret;
2676 }
2677
2678 static long btrfs_ioctl_clone_range(struct file *file, void __user *argp)
2679 {
2680         struct btrfs_ioctl_clone_range_args args;
2681
2682         if (copy_from_user(&args, argp, sizeof(args)))
2683                 return -EFAULT;
2684         return btrfs_ioctl_clone(file, args.src_fd, args.src_offset,
2685                                  args.src_length, args.dest_offset);
2686 }
2687
2688 /*
2689  * there are many ways the trans_start and trans_end ioctls can lead
2690  * to deadlocks.  They should only be used by applications that
2691  * basically own the machine, and have a very in depth understanding
2692  * of all the possible deadlocks and enospc problems.
2693  */
2694 static long btrfs_ioctl_trans_start(struct file *file)
2695 {
2696         struct inode *inode = fdentry(file)->d_inode;
2697         struct btrfs_root *root = BTRFS_I(inode)->root;
2698         struct btrfs_trans_handle *trans;
2699         int ret;
2700
2701         ret = -EPERM;
2702         if (!capable(CAP_SYS_ADMIN))
2703                 goto out;
2704
2705         ret = -EINPROGRESS;
2706         if (file->private_data)
2707                 goto out;
2708
2709         ret = -EROFS;
2710         if (btrfs_root_readonly(root))
2711                 goto out;
2712
2713         ret = mnt_want_write_file(file);
2714         if (ret)
2715                 goto out;
2716
2717         atomic_inc(&root->fs_info->open_ioctl_trans);
2718
2719         ret = -ENOMEM;
2720         trans = btrfs_start_ioctl_transaction(root);
2721         if (IS_ERR(trans))
2722                 goto out_drop;
2723
2724         file->private_data = trans;
2725         return 0;
2726
2727 out_drop:
2728         atomic_dec(&root->fs_info->open_ioctl_trans);
2729         mnt_drop_write_file(file);
2730 out:
2731         return ret;
2732 }
2733
2734 static long btrfs_ioctl_default_subvol(struct file *file, void __user *argp)
2735 {
2736         struct inode *inode = fdentry(file)->d_inode;
2737         struct btrfs_root *root = BTRFS_I(inode)->root;
2738         struct btrfs_root *new_root;
2739         struct btrfs_dir_item *di;
2740         struct btrfs_trans_handle *trans;
2741         struct btrfs_path *path;
2742         struct btrfs_key location;
2743         struct btrfs_disk_key disk_key;
2744         struct btrfs_super_block *disk_super;
2745         u64 features;
2746         u64 objectid = 0;
2747         u64 dir_id;
2748
2749         if (!capable(CAP_SYS_ADMIN))
2750                 return -EPERM;
2751
2752         if (copy_from_user(&objectid, argp, sizeof(objectid)))
2753                 return -EFAULT;
2754
2755         if (!objectid)
2756                 objectid = root->root_key.objectid;
2757
2758         location.objectid = objectid;
2759         location.type = BTRFS_ROOT_ITEM_KEY;
2760         location.offset = (u64)-1;
2761
2762         new_root = btrfs_read_fs_root_no_name(root->fs_info, &location);
2763         if (IS_ERR(new_root))
2764                 return PTR_ERR(new_root);
2765
2766         if (btrfs_root_refs(&new_root->root_item) == 0)
2767                 return -ENOENT;
2768
2769         path = btrfs_alloc_path();
2770         if (!path)
2771                 return -ENOMEM;
2772         path->leave_spinning = 1;
2773
2774         trans = btrfs_start_transaction(root, 1);
2775         if (IS_ERR(trans)) {
2776                 btrfs_free_path(path);
2777                 return PTR_ERR(trans);
2778         }
2779
2780         dir_id = btrfs_super_root_dir(root->fs_info->super_copy);
2781         di = btrfs_lookup_dir_item(trans, root->fs_info->tree_root, path,
2782                                    dir_id, "default", 7, 1);
2783         if (IS_ERR_OR_NULL(di)) {
2784                 btrfs_free_path(path);
2785                 btrfs_end_transaction(trans, root);
2786                 printk(KERN_ERR "Umm, you don't have the default dir item, "
2787                        "this isn't going to work\n");
2788                 return -ENOENT;
2789         }
2790
2791         btrfs_cpu_key_to_disk(&disk_key, &new_root->root_key);
2792         btrfs_set_dir_item_key(path->nodes[0], di, &disk_key);
2793         btrfs_mark_buffer_dirty(path->nodes[0]);
2794         btrfs_free_path(path);
2795
2796         disk_super = root->fs_info->super_copy;
2797         features = btrfs_super_incompat_flags(disk_super);
2798         if (!(features & BTRFS_FEATURE_INCOMPAT_DEFAULT_SUBVOL)) {
2799                 features |= BTRFS_FEATURE_INCOMPAT_DEFAULT_SUBVOL;
2800                 btrfs_set_super_incompat_flags(disk_super, features);
2801         }
2802         btrfs_end_transaction(trans, root);
2803
2804         return 0;
2805 }
2806
2807 static void get_block_group_info(struct list_head *groups_list,
2808                                  struct btrfs_ioctl_space_info *space)
2809 {
2810         struct btrfs_block_group_cache *block_group;
2811
2812         space->total_bytes = 0;
2813         space->used_bytes = 0;
2814         space->flags = 0;
2815         list_for_each_entry(block_group, groups_list, list) {
2816                 space->flags = block_group->flags;
2817                 space->total_bytes += block_group->key.offset;
2818                 space->used_bytes +=
2819                         btrfs_block_group_used(&block_group->item);
2820         }
2821 }
2822
2823 long btrfs_ioctl_space_info(struct btrfs_root *root, void __user *arg)
2824 {
2825         struct btrfs_ioctl_space_args space_args;
2826         struct btrfs_ioctl_space_info space;
2827         struct btrfs_ioctl_space_info *dest;
2828         struct btrfs_ioctl_space_info *dest_orig;
2829         struct btrfs_ioctl_space_info __user *user_dest;
2830         struct btrfs_space_info *info;
2831         u64 types[] = {BTRFS_BLOCK_GROUP_DATA,
2832                        BTRFS_BLOCK_GROUP_SYSTEM,
2833                        BTRFS_BLOCK_GROUP_METADATA,
2834                        BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA};
2835         int num_types = 4;
2836         int alloc_size;
2837         int ret = 0;
2838         u64 slot_count = 0;
2839         int i, c;
2840
2841         if (copy_from_user(&space_args,
2842                            (struct btrfs_ioctl_space_args __user *)arg,
2843                            sizeof(space_args)))
2844                 return -EFAULT;
2845
2846         for (i = 0; i < num_types; i++) {
2847                 struct btrfs_space_info *tmp;
2848
2849                 info = NULL;
2850                 rcu_read_lock();
2851                 list_for_each_entry_rcu(tmp, &root->fs_info->space_info,
2852                                         list) {
2853                         if (tmp->flags == types[i]) {
2854                                 info = tmp;
2855                                 break;
2856                         }
2857                 }
2858                 rcu_read_unlock();
2859
2860                 if (!info)
2861                         continue;
2862
2863                 down_read(&info->groups_sem);
2864                 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
2865                         if (!list_empty(&info->block_groups[c]))
2866                                 slot_count++;
2867                 }
2868                 up_read(&info->groups_sem);
2869         }
2870
2871         /* space_slots == 0 means they are asking for a count */
2872         if (space_args.space_slots == 0) {
2873                 space_args.total_spaces = slot_count;
2874                 goto out;
2875         }
2876
2877         slot_count = min_t(u64, space_args.space_slots, slot_count);
2878
2879         alloc_size = sizeof(*dest) * slot_count;
2880
2881         /* we generally have at most 6 or so space infos, one for each raid
2882          * level.  So, a whole page should be more than enough for everyone
2883          */
2884         if (alloc_size > PAGE_CACHE_SIZE)
2885                 return -ENOMEM;
2886
2887         space_args.total_spaces = 0;
2888         dest = kmalloc(alloc_size, GFP_NOFS);
2889         if (!dest)
2890                 return -ENOMEM;
2891         dest_orig = dest;
2892
2893         /* now we have a buffer to copy into */
2894         for (i = 0; i < num_types; i++) {
2895                 struct btrfs_space_info *tmp;
2896
2897                 if (!slot_count)
2898                         break;
2899
2900                 info = NULL;
2901                 rcu_read_lock();
2902                 list_for_each_entry_rcu(tmp, &root->fs_info->space_info,
2903                                         list) {
2904                         if (tmp->flags == types[i]) {
2905                                 info = tmp;
2906                                 break;
2907                         }
2908                 }
2909                 rcu_read_unlock();
2910
2911                 if (!info)
2912                         continue;
2913                 down_read(&info->groups_sem);
2914                 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
2915                         if (!list_empty(&info->block_groups[c])) {
2916                                 get_block_group_info(&info->block_groups[c],
2917                                                      &space);
2918                                 memcpy(dest, &space, sizeof(space));
2919                                 dest++;
2920                                 space_args.total_spaces++;
2921                                 slot_count--;
2922                         }
2923                         if (!slot_count)
2924                                 break;
2925                 }
2926                 up_read(&info->groups_sem);
2927         }
2928
2929         user_dest = (struct btrfs_ioctl_space_info __user *)
2930                 (arg + sizeof(struct btrfs_ioctl_space_args));
2931
2932         if (copy_to_user(user_dest, dest_orig, alloc_size))
2933                 ret = -EFAULT;
2934
2935         kfree(dest_orig);
2936 out:
2937         if (ret == 0 && copy_to_user(arg, &space_args, sizeof(space_args)))
2938                 ret = -EFAULT;
2939
2940         return ret;
2941 }
2942
2943 /*
2944  * there are many ways the trans_start and trans_end ioctls can lead
2945  * to deadlocks.  They should only be used by applications that
2946  * basically own the machine, and have a very in depth understanding
2947  * of all the possible deadlocks and enospc problems.
2948  */
2949 long btrfs_ioctl_trans_end(struct file *file)
2950 {
2951         struct inode *inode = fdentry(file)->d_inode;
2952         struct btrfs_root *root = BTRFS_I(inode)->root;
2953         struct btrfs_trans_handle *trans;
2954
2955         trans = file->private_data;
2956         if (!trans)
2957                 return -EINVAL;
2958         file->private_data = NULL;
2959
2960         btrfs_end_transaction(trans, root);
2961
2962         atomic_dec(&root->fs_info->open_ioctl_trans);
2963
2964         mnt_drop_write_file(file);
2965         return 0;
2966 }
2967
2968 static noinline long btrfs_ioctl_start_sync(struct file *file, void __user *argp)
2969 {
2970         struct btrfs_root *root = BTRFS_I(file->f_dentry->d_inode)->root;
2971         struct btrfs_trans_handle *trans;
2972         u64 transid;
2973         int ret;
2974
2975         trans = btrfs_start_transaction(root, 0);
2976         if (IS_ERR(trans))
2977                 return PTR_ERR(trans);
2978         transid = trans->transid;
2979         ret = btrfs_commit_transaction_async(trans, root, 0);
2980         if (ret) {
2981                 btrfs_end_transaction(trans, root);
2982                 return ret;
2983         }
2984
2985         if (argp)
2986                 if (copy_to_user(argp, &transid, sizeof(transid)))
2987                         return -EFAULT;
2988         return 0;
2989 }
2990
2991 static noinline long btrfs_ioctl_wait_sync(struct file *file, void __user *argp)
2992 {
2993         struct btrfs_root *root = BTRFS_I(file->f_dentry->d_inode)->root;
2994         u64 transid;
2995
2996         if (argp) {
2997                 if (copy_from_user(&transid, argp, sizeof(transid)))
2998                         return -EFAULT;
2999         } else {
3000                 transid = 0;  /* current trans */
3001         }
3002         return btrfs_wait_for_commit(root, transid);
3003 }
3004
3005 static long btrfs_ioctl_scrub(struct btrfs_root *root, void __user *arg)
3006 {
3007         int ret;
3008         struct btrfs_ioctl_scrub_args *sa;
3009
3010         if (!capable(CAP_SYS_ADMIN))
3011                 return -EPERM;
3012
3013         sa = memdup_user(arg, sizeof(*sa));
3014         if (IS_ERR(sa))
3015                 return PTR_ERR(sa);
3016
3017         ret = btrfs_scrub_dev(root, sa->devid, sa->start, sa->end,
3018                               &sa->progress, sa->flags & BTRFS_SCRUB_READONLY);
3019
3020         if (copy_to_user(arg, sa, sizeof(*sa)))
3021                 ret = -EFAULT;
3022
3023         kfree(sa);
3024         return ret;
3025 }
3026
3027 static long btrfs_ioctl_scrub_cancel(struct btrfs_root *root, void __user *arg)
3028 {
3029         if (!capable(CAP_SYS_ADMIN))
3030                 return -EPERM;
3031
3032         return btrfs_scrub_cancel(root);
3033 }
3034
3035 static long btrfs_ioctl_scrub_progress(struct btrfs_root *root,
3036                                        void __user *arg)
3037 {
3038         struct btrfs_ioctl_scrub_args *sa;
3039         int ret;
3040
3041         if (!capable(CAP_SYS_ADMIN))
3042                 return -EPERM;
3043
3044         sa = memdup_user(arg, sizeof(*sa));
3045         if (IS_ERR(sa))
3046                 return PTR_ERR(sa);
3047
3048         ret = btrfs_scrub_progress(root, sa->devid, &sa->progress);
3049
3050         if (copy_to_user(arg, sa, sizeof(*sa)))
3051                 ret = -EFAULT;
3052
3053         kfree(sa);
3054         return ret;
3055 }
3056
3057 static long btrfs_ioctl_get_dev_stats(struct btrfs_root *root,
3058                                       void __user *arg, int reset_after_read)
3059 {
3060         struct btrfs_ioctl_get_dev_stats *sa;
3061         int ret;
3062
3063         if (reset_after_read && !capable(CAP_SYS_ADMIN))
3064                 return -EPERM;
3065
3066         sa = memdup_user(arg, sizeof(*sa));
3067         if (IS_ERR(sa))
3068                 return PTR_ERR(sa);
3069
3070         ret = btrfs_get_dev_stats(root, sa, reset_after_read);
3071
3072         if (copy_to_user(arg, sa, sizeof(*sa)))
3073                 ret = -EFAULT;
3074
3075         kfree(sa);
3076         return ret;
3077 }
3078
3079 static long btrfs_ioctl_ino_to_path(struct btrfs_root *root, void __user *arg)
3080 {
3081         int ret = 0;
3082         int i;
3083         u64 rel_ptr;
3084         int size;
3085         struct btrfs_ioctl_ino_path_args *ipa = NULL;
3086         struct inode_fs_paths *ipath = NULL;
3087         struct btrfs_path *path;
3088
3089         if (!capable(CAP_SYS_ADMIN))
3090                 return -EPERM;
3091
3092         path = btrfs_alloc_path();
3093         if (!path) {
3094                 ret = -ENOMEM;
3095                 goto out;
3096         }
3097
3098         ipa = memdup_user(arg, sizeof(*ipa));
3099         if (IS_ERR(ipa)) {
3100                 ret = PTR_ERR(ipa);
3101                 ipa = NULL;
3102                 goto out;
3103         }
3104
3105         size = min_t(u32, ipa->size, 4096);
3106         ipath = init_ipath(size, root, path);
3107         if (IS_ERR(ipath)) {
3108                 ret = PTR_ERR(ipath);
3109                 ipath = NULL;
3110                 goto out;
3111         }
3112
3113         ret = paths_from_inode(ipa->inum, ipath);
3114         if (ret < 0)
3115                 goto out;
3116
3117         for (i = 0; i < ipath->fspath->elem_cnt; ++i) {
3118                 rel_ptr = ipath->fspath->val[i] -
3119                           (u64)(unsigned long)ipath->fspath->val;
3120                 ipath->fspath->val[i] = rel_ptr;
3121         }
3122
3123         ret = copy_to_user((void *)(unsigned long)ipa->fspath,
3124                            (void *)(unsigned long)ipath->fspath, size);
3125         if (ret) {
3126                 ret = -EFAULT;
3127                 goto out;
3128         }
3129
3130 out:
3131         btrfs_free_path(path);
3132         free_ipath(ipath);
3133         kfree(ipa);
3134
3135         return ret;
3136 }
3137
3138 static int build_ino_list(u64 inum, u64 offset, u64 root, void *ctx)
3139 {
3140         struct btrfs_data_container *inodes = ctx;
3141         const size_t c = 3 * sizeof(u64);
3142
3143         if (inodes->bytes_left >= c) {
3144                 inodes->bytes_left -= c;
3145                 inodes->val[inodes->elem_cnt] = inum;
3146                 inodes->val[inodes->elem_cnt + 1] = offset;
3147                 inodes->val[inodes->elem_cnt + 2] = root;
3148                 inodes->elem_cnt += 3;
3149         } else {
3150                 inodes->bytes_missing += c - inodes->bytes_left;
3151                 inodes->bytes_left = 0;
3152                 inodes->elem_missed += 3;
3153         }
3154
3155         return 0;
3156 }
3157
3158 static long btrfs_ioctl_logical_to_ino(struct btrfs_root *root,
3159                                         void __user *arg)
3160 {
3161         int ret = 0;
3162         int size;
3163         u64 extent_item_pos;
3164         struct btrfs_ioctl_logical_ino_args *loi;
3165         struct btrfs_data_container *inodes = NULL;
3166         struct btrfs_path *path = NULL;
3167         struct btrfs_key key;
3168
3169         if (!capable(CAP_SYS_ADMIN))
3170                 return -EPERM;
3171
3172         loi = memdup_user(arg, sizeof(*loi));
3173         if (IS_ERR(loi)) {
3174                 ret = PTR_ERR(loi);
3175                 loi = NULL;
3176                 goto out;
3177         }
3178
3179         path = btrfs_alloc_path();
3180         if (!path) {
3181                 ret = -ENOMEM;
3182                 goto out;
3183         }
3184
3185         size = min_t(u32, loi->size, 4096);
3186         inodes = init_data_container(size);
3187         if (IS_ERR(inodes)) {
3188                 ret = PTR_ERR(inodes);
3189                 inodes = NULL;
3190                 goto out;
3191         }
3192
3193         ret = extent_from_logical(root->fs_info, loi->logical, path, &key);
3194         btrfs_release_path(path);
3195
3196         if (ret & BTRFS_EXTENT_FLAG_TREE_BLOCK)
3197                 ret = -ENOENT;
3198         if (ret < 0)
3199                 goto out;
3200
3201         extent_item_pos = loi->logical - key.objectid;
3202         ret = iterate_extent_inodes(root->fs_info, key.objectid,
3203                                         extent_item_pos, 0, build_ino_list,
3204                                         inodes);
3205
3206         if (ret < 0)
3207                 goto out;
3208
3209         ret = copy_to_user((void *)(unsigned long)loi->inodes,
3210                            (void *)(unsigned long)inodes, size);
3211         if (ret)
3212                 ret = -EFAULT;
3213
3214 out:
3215         btrfs_free_path(path);
3216         kfree(inodes);
3217         kfree(loi);
3218
3219         return ret;
3220 }
3221
3222 void update_ioctl_balance_args(struct btrfs_fs_info *fs_info, int lock,
3223                                struct btrfs_ioctl_balance_args *bargs)
3224 {
3225         struct btrfs_balance_control *bctl = fs_info->balance_ctl;
3226
3227         bargs->flags = bctl->flags;
3228
3229         if (atomic_read(&fs_info->balance_running))
3230                 bargs->state |= BTRFS_BALANCE_STATE_RUNNING;
3231         if (atomic_read(&fs_info->balance_pause_req))
3232                 bargs->state |= BTRFS_BALANCE_STATE_PAUSE_REQ;
3233         if (atomic_read(&fs_info->balance_cancel_req))
3234                 bargs->state |= BTRFS_BALANCE_STATE_CANCEL_REQ;
3235
3236         memcpy(&bargs->data, &bctl->data, sizeof(bargs->data));
3237         memcpy(&bargs->meta, &bctl->meta, sizeof(bargs->meta));
3238         memcpy(&bargs->sys, &bctl->sys, sizeof(bargs->sys));
3239
3240         if (lock) {
3241                 spin_lock(&fs_info->balance_lock);
3242                 memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat));
3243                 spin_unlock(&fs_info->balance_lock);
3244         } else {
3245                 memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat));
3246         }
3247 }
3248
3249 static long btrfs_ioctl_balance(struct file *file, void __user *arg)
3250 {
3251         struct btrfs_root *root = BTRFS_I(fdentry(file)->d_inode)->root;
3252         struct btrfs_fs_info *fs_info = root->fs_info;
3253         struct btrfs_ioctl_balance_args *bargs;
3254         struct btrfs_balance_control *bctl;
3255         int ret;
3256
3257         if (!capable(CAP_SYS_ADMIN))
3258                 return -EPERM;
3259
3260         if (fs_info->sb->s_flags & MS_RDONLY)
3261                 return -EROFS;
3262
3263         ret = mnt_want_write(file->f_path.mnt);
3264         if (ret)
3265                 return ret;
3266
3267         mutex_lock(&fs_info->volume_mutex);
3268         mutex_lock(&fs_info->balance_mutex);
3269
3270         if (arg) {
3271                 bargs = memdup_user(arg, sizeof(*bargs));
3272                 if (IS_ERR(bargs)) {
3273                         ret = PTR_ERR(bargs);
3274                         goto out;
3275                 }
3276
3277                 if (bargs->flags & BTRFS_BALANCE_RESUME) {
3278                         if (!fs_info->balance_ctl) {
3279                                 ret = -ENOTCONN;
3280                                 goto out_bargs;
3281                         }
3282
3283                         bctl = fs_info->balance_ctl;
3284                         spin_lock(&fs_info->balance_lock);
3285                         bctl->flags |= BTRFS_BALANCE_RESUME;
3286                         spin_unlock(&fs_info->balance_lock);
3287
3288                         goto do_balance;
3289                 }
3290         } else {
3291                 bargs = NULL;
3292         }
3293
3294         if (fs_info->balance_ctl) {
3295                 ret = -EINPROGRESS;
3296                 goto out_bargs;
3297         }
3298
3299         bctl = kzalloc(sizeof(*bctl), GFP_NOFS);
3300         if (!bctl) {
3301                 ret = -ENOMEM;
3302                 goto out_bargs;
3303         }
3304
3305         bctl->fs_info = fs_info;
3306         if (arg) {
3307                 memcpy(&bctl->data, &bargs->data, sizeof(bctl->data));
3308                 memcpy(&bctl->meta, &bargs->meta, sizeof(bctl->meta));
3309                 memcpy(&bctl->sys, &bargs->sys, sizeof(bctl->sys));
3310
3311                 bctl->flags = bargs->flags;
3312         } else {
3313                 /* balance everything - no filters */
3314                 bctl->flags |= BTRFS_BALANCE_TYPE_MASK;
3315         }
3316
3317 do_balance:
3318         ret = btrfs_balance(bctl, bargs);
3319         /*
3320          * bctl is freed in __cancel_balance or in free_fs_info if
3321          * restriper was paused all the way until unmount
3322          */
3323         if (arg) {
3324                 if (copy_to_user(arg, bargs, sizeof(*bargs)))
3325                         ret = -EFAULT;
3326         }
3327
3328 out_bargs:
3329         kfree(bargs);
3330 out:
3331         mutex_unlock(&fs_info->balance_mutex);
3332         mutex_unlock(&fs_info->volume_mutex);
3333         mnt_drop_write(file->f_path.mnt);
3334         return ret;
3335 }
3336
3337 static long btrfs_ioctl_balance_ctl(struct btrfs_root *root, int cmd)
3338 {
3339         if (!capable(CAP_SYS_ADMIN))
3340                 return -EPERM;
3341
3342         switch (cmd) {
3343         case BTRFS_BALANCE_CTL_PAUSE:
3344                 return btrfs_pause_balance(root->fs_info);
3345         case BTRFS_BALANCE_CTL_CANCEL:
3346                 return btrfs_cancel_balance(root->fs_info);
3347         }
3348
3349         return -EINVAL;
3350 }
3351
3352 static long btrfs_ioctl_balance_progress(struct btrfs_root *root,
3353                                          void __user *arg)
3354 {
3355         struct btrfs_fs_info *fs_info = root->fs_info;
3356         struct btrfs_ioctl_balance_args *bargs;
3357         int ret = 0;
3358
3359         if (!capable(CAP_SYS_ADMIN))
3360                 return -EPERM;
3361
3362         mutex_lock(&fs_info->balance_mutex);
3363         if (!fs_info->balance_ctl) {
3364                 ret = -ENOTCONN;
3365                 goto out;
3366         }
3367
3368         bargs = kzalloc(sizeof(*bargs), GFP_NOFS);
3369         if (!bargs) {
3370                 ret = -ENOMEM;
3371                 goto out;
3372         }
3373
3374         update_ioctl_balance_args(fs_info, 1, bargs);
3375
3376         if (copy_to_user(arg, bargs, sizeof(*bargs)))
3377                 ret = -EFAULT;
3378
3379         kfree(bargs);
3380 out:
3381         mutex_unlock(&fs_info->balance_mutex);
3382         return ret;
3383 }
3384
3385 long btrfs_ioctl(struct file *file, unsigned int
3386                 cmd, unsigned long arg)
3387 {
3388         struct btrfs_root *root = BTRFS_I(fdentry(file)->d_inode)->root;
3389         void __user *argp = (void __user *)arg;
3390
3391         switch (cmd) {
3392         case FS_IOC_GETFLAGS:
3393                 return btrfs_ioctl_getflags(file, argp);
3394         case FS_IOC_SETFLAGS:
3395                 return btrfs_ioctl_setflags(file, argp);
3396         case FS_IOC_GETVERSION:
3397                 return btrfs_ioctl_getversion(file, argp);
3398         case FITRIM:
3399                 return btrfs_ioctl_fitrim(file, argp);
3400         case BTRFS_IOC_SNAP_CREATE:
3401                 return btrfs_ioctl_snap_create(file, argp, 0);
3402         case BTRFS_IOC_SNAP_CREATE_V2:
3403                 return btrfs_ioctl_snap_create_v2(file, argp, 0);
3404         case BTRFS_IOC_SUBVOL_CREATE:
3405                 return btrfs_ioctl_snap_create(file, argp, 1);
3406         case BTRFS_IOC_SNAP_DESTROY:
3407                 return btrfs_ioctl_snap_destroy(file, argp);
3408         case BTRFS_IOC_SUBVOL_GETFLAGS:
3409                 return btrfs_ioctl_subvol_getflags(file, argp);
3410         case BTRFS_IOC_SUBVOL_SETFLAGS:
3411                 return btrfs_ioctl_subvol_setflags(file, argp);
3412         case BTRFS_IOC_DEFAULT_SUBVOL:
3413                 return btrfs_ioctl_default_subvol(file, argp);
3414         case BTRFS_IOC_DEFRAG:
3415                 return btrfs_ioctl_defrag(file, NULL);
3416         case BTRFS_IOC_DEFRAG_RANGE:
3417                 return btrfs_ioctl_defrag(file, argp);
3418         case BTRFS_IOC_RESIZE:
3419                 return btrfs_ioctl_resize(root, argp);
3420         case BTRFS_IOC_ADD_DEV:
3421                 return btrfs_ioctl_add_dev(root, argp);
3422         case BTRFS_IOC_RM_DEV:
3423                 return btrfs_ioctl_rm_dev(root, argp);
3424         case BTRFS_IOC_FS_INFO:
3425                 return btrfs_ioctl_fs_info(root, argp);
3426         case BTRFS_IOC_DEV_INFO:
3427                 return btrfs_ioctl_dev_info(root, argp);
3428         case BTRFS_IOC_BALANCE:
3429                 return btrfs_ioctl_balance(file, NULL);
3430         case BTRFS_IOC_CLONE:
3431                 return btrfs_ioctl_clone(file, arg, 0, 0, 0);
3432         case BTRFS_IOC_CLONE_RANGE:
3433                 return btrfs_ioctl_clone_range(file, argp);
3434         case BTRFS_IOC_TRANS_START:
3435                 return btrfs_ioctl_trans_start(file);
3436         case BTRFS_IOC_TRANS_END:
3437                 return btrfs_ioctl_trans_end(file);
3438         case BTRFS_IOC_TREE_SEARCH:
3439                 return btrfs_ioctl_tree_search(file, argp);
3440         case BTRFS_IOC_INO_LOOKUP:
3441                 return btrfs_ioctl_ino_lookup(file, argp);
3442         case BTRFS_IOC_INO_PATHS:
3443                 return btrfs_ioctl_ino_to_path(root, argp);
3444         case BTRFS_IOC_LOGICAL_INO:
3445                 return btrfs_ioctl_logical_to_ino(root, argp);
3446         case BTRFS_IOC_SPACE_INFO:
3447                 return btrfs_ioctl_space_info(root, argp);
3448         case BTRFS_IOC_SYNC:
3449                 btrfs_sync_fs(file->f_dentry->d_sb, 1);
3450                 return 0;
3451         case BTRFS_IOC_START_SYNC:
3452                 return btrfs_ioctl_start_sync(file, argp);
3453         case BTRFS_IOC_WAIT_SYNC:
3454                 return btrfs_ioctl_wait_sync(file, argp);
3455         case BTRFS_IOC_SCRUB:
3456                 return btrfs_ioctl_scrub(root, argp);
3457         case BTRFS_IOC_SCRUB_CANCEL:
3458                 return btrfs_ioctl_scrub_cancel(root, argp);
3459         case BTRFS_IOC_SCRUB_PROGRESS:
3460                 return btrfs_ioctl_scrub_progress(root, argp);
3461         case BTRFS_IOC_BALANCE_V2:
3462                 return btrfs_ioctl_balance(file, argp);
3463         case BTRFS_IOC_BALANCE_CTL:
3464                 return btrfs_ioctl_balance_ctl(root, arg);
3465         case BTRFS_IOC_BALANCE_PROGRESS:
3466                 return btrfs_ioctl_balance_progress(root, argp);
3467         case BTRFS_IOC_GET_DEV_STATS:
3468                 return btrfs_ioctl_get_dev_stats(root, argp, 0);
3469         case BTRFS_IOC_GET_AND_RESET_DEV_STATS:
3470                 return btrfs_ioctl_get_dev_stats(root, argp, 1);
3471         }
3472
3473         return -ENOTTY;
3474 }