Pileus Git - ~andy/linux/blob - fs/btrfs/ioctl.c

   1 /*
   2  * Copyright (C) 2007 Oracle.  All rights reserved.
   3  *
   4  * This program is free software; you can redistribute it and/or
   5  * modify it under the terms of the GNU General Public
   6  * License v2 as published by the Free Software Foundation.
   7  *
   8  * This program is distributed in the hope that it will be useful,
   9  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  10  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  11  * General Public License for more details.
  12  *
  13  * You should have received a copy of the GNU General Public
  14  * License along with this program; if not, write to the
  15  * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
  16  * Boston, MA 021110-1307, USA.
  17  */
  18
  19 #include <linux/kernel.h>
  20 #include <linux/bio.h>
  21 #include <linux/buffer_head.h>
  22 #include <linux/file.h>
  23 #include <linux/fs.h>
  24 #include <linux/fsnotify.h>
  25 #include <linux/pagemap.h>
  26 #include <linux/highmem.h>
  27 #include <linux/time.h>
  28 #include <linux/init.h>
  29 #include <linux/string.h>
  30 #include <linux/backing-dev.h>
  31 #include <linux/mount.h>
  32 #include <linux/mpage.h>
  33 #include <linux/namei.h>
  34 #include <linux/swap.h>
  35 #include <linux/writeback.h>
  36 #include <linux/statfs.h>
  37 #include <linux/compat.h>
  38 #include <linux/bit_spinlock.h>
  39 #include <linux/security.h>
  40 #include <linux/xattr.h>
  41 #include <linux/vmalloc.h>
  42 #include <linux/slab.h>
  43 #include <linux/blkdev.h>
  44 #include <linux/uuid.h>
  45 #include <linux/btrfs.h>
  46 #include "compat.h"
  47 #include "ctree.h"
  48 #include "disk-io.h"
  49 #include "transaction.h"
  50 #include "btrfs_inode.h"
  51 #include "print-tree.h"
  52 #include "volumes.h"
  53 #include "locking.h"
  54 #include "inode-map.h"
  55 #include "backref.h"
  56 #include "rcu-string.h"
  57 #include "send.h"
  58 #include "dev-replace.h"
  59
  60 /* Mask out flags that are inappropriate for the given type of inode. */
  61 static inline __u32 btrfs_mask_flags(umode_t mode, __u32 flags)
  62 {
  63         if (S_ISDIR(mode))
  64                 return flags;
  65         else if (S_ISREG(mode))
  66                 return flags & ~FS_DIRSYNC_FL;
  67         else
  68                 return flags & (FS_NODUMP_FL | FS_NOATIME_FL);
  69 }
  70
  71 /*
  72  * Export inode flags to the format expected by the FS_IOC_GETFLAGS ioctl.
  73  */
  74 static unsigned int btrfs_flags_to_ioctl(unsigned int flags)
  75 {
  76         unsigned int iflags = 0;
  77
  78         if (flags & BTRFS_INODE_SYNC)
  79                 iflags |= FS_SYNC_FL;
  80         if (flags & BTRFS_INODE_IMMUTABLE)
  81                 iflags |= FS_IMMUTABLE_FL;
  82         if (flags & BTRFS_INODE_APPEND)
  83                 iflags |= FS_APPEND_FL;
  84         if (flags & BTRFS_INODE_NODUMP)
  85                 iflags |= FS_NODUMP_FL;
  86         if (flags & BTRFS_INODE_NOATIME)
  87                 iflags |= FS_NOATIME_FL;
  88         if (flags & BTRFS_INODE_DIRSYNC)
  89                 iflags |= FS_DIRSYNC_FL;
  90         if (flags & BTRFS_INODE_NODATACOW)
  91                 iflags |= FS_NOCOW_FL;
  92
  93         if ((flags & BTRFS_INODE_COMPRESS) && !(flags & BTRFS_INODE_NOCOMPRESS))
  94                 iflags |= FS_COMPR_FL;
  95         else if (flags & BTRFS_INODE_NOCOMPRESS)
  96                 iflags |= FS_NOCOMP_FL;
  97
  98         return iflags;
  99 }
 100
 101 /*
 102  * Update inode->i_flags based on the btrfs internal flags.
 103  */
 104 void btrfs_update_iflags(struct inode *inode)
 105 {
 106         struct btrfs_inode *ip = BTRFS_I(inode);
 107
 108         inode->i_flags &= ~(S_SYNC|S_APPEND|S_IMMUTABLE|S_NOATIME|S_DIRSYNC);
 109
 110         if (ip->flags & BTRFS_INODE_SYNC)
 111                 inode->i_flags |= S_SYNC;
 112         if (ip->flags & BTRFS_INODE_IMMUTABLE)
 113                 inode->i_flags |= S_IMMUTABLE;
 114         if (ip->flags & BTRFS_INODE_APPEND)
 115                 inode->i_flags |= S_APPEND;
 116         if (ip->flags & BTRFS_INODE_NOATIME)
 117                 inode->i_flags |= S_NOATIME;
 118         if (ip->flags & BTRFS_INODE_DIRSYNC)
 119                 inode->i_flags |= S_DIRSYNC;
 120 }
 121
 122 /*
 123  * Inherit flags from the parent inode.
 124  *
 125  * Currently only the compression flags and the cow flags are inherited.
 126  */
 127 void btrfs_inherit_iflags(struct inode *inode, struct inode *dir)
 128 {
 129         unsigned int flags;
 130
 131         if (!dir)
 132                 return;
 133
 134         flags = BTRFS_I(dir)->flags;
 135
 136         if (flags & BTRFS_INODE_NOCOMPRESS) {
 137                 BTRFS_I(inode)->flags &= ~BTRFS_INODE_COMPRESS;
 138                 BTRFS_I(inode)->flags |= BTRFS_INODE_NOCOMPRESS;
 139         } else if (flags & BTRFS_INODE_COMPRESS) {
 140                 BTRFS_I(inode)->flags &= ~BTRFS_INODE_NOCOMPRESS;
 141                 BTRFS_I(inode)->flags |= BTRFS_INODE_COMPRESS;
 142         }
 143
 144         if (flags & BTRFS_INODE_NODATACOW) {
 145                 BTRFS_I(inode)->flags |= BTRFS_INODE_NODATACOW;
 146                 if (S_ISREG(inode->i_mode))
 147                         BTRFS_I(inode)->flags |= BTRFS_INODE_NODATASUM;
 148         }
 149
 150         btrfs_update_iflags(inode);
 151 }
 152
 153 static int btrfs_ioctl_getflags(struct file *file, void __user *arg)
 154 {
 155         struct btrfs_inode *ip = BTRFS_I(file_inode(file));
 156         unsigned int flags = btrfs_flags_to_ioctl(ip->flags);
 157
 158         if (copy_to_user(arg, &flags, sizeof(flags)))
 159                 return -EFAULT;
 160         return 0;
 161 }
 162
 163 static int check_flags(unsigned int flags)
 164 {
 165         if (flags & ~(FS_IMMUTABLE_FL | FS_APPEND_FL | \
 166                       FS_NOATIME_FL | FS_NODUMP_FL | \
 167                       FS_SYNC_FL | FS_DIRSYNC_FL | \
 168                       FS_NOCOMP_FL | FS_COMPR_FL |
 169                       FS_NOCOW_FL))
 170                 return -EOPNOTSUPP;
 171
 172         if ((flags & FS_NOCOMP_FL) && (flags & FS_COMPR_FL))
 173                 return -EINVAL;
 174
 175         return 0;
 176 }
 177
 178 static int btrfs_ioctl_setflags(struct file *file, void __user *arg)
 179 {
 180         struct inode *inode = file_inode(file);
 181         struct btrfs_inode *ip = BTRFS_I(inode);
 182         struct btrfs_root *root = ip->root;
 183         struct btrfs_trans_handle *trans;
 184         unsigned int flags, oldflags;
 185         int ret;
 186         u64 ip_oldflags;
 187         unsigned int i_oldflags;
 188         umode_t mode;
 189
 190         if (btrfs_root_readonly(root))
 191                 return -EROFS;
 192
 193         if (copy_from_user(&flags, arg, sizeof(flags)))
 194                 return -EFAULT;
 195
 196         ret = check_flags(flags);
 197         if (ret)
 198                 return ret;
 199
 200         if (!inode_owner_or_capable(inode))
 201                 return -EACCES;
 202
 203         ret = mnt_want_write_file(file);
 204         if (ret)
 205                 return ret;
 206
 207         mutex_lock(&inode->i_mutex);
 208
 209         ip_oldflags = ip->flags;
 210         i_oldflags = inode->i_flags;
 211         mode = inode->i_mode;
 212
 213         flags = btrfs_mask_flags(inode->i_mode, flags);
 214         oldflags = btrfs_flags_to_ioctl(ip->flags);
 215         if ((flags ^ oldflags) & (FS_APPEND_FL | FS_IMMUTABLE_FL)) {
 216                 if (!capable(CAP_LINUX_IMMUTABLE)) {
 217                         ret = -EPERM;
 218                         goto out_unlock;
 219                 }
 220         }
 221
 222         if (flags & FS_SYNC_FL)
 223                 ip->flags |= BTRFS_INODE_SYNC;
 224         else
 225                 ip->flags &= ~BTRFS_INODE_SYNC;
 226         if (flags & FS_IMMUTABLE_FL)
 227                 ip->flags |= BTRFS_INODE_IMMUTABLE;
 228         else
 229                 ip->flags &= ~BTRFS_INODE_IMMUTABLE;
 230         if (flags & FS_APPEND_FL)
 231                 ip->flags |= BTRFS_INODE_APPEND;
 232         else
 233                 ip->flags &= ~BTRFS_INODE_APPEND;
 234         if (flags & FS_NODUMP_FL)
 235                 ip->flags |= BTRFS_INODE_NODUMP;
 236         else
 237                 ip->flags &= ~BTRFS_INODE_NODUMP;
 238         if (flags & FS_NOATIME_FL)
 239                 ip->flags |= BTRFS_INODE_NOATIME;
 240         else
 241                 ip->flags &= ~BTRFS_INODE_NOATIME;
 242         if (flags & FS_DIRSYNC_FL)
 243                 ip->flags |= BTRFS_INODE_DIRSYNC;
 244         else
 245                 ip->flags &= ~BTRFS_INODE_DIRSYNC;
 246         if (flags & FS_NOCOW_FL) {
 247                 if (S_ISREG(mode)) {
 248                         /*
 249                          * It's safe to turn csums off here, no extents exist.
 250                          * Otherwise we want the flag to reflect the real COW
 251                          * status of the file and will not set it.
 252                          */
 253                         if (inode->i_size == 0)
 254                                 ip->flags |= BTRFS_INODE_NODATACOW
 255                                            | BTRFS_INODE_NODATASUM;
 256                 } else {
 257                         ip->flags |= BTRFS_INODE_NODATACOW;
 258                 }
 259         } else {
 260                 /*
 261                  * Revert back under same assuptions as above
 262                  */
 263                 if (S_ISREG(mode)) {
 264                         if (inode->i_size == 0)
 265                                 ip->flags &= ~(BTRFS_INODE_NODATACOW
 266                                              | BTRFS_INODE_NODATASUM);
 267                 } else {
 268                         ip->flags &= ~BTRFS_INODE_NODATACOW;
 269                 }
 270         }
 271
 272         /*
 273          * The COMPRESS flag can only be changed by users, while the NOCOMPRESS
 274          * flag may be changed automatically if compression code won't make
 275          * things smaller.
 276          */
 277         if (flags & FS_NOCOMP_FL) {
 278                 ip->flags &= ~BTRFS_INODE_COMPRESS;
 279                 ip->flags |= BTRFS_INODE_NOCOMPRESS;
 280         } else if (flags & FS_COMPR_FL) {
 281                 ip->flags |= BTRFS_INODE_COMPRESS;
 282                 ip->flags &= ~BTRFS_INODE_NOCOMPRESS;
 283         } else {
 284                 ip->flags &= ~(BTRFS_INODE_COMPRESS | BTRFS_INODE_NOCOMPRESS);
 285         }
 286
 287         trans = btrfs_start_transaction(root, 1);
 288         if (IS_ERR(trans)) {
 289                 ret = PTR_ERR(trans);
 290                 goto out_drop;
 291         }
 292
 293         btrfs_update_iflags(inode);
 294         inode_inc_iversion(inode);
 295         inode->i_ctime = CURRENT_TIME;
 296         ret = btrfs_update_inode(trans, root, inode);
 297
 298         btrfs_end_transaction(trans, root);
 299  out_drop:
 300         if (ret) {
 301                 ip->flags = ip_oldflags;
 302                 inode->i_flags = i_oldflags;
 303         }
 304
 305  out_unlock:
 306         mutex_unlock(&inode->i_mutex);
 307         mnt_drop_write_file(file);
 308         return ret;
 309 }
 310
 311 static int btrfs_ioctl_getversion(struct file *file, int __user *arg)
 312 {
 313         struct inode *inode = file_inode(file);
 314
 315         return put_user(inode->i_generation, arg);
 316 }
 317
 318 static noinline int btrfs_ioctl_fitrim(struct file *file, void __user *arg)
 319 {
 320         struct btrfs_fs_info *fs_info = btrfs_sb(fdentry(file)->d_sb);
 321         struct btrfs_device *device;
 322         struct request_queue *q;
 323         struct fstrim_range range;
 324         u64 minlen = ULLONG_MAX;
 325         u64 num_devices = 0;
 326         u64 total_bytes = btrfs_super_total_bytes(fs_info->super_copy);
 327         int ret;
 328
 329         if (!capable(CAP_SYS_ADMIN))
 330                 return -EPERM;
 331
 332         rcu_read_lock();
 333         list_for_each_entry_rcu(device, &fs_info->fs_devices->devices,
 334                                 dev_list) {
 335                 if (!device->bdev)
 336                         continue;
 337                 q = bdev_get_queue(device->bdev);
 338                 if (blk_queue_discard(q)) {
 339                         num_devices++;
 340                         minlen = min((u64)q->limits.discard_granularity,
 341                                      minlen);
 342                 }
 343         }
 344         rcu_read_unlock();
 345
 346         if (!num_devices)
 347                 return -EOPNOTSUPP;
 348         if (copy_from_user(&range, arg, sizeof(range)))
 349                 return -EFAULT;
 350         if (range.start > total_bytes ||
 351             range.len < fs_info->sb->s_blocksize)
 352                 return -EINVAL;
 353
 354         range.len = min(range.len, total_bytes - range.start);
 355         range.minlen = max(range.minlen, minlen);
 356         ret = btrfs_trim_fs(fs_info->tree_root, &range);
 357         if (ret < 0)
 358                 return ret;
 359
 360         if (copy_to_user(arg, &range, sizeof(range)))
 361                 return -EFAULT;
 362
 363         return 0;
 364 }
 365
 366 static noinline int create_subvol(struct inode *dir,
 367                                   struct dentry *dentry,
 368                                   char *name, int namelen,
 369                                   u64 *async_transid,
 370                                   struct btrfs_qgroup_inherit *inherit)
 371 {
 372         struct btrfs_trans_handle *trans;
 373         struct btrfs_key key;
 374         struct btrfs_root_item root_item;
 375         struct btrfs_inode_item *inode_item;
 376         struct extent_buffer *leaf;
 377         struct btrfs_root *root = BTRFS_I(dir)->root;
 378         struct btrfs_root *new_root;
 379         struct btrfs_block_rsv block_rsv;
 380         struct timespec cur_time = CURRENT_TIME;
 381         int ret;
 382         int err;
 383         u64 objectid;
 384         u64 new_dirid = BTRFS_FIRST_FREE_OBJECTID;
 385         u64 index = 0;
 386         u64 qgroup_reserved;
 387         uuid_le new_uuid;
 388
 389         ret = btrfs_find_free_objectid(root->fs_info->tree_root, &objectid);
 390         if (ret)
 391                 return ret;
 392
 393         btrfs_init_block_rsv(&block_rsv, BTRFS_BLOCK_RSV_TEMP);
 394         /*
 395          * The same as the snapshot creation, please see the comment
 396          * of create_snapshot().
 397          */
 398         ret = btrfs_subvolume_reserve_metadata(root, &block_rsv,
 399                                                7, &qgroup_reserved);
 400         if (ret)
 401                 return ret;
 402
 403         trans = btrfs_start_transaction(root, 0);
 404         if (IS_ERR(trans)) {
 405                 ret = PTR_ERR(trans);
 406                 goto out;
 407         }
 408         trans->block_rsv = &block_rsv;
 409         trans->bytes_reserved = block_rsv.size;
 410
 411         ret = btrfs_qgroup_inherit(trans, root->fs_info, 0, objectid, inherit);
 412         if (ret)
 413                 goto fail;
 414
 415         leaf = btrfs_alloc_free_block(trans, root, root->leafsize,
 416                                       0, objectid, NULL, 0, 0, 0);
 417         if (IS_ERR(leaf)) {
 418                 ret = PTR_ERR(leaf);
 419                 goto fail;
 420         }
 421
 422         memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header));
 423         btrfs_set_header_bytenr(leaf, leaf->start);
 424         btrfs_set_header_generation(leaf, trans->transid);
 425         btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV);
 426         btrfs_set_header_owner(leaf, objectid);
 427
 428         write_extent_buffer(leaf, root->fs_info->fsid,
 429                             (unsigned long)btrfs_header_fsid(leaf),
 430                             BTRFS_FSID_SIZE);
 431         write_extent_buffer(leaf, root->fs_info->chunk_tree_uuid,
 432                             (unsigned long)btrfs_header_chunk_tree_uuid(leaf),
 433                             BTRFS_UUID_SIZE);
 434         btrfs_mark_buffer_dirty(leaf);
 435
 436         memset(&root_item, 0, sizeof(root_item));
 437
 438         inode_item = &root_item.inode;
 439         inode_item->generation = cpu_to_le64(1);
 440         inode_item->size = cpu_to_le64(3);
 441         inode_item->nlink = cpu_to_le32(1);
 442         inode_item->nbytes = cpu_to_le64(root->leafsize);
 443         inode_item->mode = cpu_to_le32(S_IFDIR | 0755);
 444
 445         root_item.flags = 0;
 446         root_item.byte_limit = 0;
 447         inode_item->flags = cpu_to_le64(BTRFS_INODE_ROOT_ITEM_INIT);
 448
 449         btrfs_set_root_bytenr(&root_item, leaf->start);
 450         btrfs_set_root_generation(&root_item, trans->transid);
 451         btrfs_set_root_level(&root_item, 0);
 452         btrfs_set_root_refs(&root_item, 1);
 453         btrfs_set_root_used(&root_item, leaf->len);
 454         btrfs_set_root_last_snapshot(&root_item, 0);
 455
 456         btrfs_set_root_generation_v2(&root_item,
 457                         btrfs_root_generation(&root_item));
 458         uuid_le_gen(&new_uuid);
 459         memcpy(root_item.uuid, new_uuid.b, BTRFS_UUID_SIZE);
 460         root_item.otime.sec = cpu_to_le64(cur_time.tv_sec);
 461         root_item.otime.nsec = cpu_to_le32(cur_time.tv_nsec);
 462         root_item.ctime = root_item.otime;
 463         btrfs_set_root_ctransid(&root_item, trans->transid);
 464         btrfs_set_root_otransid(&root_item, trans->transid);
 465
 466         btrfs_tree_unlock(leaf);
 467         free_extent_buffer(leaf);
 468         leaf = NULL;
 469
 470         btrfs_set_root_dirid(&root_item, new_dirid);
 471
 472         key.objectid = objectid;
 473         key.offset = 0;
 474         btrfs_set_key_type(&key, BTRFS_ROOT_ITEM_KEY);
 475         ret = btrfs_insert_root(trans, root->fs_info->tree_root, &key,
 476                                 &root_item);
 477         if (ret)
 478                 goto fail;
 479
 480         key.offset = (u64)-1;
 481         new_root = btrfs_read_fs_root_no_name(root->fs_info, &key);
 482         if (IS_ERR(new_root)) {
 483                 btrfs_abort_transaction(trans, root, PTR_ERR(new_root));
 484                 ret = PTR_ERR(new_root);
 485                 goto fail;
 486         }
 487
 488         btrfs_record_root_in_trans(trans, new_root);
 489
 490         ret = btrfs_create_subvol_root(trans, new_root, new_dirid);
 491         if (ret) {
 492                 /* We potentially lose an unused inode item here */
 493                 btrfs_abort_transaction(trans, root, ret);
 494                 goto fail;
 495         }
 496
 497         /*
 498          * insert the directory item
 499          */
 500         ret = btrfs_set_inode_index(dir, &index);
 501         if (ret) {
 502                 btrfs_abort_transaction(trans, root, ret);
 503                 goto fail;
 504         }
 505
 506         ret = btrfs_insert_dir_item(trans, root,
 507                                     name, namelen, dir, &key,
 508                                     BTRFS_FT_DIR, index);
 509         if (ret) {
 510                 btrfs_abort_transaction(trans, root, ret);
 511                 goto fail;
 512         }
 513
 514         btrfs_i_size_write(dir, dir->i_size + namelen * 2);
 515         ret = btrfs_update_inode(trans, root, dir);
 516         BUG_ON(ret);
 517
 518         ret = btrfs_add_root_ref(trans, root->fs_info->tree_root,
 519                                  objectid, root->root_key.objectid,
 520                                  btrfs_ino(dir), index, name, namelen);
 521
 522         BUG_ON(ret);
 523
 524 fail:
 525         trans->block_rsv = NULL;
 526         trans->bytes_reserved = 0;
 527         if (async_transid) {
 528                 *async_transid = trans->transid;
 529                 err = btrfs_commit_transaction_async(trans, root, 1);
 530                 if (err)
 531                         err = btrfs_commit_transaction(trans, root);
 532         } else {
 533                 err = btrfs_commit_transaction(trans, root);
 534         }
 535         if (err && !ret)
 536                 ret = err;
 537
 538         if (!ret)
 539                 d_instantiate(dentry, btrfs_lookup_dentry(dir, dentry));
 540 out:
 541         btrfs_subvolume_release_metadata(root, &block_rsv, qgroup_reserved);
 542         return ret;
 543 }
 544
 545 static int create_snapshot(struct btrfs_root *root, struct inode *dir,
 546                            struct dentry *dentry, char *name, int namelen,
 547                            u64 *async_transid, bool readonly,
 548                            struct btrfs_qgroup_inherit *inherit)
 549 {
 550         struct inode *inode;
 551         struct btrfs_pending_snapshot *pending_snapshot;
 552         struct btrfs_trans_handle *trans;
 553         int ret;
 554
 555         if (!root->ref_cows)
 556                 return -EINVAL;
 557
 558         pending_snapshot = kzalloc(sizeof(*pending_snapshot), GFP_NOFS);
 559         if (!pending_snapshot)
 560                 return -ENOMEM;
 561
 562         btrfs_init_block_rsv(&pending_snapshot->block_rsv,
 563                              BTRFS_BLOCK_RSV_TEMP);
 564         /*
 565          * 1 - parent dir inode
 566          * 2 - dir entries
 567          * 1 - root item
 568          * 2 - root ref/backref
 569          * 1 - root of snapshot
 570          */
 571         ret = btrfs_subvolume_reserve_metadata(BTRFS_I(dir)->root,
 572                                         &pending_snapshot->block_rsv, 7,
 573                                         &pending_snapshot->qgroup_reserved);
 574         if (ret)
 575                 goto out;
 576
 577         pending_snapshot->dentry = dentry;
 578         pending_snapshot->root = root;
 579         pending_snapshot->readonly = readonly;
 580         pending_snapshot->dir = dir;
 581         pending_snapshot->inherit = inherit;
 582
 583         trans = btrfs_start_transaction(root, 0);
 584         if (IS_ERR(trans)) {
 585                 ret = PTR_ERR(trans);
 586                 goto fail;
 587         }
 588
 589         spin_lock(&root->fs_info->trans_lock);
 590         list_add(&pending_snapshot->list,
 591                  &trans->transaction->pending_snapshots);
 592         spin_unlock(&root->fs_info->trans_lock);
 593         if (async_transid) {
 594                 *async_transid = trans->transid;
 595                 ret = btrfs_commit_transaction_async(trans,
 596                                      root->fs_info->extent_root, 1);
 597                 if (ret)
 598                         ret = btrfs_commit_transaction(trans, root);
 599         } else {
 600                 ret = btrfs_commit_transaction(trans,
 601                                                root->fs_info->extent_root);
 602         }
 603         if (ret)
 604                 goto fail;
 605
 606         ret = pending_snapshot->error;
 607         if (ret)
 608                 goto fail;
 609
 610         ret = btrfs_orphan_cleanup(pending_snapshot->snap);
 611         if (ret)
 612                 goto fail;
 613
 614         inode = btrfs_lookup_dentry(dentry->d_parent->d_inode, dentry);
 615         if (IS_ERR(inode)) {
 616                 ret = PTR_ERR(inode);
 617                 goto fail;
 618         }
 619         BUG_ON(!inode);
 620         d_instantiate(dentry, inode);
 621         ret = 0;
 622 fail:
 623         btrfs_subvolume_release_metadata(BTRFS_I(dir)->root,
 624                                          &pending_snapshot->block_rsv,
 625                                          pending_snapshot->qgroup_reserved);
 626 out:
 627         kfree(pending_snapshot);
 628         return ret;
 629 }
 630
 631 /*  copy of check_sticky in fs/namei.c()
 632 * It's inline, so penalty for filesystems that don't use sticky bit is
 633 * minimal.
 634 */
 635 static inline int btrfs_check_sticky(struct inode *dir, struct inode *inode)
 636 {
 637         kuid_t fsuid = current_fsuid();
 638
 639         if (!(dir->i_mode & S_ISVTX))
 640                 return 0;
 641         if (uid_eq(inode->i_uid, fsuid))
 642                 return 0;
 643         if (uid_eq(dir->i_uid, fsuid))
 644                 return 0;
 645         return !capable(CAP_FOWNER);
 646 }
 647
 648 /*  copy of may_delete in fs/namei.c()
 649  *      Check whether we can remove a link victim from directory dir, check
 650  *  whether the type of victim is right.
 651  *  1. We can't do it if dir is read-only (done in permission())
 652  *  2. We should have write and exec permissions on dir
 653  *  3. We can't remove anything from append-only dir
 654  *  4. We can't do anything with immutable dir (done in permission())
 655  *  5. If the sticky bit on dir is set we should either
 656  *      a. be owner of dir, or
 657  *      b. be owner of victim, or
 658  *      c. have CAP_FOWNER capability
 659  *  6. If the victim is append-only or immutable we can't do antyhing with
 660  *     links pointing to it.
 661  *  7. If we were asked to remove a directory and victim isn't one - ENOTDIR.
 662  *  8. If we were asked to remove a non-directory and victim isn't one - EISDIR.
 663  *  9. We can't remove a root or mountpoint.
 664  * 10. We don't allow removal of NFS sillyrenamed files; it's handled by
 665  *     nfs_async_unlink().
 666  */
 667
 668 static int btrfs_may_delete(struct inode *dir,struct dentry *victim,int isdir)
 669 {
 670         int error;
 671
 672         if (!victim->d_inode)
 673                 return -ENOENT;
 674
 675         BUG_ON(victim->d_parent->d_inode != dir);
 676         audit_inode_child(dir, victim, AUDIT_TYPE_CHILD_DELETE);
 677
 678         error = inode_permission(dir, MAY_WRITE | MAY_EXEC);
 679         if (error)
 680                 return error;
 681         if (IS_APPEND(dir))
 682                 return -EPERM;
 683         if (btrfs_check_sticky(dir, victim->d_inode)||
 684                 IS_APPEND(victim->d_inode)||
 685             IS_IMMUTABLE(victim->d_inode) || IS_SWAPFILE(victim->d_inode))
 686                 return -EPERM;
 687         if (isdir) {
 688                 if (!S_ISDIR(victim->d_inode->i_mode))
 689                         return -ENOTDIR;
 690                 if (IS_ROOT(victim))
 691                         return -EBUSY;
 692         } else if (S_ISDIR(victim->d_inode->i_mode))
 693                 return -EISDIR;
 694         if (IS_DEADDIR(dir))
 695                 return -ENOENT;
 696         if (victim->d_flags & DCACHE_NFSFS_RENAMED)
 697                 return -EBUSY;
 698         return 0;
 699 }
 700
 701 /* copy of may_create in fs/namei.c() */
 702 static inline int btrfs_may_create(struct inode *dir, struct dentry *child)
 703 {
 704         if (child->d_inode)
 705                 return -EEXIST;
 706         if (IS_DEADDIR(dir))
 707                 return -ENOENT;
 708         return inode_permission(dir, MAY_WRITE | MAY_EXEC);
 709 }
 710
 711 /*
 712  * Create a new subvolume below @parent.  This is largely modeled after
 713  * sys_mkdirat and vfs_mkdir, but we only do a single component lookup
 714  * inside this filesystem so it's quite a bit simpler.
 715  */
 716 static noinline int btrfs_mksubvol(struct path *parent,
 717                                    char *name, int namelen,
 718                                    struct btrfs_root *snap_src,
 719                                    u64 *async_transid, bool readonly,
 720                                    struct btrfs_qgroup_inherit *inherit)
 721 {
 722         struct inode *dir  = parent->dentry->d_inode;
 723         struct dentry *dentry;
 724         int error;
 725
 726         error = mutex_lock_killable_nested(&dir->i_mutex, I_MUTEX_PARENT);
 727         if (error == -EINTR)
 728                 return error;
 729
 730         dentry = lookup_one_len(name, parent->dentry, namelen);
 731         error = PTR_ERR(dentry);
 732         if (IS_ERR(dentry))
 733                 goto out_unlock;
 734
 735         error = -EEXIST;
 736         if (dentry->d_inode)
 737                 goto out_dput;
 738
 739         error = btrfs_may_create(dir, dentry);
 740         if (error)
 741                 goto out_dput;
 742
 743         /*
 744          * even if this name doesn't exist, we may get hash collisions.
 745          * check for them now when we can safely fail
 746          */
 747         error = btrfs_check_dir_item_collision(BTRFS_I(dir)->root,
 748                                                dir->i_ino, name,
 749                                                namelen);
 750         if (error)
 751                 goto out_dput;
 752
 753         down_read(&BTRFS_I(dir)->root->fs_info->subvol_sem);
 754
 755         if (btrfs_root_refs(&BTRFS_I(dir)->root->root_item) == 0)
 756                 goto out_up_read;
 757
 758         if (snap_src) {
 759                 error = create_snapshot(snap_src, dir, dentry, name, namelen,
 760                                         async_transid, readonly, inherit);
 761         } else {
 762                 error = create_subvol(dir, dentry, name, namelen,
 763                                       async_transid, inherit);
 764         }
 765         if (!error)
 766                 fsnotify_mkdir(dir, dentry);
 767 out_up_read:
 768         up_read(&BTRFS_I(dir)->root->fs_info->subvol_sem);
 769 out_dput:
 770         dput(dentry);
 771 out_unlock:
 772         mutex_unlock(&dir->i_mutex);
 773         return error;
 774 }
 775
 776 /*
 777  * When we're defragging a range, we don't want to kick it off again
 778  * if it is really just waiting for delalloc to send it down.
 779  * If we find a nice big extent or delalloc range for the bytes in the
 780  * file you want to defrag, we return 0 to let you know to skip this
 781  * part of the file
 782  */
 783 static int check_defrag_in_cache(struct inode *inode, u64 offset, int thresh)
 784 {
 785         struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
 786         struct extent_map *em = NULL;
 787         struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
 788         u64 end;
 789
 790         read_lock(&em_tree->lock);
 791         em = lookup_extent_mapping(em_tree, offset, PAGE_CACHE_SIZE);
 792         read_unlock(&em_tree->lock);
 793
 794         if (em) {
 795                 end = extent_map_end(em);
 796                 free_extent_map(em);
 797                 if (end - offset > thresh)
 798                         return 0;
 799         }
 800         /* if we already have a nice delalloc here, just stop */
 801         thresh /= 2;
 802         end = count_range_bits(io_tree, &offset, offset + thresh,
 803                                thresh, EXTENT_DELALLOC, 1);
 804         if (end >= thresh)
 805                 return 0;
 806         return 1;
 807 }
 808
 809 /*
 810  * helper function to walk through a file and find extents
 811  * newer than a specific transid, and smaller than thresh.
 812  *
 813  * This is used by the defragging code to find new and small
 814  * extents
 815  */
 816 static int find_new_extents(struct btrfs_root *root,
 817                             struct inode *inode, u64 newer_than,
 818                             u64 *off, int thresh)
 819 {
 820         struct btrfs_path *path;
 821         struct btrfs_key min_key;
 822         struct btrfs_key max_key;
 823         struct extent_buffer *leaf;
 824         struct btrfs_file_extent_item *extent;
 825         int type;
 826         int ret;
 827         u64 ino = btrfs_ino(inode);
 828
 829         path = btrfs_alloc_path();
 830         if (!path)
 831                 return -ENOMEM;
 832
 833         min_key.objectid = ino;
 834         min_key.type = BTRFS_EXTENT_DATA_KEY;
 835         min_key.offset = *off;
 836
 837         max_key.objectid = ino;
 838         max_key.type = (u8)-1;
 839         max_key.offset = (u64)-1;
 840
 841         path->keep_locks = 1;
 842
 843         while(1) {
 844                 ret = btrfs_search_forward(root, &min_key, &max_key,
 845                                            path, newer_than);
 846                 if (ret != 0)
 847                         goto none;
 848                 if (min_key.objectid != ino)
 849                         goto none;
 850                 if (min_key.type != BTRFS_EXTENT_DATA_KEY)
 851                         goto none;
 852
 853                 leaf = path->nodes[0];
 854                 extent = btrfs_item_ptr(leaf, path->slots[0],
 855                                         struct btrfs_file_extent_item);
 856
 857                 type = btrfs_file_extent_type(leaf, extent);
 858                 if (type == BTRFS_FILE_EXTENT_REG &&
 859                     btrfs_file_extent_num_bytes(leaf, extent) < thresh &&
 860                     check_defrag_in_cache(inode, min_key.offset, thresh)) {
 861                         *off = min_key.offset;
 862                         btrfs_free_path(path);
 863                         return 0;
 864                 }
 865
 866                 if (min_key.offset == (u64)-1)
 867                         goto none;
 868
 869                 min_key.offset++;
 870                 btrfs_release_path(path);
 871         }
 872 none:
 873         btrfs_free_path(path);
 874         return -ENOENT;
 875 }
 876
 877 static struct extent_map *defrag_lookup_extent(struct inode *inode, u64 start)
 878 {
 879         struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
 880         struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
 881         struct extent_map *em;
 882         u64 len = PAGE_CACHE_SIZE;
 883
 884         /*
 885          * hopefully we have this extent in the tree already, try without
 886          * the full extent lock
 887          */
 888         read_lock(&em_tree->lock);
 889         em = lookup_extent_mapping(em_tree, start, len);
 890         read_unlock(&em_tree->lock);
 891
 892         if (!em) {
 893                 /* get the big lock and read metadata off disk */
 894                 lock_extent(io_tree, start, start + len - 1);
 895                 em = btrfs_get_extent(inode, NULL, 0, start, len, 0);
 896                 unlock_extent(io_tree, start, start + len - 1);
 897
 898                 if (IS_ERR(em))
 899                         return NULL;
 900         }
 901
 902         return em;
 903 }
 904
 905 static bool defrag_check_next_extent(struct inode *inode, struct extent_map *em)
 906 {
 907         struct extent_map *next;
 908         bool ret = true;
 909
 910         /* this is the last extent */
 911         if (em->start + em->len >= i_size_read(inode))
 912                 return false;
 913
 914         next = defrag_lookup_extent(inode, em->start + em->len);
 915         if (!next || next->block_start >= EXTENT_MAP_LAST_BYTE)
 916                 ret = false;
 917
 918         free_extent_map(next);
 919         return ret;
 920 }
 921
 922 static int should_defrag_range(struct inode *inode, u64 start, int thresh,
 923                                u64 *last_len, u64 *skip, u64 *defrag_end,
 924                                int compress)
 925 {
 926         struct extent_map *em;
 927         int ret = 1;
 928         bool next_mergeable = true;
 929
 930         /*
 931          * make sure that once we start defragging an extent, we keep on
 932          * defragging it
 933          */
 934         if (start < *defrag_end)
 935                 return 1;
 936
 937         *skip = 0;
 938
 939         em = defrag_lookup_extent(inode, start);
 940         if (!em)
 941                 return 0;
 942
 943         /* this will cover holes, and inline extents */
 944         if (em->block_start >= EXTENT_MAP_LAST_BYTE) {
 945                 ret = 0;
 946                 goto out;
 947         }
 948
 949         next_mergeable = defrag_check_next_extent(inode, em);
 950
 951         /*
 952          * we hit a real extent, if it is big or the next extent is not a
 953          * real extent, don't bother defragging it
 954          */
 955         if (!compress && (*last_len == 0 || *last_len >= thresh) &&
 956             (em->len >= thresh || !next_mergeable))
 957                 ret = 0;
 958 out:
 959         /*
 960          * last_len ends up being a counter of how many bytes we've defragged.
 961          * every time we choose not to defrag an extent, we reset *last_len
 962          * so that the next tiny extent will force a defrag.
 963          *
 964          * The end result of this is that tiny extents before a single big
 965          * extent will force at least part of that big extent to be defragged.
 966          */
 967         if (ret) {
 968                 *defrag_end = extent_map_end(em);
 969         } else {
 970                 *last_len = 0;
 971                 *skip = extent_map_end(em);
 972                 *defrag_end = 0;
 973         }
 974
 975         free_extent_map(em);
 976         return ret;
 977 }
 978
 979 /*
 980  * it doesn't do much good to defrag one or two pages
 981  * at a time.  This pulls in a nice chunk of pages
 982  * to COW and defrag.
 983  *
 984  * It also makes sure the delalloc code has enough
 985  * dirty data to avoid making new small extents as part
 986  * of the defrag
 987  *
 988  * It's a good idea to start RA on this range
 989  * before calling this.
 990  */
 991 static int cluster_pages_for_defrag(struct inode *inode,
 992                                     struct page **pages,
 993                                     unsigned long start_index,
 994                                     int num_pages)
 995 {
 996         unsigned long file_end;
 997         u64 isize = i_size_read(inode);
 998         u64 page_start;
 999         u64 page_end;
1000         u64 page_cnt;
1001         int ret;
1002         int i;
1003         int i_done;
1004         struct btrfs_ordered_extent *ordered;
1005         struct extent_state *cached_state = NULL;
1006         struct extent_io_tree *tree;
1007         gfp_t mask = btrfs_alloc_write_mask(inode->i_mapping);
1008
1009         file_end = (isize - 1) >> PAGE_CACHE_SHIFT;
1010         if (!isize || start_index > file_end)
1011                 return 0;
1012
1013         page_cnt = min_t(u64, (u64)num_pages, (u64)file_end - start_index + 1);
1014
1015         ret = btrfs_delalloc_reserve_space(inode,
1016                                            page_cnt << PAGE_CACHE_SHIFT);
1017         if (ret)
1018                 return ret;
1019         i_done = 0;
1020         tree = &BTRFS_I(inode)->io_tree;
1021
1022         /* step one, lock all the pages */
1023         for (i = 0; i < page_cnt; i++) {
1024                 struct page *page;
1025 again:
1026                 page = find_or_create_page(inode->i_mapping,
1027                                            start_index + i, mask);
1028                 if (!page)
1029                         break;
1030
1031                 page_start = page_offset(page);
1032                 page_end = page_start + PAGE_CACHE_SIZE - 1;
1033                 while (1) {
1034                         lock_extent(tree, page_start, page_end);
1035                         ordered = btrfs_lookup_ordered_extent(inode,
1036                                                               page_start);
1037                         unlock_extent(tree, page_start, page_end);
1038                         if (!ordered)
1039                                 break;
1040
1041                         unlock_page(page);
1042                         btrfs_start_ordered_extent(inode, ordered, 1);
1043                         btrfs_put_ordered_extent(ordered);
1044                         lock_page(page);
1045                         /*
1046                          * we unlocked the page above, so we need check if
1047                          * it was released or not.
1048                          */
1049                         if (page->mapping != inode->i_mapping) {
1050                                 unlock_page(page);
1051                                 page_cache_release(page);
1052                                 goto again;
1053                         }
1054                 }
1055
1056                 if (!PageUptodate(page)) {
1057                         btrfs_readpage(NULL, page);
1058                         lock_page(page);
1059                         if (!PageUptodate(page)) {
1060                                 unlock_page(page);
1061                                 page_cache_release(page);
1062                                 ret = -EIO;
1063                                 break;
1064                         }
1065                 }
1066
1067                 if (page->mapping != inode->i_mapping) {
1068                         unlock_page(page);
1069                         page_cache_release(page);
1070                         goto again;
1071                 }
1072
1073                 pages[i] = page;
1074                 i_done++;
1075         }
1076         if (!i_done || ret)
1077                 goto out;
1078
1079         if (!(inode->i_sb->s_flags & MS_ACTIVE))
1080                 goto out;
1081
1082         /*
1083          * so now we have a nice long stream of locked
1084          * and up to date pages, lets wait on them
1085          */
1086         for (i = 0; i < i_done; i++)
1087                 wait_on_page_writeback(pages[i]);
1088
1089         page_start = page_offset(pages[0]);
1090         page_end = page_offset(pages[i_done - 1]) + PAGE_CACHE_SIZE;
1091
1092         lock_extent_bits(&BTRFS_I(inode)->io_tree,
1093                          page_start, page_end - 1, 0, &cached_state);
1094         clear_extent_bit(&BTRFS_I(inode)->io_tree, page_start,
1095                           page_end - 1, EXTENT_DIRTY | EXTENT_DELALLOC |
1096                           EXTENT_DO_ACCOUNTING | EXTENT_DEFRAG, 0, 0,
1097                           &cached_state, GFP_NOFS);
1098
1099         if (i_done != page_cnt) {
1100                 spin_lock(&BTRFS_I(inode)->lock);
1101                 BTRFS_I(inode)->outstanding_extents++;
1102                 spin_unlock(&BTRFS_I(inode)->lock);
1103                 btrfs_delalloc_release_space(inode,
1104                                      (page_cnt - i_done) << PAGE_CACHE_SHIFT);
1105         }
1106
1107
1108         set_extent_defrag(&BTRFS_I(inode)->io_tree, page_start, page_end - 1,
1109                           &cached_state, GFP_NOFS);
1110
1111         unlock_extent_cached(&BTRFS_I(inode)->io_tree,
1112                              page_start, page_end - 1, &cached_state,
1113                              GFP_NOFS);
1114
1115         for (i = 0; i < i_done; i++) {
1116                 clear_page_dirty_for_io(pages[i]);
1117                 ClearPageChecked(pages[i]);
1118                 set_page_extent_mapped(pages[i]);
1119                 set_page_dirty(pages[i]);
1120                 unlock_page(pages[i]);
1121                 page_cache_release(pages[i]);
1122         }
1123         return i_done;
1124 out:
1125         for (i = 0; i < i_done; i++) {
1126                 unlock_page(pages[i]);
1127                 page_cache_release(pages[i]);
1128         }
1129         btrfs_delalloc_release_space(inode, page_cnt << PAGE_CACHE_SHIFT);
1130         return ret;
1131
1132 }
1133
1134 int btrfs_defrag_file(struct inode *inode, struct file *file,
1135                       struct btrfs_ioctl_defrag_range_args *range,
1136                       u64 newer_than, unsigned long max_to_defrag)
1137 {
1138         struct btrfs_root *root = BTRFS_I(inode)->root;
1139         struct file_ra_state *ra = NULL;
1140         unsigned long last_index;
1141         u64 isize = i_size_read(inode);
1142         u64 last_len = 0;
1143         u64 skip = 0;
1144         u64 defrag_end = 0;
1145         u64 newer_off = range->start;
1146         unsigned long i;
1147         unsigned long ra_index = 0;
1148         int ret;
1149         int defrag_count = 0;
1150         int compress_type = BTRFS_COMPRESS_ZLIB;
1151         int extent_thresh = range->extent_thresh;
1152         int max_cluster = (256 * 1024) >> PAGE_CACHE_SHIFT;
1153         int cluster = max_cluster;
1154         u64 new_align = ~((u64)128 * 1024 - 1);
1155         struct page **pages = NULL;
1156
1157         if (extent_thresh == 0)
1158                 extent_thresh = 256 * 1024;
1159
1160         if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS) {
1161                 if (range->compress_type > BTRFS_COMPRESS_TYPES)
1162                         return -EINVAL;
1163                 if (range->compress_type)
1164                         compress_type = range->compress_type;
1165         }
1166
1167         if (isize == 0)
1168                 return 0;
1169
1170         /*
1171          * if we were not given a file, allocate a readahead
1172          * context
1173          */
1174         if (!file) {
1175                 ra = kzalloc(sizeof(*ra), GFP_NOFS);
1176                 if (!ra)
1177                         return -ENOMEM;
1178                 file_ra_state_init(ra, inode->i_mapping);
1179         } else {
1180                 ra = &file->f_ra;
1181         }
1182
1183         pages = kmalloc(sizeof(struct page *) * max_cluster,
1184                         GFP_NOFS);
1185         if (!pages) {
1186                 ret = -ENOMEM;
1187                 goto out_ra;
1188         }
1189
1190         /* find the last page to defrag */
1191         if (range->start + range->len > range->start) {
1192                 last_index = min_t(u64, isize - 1,
1193                          range->start + range->len - 1) >> PAGE_CACHE_SHIFT;
1194         } else {
1195                 last_index = (isize - 1) >> PAGE_CACHE_SHIFT;
1196         }
1197
1198         if (newer_than) {
1199                 ret = find_new_extents(root, inode, newer_than,
1200                                        &newer_off, 64 * 1024);
1201                 if (!ret) {
1202                         range->start = newer_off;
1203                         /*
1204                          * we always align our defrag to help keep
1205                          * the extents in the file evenly spaced
1206                          */
1207                         i = (newer_off & new_align) >> PAGE_CACHE_SHIFT;
1208                 } else
1209                         goto out_ra;
1210         } else {
1211                 i = range->start >> PAGE_CACHE_SHIFT;
1212         }
1213         if (!max_to_defrag)
1214                 max_to_defrag = last_index + 1;
1215
1216         /*
1217          * make writeback starts from i, so the defrag range can be
1218          * written sequentially.
1219          */
1220         if (i < inode->i_mapping->writeback_index)
1221                 inode->i_mapping->writeback_index = i;
1222
1223         while (i <= last_index && defrag_count < max_to_defrag &&
1224                (i < (i_size_read(inode) + PAGE_CACHE_SIZE - 1) >>
1225                 PAGE_CACHE_SHIFT)) {
1226                 /*
1227                  * make sure we stop running if someone unmounts
1228                  * the FS
1229                  */
1230                 if (!(inode->i_sb->s_flags & MS_ACTIVE))
1231                         break;
1232
1233                 if (btrfs_defrag_cancelled(root->fs_info)) {
1234                         printk(KERN_DEBUG "btrfs: defrag_file cancelled\n");
1235                         ret = -EAGAIN;
1236                         break;
1237                 }
1238
1239                 if (!should_defrag_range(inode, (u64)i << PAGE_CACHE_SHIFT,
1240                                          extent_thresh, &last_len, &skip,
1241                                          &defrag_end, range->flags &
1242                                          BTRFS_DEFRAG_RANGE_COMPRESS)) {
1243                         unsigned long next;
1244                         /*
1245                          * the should_defrag function tells us how much to skip
1246                          * bump our counter by the suggested amount
1247                          */
1248                         next = (skip + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT;
1249                         i = max(i + 1, next);
1250                         continue;
1251                 }
1252
1253                 if (!newer_than) {
1254                         cluster = (PAGE_CACHE_ALIGN(defrag_end) >>
1255                                    PAGE_CACHE_SHIFT) - i;
1256                         cluster = min(cluster, max_cluster);
1257                 } else {
1258                         cluster = max_cluster;
1259                 }
1260
1261                 if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)
1262                         BTRFS_I(inode)->force_compress = compress_type;
1263
1264                 if (i + cluster > ra_index) {
1265                         ra_index = max(i, ra_index);
1266                         btrfs_force_ra(inode->i_mapping, ra, file, ra_index,
1267                                        cluster);
1268                         ra_index += max_cluster;
1269                 }
1270
1271                 mutex_lock(&inode->i_mutex);
1272                 ret = cluster_pages_for_defrag(inode, pages, i, cluster);
1273                 if (ret < 0) {
1274                         mutex_unlock(&inode->i_mutex);
1275                         goto out_ra;
1276                 }
1277
1278                 defrag_count += ret;
1279                 balance_dirty_pages_ratelimited(inode->i_mapping);
1280                 mutex_unlock(&inode->i_mutex);
1281
1282                 if (newer_than) {
1283                         if (newer_off == (u64)-1)
1284                                 break;
1285
1286                         if (ret > 0)
1287                                 i += ret;
1288
1289                         newer_off = max(newer_off + 1,
1290                                         (u64)i << PAGE_CACHE_SHIFT);
1291
1292                         ret = find_new_extents(root, inode,
1293                                                newer_than, &newer_off,
1294                                                64 * 1024);
1295                         if (!ret) {
1296                                 range->start = newer_off;
1297                                 i = (newer_off & new_align) >> PAGE_CACHE_SHIFT;
1298                         } else {
1299                                 break;
1300                         }
1301                 } else {
1302                         if (ret > 0) {
1303                                 i += ret;
1304                                 last_len += ret << PAGE_CACHE_SHIFT;
1305                         } else {
1306                                 i++;
1307                                 last_len = 0;
1308                         }
1309                 }
1310         }
1311
1312         if ((range->flags & BTRFS_DEFRAG_RANGE_START_IO))
1313                 filemap_flush(inode->i_mapping);
1314
1315         if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
1316                 /* the filemap_flush will queue IO into the worker threads, but
1317                  * we have to make sure the IO is actually started and that
1318                  * ordered extents get created before we return
1319                  */
1320                 atomic_inc(&root->fs_info->async_submit_draining);
1321                 while (atomic_read(&root->fs_info->nr_async_submits) ||
1322                       atomic_read(&root->fs_info->async_delalloc_pages)) {
1323                         wait_event(root->fs_info->async_submit_wait,
1324                            (atomic_read(&root->fs_info->nr_async_submits) == 0 &&
1325                             atomic_read(&root->fs_info->async_delalloc_pages) == 0));
1326                 }
1327                 atomic_dec(&root->fs_info->async_submit_draining);
1328
1329                 mutex_lock(&inode->i_mutex);
1330                 BTRFS_I(inode)->force_compress = BTRFS_COMPRESS_NONE;
1331                 mutex_unlock(&inode->i_mutex);
1332         }
1333
1334         if (range->compress_type == BTRFS_COMPRESS_LZO) {
1335                 btrfs_set_fs_incompat(root->fs_info, COMPRESS_LZO);
1336         }
1337
1338         ret = defrag_count;
1339
1340 out_ra:
1341         if (!file)
1342                 kfree(ra);
1343         kfree(pages);
1344         return ret;
1345 }
1346
1347 static noinline int btrfs_ioctl_resize(struct file *file,
1348                                         void __user *arg)
1349 {
1350         u64 new_size;
1351         u64 old_size;
1352         u64 devid = 1;
1353         struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
1354         struct btrfs_ioctl_vol_args *vol_args;
1355         struct btrfs_trans_handle *trans;
1356         struct btrfs_device *device = NULL;
1357         char *sizestr;
1358         char *devstr = NULL;
1359         int ret = 0;
1360         int mod = 0;
1361
1362         if (!capable(CAP_SYS_ADMIN))
1363                 return -EPERM;
1364
1365         ret = mnt_want_write_file(file);
1366         if (ret)
1367                 return ret;
1368
1369         if (atomic_xchg(&root->fs_info->mutually_exclusive_operation_running,
1370                         1)) {
1371                 pr_info("btrfs: dev add/delete/balance/replace/resize operation in progress\n");
1372                 mnt_drop_write_file(file);
1373                 return -EINVAL;
1374         }
1375
1376         mutex_lock(&root->fs_info->volume_mutex);
1377         vol_args = memdup_user(arg, sizeof(*vol_args));
1378         if (IS_ERR(vol_args)) {
1379                 ret = PTR_ERR(vol_args);
1380                 goto out;
1381         }
1382
1383         vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1384
1385         sizestr = vol_args->name;
1386         devstr = strchr(sizestr, ':');
1387         if (devstr) {
1388                 char *end;
1389                 sizestr = devstr + 1;
1390                 *devstr = '\0';
1391                 devstr = vol_args->name;
1392                 devid = simple_strtoull(devstr, &end, 10);
1393                 if (!devid) {
1394                         ret = -EINVAL;
1395                         goto out_free;
1396                 }
1397                 printk(KERN_INFO "btrfs: resizing devid %llu\n",
1398                        (unsigned long long)devid);
1399         }
1400
1401         device = btrfs_find_device(root->fs_info, devid, NULL, NULL);
1402         if (!device) {
1403                 printk(KERN_INFO "btrfs: resizer unable to find device %llu\n",
1404                        (unsigned long long)devid);
1405                 ret = -ENODEV;
1406                 goto out_free;
1407         }
1408
1409         if (!device->writeable) {
1410                 printk(KERN_INFO "btrfs: resizer unable to apply on "
1411                        "readonly device %llu\n",
1412                        (unsigned long long)devid);
1413                 ret = -EPERM;
1414                 goto out_free;
1415         }
1416
1417         if (!strcmp(sizestr, "max"))
1418                 new_size = device->bdev->bd_inode->i_size;
1419         else {
1420                 if (sizestr[0] == '-') {
1421                         mod = -1;
1422                         sizestr++;
1423                 } else if (sizestr[0] == '+') {
1424                         mod = 1;
1425                         sizestr++;
1426                 }
1427                 new_size = memparse(sizestr, NULL);
1428                 if (new_size == 0) {
1429                         ret = -EINVAL;
1430                         goto out_free;
1431                 }
1432         }
1433
1434         if (device->is_tgtdev_for_dev_replace) {
1435                 ret = -EPERM;
1436                 goto out_free;
1437         }
1438
1439         old_size = device->total_bytes;
1440
1441         if (mod < 0) {
1442                 if (new_size > old_size) {
1443                         ret = -EINVAL;
1444                         goto out_free;
1445                 }
1446                 new_size = old_size - new_size;
1447         } else if (mod > 0) {
1448                 new_size = old_size + new_size;
1449         }
1450
1451         if (new_size < 256 * 1024 * 1024) {
1452                 ret = -EINVAL;
1453                 goto out_free;
1454         }
1455         if (new_size > device->bdev->bd_inode->i_size) {
1456                 ret = -EFBIG;
1457                 goto out_free;
1458         }
1459
1460         do_div(new_size, root->sectorsize);
1461         new_size *= root->sectorsize;
1462
1463         printk_in_rcu(KERN_INFO "btrfs: new size for %s is %llu\n",
1464                       rcu_str_deref(device->name),
1465                       (unsigned long long)new_size);
1466
1467         if (new_size > old_size) {
1468                 trans = btrfs_start_transaction(root, 0);
1469                 if (IS_ERR(trans)) {
1470                         ret = PTR_ERR(trans);
1471                         goto out_free;
1472                 }
1473                 ret = btrfs_grow_device(trans, device, new_size);
1474                 btrfs_commit_transaction(trans, root);
1475         } else if (new_size < old_size) {
1476                 ret = btrfs_shrink_device(device, new_size);
1477         } /* equal, nothing need to do */
1478
1479 out_free:
1480         kfree(vol_args);
1481 out:
1482         mutex_unlock(&root->fs_info->volume_mutex);
1483         atomic_set(&root->fs_info->mutually_exclusive_operation_running, 0);
1484         mnt_drop_write_file(file);
1485         return ret;
1486 }
1487
1488 static noinline int btrfs_ioctl_snap_create_transid(struct file *file,
1489                                 char *name, unsigned long fd, int subvol,
1490                                 u64 *transid, bool readonly,
1491                                 struct btrfs_qgroup_inherit *inherit)
1492 {
1493         int namelen;
1494         int ret = 0;
1495
1496         ret = mnt_want_write_file(file);
1497         if (ret)
1498                 goto out;
1499
1500         namelen = strlen(name);
1501         if (strchr(name, '/')) {
1502                 ret = -EINVAL;
1503                 goto out_drop_write;
1504         }
1505
1506         if (name[0] == '.' &&
1507            (namelen == 1 || (name[1] == '.' && namelen == 2))) {
1508                 ret = -EEXIST;
1509                 goto out_drop_write;
1510         }
1511
1512         if (subvol) {
1513                 ret = btrfs_mksubvol(&file->f_path, name, namelen,
1514                                      NULL, transid, readonly, inherit);
1515         } else {
1516                 struct fd src = fdget(fd);
1517                 struct inode *src_inode;
1518                 if (!src.file) {
1519                         ret = -EINVAL;
1520                         goto out_drop_write;
1521                 }
1522
1523                 src_inode = file_inode(src.file);
1524                 if (src_inode->i_sb != file_inode(file)->i_sb) {
1525                         printk(KERN_INFO "btrfs: Snapshot src from "
1526                                "another FS\n");
1527                         ret = -EINVAL;
1528                 } else {
1529                         ret = btrfs_mksubvol(&file->f_path, name, namelen,
1530                                              BTRFS_I(src_inode)->root,
1531                                              transid, readonly, inherit);
1532                 }
1533                 fdput(src);
1534         }
1535 out_drop_write:
1536         mnt_drop_write_file(file);
1537 out:
1538         return ret;
1539 }
1540
1541 static noinline int btrfs_ioctl_snap_create(struct file *file,
1542                                             void __user *arg, int subvol)
1543 {
1544         struct btrfs_ioctl_vol_args *vol_args;
1545         int ret;
1546
1547         vol_args = memdup_user(arg, sizeof(*vol_args));
1548         if (IS_ERR(vol_args))
1549                 return PTR_ERR(vol_args);
1550         vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1551
1552         ret = btrfs_ioctl_snap_create_transid(file, vol_args->name,
1553                                               vol_args->fd, subvol,
1554                                               NULL, false, NULL);
1555
1556         kfree(vol_args);
1557         return ret;
1558 }
1559
1560 static noinline int btrfs_ioctl_snap_create_v2(struct file *file,
1561                                                void __user *arg, int subvol)
1562 {
1563         struct btrfs_ioctl_vol_args_v2 *vol_args;
1564         int ret;
1565         u64 transid = 0;
1566         u64 *ptr = NULL;
1567         bool readonly = false;
1568         struct btrfs_qgroup_inherit *inherit = NULL;
1569
1570         vol_args = memdup_user(arg, sizeof(*vol_args));
1571         if (IS_ERR(vol_args))
1572                 return PTR_ERR(vol_args);
1573         vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0';
1574
1575         if (vol_args->flags &
1576             ~(BTRFS_SUBVOL_CREATE_ASYNC | BTRFS_SUBVOL_RDONLY |
1577               BTRFS_SUBVOL_QGROUP_INHERIT)) {
1578                 ret = -EOPNOTSUPP;
1579                 goto out;
1580         }
1581
1582         if (vol_args->flags & BTRFS_SUBVOL_CREATE_ASYNC)
1583                 ptr = &transid;
1584         if (vol_args->flags & BTRFS_SUBVOL_RDONLY)
1585                 readonly = true;
1586         if (vol_args->flags & BTRFS_SUBVOL_QGROUP_INHERIT) {
1587                 if (vol_args->size > PAGE_CACHE_SIZE) {
1588                         ret = -EINVAL;
1589                         goto out;
1590                 }
1591                 inherit = memdup_user(vol_args->qgroup_inherit, vol_args->size);
1592                 if (IS_ERR(inherit)) {
1593                         ret = PTR_ERR(inherit);
1594                         goto out;
1595                 }
1596         }
1597
1598         ret = btrfs_ioctl_snap_create_transid(file, vol_args->name,
1599                                               vol_args->fd, subvol, ptr,
1600                                               readonly, inherit);
1601
1602         if (ret == 0 && ptr &&
1603             copy_to_user(arg +
1604                          offsetof(struct btrfs_ioctl_vol_args_v2,
1605                                   transid), ptr, sizeof(*ptr)))
1606                 ret = -EFAULT;
1607 out:
1608         kfree(vol_args);
1609         kfree(inherit);
1610         return ret;
1611 }
1612
1613 static noinline int btrfs_ioctl_subvol_getflags(struct file *file,
1614                                                 void __user *arg)
1615 {
1616         struct inode *inode = file_inode(file);
1617         struct btrfs_root *root = BTRFS_I(inode)->root;
1618         int ret = 0;
1619         u64 flags = 0;
1620
1621         if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID)
1622                 return -EINVAL;
1623
1624         down_read(&root->fs_info->subvol_sem);
1625         if (btrfs_root_readonly(root))
1626                 flags |= BTRFS_SUBVOL_RDONLY;
1627         up_read(&root->fs_info->subvol_sem);
1628
1629         if (copy_to_user(arg, &flags, sizeof(flags)))
1630                 ret = -EFAULT;
1631
1632         return ret;
1633 }
1634
1635 static noinline int btrfs_ioctl_subvol_setflags(struct file *file,
1636                                               void __user *arg)
1637 {
1638         struct inode *inode = file_inode(file);
1639         struct btrfs_root *root = BTRFS_I(inode)->root;
1640         struct btrfs_trans_handle *trans;
1641         u64 root_flags;
1642         u64 flags;
1643         int ret = 0;
1644
1645         ret = mnt_want_write_file(file);
1646         if (ret)
1647                 goto out;
1648
1649         if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) {
1650                 ret = -EINVAL;
1651                 goto out_drop_write;
1652         }
1653
1654         if (copy_from_user(&flags, arg, sizeof(flags))) {
1655                 ret = -EFAULT;
1656                 goto out_drop_write;
1657         }
1658
1659         if (flags & BTRFS_SUBVOL_CREATE_ASYNC) {
1660                 ret = -EINVAL;
1661                 goto out_drop_write;
1662         }
1663
1664         if (flags & ~BTRFS_SUBVOL_RDONLY) {
1665                 ret = -EOPNOTSUPP;
1666                 goto out_drop_write;
1667         }
1668
1669         if (!inode_owner_or_capable(inode)) {
1670                 ret = -EACCES;
1671                 goto out_drop_write;
1672         }
1673
1674         down_write(&root->fs_info->subvol_sem);
1675
1676         /* nothing to do */
1677         if (!!(flags & BTRFS_SUBVOL_RDONLY) == btrfs_root_readonly(root))
1678                 goto out_drop_sem;
1679
1680         root_flags = btrfs_root_flags(&root->root_item);
1681         if (flags & BTRFS_SUBVOL_RDONLY)
1682                 btrfs_set_root_flags(&root->root_item,
1683                                      root_flags | BTRFS_ROOT_SUBVOL_RDONLY);
1684         else
1685                 btrfs_set_root_flags(&root->root_item,
1686                                      root_flags & ~BTRFS_ROOT_SUBVOL_RDONLY);
1687
1688         trans = btrfs_start_transaction(root, 1);
1689         if (IS_ERR(trans)) {
1690                 ret = PTR_ERR(trans);
1691                 goto out_reset;
1692         }
1693
1694         ret = btrfs_update_root(trans, root->fs_info->tree_root,
1695                                 &root->root_key, &root->root_item);
1696
1697         btrfs_commit_transaction(trans, root);
1698 out_reset:
1699         if (ret)
1700                 btrfs_set_root_flags(&root->root_item, root_flags);
1701 out_drop_sem:
1702         up_write(&root->fs_info->subvol_sem);
1703 out_drop_write:
1704         mnt_drop_write_file(file);
1705 out:
1706         return ret;
1707 }
1708
1709 /*
1710  * helper to check if the subvolume references other subvolumes
1711  */
1712 static noinline int may_destroy_subvol(struct btrfs_root *root)
1713 {
1714         struct btrfs_path *path;
1715         struct btrfs_key key;
1716         int ret;
1717
1718         path = btrfs_alloc_path();
1719         if (!path)
1720                 return -ENOMEM;
1721
1722         key.objectid = root->root_key.objectid;
1723         key.type = BTRFS_ROOT_REF_KEY;
1724         key.offset = (u64)-1;
1725
1726         ret = btrfs_search_slot(NULL, root->fs_info->tree_root,
1727                                 &key, path, 0, 0);
1728         if (ret < 0)
1729                 goto out;
1730         BUG_ON(ret == 0);
1731
1732         ret = 0;
1733         if (path->slots[0] > 0) {
1734                 path->slots[0]--;
1735                 btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
1736                 if (key.objectid == root->root_key.objectid &&
1737                     key.type == BTRFS_ROOT_REF_KEY)
1738                         ret = -ENOTEMPTY;
1739         }
1740 out:
1741         btrfs_free_path(path);
1742         return ret;
1743 }
1744
1745 static noinline int key_in_sk(struct btrfs_key *key,
1746                               struct btrfs_ioctl_search_key *sk)
1747 {
1748         struct btrfs_key test;
1749         int ret;
1750
1751         test.objectid = sk->min_objectid;
1752         test.type = sk->min_type;
1753         test.offset = sk->min_offset;
1754
1755         ret = btrfs_comp_cpu_keys(key, &test);
1756         if (ret < 0)
1757                 return 0;
1758
1759         test.objectid = sk->max_objectid;
1760         test.type = sk->max_type;
1761         test.offset = sk->max_offset;
1762
1763         ret = btrfs_comp_cpu_keys(key, &test);
1764         if (ret > 0)
1765                 return 0;
1766         return 1;
1767 }
1768
1769 static noinline int copy_to_sk(struct btrfs_root *root,
1770                                struct btrfs_path *path,
1771                                struct btrfs_key *key,
1772                                struct btrfs_ioctl_search_key *sk,
1773                                char *buf,
1774                                unsigned long *sk_offset,
1775                                int *num_found)
1776 {
1777         u64 found_transid;
1778         struct extent_buffer *leaf;
1779         struct btrfs_ioctl_search_header sh;
1780         unsigned long item_off;
1781         unsigned long item_len;
1782         int nritems;
1783         int i;
1784         int slot;
1785         int ret = 0;
1786
1787         leaf = path->nodes[0];
1788         slot = path->slots[0];
1789         nritems = btrfs_header_nritems(leaf);
1790
1791         if (btrfs_header_generation(leaf) > sk->max_transid) {
1792                 i = nritems;
1793                 goto advance_key;
1794         }
1795         found_transid = btrfs_header_generation(leaf);
1796
1797         for (i = slot; i < nritems; i++) {
1798                 item_off = btrfs_item_ptr_offset(leaf, i);
1799                 item_len = btrfs_item_size_nr(leaf, i);
1800
1801                 if (item_len > BTRFS_SEARCH_ARGS_BUFSIZE)
1802                         item_len = 0;
1803
1804                 if (sizeof(sh) + item_len + *sk_offset >
1805                     BTRFS_SEARCH_ARGS_BUFSIZE) {
1806                         ret = 1;
1807                         goto overflow;
1808                 }
1809
1810                 btrfs_item_key_to_cpu(leaf, key, i);
1811                 if (!key_in_sk(key, sk))
1812                         continue;
1813
1814                 sh.objectid = key->objectid;
1815                 sh.offset = key->offset;
1816                 sh.type = key->type;
1817                 sh.len = item_len;
1818                 sh.transid = found_transid;
1819
1820                 /* copy search result header */
1821                 memcpy(buf + *sk_offset, &sh, sizeof(sh));
1822                 *sk_offset += sizeof(sh);
1823
1824                 if (item_len) {
1825                         char *p = buf + *sk_offset;
1826                         /* copy the item */
1827                         read_extent_buffer(leaf, p,
1828                                            item_off, item_len);
1829                         *sk_offset += item_len;
1830                 }
1831                 (*num_found)++;
1832
1833                 if (*num_found >= sk->nr_items)
1834                         break;
1835         }
1836 advance_key:
1837         ret = 0;
1838         if (key->offset < (u64)-1 && key->offset < sk->max_offset)
1839                 key->offset++;
1840         else if (key->type < (u8)-1 && key->type < sk->max_type) {
1841                 key->offset = 0;
1842                 key->type++;
1843         } else if (key->objectid < (u64)-1 && key->objectid < sk->max_objectid) {
1844                 key->offset = 0;
1845                 key->type = 0;
1846                 key->objectid++;
1847         } else
1848                 ret = 1;
1849 overflow:
1850         return ret;
1851 }
1852
1853 static noinline int search_ioctl(struct inode *inode,
1854                                  struct btrfs_ioctl_search_args *args)
1855 {
1856         struct btrfs_root *root;
1857         struct btrfs_key key;
1858         struct btrfs_key max_key;
1859         struct btrfs_path *path;
1860         struct btrfs_ioctl_search_key *sk = &args->key;
1861         struct btrfs_fs_info *info = BTRFS_I(inode)->root->fs_info;
1862         int ret;
1863         int num_found = 0;
1864         unsigned long sk_offset = 0;
1865
1866         path = btrfs_alloc_path();
1867         if (!path)
1868                 return -ENOMEM;
1869
1870         if (sk->tree_id == 0) {
1871                 /* search the root of the inode that was passed */
1872                 root = BTRFS_I(inode)->root;
1873         } else {
1874                 key.objectid = sk->tree_id;
1875                 key.type = BTRFS_ROOT_ITEM_KEY;
1876                 key.offset = (u64)-1;
1877                 root = btrfs_read_fs_root_no_name(info, &key);
1878                 if (IS_ERR(root)) {
1879                         printk(KERN_ERR "could not find root %llu\n",
1880                                sk->tree_id);
1881                         btrfs_free_path(path);
1882                         return -ENOENT;
1883                 }
1884         }
1885
1886         key.objectid = sk->min_objectid;
1887         key.type = sk->min_type;
1888         key.offset = sk->min_offset;
1889
1890         max_key.objectid = sk->max_objectid;
1891         max_key.type = sk->max_type;
1892         max_key.offset = sk->max_offset;
1893
1894         path->keep_locks = 1;
1895
1896         while(1) {
1897                 ret = btrfs_search_forward(root, &key, &max_key, path,
1898                                            sk->min_transid);
1899                 if (ret != 0) {
1900                         if (ret > 0)
1901                                 ret = 0;
1902                         goto err;
1903                 }
1904                 ret = copy_to_sk(root, path, &key, sk, args->buf,
1905                                  &sk_offset, &num_found);
1906                 btrfs_release_path(path);
1907                 if (ret || num_found >= sk->nr_items)
1908                         break;
1909
1910         }
1911         ret = 0;
1912 err:
1913         sk->nr_items = num_found;
1914         btrfs_free_path(path);
1915         return ret;
1916 }
1917
1918 static noinline int btrfs_ioctl_tree_search(struct file *file,
1919                                            void __user *argp)
1920 {
1921          struct btrfs_ioctl_search_args *args;
1922          struct inode *inode;
1923          int ret;
1924
1925         if (!capable(CAP_SYS_ADMIN))
1926                 return -EPERM;
1927
1928         args = memdup_user(argp, sizeof(*args));
1929         if (IS_ERR(args))
1930                 return PTR_ERR(args);
1931
1932         inode = file_inode(file);
1933         ret = search_ioctl(inode, args);
1934         if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
1935                 ret = -EFAULT;
1936         kfree(args);
1937         return ret;
1938 }
1939
1940 /*
1941  * Search INODE_REFs to identify path name of 'dirid' directory
1942  * in a 'tree_id' tree. and sets path name to 'name'.
1943  */
1944 static noinline int btrfs_search_path_in_tree(struct btrfs_fs_info *info,
1945                                 u64 tree_id, u64 dirid, char *name)
1946 {
1947         struct btrfs_root *root;
1948         struct btrfs_key key;
1949         char *ptr;
1950         int ret = -1;
1951         int slot;
1952         int len;
1953         int total_len = 0;
1954         struct btrfs_inode_ref *iref;
1955         struct extent_buffer *l;
1956         struct btrfs_path *path;
1957
1958         if (dirid == BTRFS_FIRST_FREE_OBJECTID) {
1959                 name[0]='\0';
1960                 return 0;
1961         }
1962
1963         path = btrfs_alloc_path();
1964         if (!path)
1965                 return -ENOMEM;
1966
1967         ptr = &name[BTRFS_INO_LOOKUP_PATH_MAX];
1968
1969         key.objectid = tree_id;
1970         key.type = BTRFS_ROOT_ITEM_KEY;
1971         key.offset = (u64)-1;
1972         root = btrfs_read_fs_root_no_name(info, &key);
1973         if (IS_ERR(root)) {
1974                 printk(KERN_ERR "could not find root %llu\n", tree_id);
1975                 ret = -ENOENT;
1976                 goto out;
1977         }
1978
1979         key.objectid = dirid;
1980         key.type = BTRFS_INODE_REF_KEY;
1981         key.offset = (u64)-1;
1982
1983         while(1) {
1984                 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
1985                 if (ret < 0)
1986                         goto out;
1987
1988                 l = path->nodes[0];
1989                 slot = path->slots[0];
1990                 if (ret > 0 && slot > 0)
1991                         slot--;
1992                 btrfs_item_key_to_cpu(l, &key, slot);
1993
1994                 if (ret > 0 && (key.objectid != dirid ||
1995                                 key.type != BTRFS_INODE_REF_KEY)) {
1996                         ret = -ENOENT;
1997                         goto out;
1998                 }
1999
2000                 iref = btrfs_item_ptr(l, slot, struct btrfs_inode_ref);
2001                 len = btrfs_inode_ref_name_len(l, iref);
2002                 ptr -= len + 1;
2003                 total_len += len + 1;
2004                 if (ptr < name)
2005                         goto out;
2006
2007                 *(ptr + len) = '/';
2008                 read_extent_buffer(l, ptr,(unsigned long)(iref + 1), len);
2009
2010                 if (key.offset == BTRFS_FIRST_FREE_OBJECTID)
2011                         break;
2012
2013                 btrfs_release_path(path);
2014                 key.objectid = key.offset;
2015                 key.offset = (u64)-1;
2016                 dirid = key.objectid;
2017         }
2018         if (ptr < name)
2019                 goto out;
2020         memmove(name, ptr, total_len);
2021         name[total_len]='\0';
2022         ret = 0;
2023 out:
2024         btrfs_free_path(path);
2025         return ret;
2026 }
2027
2028 static noinline int btrfs_ioctl_ino_lookup(struct file *file,
2029                                            void __user *argp)
2030 {
2031          struct btrfs_ioctl_ino_lookup_args *args;
2032          struct inode *inode;
2033          int ret;
2034
2035         if (!capable(CAP_SYS_ADMIN))
2036                 return -EPERM;
2037
2038         args = memdup_user(argp, sizeof(*args));
2039         if (IS_ERR(args))
2040                 return PTR_ERR(args);
2041
2042         inode = file_inode(file);
2043
2044         if (args->treeid == 0)
2045                 args->treeid = BTRFS_I(inode)->root->root_key.objectid;
2046
2047         ret = btrfs_search_path_in_tree(BTRFS_I(inode)->root->fs_info,
2048                                         args->treeid, args->objectid,
2049                                         args->name);
2050
2051         if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
2052                 ret = -EFAULT;
2053
2054         kfree(args);
2055         return ret;
2056 }
2057
2058 static noinline int btrfs_ioctl_snap_destroy(struct file *file,
2059                                              void __user *arg)
2060 {
2061         struct dentry *parent = fdentry(file);
2062         struct dentry *dentry;
2063         struct inode *dir = parent->d_inode;
2064         struct inode *inode;
2065         struct btrfs_root *root = BTRFS_I(dir)->root;
2066         struct btrfs_root *dest = NULL;
2067         struct btrfs_ioctl_vol_args *vol_args;
2068         struct btrfs_trans_handle *trans;
2069         struct btrfs_block_rsv block_rsv;
2070         u64 qgroup_reserved;
2071         int namelen;
2072         int ret;
2073         int err = 0;
2074
2075         vol_args = memdup_user(arg, sizeof(*vol_args));
2076         if (IS_ERR(vol_args))
2077                 return PTR_ERR(vol_args);
2078
2079         vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2080         namelen = strlen(vol_args->name);
2081         if (strchr(vol_args->name, '/') ||
2082             strncmp(vol_args->name, "..", namelen) == 0) {
2083                 err = -EINVAL;
2084                 goto out;
2085         }
2086
2087         err = mnt_want_write_file(file);
2088         if (err)
2089                 goto out;
2090
2091         err = mutex_lock_killable_nested(&dir->i_mutex, I_MUTEX_PARENT);
2092         if (err == -EINTR)
2093                 goto out;
2094         dentry = lookup_one_len(vol_args->name, parent, namelen);
2095         if (IS_ERR(dentry)) {
2096                 err = PTR_ERR(dentry);
2097                 goto out_unlock_dir;
2098         }
2099
2100         if (!dentry->d_inode) {
2101                 err = -ENOENT;
2102                 goto out_dput;
2103         }
2104
2105         inode = dentry->d_inode;
2106         dest = BTRFS_I(inode)->root;
2107         if (!capable(CAP_SYS_ADMIN)){
2108                 /*
2109                  * Regular user.  Only allow this with a special mount
2110                  * option, when the user has write+exec access to the
2111                  * subvol root, and when rmdir(2) would have been
2112                  * allowed.
2113                  *
2114                  * Note that this is _not_ check that the subvol is
2115                  * empty or doesn't contain data that we wouldn't
2116                  * otherwise be able to delete.
2117                  *
2118                  * Users who want to delete empty subvols should try
2119                  * rmdir(2).
2120                  */
2121                 err = -EPERM;
2122                 if (!btrfs_test_opt(root, USER_SUBVOL_RM_ALLOWED))
2123                         goto out_dput;
2124
2125                 /*
2126                  * Do not allow deletion if the parent dir is the same
2127                  * as the dir to be deleted.  That means the ioctl
2128                  * must be called on the dentry referencing the root
2129                  * of the subvol, not a random directory contained
2130                  * within it.
2131                  */
2132                 err = -EINVAL;
2133                 if (root == dest)
2134                         goto out_dput;
2135
2136                 err = inode_permission(inode, MAY_WRITE | MAY_EXEC);
2137                 if (err)
2138                         goto out_dput;
2139         }
2140
2141         /* check if subvolume may be deleted by a user */
2142         err = btrfs_may_delete(dir, dentry, 1);
2143         if (err)
2144                 goto out_dput;
2145
2146         if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) {
2147                 err = -EINVAL;
2148                 goto out_dput;
2149         }
2150
2151         mutex_lock(&inode->i_mutex);
2152         err = d_invalidate(dentry);
2153         if (err)
2154                 goto out_unlock;
2155
2156         down_write(&root->fs_info->subvol_sem);
2157
2158         err = may_destroy_subvol(dest);
2159         if (err)
2160                 goto out_up_write;
2161
2162         btrfs_init_block_rsv(&block_rsv, BTRFS_BLOCK_RSV_TEMP);
2163         /*
2164          * One for dir inode, two for dir entries, two for root
2165          * ref/backref.
2166          */
2167         err = btrfs_subvolume_reserve_metadata(root, &block_rsv,
2168                                                5, &qgroup_reserved);
2169         if (err)
2170                 goto out_up_write;
2171
2172         trans = btrfs_start_transaction(root, 0);
2173         if (IS_ERR(trans)) {
2174                 err = PTR_ERR(trans);
2175                 goto out_release;
2176         }
2177         trans->block_rsv = &block_rsv;
2178         trans->bytes_reserved = block_rsv.size;
2179
2180         ret = btrfs_unlink_subvol(trans, root, dir,
2181                                 dest->root_key.objectid,
2182                                 dentry->d_name.name,
2183                                 dentry->d_name.len);
2184         if (ret) {
2185                 err = ret;
2186                 btrfs_abort_transaction(trans, root, ret);
2187                 goto out_end_trans;
2188         }
2189
2190         btrfs_record_root_in_trans(trans, dest);
2191
2192         memset(&dest->root_item.drop_progress, 0,
2193                 sizeof(dest->root_item.drop_progress));
2194         dest->root_item.drop_level = 0;
2195         btrfs_set_root_refs(&dest->root_item, 0);
2196
2197         if (!xchg(&dest->orphan_item_inserted, 1)) {
2198                 ret = btrfs_insert_orphan_item(trans,
2199                                         root->fs_info->tree_root,
2200                                         dest->root_key.objectid);
2201                 if (ret) {
2202                         btrfs_abort_transaction(trans, root, ret);
2203                         err = ret;
2204                         goto out_end_trans;
2205                 }
2206         }
2207 out_end_trans:
2208         trans->block_rsv = NULL;
2209         trans->bytes_reserved = 0;
2210         ret = btrfs_end_transaction(trans, root);
2211         if (ret && !err)
2212                 err = ret;
2213         inode->i_flags |= S_DEAD;
2214 out_release:
2215         btrfs_subvolume_release_metadata(root, &block_rsv, qgroup_reserved);
2216 out_up_write:
2217         up_write(&root->fs_info->subvol_sem);
2218 out_unlock:
2219         mutex_unlock(&inode->i_mutex);
2220         if (!err) {
2221                 shrink_dcache_sb(root->fs_info->sb);
2222                 btrfs_invalidate_inodes(dest);
2223                 d_delete(dentry);
2224
2225                 /* the last ref */
2226                 if (dest->cache_inode) {
2227                         iput(dest->cache_inode);
2228                         dest->cache_inode = NULL;
2229                 }
2230         }
2231 out_dput:
2232         dput(dentry);
2233 out_unlock_dir:
2234         mutex_unlock(&dir->i_mutex);
2235         mnt_drop_write_file(file);
2236 out:
2237         kfree(vol_args);
2238         return err;
2239 }
2240
2241 static int btrfs_ioctl_defrag(struct file *file, void __user *argp)
2242 {
2243         struct inode *inode = file_inode(file);
2244         struct btrfs_root *root = BTRFS_I(inode)->root;
2245         struct btrfs_ioctl_defrag_range_args *range;
2246         int ret;
2247
2248         ret = mnt_want_write_file(file);
2249         if (ret)
2250                 return ret;
2251
2252         if (btrfs_root_readonly(root)) {
2253                 ret = -EROFS;
2254                 goto out;
2255         }
2256
2257         switch (inode->i_mode & S_IFMT) {
2258         case S_IFDIR:
2259                 if (!capable(CAP_SYS_ADMIN)) {
2260                         ret = -EPERM;
2261                         goto out;
2262                 }
2263                 ret = btrfs_defrag_root(root);
2264                 if (ret)
2265                         goto out;
2266                 ret = btrfs_defrag_root(root->fs_info->extent_root);
2267                 break;
2268         case S_IFREG:
2269                 if (!(file->f_mode & FMODE_WRITE)) {
2270                         ret = -EINVAL;
2271                         goto out;
2272                 }
2273
2274                 range = kzalloc(sizeof(*range), GFP_KERNEL);
2275                 if (!range) {
2276                         ret = -ENOMEM;
2277                         goto out;
2278                 }
2279
2280                 if (argp) {
2281                         if (copy_from_user(range, argp,
2282                                            sizeof(*range))) {
2283                                 ret = -EFAULT;
2284                                 kfree(range);
2285                                 goto out;
2286                         }
2287                         /* compression requires us to start the IO */
2288                         if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
2289                                 range->flags |= BTRFS_DEFRAG_RANGE_START_IO;
2290                                 range->extent_thresh = (u32)-1;
2291                         }
2292                 } else {
2293                         /* the rest are all set to zero by kzalloc */
2294                         range->len = (u64)-1;
2295                 }
2296                 ret = btrfs_defrag_file(file_inode(file), file,
2297                                         range, 0, 0);
2298                 if (ret > 0)
2299                         ret = 0;
2300                 kfree(range);
2301                 break;
2302         default:
2303                 ret = -EINVAL;
2304         }
2305 out:
2306         mnt_drop_write_file(file);
2307         return ret;
2308 }
2309
2310 static long btrfs_ioctl_add_dev(struct btrfs_root *root, void __user *arg)
2311 {
2312         struct btrfs_ioctl_vol_args *vol_args;
2313         int ret;
2314
2315         if (!capable(CAP_SYS_ADMIN))
2316                 return -EPERM;
2317
2318         if (atomic_xchg(&root->fs_info->mutually_exclusive_operation_running,
2319                         1)) {
2320                 pr_info("btrfs: dev add/delete/balance/replace/resize operation in progress\n");
2321                 return -EINVAL;
2322         }
2323
2324         mutex_lock(&root->fs_info->volume_mutex);
2325         vol_args = memdup_user(arg, sizeof(*vol_args));
2326         if (IS_ERR(vol_args)) {
2327                 ret = PTR_ERR(vol_args);
2328                 goto out;
2329         }
2330
2331         vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2332         ret = btrfs_init_new_device(root, vol_args->name);
2333
2334         kfree(vol_args);
2335 out:
2336         mutex_unlock(&root->fs_info->volume_mutex);
2337         atomic_set(&root->fs_info->mutually_exclusive_operation_running, 0);
2338         return ret;
2339 }
2340
2341 static long btrfs_ioctl_rm_dev(struct file *file, void __user *arg)
2342 {
2343         struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
2344         struct btrfs_ioctl_vol_args *vol_args;
2345         int ret;
2346
2347         if (!capable(CAP_SYS_ADMIN))
2348                 return -EPERM;
2349
2350         ret = mnt_want_write_file(file);
2351         if (ret)
2352                 return ret;
2353
2354         if (atomic_xchg(&root->fs_info->mutually_exclusive_operation_running,
2355                         1)) {
2356                 pr_info("btrfs: dev add/delete/balance/replace/resize operation in progress\n");
2357                 mnt_drop_write_file(file);
2358                 return -EINVAL;
2359         }
2360
2361         mutex_lock(&root->fs_info->volume_mutex);
2362         vol_args = memdup_user(arg, sizeof(*vol_args));
2363         if (IS_ERR(vol_args)) {
2364                 ret = PTR_ERR(vol_args);
2365                 goto out;
2366         }
2367
2368         vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2369         ret = btrfs_rm_device(root, vol_args->name);
2370
2371         kfree(vol_args);
2372 out:
2373         mutex_unlock(&root->fs_info->volume_mutex);
2374         atomic_set(&root->fs_info->mutually_exclusive_operation_running, 0);
2375         mnt_drop_write_file(file);
2376         return ret;
2377 }
2378
2379 static long btrfs_ioctl_fs_info(struct btrfs_root *root, void __user *arg)
2380 {
2381         struct btrfs_ioctl_fs_info_args *fi_args;
2382         struct btrfs_device *device;
2383         struct btrfs_device *next;
2384         struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
2385         int ret = 0;
2386
2387         if (!capable(CAP_SYS_ADMIN))
2388                 return -EPERM;
2389
2390         fi_args = kzalloc(sizeof(*fi_args), GFP_KERNEL);
2391         if (!fi_args)
2392                 return -ENOMEM;
2393
2394         fi_args->num_devices = fs_devices->num_devices;
2395         memcpy(&fi_args->fsid, root->fs_info->fsid, sizeof(fi_args->fsid));
2396
2397         mutex_lock(&fs_devices->device_list_mutex);
2398         list_for_each_entry_safe(device, next, &fs_devices->devices, dev_list) {
2399                 if (device->devid > fi_args->max_id)
2400                         fi_args->max_id = device->devid;
2401         }
2402         mutex_unlock(&fs_devices->device_list_mutex);
2403
2404         if (copy_to_user(arg, fi_args, sizeof(*fi_args)))
2405                 ret = -EFAULT;
2406
2407         kfree(fi_args);
2408         return ret;
2409 }
2410
2411 static long btrfs_ioctl_dev_info(struct btrfs_root *root, void __user *arg)
2412 {
2413         struct btrfs_ioctl_dev_info_args *di_args;
2414         struct btrfs_device *dev;
2415         struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
2416         int ret = 0;
2417         char *s_uuid = NULL;
2418         char empty_uuid[BTRFS_UUID_SIZE] = {0};
2419
2420         if (!capable(CAP_SYS_ADMIN))
2421                 return -EPERM;
2422
2423         di_args = memdup_user(arg, sizeof(*di_args));
2424         if (IS_ERR(di_args))
2425                 return PTR_ERR(di_args);
2426
2427         if (memcmp(empty_uuid, di_args->uuid, BTRFS_UUID_SIZE) != 0)
2428                 s_uuid = di_args->uuid;
2429
2430         mutex_lock(&fs_devices->device_list_mutex);
2431         dev = btrfs_find_device(root->fs_info, di_args->devid, s_uuid, NULL);
2432         mutex_unlock(&fs_devices->device_list_mutex);
2433
2434         if (!dev) {
2435                 ret = -ENODEV;
2436                 goto out;
2437         }
2438
2439         di_args->devid = dev->devid;
2440         di_args->bytes_used = dev->bytes_used;
2441         di_args->total_bytes = dev->total_bytes;
2442         memcpy(di_args->uuid, dev->uuid, sizeof(di_args->uuid));
2443         if (dev->name) {
2444                 struct rcu_string *name;
2445
2446                 rcu_read_lock();
2447                 name = rcu_dereference(dev->name);
2448                 strncpy(di_args->path, name->str, sizeof(di_args->path));
2449                 rcu_read_unlock();
2450                 di_args->path[sizeof(di_args->path) - 1] = 0;
2451         } else {
2452                 di_args->path[0] = '\0';
2453         }
2454
2455 out:
2456         if (ret == 0 && copy_to_user(arg, di_args, sizeof(*di_args)))
2457                 ret = -EFAULT;
2458
2459         kfree(di_args);
2460         return ret;
2461 }
2462
2463 static noinline long btrfs_ioctl_clone(struct file *file, unsigned long srcfd,
2464                                        u64 off, u64 olen, u64 destoff)
2465 {
2466         struct inode *inode = file_inode(file);
2467         struct btrfs_root *root = BTRFS_I(inode)->root;
2468         struct fd src_file;
2469         struct inode *src;
2470         struct btrfs_trans_handle *trans;
2471         struct btrfs_path *path;
2472         struct extent_buffer *leaf;
2473         char *buf;
2474         struct btrfs_key key;
2475         u32 nritems;
2476         int slot;
2477         int ret;
2478         u64 len = olen;
2479         u64 bs = root->fs_info->sb->s_blocksize;
2480
2481         /*
2482          * TODO:
2483          * - split compressed inline extents.  annoying: we need to
2484          *   decompress into destination's address_space (the file offset
2485          *   may change, so source mapping won't do), then recompress (or
2486          *   otherwise reinsert) a subrange.
2487          * - allow ranges within the same file to be cloned (provided
2488          *   they don't overlap)?
2489          */
2490
2491         /* the destination must be opened for writing */
2492         if (!(file->f_mode & FMODE_WRITE) || (file->f_flags & O_APPEND))
2493                 return -EINVAL;
2494
2495         if (btrfs_root_readonly(root))
2496                 return -EROFS;
2497
2498         ret = mnt_want_write_file(file);
2499         if (ret)
2500                 return ret;
2501
2502         src_file = fdget(srcfd);
2503         if (!src_file.file) {
2504                 ret = -EBADF;
2505                 goto out_drop_write;
2506         }
2507
2508         ret = -EXDEV;
2509         if (src_file.file->f_path.mnt != file->f_path.mnt)
2510                 goto out_fput;
2511
2512         src = file_inode(src_file.file);
2513
2514         ret = -EINVAL;
2515         if (src == inode)
2516                 goto out_fput;
2517
2518         /* the src must be open for reading */
2519         if (!(src_file.file->f_mode & FMODE_READ))
2520                 goto out_fput;
2521
2522         /* don't make the dst file partly checksummed */
2523         if ((BTRFS_I(src)->flags & BTRFS_INODE_NODATASUM) !=
2524             (BTRFS_I(inode)->flags & BTRFS_INODE_NODATASUM))
2525                 goto out_fput;
2526
2527         ret = -EISDIR;
2528         if (S_ISDIR(src->i_mode) || S_ISDIR(inode->i_mode))
2529                 goto out_fput;
2530
2531         ret = -EXDEV;
2532         if (src->i_sb != inode->i_sb)
2533                 goto out_fput;
2534
2535         ret = -ENOMEM;
2536         buf = vmalloc(btrfs_level_size(root, 0));
2537         if (!buf)
2538                 goto out_fput;
2539
2540         path = btrfs_alloc_path();
2541         if (!path) {
2542                 vfree(buf);
2543                 goto out_fput;
2544         }
2545         path->reada = 2;
2546
2547         if (inode < src) {
2548                 mutex_lock_nested(&inode->i_mutex, I_MUTEX_PARENT);
2549                 mutex_lock_nested(&src->i_mutex, I_MUTEX_CHILD);
2550         } else {
2551                 mutex_lock_nested(&src->i_mutex, I_MUTEX_PARENT);
2552                 mutex_lock_nested(&inode->i_mutex, I_MUTEX_CHILD);
2553         }
2554
2555         /* determine range to clone */
2556         ret = -EINVAL;
2557         if (off + len > src->i_size || off + len < off)
2558                 goto out_unlock;
2559         if (len == 0)
2560                 olen = len = src->i_size - off;
2561         /* if we extend to eof, continue to block boundary */
2562         if (off + len == src->i_size)
2563                 len = ALIGN(src->i_size, bs) - off;
2564
2565         /* verify the end result is block aligned */
2566         if (!IS_ALIGNED(off, bs) || !IS_ALIGNED(off + len, bs) ||
2567             !IS_ALIGNED(destoff, bs))
2568                 goto out_unlock;
2569
2570         if (destoff > inode->i_size) {
2571                 ret = btrfs_cont_expand(inode, inode->i_size, destoff);
2572                 if (ret)
2573                         goto out_unlock;
2574         }
2575
2576         /* truncate page cache pages from target inode range */
2577         truncate_inode_pages_range(&inode->i_data, destoff,
2578                                    PAGE_CACHE_ALIGN(destoff + len) - 1);
2579
2580         /* do any pending delalloc/csum calc on src, one way or
2581            another, and lock file content */
2582         while (1) {
2583                 struct btrfs_ordered_extent *ordered;
2584                 lock_extent(&BTRFS_I(src)->io_tree, off, off + len - 1);
2585                 ordered = btrfs_lookup_first_ordered_extent(src, off + len - 1);
2586                 if (!ordered &&
2587                     !test_range_bit(&BTRFS_I(src)->io_tree, off, off + len - 1,
2588                                     EXTENT_DELALLOC, 0, NULL))
2589                         break;
2590                 unlock_extent(&BTRFS_I(src)->io_tree, off, off + len - 1);
2591                 if (ordered)
2592                         btrfs_put_ordered_extent(ordered);
2593                 btrfs_wait_ordered_range(src, off, len);
2594         }
2595
2596         /* clone data */
2597         key.objectid = btrfs_ino(src);
2598         key.type = BTRFS_EXTENT_DATA_KEY;
2599         key.offset = 0;
2600
2601         while (1) {
2602                 /*
2603                  * note the key will change type as we walk through the
2604                  * tree.
2605                  */
2606                 ret = btrfs_search_slot(NULL, BTRFS_I(src)->root, &key, path,
2607                                 0, 0);
2608                 if (ret < 0)
2609                         goto out;
2610
2611                 nritems = btrfs_header_nritems(path->nodes[0]);
2612                 if (path->slots[0] >= nritems) {
2613                         ret = btrfs_next_leaf(BTRFS_I(src)->root, path);
2614                         if (ret < 0)
2615                                 goto out;
2616                         if (ret > 0)
2617                                 break;
2618                         nritems = btrfs_header_nritems(path->nodes[0]);
2619                 }
2620                 leaf = path->nodes[0];
2621                 slot = path->slots[0];
2622
2623                 btrfs_item_key_to_cpu(leaf, &key, slot);
2624                 if (btrfs_key_type(&key) > BTRFS_EXTENT_DATA_KEY ||
2625                     key.objectid != btrfs_ino(src))
2626                         break;
2627
2628                 if (btrfs_key_type(&key) == BTRFS_EXTENT_DATA_KEY) {
2629                         struct btrfs_file_extent_item *extent;
2630                         int type;
2631                         u32 size;
2632                         struct btrfs_key new_key;
2633                         u64 disko = 0, diskl = 0;
2634                         u64 datao = 0, datal = 0;
2635                         u8 comp;
2636                         u64 endoff;
2637
2638                         size = btrfs_item_size_nr(leaf, slot);
2639                         read_extent_buffer(leaf, buf,
2640                                            btrfs_item_ptr_offset(leaf, slot),
2641                                            size);
2642
2643                         extent = btrfs_item_ptr(leaf, slot,
2644                                                 struct btrfs_file_extent_item);
2645                         comp = btrfs_file_extent_compression(leaf, extent);
2646                         type = btrfs_file_extent_type(leaf, extent);
2647                         if (type == BTRFS_FILE_EXTENT_REG ||
2648                             type == BTRFS_FILE_EXTENT_PREALLOC) {
2649                                 disko = btrfs_file_extent_disk_bytenr(leaf,
2650                                                                       extent);
2651                                 diskl = btrfs_file_extent_disk_num_bytes(leaf,
2652                                                                  extent);
2653                                 datao = btrfs_file_extent_offset(leaf, extent);
2654                                 datal = btrfs_file_extent_num_bytes(leaf,
2655                                                                     extent);
2656                         } else if (type == BTRFS_FILE_EXTENT_INLINE) {
2657                                 /* take upper bound, may be compressed */
2658                                 datal = btrfs_file_extent_ram_bytes(leaf,
2659                                                                     extent);
2660                         }
2661                         btrfs_release_path(path);
2662
2663                         if (key.offset + datal <= off ||
2664                             key.offset >= off + len - 1)
2665                                 goto next;
2666
2667                         memcpy(&new_key, &key, sizeof(new_key));
2668                         new_key.objectid = btrfs_ino(inode);
2669                         if (off <= key.offset)
2670                                 new_key.offset = key.offset + destoff - off;
2671                         else
2672                                 new_key.offset = destoff;
2673
2674                         /*
2675                          * 1 - adjusting old extent (we may have to split it)
2676                          * 1 - add new extent
2677                          * 1 - inode update
2678                          */
2679                         trans = btrfs_start_transaction(root, 3);
2680                         if (IS_ERR(trans)) {
2681                                 ret = PTR_ERR(trans);
2682                                 goto out;
2683                         }
2684
2685                         if (type == BTRFS_FILE_EXTENT_REG ||
2686                             type == BTRFS_FILE_EXTENT_PREALLOC) {
2687                                 /*
2688                                  *    a  | --- range to clone ---|  b
2689                                  * | ------------- extent ------------- |
2690                                  */
2691
2692                                 /* substract range b */
2693                                 if (key.offset + datal > off + len)
2694                                         datal = off + len - key.offset;
2695
2696                                 /* substract range a */
2697                                 if (off > key.offset) {
2698                                         datao += off - key.offset;
2699                                         datal -= off - key.offset;
2700                                 }
2701
2702                                 ret = btrfs_drop_extents(trans, root, inode,
2703                                                          new_key.offset,
2704                                                          new_key.offset + datal,
2705                                                          1);
2706                                 if (ret) {
2707                                         btrfs_abort_transaction(trans, root,
2708                                                                 ret);
2709                                         btrfs_end_transaction(trans, root);
2710                                         goto out;
2711                                 }
2712
2713                                 ret = btrfs_insert_empty_item(trans, root, path,
2714                                                               &new_key, size);
2715                                 if (ret) {
2716                                         btrfs_abort_transaction(trans, root,
2717                                                                 ret);
2718                                         btrfs_end_transaction(trans, root);
2719                                         goto out;
2720                                 }
2721
2722                                 leaf = path->nodes[0];
2723                                 slot = path->slots[0];
2724                                 write_extent_buffer(leaf, buf,
2725                                             btrfs_item_ptr_offset(leaf, slot),
2726                                             size);
2727
2728                                 extent = btrfs_item_ptr(leaf, slot,
2729                                                 struct btrfs_file_extent_item);
2730
2731                                 /* disko == 0 means it's a hole */
2732                                 if (!disko)
2733                                         datao = 0;
2734
2735                                 btrfs_set_file_extent_offset(leaf, extent,
2736                                                              datao);
2737                                 btrfs_set_file_extent_num_bytes(leaf, extent,
2738                                                                 datal);
2739                                 if (disko) {
2740                                         inode_add_bytes(inode, datal);
2741                                         ret = btrfs_inc_extent_ref(trans, root,
2742                                                         disko, diskl, 0,
2743                                                         root->root_key.objectid,
2744                                                         btrfs_ino(inode),
2745                                                         new_key.offset - datao,
2746                                                         0);
2747                                         if (ret) {
2748                                                 btrfs_abort_transaction(trans,
2749                                                                         root,
2750                                                                         ret);
2751                                                 btrfs_end_transaction(trans,
2752                                                                       root);
2753                                                 goto out;
2754
2755                                         }
2756                                 }
2757                         } else if (type == BTRFS_FILE_EXTENT_INLINE) {
2758                                 u64 skip = 0;
2759                                 u64 trim = 0;
2760                                 if (off > key.offset) {
2761                                         skip = off - key.offset;
2762                                         new_key.offset += skip;
2763                                 }
2764
2765                                 if (key.offset + datal > off + len)
2766                                         trim = key.offset + datal - (off + len);
2767
2768                                 if (comp && (skip || trim)) {
2769                                         ret = -EINVAL;
2770                                         btrfs_end_transaction(trans, root);
2771                                         goto out;
2772                                 }
2773                                 size -= skip + trim;
2774                                 datal -= skip + trim;
2775
2776                                 ret = btrfs_drop_extents(trans, root, inode,
2777                                                          new_key.offset,
2778                                                          new_key.offset + datal,
2779                                                          1);
2780                                 if (ret) {
2781                                         btrfs_abort_transaction(trans, root,
2782                                                                 ret);
2783                                         btrfs_end_transaction(trans, root);
2784                                         goto out;
2785                                 }
2786
2787                                 ret = btrfs_insert_empty_item(trans, root, path,
2788                                                               &new_key, size);
2789                                 if (ret) {
2790                                         btrfs_abort_transaction(trans, root,
2791                                                                 ret);
2792                                         btrfs_end_transaction(trans, root);
2793                                         goto out;
2794                                 }
2795
2796                                 if (skip) {
2797                                         u32 start =
2798                                           btrfs_file_extent_calc_inline_size(0);
2799                                         memmove(buf+start, buf+start+skip,
2800                                                 datal);
2801                                 }
2802
2803                                 leaf = path->nodes[0];
2804                                 slot = path->slots[0];
2805                                 write_extent_buffer(leaf, buf,
2806                                             btrfs_item_ptr_offset(leaf, slot),
2807                                             size);
2808                                 inode_add_bytes(inode, datal);
2809                         }
2810
2811                         btrfs_mark_buffer_dirty(leaf);
2812                         btrfs_release_path(path);
2813
2814                         inode_inc_iversion(inode);
2815                         inode->i_mtime = inode->i_ctime = CURRENT_TIME;
2816
2817                         /*
2818                          * we round up to the block size at eof when
2819                          * determining which extents to clone above,
2820                          * but shouldn't round up the file size
2821                          */
2822                         endoff = new_key.offset + datal;
2823                         if (endoff > destoff+olen)
2824                                 endoff = destoff+olen;
2825                         if (endoff > inode->i_size)
2826                                 btrfs_i_size_write(inode, endoff);
2827
2828                         ret = btrfs_update_inode(trans, root, inode);
2829                         if (ret) {
2830                                 btrfs_abort_transaction(trans, root, ret);
2831                                 btrfs_end_transaction(trans, root);
2832                                 goto out;
2833                         }
2834                         ret = btrfs_end_transaction(trans, root);
2835                 }
2836 next:
2837                 btrfs_release_path(path);
2838                 key.offset++;
2839         }
2840         ret = 0;
2841 out:
2842         btrfs_release_path(path);
2843         unlock_extent(&BTRFS_I(src)->io_tree, off, off + len - 1);
2844 out_unlock:
2845         mutex_unlock(&src->i_mutex);
2846         mutex_unlock(&inode->i_mutex);
2847         vfree(buf);
2848         btrfs_free_path(path);
2849 out_fput:
2850         fdput(src_file);
2851 out_drop_write:
2852         mnt_drop_write_file(file);
2853         return ret;
2854 }
2855
2856 static long btrfs_ioctl_clone_range(struct file *file, void __user *argp)
2857 {
2858         struct btrfs_ioctl_clone_range_args args;
2859
2860         if (copy_from_user(&args, argp, sizeof(args)))
2861                 return -EFAULT;
2862         return btrfs_ioctl_clone(file, args.src_fd, args.src_offset,
2863                                  args.src_length, args.dest_offset);
2864 }
2865
2866 /*
2867  * there are many ways the trans_start and trans_end ioctls can lead
2868  * to deadlocks.  They should only be used by applications that
2869  * basically own the machine, and have a very in depth understanding
2870  * of all the possible deadlocks and enospc problems.
2871  */
2872 static long btrfs_ioctl_trans_start(struct file *file)
2873 {
2874         struct inode *inode = file_inode(file);
2875         struct btrfs_root *root = BTRFS_I(inode)->root;
2876         struct btrfs_trans_handle *trans;
2877         int ret;
2878
2879         ret = -EPERM;
2880         if (!capable(CAP_SYS_ADMIN))
2881                 goto out;
2882
2883         ret = -EINPROGRESS;
2884         if (file->private_data)
2885                 goto out;
2886
2887         ret = -EROFS;
2888         if (btrfs_root_readonly(root))
2889                 goto out;
2890
2891         ret = mnt_want_write_file(file);
2892         if (ret)
2893                 goto out;
2894
2895         atomic_inc(&root->fs_info->open_ioctl_trans);
2896
2897         ret = -ENOMEM;
2898         trans = btrfs_start_ioctl_transaction(root);
2899         if (IS_ERR(trans))
2900                 goto out_drop;
2901
2902         file->private_data = trans;
2903         return 0;
2904
2905 out_drop:
2906         atomic_dec(&root->fs_info->open_ioctl_trans);
2907         mnt_drop_write_file(file);
2908 out:
2909         return ret;
2910 }
2911
2912 static long btrfs_ioctl_default_subvol(struct file *file, void __user *argp)
2913 {
2914         struct inode *inode = file_inode(file);
2915         struct btrfs_root *root = BTRFS_I(inode)->root;
2916         struct btrfs_root *new_root;
2917         struct btrfs_dir_item *di;
2918         struct btrfs_trans_handle *trans;
2919         struct btrfs_path *path;
2920         struct btrfs_key location;
2921         struct btrfs_disk_key disk_key;
2922         u64 objectid = 0;
2923         u64 dir_id;
2924         int ret;
2925
2926         if (!capable(CAP_SYS_ADMIN))
2927                 return -EPERM;
2928
2929         ret = mnt_want_write_file(file);
2930         if (ret)
2931                 return ret;
2932
2933         if (copy_from_user(&objectid, argp, sizeof(objectid))) {
2934                 ret = -EFAULT;
2935                 goto out;
2936         }
2937
2938         if (!objectid)
2939                 objectid = root->root_key.objectid;
2940
2941         location.objectid = objectid;
2942         location.type = BTRFS_ROOT_ITEM_KEY;
2943         location.offset = (u64)-1;
2944
2945         new_root = btrfs_read_fs_root_no_name(root->fs_info, &location);
2946         if (IS_ERR(new_root)) {
2947                 ret = PTR_ERR(new_root);
2948                 goto out;
2949         }
2950
2951         if (btrfs_root_refs(&new_root->root_item) == 0) {
2952                 ret = -ENOENT;
2953                 goto out;
2954         }
2955
2956         path = btrfs_alloc_path();
2957         if (!path) {
2958                 ret = -ENOMEM;
2959                 goto out;
2960         }
2961         path->leave_spinning = 1;
2962
2963         trans = btrfs_start_transaction(root, 1);
2964         if (IS_ERR(trans)) {
2965                 btrfs_free_path(path);
2966                 ret = PTR_ERR(trans);
2967                 goto out;
2968         }
2969
2970         dir_id = btrfs_super_root_dir(root->fs_info->super_copy);
2971         di = btrfs_lookup_dir_item(trans, root->fs_info->tree_root, path,
2972                                    dir_id, "default", 7, 1);
2973         if (IS_ERR_OR_NULL(di)) {
2974                 btrfs_free_path(path);
2975                 btrfs_end_transaction(trans, root);
2976                 printk(KERN_ERR "Umm, you don't have the default dir item, "
2977                        "this isn't going to work\n");
2978                 ret = -ENOENT;
2979                 goto out;
2980         }
2981
2982         btrfs_cpu_key_to_disk(&disk_key, &new_root->root_key);
2983         btrfs_set_dir_item_key(path->nodes[0], di, &disk_key);
2984         btrfs_mark_buffer_dirty(path->nodes[0]);
2985         btrfs_free_path(path);
2986
2987         btrfs_set_fs_incompat(root->fs_info, DEFAULT_SUBVOL);
2988         btrfs_end_transaction(trans, root);
2989 out:
2990         mnt_drop_write_file(file);
2991         return ret;
2992 }
2993
2994 void btrfs_get_block_group_info(struct list_head *groups_list,
2995                                 struct btrfs_ioctl_space_info *space)
2996 {
2997         struct btrfs_block_group_cache *block_group;
2998
2999         space->total_bytes = 0;
3000         space->used_bytes = 0;
3001         space->flags = 0;
3002         list_for_each_entry(block_group, groups_list, list) {
3003                 space->flags = block_group->flags;
3004                 space->total_bytes += block_group->key.offset;
3005                 space->used_bytes +=
3006                         btrfs_block_group_used(&block_group->item);
3007         }
3008 }
3009
3010 long btrfs_ioctl_space_info(struct btrfs_root *root, void __user *arg)
3011 {
3012         struct btrfs_ioctl_space_args space_args;
3013         struct btrfs_ioctl_space_info space;
3014         struct btrfs_ioctl_space_info *dest;
3015         struct btrfs_ioctl_space_info *dest_orig;
3016         struct btrfs_ioctl_space_info __user *user_dest;
3017         struct btrfs_space_info *info;
3018         u64 types[] = {BTRFS_BLOCK_GROUP_DATA,
3019                        BTRFS_BLOCK_GROUP_SYSTEM,
3020                        BTRFS_BLOCK_GROUP_METADATA,
3021                        BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA};
3022         int num_types = 4;
3023         int alloc_size;
3024         int ret = 0;
3025         u64 slot_count = 0;
3026         int i, c;
3027
3028         if (copy_from_user(&space_args,
3029                            (struct btrfs_ioctl_space_args __user *)arg,
3030                            sizeof(space_args)))
3031                 return -EFAULT;
3032
3033         for (i = 0; i < num_types; i++) {
3034                 struct btrfs_space_info *tmp;
3035
3036                 info = NULL;
3037                 rcu_read_lock();
3038                 list_for_each_entry_rcu(tmp, &root->fs_info->space_info,
3039                                         list) {
3040                         if (tmp->flags == types[i]) {
3041                                 info = tmp;
3042                                 break;
3043                         }
3044                 }
3045                 rcu_read_unlock();
3046
3047                 if (!info)
3048                         continue;
3049
3050                 down_read(&info->groups_sem);
3051                 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
3052                         if (!list_empty(&info->block_groups[c]))
3053                                 slot_count++;
3054                 }
3055                 up_read(&info->groups_sem);
3056         }
3057
3058         /* space_slots == 0 means they are asking for a count */
3059         if (space_args.space_slots == 0) {
3060                 space_args.total_spaces = slot_count;
3061                 goto out;
3062         }
3063
3064         slot_count = min_t(u64, space_args.space_slots, slot_count);
3065
3066         alloc_size = sizeof(*dest) * slot_count;
3067
3068         /* we generally have at most 6 or so space infos, one for each raid
3069          * level.  So, a whole page should be more than enough for everyone
3070          */
3071         if (alloc_size > PAGE_CACHE_SIZE)
3072                 return -ENOMEM;
3073
3074         space_args.total_spaces = 0;
3075         dest = kmalloc(alloc_size, GFP_NOFS);
3076         if (!dest)
3077                 return -ENOMEM;
3078         dest_orig = dest;
3079
3080         /* now we have a buffer to copy into */
3081         for (i = 0; i < num_types; i++) {
3082                 struct btrfs_space_info *tmp;
3083
3084                 if (!slot_count)
3085                         break;
3086
3087                 info = NULL;
3088                 rcu_read_lock();
3089                 list_for_each_entry_rcu(tmp, &root->fs_info->space_info,
3090                                         list) {
3091                         if (tmp->flags == types[i]) {
3092                                 info = tmp;
3093                                 break;
3094                         }
3095                 }
3096                 rcu_read_unlock();
3097
3098                 if (!info)
3099                         continue;
3100                 down_read(&info->groups_sem);
3101                 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
3102                         if (!list_empty(&info->block_groups[c])) {
3103                                 btrfs_get_block_group_info(
3104                                         &info->block_groups[c], &space);
3105                                 memcpy(dest, &space, sizeof(space));
3106                                 dest++;
3107                                 space_args.total_spaces++;
3108                                 slot_count--;
3109                         }
3110                         if (!slot_count)
3111                                 break;
3112                 }
3113                 up_read(&info->groups_sem);
3114         }
3115
3116         user_dest = (struct btrfs_ioctl_space_info __user *)
3117                 (arg + sizeof(struct btrfs_ioctl_space_args));
3118
3119         if (copy_to_user(user_dest, dest_orig, alloc_size))
3120                 ret = -EFAULT;
3121
3122         kfree(dest_orig);
3123 out:
3124         if (ret == 0 && copy_to_user(arg, &space_args, sizeof(space_args)))
3125                 ret = -EFAULT;
3126
3127         return ret;
3128 }
3129
3130 /*
3131  * there are many ways the trans_start and trans_end ioctls can lead
3132  * to deadlocks.  They should only be used by applications that
3133  * basically own the machine, and have a very in depth understanding
3134  * of all the possible deadlocks and enospc problems.
3135  */
3136 long btrfs_ioctl_trans_end(struct file *file)
3137 {
3138         struct inode *inode = file_inode(file);
3139         struct btrfs_root *root = BTRFS_I(inode)->root;
3140         struct btrfs_trans_handle *trans;
3141
3142         trans = file->private_data;
3143         if (!trans)
3144                 return -EINVAL;
3145         file->private_data = NULL;
3146
3147         btrfs_end_transaction(trans, root);
3148
3149         atomic_dec(&root->fs_info->open_ioctl_trans);
3150
3151         mnt_drop_write_file(file);
3152         return 0;
3153 }
3154
3155 static noinline long btrfs_ioctl_start_sync(struct btrfs_root *root,
3156                                             void __user *argp)
3157 {
3158         struct btrfs_trans_handle *trans;
3159         u64 transid;
3160         int ret;
3161
3162         trans = btrfs_attach_transaction_barrier(root);
3163         if (IS_ERR(trans)) {
3164                 if (PTR_ERR(trans) != -ENOENT)
3165                         return PTR_ERR(trans);
3166
3167                 /* No running transaction, don't bother */
3168                 transid = root->fs_info->last_trans_committed;
3169                 goto out;
3170         }
3171         transid = trans->transid;
3172         ret = btrfs_commit_transaction_async(trans, root, 0);
3173         if (ret) {
3174                 btrfs_end_transaction(trans, root);
3175                 return ret;
3176         }
3177 out:
3178         if (argp)
3179                 if (copy_to_user(argp, &transid, sizeof(transid)))
3180                         return -EFAULT;
3181         return 0;
3182 }
3183
3184 static noinline long btrfs_ioctl_wait_sync(struct btrfs_root *root,
3185                                            void __user *argp)
3186 {
3187         u64 transid;
3188
3189         if (argp) {
3190                 if (copy_from_user(&transid, argp, sizeof(transid)))
3191                         return -EFAULT;
3192         } else {
3193                 transid = 0;  /* current trans */
3194         }
3195         return btrfs_wait_for_commit(root, transid);
3196 }
3197
3198 static long btrfs_ioctl_scrub(struct file *file, void __user *arg)
3199 {
3200         struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
3201         struct btrfs_ioctl_scrub_args *sa;
3202         int ret;
3203
3204         if (!capable(CAP_SYS_ADMIN))
3205                 return -EPERM;
3206
3207         sa = memdup_user(arg, sizeof(*sa));
3208         if (IS_ERR(sa))
3209                 return PTR_ERR(sa);
3210
3211         if (!(sa->flags & BTRFS_SCRUB_READONLY)) {
3212                 ret = mnt_want_write_file(file);
3213                 if (ret)
3214                         goto out;
3215         }
3216
3217         ret = btrfs_scrub_dev(root->fs_info, sa->devid, sa->start, sa->end,
3218                               &sa->progress, sa->flags & BTRFS_SCRUB_READONLY,
3219                               0);
3220
3221         if (copy_to_user(arg, sa, sizeof(*sa)))
3222                 ret = -EFAULT;
3223
3224         if (!(sa->flags & BTRFS_SCRUB_READONLY))
3225                 mnt_drop_write_file(file);
3226 out:
3227         kfree(sa);
3228         return ret;
3229 }
3230
3231 static long btrfs_ioctl_scrub_cancel(struct btrfs_root *root, void __user *arg)
3232 {
3233         if (!capable(CAP_SYS_ADMIN))
3234                 return -EPERM;
3235
3236         return btrfs_scrub_cancel(root->fs_info);
3237 }
3238
3239 static long btrfs_ioctl_scrub_progress(struct btrfs_root *root,
3240                                        void __user *arg)
3241 {
3242         struct btrfs_ioctl_scrub_args *sa;
3243         int ret;
3244
3245         if (!capable(CAP_SYS_ADMIN))
3246                 return -EPERM;
3247
3248         sa = memdup_user(arg, sizeof(*sa));
3249         if (IS_ERR(sa))
3250                 return PTR_ERR(sa);
3251
3252         ret = btrfs_scrub_progress(root, sa->devid, &sa->progress);
3253
3254         if (copy_to_user(arg, sa, sizeof(*sa)))
3255                 ret = -EFAULT;
3256
3257         kfree(sa);
3258         return ret;
3259 }
3260
3261 static long btrfs_ioctl_get_dev_stats(struct btrfs_root *root,
3262                                       void __user *arg)
3263 {
3264         struct btrfs_ioctl_get_dev_stats *sa;
3265         int ret;
3266
3267         sa = memdup_user(arg, sizeof(*sa));
3268         if (IS_ERR(sa))
3269                 return PTR_ERR(sa);
3270
3271         if ((sa->flags & BTRFS_DEV_STATS_RESET) && !capable(CAP_SYS_ADMIN)) {
3272                 kfree(sa);
3273                 return -EPERM;
3274         }
3275
3276         ret = btrfs_get_dev_stats(root, sa);
3277
3278         if (copy_to_user(arg, sa, sizeof(*sa)))
3279                 ret = -EFAULT;
3280
3281         kfree(sa);
3282         return ret;
3283 }
3284
3285 static long btrfs_ioctl_dev_replace(struct btrfs_root *root, void __user *arg)
3286 {
3287         struct btrfs_ioctl_dev_replace_args *p;
3288         int ret;
3289
3290         if (!capable(CAP_SYS_ADMIN))
3291                 return -EPERM;
3292
3293         p = memdup_user(arg, sizeof(*p));
3294         if (IS_ERR(p))
3295                 return PTR_ERR(p);
3296
3297         switch (p->cmd) {
3298         case BTRFS_IOCTL_DEV_REPLACE_CMD_START:
3299                 if (atomic_xchg(
3300                         &root->fs_info->mutually_exclusive_operation_running,
3301                         1)) {
3302                         pr_info("btrfs: dev add/delete/balance/replace/resize operation in progress\n");
3303                         ret = -EINPROGRESS;
3304                 } else {
3305                         ret = btrfs_dev_replace_start(root, p);
3306                         atomic_set(
3307                          &root->fs_info->mutually_exclusive_operation_running,
3308                          0);
3309                 }
3310                 break;
3311         case BTRFS_IOCTL_DEV_REPLACE_CMD_STATUS:
3312                 btrfs_dev_replace_status(root->fs_info, p);
3313                 ret = 0;
3314                 break;
3315         case BTRFS_IOCTL_DEV_REPLACE_CMD_CANCEL:
3316                 ret = btrfs_dev_replace_cancel(root->fs_info, p);
3317                 break;
3318         default:
3319                 ret = -EINVAL;
3320                 break;
3321         }
3322
3323         if (copy_to_user(arg, p, sizeof(*p)))
3324                 ret = -EFAULT;
3325
3326         kfree(p);
3327         return ret;
3328 }
3329
3330 static long btrfs_ioctl_ino_to_path(struct btrfs_root *root, void __user *arg)
3331 {
3332         int ret = 0;
3333         int i;
3334         u64 rel_ptr;
3335         int size;
3336         struct btrfs_ioctl_ino_path_args *ipa = NULL;
3337         struct inode_fs_paths *ipath = NULL;
3338         struct btrfs_path *path;
3339
3340         if (!capable(CAP_DAC_READ_SEARCH))
3341                 return -EPERM;
3342
3343         path = btrfs_alloc_path();
3344         if (!path) {
3345                 ret = -ENOMEM;
3346                 goto out;
3347         }
3348
3349         ipa = memdup_user(arg, sizeof(*ipa));
3350         if (IS_ERR(ipa)) {
3351                 ret = PTR_ERR(ipa);
3352                 ipa = NULL;
3353                 goto out;
3354         }
3355
3356         size = min_t(u32, ipa->size, 4096);
3357         ipath = init_ipath(size, root, path);
3358         if (IS_ERR(ipath)) {
3359                 ret = PTR_ERR(ipath);
3360                 ipath = NULL;
3361                 goto out;
3362         }
3363
3364         ret = paths_from_inode(ipa->inum, ipath);
3365         if (ret < 0)
3366                 goto out;
3367
3368         for (i = 0; i < ipath->fspath->elem_cnt; ++i) {
3369                 rel_ptr = ipath->fspath->val[i] -
3370                           (u64)(unsigned long)ipath->fspath->val;
3371                 ipath->fspath->val[i] = rel_ptr;
3372         }
3373
3374         ret = copy_to_user((void *)(unsigned long)ipa->fspath,
3375                            (void *)(unsigned long)ipath->fspath, size);
3376         if (ret) {
3377                 ret = -EFAULT;
3378                 goto out;
3379         }
3380
3381 out:
3382         btrfs_free_path(path);
3383         free_ipath(ipath);
3384         kfree(ipa);
3385
3386         return ret;
3387 }
3388
3389 static int build_ino_list(u64 inum, u64 offset, u64 root, void *ctx)
3390 {
3391         struct btrfs_data_container *inodes = ctx;
3392         const size_t c = 3 * sizeof(u64);
3393
3394         if (inodes->bytes_left >= c) {
3395                 inodes->bytes_left -= c;
3396                 inodes->val[inodes->elem_cnt] = inum;
3397                 inodes->val[inodes->elem_cnt + 1] = offset;
3398                 inodes->val[inodes->elem_cnt + 2] = root;
3399                 inodes->elem_cnt += 3;
3400         } else {
3401                 inodes->bytes_missing += c - inodes->bytes_left;
3402                 inodes->bytes_left = 0;
3403                 inodes->elem_missed += 3;
3404         }
3405
3406         return 0;
3407 }
3408
3409 static long btrfs_ioctl_logical_to_ino(struct btrfs_root *root,
3410                                         void __user *arg)
3411 {
3412         int ret = 0;
3413         int size;
3414         struct btrfs_ioctl_logical_ino_args *loi;
3415         struct btrfs_data_container *inodes = NULL;
3416         struct btrfs_path *path = NULL;
3417
3418         if (!capable(CAP_SYS_ADMIN))
3419                 return -EPERM;
3420
3421         loi = memdup_user(arg, sizeof(*loi));
3422         if (IS_ERR(loi)) {
3423                 ret = PTR_ERR(loi);
3424                 loi = NULL;
3425                 goto out;
3426         }
3427
3428         path = btrfs_alloc_path();
3429         if (!path) {
3430                 ret = -ENOMEM;
3431                 goto out;
3432         }
3433
3434         size = min_t(u32, loi->size, 64 * 1024);
3435         inodes = init_data_container(size);
3436         if (IS_ERR(inodes)) {
3437                 ret = PTR_ERR(inodes);
3438                 inodes = NULL;
3439                 goto out;
3440         }
3441
3442         ret = iterate_inodes_from_logical(loi->logical, root->fs_info, path,
3443                                           build_ino_list, inodes);
3444         if (ret == -EINVAL)
3445                 ret = -ENOENT;
3446         if (ret < 0)
3447                 goto out;
3448
3449         ret = copy_to_user((void *)(unsigned long)loi->inodes,
3450                            (void *)(unsigned long)inodes, size);
3451         if (ret)
3452                 ret = -EFAULT;
3453
3454 out:
3455         btrfs_free_path(path);
3456         vfree(inodes);
3457         kfree(loi);
3458
3459         return ret;
3460 }
3461
3462 void update_ioctl_balance_args(struct btrfs_fs_info *fs_info, int lock,
3463                                struct btrfs_ioctl_balance_args *bargs)
3464 {
3465         struct btrfs_balance_control *bctl = fs_info->balance_ctl;
3466
3467         bargs->flags = bctl->flags;
3468
3469         if (atomic_read(&fs_info->balance_running))
3470                 bargs->state |= BTRFS_BALANCE_STATE_RUNNING;
3471         if (atomic_read(&fs_info->balance_pause_req))
3472                 bargs->state |= BTRFS_BALANCE_STATE_PAUSE_REQ;
3473         if (atomic_read(&fs_info->balance_cancel_req))
3474                 bargs->state |= BTRFS_BALANCE_STATE_CANCEL_REQ;
3475
3476         memcpy(&bargs->data, &bctl->data, sizeof(bargs->data));
3477         memcpy(&bargs->meta, &bctl->meta, sizeof(bargs->meta));
3478         memcpy(&bargs->sys, &bctl->sys, sizeof(bargs->sys));
3479
3480         if (lock) {
3481                 spin_lock(&fs_info->balance_lock);
3482                 memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat));
3483                 spin_unlock(&fs_info->balance_lock);
3484         } else {
3485                 memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat));
3486         }
3487 }
3488
3489 static long btrfs_ioctl_balance(struct file *file, void __user *arg)
3490 {
3491         struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
3492         struct btrfs_fs_info *fs_info = root->fs_info;
3493         struct btrfs_ioctl_balance_args *bargs;
3494         struct btrfs_balance_control *bctl;
3495         bool need_unlock; /* for mut. excl. ops lock */
3496         int ret;
3497
3498         if (!capable(CAP_SYS_ADMIN))
3499                 return -EPERM;
3500
3501         ret = mnt_want_write_file(file);
3502         if (ret)
3503                 return ret;
3504
3505 again:
3506         if (!atomic_xchg(&fs_info->mutually_exclusive_operation_running, 1)) {
3507                 mutex_lock(&fs_info->volume_mutex);
3508                 mutex_lock(&fs_info->balance_mutex);
3509                 need_unlock = true;
3510                 goto locked;
3511         }
3512
3513         /*
3514          * mut. excl. ops lock is locked.  Three possibilites:
3515          *   (1) some other op is running
3516          *   (2) balance is running
3517          *   (3) balance is paused -- special case (think resume)
3518          */
3519         mutex_lock(&fs_info->balance_mutex);
3520         if (fs_info->balance_ctl) {
3521                 /* this is either (2) or (3) */
3522                 if (!atomic_read(&fs_info->balance_running)) {
3523                         mutex_unlock(&fs_info->balance_mutex);
3524                         if (!mutex_trylock(&fs_info->volume_mutex))
3525                                 goto again;
3526                         mutex_lock(&fs_info->balance_mutex);
3527
3528                         if (fs_info->balance_ctl &&
3529                             !atomic_read(&fs_info->balance_running)) {
3530                                 /* this is (3) */
3531                                 need_unlock = false;
3532                                 goto locked;
3533                         }
3534
3535                         mutex_unlock(&fs_info->balance_mutex);
3536                         mutex_unlock(&fs_info->volume_mutex);
3537                         goto again;
3538                 } else {
3539                         /* this is (2) */
3540                         mutex_unlock(&fs_info->balance_mutex);
3541                         ret = -EINPROGRESS;
3542                         goto out;
3543                 }
3544         } else {
3545                 /* this is (1) */
3546                 mutex_unlock(&fs_info->balance_mutex);
3547                 pr_info("btrfs: dev add/delete/balance/replace/resize operation in progress\n");
3548                 ret = -EINVAL;
3549                 goto out;
3550         }
3551
3552 locked:
3553         BUG_ON(!atomic_read(&fs_info->mutually_exclusive_operation_running));
3554
3555         if (arg) {
3556                 bargs = memdup_user(arg, sizeof(*bargs));
3557                 if (IS_ERR(bargs)) {
3558                         ret = PTR_ERR(bargs);
3559                         goto out_unlock;
3560                 }
3561
3562                 if (bargs->flags & BTRFS_BALANCE_RESUME) {
3563                         if (!fs_info->balance_ctl) {
3564                                 ret = -ENOTCONN;
3565                                 goto out_bargs;
3566                         }
3567
3568                         bctl = fs_info->balance_ctl;
3569                         spin_lock(&fs_info->balance_lock);
3570                         bctl->flags |= BTRFS_BALANCE_RESUME;
3571                         spin_unlock(&fs_info->balance_lock);
3572
3573                         goto do_balance;
3574                 }
3575         } else {
3576                 bargs = NULL;
3577         }
3578
3579         if (fs_info->balance_ctl) {
3580                 ret = -EINPROGRESS;
3581                 goto out_bargs;
3582         }
3583
3584         bctl = kzalloc(sizeof(*bctl), GFP_NOFS);
3585         if (!bctl) {
3586                 ret = -ENOMEM;
3587                 goto out_bargs;
3588         }
3589
3590         bctl->fs_info = fs_info;
3591         if (arg) {
3592                 memcpy(&bctl->data, &bargs->data, sizeof(bctl->data));
3593                 memcpy(&bctl->meta, &bargs->meta, sizeof(bctl->meta));
3594                 memcpy(&bctl->sys, &bargs->sys, sizeof(bctl->sys));
3595
3596                 bctl->flags = bargs->flags;
3597         } else {
3598                 /* balance everything - no filters */
3599                 bctl->flags |= BTRFS_BALANCE_TYPE_MASK;
3600         }
3601
3602 do_balance:
3603         /*
3604          * Ownership of bctl and mutually_exclusive_operation_running
3605          * goes to to btrfs_balance.  bctl is freed in __cancel_balance,
3606          * or, if restriper was paused all the way until unmount, in
3607          * free_fs_info.  mutually_exclusive_operation_running is
3608          * cleared in __cancel_balance.
3609          */
3610         need_unlock = false;
3611
3612         ret = btrfs_balance(bctl, bargs);
3613
3614         if (arg) {
3615                 if (copy_to_user(arg, bargs, sizeof(*bargs)))
3616                         ret = -EFAULT;
3617         }
3618
3619 out_bargs:
3620         kfree(bargs);
3621 out_unlock:
3622         mutex_unlock(&fs_info->balance_mutex);
3623         mutex_unlock(&fs_info->volume_mutex);
3624         if (need_unlock)
3625                 atomic_set(&fs_info->mutually_exclusive_operation_running, 0);
3626 out:
3627         mnt_drop_write_file(file);
3628         return ret;
3629 }
3630
3631 static long btrfs_ioctl_balance_ctl(struct btrfs_root *root, int cmd)
3632 {
3633         if (!capable(CAP_SYS_ADMIN))
3634                 return -EPERM;
3635
3636         switch (cmd) {
3637         case BTRFS_BALANCE_CTL_PAUSE:
3638                 return btrfs_pause_balance(root->fs_info);
3639         case BTRFS_BALANCE_CTL_CANCEL:
3640                 return btrfs_cancel_balance(root->fs_info);
3641         }
3642
3643         return -EINVAL;
3644 }
3645
3646 static long btrfs_ioctl_balance_progress(struct btrfs_root *root,
3647                                          void __user *arg)
3648 {
3649         struct btrfs_fs_info *fs_info = root->fs_info;
3650         struct btrfs_ioctl_balance_args *bargs;
3651         int ret = 0;
3652
3653         if (!capable(CAP_SYS_ADMIN))
3654                 return -EPERM;
3655
3656         mutex_lock(&fs_info->balance_mutex);
3657         if (!fs_info->balance_ctl) {
3658                 ret = -ENOTCONN;
3659                 goto out;
3660         }
3661
3662         bargs = kzalloc(sizeof(*bargs), GFP_NOFS);
3663         if (!bargs) {
3664                 ret = -ENOMEM;
3665                 goto out;
3666         }
3667
3668         update_ioctl_balance_args(fs_info, 1, bargs);
3669
3670         if (copy_to_user(arg, bargs, sizeof(*bargs)))
3671                 ret = -EFAULT;
3672
3673         kfree(bargs);
3674 out:
3675         mutex_unlock(&fs_info->balance_mutex);
3676         return ret;
3677 }
3678
3679 static long btrfs_ioctl_quota_ctl(struct file *file, void __user *arg)
3680 {
3681         struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
3682         struct btrfs_ioctl_quota_ctl_args *sa;
3683         struct btrfs_trans_handle *trans = NULL;
3684         int ret;
3685         int err;
3686
3687         if (!capable(CAP_SYS_ADMIN))
3688                 return -EPERM;
3689
3690         ret = mnt_want_write_file(file);
3691         if (ret)
3692                 return ret;
3693
3694         sa = memdup_user(arg, sizeof(*sa));
3695         if (IS_ERR(sa)) {
3696                 ret = PTR_ERR(sa);
3697                 goto drop_write;
3698         }
3699
3700         if (sa->cmd != BTRFS_QUOTA_CTL_RESCAN) {
3701                 trans = btrfs_start_transaction(root, 2);
3702                 if (IS_ERR(trans)) {
3703                         ret = PTR_ERR(trans);
3704                         goto out;
3705                 }
3706         }
3707
3708         switch (sa->cmd) {
3709         case BTRFS_QUOTA_CTL_ENABLE:
3710                 ret = btrfs_quota_enable(trans, root->fs_info);
3711                 break;
3712         case BTRFS_QUOTA_CTL_DISABLE:
3713                 ret = btrfs_quota_disable(trans, root->fs_info);
3714                 break;
3715         case BTRFS_QUOTA_CTL_RESCAN:
3716                 ret = btrfs_quota_rescan(root->fs_info);
3717                 break;
3718         default:
3719                 ret = -EINVAL;
3720                 break;
3721         }
3722
3723         if (copy_to_user(arg, sa, sizeof(*sa)))
3724                 ret = -EFAULT;
3725
3726         if (trans) {
3727                 err = btrfs_commit_transaction(trans, root);
3728                 if (err && !ret)
3729                         ret = err;
3730         }
3731 out:
3732         kfree(sa);
3733 drop_write:
3734         mnt_drop_write_file(file);
3735         return ret;
3736 }
3737
3738 static long btrfs_ioctl_qgroup_assign(struct file *file, void __user *arg)
3739 {
3740         struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
3741         struct btrfs_ioctl_qgroup_assign_args *sa;
3742         struct btrfs_trans_handle *trans;
3743         int ret;
3744         int err;
3745
3746         if (!capable(CAP_SYS_ADMIN))
3747                 return -EPERM;
3748
3749         ret = mnt_want_write_file(file);
3750         if (ret)
3751                 return ret;
3752
3753         sa = memdup_user(arg, sizeof(*sa));
3754         if (IS_ERR(sa)) {
3755                 ret = PTR_ERR(sa);
3756                 goto drop_write;
3757         }
3758
3759         trans = btrfs_join_transaction(root);
3760         if (IS_ERR(trans)) {
3761                 ret = PTR_ERR(trans);
3762                 goto out;
3763         }
3764
3765         /* FIXME: check if the IDs really exist */
3766         if (sa->assign) {
3767                 ret = btrfs_add_qgroup_relation(trans, root->fs_info,
3768                                                 sa->src, sa->dst);
3769         } else {
3770                 ret = btrfs_del_qgroup_relation(trans, root->fs_info,
3771                                                 sa->src, sa->dst);
3772         }
3773
3774         err = btrfs_end_transaction(trans, root);
3775         if (err && !ret)
3776                 ret = err;
3777
3778 out:
3779         kfree(sa);
3780 drop_write:
3781         mnt_drop_write_file(file);
3782         return ret;
3783 }
3784
3785 static long btrfs_ioctl_qgroup_create(struct file *file, void __user *arg)
3786 {
3787         struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
3788         struct btrfs_ioctl_qgroup_create_args *sa;
3789         struct btrfs_trans_handle *trans;
3790         int ret;
3791         int err;
3792
3793         if (!capable(CAP_SYS_ADMIN))
3794                 return -EPERM;
3795
3796         ret = mnt_want_write_file(file);
3797         if (ret)
3798                 return ret;
3799
3800         sa = memdup_user(arg, sizeof(*sa));
3801         if (IS_ERR(sa)) {
3802                 ret = PTR_ERR(sa);
3803                 goto drop_write;
3804         }
3805
3806         if (!sa->qgroupid) {
3807                 ret = -EINVAL;
3808                 goto out;
3809         }
3810
3811         trans = btrfs_join_transaction(root);
3812         if (IS_ERR(trans)) {
3813                 ret = PTR_ERR(trans);
3814                 goto out;
3815         }
3816
3817         /* FIXME: check if the IDs really exist */
3818         if (sa->create) {
3819                 ret = btrfs_create_qgroup(trans, root->fs_info, sa->qgroupid,
3820                                           NULL);
3821         } else {
3822                 ret = btrfs_remove_qgroup(trans, root->fs_info, sa->qgroupid);
3823         }
3824
3825         err = btrfs_end_transaction(trans, root);
3826         if (err && !ret)
3827                 ret = err;
3828
3829 out:
3830         kfree(sa);
3831 drop_write:
3832         mnt_drop_write_file(file);
3833         return ret;
3834 }
3835
3836 static long btrfs_ioctl_qgroup_limit(struct file *file, void __user *arg)
3837 {
3838         struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
3839         struct btrfs_ioctl_qgroup_limit_args *sa;
3840         struct btrfs_trans_handle *trans;
3841         int ret;
3842         int err;
3843         u64 qgroupid;
3844
3845         if (!capable(CAP_SYS_ADMIN))
3846                 return -EPERM;
3847
3848         ret = mnt_want_write_file(file);
3849         if (ret)
3850                 return ret;
3851
3852         sa = memdup_user(arg, sizeof(*sa));
3853         if (IS_ERR(sa)) {
3854                 ret = PTR_ERR(sa);
3855                 goto drop_write;
3856         }
3857
3858         trans = btrfs_join_transaction(root);
3859         if (IS_ERR(trans)) {
3860                 ret = PTR_ERR(trans);
3861                 goto out;
3862         }
3863
3864         qgroupid = sa->qgroupid;
3865         if (!qgroupid) {
3866                 /* take the current subvol as qgroup */
3867                 qgroupid = root->root_key.objectid;
3868         }
3869
3870         /* FIXME: check if the IDs really exist */
3871         ret = btrfs_limit_qgroup(trans, root->fs_info, qgroupid, &sa->lim);
3872
3873         err = btrfs_end_transaction(trans, root);
3874         if (err && !ret)
3875                 ret = err;
3876
3877 out:
3878         kfree(sa);
3879 drop_write:
3880         mnt_drop_write_file(file);
3881         return ret;
3882 }
3883
3884 static long btrfs_ioctl_set_received_subvol(struct file *file,
3885                                             void __user *arg)
3886 {
3887         struct btrfs_ioctl_received_subvol_args *sa = NULL;
3888         struct inode *inode = file_inode(file);
3889         struct btrfs_root *root = BTRFS_I(inode)->root;
3890         struct btrfs_root_item *root_item = &root->root_item;
3891         struct btrfs_trans_handle *trans;
3892         struct timespec ct = CURRENT_TIME;
3893         int ret = 0;
3894
3895         ret = mnt_want_write_file(file);
3896         if (ret < 0)
3897                 return ret;
3898
3899         down_write(&root->fs_info->subvol_sem);
3900
3901         if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) {
3902                 ret = -EINVAL;
3903                 goto out;
3904         }
3905
3906         if (btrfs_root_readonly(root)) {
3907                 ret = -EROFS;
3908                 goto out;
3909         }
3910
3911         if (!inode_owner_or_capable(inode)) {
3912                 ret = -EACCES;
3913                 goto out;
3914         }
3915
3916         sa = memdup_user(arg, sizeof(*sa));
3917         if (IS_ERR(sa)) {
3918                 ret = PTR_ERR(sa);
3919                 sa = NULL;
3920                 goto out;
3921         }
3922
3923         trans = btrfs_start_transaction(root, 1);
3924         if (IS_ERR(trans)) {
3925                 ret = PTR_ERR(trans);
3926                 trans = NULL;
3927                 goto out;
3928         }
3929
3930         sa->rtransid = trans->transid;
3931         sa->rtime.sec = ct.tv_sec;
3932         sa->rtime.nsec = ct.tv_nsec;
3933
3934         memcpy(root_item->received_uuid, sa->uuid, BTRFS_UUID_SIZE);
3935         btrfs_set_root_stransid(root_item, sa->stransid);
3936         btrfs_set_root_rtransid(root_item, sa->rtransid);
3937         root_item->stime.sec = cpu_to_le64(sa->stime.sec);
3938         root_item->stime.nsec = cpu_to_le32(sa->stime.nsec);
3939         root_item->rtime.sec = cpu_to_le64(sa->rtime.sec);
3940         root_item->rtime.nsec = cpu_to_le32(sa->rtime.nsec);
3941
3942         ret = btrfs_update_root(trans, root->fs_info->tree_root,
3943                                 &root->root_key, &root->root_item);
3944         if (ret < 0) {
3945                 btrfs_end_transaction(trans, root);
3946                 trans = NULL;
3947                 goto out;
3948         } else {
3949                 ret = btrfs_commit_transaction(trans, root);
3950                 if (ret < 0)
3951                         goto out;
3952         }
3953
3954         ret = copy_to_user(arg, sa, sizeof(*sa));
3955         if (ret)
3956                 ret = -EFAULT;
3957
3958 out:
3959         kfree(sa);
3960         up_write(&root->fs_info->subvol_sem);
3961         mnt_drop_write_file(file);
3962         return ret;
3963 }
3964
3965 static int btrfs_ioctl_get_fslabel(struct file *file, void __user *arg)
3966 {
3967         struct btrfs_root *root = BTRFS_I(fdentry(file)->d_inode)->root;
3968         const char *label = root->fs_info->super_copy->label;
3969         size_t len = strnlen(label, BTRFS_LABEL_SIZE);
3970         int ret;
3971
3972         if (len == BTRFS_LABEL_SIZE) {
3973                 pr_warn("btrfs: label is too long, return the first %zu bytes\n",
3974                         --len);
3975         }
3976
3977         mutex_lock(&root->fs_info->volume_mutex);
3978         ret = copy_to_user(arg, label, len);
3979         mutex_unlock(&root->fs_info->volume_mutex);
3980
3981         return ret ? -EFAULT : 0;
3982 }
3983
3984 static int btrfs_ioctl_set_fslabel(struct file *file, void __user *arg)
3985 {
3986         struct btrfs_root *root = BTRFS_I(fdentry(file)->d_inode)->root;
3987         struct btrfs_super_block *super_block = root->fs_info->super_copy;
3988         struct btrfs_trans_handle *trans;
3989         char label[BTRFS_LABEL_SIZE];
3990         int ret;
3991
3992         if (!capable(CAP_SYS_ADMIN))
3993                 return -EPERM;
3994
3995         if (copy_from_user(label, arg, sizeof(label)))
3996                 return -EFAULT;
3997
3998         if (strnlen(label, BTRFS_LABEL_SIZE) == BTRFS_LABEL_SIZE) {
3999                 pr_err("btrfs: unable to set label with more than %d bytes\n",
4000                        BTRFS_LABEL_SIZE - 1);
4001                 return -EINVAL;
4002         }
4003
4004         ret = mnt_want_write_file(file);
4005         if (ret)
4006                 return ret;
4007
4008         mutex_lock(&root->fs_info->volume_mutex);
4009         trans = btrfs_start_transaction(root, 0);
4010         if (IS_ERR(trans)) {
4011                 ret = PTR_ERR(trans);
4012                 goto out_unlock;
4013         }
4014
4015         strcpy(super_block->label, label);
4016         ret = btrfs_end_transaction(trans, root);
4017
4018 out_unlock:
4019         mutex_unlock(&root->fs_info->volume_mutex);
4020         mnt_drop_write_file(file);
4021         return ret;
4022 }
4023
4024 long btrfs_ioctl(struct file *file, unsigned int
4025                 cmd, unsigned long arg)
4026 {
4027         struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4028         void __user *argp = (void __user *)arg;
4029
4030         switch (cmd) {
4031         case FS_IOC_GETFLAGS:
4032                 return btrfs_ioctl_getflags(file, argp);
4033         case FS_IOC_SETFLAGS:
4034                 return btrfs_ioctl_setflags(file, argp);
4035         case FS_IOC_GETVERSION:
4036                 return btrfs_ioctl_getversion(file, argp);
4037         case FITRIM:
4038                 return btrfs_ioctl_fitrim(file, argp);
4039         case BTRFS_IOC_SNAP_CREATE:
4040                 return btrfs_ioctl_snap_create(file, argp, 0);
4041         case BTRFS_IOC_SNAP_CREATE_V2:
4042                 return btrfs_ioctl_snap_create_v2(file, argp, 0);
4043         case BTRFS_IOC_SUBVOL_CREATE:
4044                 return btrfs_ioctl_snap_create(file, argp, 1);
4045         case BTRFS_IOC_SUBVOL_CREATE_V2:
4046                 return btrfs_ioctl_snap_create_v2(file, argp, 1);
4047         case BTRFS_IOC_SNAP_DESTROY:
4048                 return btrfs_ioctl_snap_destroy(file, argp);
4049         case BTRFS_IOC_SUBVOL_GETFLAGS:
4050                 return btrfs_ioctl_subvol_getflags(file, argp);
4051         case BTRFS_IOC_SUBVOL_SETFLAGS:
4052                 return btrfs_ioctl_subvol_setflags(file, argp);
4053         case BTRFS_IOC_DEFAULT_SUBVOL:
4054                 return btrfs_ioctl_default_subvol(file, argp);
4055         case BTRFS_IOC_DEFRAG:
4056                 return btrfs_ioctl_defrag(file, NULL);
4057         case BTRFS_IOC_DEFRAG_RANGE:
4058                 return btrfs_ioctl_defrag(file, argp);
4059         case BTRFS_IOC_RESIZE:
4060                 return btrfs_ioctl_resize(file, argp);
4061         case BTRFS_IOC_ADD_DEV:
4062                 return btrfs_ioctl_add_dev(root, argp);
4063         case BTRFS_IOC_RM_DEV:
4064                 return btrfs_ioctl_rm_dev(file, argp);
4065         case BTRFS_IOC_FS_INFO:
4066                 return btrfs_ioctl_fs_info(root, argp);
4067         case BTRFS_IOC_DEV_INFO:
4068                 return btrfs_ioctl_dev_info(root, argp);
4069         case BTRFS_IOC_BALANCE:
4070                 return btrfs_ioctl_balance(file, NULL);
4071         case BTRFS_IOC_CLONE:
4072                 return btrfs_ioctl_clone(file, arg, 0, 0, 0);
4073         case BTRFS_IOC_CLONE_RANGE:
4074                 return btrfs_ioctl_clone_range(file, argp);
4075         case BTRFS_IOC_TRANS_START:
4076                 return btrfs_ioctl_trans_start(file);
4077         case BTRFS_IOC_TRANS_END:
4078                 return btrfs_ioctl_trans_end(file);
4079         case BTRFS_IOC_TREE_SEARCH:
4080                 return btrfs_ioctl_tree_search(file, argp);
4081         case BTRFS_IOC_INO_LOOKUP:
4082                 return btrfs_ioctl_ino_lookup(file, argp);
4083         case BTRFS_IOC_INO_PATHS:
4084                 return btrfs_ioctl_ino_to_path(root, argp);
4085         case BTRFS_IOC_LOGICAL_INO:
4086                 return btrfs_ioctl_logical_to_ino(root, argp);
4087         case BTRFS_IOC_SPACE_INFO:
4088                 return btrfs_ioctl_space_info(root, argp);
4089         case BTRFS_IOC_SYNC:
4090                 btrfs_sync_fs(file->f_dentry->d_sb, 1);
4091                 return 0;
4092         case BTRFS_IOC_START_SYNC:
4093                 return btrfs_ioctl_start_sync(root, argp);
4094         case BTRFS_IOC_WAIT_SYNC:
4095                 return btrfs_ioctl_wait_sync(root, argp);
4096         case BTRFS_IOC_SCRUB:
4097                 return btrfs_ioctl_scrub(file, argp);
4098         case BTRFS_IOC_SCRUB_CANCEL:
4099                 return btrfs_ioctl_scrub_cancel(root, argp);
4100         case BTRFS_IOC_SCRUB_PROGRESS:
4101                 return btrfs_ioctl_scrub_progress(root, argp);
4102         case BTRFS_IOC_BALANCE_V2:
4103                 return btrfs_ioctl_balance(file, argp);
4104         case BTRFS_IOC_BALANCE_CTL:
4105                 return btrfs_ioctl_balance_ctl(root, arg);
4106         case BTRFS_IOC_BALANCE_PROGRESS:
4107                 return btrfs_ioctl_balance_progress(root, argp);
4108         case BTRFS_IOC_SET_RECEIVED_SUBVOL:
4109                 return btrfs_ioctl_set_received_subvol(file, argp);
4110         case BTRFS_IOC_SEND:
4111                 return btrfs_ioctl_send(file, argp);
4112         case BTRFS_IOC_GET_DEV_STATS:
4113                 return btrfs_ioctl_get_dev_stats(root, argp);
4114         case BTRFS_IOC_QUOTA_CTL:
4115                 return btrfs_ioctl_quota_ctl(file, argp);
4116         case BTRFS_IOC_QGROUP_ASSIGN:
4117                 return btrfs_ioctl_qgroup_assign(file, argp);
4118         case BTRFS_IOC_QGROUP_CREATE:
4119                 return btrfs_ioctl_qgroup_create(file, argp);
4120         case BTRFS_IOC_QGROUP_LIMIT:
4121                 return btrfs_ioctl_qgroup_limit(file, argp);
4122         case BTRFS_IOC_DEV_REPLACE:
4123                 return btrfs_ioctl_dev_replace(root, argp);
4124         case BTRFS_IOC_GET_FSLABEL:
4125                 return btrfs_ioctl_get_fslabel(file, argp);
4126         case BTRFS_IOC_SET_FSLABEL:
4127                 return btrfs_ioctl_set_fslabel(file, argp);
4128         }
4129
4130         return -ENOTTY;
4131 }