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, false);
 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         btrfs_set_stack_inode_generation(inode_item, 1);
 440         btrfs_set_stack_inode_size(inode_item, 3);
 441         btrfs_set_stack_inode_nlink(inode_item, 1);
 442         btrfs_set_stack_inode_nbytes(inode_item, root->leafsize);
 443         btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755);
 444
 445         btrfs_set_root_flags(&root_item, 0);
 446         btrfs_set_root_limit(&root_item, 0);
 447         btrfs_set_stack_inode_flags(inode_item, 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         btrfs_set_stack_timespec_sec(&root_item.otime, cur_time.tv_sec);
 461         btrfs_set_stack_timespec_nsec(&root_item.otime, 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         ret = btrfs_start_delalloc_inodes(root, 0);
 559         if (ret)
 560                 return ret;
 561
 562         btrfs_wait_ordered_extents(root, 0);
 563
 564         pending_snapshot = kzalloc(sizeof(*pending_snapshot), GFP_NOFS);
 565         if (!pending_snapshot)
 566                 return -ENOMEM;
 567
 568         btrfs_init_block_rsv(&pending_snapshot->block_rsv,
 569                              BTRFS_BLOCK_RSV_TEMP);
 570         /*
 571          * 1 - parent dir inode
 572          * 2 - dir entries
 573          * 1 - root item
 574          * 2 - root ref/backref
 575          * 1 - root of snapshot
 576          */
 577         ret = btrfs_subvolume_reserve_metadata(BTRFS_I(dir)->root,
 578                                         &pending_snapshot->block_rsv, 7,
 579                                         &pending_snapshot->qgroup_reserved,
 580                                         false);
 581         if (ret)
 582                 goto out;
 583
 584         pending_snapshot->dentry = dentry;
 585         pending_snapshot->root = root;
 586         pending_snapshot->readonly = readonly;
 587         pending_snapshot->dir = dir;
 588         pending_snapshot->inherit = inherit;
 589
 590         trans = btrfs_start_transaction(root, 0);
 591         if (IS_ERR(trans)) {
 592                 ret = PTR_ERR(trans);
 593                 goto fail;
 594         }
 595
 596         spin_lock(&root->fs_info->trans_lock);
 597         list_add(&pending_snapshot->list,
 598                  &trans->transaction->pending_snapshots);
 599         spin_unlock(&root->fs_info->trans_lock);
 600         if (async_transid) {
 601                 *async_transid = trans->transid;
 602                 ret = btrfs_commit_transaction_async(trans,
 603                                      root->fs_info->extent_root, 1);
 604                 if (ret)
 605                         ret = btrfs_commit_transaction(trans, root);
 606         } else {
 607                 ret = btrfs_commit_transaction(trans,
 608                                                root->fs_info->extent_root);
 609         }
 610         if (ret)
 611                 goto fail;
 612
 613         ret = pending_snapshot->error;
 614         if (ret)
 615                 goto fail;
 616
 617         ret = btrfs_orphan_cleanup(pending_snapshot->snap);
 618         if (ret)
 619                 goto fail;
 620
 621         inode = btrfs_lookup_dentry(dentry->d_parent->d_inode, dentry);
 622         if (IS_ERR(inode)) {
 623                 ret = PTR_ERR(inode);
 624                 goto fail;
 625         }
 626         BUG_ON(!inode);
 627         d_instantiate(dentry, inode);
 628         ret = 0;
 629 fail:
 630         btrfs_subvolume_release_metadata(BTRFS_I(dir)->root,
 631                                          &pending_snapshot->block_rsv,
 632                                          pending_snapshot->qgroup_reserved);
 633 out:
 634         kfree(pending_snapshot);
 635         return ret;
 636 }
 637
 638 /*  copy of check_sticky in fs/namei.c()
 639 * It's inline, so penalty for filesystems that don't use sticky bit is
 640 * minimal.
 641 */
 642 static inline int btrfs_check_sticky(struct inode *dir, struct inode *inode)
 643 {
 644         kuid_t fsuid = current_fsuid();
 645
 646         if (!(dir->i_mode & S_ISVTX))
 647                 return 0;
 648         if (uid_eq(inode->i_uid, fsuid))
 649                 return 0;
 650         if (uid_eq(dir->i_uid, fsuid))
 651                 return 0;
 652         return !capable(CAP_FOWNER);
 653 }
 654
 655 /*  copy of may_delete in fs/namei.c()
 656  *      Check whether we can remove a link victim from directory dir, check
 657  *  whether the type of victim is right.
 658  *  1. We can't do it if dir is read-only (done in permission())
 659  *  2. We should have write and exec permissions on dir
 660  *  3. We can't remove anything from append-only dir
 661  *  4. We can't do anything with immutable dir (done in permission())
 662  *  5. If the sticky bit on dir is set we should either
 663  *      a. be owner of dir, or
 664  *      b. be owner of victim, or
 665  *      c. have CAP_FOWNER capability
 666  *  6. If the victim is append-only or immutable we can't do antyhing with
 667  *     links pointing to it.
 668  *  7. If we were asked to remove a directory and victim isn't one - ENOTDIR.
 669  *  8. If we were asked to remove a non-directory and victim isn't one - EISDIR.
 670  *  9. We can't remove a root or mountpoint.
 671  * 10. We don't allow removal of NFS sillyrenamed files; it's handled by
 672  *     nfs_async_unlink().
 673  */
 674
 675 static int btrfs_may_delete(struct inode *dir,struct dentry *victim,int isdir)
 676 {
 677         int error;
 678
 679         if (!victim->d_inode)
 680                 return -ENOENT;
 681
 682         BUG_ON(victim->d_parent->d_inode != dir);
 683         audit_inode_child(dir, victim, AUDIT_TYPE_CHILD_DELETE);
 684
 685         error = inode_permission(dir, MAY_WRITE | MAY_EXEC);
 686         if (error)
 687                 return error;
 688         if (IS_APPEND(dir))
 689                 return -EPERM;
 690         if (btrfs_check_sticky(dir, victim->d_inode)||
 691                 IS_APPEND(victim->d_inode)||
 692             IS_IMMUTABLE(victim->d_inode) || IS_SWAPFILE(victim->d_inode))
 693                 return -EPERM;
 694         if (isdir) {
 695                 if (!S_ISDIR(victim->d_inode->i_mode))
 696                         return -ENOTDIR;
 697                 if (IS_ROOT(victim))
 698                         return -EBUSY;
 699         } else if (S_ISDIR(victim->d_inode->i_mode))
 700                 return -EISDIR;
 701         if (IS_DEADDIR(dir))
 702                 return -ENOENT;
 703         if (victim->d_flags & DCACHE_NFSFS_RENAMED)
 704                 return -EBUSY;
 705         return 0;
 706 }
 707
 708 /* copy of may_create in fs/namei.c() */
 709 static inline int btrfs_may_create(struct inode *dir, struct dentry *child)
 710 {
 711         if (child->d_inode)
 712                 return -EEXIST;
 713         if (IS_DEADDIR(dir))
 714                 return -ENOENT;
 715         return inode_permission(dir, MAY_WRITE | MAY_EXEC);
 716 }
 717
 718 /*
 719  * Create a new subvolume below @parent.  This is largely modeled after
 720  * sys_mkdirat and vfs_mkdir, but we only do a single component lookup
 721  * inside this filesystem so it's quite a bit simpler.
 722  */
 723 static noinline int btrfs_mksubvol(struct path *parent,
 724                                    char *name, int namelen,
 725                                    struct btrfs_root *snap_src,
 726                                    u64 *async_transid, bool readonly,
 727                                    struct btrfs_qgroup_inherit *inherit)
 728 {
 729         struct inode *dir  = parent->dentry->d_inode;
 730         struct dentry *dentry;
 731         int error;
 732
 733         error = mutex_lock_killable_nested(&dir->i_mutex, I_MUTEX_PARENT);
 734         if (error == -EINTR)
 735                 return error;
 736
 737         dentry = lookup_one_len(name, parent->dentry, namelen);
 738         error = PTR_ERR(dentry);
 739         if (IS_ERR(dentry))
 740                 goto out_unlock;
 741
 742         error = -EEXIST;
 743         if (dentry->d_inode)
 744                 goto out_dput;
 745
 746         error = btrfs_may_create(dir, dentry);
 747         if (error)
 748                 goto out_dput;
 749
 750         /*
 751          * even if this name doesn't exist, we may get hash collisions.
 752          * check for them now when we can safely fail
 753          */
 754         error = btrfs_check_dir_item_collision(BTRFS_I(dir)->root,
 755                                                dir->i_ino, name,
 756                                                namelen);
 757         if (error)
 758                 goto out_dput;
 759
 760         down_read(&BTRFS_I(dir)->root->fs_info->subvol_sem);
 761
 762         if (btrfs_root_refs(&BTRFS_I(dir)->root->root_item) == 0)
 763                 goto out_up_read;
 764
 765         if (snap_src) {
 766                 error = create_snapshot(snap_src, dir, dentry, name, namelen,
 767                                         async_transid, readonly, inherit);
 768         } else {
 769                 error = create_subvol(dir, dentry, name, namelen,
 770                                       async_transid, inherit);
 771         }
 772         if (!error)
 773                 fsnotify_mkdir(dir, dentry);
 774 out_up_read:
 775         up_read(&BTRFS_I(dir)->root->fs_info->subvol_sem);
 776 out_dput:
 777         dput(dentry);
 778 out_unlock:
 779         mutex_unlock(&dir->i_mutex);
 780         return error;
 781 }
 782
 783 /*
 784  * When we're defragging a range, we don't want to kick it off again
 785  * if it is really just waiting for delalloc to send it down.
 786  * If we find a nice big extent or delalloc range for the bytes in the
 787  * file you want to defrag, we return 0 to let you know to skip this
 788  * part of the file
 789  */
 790 static int check_defrag_in_cache(struct inode *inode, u64 offset, int thresh)
 791 {
 792         struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
 793         struct extent_map *em = NULL;
 794         struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
 795         u64 end;
 796
 797         read_lock(&em_tree->lock);
 798         em = lookup_extent_mapping(em_tree, offset, PAGE_CACHE_SIZE);
 799         read_unlock(&em_tree->lock);
 800
 801         if (em) {
 802                 end = extent_map_end(em);
 803                 free_extent_map(em);
 804                 if (end - offset > thresh)
 805                         return 0;
 806         }
 807         /* if we already have a nice delalloc here, just stop */
 808         thresh /= 2;
 809         end = count_range_bits(io_tree, &offset, offset + thresh,
 810                                thresh, EXTENT_DELALLOC, 1);
 811         if (end >= thresh)
 812                 return 0;
 813         return 1;
 814 }
 815
 816 /*
 817  * helper function to walk through a file and find extents
 818  * newer than a specific transid, and smaller than thresh.
 819  *
 820  * This is used by the defragging code to find new and small
 821  * extents
 822  */
 823 static int find_new_extents(struct btrfs_root *root,
 824                             struct inode *inode, u64 newer_than,
 825                             u64 *off, int thresh)
 826 {
 827         struct btrfs_path *path;
 828         struct btrfs_key min_key;
 829         struct btrfs_key max_key;
 830         struct extent_buffer *leaf;
 831         struct btrfs_file_extent_item *extent;
 832         int type;
 833         int ret;
 834         u64 ino = btrfs_ino(inode);
 835
 836         path = btrfs_alloc_path();
 837         if (!path)
 838                 return -ENOMEM;
 839
 840         min_key.objectid = ino;
 841         min_key.type = BTRFS_EXTENT_DATA_KEY;
 842         min_key.offset = *off;
 843
 844         max_key.objectid = ino;
 845         max_key.type = (u8)-1;
 846         max_key.offset = (u64)-1;
 847
 848         path->keep_locks = 1;
 849
 850         while(1) {
 851                 ret = btrfs_search_forward(root, &min_key, &max_key,
 852                                            path, newer_than);
 853                 if (ret != 0)
 854                         goto none;
 855                 if (min_key.objectid != ino)
 856                         goto none;
 857                 if (min_key.type != BTRFS_EXTENT_DATA_KEY)
 858                         goto none;
 859
 860                 leaf = path->nodes[0];
 861                 extent = btrfs_item_ptr(leaf, path->slots[0],
 862                                         struct btrfs_file_extent_item);
 863
 864                 type = btrfs_file_extent_type(leaf, extent);
 865                 if (type == BTRFS_FILE_EXTENT_REG &&
 866                     btrfs_file_extent_num_bytes(leaf, extent) < thresh &&
 867                     check_defrag_in_cache(inode, min_key.offset, thresh)) {
 868                         *off = min_key.offset;
 869                         btrfs_free_path(path);
 870                         return 0;
 871                 }
 872
 873                 if (min_key.offset == (u64)-1)
 874                         goto none;
 875
 876                 min_key.offset++;
 877                 btrfs_release_path(path);
 878         }
 879 none:
 880         btrfs_free_path(path);
 881         return -ENOENT;
 882 }
 883
 884 static struct extent_map *defrag_lookup_extent(struct inode *inode, u64 start)
 885 {
 886         struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
 887         struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
 888         struct extent_map *em;
 889         u64 len = PAGE_CACHE_SIZE;
 890
 891         /*
 892          * hopefully we have this extent in the tree already, try without
 893          * the full extent lock
 894          */
 895         read_lock(&em_tree->lock);
 896         em = lookup_extent_mapping(em_tree, start, len);
 897         read_unlock(&em_tree->lock);
 898
 899         if (!em) {
 900                 /* get the big lock and read metadata off disk */
 901                 lock_extent(io_tree, start, start + len - 1);
 902                 em = btrfs_get_extent(inode, NULL, 0, start, len, 0);
 903                 unlock_extent(io_tree, start, start + len - 1);
 904
 905                 if (IS_ERR(em))
 906                         return NULL;
 907         }
 908
 909         return em;
 910 }
 911
 912 static bool defrag_check_next_extent(struct inode *inode, struct extent_map *em)
 913 {
 914         struct extent_map *next;
 915         bool ret = true;
 916
 917         /* this is the last extent */
 918         if (em->start + em->len >= i_size_read(inode))
 919                 return false;
 920
 921         next = defrag_lookup_extent(inode, em->start + em->len);
 922         if (!next || next->block_start >= EXTENT_MAP_LAST_BYTE)
 923                 ret = false;
 924
 925         free_extent_map(next);
 926         return ret;
 927 }
 928
 929 static int should_defrag_range(struct inode *inode, u64 start, int thresh,
 930                                u64 *last_len, u64 *skip, u64 *defrag_end,
 931                                int compress)
 932 {
 933         struct extent_map *em;
 934         int ret = 1;
 935         bool next_mergeable = true;
 936
 937         /*
 938          * make sure that once we start defragging an extent, we keep on
 939          * defragging it
 940          */
 941         if (start < *defrag_end)
 942                 return 1;
 943
 944         *skip = 0;
 945
 946         em = defrag_lookup_extent(inode, start);
 947         if (!em)
 948                 return 0;
 949
 950         /* this will cover holes, and inline extents */
 951         if (em->block_start >= EXTENT_MAP_LAST_BYTE) {
 952                 ret = 0;
 953                 goto out;
 954         }
 955
 956         next_mergeable = defrag_check_next_extent(inode, em);
 957
 958         /*
 959          * we hit a real extent, if it is big or the next extent is not a
 960          * real extent, don't bother defragging it
 961          */
 962         if (!compress && (*last_len == 0 || *last_len >= thresh) &&
 963             (em->len >= thresh || !next_mergeable))
 964                 ret = 0;
 965 out:
 966         /*
 967          * last_len ends up being a counter of how many bytes we've defragged.
 968          * every time we choose not to defrag an extent, we reset *last_len
 969          * so that the next tiny extent will force a defrag.
 970          *
 971          * The end result of this is that tiny extents before a single big
 972          * extent will force at least part of that big extent to be defragged.
 973          */
 974         if (ret) {
 975                 *defrag_end = extent_map_end(em);
 976         } else {
 977                 *last_len = 0;
 978                 *skip = extent_map_end(em);
 979                 *defrag_end = 0;
 980         }
 981
 982         free_extent_map(em);
 983         return ret;
 984 }
 985
 986 /*
 987  * it doesn't do much good to defrag one or two pages
 988  * at a time.  This pulls in a nice chunk of pages
 989  * to COW and defrag.
 990  *
 991  * It also makes sure the delalloc code has enough
 992  * dirty data to avoid making new small extents as part
 993  * of the defrag
 994  *
 995  * It's a good idea to start RA on this range
 996  * before calling this.
 997  */
 998 static int cluster_pages_for_defrag(struct inode *inode,
 999                                     struct page **pages,
1000                                     unsigned long start_index,
1001                                     int num_pages)
1002 {
1003         unsigned long file_end;
1004         u64 isize = i_size_read(inode);
1005         u64 page_start;
1006         u64 page_end;
1007         u64 page_cnt;
1008         int ret;
1009         int i;
1010         int i_done;
1011         struct btrfs_ordered_extent *ordered;
1012         struct extent_state *cached_state = NULL;
1013         struct extent_io_tree *tree;
1014         gfp_t mask = btrfs_alloc_write_mask(inode->i_mapping);
1015
1016         file_end = (isize - 1) >> PAGE_CACHE_SHIFT;
1017         if (!isize || start_index > file_end)
1018                 return 0;
1019
1020         page_cnt = min_t(u64, (u64)num_pages, (u64)file_end - start_index + 1);
1021
1022         ret = btrfs_delalloc_reserve_space(inode,
1023                                            page_cnt << PAGE_CACHE_SHIFT);
1024         if (ret)
1025                 return ret;
1026         i_done = 0;
1027         tree = &BTRFS_I(inode)->io_tree;
1028
1029         /* step one, lock all the pages */
1030         for (i = 0; i < page_cnt; i++) {
1031                 struct page *page;
1032 again:
1033                 page = find_or_create_page(inode->i_mapping,
1034                                            start_index + i, mask);
1035                 if (!page)
1036                         break;
1037
1038                 page_start = page_offset(page);
1039                 page_end = page_start + PAGE_CACHE_SIZE - 1;
1040                 while (1) {
1041                         lock_extent(tree, page_start, page_end);
1042                         ordered = btrfs_lookup_ordered_extent(inode,
1043                                                               page_start);
1044                         unlock_extent(tree, page_start, page_end);
1045                         if (!ordered)
1046                                 break;
1047
1048                         unlock_page(page);
1049                         btrfs_start_ordered_extent(inode, ordered, 1);
1050                         btrfs_put_ordered_extent(ordered);
1051                         lock_page(page);
1052                         /*
1053                          * we unlocked the page above, so we need check if
1054                          * it was released or not.
1055                          */
1056                         if (page->mapping != inode->i_mapping) {
1057                                 unlock_page(page);
1058                                 page_cache_release(page);
1059                                 goto again;
1060                         }
1061                 }
1062
1063                 if (!PageUptodate(page)) {
1064                         btrfs_readpage(NULL, page);
1065                         lock_page(page);
1066                         if (!PageUptodate(page)) {
1067                                 unlock_page(page);
1068                                 page_cache_release(page);
1069                                 ret = -EIO;
1070                                 break;
1071                         }
1072                 }
1073
1074                 if (page->mapping != inode->i_mapping) {
1075                         unlock_page(page);
1076                         page_cache_release(page);
1077                         goto again;
1078                 }
1079
1080                 pages[i] = page;
1081                 i_done++;
1082         }
1083         if (!i_done || ret)
1084                 goto out;
1085
1086         if (!(inode->i_sb->s_flags & MS_ACTIVE))
1087                 goto out;
1088
1089         /*
1090          * so now we have a nice long stream of locked
1091          * and up to date pages, lets wait on them
1092          */
1093         for (i = 0; i < i_done; i++)
1094                 wait_on_page_writeback(pages[i]);
1095
1096         page_start = page_offset(pages[0]);
1097         page_end = page_offset(pages[i_done - 1]) + PAGE_CACHE_SIZE;
1098
1099         lock_extent_bits(&BTRFS_I(inode)->io_tree,
1100                          page_start, page_end - 1, 0, &cached_state);
1101         clear_extent_bit(&BTRFS_I(inode)->io_tree, page_start,
1102                           page_end - 1, EXTENT_DIRTY | EXTENT_DELALLOC |
1103                           EXTENT_DO_ACCOUNTING | EXTENT_DEFRAG, 0, 0,
1104                           &cached_state, GFP_NOFS);
1105
1106         if (i_done != page_cnt) {
1107                 spin_lock(&BTRFS_I(inode)->lock);
1108                 BTRFS_I(inode)->outstanding_extents++;
1109                 spin_unlock(&BTRFS_I(inode)->lock);
1110                 btrfs_delalloc_release_space(inode,
1111                                      (page_cnt - i_done) << PAGE_CACHE_SHIFT);
1112         }
1113
1114
1115         set_extent_defrag(&BTRFS_I(inode)->io_tree, page_start, page_end - 1,
1116                           &cached_state, GFP_NOFS);
1117
1118         unlock_extent_cached(&BTRFS_I(inode)->io_tree,
1119                              page_start, page_end - 1, &cached_state,
1120                              GFP_NOFS);
1121
1122         for (i = 0; i < i_done; i++) {
1123                 clear_page_dirty_for_io(pages[i]);
1124                 ClearPageChecked(pages[i]);
1125                 set_page_extent_mapped(pages[i]);
1126                 set_page_dirty(pages[i]);
1127                 unlock_page(pages[i]);
1128                 page_cache_release(pages[i]);
1129         }
1130         return i_done;
1131 out:
1132         for (i = 0; i < i_done; i++) {
1133                 unlock_page(pages[i]);
1134                 page_cache_release(pages[i]);
1135         }
1136         btrfs_delalloc_release_space(inode, page_cnt << PAGE_CACHE_SHIFT);
1137         return ret;
1138
1139 }
1140
1141 int btrfs_defrag_file(struct inode *inode, struct file *file,
1142                       struct btrfs_ioctl_defrag_range_args *range,
1143                       u64 newer_than, unsigned long max_to_defrag)
1144 {
1145         struct btrfs_root *root = BTRFS_I(inode)->root;
1146         struct file_ra_state *ra = NULL;
1147         unsigned long last_index;
1148         u64 isize = i_size_read(inode);
1149         u64 last_len = 0;
1150         u64 skip = 0;
1151         u64 defrag_end = 0;
1152         u64 newer_off = range->start;
1153         unsigned long i;
1154         unsigned long ra_index = 0;
1155         int ret;
1156         int defrag_count = 0;
1157         int compress_type = BTRFS_COMPRESS_ZLIB;
1158         int extent_thresh = range->extent_thresh;
1159         int max_cluster = (256 * 1024) >> PAGE_CACHE_SHIFT;
1160         int cluster = max_cluster;
1161         u64 new_align = ~((u64)128 * 1024 - 1);
1162         struct page **pages = NULL;
1163
1164         if (isize == 0)
1165                 return 0;
1166
1167         if (range->start >= isize)
1168                 return -EINVAL;
1169
1170         if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS) {
1171                 if (range->compress_type > BTRFS_COMPRESS_TYPES)
1172                         return -EINVAL;
1173                 if (range->compress_type)
1174                         compress_type = range->compress_type;
1175         }
1176
1177         if (extent_thresh == 0)
1178                 extent_thresh = 256 * 1024;
1179
1180         /*
1181          * if we were not given a file, allocate a readahead
1182          * context
1183          */
1184         if (!file) {
1185                 ra = kzalloc(sizeof(*ra), GFP_NOFS);
1186                 if (!ra)
1187                         return -ENOMEM;
1188                 file_ra_state_init(ra, inode->i_mapping);
1189         } else {
1190                 ra = &file->f_ra;
1191         }
1192
1193         pages = kmalloc(sizeof(struct page *) * max_cluster,
1194                         GFP_NOFS);
1195         if (!pages) {
1196                 ret = -ENOMEM;
1197                 goto out_ra;
1198         }
1199
1200         /* find the last page to defrag */
1201         if (range->start + range->len > range->start) {
1202                 last_index = min_t(u64, isize - 1,
1203                          range->start + range->len - 1) >> PAGE_CACHE_SHIFT;
1204         } else {
1205                 last_index = (isize - 1) >> PAGE_CACHE_SHIFT;
1206         }
1207
1208         if (newer_than) {
1209                 ret = find_new_extents(root, inode, newer_than,
1210                                        &newer_off, 64 * 1024);
1211                 if (!ret) {
1212                         range->start = newer_off;
1213                         /*
1214                          * we always align our defrag to help keep
1215                          * the extents in the file evenly spaced
1216                          */
1217                         i = (newer_off & new_align) >> PAGE_CACHE_SHIFT;
1218                 } else
1219                         goto out_ra;
1220         } else {
1221                 i = range->start >> PAGE_CACHE_SHIFT;
1222         }
1223         if (!max_to_defrag)
1224                 max_to_defrag = last_index + 1;
1225
1226         /*
1227          * make writeback starts from i, so the defrag range can be
1228          * written sequentially.
1229          */
1230         if (i < inode->i_mapping->writeback_index)
1231                 inode->i_mapping->writeback_index = i;
1232
1233         while (i <= last_index && defrag_count < max_to_defrag &&
1234                (i < (i_size_read(inode) + PAGE_CACHE_SIZE - 1) >>
1235                 PAGE_CACHE_SHIFT)) {
1236                 /*
1237                  * make sure we stop running if someone unmounts
1238                  * the FS
1239                  */
1240                 if (!(inode->i_sb->s_flags & MS_ACTIVE))
1241                         break;
1242
1243                 if (btrfs_defrag_cancelled(root->fs_info)) {
1244                         printk(KERN_DEBUG "btrfs: defrag_file cancelled\n");
1245                         ret = -EAGAIN;
1246                         break;
1247                 }
1248
1249                 if (!should_defrag_range(inode, (u64)i << PAGE_CACHE_SHIFT,
1250                                          extent_thresh, &last_len, &skip,
1251                                          &defrag_end, range->flags &
1252                                          BTRFS_DEFRAG_RANGE_COMPRESS)) {
1253                         unsigned long next;
1254                         /*
1255                          * the should_defrag function tells us how much to skip
1256                          * bump our counter by the suggested amount
1257                          */
1258                         next = (skip + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT;
1259                         i = max(i + 1, next);
1260                         continue;
1261                 }
1262
1263                 if (!newer_than) {
1264                         cluster = (PAGE_CACHE_ALIGN(defrag_end) >>
1265                                    PAGE_CACHE_SHIFT) - i;
1266                         cluster = min(cluster, max_cluster);
1267                 } else {
1268                         cluster = max_cluster;
1269                 }
1270
1271                 if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)
1272                         BTRFS_I(inode)->force_compress = compress_type;
1273
1274                 if (i + cluster > ra_index) {
1275                         ra_index = max(i, ra_index);
1276                         btrfs_force_ra(inode->i_mapping, ra, file, ra_index,
1277                                        cluster);
1278                         ra_index += max_cluster;
1279                 }
1280
1281                 mutex_lock(&inode->i_mutex);
1282                 ret = cluster_pages_for_defrag(inode, pages, i, cluster);
1283                 if (ret < 0) {
1284                         mutex_unlock(&inode->i_mutex);
1285                         goto out_ra;
1286                 }
1287
1288                 defrag_count += ret;
1289                 balance_dirty_pages_ratelimited(inode->i_mapping);
1290                 mutex_unlock(&inode->i_mutex);
1291
1292                 if (newer_than) {
1293                         if (newer_off == (u64)-1)
1294                                 break;
1295
1296                         if (ret > 0)
1297                                 i += ret;
1298
1299                         newer_off = max(newer_off + 1,
1300                                         (u64)i << PAGE_CACHE_SHIFT);
1301
1302                         ret = find_new_extents(root, inode,
1303                                                newer_than, &newer_off,
1304                                                64 * 1024);
1305                         if (!ret) {
1306                                 range->start = newer_off;
1307                                 i = (newer_off & new_align) >> PAGE_CACHE_SHIFT;
1308                         } else {
1309                                 break;
1310                         }
1311                 } else {
1312                         if (ret > 0) {
1313                                 i += ret;
1314                                 last_len += ret << PAGE_CACHE_SHIFT;
1315                         } else {
1316                                 i++;
1317                                 last_len = 0;
1318                         }
1319                 }
1320         }
1321
1322         if ((range->flags & BTRFS_DEFRAG_RANGE_START_IO))
1323                 filemap_flush(inode->i_mapping);
1324
1325         if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
1326                 /* the filemap_flush will queue IO into the worker threads, but
1327                  * we have to make sure the IO is actually started and that
1328                  * ordered extents get created before we return
1329                  */
1330                 atomic_inc(&root->fs_info->async_submit_draining);
1331                 while (atomic_read(&root->fs_info->nr_async_submits) ||
1332                       atomic_read(&root->fs_info->async_delalloc_pages)) {
1333                         wait_event(root->fs_info->async_submit_wait,
1334                            (atomic_read(&root->fs_info->nr_async_submits) == 0 &&
1335                             atomic_read(&root->fs_info->async_delalloc_pages) == 0));
1336                 }
1337                 atomic_dec(&root->fs_info->async_submit_draining);
1338
1339                 mutex_lock(&inode->i_mutex);
1340                 BTRFS_I(inode)->force_compress = BTRFS_COMPRESS_NONE;
1341                 mutex_unlock(&inode->i_mutex);
1342         }
1343
1344         if (range->compress_type == BTRFS_COMPRESS_LZO) {
1345                 btrfs_set_fs_incompat(root->fs_info, COMPRESS_LZO);
1346         }
1347
1348         ret = defrag_count;
1349
1350 out_ra:
1351         if (!file)
1352                 kfree(ra);
1353         kfree(pages);
1354         return ret;
1355 }
1356
1357 static noinline int btrfs_ioctl_resize(struct file *file,
1358                                         void __user *arg)
1359 {
1360         u64 new_size;
1361         u64 old_size;
1362         u64 devid = 1;
1363         struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
1364         struct btrfs_ioctl_vol_args *vol_args;
1365         struct btrfs_trans_handle *trans;
1366         struct btrfs_device *device = NULL;
1367         char *sizestr;
1368         char *devstr = NULL;
1369         int ret = 0;
1370         int mod = 0;
1371
1372         if (!capable(CAP_SYS_ADMIN))
1373                 return -EPERM;
1374
1375         ret = mnt_want_write_file(file);
1376         if (ret)
1377                 return ret;
1378
1379         if (atomic_xchg(&root->fs_info->mutually_exclusive_operation_running,
1380                         1)) {
1381                 pr_info("btrfs: dev add/delete/balance/replace/resize operation in progress\n");
1382                 mnt_drop_write_file(file);
1383                 return -EINVAL;
1384         }
1385
1386         mutex_lock(&root->fs_info->volume_mutex);
1387         vol_args = memdup_user(arg, sizeof(*vol_args));
1388         if (IS_ERR(vol_args)) {
1389                 ret = PTR_ERR(vol_args);
1390                 goto out;
1391         }
1392
1393         vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1394
1395         sizestr = vol_args->name;
1396         devstr = strchr(sizestr, ':');
1397         if (devstr) {
1398                 char *end;
1399                 sizestr = devstr + 1;
1400                 *devstr = '\0';
1401                 devstr = vol_args->name;
1402                 devid = simple_strtoull(devstr, &end, 10);
1403                 if (!devid) {
1404                         ret = -EINVAL;
1405                         goto out_free;
1406                 }
1407                 printk(KERN_INFO "btrfs: resizing devid %llu\n",
1408                        (unsigned long long)devid);
1409         }
1410
1411         device = btrfs_find_device(root->fs_info, devid, NULL, NULL);
1412         if (!device) {
1413                 printk(KERN_INFO "btrfs: resizer unable to find device %llu\n",
1414                        (unsigned long long)devid);
1415                 ret = -ENODEV;
1416                 goto out_free;
1417         }
1418
1419         if (!device->writeable) {
1420                 printk(KERN_INFO "btrfs: resizer unable to apply on "
1421                        "readonly device %llu\n",
1422                        (unsigned long long)devid);
1423                 ret = -EPERM;
1424                 goto out_free;
1425         }
1426
1427         if (!strcmp(sizestr, "max"))
1428                 new_size = device->bdev->bd_inode->i_size;
1429         else {
1430                 if (sizestr[0] == '-') {
1431                         mod = -1;
1432                         sizestr++;
1433                 } else if (sizestr[0] == '+') {
1434                         mod = 1;
1435                         sizestr++;
1436                 }
1437                 new_size = memparse(sizestr, NULL);
1438                 if (new_size == 0) {
1439                         ret = -EINVAL;
1440                         goto out_free;
1441                 }
1442         }
1443
1444         if (device->is_tgtdev_for_dev_replace) {
1445                 ret = -EPERM;
1446                 goto out_free;
1447         }
1448
1449         old_size = device->total_bytes;
1450
1451         if (mod < 0) {
1452                 if (new_size > old_size) {
1453                         ret = -EINVAL;
1454                         goto out_free;
1455                 }
1456                 new_size = old_size - new_size;
1457         } else if (mod > 0) {
1458                 new_size = old_size + new_size;
1459         }
1460
1461         if (new_size < 256 * 1024 * 1024) {
1462                 ret = -EINVAL;
1463                 goto out_free;
1464         }
1465         if (new_size > device->bdev->bd_inode->i_size) {
1466                 ret = -EFBIG;
1467                 goto out_free;
1468         }
1469
1470         do_div(new_size, root->sectorsize);
1471         new_size *= root->sectorsize;
1472
1473         printk_in_rcu(KERN_INFO "btrfs: new size for %s is %llu\n",
1474                       rcu_str_deref(device->name),
1475                       (unsigned long long)new_size);
1476
1477         if (new_size > old_size) {
1478                 trans = btrfs_start_transaction(root, 0);
1479                 if (IS_ERR(trans)) {
1480                         ret = PTR_ERR(trans);
1481                         goto out_free;
1482                 }
1483                 ret = btrfs_grow_device(trans, device, new_size);
1484                 btrfs_commit_transaction(trans, root);
1485         } else if (new_size < old_size) {
1486                 ret = btrfs_shrink_device(device, new_size);
1487         } /* equal, nothing need to do */
1488
1489 out_free:
1490         kfree(vol_args);
1491 out:
1492         mutex_unlock(&root->fs_info->volume_mutex);
1493         atomic_set(&root->fs_info->mutually_exclusive_operation_running, 0);
1494         mnt_drop_write_file(file);
1495         return ret;
1496 }
1497
1498 static noinline int btrfs_ioctl_snap_create_transid(struct file *file,
1499                                 char *name, unsigned long fd, int subvol,
1500                                 u64 *transid, bool readonly,
1501                                 struct btrfs_qgroup_inherit *inherit)
1502 {
1503         int namelen;
1504         int ret = 0;
1505
1506         ret = mnt_want_write_file(file);
1507         if (ret)
1508                 goto out;
1509
1510         namelen = strlen(name);
1511         if (strchr(name, '/')) {
1512                 ret = -EINVAL;
1513                 goto out_drop_write;
1514         }
1515
1516         if (name[0] == '.' &&
1517            (namelen == 1 || (name[1] == '.' && namelen == 2))) {
1518                 ret = -EEXIST;
1519                 goto out_drop_write;
1520         }
1521
1522         if (subvol) {
1523                 ret = btrfs_mksubvol(&file->f_path, name, namelen,
1524                                      NULL, transid, readonly, inherit);
1525         } else {
1526                 struct fd src = fdget(fd);
1527                 struct inode *src_inode;
1528                 if (!src.file) {
1529                         ret = -EINVAL;
1530                         goto out_drop_write;
1531                 }
1532
1533                 src_inode = file_inode(src.file);
1534                 if (src_inode->i_sb != file_inode(file)->i_sb) {
1535                         printk(KERN_INFO "btrfs: Snapshot src from "
1536                                "another FS\n");
1537                         ret = -EINVAL;
1538                 } else {
1539                         ret = btrfs_mksubvol(&file->f_path, name, namelen,
1540                                              BTRFS_I(src_inode)->root,
1541                                              transid, readonly, inherit);
1542                 }
1543                 fdput(src);
1544         }
1545 out_drop_write:
1546         mnt_drop_write_file(file);
1547 out:
1548         return ret;
1549 }
1550
1551 static noinline int btrfs_ioctl_snap_create(struct file *file,
1552                                             void __user *arg, int subvol)
1553 {
1554         struct btrfs_ioctl_vol_args *vol_args;
1555         int ret;
1556
1557         vol_args = memdup_user(arg, sizeof(*vol_args));
1558         if (IS_ERR(vol_args))
1559                 return PTR_ERR(vol_args);
1560         vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1561
1562         ret = btrfs_ioctl_snap_create_transid(file, vol_args->name,
1563                                               vol_args->fd, subvol,
1564                                               NULL, false, NULL);
1565
1566         kfree(vol_args);
1567         return ret;
1568 }
1569
1570 static noinline int btrfs_ioctl_snap_create_v2(struct file *file,
1571                                                void __user *arg, int subvol)
1572 {
1573         struct btrfs_ioctl_vol_args_v2 *vol_args;
1574         int ret;
1575         u64 transid = 0;
1576         u64 *ptr = NULL;
1577         bool readonly = false;
1578         struct btrfs_qgroup_inherit *inherit = NULL;
1579
1580         vol_args = memdup_user(arg, sizeof(*vol_args));
1581         if (IS_ERR(vol_args))
1582                 return PTR_ERR(vol_args);
1583         vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0';
1584
1585         if (vol_args->flags &
1586             ~(BTRFS_SUBVOL_CREATE_ASYNC | BTRFS_SUBVOL_RDONLY |
1587               BTRFS_SUBVOL_QGROUP_INHERIT)) {
1588                 ret = -EOPNOTSUPP;
1589                 goto out;
1590         }
1591
1592         if (vol_args->flags & BTRFS_SUBVOL_CREATE_ASYNC)
1593                 ptr = &transid;
1594         if (vol_args->flags & BTRFS_SUBVOL_RDONLY)
1595                 readonly = true;
1596         if (vol_args->flags & BTRFS_SUBVOL_QGROUP_INHERIT) {
1597                 if (vol_args->size > PAGE_CACHE_SIZE) {
1598                         ret = -EINVAL;
1599                         goto out;
1600                 }
1601                 inherit = memdup_user(vol_args->qgroup_inherit, vol_args->size);
1602                 if (IS_ERR(inherit)) {
1603                         ret = PTR_ERR(inherit);
1604                         goto out;
1605                 }
1606         }
1607
1608         ret = btrfs_ioctl_snap_create_transid(file, vol_args->name,
1609                                               vol_args->fd, subvol, ptr,
1610                                               readonly, inherit);
1611
1612         if (ret == 0 && ptr &&
1613             copy_to_user(arg +
1614                          offsetof(struct btrfs_ioctl_vol_args_v2,
1615                                   transid), ptr, sizeof(*ptr)))
1616                 ret = -EFAULT;
1617 out:
1618         kfree(vol_args);
1619         kfree(inherit);
1620         return ret;
1621 }
1622
1623 static noinline int btrfs_ioctl_subvol_getflags(struct file *file,
1624                                                 void __user *arg)
1625 {
1626         struct inode *inode = file_inode(file);
1627         struct btrfs_root *root = BTRFS_I(inode)->root;
1628         int ret = 0;
1629         u64 flags = 0;
1630
1631         if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID)
1632                 return -EINVAL;
1633
1634         down_read(&root->fs_info->subvol_sem);
1635         if (btrfs_root_readonly(root))
1636                 flags |= BTRFS_SUBVOL_RDONLY;
1637         up_read(&root->fs_info->subvol_sem);
1638
1639         if (copy_to_user(arg, &flags, sizeof(flags)))
1640                 ret = -EFAULT;
1641
1642         return ret;
1643 }
1644
1645 static noinline int btrfs_ioctl_subvol_setflags(struct file *file,
1646                                               void __user *arg)
1647 {
1648         struct inode *inode = file_inode(file);
1649         struct btrfs_root *root = BTRFS_I(inode)->root;
1650         struct btrfs_trans_handle *trans;
1651         u64 root_flags;
1652         u64 flags;
1653         int ret = 0;
1654
1655         ret = mnt_want_write_file(file);
1656         if (ret)
1657                 goto out;
1658
1659         if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) {
1660                 ret = -EINVAL;
1661                 goto out_drop_write;
1662         }
1663
1664         if (copy_from_user(&flags, arg, sizeof(flags))) {
1665                 ret = -EFAULT;
1666                 goto out_drop_write;
1667         }
1668
1669         if (flags & BTRFS_SUBVOL_CREATE_ASYNC) {
1670                 ret = -EINVAL;
1671                 goto out_drop_write;
1672         }
1673
1674         if (flags & ~BTRFS_SUBVOL_RDONLY) {
1675                 ret = -EOPNOTSUPP;
1676                 goto out_drop_write;
1677         }
1678
1679         if (!inode_owner_or_capable(inode)) {
1680                 ret = -EACCES;
1681                 goto out_drop_write;
1682         }
1683
1684         down_write(&root->fs_info->subvol_sem);
1685
1686         /* nothing to do */
1687         if (!!(flags & BTRFS_SUBVOL_RDONLY) == btrfs_root_readonly(root))
1688                 goto out_drop_sem;
1689
1690         root_flags = btrfs_root_flags(&root->root_item);
1691         if (flags & BTRFS_SUBVOL_RDONLY)
1692                 btrfs_set_root_flags(&root->root_item,
1693                                      root_flags | BTRFS_ROOT_SUBVOL_RDONLY);
1694         else
1695                 btrfs_set_root_flags(&root->root_item,
1696                                      root_flags & ~BTRFS_ROOT_SUBVOL_RDONLY);
1697
1698         trans = btrfs_start_transaction(root, 1);
1699         if (IS_ERR(trans)) {
1700                 ret = PTR_ERR(trans);
1701                 goto out_reset;
1702         }
1703
1704         ret = btrfs_update_root(trans, root->fs_info->tree_root,
1705                                 &root->root_key, &root->root_item);
1706
1707         btrfs_commit_transaction(trans, root);
1708 out_reset:
1709         if (ret)
1710                 btrfs_set_root_flags(&root->root_item, root_flags);
1711 out_drop_sem:
1712         up_write(&root->fs_info->subvol_sem);
1713 out_drop_write:
1714         mnt_drop_write_file(file);
1715 out:
1716         return ret;
1717 }
1718
1719 /*
1720  * helper to check if the subvolume references other subvolumes
1721  */
1722 static noinline int may_destroy_subvol(struct btrfs_root *root)
1723 {
1724         struct btrfs_path *path;
1725         struct btrfs_key key;
1726         int ret;
1727
1728         path = btrfs_alloc_path();
1729         if (!path)
1730                 return -ENOMEM;
1731
1732         key.objectid = root->root_key.objectid;
1733         key.type = BTRFS_ROOT_REF_KEY;
1734         key.offset = (u64)-1;
1735
1736         ret = btrfs_search_slot(NULL, root->fs_info->tree_root,
1737                                 &key, path, 0, 0);
1738         if (ret < 0)
1739                 goto out;
1740         BUG_ON(ret == 0);
1741
1742         ret = 0;
1743         if (path->slots[0] > 0) {
1744                 path->slots[0]--;
1745                 btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
1746                 if (key.objectid == root->root_key.objectid &&
1747                     key.type == BTRFS_ROOT_REF_KEY)
1748                         ret = -ENOTEMPTY;
1749         }
1750 out:
1751         btrfs_free_path(path);
1752         return ret;
1753 }
1754
1755 static noinline int key_in_sk(struct btrfs_key *key,
1756                               struct btrfs_ioctl_search_key *sk)
1757 {
1758         struct btrfs_key test;
1759         int ret;
1760
1761         test.objectid = sk->min_objectid;
1762         test.type = sk->min_type;
1763         test.offset = sk->min_offset;
1764
1765         ret = btrfs_comp_cpu_keys(key, &test);
1766         if (ret < 0)
1767                 return 0;
1768
1769         test.objectid = sk->max_objectid;
1770         test.type = sk->max_type;
1771         test.offset = sk->max_offset;
1772
1773         ret = btrfs_comp_cpu_keys(key, &test);
1774         if (ret > 0)
1775                 return 0;
1776         return 1;
1777 }
1778
1779 static noinline int copy_to_sk(struct btrfs_root *root,
1780                                struct btrfs_path *path,
1781                                struct btrfs_key *key,
1782                                struct btrfs_ioctl_search_key *sk,
1783                                char *buf,
1784                                unsigned long *sk_offset,
1785                                int *num_found)
1786 {
1787         u64 found_transid;
1788         struct extent_buffer *leaf;
1789         struct btrfs_ioctl_search_header sh;
1790         unsigned long item_off;
1791         unsigned long item_len;
1792         int nritems;
1793         int i;
1794         int slot;
1795         int ret = 0;
1796
1797         leaf = path->nodes[0];
1798         slot = path->slots[0];
1799         nritems = btrfs_header_nritems(leaf);
1800
1801         if (btrfs_header_generation(leaf) > sk->max_transid) {
1802                 i = nritems;
1803                 goto advance_key;
1804         }
1805         found_transid = btrfs_header_generation(leaf);
1806
1807         for (i = slot; i < nritems; i++) {
1808                 item_off = btrfs_item_ptr_offset(leaf, i);
1809                 item_len = btrfs_item_size_nr(leaf, i);
1810
1811                 btrfs_item_key_to_cpu(leaf, key, i);
1812                 if (!key_in_sk(key, sk))
1813                         continue;
1814
1815                 if (sizeof(sh) + item_len > BTRFS_SEARCH_ARGS_BUFSIZE)
1816                         item_len = 0;
1817
1818                 if (sizeof(sh) + item_len + *sk_offset >
1819                     BTRFS_SEARCH_ARGS_BUFSIZE) {
1820                         ret = 1;
1821                         goto overflow;
1822                 }
1823
1824                 sh.objectid = key->objectid;
1825                 sh.offset = key->offset;
1826                 sh.type = key->type;
1827                 sh.len = item_len;
1828                 sh.transid = found_transid;
1829
1830                 /* copy search result header */
1831                 memcpy(buf + *sk_offset, &sh, sizeof(sh));
1832                 *sk_offset += sizeof(sh);
1833
1834                 if (item_len) {
1835                         char *p = buf + *sk_offset;
1836                         /* copy the item */
1837                         read_extent_buffer(leaf, p,
1838                                            item_off, item_len);
1839                         *sk_offset += item_len;
1840                 }
1841                 (*num_found)++;
1842
1843                 if (*num_found >= sk->nr_items)
1844                         break;
1845         }
1846 advance_key:
1847         ret = 0;
1848         if (key->offset < (u64)-1 && key->offset < sk->max_offset)
1849                 key->offset++;
1850         else if (key->type < (u8)-1 && key->type < sk->max_type) {
1851                 key->offset = 0;
1852                 key->type++;
1853         } else if (key->objectid < (u64)-1 && key->objectid < sk->max_objectid) {
1854                 key->offset = 0;
1855                 key->type = 0;
1856                 key->objectid++;
1857         } else
1858                 ret = 1;
1859 overflow:
1860         return ret;
1861 }
1862
1863 static noinline int search_ioctl(struct inode *inode,
1864                                  struct btrfs_ioctl_search_args *args)
1865 {
1866         struct btrfs_root *root;
1867         struct btrfs_key key;
1868         struct btrfs_key max_key;
1869         struct btrfs_path *path;
1870         struct btrfs_ioctl_search_key *sk = &args->key;
1871         struct btrfs_fs_info *info = BTRFS_I(inode)->root->fs_info;
1872         int ret;
1873         int num_found = 0;
1874         unsigned long sk_offset = 0;
1875
1876         path = btrfs_alloc_path();
1877         if (!path)
1878                 return -ENOMEM;
1879
1880         if (sk->tree_id == 0) {
1881                 /* search the root of the inode that was passed */
1882                 root = BTRFS_I(inode)->root;
1883         } else {
1884                 key.objectid = sk->tree_id;
1885                 key.type = BTRFS_ROOT_ITEM_KEY;
1886                 key.offset = (u64)-1;
1887                 root = btrfs_read_fs_root_no_name(info, &key);
1888                 if (IS_ERR(root)) {
1889                         printk(KERN_ERR "could not find root %llu\n",
1890                                sk->tree_id);
1891                         btrfs_free_path(path);
1892                         return -ENOENT;
1893                 }
1894         }
1895
1896         key.objectid = sk->min_objectid;
1897         key.type = sk->min_type;
1898         key.offset = sk->min_offset;
1899
1900         max_key.objectid = sk->max_objectid;
1901         max_key.type = sk->max_type;
1902         max_key.offset = sk->max_offset;
1903
1904         path->keep_locks = 1;
1905
1906         while(1) {
1907                 ret = btrfs_search_forward(root, &key, &max_key, path,
1908                                            sk->min_transid);
1909                 if (ret != 0) {
1910                         if (ret > 0)
1911                                 ret = 0;
1912                         goto err;
1913                 }
1914                 ret = copy_to_sk(root, path, &key, sk, args->buf,
1915                                  &sk_offset, &num_found);
1916                 btrfs_release_path(path);
1917                 if (ret || num_found >= sk->nr_items)
1918                         break;
1919
1920         }
1921         ret = 0;
1922 err:
1923         sk->nr_items = num_found;
1924         btrfs_free_path(path);
1925         return ret;
1926 }
1927
1928 static noinline int btrfs_ioctl_tree_search(struct file *file,
1929                                            void __user *argp)
1930 {
1931          struct btrfs_ioctl_search_args *args;
1932          struct inode *inode;
1933          int ret;
1934
1935         if (!capable(CAP_SYS_ADMIN))
1936                 return -EPERM;
1937
1938         args = memdup_user(argp, sizeof(*args));
1939         if (IS_ERR(args))
1940                 return PTR_ERR(args);
1941
1942         inode = file_inode(file);
1943         ret = search_ioctl(inode, args);
1944         if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
1945                 ret = -EFAULT;
1946         kfree(args);
1947         return ret;
1948 }
1949
1950 /*
1951  * Search INODE_REFs to identify path name of 'dirid' directory
1952  * in a 'tree_id' tree. and sets path name to 'name'.
1953  */
1954 static noinline int btrfs_search_path_in_tree(struct btrfs_fs_info *info,
1955                                 u64 tree_id, u64 dirid, char *name)
1956 {
1957         struct btrfs_root *root;
1958         struct btrfs_key key;
1959         char *ptr;
1960         int ret = -1;
1961         int slot;
1962         int len;
1963         int total_len = 0;
1964         struct btrfs_inode_ref *iref;
1965         struct extent_buffer *l;
1966         struct btrfs_path *path;
1967
1968         if (dirid == BTRFS_FIRST_FREE_OBJECTID) {
1969                 name[0]='\0';
1970                 return 0;
1971         }
1972
1973         path = btrfs_alloc_path();
1974         if (!path)
1975                 return -ENOMEM;
1976
1977         ptr = &name[BTRFS_INO_LOOKUP_PATH_MAX];
1978
1979         key.objectid = tree_id;
1980         key.type = BTRFS_ROOT_ITEM_KEY;
1981         key.offset = (u64)-1;
1982         root = btrfs_read_fs_root_no_name(info, &key);
1983         if (IS_ERR(root)) {
1984                 printk(KERN_ERR "could not find root %llu\n", tree_id);
1985                 ret = -ENOENT;
1986                 goto out;
1987         }
1988
1989         key.objectid = dirid;
1990         key.type = BTRFS_INODE_REF_KEY;
1991         key.offset = (u64)-1;
1992
1993         while(1) {
1994                 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
1995                 if (ret < 0)
1996                         goto out;
1997
1998                 l = path->nodes[0];
1999                 slot = path->slots[0];
2000                 if (ret > 0 && slot > 0)
2001                         slot--;
2002                 btrfs_item_key_to_cpu(l, &key, slot);
2003
2004                 if (ret > 0 && (key.objectid != dirid ||
2005                                 key.type != BTRFS_INODE_REF_KEY)) {
2006                         ret = -ENOENT;
2007                         goto out;
2008                 }
2009
2010                 iref = btrfs_item_ptr(l, slot, struct btrfs_inode_ref);
2011                 len = btrfs_inode_ref_name_len(l, iref);
2012                 ptr -= len + 1;
2013                 total_len += len + 1;
2014                 if (ptr < name)
2015                         goto out;
2016
2017                 *(ptr + len) = '/';
2018                 read_extent_buffer(l, ptr,(unsigned long)(iref + 1), len);
2019
2020                 if (key.offset == BTRFS_FIRST_FREE_OBJECTID)
2021                         break;
2022
2023                 btrfs_release_path(path);
2024                 key.objectid = key.offset;
2025                 key.offset = (u64)-1;
2026                 dirid = key.objectid;
2027         }
2028         if (ptr < name)
2029                 goto out;
2030         memmove(name, ptr, total_len);
2031         name[total_len]='\0';
2032         ret = 0;
2033 out:
2034         btrfs_free_path(path);
2035         return ret;
2036 }
2037
2038 static noinline int btrfs_ioctl_ino_lookup(struct file *file,
2039                                            void __user *argp)
2040 {
2041          struct btrfs_ioctl_ino_lookup_args *args;
2042          struct inode *inode;
2043          int ret;
2044
2045         if (!capable(CAP_SYS_ADMIN))
2046                 return -EPERM;
2047
2048         args = memdup_user(argp, sizeof(*args));
2049         if (IS_ERR(args))
2050                 return PTR_ERR(args);
2051
2052         inode = file_inode(file);
2053
2054         if (args->treeid == 0)
2055                 args->treeid = BTRFS_I(inode)->root->root_key.objectid;
2056
2057         ret = btrfs_search_path_in_tree(BTRFS_I(inode)->root->fs_info,
2058                                         args->treeid, args->objectid,
2059                                         args->name);
2060
2061         if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
2062                 ret = -EFAULT;
2063
2064         kfree(args);
2065         return ret;
2066 }
2067
2068 static noinline int btrfs_ioctl_snap_destroy(struct file *file,
2069                                              void __user *arg)
2070 {
2071         struct dentry *parent = fdentry(file);
2072         struct dentry *dentry;
2073         struct inode *dir = parent->d_inode;
2074         struct inode *inode;
2075         struct btrfs_root *root = BTRFS_I(dir)->root;
2076         struct btrfs_root *dest = NULL;
2077         struct btrfs_ioctl_vol_args *vol_args;
2078         struct btrfs_trans_handle *trans;
2079         struct btrfs_block_rsv block_rsv;
2080         u64 qgroup_reserved;
2081         int namelen;
2082         int ret;
2083         int err = 0;
2084
2085         vol_args = memdup_user(arg, sizeof(*vol_args));
2086         if (IS_ERR(vol_args))
2087                 return PTR_ERR(vol_args);
2088
2089         vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2090         namelen = strlen(vol_args->name);
2091         if (strchr(vol_args->name, '/') ||
2092             strncmp(vol_args->name, "..", namelen) == 0) {
2093                 err = -EINVAL;
2094                 goto out;
2095         }
2096
2097         err = mnt_want_write_file(file);
2098         if (err)
2099                 goto out;
2100
2101         err = mutex_lock_killable_nested(&dir->i_mutex, I_MUTEX_PARENT);
2102         if (err == -EINTR)
2103                 goto out;
2104         dentry = lookup_one_len(vol_args->name, parent, namelen);
2105         if (IS_ERR(dentry)) {
2106                 err = PTR_ERR(dentry);
2107                 goto out_unlock_dir;
2108         }
2109
2110         if (!dentry->d_inode) {
2111                 err = -ENOENT;
2112                 goto out_dput;
2113         }
2114
2115         inode = dentry->d_inode;
2116         dest = BTRFS_I(inode)->root;
2117         if (!capable(CAP_SYS_ADMIN)){
2118                 /*
2119                  * Regular user.  Only allow this with a special mount
2120                  * option, when the user has write+exec access to the
2121                  * subvol root, and when rmdir(2) would have been
2122                  * allowed.
2123                  *
2124                  * Note that this is _not_ check that the subvol is
2125                  * empty or doesn't contain data that we wouldn't
2126                  * otherwise be able to delete.
2127                  *
2128                  * Users who want to delete empty subvols should try
2129                  * rmdir(2).
2130                  */
2131                 err = -EPERM;
2132                 if (!btrfs_test_opt(root, USER_SUBVOL_RM_ALLOWED))
2133                         goto out_dput;
2134
2135                 /*
2136                  * Do not allow deletion if the parent dir is the same
2137                  * as the dir to be deleted.  That means the ioctl
2138                  * must be called on the dentry referencing the root
2139                  * of the subvol, not a random directory contained
2140                  * within it.
2141                  */
2142                 err = -EINVAL;
2143                 if (root == dest)
2144                         goto out_dput;
2145
2146                 err = inode_permission(inode, MAY_WRITE | MAY_EXEC);
2147                 if (err)
2148                         goto out_dput;
2149         }
2150
2151         /* check if subvolume may be deleted by a user */
2152         err = btrfs_may_delete(dir, dentry, 1);
2153         if (err)
2154                 goto out_dput;
2155
2156         if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) {
2157                 err = -EINVAL;
2158                 goto out_dput;
2159         }
2160
2161         mutex_lock(&inode->i_mutex);
2162         err = d_invalidate(dentry);
2163         if (err)
2164                 goto out_unlock;
2165
2166         down_write(&root->fs_info->subvol_sem);
2167
2168         err = may_destroy_subvol(dest);
2169         if (err)
2170                 goto out_up_write;
2171
2172         btrfs_init_block_rsv(&block_rsv, BTRFS_BLOCK_RSV_TEMP);
2173         /*
2174          * One for dir inode, two for dir entries, two for root
2175          * ref/backref.
2176          */
2177         err = btrfs_subvolume_reserve_metadata(root, &block_rsv,
2178                                                5, &qgroup_reserved, true);
2179         if (err)
2180                 goto out_up_write;
2181
2182         trans = btrfs_start_transaction(root, 0);
2183         if (IS_ERR(trans)) {
2184                 err = PTR_ERR(trans);
2185                 goto out_release;
2186         }
2187         trans->block_rsv = &block_rsv;
2188         trans->bytes_reserved = block_rsv.size;
2189
2190         ret = btrfs_unlink_subvol(trans, root, dir,
2191                                 dest->root_key.objectid,
2192                                 dentry->d_name.name,
2193                                 dentry->d_name.len);
2194         if (ret) {
2195                 err = ret;
2196                 btrfs_abort_transaction(trans, root, ret);
2197                 goto out_end_trans;
2198         }
2199
2200         btrfs_record_root_in_trans(trans, dest);
2201
2202         memset(&dest->root_item.drop_progress, 0,
2203                 sizeof(dest->root_item.drop_progress));
2204         dest->root_item.drop_level = 0;
2205         btrfs_set_root_refs(&dest->root_item, 0);
2206
2207         if (!xchg(&dest->orphan_item_inserted, 1)) {
2208                 ret = btrfs_insert_orphan_item(trans,
2209                                         root->fs_info->tree_root,
2210                                         dest->root_key.objectid);
2211                 if (ret) {
2212                         btrfs_abort_transaction(trans, root, ret);
2213                         err = ret;
2214                         goto out_end_trans;
2215                 }
2216         }
2217 out_end_trans:
2218         trans->block_rsv = NULL;
2219         trans->bytes_reserved = 0;
2220         ret = btrfs_end_transaction(trans, root);
2221         if (ret && !err)
2222                 err = ret;
2223         inode->i_flags |= S_DEAD;
2224 out_release:
2225         btrfs_subvolume_release_metadata(root, &block_rsv, qgroup_reserved);
2226 out_up_write:
2227         up_write(&root->fs_info->subvol_sem);
2228 out_unlock:
2229         mutex_unlock(&inode->i_mutex);
2230         if (!err) {
2231                 shrink_dcache_sb(root->fs_info->sb);
2232                 btrfs_invalidate_inodes(dest);
2233                 d_delete(dentry);
2234
2235                 /* the last ref */
2236                 if (dest->cache_inode) {
2237                         iput(dest->cache_inode);
2238                         dest->cache_inode = NULL;
2239                 }
2240         }
2241 out_dput:
2242         dput(dentry);
2243 out_unlock_dir:
2244         mutex_unlock(&dir->i_mutex);
2245         mnt_drop_write_file(file);
2246 out:
2247         kfree(vol_args);
2248         return err;
2249 }
2250
2251 static int btrfs_ioctl_defrag(struct file *file, void __user *argp)
2252 {
2253         struct inode *inode = file_inode(file);
2254         struct btrfs_root *root = BTRFS_I(inode)->root;
2255         struct btrfs_ioctl_defrag_range_args *range;
2256         int ret;
2257
2258         ret = mnt_want_write_file(file);
2259         if (ret)
2260                 return ret;
2261
2262         if (btrfs_root_readonly(root)) {
2263                 ret = -EROFS;
2264                 goto out;
2265         }
2266
2267         switch (inode->i_mode & S_IFMT) {
2268         case S_IFDIR:
2269                 if (!capable(CAP_SYS_ADMIN)) {
2270                         ret = -EPERM;
2271                         goto out;
2272                 }
2273                 ret = btrfs_defrag_root(root);
2274                 if (ret)
2275                         goto out;
2276                 ret = btrfs_defrag_root(root->fs_info->extent_root);
2277                 break;
2278         case S_IFREG:
2279                 if (!(file->f_mode & FMODE_WRITE)) {
2280                         ret = -EINVAL;
2281                         goto out;
2282                 }
2283
2284                 range = kzalloc(sizeof(*range), GFP_KERNEL);
2285                 if (!range) {
2286                         ret = -ENOMEM;
2287                         goto out;
2288                 }
2289
2290                 if (argp) {
2291                         if (copy_from_user(range, argp,
2292                                            sizeof(*range))) {
2293                                 ret = -EFAULT;
2294                                 kfree(range);
2295                                 goto out;
2296                         }
2297                         /* compression requires us to start the IO */
2298                         if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
2299                                 range->flags |= BTRFS_DEFRAG_RANGE_START_IO;
2300                                 range->extent_thresh = (u32)-1;
2301                         }
2302                 } else {
2303                         /* the rest are all set to zero by kzalloc */
2304                         range->len = (u64)-1;
2305                 }
2306                 ret = btrfs_defrag_file(file_inode(file), file,
2307                                         range, 0, 0);
2308                 if (ret > 0)
2309                         ret = 0;
2310                 kfree(range);
2311                 break;
2312         default:
2313                 ret = -EINVAL;
2314         }
2315 out:
2316         mnt_drop_write_file(file);
2317         return ret;
2318 }
2319
2320 static long btrfs_ioctl_add_dev(struct btrfs_root *root, void __user *arg)
2321 {
2322         struct btrfs_ioctl_vol_args *vol_args;
2323         int ret;
2324
2325         if (!capable(CAP_SYS_ADMIN))
2326                 return -EPERM;
2327
2328         if (atomic_xchg(&root->fs_info->mutually_exclusive_operation_running,
2329                         1)) {
2330                 pr_info("btrfs: dev add/delete/balance/replace/resize operation in progress\n");
2331                 return -EINVAL;
2332         }
2333
2334         mutex_lock(&root->fs_info->volume_mutex);
2335         vol_args = memdup_user(arg, sizeof(*vol_args));
2336         if (IS_ERR(vol_args)) {
2337                 ret = PTR_ERR(vol_args);
2338                 goto out;
2339         }
2340
2341         vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2342         ret = btrfs_init_new_device(root, vol_args->name);
2343
2344         kfree(vol_args);
2345 out:
2346         mutex_unlock(&root->fs_info->volume_mutex);
2347         atomic_set(&root->fs_info->mutually_exclusive_operation_running, 0);
2348         return ret;
2349 }
2350
2351 static long btrfs_ioctl_rm_dev(struct file *file, void __user *arg)
2352 {
2353         struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
2354         struct btrfs_ioctl_vol_args *vol_args;
2355         int ret;
2356
2357         if (!capable(CAP_SYS_ADMIN))
2358                 return -EPERM;
2359
2360         ret = mnt_want_write_file(file);
2361         if (ret)
2362                 return ret;
2363
2364         vol_args = memdup_user(arg, sizeof(*vol_args));
2365         if (IS_ERR(vol_args)) {
2366                 ret = PTR_ERR(vol_args);
2367                 goto out;
2368         }
2369
2370         vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2371
2372         if (atomic_xchg(&root->fs_info->mutually_exclusive_operation_running,
2373                         1)) {
2374                 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
2375                 goto out;
2376         }
2377
2378         mutex_lock(&root->fs_info->volume_mutex);
2379         ret = btrfs_rm_device(root, vol_args->name);
2380         mutex_unlock(&root->fs_info->volume_mutex);
2381         atomic_set(&root->fs_info->mutually_exclusive_operation_running, 0);
2382
2383 out:
2384         kfree(vol_args);
2385         mnt_drop_write_file(file);
2386         return ret;
2387 }
2388
2389 static long btrfs_ioctl_fs_info(struct btrfs_root *root, void __user *arg)
2390 {
2391         struct btrfs_ioctl_fs_info_args *fi_args;
2392         struct btrfs_device *device;
2393         struct btrfs_device *next;
2394         struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
2395         int ret = 0;
2396
2397         if (!capable(CAP_SYS_ADMIN))
2398                 return -EPERM;
2399
2400         fi_args = kzalloc(sizeof(*fi_args), GFP_KERNEL);
2401         if (!fi_args)
2402                 return -ENOMEM;
2403
2404         fi_args->num_devices = fs_devices->num_devices;
2405         memcpy(&fi_args->fsid, root->fs_info->fsid, sizeof(fi_args->fsid));
2406
2407         mutex_lock(&fs_devices->device_list_mutex);
2408         list_for_each_entry_safe(device, next, &fs_devices->devices, dev_list) {
2409                 if (device->devid > fi_args->max_id)
2410                         fi_args->max_id = device->devid;
2411         }
2412         mutex_unlock(&fs_devices->device_list_mutex);
2413
2414         if (copy_to_user(arg, fi_args, sizeof(*fi_args)))
2415                 ret = -EFAULT;
2416
2417         kfree(fi_args);
2418         return ret;
2419 }
2420
2421 static long btrfs_ioctl_dev_info(struct btrfs_root *root, void __user *arg)
2422 {
2423         struct btrfs_ioctl_dev_info_args *di_args;
2424         struct btrfs_device *dev;
2425         struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
2426         int ret = 0;
2427         char *s_uuid = NULL;
2428         char empty_uuid[BTRFS_UUID_SIZE] = {0};
2429
2430         if (!capable(CAP_SYS_ADMIN))
2431                 return -EPERM;
2432
2433         di_args = memdup_user(arg, sizeof(*di_args));
2434         if (IS_ERR(di_args))
2435                 return PTR_ERR(di_args);
2436
2437         if (memcmp(empty_uuid, di_args->uuid, BTRFS_UUID_SIZE) != 0)
2438                 s_uuid = di_args->uuid;
2439
2440         mutex_lock(&fs_devices->device_list_mutex);
2441         dev = btrfs_find_device(root->fs_info, di_args->devid, s_uuid, NULL);
2442
2443         if (!dev) {
2444                 ret = -ENODEV;
2445                 goto out;
2446         }
2447
2448         di_args->devid = dev->devid;
2449         di_args->bytes_used = dev->bytes_used;
2450         di_args->total_bytes = dev->total_bytes;
2451         memcpy(di_args->uuid, dev->uuid, sizeof(di_args->uuid));
2452         if (dev->name) {
2453                 struct rcu_string *name;
2454
2455                 rcu_read_lock();
2456                 name = rcu_dereference(dev->name);
2457                 strncpy(di_args->path, name->str, sizeof(di_args->path));
2458                 rcu_read_unlock();
2459                 di_args->path[sizeof(di_args->path) - 1] = 0;
2460         } else {
2461                 di_args->path[0] = '\0';
2462         }
2463
2464 out:
2465         mutex_unlock(&fs_devices->device_list_mutex);
2466         if (ret == 0 && copy_to_user(arg, di_args, sizeof(*di_args)))
2467                 ret = -EFAULT;
2468
2469         kfree(di_args);
2470         return ret;
2471 }
2472
2473 static noinline long btrfs_ioctl_clone(struct file *file, unsigned long srcfd,
2474                                        u64 off, u64 olen, u64 destoff)
2475 {
2476         struct inode *inode = file_inode(file);
2477         struct btrfs_root *root = BTRFS_I(inode)->root;
2478         struct fd src_file;
2479         struct inode *src;
2480         struct btrfs_trans_handle *trans;
2481         struct btrfs_path *path;
2482         struct extent_buffer *leaf;
2483         char *buf;
2484         struct btrfs_key key;
2485         u32 nritems;
2486         int slot;
2487         int ret;
2488         u64 len = olen;
2489         u64 bs = root->fs_info->sb->s_blocksize;
2490         int same_inode = 0;
2491
2492         /*
2493          * TODO:
2494          * - split compressed inline extents.  annoying: we need to
2495          *   decompress into destination's address_space (the file offset
2496          *   may change, so source mapping won't do), then recompress (or
2497          *   otherwise reinsert) a subrange.
2498          * - allow ranges within the same file to be cloned (provided
2499          *   they don't overlap)?
2500          */
2501
2502         /* the destination must be opened for writing */
2503         if (!(file->f_mode & FMODE_WRITE) || (file->f_flags & O_APPEND))
2504                 return -EINVAL;
2505
2506         if (btrfs_root_readonly(root))
2507                 return -EROFS;
2508
2509         ret = mnt_want_write_file(file);
2510         if (ret)
2511                 return ret;
2512
2513         src_file = fdget(srcfd);
2514         if (!src_file.file) {
2515                 ret = -EBADF;
2516                 goto out_drop_write;
2517         }
2518
2519         ret = -EXDEV;
2520         if (src_file.file->f_path.mnt != file->f_path.mnt)
2521                 goto out_fput;
2522
2523         src = file_inode(src_file.file);
2524
2525         ret = -EINVAL;
2526         if (src == inode)
2527                 same_inode = 1;
2528
2529         /* the src must be open for reading */
2530         if (!(src_file.file->f_mode & FMODE_READ))
2531                 goto out_fput;
2532
2533         /* don't make the dst file partly checksummed */
2534         if ((BTRFS_I(src)->flags & BTRFS_INODE_NODATASUM) !=
2535             (BTRFS_I(inode)->flags & BTRFS_INODE_NODATASUM))
2536                 goto out_fput;
2537
2538         ret = -EISDIR;
2539         if (S_ISDIR(src->i_mode) || S_ISDIR(inode->i_mode))
2540                 goto out_fput;
2541
2542         ret = -EXDEV;
2543         if (src->i_sb != inode->i_sb)
2544                 goto out_fput;
2545
2546         ret = -ENOMEM;
2547         buf = vmalloc(btrfs_level_size(root, 0));
2548         if (!buf)
2549                 goto out_fput;
2550
2551         path = btrfs_alloc_path();
2552         if (!path) {
2553                 vfree(buf);
2554                 goto out_fput;
2555         }
2556         path->reada = 2;
2557
2558         if (!same_inode) {
2559                 if (inode < src) {
2560                         mutex_lock_nested(&inode->i_mutex, I_MUTEX_PARENT);
2561                         mutex_lock_nested(&src->i_mutex, I_MUTEX_CHILD);
2562                 } else {
2563                         mutex_lock_nested(&src->i_mutex, I_MUTEX_PARENT);
2564                         mutex_lock_nested(&inode->i_mutex, I_MUTEX_CHILD);
2565                 }
2566         } else {
2567                 mutex_lock(&src->i_mutex);
2568         }
2569
2570         /* determine range to clone */
2571         ret = -EINVAL;
2572         if (off + len > src->i_size || off + len < off)
2573                 goto out_unlock;
2574         if (len == 0)
2575                 olen = len = src->i_size - off;
2576         /* if we extend to eof, continue to block boundary */
2577         if (off + len == src->i_size)
2578                 len = ALIGN(src->i_size, bs) - off;
2579
2580         /* verify the end result is block aligned */
2581         if (!IS_ALIGNED(off, bs) || !IS_ALIGNED(off + len, bs) ||
2582             !IS_ALIGNED(destoff, bs))
2583                 goto out_unlock;
2584
2585         /* verify if ranges are overlapped within the same file */
2586         if (same_inode) {
2587                 if (destoff + len > off && destoff < off + len)
2588                         goto out_unlock;
2589         }
2590
2591         if (destoff > inode->i_size) {
2592                 ret = btrfs_cont_expand(inode, inode->i_size, destoff);
2593                 if (ret)
2594                         goto out_unlock;
2595         }
2596
2597         /* truncate page cache pages from target inode range */
2598         truncate_inode_pages_range(&inode->i_data, destoff,
2599                                    PAGE_CACHE_ALIGN(destoff + len) - 1);
2600
2601         /* do any pending delalloc/csum calc on src, one way or
2602            another, and lock file content */
2603         while (1) {
2604                 struct btrfs_ordered_extent *ordered;
2605                 lock_extent(&BTRFS_I(src)->io_tree, off, off + len - 1);
2606                 ordered = btrfs_lookup_first_ordered_extent(src, off + len - 1);
2607                 if (!ordered &&
2608                     !test_range_bit(&BTRFS_I(src)->io_tree, off, off + len - 1,
2609                                     EXTENT_DELALLOC, 0, NULL))
2610                         break;
2611                 unlock_extent(&BTRFS_I(src)->io_tree, off, off + len - 1);
2612                 if (ordered)
2613                         btrfs_put_ordered_extent(ordered);
2614                 btrfs_wait_ordered_range(src, off, len);
2615         }
2616
2617         /* clone data */
2618         key.objectid = btrfs_ino(src);
2619         key.type = BTRFS_EXTENT_DATA_KEY;
2620         key.offset = 0;
2621
2622         while (1) {
2623                 /*
2624                  * note the key will change type as we walk through the
2625                  * tree.
2626                  */
2627                 ret = btrfs_search_slot(NULL, BTRFS_I(src)->root, &key, path,
2628                                 0, 0);
2629                 if (ret < 0)
2630                         goto out;
2631
2632                 nritems = btrfs_header_nritems(path->nodes[0]);
2633                 if (path->slots[0] >= nritems) {
2634                         ret = btrfs_next_leaf(BTRFS_I(src)->root, path);
2635                         if (ret < 0)
2636                                 goto out;
2637                         if (ret > 0)
2638                                 break;
2639                         nritems = btrfs_header_nritems(path->nodes[0]);
2640                 }
2641                 leaf = path->nodes[0];
2642                 slot = path->slots[0];
2643
2644                 btrfs_item_key_to_cpu(leaf, &key, slot);
2645                 if (btrfs_key_type(&key) > BTRFS_EXTENT_DATA_KEY ||
2646                     key.objectid != btrfs_ino(src))
2647                         break;
2648
2649                 if (btrfs_key_type(&key) == BTRFS_EXTENT_DATA_KEY) {
2650                         struct btrfs_file_extent_item *extent;
2651                         int type;
2652                         u32 size;
2653                         struct btrfs_key new_key;
2654                         u64 disko = 0, diskl = 0;
2655                         u64 datao = 0, datal = 0;
2656                         u8 comp;
2657                         u64 endoff;
2658
2659                         size = btrfs_item_size_nr(leaf, slot);
2660                         read_extent_buffer(leaf, buf,
2661                                            btrfs_item_ptr_offset(leaf, slot),
2662                                            size);
2663
2664                         extent = btrfs_item_ptr(leaf, slot,
2665                                                 struct btrfs_file_extent_item);
2666                         comp = btrfs_file_extent_compression(leaf, extent);
2667                         type = btrfs_file_extent_type(leaf, extent);
2668                         if (type == BTRFS_FILE_EXTENT_REG ||
2669                             type == BTRFS_FILE_EXTENT_PREALLOC) {
2670                                 disko = btrfs_file_extent_disk_bytenr(leaf,
2671                                                                       extent);
2672                                 diskl = btrfs_file_extent_disk_num_bytes(leaf,
2673                                                                  extent);
2674                                 datao = btrfs_file_extent_offset(leaf, extent);
2675                                 datal = btrfs_file_extent_num_bytes(leaf,
2676                                                                     extent);
2677                         } else if (type == BTRFS_FILE_EXTENT_INLINE) {
2678                                 /* take upper bound, may be compressed */
2679                                 datal = btrfs_file_extent_ram_bytes(leaf,
2680                                                                     extent);
2681                         }
2682                         btrfs_release_path(path);
2683
2684                         if (key.offset + datal <= off ||
2685                             key.offset >= off + len - 1)
2686                                 goto next;
2687
2688                         memcpy(&new_key, &key, sizeof(new_key));
2689                         new_key.objectid = btrfs_ino(inode);
2690                         if (off <= key.offset)
2691                                 new_key.offset = key.offset + destoff - off;
2692                         else
2693                                 new_key.offset = destoff;
2694
2695                         /*
2696                          * 1 - adjusting old extent (we may have to split it)
2697                          * 1 - add new extent
2698                          * 1 - inode update
2699                          */
2700                         trans = btrfs_start_transaction(root, 3);
2701                         if (IS_ERR(trans)) {
2702                                 ret = PTR_ERR(trans);
2703                                 goto out;
2704                         }
2705
2706                         if (type == BTRFS_FILE_EXTENT_REG ||
2707                             type == BTRFS_FILE_EXTENT_PREALLOC) {
2708                                 /*
2709                                  *    a  | --- range to clone ---|  b
2710                                  * | ------------- extent ------------- |
2711                                  */
2712
2713                                 /* substract range b */
2714                                 if (key.offset + datal > off + len)
2715                                         datal = off + len - key.offset;
2716
2717                                 /* substract range a */
2718                                 if (off > key.offset) {
2719                                         datao += off - key.offset;
2720                                         datal -= off - key.offset;
2721                                 }
2722
2723                                 ret = btrfs_drop_extents(trans, root, inode,
2724                                                          new_key.offset,
2725                                                          new_key.offset + datal,
2726                                                          1);
2727                                 if (ret) {
2728                                         btrfs_abort_transaction(trans, root,
2729                                                                 ret);
2730                                         btrfs_end_transaction(trans, root);
2731                                         goto out;
2732                                 }
2733
2734                                 ret = btrfs_insert_empty_item(trans, root, path,
2735                                                               &new_key, size);
2736                                 if (ret) {
2737                                         btrfs_abort_transaction(trans, root,
2738                                                                 ret);
2739                                         btrfs_end_transaction(trans, root);
2740                                         goto out;
2741                                 }
2742
2743                                 leaf = path->nodes[0];
2744                                 slot = path->slots[0];
2745                                 write_extent_buffer(leaf, buf,
2746                                             btrfs_item_ptr_offset(leaf, slot),
2747                                             size);
2748
2749                                 extent = btrfs_item_ptr(leaf, slot,
2750                                                 struct btrfs_file_extent_item);
2751
2752                                 /* disko == 0 means it's a hole */
2753                                 if (!disko)
2754                                         datao = 0;
2755
2756                                 btrfs_set_file_extent_offset(leaf, extent,
2757                                                              datao);
2758                                 btrfs_set_file_extent_num_bytes(leaf, extent,
2759                                                                 datal);
2760                                 if (disko) {
2761                                         inode_add_bytes(inode, datal);
2762                                         ret = btrfs_inc_extent_ref(trans, root,
2763                                                         disko, diskl, 0,
2764                                                         root->root_key.objectid,
2765                                                         btrfs_ino(inode),
2766                                                         new_key.offset - datao,
2767                                                         0);
2768                                         if (ret) {
2769                                                 btrfs_abort_transaction(trans,
2770                                                                         root,
2771                                                                         ret);
2772                                                 btrfs_end_transaction(trans,
2773                                                                       root);
2774                                                 goto out;
2775
2776                                         }
2777                                 }
2778                         } else if (type == BTRFS_FILE_EXTENT_INLINE) {
2779                                 u64 skip = 0;
2780                                 u64 trim = 0;
2781                                 if (off > key.offset) {
2782                                         skip = off - key.offset;
2783                                         new_key.offset += skip;
2784                                 }
2785
2786                                 if (key.offset + datal > off + len)
2787                                         trim = key.offset + datal - (off + len);
2788
2789                                 if (comp && (skip || trim)) {
2790                                         ret = -EINVAL;
2791                                         btrfs_end_transaction(trans, root);
2792                                         goto out;
2793                                 }
2794                                 size -= skip + trim;
2795                                 datal -= skip + trim;
2796
2797                                 ret = btrfs_drop_extents(trans, root, inode,
2798                                                          new_key.offset,
2799                                                          new_key.offset + datal,
2800                                                          1);
2801                                 if (ret) {
2802                                         btrfs_abort_transaction(trans, root,
2803                                                                 ret);
2804                                         btrfs_end_transaction(trans, root);
2805                                         goto out;
2806                                 }
2807
2808                                 ret = btrfs_insert_empty_item(trans, root, path,
2809                                                               &new_key, size);
2810                                 if (ret) {
2811                                         btrfs_abort_transaction(trans, root,
2812                                                                 ret);
2813                                         btrfs_end_transaction(trans, root);
2814                                         goto out;
2815                                 }
2816
2817                                 if (skip) {
2818                                         u32 start =
2819                                           btrfs_file_extent_calc_inline_size(0);
2820                                         memmove(buf+start, buf+start+skip,
2821                                                 datal);
2822                                 }
2823
2824                                 leaf = path->nodes[0];
2825                                 slot = path->slots[0];
2826                                 write_extent_buffer(leaf, buf,
2827                                             btrfs_item_ptr_offset(leaf, slot),
2828                                             size);
2829                                 inode_add_bytes(inode, datal);
2830                         }
2831
2832                         btrfs_mark_buffer_dirty(leaf);
2833                         btrfs_release_path(path);
2834
2835                         inode_inc_iversion(inode);
2836                         inode->i_mtime = inode->i_ctime = CURRENT_TIME;
2837
2838                         /*
2839                          * we round up to the block size at eof when
2840                          * determining which extents to clone above,
2841                          * but shouldn't round up the file size
2842                          */
2843                         endoff = new_key.offset + datal;
2844                         if (endoff > destoff+olen)
2845                                 endoff = destoff+olen;
2846                         if (endoff > inode->i_size)
2847                                 btrfs_i_size_write(inode, endoff);
2848
2849                         ret = btrfs_update_inode(trans, root, inode);
2850                         if (ret) {
2851                                 btrfs_abort_transaction(trans, root, ret);
2852                                 btrfs_end_transaction(trans, root);
2853                                 goto out;
2854                         }
2855                         ret = btrfs_end_transaction(trans, root);
2856                 }
2857 next:
2858                 btrfs_release_path(path);
2859                 key.offset++;
2860         }
2861         ret = 0;
2862 out:
2863         btrfs_release_path(path);
2864         unlock_extent(&BTRFS_I(src)->io_tree, off, off + len - 1);
2865 out_unlock:
2866         mutex_unlock(&src->i_mutex);
2867         if (!same_inode)
2868                 mutex_unlock(&inode->i_mutex);
2869         vfree(buf);
2870         btrfs_free_path(path);
2871 out_fput:
2872         fdput(src_file);
2873 out_drop_write:
2874         mnt_drop_write_file(file);
2875         return ret;
2876 }
2877
2878 static long btrfs_ioctl_clone_range(struct file *file, void __user *argp)
2879 {
2880         struct btrfs_ioctl_clone_range_args args;
2881
2882         if (copy_from_user(&args, argp, sizeof(args)))
2883                 return -EFAULT;
2884         return btrfs_ioctl_clone(file, args.src_fd, args.src_offset,
2885                                  args.src_length, args.dest_offset);
2886 }
2887
2888 /*
2889  * there are many ways the trans_start and trans_end ioctls can lead
2890  * to deadlocks.  They should only be used by applications that
2891  * basically own the machine, and have a very in depth understanding
2892  * of all the possible deadlocks and enospc problems.
2893  */
2894 static long btrfs_ioctl_trans_start(struct file *file)
2895 {
2896         struct inode *inode = file_inode(file);
2897         struct btrfs_root *root = BTRFS_I(inode)->root;
2898         struct btrfs_trans_handle *trans;
2899         int ret;
2900
2901         ret = -EPERM;
2902         if (!capable(CAP_SYS_ADMIN))
2903                 goto out;
2904
2905         ret = -EINPROGRESS;
2906         if (file->private_data)
2907                 goto out;
2908
2909         ret = -EROFS;
2910         if (btrfs_root_readonly(root))
2911                 goto out;
2912
2913         ret = mnt_want_write_file(file);
2914         if (ret)
2915                 goto out;
2916
2917         atomic_inc(&root->fs_info->open_ioctl_trans);
2918
2919         ret = -ENOMEM;
2920         trans = btrfs_start_ioctl_transaction(root);
2921         if (IS_ERR(trans))
2922                 goto out_drop;
2923
2924         file->private_data = trans;
2925         return 0;
2926
2927 out_drop:
2928         atomic_dec(&root->fs_info->open_ioctl_trans);
2929         mnt_drop_write_file(file);
2930 out:
2931         return ret;
2932 }
2933
2934 static long btrfs_ioctl_default_subvol(struct file *file, void __user *argp)
2935 {
2936         struct inode *inode = file_inode(file);
2937         struct btrfs_root *root = BTRFS_I(inode)->root;
2938         struct btrfs_root *new_root;
2939         struct btrfs_dir_item *di;
2940         struct btrfs_trans_handle *trans;
2941         struct btrfs_path *path;
2942         struct btrfs_key location;
2943         struct btrfs_disk_key disk_key;
2944         u64 objectid = 0;
2945         u64 dir_id;
2946         int ret;
2947
2948         if (!capable(CAP_SYS_ADMIN))
2949                 return -EPERM;
2950
2951         ret = mnt_want_write_file(file);
2952         if (ret)
2953                 return ret;
2954
2955         if (copy_from_user(&objectid, argp, sizeof(objectid))) {
2956                 ret = -EFAULT;
2957                 goto out;
2958         }
2959
2960         if (!objectid)
2961                 objectid = root->root_key.objectid;
2962
2963         location.objectid = objectid;
2964         location.type = BTRFS_ROOT_ITEM_KEY;
2965         location.offset = (u64)-1;
2966
2967         new_root = btrfs_read_fs_root_no_name(root->fs_info, &location);
2968         if (IS_ERR(new_root)) {
2969                 ret = PTR_ERR(new_root);
2970                 goto out;
2971         }
2972
2973         path = btrfs_alloc_path();
2974         if (!path) {
2975                 ret = -ENOMEM;
2976                 goto out;
2977         }
2978         path->leave_spinning = 1;
2979
2980         trans = btrfs_start_transaction(root, 1);
2981         if (IS_ERR(trans)) {
2982                 btrfs_free_path(path);
2983                 ret = PTR_ERR(trans);
2984                 goto out;
2985         }
2986
2987         dir_id = btrfs_super_root_dir(root->fs_info->super_copy);
2988         di = btrfs_lookup_dir_item(trans, root->fs_info->tree_root, path,
2989                                    dir_id, "default", 7, 1);
2990         if (IS_ERR_OR_NULL(di)) {
2991                 btrfs_free_path(path);
2992                 btrfs_end_transaction(trans, root);
2993                 printk(KERN_ERR "Umm, you don't have the default dir item, "
2994                        "this isn't going to work\n");
2995                 ret = -ENOENT;
2996                 goto out;
2997         }
2998
2999         btrfs_cpu_key_to_disk(&disk_key, &new_root->root_key);
3000         btrfs_set_dir_item_key(path->nodes[0], di, &disk_key);
3001         btrfs_mark_buffer_dirty(path->nodes[0]);
3002         btrfs_free_path(path);
3003
3004         btrfs_set_fs_incompat(root->fs_info, DEFAULT_SUBVOL);
3005         btrfs_end_transaction(trans, root);
3006 out:
3007         mnt_drop_write_file(file);
3008         return ret;
3009 }
3010
3011 void btrfs_get_block_group_info(struct list_head *groups_list,
3012                                 struct btrfs_ioctl_space_info *space)
3013 {
3014         struct btrfs_block_group_cache *block_group;
3015
3016         space->total_bytes = 0;
3017         space->used_bytes = 0;
3018         space->flags = 0;
3019         list_for_each_entry(block_group, groups_list, list) {
3020                 space->flags = block_group->flags;
3021                 space->total_bytes += block_group->key.offset;
3022                 space->used_bytes +=
3023                         btrfs_block_group_used(&block_group->item);
3024         }
3025 }
3026
3027 static long btrfs_ioctl_space_info(struct btrfs_root *root, void __user *arg)
3028 {
3029         struct btrfs_ioctl_space_args space_args;
3030         struct btrfs_ioctl_space_info space;
3031         struct btrfs_ioctl_space_info *dest;
3032         struct btrfs_ioctl_space_info *dest_orig;
3033         struct btrfs_ioctl_space_info __user *user_dest;
3034         struct btrfs_space_info *info;
3035         u64 types[] = {BTRFS_BLOCK_GROUP_DATA,
3036                        BTRFS_BLOCK_GROUP_SYSTEM,
3037                        BTRFS_BLOCK_GROUP_METADATA,
3038                        BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA};
3039         int num_types = 4;
3040         int alloc_size;
3041         int ret = 0;
3042         u64 slot_count = 0;
3043         int i, c;
3044
3045         if (copy_from_user(&space_args,
3046                            (struct btrfs_ioctl_space_args __user *)arg,
3047                            sizeof(space_args)))
3048                 return -EFAULT;
3049
3050         for (i = 0; i < num_types; i++) {
3051                 struct btrfs_space_info *tmp;
3052
3053                 info = NULL;
3054                 rcu_read_lock();
3055                 list_for_each_entry_rcu(tmp, &root->fs_info->space_info,
3056                                         list) {
3057                         if (tmp->flags == types[i]) {
3058                                 info = tmp;
3059                                 break;
3060                         }
3061                 }
3062                 rcu_read_unlock();
3063
3064                 if (!info)
3065                         continue;
3066
3067                 down_read(&info->groups_sem);
3068                 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
3069                         if (!list_empty(&info->block_groups[c]))
3070                                 slot_count++;
3071                 }
3072                 up_read(&info->groups_sem);
3073         }
3074
3075         /* space_slots == 0 means they are asking for a count */
3076         if (space_args.space_slots == 0) {
3077                 space_args.total_spaces = slot_count;
3078                 goto out;
3079         }
3080
3081         slot_count = min_t(u64, space_args.space_slots, slot_count);
3082
3083         alloc_size = sizeof(*dest) * slot_count;
3084
3085         /* we generally have at most 6 or so space infos, one for each raid
3086          * level.  So, a whole page should be more than enough for everyone
3087          */
3088         if (alloc_size > PAGE_CACHE_SIZE)
3089                 return -ENOMEM;
3090
3091         space_args.total_spaces = 0;
3092         dest = kmalloc(alloc_size, GFP_NOFS);
3093         if (!dest)
3094                 return -ENOMEM;
3095         dest_orig = dest;
3096
3097         /* now we have a buffer to copy into */
3098         for (i = 0; i < num_types; i++) {
3099                 struct btrfs_space_info *tmp;
3100
3101                 if (!slot_count)
3102                         break;
3103
3104                 info = NULL;
3105                 rcu_read_lock();
3106                 list_for_each_entry_rcu(tmp, &root->fs_info->space_info,
3107                                         list) {
3108                         if (tmp->flags == types[i]) {
3109                                 info = tmp;
3110                                 break;
3111                         }
3112                 }
3113                 rcu_read_unlock();
3114
3115                 if (!info)
3116                         continue;
3117                 down_read(&info->groups_sem);
3118                 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
3119                         if (!list_empty(&info->block_groups[c])) {
3120                                 btrfs_get_block_group_info(
3121                                         &info->block_groups[c], &space);
3122                                 memcpy(dest, &space, sizeof(space));
3123                                 dest++;
3124                                 space_args.total_spaces++;
3125                                 slot_count--;
3126                         }
3127                         if (!slot_count)
3128                                 break;
3129                 }
3130                 up_read(&info->groups_sem);
3131         }
3132
3133         user_dest = (struct btrfs_ioctl_space_info __user *)
3134                 (arg + sizeof(struct btrfs_ioctl_space_args));
3135
3136         if (copy_to_user(user_dest, dest_orig, alloc_size))
3137                 ret = -EFAULT;
3138
3139         kfree(dest_orig);
3140 out:
3141         if (ret == 0 && copy_to_user(arg, &space_args, sizeof(space_args)))
3142                 ret = -EFAULT;
3143
3144         return ret;
3145 }
3146
3147 /*
3148  * there are many ways the trans_start and trans_end ioctls can lead
3149  * to deadlocks.  They should only be used by applications that
3150  * basically own the machine, and have a very in depth understanding
3151  * of all the possible deadlocks and enospc problems.
3152  */
3153 long btrfs_ioctl_trans_end(struct file *file)
3154 {
3155         struct inode *inode = file_inode(file);
3156         struct btrfs_root *root = BTRFS_I(inode)->root;
3157         struct btrfs_trans_handle *trans;
3158
3159         trans = file->private_data;
3160         if (!trans)
3161                 return -EINVAL;
3162         file->private_data = NULL;
3163
3164         btrfs_end_transaction(trans, root);
3165
3166         atomic_dec(&root->fs_info->open_ioctl_trans);
3167
3168         mnt_drop_write_file(file);
3169         return 0;
3170 }
3171
3172 static noinline long btrfs_ioctl_start_sync(struct btrfs_root *root,
3173                                             void __user *argp)
3174 {
3175         struct btrfs_trans_handle *trans;
3176         u64 transid;
3177         int ret;
3178
3179         trans = btrfs_attach_transaction_barrier(root);
3180         if (IS_ERR(trans)) {
3181                 if (PTR_ERR(trans) != -ENOENT)
3182                         return PTR_ERR(trans);
3183
3184                 /* No running transaction, don't bother */
3185                 transid = root->fs_info->last_trans_committed;
3186                 goto out;
3187         }
3188         transid = trans->transid;
3189         ret = btrfs_commit_transaction_async(trans, root, 0);
3190         if (ret) {
3191                 btrfs_end_transaction(trans, root);
3192                 return ret;
3193         }
3194 out:
3195         if (argp)
3196                 if (copy_to_user(argp, &transid, sizeof(transid)))
3197                         return -EFAULT;
3198         return 0;
3199 }
3200
3201 static noinline long btrfs_ioctl_wait_sync(struct btrfs_root *root,
3202                                            void __user *argp)
3203 {
3204         u64 transid;
3205
3206         if (argp) {
3207                 if (copy_from_user(&transid, argp, sizeof(transid)))
3208                         return -EFAULT;
3209         } else {
3210                 transid = 0;  /* current trans */
3211         }
3212         return btrfs_wait_for_commit(root, transid);
3213 }
3214
3215 static long btrfs_ioctl_scrub(struct file *file, void __user *arg)
3216 {
3217         struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
3218         struct btrfs_ioctl_scrub_args *sa;
3219         int ret;
3220
3221         if (!capable(CAP_SYS_ADMIN))
3222                 return -EPERM;
3223
3224         sa = memdup_user(arg, sizeof(*sa));
3225         if (IS_ERR(sa))
3226                 return PTR_ERR(sa);
3227
3228         if (!(sa->flags & BTRFS_SCRUB_READONLY)) {
3229                 ret = mnt_want_write_file(file);
3230                 if (ret)
3231                         goto out;
3232         }
3233
3234         ret = btrfs_scrub_dev(root->fs_info, sa->devid, sa->start, sa->end,
3235                               &sa->progress, sa->flags & BTRFS_SCRUB_READONLY,
3236                               0);
3237
3238         if (copy_to_user(arg, sa, sizeof(*sa)))
3239                 ret = -EFAULT;
3240
3241         if (!(sa->flags & BTRFS_SCRUB_READONLY))
3242                 mnt_drop_write_file(file);
3243 out:
3244         kfree(sa);
3245         return ret;
3246 }
3247
3248 static long btrfs_ioctl_scrub_cancel(struct btrfs_root *root, void __user *arg)
3249 {
3250         if (!capable(CAP_SYS_ADMIN))
3251                 return -EPERM;
3252
3253         return btrfs_scrub_cancel(root->fs_info);
3254 }
3255
3256 static long btrfs_ioctl_scrub_progress(struct btrfs_root *root,
3257                                        void __user *arg)
3258 {
3259         struct btrfs_ioctl_scrub_args *sa;
3260         int ret;
3261
3262         if (!capable(CAP_SYS_ADMIN))
3263                 return -EPERM;
3264
3265         sa = memdup_user(arg, sizeof(*sa));
3266         if (IS_ERR(sa))
3267                 return PTR_ERR(sa);
3268
3269         ret = btrfs_scrub_progress(root, sa->devid, &sa->progress);
3270
3271         if (copy_to_user(arg, sa, sizeof(*sa)))
3272                 ret = -EFAULT;
3273
3274         kfree(sa);
3275         return ret;
3276 }
3277
3278 static long btrfs_ioctl_get_dev_stats(struct btrfs_root *root,
3279                                       void __user *arg)
3280 {
3281         struct btrfs_ioctl_get_dev_stats *sa;
3282         int ret;
3283
3284         sa = memdup_user(arg, sizeof(*sa));
3285         if (IS_ERR(sa))
3286                 return PTR_ERR(sa);
3287
3288         if ((sa->flags & BTRFS_DEV_STATS_RESET) && !capable(CAP_SYS_ADMIN)) {
3289                 kfree(sa);
3290                 return -EPERM;
3291         }
3292
3293         ret = btrfs_get_dev_stats(root, sa);
3294
3295         if (copy_to_user(arg, sa, sizeof(*sa)))
3296                 ret = -EFAULT;
3297
3298         kfree(sa);
3299         return ret;
3300 }
3301
3302 static long btrfs_ioctl_dev_replace(struct btrfs_root *root, void __user *arg)
3303 {
3304         struct btrfs_ioctl_dev_replace_args *p;
3305         int ret;
3306
3307         if (!capable(CAP_SYS_ADMIN))
3308                 return -EPERM;
3309
3310         p = memdup_user(arg, sizeof(*p));
3311         if (IS_ERR(p))
3312                 return PTR_ERR(p);
3313
3314         switch (p->cmd) {
3315         case BTRFS_IOCTL_DEV_REPLACE_CMD_START:
3316                 if (atomic_xchg(
3317                         &root->fs_info->mutually_exclusive_operation_running,
3318                         1)) {
3319                         pr_info("btrfs: dev add/delete/balance/replace/resize operation in progress\n");
3320                         ret = -EINPROGRESS;
3321                 } else {
3322                         ret = btrfs_dev_replace_start(root, p);
3323                         atomic_set(
3324                          &root->fs_info->mutually_exclusive_operation_running,
3325                          0);
3326                 }
3327                 break;
3328         case BTRFS_IOCTL_DEV_REPLACE_CMD_STATUS:
3329                 btrfs_dev_replace_status(root->fs_info, p);
3330                 ret = 0;
3331                 break;
3332         case BTRFS_IOCTL_DEV_REPLACE_CMD_CANCEL:
3333                 ret = btrfs_dev_replace_cancel(root->fs_info, p);
3334                 break;
3335         default:
3336                 ret = -EINVAL;
3337                 break;
3338         }
3339
3340         if (copy_to_user(arg, p, sizeof(*p)))
3341                 ret = -EFAULT;
3342
3343         kfree(p);
3344         return ret;
3345 }
3346
3347 static long btrfs_ioctl_ino_to_path(struct btrfs_root *root, void __user *arg)
3348 {
3349         int ret = 0;
3350         int i;
3351         u64 rel_ptr;
3352         int size;
3353         struct btrfs_ioctl_ino_path_args *ipa = NULL;
3354         struct inode_fs_paths *ipath = NULL;
3355         struct btrfs_path *path;
3356
3357         if (!capable(CAP_DAC_READ_SEARCH))
3358                 return -EPERM;
3359
3360         path = btrfs_alloc_path();
3361         if (!path) {
3362                 ret = -ENOMEM;
3363                 goto out;
3364         }
3365
3366         ipa = memdup_user(arg, sizeof(*ipa));
3367         if (IS_ERR(ipa)) {
3368                 ret = PTR_ERR(ipa);
3369                 ipa = NULL;
3370                 goto out;
3371         }
3372
3373         size = min_t(u32, ipa->size, 4096);
3374         ipath = init_ipath(size, root, path);
3375         if (IS_ERR(ipath)) {
3376                 ret = PTR_ERR(ipath);
3377                 ipath = NULL;
3378                 goto out;
3379         }
3380
3381         ret = paths_from_inode(ipa->inum, ipath);
3382         if (ret < 0)
3383                 goto out;
3384
3385         for (i = 0; i < ipath->fspath->elem_cnt; ++i) {
3386                 rel_ptr = ipath->fspath->val[i] -
3387                           (u64)(unsigned long)ipath->fspath->val;
3388                 ipath->fspath->val[i] = rel_ptr;
3389         }
3390
3391         ret = copy_to_user((void *)(unsigned long)ipa->fspath,
3392                            (void *)(unsigned long)ipath->fspath, size);
3393         if (ret) {
3394                 ret = -EFAULT;
3395                 goto out;
3396         }
3397
3398 out:
3399         btrfs_free_path(path);
3400         free_ipath(ipath);
3401         kfree(ipa);
3402
3403         return ret;
3404 }
3405
3406 static int build_ino_list(u64 inum, u64 offset, u64 root, void *ctx)
3407 {
3408         struct btrfs_data_container *inodes = ctx;
3409         const size_t c = 3 * sizeof(u64);
3410
3411         if (inodes->bytes_left >= c) {
3412                 inodes->bytes_left -= c;
3413                 inodes->val[inodes->elem_cnt] = inum;
3414                 inodes->val[inodes->elem_cnt + 1] = offset;
3415                 inodes->val[inodes->elem_cnt + 2] = root;
3416                 inodes->elem_cnt += 3;
3417         } else {
3418                 inodes->bytes_missing += c - inodes->bytes_left;
3419                 inodes->bytes_left = 0;
3420                 inodes->elem_missed += 3;
3421         }
3422
3423         return 0;
3424 }
3425
3426 static long btrfs_ioctl_logical_to_ino(struct btrfs_root *root,
3427                                         void __user *arg)
3428 {
3429         int ret = 0;
3430         int size;
3431         struct btrfs_ioctl_logical_ino_args *loi;
3432         struct btrfs_data_container *inodes = NULL;
3433         struct btrfs_path *path = NULL;
3434
3435         if (!capable(CAP_SYS_ADMIN))
3436                 return -EPERM;
3437
3438         loi = memdup_user(arg, sizeof(*loi));
3439         if (IS_ERR(loi)) {
3440                 ret = PTR_ERR(loi);
3441                 loi = NULL;
3442                 goto out;
3443         }
3444
3445         path = btrfs_alloc_path();
3446         if (!path) {
3447                 ret = -ENOMEM;
3448                 goto out;
3449         }
3450
3451         size = min_t(u32, loi->size, 64 * 1024);
3452         inodes = init_data_container(size);
3453         if (IS_ERR(inodes)) {
3454                 ret = PTR_ERR(inodes);
3455                 inodes = NULL;
3456                 goto out;
3457         }
3458
3459         ret = iterate_inodes_from_logical(loi->logical, root->fs_info, path,
3460                                           build_ino_list, inodes);
3461         if (ret == -EINVAL)
3462                 ret = -ENOENT;
3463         if (ret < 0)
3464                 goto out;
3465
3466         ret = copy_to_user((void *)(unsigned long)loi->inodes,
3467                            (void *)(unsigned long)inodes, size);
3468         if (ret)
3469                 ret = -EFAULT;
3470
3471 out:
3472         btrfs_free_path(path);
3473         vfree(inodes);
3474         kfree(loi);
3475
3476         return ret;
3477 }
3478
3479 void update_ioctl_balance_args(struct btrfs_fs_info *fs_info, int lock,
3480                                struct btrfs_ioctl_balance_args *bargs)
3481 {
3482         struct btrfs_balance_control *bctl = fs_info->balance_ctl;
3483
3484         bargs->flags = bctl->flags;
3485
3486         if (atomic_read(&fs_info->balance_running))
3487                 bargs->state |= BTRFS_BALANCE_STATE_RUNNING;
3488         if (atomic_read(&fs_info->balance_pause_req))
3489                 bargs->state |= BTRFS_BALANCE_STATE_PAUSE_REQ;
3490         if (atomic_read(&fs_info->balance_cancel_req))
3491                 bargs->state |= BTRFS_BALANCE_STATE_CANCEL_REQ;
3492
3493         memcpy(&bargs->data, &bctl->data, sizeof(bargs->data));
3494         memcpy(&bargs->meta, &bctl->meta, sizeof(bargs->meta));
3495         memcpy(&bargs->sys, &bctl->sys, sizeof(bargs->sys));
3496
3497         if (lock) {
3498                 spin_lock(&fs_info->balance_lock);
3499                 memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat));
3500                 spin_unlock(&fs_info->balance_lock);
3501         } else {
3502                 memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat));
3503         }
3504 }
3505
3506 static long btrfs_ioctl_balance(struct file *file, void __user *arg)
3507 {
3508         struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
3509         struct btrfs_fs_info *fs_info = root->fs_info;
3510         struct btrfs_ioctl_balance_args *bargs;
3511         struct btrfs_balance_control *bctl;
3512         bool need_unlock; /* for mut. excl. ops lock */
3513         int ret;
3514
3515         if (!capable(CAP_SYS_ADMIN))
3516                 return -EPERM;
3517
3518         ret = mnt_want_write_file(file);
3519         if (ret)
3520                 return ret;
3521
3522 again:
3523         if (!atomic_xchg(&fs_info->mutually_exclusive_operation_running, 1)) {
3524                 mutex_lock(&fs_info->volume_mutex);
3525                 mutex_lock(&fs_info->balance_mutex);
3526                 need_unlock = true;
3527                 goto locked;
3528         }
3529
3530         /*
3531          * mut. excl. ops lock is locked.  Three possibilites:
3532          *   (1) some other op is running
3533          *   (2) balance is running
3534          *   (3) balance is paused -- special case (think resume)
3535          */
3536         mutex_lock(&fs_info->balance_mutex);
3537         if (fs_info->balance_ctl) {
3538                 /* this is either (2) or (3) */
3539                 if (!atomic_read(&fs_info->balance_running)) {
3540                         mutex_unlock(&fs_info->balance_mutex);
3541                         if (!mutex_trylock(&fs_info->volume_mutex))
3542                                 goto again;
3543                         mutex_lock(&fs_info->balance_mutex);
3544
3545                         if (fs_info->balance_ctl &&
3546                             !atomic_read(&fs_info->balance_running)) {
3547                                 /* this is (3) */
3548                                 need_unlock = false;
3549                                 goto locked;
3550                         }
3551
3552                         mutex_unlock(&fs_info->balance_mutex);
3553                         mutex_unlock(&fs_info->volume_mutex);
3554                         goto again;
3555                 } else {
3556                         /* this is (2) */
3557                         mutex_unlock(&fs_info->balance_mutex);
3558                         ret = -EINPROGRESS;
3559                         goto out;
3560                 }
3561         } else {
3562                 /* this is (1) */
3563                 mutex_unlock(&fs_info->balance_mutex);
3564                 pr_info("btrfs: dev add/delete/balance/replace/resize operation in progress\n");
3565                 ret = -EINVAL;
3566                 goto out;
3567         }
3568
3569 locked:
3570         BUG_ON(!atomic_read(&fs_info->mutually_exclusive_operation_running));
3571
3572         if (arg) {
3573                 bargs = memdup_user(arg, sizeof(*bargs));
3574                 if (IS_ERR(bargs)) {
3575                         ret = PTR_ERR(bargs);
3576                         goto out_unlock;
3577                 }
3578
3579                 if (bargs->flags & BTRFS_BALANCE_RESUME) {
3580                         if (!fs_info->balance_ctl) {
3581                                 ret = -ENOTCONN;
3582                                 goto out_bargs;
3583                         }
3584
3585                         bctl = fs_info->balance_ctl;
3586                         spin_lock(&fs_info->balance_lock);
3587                         bctl->flags |= BTRFS_BALANCE_RESUME;
3588                         spin_unlock(&fs_info->balance_lock);
3589
3590                         goto do_balance;
3591                 }
3592         } else {
3593                 bargs = NULL;
3594         }
3595
3596         if (fs_info->balance_ctl) {
3597                 ret = -EINPROGRESS;
3598                 goto out_bargs;
3599         }
3600
3601         bctl = kzalloc(sizeof(*bctl), GFP_NOFS);
3602         if (!bctl) {
3603                 ret = -ENOMEM;
3604                 goto out_bargs;
3605         }
3606
3607         bctl->fs_info = fs_info;
3608         if (arg) {
3609                 memcpy(&bctl->data, &bargs->data, sizeof(bctl->data));
3610                 memcpy(&bctl->meta, &bargs->meta, sizeof(bctl->meta));
3611                 memcpy(&bctl->sys, &bargs->sys, sizeof(bctl->sys));
3612
3613                 bctl->flags = bargs->flags;
3614         } else {
3615                 /* balance everything - no filters */
3616                 bctl->flags |= BTRFS_BALANCE_TYPE_MASK;
3617         }
3618
3619 do_balance:
3620         /*
3621          * Ownership of bctl and mutually_exclusive_operation_running
3622          * goes to to btrfs_balance.  bctl is freed in __cancel_balance,
3623          * or, if restriper was paused all the way until unmount, in
3624          * free_fs_info.  mutually_exclusive_operation_running is
3625          * cleared in __cancel_balance.
3626          */
3627         need_unlock = false;
3628
3629         ret = btrfs_balance(bctl, bargs);
3630
3631         if (arg) {
3632                 if (copy_to_user(arg, bargs, sizeof(*bargs)))
3633                         ret = -EFAULT;
3634         }
3635
3636 out_bargs:
3637         kfree(bargs);
3638 out_unlock:
3639         mutex_unlock(&fs_info->balance_mutex);
3640         mutex_unlock(&fs_info->volume_mutex);
3641         if (need_unlock)
3642                 atomic_set(&fs_info->mutually_exclusive_operation_running, 0);
3643 out:
3644         mnt_drop_write_file(file);
3645         return ret;
3646 }
3647
3648 static long btrfs_ioctl_balance_ctl(struct btrfs_root *root, int cmd)
3649 {
3650         if (!capable(CAP_SYS_ADMIN))
3651                 return -EPERM;
3652
3653         switch (cmd) {
3654         case BTRFS_BALANCE_CTL_PAUSE:
3655                 return btrfs_pause_balance(root->fs_info);
3656         case BTRFS_BALANCE_CTL_CANCEL:
3657                 return btrfs_cancel_balance(root->fs_info);
3658         }
3659
3660         return -EINVAL;
3661 }
3662
3663 static long btrfs_ioctl_balance_progress(struct btrfs_root *root,
3664                                          void __user *arg)
3665 {
3666         struct btrfs_fs_info *fs_info = root->fs_info;
3667         struct btrfs_ioctl_balance_args *bargs;
3668         int ret = 0;
3669
3670         if (!capable(CAP_SYS_ADMIN))
3671                 return -EPERM;
3672
3673         mutex_lock(&fs_info->balance_mutex);
3674         if (!fs_info->balance_ctl) {
3675                 ret = -ENOTCONN;
3676                 goto out;
3677         }
3678
3679         bargs = kzalloc(sizeof(*bargs), GFP_NOFS);
3680         if (!bargs) {
3681                 ret = -ENOMEM;
3682                 goto out;
3683         }
3684
3685         update_ioctl_balance_args(fs_info, 1, bargs);
3686
3687         if (copy_to_user(arg, bargs, sizeof(*bargs)))
3688                 ret = -EFAULT;
3689
3690         kfree(bargs);
3691 out:
3692         mutex_unlock(&fs_info->balance_mutex);
3693         return ret;
3694 }
3695
3696 static long btrfs_ioctl_quota_ctl(struct file *file, void __user *arg)
3697 {
3698         struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
3699         struct btrfs_ioctl_quota_ctl_args *sa;
3700         struct btrfs_trans_handle *trans = NULL;
3701         int ret;
3702         int err;
3703
3704         if (!capable(CAP_SYS_ADMIN))
3705                 return -EPERM;
3706
3707         ret = mnt_want_write_file(file);
3708         if (ret)
3709                 return ret;
3710
3711         sa = memdup_user(arg, sizeof(*sa));
3712         if (IS_ERR(sa)) {
3713                 ret = PTR_ERR(sa);
3714                 goto drop_write;
3715         }
3716
3717         down_write(&root->fs_info->subvol_sem);
3718         trans = btrfs_start_transaction(root->fs_info->tree_root, 2);
3719         if (IS_ERR(trans)) {
3720                 ret = PTR_ERR(trans);
3721                 goto out;
3722         }
3723
3724         switch (sa->cmd) {
3725         case BTRFS_QUOTA_CTL_ENABLE:
3726                 ret = btrfs_quota_enable(trans, root->fs_info);
3727                 break;
3728         case BTRFS_QUOTA_CTL_DISABLE:
3729                 ret = btrfs_quota_disable(trans, root->fs_info);
3730                 break;
3731         default:
3732                 ret = -EINVAL;
3733                 break;
3734         }
3735
3736         err = btrfs_commit_transaction(trans, root->fs_info->tree_root);
3737         if (err && !ret)
3738                 ret = err;
3739 out:
3740         kfree(sa);
3741         up_write(&root->fs_info->subvol_sem);
3742 drop_write:
3743         mnt_drop_write_file(file);
3744         return ret;
3745 }
3746
3747 static long btrfs_ioctl_qgroup_assign(struct file *file, void __user *arg)
3748 {
3749         struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
3750         struct btrfs_ioctl_qgroup_assign_args *sa;
3751         struct btrfs_trans_handle *trans;
3752         int ret;
3753         int err;
3754
3755         if (!capable(CAP_SYS_ADMIN))
3756                 return -EPERM;
3757
3758         ret = mnt_want_write_file(file);
3759         if (ret)
3760                 return ret;
3761
3762         sa = memdup_user(arg, sizeof(*sa));
3763         if (IS_ERR(sa)) {
3764                 ret = PTR_ERR(sa);
3765                 goto drop_write;
3766         }
3767
3768         trans = btrfs_join_transaction(root);
3769         if (IS_ERR(trans)) {
3770                 ret = PTR_ERR(trans);
3771                 goto out;
3772         }
3773
3774         /* FIXME: check if the IDs really exist */
3775         if (sa->assign) {
3776                 ret = btrfs_add_qgroup_relation(trans, root->fs_info,
3777                                                 sa->src, sa->dst);
3778         } else {
3779                 ret = btrfs_del_qgroup_relation(trans, root->fs_info,
3780                                                 sa->src, sa->dst);
3781         }
3782
3783         err = btrfs_end_transaction(trans, root);
3784         if (err && !ret)
3785                 ret = err;
3786
3787 out:
3788         kfree(sa);
3789 drop_write:
3790         mnt_drop_write_file(file);
3791         return ret;
3792 }
3793
3794 static long btrfs_ioctl_qgroup_create(struct file *file, void __user *arg)
3795 {
3796         struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
3797         struct btrfs_ioctl_qgroup_create_args *sa;
3798         struct btrfs_trans_handle *trans;
3799         int ret;
3800         int err;
3801
3802         if (!capable(CAP_SYS_ADMIN))
3803                 return -EPERM;
3804
3805         ret = mnt_want_write_file(file);
3806         if (ret)
3807                 return ret;
3808
3809         sa = memdup_user(arg, sizeof(*sa));
3810         if (IS_ERR(sa)) {
3811                 ret = PTR_ERR(sa);
3812                 goto drop_write;
3813         }
3814
3815         if (!sa->qgroupid) {
3816                 ret = -EINVAL;
3817                 goto out;
3818         }
3819
3820         trans = btrfs_join_transaction(root);
3821         if (IS_ERR(trans)) {
3822                 ret = PTR_ERR(trans);
3823                 goto out;
3824         }
3825
3826         /* FIXME: check if the IDs really exist */
3827         if (sa->create) {
3828                 ret = btrfs_create_qgroup(trans, root->fs_info, sa->qgroupid,
3829                                           NULL);
3830         } else {
3831                 ret = btrfs_remove_qgroup(trans, root->fs_info, sa->qgroupid);
3832         }
3833
3834         err = btrfs_end_transaction(trans, root);
3835         if (err && !ret)
3836                 ret = err;
3837
3838 out:
3839         kfree(sa);
3840 drop_write:
3841         mnt_drop_write_file(file);
3842         return ret;
3843 }
3844
3845 static long btrfs_ioctl_qgroup_limit(struct file *file, void __user *arg)
3846 {
3847         struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
3848         struct btrfs_ioctl_qgroup_limit_args *sa;
3849         struct btrfs_trans_handle *trans;
3850         int ret;
3851         int err;
3852         u64 qgroupid;
3853
3854         if (!capable(CAP_SYS_ADMIN))
3855                 return -EPERM;
3856
3857         ret = mnt_want_write_file(file);
3858         if (ret)
3859                 return ret;
3860
3861         sa = memdup_user(arg, sizeof(*sa));
3862         if (IS_ERR(sa)) {
3863                 ret = PTR_ERR(sa);
3864                 goto drop_write;
3865         }
3866
3867         trans = btrfs_join_transaction(root);
3868         if (IS_ERR(trans)) {
3869                 ret = PTR_ERR(trans);
3870                 goto out;
3871         }
3872
3873         qgroupid = sa->qgroupid;
3874         if (!qgroupid) {
3875                 /* take the current subvol as qgroup */
3876                 qgroupid = root->root_key.objectid;
3877         }
3878
3879         /* FIXME: check if the IDs really exist */
3880         ret = btrfs_limit_qgroup(trans, root->fs_info, qgroupid, &sa->lim);
3881
3882         err = btrfs_end_transaction(trans, root);
3883         if (err && !ret)
3884                 ret = err;
3885
3886 out:
3887         kfree(sa);
3888 drop_write:
3889         mnt_drop_write_file(file);
3890         return ret;
3891 }
3892
3893 static long btrfs_ioctl_quota_rescan(struct file *file, void __user *arg)
3894 {
3895         struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
3896         struct btrfs_ioctl_quota_rescan_args *qsa;
3897         int ret;
3898
3899         if (!capable(CAP_SYS_ADMIN))
3900                 return -EPERM;
3901
3902         ret = mnt_want_write_file(file);
3903         if (ret)
3904                 return ret;
3905
3906         qsa = memdup_user(arg, sizeof(*qsa));
3907         if (IS_ERR(qsa)) {
3908                 ret = PTR_ERR(qsa);
3909                 goto drop_write;
3910         }
3911
3912         if (qsa->flags) {
3913                 ret = -EINVAL;
3914                 goto out;
3915         }
3916
3917         ret = btrfs_qgroup_rescan(root->fs_info);
3918
3919 out:
3920         kfree(qsa);
3921 drop_write:
3922         mnt_drop_write_file(file);
3923         return ret;
3924 }
3925
3926 static long btrfs_ioctl_quota_rescan_status(struct file *file, void __user *arg)
3927 {
3928         struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
3929         struct btrfs_ioctl_quota_rescan_args *qsa;
3930         int ret = 0;
3931
3932         if (!capable(CAP_SYS_ADMIN))
3933                 return -EPERM;
3934
3935         qsa = kzalloc(sizeof(*qsa), GFP_NOFS);
3936         if (!qsa)
3937                 return -ENOMEM;
3938
3939         if (root->fs_info->qgroup_flags & BTRFS_QGROUP_STATUS_FLAG_RESCAN) {
3940                 qsa->flags = 1;
3941                 qsa->progress = root->fs_info->qgroup_rescan_progress.objectid;
3942         }
3943
3944         if (copy_to_user(arg, qsa, sizeof(*qsa)))
3945                 ret = -EFAULT;
3946
3947         kfree(qsa);
3948         return ret;
3949 }
3950
3951 static long btrfs_ioctl_quota_rescan_wait(struct file *file, void __user *arg)
3952 {
3953         struct btrfs_root *root = BTRFS_I(fdentry(file)->d_inode)->root;
3954
3955         if (!capable(CAP_SYS_ADMIN))
3956                 return -EPERM;
3957
3958         return btrfs_qgroup_wait_for_completion(root->fs_info);
3959 }
3960
3961 static long btrfs_ioctl_set_received_subvol(struct file *file,
3962                                             void __user *arg)
3963 {
3964         struct btrfs_ioctl_received_subvol_args *sa = NULL;
3965         struct inode *inode = file_inode(file);
3966         struct btrfs_root *root = BTRFS_I(inode)->root;
3967         struct btrfs_root_item *root_item = &root->root_item;
3968         struct btrfs_trans_handle *trans;
3969         struct timespec ct = CURRENT_TIME;
3970         int ret = 0;
3971
3972         ret = mnt_want_write_file(file);
3973         if (ret < 0)
3974                 return ret;
3975
3976         down_write(&root->fs_info->subvol_sem);
3977
3978         if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) {
3979                 ret = -EINVAL;
3980                 goto out;
3981         }
3982
3983         if (btrfs_root_readonly(root)) {
3984                 ret = -EROFS;
3985                 goto out;
3986         }
3987
3988         if (!inode_owner_or_capable(inode)) {
3989                 ret = -EACCES;
3990                 goto out;
3991         }
3992
3993         sa = memdup_user(arg, sizeof(*sa));
3994         if (IS_ERR(sa)) {
3995                 ret = PTR_ERR(sa);
3996                 sa = NULL;
3997                 goto out;
3998         }
3999
4000         trans = btrfs_start_transaction(root, 1);
4001         if (IS_ERR(trans)) {
4002                 ret = PTR_ERR(trans);
4003                 trans = NULL;
4004                 goto out;
4005         }
4006
4007         sa->rtransid = trans->transid;
4008         sa->rtime.sec = ct.tv_sec;
4009         sa->rtime.nsec = ct.tv_nsec;
4010
4011         memcpy(root_item->received_uuid, sa->uuid, BTRFS_UUID_SIZE);
4012         btrfs_set_root_stransid(root_item, sa->stransid);
4013         btrfs_set_root_rtransid(root_item, sa->rtransid);
4014         btrfs_set_stack_timespec_sec(&root_item->stime, sa->stime.sec);
4015         btrfs_set_stack_timespec_nsec(&root_item->stime, sa->stime.nsec);
4016         btrfs_set_stack_timespec_sec(&root_item->rtime, sa->rtime.sec);
4017         btrfs_set_stack_timespec_nsec(&root_item->rtime, sa->rtime.nsec);
4018
4019         ret = btrfs_update_root(trans, root->fs_info->tree_root,
4020                                 &root->root_key, &root->root_item);
4021         if (ret < 0) {
4022                 btrfs_end_transaction(trans, root);
4023                 trans = NULL;
4024                 goto out;
4025         } else {
4026                 ret = btrfs_commit_transaction(trans, root);
4027                 if (ret < 0)
4028                         goto out;
4029         }
4030
4031         ret = copy_to_user(arg, sa, sizeof(*sa));
4032         if (ret)
4033                 ret = -EFAULT;
4034
4035 out:
4036         kfree(sa);
4037         up_write(&root->fs_info->subvol_sem);
4038         mnt_drop_write_file(file);
4039         return ret;
4040 }
4041
4042 static int btrfs_ioctl_get_fslabel(struct file *file, void __user *arg)
4043 {
4044         struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4045         const char *label = root->fs_info->super_copy->label;
4046         size_t len = strnlen(label, BTRFS_LABEL_SIZE);
4047         int ret;
4048
4049         if (len == BTRFS_LABEL_SIZE) {
4050                 pr_warn("btrfs: label is too long, return the first %zu bytes\n",
4051                         --len);
4052         }
4053
4054         mutex_lock(&root->fs_info->volume_mutex);
4055         ret = copy_to_user(arg, label, len);
4056         mutex_unlock(&root->fs_info->volume_mutex);
4057
4058         return ret ? -EFAULT : 0;
4059 }
4060
4061 static int btrfs_ioctl_set_fslabel(struct file *file, void __user *arg)
4062 {
4063         struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4064         struct btrfs_super_block *super_block = root->fs_info->super_copy;
4065         struct btrfs_trans_handle *trans;
4066         char label[BTRFS_LABEL_SIZE];
4067         int ret;
4068
4069         if (!capable(CAP_SYS_ADMIN))
4070                 return -EPERM;
4071
4072         if (copy_from_user(label, arg, sizeof(label)))
4073                 return -EFAULT;
4074
4075         if (strnlen(label, BTRFS_LABEL_SIZE) == BTRFS_LABEL_SIZE) {
4076                 pr_err("btrfs: unable to set label with more than %d bytes\n",
4077                        BTRFS_LABEL_SIZE - 1);
4078                 return -EINVAL;
4079         }
4080
4081         ret = mnt_want_write_file(file);
4082         if (ret)
4083                 return ret;
4084
4085         mutex_lock(&root->fs_info->volume_mutex);
4086         trans = btrfs_start_transaction(root, 0);
4087         if (IS_ERR(trans)) {
4088                 ret = PTR_ERR(trans);
4089                 goto out_unlock;
4090         }
4091
4092         strcpy(super_block->label, label);
4093         ret = btrfs_end_transaction(trans, root);
4094
4095 out_unlock:
4096         mutex_unlock(&root->fs_info->volume_mutex);
4097         mnt_drop_write_file(file);
4098         return ret;
4099 }
4100
4101 long btrfs_ioctl(struct file *file, unsigned int
4102                 cmd, unsigned long arg)
4103 {
4104         struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4105         void __user *argp = (void __user *)arg;
4106
4107         switch (cmd) {
4108         case FS_IOC_GETFLAGS:
4109                 return btrfs_ioctl_getflags(file, argp);
4110         case FS_IOC_SETFLAGS:
4111                 return btrfs_ioctl_setflags(file, argp);
4112         case FS_IOC_GETVERSION:
4113                 return btrfs_ioctl_getversion(file, argp);
4114         case FITRIM:
4115                 return btrfs_ioctl_fitrim(file, argp);
4116         case BTRFS_IOC_SNAP_CREATE:
4117                 return btrfs_ioctl_snap_create(file, argp, 0);
4118         case BTRFS_IOC_SNAP_CREATE_V2:
4119                 return btrfs_ioctl_snap_create_v2(file, argp, 0);
4120         case BTRFS_IOC_SUBVOL_CREATE:
4121                 return btrfs_ioctl_snap_create(file, argp, 1);
4122         case BTRFS_IOC_SUBVOL_CREATE_V2:
4123                 return btrfs_ioctl_snap_create_v2(file, argp, 1);
4124         case BTRFS_IOC_SNAP_DESTROY:
4125                 return btrfs_ioctl_snap_destroy(file, argp);
4126         case BTRFS_IOC_SUBVOL_GETFLAGS:
4127                 return btrfs_ioctl_subvol_getflags(file, argp);
4128         case BTRFS_IOC_SUBVOL_SETFLAGS:
4129                 return btrfs_ioctl_subvol_setflags(file, argp);
4130         case BTRFS_IOC_DEFAULT_SUBVOL:
4131                 return btrfs_ioctl_default_subvol(file, argp);
4132         case BTRFS_IOC_DEFRAG:
4133                 return btrfs_ioctl_defrag(file, NULL);
4134         case BTRFS_IOC_DEFRAG_RANGE:
4135                 return btrfs_ioctl_defrag(file, argp);
4136         case BTRFS_IOC_RESIZE:
4137                 return btrfs_ioctl_resize(file, argp);
4138         case BTRFS_IOC_ADD_DEV:
4139                 return btrfs_ioctl_add_dev(root, argp);
4140         case BTRFS_IOC_RM_DEV:
4141                 return btrfs_ioctl_rm_dev(file, argp);
4142         case BTRFS_IOC_FS_INFO:
4143                 return btrfs_ioctl_fs_info(root, argp);
4144         case BTRFS_IOC_DEV_INFO:
4145                 return btrfs_ioctl_dev_info(root, argp);
4146         case BTRFS_IOC_BALANCE:
4147                 return btrfs_ioctl_balance(file, NULL);
4148         case BTRFS_IOC_CLONE:
4149                 return btrfs_ioctl_clone(file, arg, 0, 0, 0);
4150         case BTRFS_IOC_CLONE_RANGE:
4151                 return btrfs_ioctl_clone_range(file, argp);
4152         case BTRFS_IOC_TRANS_START:
4153                 return btrfs_ioctl_trans_start(file);
4154         case BTRFS_IOC_TRANS_END:
4155                 return btrfs_ioctl_trans_end(file);
4156         case BTRFS_IOC_TREE_SEARCH:
4157                 return btrfs_ioctl_tree_search(file, argp);
4158         case BTRFS_IOC_INO_LOOKUP:
4159                 return btrfs_ioctl_ino_lookup(file, argp);
4160         case BTRFS_IOC_INO_PATHS:
4161                 return btrfs_ioctl_ino_to_path(root, argp);
4162         case BTRFS_IOC_LOGICAL_INO:
4163                 return btrfs_ioctl_logical_to_ino(root, argp);
4164         case BTRFS_IOC_SPACE_INFO:
4165                 return btrfs_ioctl_space_info(root, argp);
4166         case BTRFS_IOC_SYNC:
4167                 btrfs_sync_fs(file->f_dentry->d_sb, 1);
4168                 return 0;
4169         case BTRFS_IOC_START_SYNC:
4170                 return btrfs_ioctl_start_sync(root, argp);
4171         case BTRFS_IOC_WAIT_SYNC:
4172                 return btrfs_ioctl_wait_sync(root, argp);
4173         case BTRFS_IOC_SCRUB:
4174                 return btrfs_ioctl_scrub(file, argp);
4175         case BTRFS_IOC_SCRUB_CANCEL:
4176                 return btrfs_ioctl_scrub_cancel(root, argp);
4177         case BTRFS_IOC_SCRUB_PROGRESS:
4178                 return btrfs_ioctl_scrub_progress(root, argp);
4179         case BTRFS_IOC_BALANCE_V2:
4180                 return btrfs_ioctl_balance(file, argp);
4181         case BTRFS_IOC_BALANCE_CTL:
4182                 return btrfs_ioctl_balance_ctl(root, arg);
4183         case BTRFS_IOC_BALANCE_PROGRESS:
4184                 return btrfs_ioctl_balance_progress(root, argp);
4185         case BTRFS_IOC_SET_RECEIVED_SUBVOL:
4186                 return btrfs_ioctl_set_received_subvol(file, argp);
4187         case BTRFS_IOC_SEND:
4188                 return btrfs_ioctl_send(file, argp);
4189         case BTRFS_IOC_GET_DEV_STATS:
4190                 return btrfs_ioctl_get_dev_stats(root, argp);
4191         case BTRFS_IOC_QUOTA_CTL:
4192                 return btrfs_ioctl_quota_ctl(file, argp);
4193         case BTRFS_IOC_QGROUP_ASSIGN:
4194                 return btrfs_ioctl_qgroup_assign(file, argp);
4195         case BTRFS_IOC_QGROUP_CREATE:
4196                 return btrfs_ioctl_qgroup_create(file, argp);
4197         case BTRFS_IOC_QGROUP_LIMIT:
4198                 return btrfs_ioctl_qgroup_limit(file, argp);
4199         case BTRFS_IOC_QUOTA_RESCAN:
4200                 return btrfs_ioctl_quota_rescan(file, argp);
4201         case BTRFS_IOC_QUOTA_RESCAN_STATUS:
4202                 return btrfs_ioctl_quota_rescan_status(file, argp);
4203         case BTRFS_IOC_QUOTA_RESCAN_WAIT:
4204                 return btrfs_ioctl_quota_rescan_wait(file, argp);
4205         case BTRFS_IOC_DEV_REPLACE:
4206                 return btrfs_ioctl_dev_replace(root, argp);
4207         case BTRFS_IOC_GET_FSLABEL:
4208                 return btrfs_ioctl_get_fslabel(file, argp);
4209         case BTRFS_IOC_SET_FSLABEL:
4210                 return btrfs_ioctl_set_fslabel(file, argp);
4211         }
4212
4213         return -ENOTTY;
4214 }