fs/btrfs/ioctl.c

   1 /*
   2  * Copyright (C) 2007 Oracle.  All rights reserved.
   3  *
   4  * This program is free software; you can redistribute it and/or
   5  * modify it under the terms of the GNU General Public
   6  * License v2 as published by the Free Software Foundation.
   7  *
   8  * This program is distributed in the hope that it will be useful,
   9  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  10  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  11  * General Public License for more details.
  12  *
  13  * You should have received a copy of the GNU General Public
  14  * License along with this program; if not, write to the
  15  * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
  16  * Boston, MA 021110-1307, USA.
  17  */
  18
  19 #include <linux/kernel.h>
  20 #include <linux/bio.h>
  21 #include <linux/buffer_head.h>
  22 #include <linux/file.h>
  23 #include <linux/fs.h>
  24 #include <linux/fsnotify.h>
  25 #include <linux/pagemap.h>
  26 #include <linux/highmem.h>
  27 #include <linux/time.h>
  28 #include <linux/init.h>
  29 #include <linux/string.h>
  30 #include <linux/backing-dev.h>
  31 #include <linux/mount.h>
  32 #include <linux/mpage.h>
  33 #include <linux/namei.h>
  34 #include <linux/swap.h>
  35 #include <linux/writeback.h>
  36 #include <linux/statfs.h>
  37 #include <linux/compat.h>
  38 #include <linux/bit_spinlock.h>
  39 #include <linux/security.h>
  40 #include <linux/xattr.h>
  41 #include <linux/vmalloc.h>
  42 #include <linux/slab.h>
  43 #include <linux/blkdev.h>
  44 #include <linux/uuid.h>
  45 #include <linux/btrfs.h>
  46 #include "compat.h"
  47 #include "ctree.h"
  48 #include "disk-io.h"
  49 #include "transaction.h"
  50 #include "btrfs_inode.h"
  51 #include "print-tree.h"
  52 #include "volumes.h"
  53 #include "locking.h"
  54 #include "inode-map.h"
  55 #include "backref.h"
  56 #include "rcu-string.h"
  57 #include "send.h"
  58 #include "dev-replace.h"
  59
  60 /* Mask out flags that are inappropriate for the given type of inode. */
  61 static inline __u32 btrfs_mask_flags(umode_t mode, __u32 flags)
  62 {
  63         if (S_ISDIR(mode))
  64                 return flags;
  65         else if (S_ISREG(mode))
  66                 return flags & ~FS_DIRSYNC_FL;
  67         else
  68                 return flags & (FS_NODUMP_FL | FS_NOATIME_FL);
  69 }
  70
  71 /*
  72  * Export inode flags to the format expected by the FS_IOC_GETFLAGS ioctl.
  73  */
  74 static unsigned int btrfs_flags_to_ioctl(unsigned int flags)
  75 {
  76         unsigned int iflags = 0;
  77
  78         if (flags & BTRFS_INODE_SYNC)
  79                 iflags |= FS_SYNC_FL;
  80         if (flags & BTRFS_INODE_IMMUTABLE)
  81                 iflags |= FS_IMMUTABLE_FL;
  82         if (flags & BTRFS_INODE_APPEND)
  83                 iflags |= FS_APPEND_FL;
  84         if (flags & BTRFS_INODE_NODUMP)
  85                 iflags |= FS_NODUMP_FL;
  86         if (flags & BTRFS_INODE_NOATIME)
  87                 iflags |= FS_NOATIME_FL;
  88         if (flags & BTRFS_INODE_DIRSYNC)
  89                 iflags |= FS_DIRSYNC_FL;
  90         if (flags & BTRFS_INODE_NODATACOW)
  91                 iflags |= FS_NOCOW_FL;
  92
  93         if ((flags & BTRFS_INODE_COMPRESS) && !(flags & BTRFS_INODE_NOCOMPRESS))
  94                 iflags |= FS_COMPR_FL;
  95         else if (flags & BTRFS_INODE_NOCOMPRESS)
  96                 iflags |= FS_NOCOMP_FL;
  97
  98         return iflags;
  99 }
 100
 101 /*
 102  * Update inode->i_flags based on the btrfs internal flags.
 103  */
 104 void btrfs_update_iflags(struct inode *inode)
 105 {
 106         struct btrfs_inode *ip = BTRFS_I(inode);
 107
 108         inode->i_flags &= ~(S_SYNC|S_APPEND|S_IMMUTABLE|S_NOATIME|S_DIRSYNC);
 109
 110         if (ip->flags & BTRFS_INODE_SYNC)
 111                 inode->i_flags |= S_SYNC;
 112         if (ip->flags & BTRFS_INODE_IMMUTABLE)
 113                 inode->i_flags |= S_IMMUTABLE;
 114         if (ip->flags & BTRFS_INODE_APPEND)
 115                 inode->i_flags |= S_APPEND;
 116         if (ip->flags & BTRFS_INODE_NOATIME)
 117                 inode->i_flags |= S_NOATIME;
 118         if (ip->flags & BTRFS_INODE_DIRSYNC)
 119                 inode->i_flags |= S_DIRSYNC;
 120 }
 121
 122 /*
 123  * Inherit flags from the parent inode.
 124  *
 125  * Currently only the compression flags and the cow flags are inherited.
 126  */
 127 void btrfs_inherit_iflags(struct inode *inode, struct inode *dir)
 128 {
 129         unsigned int flags;
 130
 131         if (!dir)
 132                 return;
 133
 134         flags = BTRFS_I(dir)->flags;
 135
 136         if (flags & BTRFS_INODE_NOCOMPRESS) {
 137                 BTRFS_I(inode)->flags &= ~BTRFS_INODE_COMPRESS;
 138                 BTRFS_I(inode)->flags |= BTRFS_INODE_NOCOMPRESS;
 139         } else if (flags & BTRFS_INODE_COMPRESS) {
 140                 BTRFS_I(inode)->flags &= ~BTRFS_INODE_NOCOMPRESS;
 141                 BTRFS_I(inode)->flags |= BTRFS_INODE_COMPRESS;
 142         }
 143
 144         if (flags & BTRFS_INODE_NODATACOW) {
 145                 BTRFS_I(inode)->flags |= BTRFS_INODE_NODATACOW;
 146                 if (S_ISREG(inode->i_mode))
 147                         BTRFS_I(inode)->flags |= BTRFS_INODE_NODATASUM;
 148         }
 149
 150         btrfs_update_iflags(inode);
 151 }
 152
 153 static int btrfs_ioctl_getflags(struct file *file, void __user *arg)
 154 {
 155         struct btrfs_inode *ip = BTRFS_I(file_inode(file));
 156         unsigned int flags = btrfs_flags_to_ioctl(ip->flags);
 157
 158         if (copy_to_user(arg, &flags, sizeof(flags)))
 159                 return -EFAULT;
 160         return 0;
 161 }
 162
 163 static int check_flags(unsigned int flags)
 164 {
 165         if (flags & ~(FS_IMMUTABLE_FL | FS_APPEND_FL | \
 166                       FS_NOATIME_FL | FS_NODUMP_FL | \
 167                       FS_SYNC_FL | FS_DIRSYNC_FL | \
 168                       FS_NOCOMP_FL | FS_COMPR_FL |
 169                       FS_NOCOW_FL))
 170                 return -EOPNOTSUPP;
 171
 172         if ((flags & FS_NOCOMP_FL) && (flags & FS_COMPR_FL))
 173                 return -EINVAL;
 174
 175         return 0;
 176 }
 177
 178 static int btrfs_ioctl_setflags(struct file *file, void __user *arg)
 179 {
 180         struct inode *inode = file_inode(file);
 181         struct btrfs_inode *ip = BTRFS_I(inode);
 182         struct btrfs_root *root = ip->root;
 183         struct btrfs_trans_handle *trans;
 184         unsigned int flags, oldflags;
 185         int ret;
 186         u64 ip_oldflags;
 187         unsigned int i_oldflags;
 188         umode_t mode;
 189
 190         if (btrfs_root_readonly(root))
 191                 return -EROFS;
 192
 193         if (copy_from_user(&flags, arg, sizeof(flags)))
 194                 return -EFAULT;
 195
 196         ret = check_flags(flags);
 197         if (ret)
 198                 return ret;
 199
 200         if (!inode_owner_or_capable(inode))
 201                 return -EACCES;
 202
 203         ret = mnt_want_write_file(file);
 204         if (ret)
 205                 return ret;
 206
 207         mutex_lock(&inode->i_mutex);
 208
 209         ip_oldflags = ip->flags;
 210         i_oldflags = inode->i_flags;
 211         mode = inode->i_mode;
 212
 213         flags = btrfs_mask_flags(inode->i_mode, flags);
 214         oldflags = btrfs_flags_to_ioctl(ip->flags);
 215         if ((flags ^ oldflags) & (FS_APPEND_FL | FS_IMMUTABLE_FL)) {
 216                 if (!capable(CAP_LINUX_IMMUTABLE)) {
 217                         ret = -EPERM;
 218                         goto out_unlock;
 219                 }
 220         }
 221
 222         if (flags & FS_SYNC_FL)
 223                 ip->flags |= BTRFS_INODE_SYNC;
 224         else
 225                 ip->flags &= ~BTRFS_INODE_SYNC;
 226         if (flags & FS_IMMUTABLE_FL)
 227                 ip->flags |= BTRFS_INODE_IMMUTABLE;
 228         else
 229                 ip->flags &= ~BTRFS_INODE_IMMUTABLE;
 230         if (flags & FS_APPEND_FL)
 231                 ip->flags |= BTRFS_INODE_APPEND;
 232         else
 233                 ip->flags &= ~BTRFS_INODE_APPEND;
 234         if (flags & FS_NODUMP_FL)
 235                 ip->flags |= BTRFS_INODE_NODUMP;
 236         else
 237                 ip->flags &= ~BTRFS_INODE_NODUMP;
 238         if (flags & FS_NOATIME_FL)
 239                 ip->flags |= BTRFS_INODE_NOATIME;
 240         else
 241                 ip->flags &= ~BTRFS_INODE_NOATIME;
 242         if (flags & FS_DIRSYNC_FL)
 243                 ip->flags |= BTRFS_INODE_DIRSYNC;
 244         else
 245                 ip->flags &= ~BTRFS_INODE_DIRSYNC;
 246         if (flags & FS_NOCOW_FL) {
 247                 if (S_ISREG(mode)) {
 248                         /*
 249                          * It's safe to turn csums off here, no extents exist.
 250                          * Otherwise we want the flag to reflect the real COW
 251                          * status of the file and will not set it.
 252                          */
 253                         if (inode->i_size == 0)
 254                                 ip->flags |= BTRFS_INODE_NODATACOW
 255                                            | BTRFS_INODE_NODATASUM;
 256                 } else {
 257                         ip->flags |= BTRFS_INODE_NODATACOW;
 258                 }
 259         } else {
 260                 /*
 261                  * Revert back under same assuptions as above
 262                  */
 263                 if (S_ISREG(mode)) {
 264                         if (inode->i_size == 0)
 265                                 ip->flags &= ~(BTRFS_INODE_NODATACOW
 266                                              | BTRFS_INODE_NODATASUM);
 267                 } else {
 268                         ip->flags &= ~BTRFS_INODE_NODATACOW;
 269                 }
 270         }
 271
 272         /*
 273          * The COMPRESS flag can only be changed by users, while the NOCOMPRESS
 274          * flag may be changed automatically if compression code won't make
 275          * things smaller.
 276          */
 277         if (flags & FS_NOCOMP_FL) {
 278                 ip->flags &= ~BTRFS_INODE_COMPRESS;
 279                 ip->flags |= BTRFS_INODE_NOCOMPRESS;
 280         } else if (flags & FS_COMPR_FL) {
 281                 ip->flags |= BTRFS_INODE_COMPRESS;
 282                 ip->flags &= ~BTRFS_INODE_NOCOMPRESS;
 283         } else {
 284                 ip->flags &= ~(BTRFS_INODE_COMPRESS | BTRFS_INODE_NOCOMPRESS);
 285         }
 286
 287         trans = btrfs_start_transaction(root, 1);
 288         if (IS_ERR(trans)) {
 289                 ret = PTR_ERR(trans);
 290                 goto out_drop;
 291         }
 292
 293         btrfs_update_iflags(inode);
 294         inode_inc_iversion(inode);
 295         inode->i_ctime = CURRENT_TIME;
 296         ret = btrfs_update_inode(trans, root, inode);
 297
 298         btrfs_end_transaction(trans, root);
 299  out_drop:
 300         if (ret) {
 301                 ip->flags = ip_oldflags;
 302                 inode->i_flags = i_oldflags;
 303         }
 304
 305  out_unlock:
 306         mutex_unlock(&inode->i_mutex);
 307         mnt_drop_write_file(file);
 308         return ret;
 309 }
 310
 311 static int btrfs_ioctl_getversion(struct file *file, int __user *arg)
 312 {
 313         struct inode *inode = file_inode(file);
 314
 315         return put_user(inode->i_generation, arg);
 316 }
 317
 318 static noinline int btrfs_ioctl_fitrim(struct file *file, void __user *arg)
 319 {
 320         struct btrfs_fs_info *fs_info = btrfs_sb(fdentry(file)->d_sb);
 321         struct btrfs_device *device;
 322         struct request_queue *q;
 323         struct fstrim_range range;
 324         u64 minlen = ULLONG_MAX;
 325         u64 num_devices = 0;
 326         u64 total_bytes = btrfs_super_total_bytes(fs_info->super_copy);
 327         int ret;
 328
 329         if (!capable(CAP_SYS_ADMIN))
 330                 return -EPERM;
 331
 332         rcu_read_lock();
 333         list_for_each_entry_rcu(device, &fs_info->fs_devices->devices,
 334                                 dev_list) {
 335                 if (!device->bdev)
 336                         continue;
 337                 q = bdev_get_queue(device->bdev);
 338                 if (blk_queue_discard(q)) {
 339                         num_devices++;
 340                         minlen = min((u64)q->limits.discard_granularity,
 341                                      minlen);
 342                 }
 343         }
 344         rcu_read_unlock();
 345
 346         if (!num_devices)
 347                 return -EOPNOTSUPP;
 348         if (copy_from_user(&range, arg, sizeof(range)))
 349                 return -EFAULT;
 350         if (range.start > total_bytes ||
 351             range.len < fs_info->sb->s_blocksize)
 352                 return -EINVAL;
 353
 354         range.len = min(range.len, total_bytes - range.start);
 355         range.minlen = max(range.minlen, minlen);
 356         ret = btrfs_trim_fs(fs_info->tree_root, &range);
 357         if (ret < 0)
 358                 return ret;
 359
 360         if (copy_to_user(arg, &range, sizeof(range)))
 361                 return -EFAULT;
 362
 363         return 0;
 364 }
 365
 366 static noinline int create_subvol(struct inode *dir,
 367                                   struct dentry *dentry,
 368                                   char *name, int namelen,
 369                                   u64 *async_transid,
 370                                   struct btrfs_qgroup_inherit *inherit)
 371 {
 372         struct btrfs_trans_handle *trans;
 373         struct btrfs_key key;
 374         struct btrfs_root_item root_item;
 375         struct btrfs_inode_item *inode_item;
 376         struct extent_buffer *leaf;
 377         struct btrfs_root *root = BTRFS_I(dir)->root;
 378         struct btrfs_root *new_root;
 379         struct btrfs_block_rsv block_rsv;
 380         struct timespec cur_time = CURRENT_TIME;
 381         int ret;
 382         int err;
 383         u64 objectid;
 384         u64 new_dirid = BTRFS_FIRST_FREE_OBJECTID;
 385         u64 index = 0;
 386         u64 qgroup_reserved;
 387         uuid_le new_uuid;
 388
 389         ret = btrfs_find_free_objectid(root->fs_info->tree_root, &objectid);
 390         if (ret)
 391                 return ret;
 392
 393         btrfs_init_block_rsv(&block_rsv, BTRFS_BLOCK_RSV_TEMP);
 394         /*
 395          * The same as the snapshot creation, please see the comment
 396          * of create_snapshot().
 397          */
 398         ret = btrfs_subvolume_reserve_metadata(root, &block_rsv,
 399                                                7, &qgroup_reserved);
 400         if (ret)
 401                 return ret;
 402
 403         trans = btrfs_start_transaction(root, 0);
 404         if (IS_ERR(trans)) {
 405                 ret = PTR_ERR(trans);
 406                 goto out;
 407         }
 408         trans->block_rsv = &block_rsv;
 409         trans->bytes_reserved = block_rsv.size;
 410
 411         ret = btrfs_qgroup_inherit(trans, root->fs_info, 0, objectid, inherit);
 412         if (ret)
 413                 goto fail;
 414
 415         leaf = btrfs_alloc_free_block(trans, root, root->leafsize,
 416                                       0, objectid, NULL, 0, 0, 0);
 417         if (IS_ERR(leaf)) {
 418                 ret = PTR_ERR(leaf);
 419                 goto fail;
 420         }
 421
 422         memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header));
 423         btrfs_set_header_bytenr(leaf, leaf->start);
 424         btrfs_set_header_generation(leaf, trans->transid);
 425         btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV);
 426         btrfs_set_header_owner(leaf, objectid);
 427
 428         write_extent_buffer(leaf, root->fs_info->fsid,
 429                             (unsigned long)btrfs_header_fsid(leaf),
 430                             BTRFS_FSID_SIZE);
 431         write_extent_buffer(leaf, root->fs_info->chunk_tree_uuid,
 432                             (unsigned long)btrfs_header_chunk_tree_uuid(leaf),
 433                             BTRFS_UUID_SIZE);
 434         btrfs_mark_buffer_dirty(leaf);
 435
 436         memset(&root_item, 0, sizeof(root_item));
 437
 438         inode_item = &root_item.inode;
 439         inode_item->generation = cpu_to_le64(1);
 440         inode_item->size = cpu_to_le64(3);
 441         inode_item->nlink = cpu_to_le32(1);
 442         inode_item->nbytes = cpu_to_le64(root->leafsize);
 443         inode_item->mode = cpu_to_le32(S_IFDIR | 0755);
 444
 445         root_item.flags = 0;
 446         root_item.byte_limit = 0;
 447         inode_item->flags = cpu_to_le64(BTRFS_INODE_ROOT_ITEM_INIT);
 448
 449         btrfs_set_root_bytenr(&root_item, leaf->start);
 450         btrfs_set_root_generation(&root_item, trans->transid);
 451         btrfs_set_root_level(&root_item, 0);
 452         btrfs_set_root_refs(&root_item, 1);
 453         btrfs_set_root_used(&root_item, leaf->len);
 454         btrfs_set_root_last_snapshot(&root_item, 0);
 455
 456         btrfs_set_root_generation_v2(&root_item,
 457                         btrfs_root_generation(&root_item));
 458         uuid_le_gen(&new_uuid);
 459         memcpy(root_item.uuid, new_uuid.b, BTRFS_UUID_SIZE);
 460         root_item.otime.sec = cpu_to_le64(cur_time.tv_sec);
 461         root_item.otime.nsec = cpu_to_le32(cur_time.tv_nsec);
 462         root_item.ctime = root_item.otime;
 463         btrfs_set_root_ctransid(&root_item, trans->transid);
 464         btrfs_set_root_otransid(&root_item, trans->transid);
 465
 466         btrfs_tree_unlock(leaf);
 467         free_extent_buffer(leaf);
 468         leaf = NULL;
 469
 470         btrfs_set_root_dirid(&root_item, new_dirid);
 471
 472         key.objectid = objectid;
 473         key.offset = 0;
 474         btrfs_set_key_type(&key, BTRFS_ROOT_ITEM_KEY);
 475         ret = btrfs_insert_root(trans, root->fs_info->tree_root, &key,
 476                                 &root_item);
 477         if (ret)
 478                 goto fail;
 479
 480         key.offset = (u64)-1;
 481         new_root = btrfs_read_fs_root_no_name(root->fs_info, &key);
 482         if (IS_ERR(new_root)) {
 483                 btrfs_abort_transaction(trans, root, PTR_ERR(new_root));
 484                 ret = PTR_ERR(new_root);
 485                 goto fail;
 486         }
 487
 488         btrfs_record_root_in_trans(trans, new_root);
 489
 490         ret = btrfs_create_subvol_root(trans, new_root, new_dirid);
 491         if (ret) {
 492                 /* We potentially lose an unused inode item here */
 493                 btrfs_abort_transaction(trans, root, ret);
 494                 goto fail;
 495         }
 496
 497         /*
 498          * insert the directory item
 499          */
 500         ret = btrfs_set_inode_index(dir, &index);
 501         if (ret) {
 502                 btrfs_abort_transaction(trans, root, ret);
 503                 goto fail;
 504         }
 505
 506         ret = btrfs_insert_dir_item(trans, root,
 507                                     name, namelen, dir, &key,
 508                                     BTRFS_FT_DIR, index);
 509         if (ret) {
 510                 btrfs_abort_transaction(trans, root, ret);
 511                 goto fail;
 512         }
 513
 514         btrfs_i_size_write(dir, dir->i_size + namelen * 2);
 515         ret = btrfs_update_inode(trans, root, dir);
 516         BUG_ON(ret);
 517
 518         ret = btrfs_add_root_ref(trans, root->fs_info->tree_root,
 519                                  objectid, root->root_key.objectid,
 520                                  btrfs_ino(dir), index, name, namelen);
 521
 522         BUG_ON(ret);
 523
 524 fail:
 525         trans->block_rsv = NULL;
 526         trans->bytes_reserved = 0;
 527         if (async_transid) {
 528                 *async_transid = trans->transid;
 529                 err = btrfs_commit_transaction_async(trans, root, 1);
 530                 if (err)
 531                         err = btrfs_commit_transaction(trans, root);
 532         } else {
 533                 err = btrfs_commit_transaction(trans, root);
 534         }
 535         if (err && !ret)
 536                 ret = err;
 537
 538         if (!ret)
 539                 d_instantiate(dentry, btrfs_lookup_dentry(dir, dentry));
 540 out:
 541         btrfs_subvolume_release_metadata(root, &block_rsv, qgroup_reserved);
 542         return ret;
 543 }
 544
 545 static int create_snapshot(struct btrfs_root *root, struct inode *dir,
 546                            struct dentry *dentry, char *name, int namelen,
 547                            u64 *async_transid, bool readonly,
 548                            struct btrfs_qgroup_inherit *inherit)
 549 {
 550         struct inode *inode;
 551         struct btrfs_pending_snapshot *pending_snapshot;
 552         struct btrfs_trans_handle *trans;
 553         int ret;
 554
 555         if (!root->ref_cows)
 556                 return -EINVAL;
 557
 558         pending_snapshot = kzalloc(sizeof(*pending_snapshot), GFP_NOFS);
 559         if (!pending_snapshot)
 560                 return -ENOMEM;
 561
 562         btrfs_init_block_rsv(&pending_snapshot->block_rsv,
 563                              BTRFS_BLOCK_RSV_TEMP);
 564         /*
 565          * 1 - parent dir inode
 566          * 2 - dir entries
 567          * 1 - root item
 568          * 2 - root ref/backref
 569          * 1 - root of snapshot
 570          */
 571         ret = btrfs_subvolume_reserve_metadata(BTRFS_I(dir)->root,
 572                                         &pending_snapshot->block_rsv, 7,
 573                                         &pending_snapshot->qgroup_reserved);
 574         if (ret)
 575                 goto out;
 576
 577         pending_snapshot->dentry = dentry;
 578         pending_snapshot->root = root;
 579         pending_snapshot->readonly = readonly;
 580         pending_snapshot->dir = dir;
 581         pending_snapshot->inherit = inherit;
 582
 583         trans = btrfs_start_transaction(root, 0);
 584         if (IS_ERR(trans)) {
 585                 ret = PTR_ERR(trans);
 586                 goto fail;
 587         }
 588
 589         spin_lock(&root->fs_info->trans_lock);
 590         list_add(&pending_snapshot->list,
 591                  &trans->transaction->pending_snapshots);
 592         spin_unlock(&root->fs_info->trans_lock);
 593         if (async_transid) {
 594                 *async_transid = trans->transid;
 595                 ret = btrfs_commit_transaction_async(trans,
 596                                      root->fs_info->extent_root, 1);
 597                 if (ret)
 598                         ret = btrfs_commit_transaction(trans, root);
 599         } else {
 600                 ret = btrfs_commit_transaction(trans,
 601                                                root->fs_info->extent_root);
 602         }
 603         if (ret)
 604                 goto fail;
 605
 606         ret = pending_snapshot->error;
 607         if (ret)
 608                 goto fail;
 609
 610         ret = btrfs_orphan_cleanup(pending_snapshot->snap);
 611         if (ret)
 612                 goto fail;
 613
 614         inode = btrfs_lookup_dentry(dentry->d_parent->d_inode, dentry);
 615         if (IS_ERR(inode)) {
 616                 ret = PTR_ERR(inode);
 617                 goto fail;
 618         }
 619         BUG_ON(!inode);
 620         d_instantiate(dentry, inode);
 621         ret = 0;
 622 fail:
 623         btrfs_subvolume_release_metadata(BTRFS_I(dir)->root,
 624                                          &pending_snapshot->block_rsv,
 625                                          pending_snapshot->qgroup_reserved);
 626 out:
 627         kfree(pending_snapshot);
 628         return ret;
 629 }
 630
 631 /*  copy of check_sticky in fs/namei.c()
 632 * It's inline, so penalty for filesystems that don't use sticky bit is
 633 * minimal.
 634 */
 635 static inline int btrfs_check_sticky(struct inode *dir, struct inode *inode)
 636 {
 637         kuid_t fsuid = current_fsuid();
 638
 639         if (!(dir->i_mode & S_ISVTX))
 640                 return 0;
 641         if (uid_eq(inode->i_uid, fsuid))
 642                 return 0;
 643         if (uid_eq(dir->i_uid, fsuid))
 644                 return 0;
 645         return !capable(CAP_FOWNER);
 646 }
 647
 648 /*  copy of may_delete in fs/namei.c()
 649  *      Check whether we can remove a link victim from directory dir, check
 650  *  whether the type of victim is right.
 651  *  1. We can't do it if dir is read-only (done in permission())
 652  *  2. We should have write and exec permissions on dir
 653  *  3. We can't remove anything from append-only dir
 654  *  4. We can't do anything with immutable dir (done in permission())
 655  *  5. If the sticky bit on dir is set we should either
 656  *      a. be owner of dir, or
 657  *      b. be owner of victim, or
 658  *      c. have CAP_FOWNER capability
 659  *  6. If the victim is append-only or immutable we can't do antyhing with
 660  *     links pointing to it.
 661  *  7. If we were asked to remove a directory and victim isn't one - ENOTDIR.
 662  *  8. If we were asked to remove a non-directory and victim isn't one - EISDIR.
 663  *  9. We can't remove a root or mountpoint.
 664  * 10. We don't allow removal of NFS sillyrenamed files; it's handled by
 665  *     nfs_async_unlink().
 666  */
 667
 668 static int btrfs_may_delete(struct inode *dir,struct dentry *victim,int isdir)
 669 {
 670         int error;
 671
 672         if (!victim->d_inode)
 673                 return -ENOENT;
 674
 675         BUG_ON(victim->d_parent->d_inode != dir);
 676         audit_inode_child(dir, victim, AUDIT_TYPE_CHILD_DELETE);
 677
 678         error = inode_permission(dir, MAY_WRITE | MAY_EXEC);
 679         if (error)
 680                 return error;
 681         if (IS_APPEND(dir))
 682                 return -EPERM;
 683         if (btrfs_check_sticky(dir, victim->d_inode)||
 684                 IS_APPEND(victim->d_inode)||
 685             IS_IMMUTABLE(victim->d_inode) || IS_SWAPFILE(victim->d_inode))
 686                 return -EPERM;
 687         if (isdir) {
 688                 if (!S_ISDIR(victim->d_inode->i_mode))
 689                         return -ENOTDIR;
 690                 if (IS_ROOT(victim))
 691                         return -EBUSY;
 692         } else if (S_ISDIR(victim->d_inode->i_mode))
 693                 return -EISDIR;
 694         if (IS_DEADDIR(dir))
 695                 return -ENOENT;
 696         if (victim->d_flags & DCACHE_NFSFS_RENAMED)
 697                 return -EBUSY;
 698         return 0;
 699 }
 700
 701 /* copy of may_create in fs/namei.c() */
 702 static inline int btrfs_may_create(struct inode *dir, struct dentry *child)
 703 {
 704         if (child->d_inode)
 705                 return -EEXIST;
 706         if (IS_DEADDIR(dir))
 707                 return -ENOENT;
 708         return inode_permission(dir, MAY_WRITE | MAY_EXEC);
 709 }
 710
 711 /*
 712  * Create a new subvolume below @parent.  This is largely modeled after
 713  * sys_mkdirat and vfs_mkdir, but we only do a single component lookup
 714  * inside this filesystem so it's quite a bit simpler.
 715  */
 716 static noinline int btrfs_mksubvol(struct path *parent,
 717                                    char *name, int namelen,
 718                                    struct btrfs_root *snap_src,
 719                                    u64 *async_transid, bool readonly,
 720                                    struct btrfs_qgroup_inherit *inherit)
 721 {
 722         struct inode *dir  = parent->dentry->d_inode;
 723         struct dentry *dentry;
 724         int error;
 725
 726         error = mutex_lock_killable_nested(&dir->i_mutex, I_MUTEX_PARENT);
 727         if (error == -EINTR)
 728                 return error;
 729
 730         dentry = lookup_one_len(name, parent->dentry, namelen);
 731         error = PTR_ERR(dentry);
 732         if (IS_ERR(dentry))
 733                 goto out_unlock;
 734
 735         error = -EEXIST;
 736         if (dentry->d_inode)
 737                 goto out_dput;
 738
 739         error = btrfs_may_create(dir, dentry);
 740         if (error)
 741                 goto out_dput;
 742
 743         /*
 744          * even if this name doesn't exist, we may get hash collisions.
 745          * check for them now when we can safely fail
 746          */
 747         error = btrfs_check_dir_item_collision(BTRFS_I(dir)->root,
 748                                                dir->i_ino, name,
 749                                                namelen);
 750         if (error)
 751                 goto out_dput;
 752
 753         down_read(&BTRFS_I(dir)->root->fs_info->subvol_sem);
 754
 755         if (btrfs_root_refs(&BTRFS_I(dir)->root->root_item) == 0)
 756                 goto out_up_read;
 757
 758         if (snap_src) {
 759                 error = create_snapshot(snap_src, dir, dentry, name, namelen,
 760                                         async_transid, readonly, inherit);
 761         } else {
 762                 error = create_subvol(dir, dentry, name, namelen,
 763                                       async_transid, inherit);
 764         }
 765         if (!error)
 766                 fsnotify_mkdir(dir, dentry);
 767 out_up_read:
 768         up_read(&BTRFS_I(dir)->root->fs_info->subvol_sem);
 769 out_dput:
 770         dput(dentry);
 771 out_unlock:
 772         mutex_unlock(&dir->i_mutex);
 773         return error;
 774 }
 775
 776 /*
 777  * When we're defragging a range, we don't want to kick it off again
 778  * if it is really just waiting for delalloc to send it down.
 779  * If we find a nice big extent or delalloc range for the bytes in the
 780  * file you want to defrag, we return 0 to let you know to skip this
 781  * part of the file
 782  */
 783 static int check_defrag_in_cache(struct inode *inode, u64 offset, int thresh)
 784 {
 785         struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
 786         struct extent_map *em = NULL;
 787         struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
 788         u64 end;
 789
 790         read_lock(&em_tree->lock);
 791         em = lookup_extent_mapping(em_tree, offset, PAGE_CACHE_SIZE);
 792         read_unlock(&em_tree->lock);
 793
 794         if (em) {
 795                 end = extent_map_end(em);
 796                 free_extent_map(em);
 797                 if (end - offset > thresh)
 798                         return 0;
 799         }
 800         /* if we already have a nice delalloc here, just stop */
 801         thresh /= 2;
 802         end = count_range_bits(io_tree, &offset, offset + thresh,
 803                                thresh, EXTENT_DELALLOC, 1);
 804         if (end >= thresh)
 805                 return 0;
 806         return 1;
 807 }
 808
 809 /*
 810  * helper function to walk through a file and find extents
 811  * newer than a specific transid, and smaller than thresh.
 812  *
 813  * This is used by the defragging code to find new and small
 814  * extents
 815  */
 816 static int find_new_extents(struct btrfs_root *root,
 817                             struct inode *inode, u64 newer_than,
 818                             u64 *off, int thresh)
 819 {
 820         struct btrfs_path *path;
 821         struct btrfs_key min_key;
 822         struct btrfs_key max_key;
 823         struct extent_buffer *leaf;
 824         struct btrfs_file_extent_item *extent;
 825         int type;
 826         int ret;
 827         u64 ino = btrfs_ino(inode);
 828
 829         path = btrfs_alloc_path();
 830         if (!path)
 831                 return -ENOMEM;
 832
 833         min_key.objectid = ino;
 834         min_key.type = BTRFS_EXTENT_DATA_KEY;
 835         min_key.offset = *off;
 836
 837         max_key.objectid = ino;
 838         max_key.type = (u8)-1;
 839         max_key.offset = (u64)-1;
 840
 841         path->keep_locks = 1;
 842
 843         while(1) {
 844                 ret = btrfs_search_forward(root, &min_key, &max_key,
 845                                            path, newer_than);
 846                 if (ret != 0)
 847                         goto none;
 848                 if (min_key.objectid != ino)
 849                         goto none;
 850                 if (min_key.type != BTRFS_EXTENT_DATA_KEY)
 851                         goto none;
 852
 853                 leaf = path->nodes[0];
 854                 extent = btrfs_item_ptr(leaf, path->slots[0],
 855                                         struct btrfs_file_extent_item);
 856
 857                 type = btrfs_file_extent_type(leaf, extent);
 858                 if (type == BTRFS_FILE_EXTENT_REG &&
 859                     btrfs_file_extent_num_bytes(leaf, extent) < thresh &&
 860                     check_defrag_in_cache(inode, min_key.offset, thresh)) {
 861                         *off = min_key.offset;
 862                         btrfs_free_path(path);
 863                         return 0;
 864                 }
 865
 866                 if (min_key.offset == (u64)-1)
 867                         goto none;
 868
 869                 min_key.offset++;
 870                 btrfs_release_path(path);
 871         }
 872 none:
 873         btrfs_free_path(path);
 874         return -ENOENT;
 875 }
 876
 877 static struct extent_map *defrag_lookup_extent(struct inode *inode, u64 start)
 878 {
 879         struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
 880         struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
 881         struct extent_map *em;
 882         u64 len = PAGE_CACHE_SIZE;
 883
 884         /*
 885          * hopefully we have this extent in the tree already, try without
 886          * the full extent lock
 887          */
 888         read_lock(&em_tree->lock);
 889         em = lookup_extent_mapping(em_tree, start, len);
 890         read_unlock(&em_tree->lock);
 891
 892         if (!em) {
 893                 /* get the big lock and read metadata off disk */
 894                 lock_extent(io_tree, start, start + len - 1);
 895                 em = btrfs_get_extent(inode, NULL, 0, start, len, 0);
 896                 unlock_extent(io_tree, start, start + len - 1);
 897
 898                 if (IS_ERR(em))
 899                         return NULL;
 900         }
 901
 902         return em;
 903 }
 904
 905 static bool defrag_check_next_extent(struct inode *inode, struct extent_map *em)
 906 {
 907         struct extent_map *next;
 908         bool ret = true;
 909
 910         /* this is the last extent */
 911         if (em->start + em->len >= i_size_read(inode))
 912                 return false;
 913
 914         next = defrag_lookup_extent(inode, em->start + em->len);
 915         if (!next || next->block_start >= EXTENT_MAP_LAST_BYTE)
 916                 ret = false;
 917
 918         free_extent_map(next);
 919         return ret;
 920 }
 921
 922 static int should_defrag_range(struct inode *inode, u64 start, int thresh,
 923                                u64 *last_len, u64 *skip, u64 *defrag_end,
 924                                int compress)
 925 {
 926         struct extent_map *em;
 927         int ret = 1;
 928         bool next_mergeable = true;
 929
 930         /*
 931          * make sure that once we start defragging an extent, we keep on
 932          * defragging it
 933          */
 934         if (start < *defrag_end)
 935                 return 1;
 936
 937         *skip = 0;
 938
 939         em = defrag_lookup_extent(inode, start);
 940         if (!em)
 941                 return 0;
 942
 943         /* this will cover holes, and inline extents */
 944         if (em->block_start >= EXTENT_MAP_LAST_BYTE) {
 945                 ret = 0;
 946                 goto out;
 947         }
 948
 949         next_mergeable = defrag_check_next_extent(inode, em);
 950
 951         /*
 952          * we hit a real extent, if it is big or the next extent is not a
 953          * real extent, don't bother defragging it
 954          */
 955         if (!compress && (*last_len == 0 || *last_len >= thresh) &&
 956             (em->len >= thresh || !next_mergeable))
 957                 ret = 0;
 958 out:
 959         /*
 960          * last_len ends up being a counter of how many bytes we've defragged.
 961          * every time we choose not to defrag an extent, we reset *last_len
 962          * so that the next tiny extent will force a defrag.
 963          *
 964          * The end result of this is that tiny extents before a single big
 965          * extent will force at least part of that big extent to be defragged.
 966          */
 967         if (ret) {
 968                 *defrag_end = extent_map_end(em);
 969         } else {
 970                 *last_len = 0;
 971                 *skip = extent_map_end(em);
 972                 *defrag_end = 0;
 973         }
 974
 975         free_extent_map(em);
 976         return ret;
 977 }
 978
 979 /*
 980  * it doesn't do much good to defrag one or two pages
 981  * at a time.  This pulls in a nice chunk of pages
 982  * to COW and defrag.
 983  *
 984  * It also makes sure the delalloc code has enough
 985  * dirty data to avoid making new small extents as part
 986  * of the defrag
 987  *
 988  * It's a good idea to start RA on this range
 989  * before calling this.
 990  */
 991 static int cluster_pages_for_defrag(struct inode *inode,
 992                                     struct page **pages,
 993                                     unsigned long start_index,
 994                                     int num_pages)
 995 {
 996         unsigned long file_end;
 997         u64 isize = i_size_read(inode);
 998         u64 page_start;
 999         u64 page_end;
1000         u64 page_cnt;
1001         int ret;
1002         int i;
1003         int i_done;
1004         struct btrfs_ordered_extent *ordered;
1005         struct extent_state *cached_state = NULL;
1006         struct extent_io_tree *tree;
1007         gfp_t mask = btrfs_alloc_write_mask(inode->i_mapping);
1008
1009         file_end = (isize - 1) >> PAGE_CACHE_SHIFT;
1010         if (!isize || start_index > file_end)
1011                 return 0;
1012
1013         page_cnt = min_t(u64, (u64)num_pages, (u64)file_end - start_index + 1);
1014
1015         ret = btrfs_delalloc_reserve_space(inode,
1016                                            page_cnt << PAGE_CACHE_SHIFT);
1017         if (ret)
1018                 return ret;
1019         i_done = 0;
1020         tree = &BTRFS_I(inode)->io_tree;
1021
1022         /* step one, lock all the pages */
1023         for (i = 0; i < page_cnt; i++) {
1024                 struct page *page;
1025 again:
1026                 page = find_or_create_page(inode->i_mapping,
1027                                            start_index + i, mask);
1028                 if (!page)
1029                         break;
1030
1031                 page_start = page_offset(page);
1032                 page_end = page_start + PAGE_CACHE_SIZE - 1;
1033                 while (1) {
1034                         lock_extent(tree, page_start, page_end);
1035                         ordered = btrfs_lookup_ordered_extent(inode,
1036                                                               page_start);
1037                         unlock_extent(tree, page_start, page_end);
1038                         if (!ordered)
1039                                 break;
1040
1041                         unlock_page(page);
1042                         btrfs_start_ordered_extent(inode, ordered, 1);
1043                         btrfs_put_ordered_extent(ordered);
1044                         lock_page(page);
1045                         /*
1046                          * we unlocked the page above, so we need check if
1047                          * it was released or not.
1048                          */
1049                         if (page->mapping != inode->i_mapping) {
1050                                 unlock_page(page);
1051                                 page_cache_release(page);
1052                                 goto again;
1053                         }
1054                 }
1055
1056                 if (!PageUptodate(page)) {
1057                         btrfs_readpage(NULL, page);
1058                         lock_page(page);
1059                         if (!PageUptodate(page)) {
1060                                 unlock_page(page);
1061                                 page_cache_release(page);
1062                                 ret = -EIO;
1063                                 break;
1064                         }
1065                 }
1066
1067                 if (page->mapping != inode->i_mapping) {
1068                         unlock_page(page);
1069                         page_cache_release(page);
1070                         goto again;
1071                 }
1072
1073                 pages[i] = page;
1074                 i_done++;
1075         }
1076         if (!i_done || ret)
1077                 goto out;
1078
1079         if (!(inode->i_sb->s_flags & MS_ACTIVE))
1080                 goto out;
1081
1082         /*
1083          * so now we have a nice long stream of locked
1084          * and up to date pages, lets wait on them
1085          */
1086         for (i = 0; i < i_done; i++)
1087                 wait_on_page_writeback(pages[i]);
1088
1089         page_start = page_offset(pages[0]);
1090         page_end = page_offset(pages[i_done - 1]) + PAGE_CACHE_SIZE;
1091
1092         lock_extent_bits(&BTRFS_I(inode)->io_tree,
1093                          page_start, page_end - 1, 0, &cached_state);
1094         clear_extent_bit(&BTRFS_I(inode)->io_tree, page_start,
1095                           page_end - 1, EXTENT_DIRTY | EXTENT_DELALLOC |
1096                           EXTENT_DO_ACCOUNTING | EXTENT_DEFRAG, 0, 0,
1097                           &cached_state, GFP_NOFS);
1098
1099         if (i_done != page_cnt) {
1100                 spin_lock(&BTRFS_I(inode)->lock);
1101                 BTRFS_I(inode)->outstanding_extents++;
1102                 spin_unlock(&BTRFS_I(inode)->lock);
1103                 btrfs_delalloc_release_space(inode,
1104                                      (page_cnt - i_done) << PAGE_CACHE_SHIFT);
1105         }
1106
1107
1108         set_extent_defrag(&BTRFS_I(inode)->io_tree, page_start, page_end - 1,
1109                           &cached_state, GFP_NOFS);
1110
1111         unlock_extent_cached(&BTRFS_I(inode)->io_tree,
1112                              page_start, page_end - 1, &cached_state,
1113                              GFP_NOFS);
1114
1115         for (i = 0; i < i_done; i++) {
1116                 clear_page_dirty_for_io(pages[i]);
1117                 ClearPageChecked(pages[i]);
1118                 set_page_extent_mapped(pages[i]);
1119                 set_page_dirty(pages[i]);
1120                 unlock_page(pages[i]);
1121                 page_cache_release(pages[i]);
1122         }
1123         return i_done;
1124 out:
1125         for (i = 0; i < i_done; i++) {
1126                 unlock_page(pages[i]);
1127                 page_cache_release(pages[i]);
1128         }
1129         btrfs_delalloc_release_space(inode, page_cnt << PAGE_CACHE_SHIFT);
1130         return ret;
1131
1132 }
1133
1134 int btrfs_defrag_file(struct inode *inode, struct file *file,
1135                       struct btrfs_ioctl_defrag_range_args *range,
1136                       u64 newer_than, unsigned long max_to_defrag)
1137 {
1138         struct btrfs_root *root = BTRFS_I(inode)->root;
1139         struct file_ra_state *ra = NULL;
1140         unsigned long last_index;
1141         u64 isize = i_size_read(inode);
1142         u64 last_len = 0;
1143         u64 skip = 0;
1144         u64 defrag_end = 0;
1145         u64 newer_off = range->start;
1146         unsigned long i;
1147         unsigned long ra_index = 0;
1148         int ret;
1149         int defrag_count = 0;
1150         int compress_type = BTRFS_COMPRESS_ZLIB;
1151         int extent_thresh = range->extent_thresh;
1152         int max_cluster = (256 * 1024) >> PAGE_CACHE_SHIFT;
1153         int cluster = max_cluster;
1154         u64 new_align = ~((u64)128 * 1024 - 1);
1155         struct page **pages = NULL;
1156
1157         if (isize == 0)
1158                 return 0;
1159
1160         if (range->start >= isize)
1161                 return -EINVAL;
1162
1163         if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS) {
1164                 if (range->compress_type > BTRFS_COMPRESS_TYPES)
1165                         return -EINVAL;
1166                 if (range->compress_type)
1167                         compress_type = range->compress_type;
1168         }
1169
1170         if (extent_thresh == 0)
1171                 extent_thresh = 256 * 1024;
1172
1173         /*
1174          * if we were not given a file, allocate a readahead
1175          * context
1176          */
1177         if (!file) {
1178                 ra = kzalloc(sizeof(*ra), GFP_NOFS);
1179                 if (!ra)
1180                         return -ENOMEM;
1181                 file_ra_state_init(ra, inode->i_mapping);
1182         } else {
1183                 ra = &file->f_ra;
1184         }
1185
1186         pages = kmalloc(sizeof(struct page *) * max_cluster,
1187                         GFP_NOFS);
1188         if (!pages) {
1189                 ret = -ENOMEM;
1190                 goto out_ra;
1191         }
1192
1193         /* find the last page to defrag */
1194         if (range->start + range->len > range->start) {
1195                 last_index = min_t(u64, isize - 1,
1196                          range->start + range->len - 1) >> PAGE_CACHE_SHIFT;
1197         } else {
1198                 last_index = (isize - 1) >> PAGE_CACHE_SHIFT;
1199         }
1200
1201         if (newer_than) {
1202                 ret = find_new_extents(root, inode, newer_than,
1203                                        &newer_off, 64 * 1024);
1204                 if (!ret) {
1205                         range->start = newer_off;
1206                         /*
1207                          * we always align our defrag to help keep
1208                          * the extents in the file evenly spaced
1209                          */
1210                         i = (newer_off & new_align) >> PAGE_CACHE_SHIFT;
1211                 } else
1212                         goto out_ra;
1213         } else {
1214                 i = range->start >> PAGE_CACHE_SHIFT;
1215         }
1216         if (!max_to_defrag)
1217                 max_to_defrag = last_index + 1;
1218
1219         /*
1220          * make writeback starts from i, so the defrag range can be
1221          * written sequentially.
1222          */
1223         if (i < inode->i_mapping->writeback_index)
1224                 inode->i_mapping->writeback_index = i;
1225
1226         while (i <= last_index && defrag_count < max_to_defrag &&
1227                (i < (i_size_read(inode) + PAGE_CACHE_SIZE - 1) >>
1228                 PAGE_CACHE_SHIFT)) {
1229                 /*
1230                  * make sure we stop running if someone unmounts
1231                  * the FS
1232                  */
1233                 if (!(inode->i_sb->s_flags & MS_ACTIVE))
1234                         break;
1235
1236                 if (btrfs_defrag_cancelled(root->fs_info)) {
1237                         printk(KERN_DEBUG "btrfs: defrag_file cancelled\n");
1238                         ret = -EAGAIN;
1239                         break;
1240                 }
1241
1242                 if (!should_defrag_range(inode, (u64)i << PAGE_CACHE_SHIFT,
1243                                          extent_thresh, &last_len, &skip,
1244                                          &defrag_end, range->flags &
1245                                          BTRFS_DEFRAG_RANGE_COMPRESS)) {
1246                         unsigned long next;
1247                         /*
1248                          * the should_defrag function tells us how much to skip
1249                          * bump our counter by the suggested amount
1250                          */
1251                         next = (skip + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT;
1252                         i = max(i + 1, next);
1253                         continue;
1254                 }
1255
1256                 if (!newer_than) {
1257                         cluster = (PAGE_CACHE_ALIGN(defrag_end) >>
1258                                    PAGE_CACHE_SHIFT) - i;
1259                         cluster = min(cluster, max_cluster);
1260                 } else {
1261                         cluster = max_cluster;
1262                 }
1263
1264                 if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)
1265                         BTRFS_I(inode)->force_compress = compress_type;
1266
1267                 if (i + cluster > ra_index) {
1268                         ra_index = max(i, ra_index);
1269                         btrfs_force_ra(inode->i_mapping, ra, file, ra_index,
1270                                        cluster);
1271                         ra_index += max_cluster;
1272                 }
1273
1274                 mutex_lock(&inode->i_mutex);
1275                 ret = cluster_pages_for_defrag(inode, pages, i, cluster);
1276                 if (ret < 0) {
1277                         mutex_unlock(&inode->i_mutex);
1278                         goto out_ra;
1279                 }
1280
1281                 defrag_count += ret;
1282                 balance_dirty_pages_ratelimited(inode->i_mapping);
1283                 mutex_unlock(&inode->i_mutex);
1284
1285                 if (newer_than) {
1286                         if (newer_off == (u64)-1)
1287                                 break;
1288
1289                         if (ret > 0)
1290                                 i += ret;
1291
1292                         newer_off = max(newer_off + 1,
1293                                         (u64)i << PAGE_CACHE_SHIFT);
1294
1295                         ret = find_new_extents(root, inode,
1296                                                newer_than, &newer_off,
1297                                                64 * 1024);
1298                         if (!ret) {
1299                                 range->start = newer_off;
1300                                 i = (newer_off & new_align) >> PAGE_CACHE_SHIFT;
1301                         } else {
1302                                 break;
1303                         }
1304                 } else {
1305                         if (ret > 0) {
1306                                 i += ret;
1307                                 last_len += ret << PAGE_CACHE_SHIFT;
1308                         } else {
1309                                 i++;
1310                                 last_len = 0;
1311                         }
1312                 }
1313         }
1314
1315         if ((range->flags & BTRFS_DEFRAG_RANGE_START_IO))
1316                 filemap_flush(inode->i_mapping);
1317
1318         if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
1319                 /* the filemap_flush will queue IO into the worker threads, but
1320                  * we have to make sure the IO is actually started and that
1321                  * ordered extents get created before we return
1322                  */
1323                 atomic_inc(&root->fs_info->async_submit_draining);
1324                 while (atomic_read(&root->fs_info->nr_async_submits) ||
1325                       atomic_read(&root->fs_info->async_delalloc_pages)) {
1326                         wait_event(root->fs_info->async_submit_wait,
1327                            (atomic_read(&root->fs_info->nr_async_submits) == 0 &&
1328                             atomic_read(&root->fs_info->async_delalloc_pages) == 0));
1329                 }
1330                 atomic_dec(&root->fs_info->async_submit_draining);
1331
1332                 mutex_lock(&inode->i_mutex);
1333                 BTRFS_I(inode)->force_compress = BTRFS_COMPRESS_NONE;
1334                 mutex_unlock(&inode->i_mutex);
1335         }
1336
1337         if (range->compress_type == BTRFS_COMPRESS_LZO) {
1338                 btrfs_set_fs_incompat(root->fs_info, COMPRESS_LZO);
1339         }
1340
1341         ret = defrag_count;
1342
1343 out_ra:
1344         if (!file)
1345                 kfree(ra);
1346         kfree(pages);
1347         return ret;
1348 }
1349
1350 static noinline int btrfs_ioctl_resize(struct file *file,
1351                                         void __user *arg)
1352 {
1353         u64 new_size;
1354         u64 old_size;
1355         u64 devid = 1;
1356         struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
1357         struct btrfs_ioctl_vol_args *vol_args;
1358         struct btrfs_trans_handle *trans;
1359         struct btrfs_device *device = NULL;
1360         char *sizestr;
1361         char *devstr = NULL;
1362         int ret = 0;
1363         int mod = 0;
1364
1365         if (!capable(CAP_SYS_ADMIN))
1366                 return -EPERM;
1367
1368         ret = mnt_want_write_file(file);
1369         if (ret)
1370                 return ret;
1371
1372         if (atomic_xchg(&root->fs_info->mutually_exclusive_operation_running,
1373                         1)) {
1374                 pr_info("btrfs: dev add/delete/balance/replace/resize operation in progress\n");
1375                 mnt_drop_write_file(file);
1376                 return -EINVAL;
1377         }
1378
1379         mutex_lock(&root->fs_info->volume_mutex);
1380         vol_args = memdup_user(arg, sizeof(*vol_args));
1381         if (IS_ERR(vol_args)) {
1382                 ret = PTR_ERR(vol_args);
1383                 goto out;
1384         }
1385
1386         vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1387
1388         sizestr = vol_args->name;
1389         devstr = strchr(sizestr, ':');
1390         if (devstr) {
1391                 char *end;
1392                 sizestr = devstr + 1;
1393                 *devstr = '\0';
1394                 devstr = vol_args->name;
1395                 devid = simple_strtoull(devstr, &end, 10);
1396                 if (!devid) {
1397                         ret = -EINVAL;
1398                         goto out_free;
1399                 }
1400                 printk(KERN_INFO "btrfs: resizing devid %llu\n",
1401                        (unsigned long long)devid);
1402         }
1403
1404         device = btrfs_find_device(root->fs_info, devid, NULL, NULL);
1405         if (!device) {
1406                 printk(KERN_INFO "btrfs: resizer unable to find device %llu\n",
1407                        (unsigned long long)devid);
1408                 ret = -ENODEV;
1409                 goto out_free;
1410         }
1411
1412         if (!device->writeable) {
1413                 printk(KERN_INFO "btrfs: resizer unable to apply on "
1414                        "readonly device %llu\n",
1415                        (unsigned long long)devid);
1416                 ret = -EPERM;
1417                 goto out_free;
1418         }
1419
1420         if (!strcmp(sizestr, "max"))
1421                 new_size = device->bdev->bd_inode->i_size;
1422         else {
1423                 if (sizestr[0] == '-') {
1424                         mod = -1;
1425                         sizestr++;
1426                 } else if (sizestr[0] == '+') {
1427                         mod = 1;
1428                         sizestr++;
1429                 }
1430                 new_size = memparse(sizestr, NULL);
1431                 if (new_size == 0) {
1432                         ret = -EINVAL;
1433                         goto out_free;
1434                 }
1435         }
1436
1437         if (device->is_tgtdev_for_dev_replace) {
1438                 ret = -EPERM;
1439                 goto out_free;
1440         }
1441
1442         old_size = device->total_bytes;
1443
1444         if (mod < 0) {
1445                 if (new_size > old_size) {
1446                         ret = -EINVAL;
1447                         goto out_free;
1448                 }
1449                 new_size = old_size - new_size;
1450         } else if (mod > 0) {
1451                 new_size = old_size + new_size;
1452         }
1453
1454         if (new_size < 256 * 1024 * 1024) {
1455                 ret = -EINVAL;
1456                 goto out_free;
1457         }
1458         if (new_size > device->bdev->bd_inode->i_size) {
1459                 ret = -EFBIG;
1460                 goto out_free;
1461         }
1462
1463         do_div(new_size, root->sectorsize);
1464         new_size *= root->sectorsize;
1465
1466         printk_in_rcu(KERN_INFO "btrfs: new size for %s is %llu\n",
1467                       rcu_str_deref(device->name),
1468                       (unsigned long long)new_size);
1469
1470         if (new_size > old_size) {
1471                 trans = btrfs_start_transaction(root, 0);
1472                 if (IS_ERR(trans)) {
1473                         ret = PTR_ERR(trans);
1474                         goto out_free;
1475                 }
1476                 ret = btrfs_grow_device(trans, device, new_size);
1477                 btrfs_commit_transaction(trans, root);
1478         } else if (new_size < old_size) {
1479                 ret = btrfs_shrink_device(device, new_size);
1480         } /* equal, nothing need to do */
1481
1482 out_free:
1483         kfree(vol_args);
1484 out:
1485         mutex_unlock(&root->fs_info->volume_mutex);
1486         atomic_set(&root->fs_info->mutually_exclusive_operation_running, 0);
1487         mnt_drop_write_file(file);
1488         return ret;
1489 }
1490
1491 static noinline int btrfs_ioctl_snap_create_transid(struct file *file,
1492                                 char *name, unsigned long fd, int subvol,
1493                                 u64 *transid, bool readonly,
1494                                 struct btrfs_qgroup_inherit *inherit)
1495 {
1496         int namelen;
1497         int ret = 0;
1498
1499         ret = mnt_want_write_file(file);
1500         if (ret)
1501                 goto out;
1502
1503         namelen = strlen(name);
1504         if (strchr(name, '/')) {
1505                 ret = -EINVAL;
1506                 goto out_drop_write;
1507         }
1508
1509         if (name[0] == '.' &&
1510            (namelen == 1 || (name[1] == '.' && namelen == 2))) {
1511                 ret = -EEXIST;
1512                 goto out_drop_write;
1513         }
1514
1515         if (subvol) {
1516                 ret = btrfs_mksubvol(&file->f_path, name, namelen,
1517                                      NULL, transid, readonly, inherit);
1518         } else {
1519                 struct fd src = fdget(fd);
1520                 struct inode *src_inode;
1521                 if (!src.file) {
1522                         ret = -EINVAL;
1523                         goto out_drop_write;
1524                 }
1525
1526                 src_inode = file_inode(src.file);
1527                 if (src_inode->i_sb != file_inode(file)->i_sb) {
1528                         printk(KERN_INFO "btrfs: Snapshot src from "
1529                                "another FS\n");
1530                         ret = -EINVAL;
1531                 } else if (!inode_owner_or_capable(src_inode)) {
1532                         /*
1533                          * Subvolume creation is not restricted, but snapshots
1534                          * are limited to own subvolumes only
1535                          */
1536                         ret = -EPERM;
1537                 } else {
1538                         ret = btrfs_mksubvol(&file->f_path, name, namelen,
1539                                              BTRFS_I(src_inode)->root,
1540                                              transid, readonly, inherit);
1541                 }
1542                 fdput(src);
1543         }
1544 out_drop_write:
1545         mnt_drop_write_file(file);
1546 out:
1547         return ret;
1548 }
1549
1550 static noinline int btrfs_ioctl_snap_create(struct file *file,
1551                                             void __user *arg, int subvol)
1552 {
1553         struct btrfs_ioctl_vol_args *vol_args;
1554         int ret;
1555
1556         vol_args = memdup_user(arg, sizeof(*vol_args));
1557         if (IS_ERR(vol_args))
1558                 return PTR_ERR(vol_args);
1559         vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1560
1561         ret = btrfs_ioctl_snap_create_transid(file, vol_args->name,
1562                                               vol_args->fd, subvol,
1563                                               NULL, false, NULL);
1564
1565         kfree(vol_args);
1566         return ret;
1567 }
1568
1569 static noinline int btrfs_ioctl_snap_create_v2(struct file *file,
1570                                                void __user *arg, int subvol)
1571 {
1572         struct btrfs_ioctl_vol_args_v2 *vol_args;
1573         int ret;
1574         u64 transid = 0;
1575         u64 *ptr = NULL;
1576         bool readonly = false;
1577         struct btrfs_qgroup_inherit *inherit = NULL;
1578
1579         vol_args = memdup_user(arg, sizeof(*vol_args));
1580         if (IS_ERR(vol_args))
1581                 return PTR_ERR(vol_args);
1582         vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0';
1583
1584         if (vol_args->flags &
1585             ~(BTRFS_SUBVOL_CREATE_ASYNC | BTRFS_SUBVOL_RDONLY |
1586               BTRFS_SUBVOL_QGROUP_INHERIT)) {
1587                 ret = -EOPNOTSUPP;
1588                 goto out;
1589         }
1590
1591         if (vol_args->flags & BTRFS_SUBVOL_CREATE_ASYNC)
1592                 ptr = &transid;
1593         if (vol_args->flags & BTRFS_SUBVOL_RDONLY)
1594                 readonly = true;
1595         if (vol_args->flags & BTRFS_SUBVOL_QGROUP_INHERIT) {
1596                 if (vol_args->size > PAGE_CACHE_SIZE) {
1597                         ret = -EINVAL;
1598                         goto out;
1599                 }
1600                 inherit = memdup_user(vol_args->qgroup_inherit, vol_args->size);
1601                 if (IS_ERR(inherit)) {
1602                         ret = PTR_ERR(inherit);
1603                         goto out;
1604                 }
1605         }
1606
1607         ret = btrfs_ioctl_snap_create_transid(file, vol_args->name,
1608                                               vol_args->fd, subvol, ptr,
1609                                               readonly, inherit);
1610
1611         if (ret == 0 && ptr &&
1612             copy_to_user(arg +
1613                          offsetof(struct btrfs_ioctl_vol_args_v2,
1614                                   transid), ptr, sizeof(*ptr)))
1615                 ret = -EFAULT;
1616 out:
1617         kfree(vol_args);
1618         kfree(inherit);
1619         return ret;
1620 }
1621
1622 static noinline int btrfs_ioctl_subvol_getflags(struct file *file,
1623                                                 void __user *arg)
1624 {
1625         struct inode *inode = file_inode(file);
1626         struct btrfs_root *root = BTRFS_I(inode)->root;
1627         int ret = 0;
1628         u64 flags = 0;
1629
1630         if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID)
1631                 return -EINVAL;
1632
1633         down_read(&root->fs_info->subvol_sem);
1634         if (btrfs_root_readonly(root))
1635                 flags |= BTRFS_SUBVOL_RDONLY;
1636         up_read(&root->fs_info->subvol_sem);
1637
1638         if (copy_to_user(arg, &flags, sizeof(flags)))
1639                 ret = -EFAULT;
1640
1641         return ret;
1642 }
1643
1644 static noinline int btrfs_ioctl_subvol_setflags(struct file *file,
1645                                               void __user *arg)
1646 {
1647         struct inode *inode = file_inode(file);
1648         struct btrfs_root *root = BTRFS_I(inode)->root;
1649         struct btrfs_trans_handle *trans;
1650         u64 root_flags;
1651         u64 flags;
1652         int ret = 0;
1653
1654         ret = mnt_want_write_file(file);
1655         if (ret)
1656                 goto out;
1657
1658         if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) {
1659                 ret = -EINVAL;
1660                 goto out_drop_write;
1661         }
1662
1663         if (copy_from_user(&flags, arg, sizeof(flags))) {
1664                 ret = -EFAULT;
1665                 goto out_drop_write;
1666         }
1667
1668         if (flags & BTRFS_SUBVOL_CREATE_ASYNC) {
1669                 ret = -EINVAL;
1670                 goto out_drop_write;
1671         }
1672
1673         if (flags & ~BTRFS_SUBVOL_RDONLY) {
1674                 ret = -EOPNOTSUPP;
1675                 goto out_drop_write;
1676         }
1677
1678         if (!inode_owner_or_capable(inode)) {
1679                 ret = -EACCES;
1680                 goto out_drop_write;
1681         }
1682
1683         down_write(&root->fs_info->subvol_sem);
1684
1685         /* nothing to do */
1686         if (!!(flags & BTRFS_SUBVOL_RDONLY) == btrfs_root_readonly(root))
1687                 goto out_drop_sem;
1688
1689         root_flags = btrfs_root_flags(&root->root_item);
1690         if (flags & BTRFS_SUBVOL_RDONLY)
1691                 btrfs_set_root_flags(&root->root_item,
1692                                      root_flags | BTRFS_ROOT_SUBVOL_RDONLY);
1693         else
1694                 btrfs_set_root_flags(&root->root_item,
1695                                      root_flags & ~BTRFS_ROOT_SUBVOL_RDONLY);
1696
1697         trans = btrfs_start_transaction(root, 1);
1698         if (IS_ERR(trans)) {
1699                 ret = PTR_ERR(trans);
1700                 goto out_reset;
1701         }
1702
1703         ret = btrfs_update_root(trans, root->fs_info->tree_root,
1704                                 &root->root_key, &root->root_item);
1705
1706         btrfs_commit_transaction(trans, root);
1707 out_reset:
1708         if (ret)
1709                 btrfs_set_root_flags(&root->root_item, root_flags);
1710 out_drop_sem:
1711         up_write(&root->fs_info->subvol_sem);
1712 out_drop_write:
1713         mnt_drop_write_file(file);
1714 out:
1715         return ret;
1716 }
1717
1718 /*
1719  * helper to check if the subvolume references other subvolumes
1720  */
1721 static noinline int may_destroy_subvol(struct btrfs_root *root)
1722 {
1723         struct btrfs_path *path;
1724         struct btrfs_key key;
1725         int ret;
1726
1727         path = btrfs_alloc_path();
1728         if (!path)
1729                 return -ENOMEM;
1730
1731         key.objectid = root->root_key.objectid;
1732         key.type = BTRFS_ROOT_REF_KEY;
1733         key.offset = (u64)-1;
1734
1735         ret = btrfs_search_slot(NULL, root->fs_info->tree_root,
1736                                 &key, path, 0, 0);
1737         if (ret < 0)
1738                 goto out;
1739         BUG_ON(ret == 0);
1740
1741         ret = 0;
1742         if (path->slots[0] > 0) {
1743                 path->slots[0]--;
1744                 btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
1745                 if (key.objectid == root->root_key.objectid &&
1746                     key.type == BTRFS_ROOT_REF_KEY)
1747                         ret = -ENOTEMPTY;
1748         }
1749 out:
1750         btrfs_free_path(path);
1751         return ret;
1752 }
1753
1754 static noinline int key_in_sk(struct btrfs_key *key,
1755                               struct btrfs_ioctl_search_key *sk)
1756 {
1757         struct btrfs_key test;
1758         int ret;
1759
1760         test.objectid = sk->min_objectid;
1761         test.type = sk->min_type;
1762         test.offset = sk->min_offset;
1763
1764         ret = btrfs_comp_cpu_keys(key, &test);
1765         if (ret < 0)
1766                 return 0;
1767
1768         test.objectid = sk->max_objectid;
1769         test.type = sk->max_type;
1770         test.offset = sk->max_offset;
1771
1772         ret = btrfs_comp_cpu_keys(key, &test);
1773         if (ret > 0)
1774                 return 0;
1775         return 1;
1776 }
1777
1778 static noinline int copy_to_sk(struct btrfs_root *root,
1779                                struct btrfs_path *path,
1780                                struct btrfs_key *key,
1781                                struct btrfs_ioctl_search_key *sk,
1782                                char *buf,
1783                                unsigned long *sk_offset,
1784                                int *num_found)
1785 {
1786         u64 found_transid;
1787         struct extent_buffer *leaf;
1788         struct btrfs_ioctl_search_header sh;
1789         unsigned long item_off;
1790         unsigned long item_len;
1791         int nritems;
1792         int i;
1793         int slot;
1794         int ret = 0;
1795
1796         leaf = path->nodes[0];
1797         slot = path->slots[0];
1798         nritems = btrfs_header_nritems(leaf);
1799
1800         if (btrfs_header_generation(leaf) > sk->max_transid) {
1801                 i = nritems;
1802                 goto advance_key;
1803         }
1804         found_transid = btrfs_header_generation(leaf);
1805
1806         for (i = slot; i < nritems; i++) {
1807                 item_off = btrfs_item_ptr_offset(leaf, i);
1808                 item_len = btrfs_item_size_nr(leaf, i);
1809
1810                 btrfs_item_key_to_cpu(leaf, key, i);
1811                 if (!key_in_sk(key, sk))
1812                         continue;
1813
1814                 if (sizeof(sh) + item_len > BTRFS_SEARCH_ARGS_BUFSIZE)
1815                         item_len = 0;
1816
1817                 if (sizeof(sh) + item_len + *sk_offset >
1818                     BTRFS_SEARCH_ARGS_BUFSIZE) {
1819                         ret = 1;
1820                         goto overflow;
1821                 }
1822
1823                 sh.objectid = key->objectid;
1824                 sh.offset = key->offset;
1825                 sh.type = key->type;
1826                 sh.len = item_len;
1827                 sh.transid = found_transid;
1828
1829                 /* copy search result header */
1830                 memcpy(buf + *sk_offset, &sh, sizeof(sh));
1831                 *sk_offset += sizeof(sh);
1832
1833                 if (item_len) {
1834                         char *p = buf + *sk_offset;
1835                         /* copy the item */
1836                         read_extent_buffer(leaf, p,
1837                                            item_off, item_len);
1838                         *sk_offset += item_len;
1839                 }
1840                 (*num_found)++;
1841
1842                 if (*num_found >= sk->nr_items)
1843                         break;
1844         }
1845 advance_key:
1846         ret = 0;
1847         if (key->offset < (u64)-1 && key->offset < sk->max_offset)
1848                 key->offset++;
1849         else if (key->type < (u8)-1 && key->type < sk->max_type) {
1850                 key->offset = 0;
1851                 key->type++;
1852         } else if (key->objectid < (u64)-1 && key->objectid < sk->max_objectid) {
1853                 key->offset = 0;
1854                 key->type = 0;
1855                 key->objectid++;
1856         } else
1857                 ret = 1;
1858 overflow:
1859         return ret;
1860 }
1861
1862 static noinline int search_ioctl(struct inode *inode,
1863                                  struct btrfs_ioctl_search_args *args)
1864 {
1865         struct btrfs_root *root;
1866         struct btrfs_key key;
1867         struct btrfs_key max_key;
1868         struct btrfs_path *path;
1869         struct btrfs_ioctl_search_key *sk = &args->key;
1870         struct btrfs_fs_info *info = BTRFS_I(inode)->root->fs_info;
1871         int ret;
1872         int num_found = 0;
1873         unsigned long sk_offset = 0;
1874
1875         path = btrfs_alloc_path();
1876         if (!path)
1877                 return -ENOMEM;
1878
1879         if (sk->tree_id == 0) {
1880                 /* search the root of the inode that was passed */
1881                 root = BTRFS_I(inode)->root;
1882         } else {
1883                 key.objectid = sk->tree_id;
1884                 key.type = BTRFS_ROOT_ITEM_KEY;
1885                 key.offset = (u64)-1;
1886                 root = btrfs_read_fs_root_no_name(info, &key);
1887                 if (IS_ERR(root)) {
1888                         printk(KERN_ERR "could not find root %llu\n",
1889                                sk->tree_id);
1890                         btrfs_free_path(path);
1891                         return -ENOENT;
1892                 }
1893         }
1894
1895         key.objectid = sk->min_objectid;
1896         key.type = sk->min_type;
1897         key.offset = sk->min_offset;
1898
1899         max_key.objectid = sk->max_objectid;
1900         max_key.type = sk->max_type;
1901         max_key.offset = sk->max_offset;
1902
1903         path->keep_locks = 1;
1904
1905         while(1) {
1906                 ret = btrfs_search_forward(root, &key, &max_key, path,
1907                                            sk->min_transid);
1908                 if (ret != 0) {
1909                         if (ret > 0)
1910                                 ret = 0;
1911                         goto err;
1912                 }
1913                 ret = copy_to_sk(root, path, &key, sk, args->buf,
1914                                  &sk_offset, &num_found);
1915                 btrfs_release_path(path);
1916                 if (ret || num_found >= sk->nr_items)
1917                         break;
1918
1919         }
1920         ret = 0;
1921 err:
1922         sk->nr_items = num_found;
1923         btrfs_free_path(path);
1924         return ret;
1925 }
1926
1927 static noinline int btrfs_ioctl_tree_search(struct file *file,
1928                                            void __user *argp)
1929 {
1930          struct btrfs_ioctl_search_args *args;
1931          struct inode *inode;
1932          int ret;
1933
1934         if (!capable(CAP_SYS_ADMIN))
1935                 return -EPERM;
1936
1937         args = memdup_user(argp, sizeof(*args));
1938         if (IS_ERR(args))
1939                 return PTR_ERR(args);
1940
1941         inode = file_inode(file);
1942         ret = search_ioctl(inode, args);
1943         if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
1944                 ret = -EFAULT;
1945         kfree(args);
1946         return ret;
1947 }
1948
1949 /*
1950  * Search INODE_REFs to identify path name of 'dirid' directory
1951  * in a 'tree_id' tree. and sets path name to 'name'.
1952  */
1953 static noinline int btrfs_search_path_in_tree(struct btrfs_fs_info *info,
1954                                 u64 tree_id, u64 dirid, char *name)
1955 {
1956         struct btrfs_root *root;
1957         struct btrfs_key key;
1958         char *ptr;
1959         int ret = -1;
1960         int slot;
1961         int len;
1962         int total_len = 0;
1963         struct btrfs_inode_ref *iref;
1964         struct extent_buffer *l;
1965         struct btrfs_path *path;
1966
1967         if (dirid == BTRFS_FIRST_FREE_OBJECTID) {
1968                 name[0]='\0';
1969                 return 0;
1970         }
1971
1972         path = btrfs_alloc_path();
1973         if (!path)
1974                 return -ENOMEM;
1975
1976         ptr = &name[BTRFS_INO_LOOKUP_PATH_MAX];
1977
1978         key.objectid = tree_id;
1979         key.type = BTRFS_ROOT_ITEM_KEY;
1980         key.offset = (u64)-1;
1981         root = btrfs_read_fs_root_no_name(info, &key);
1982         if (IS_ERR(root)) {
1983                 printk(KERN_ERR "could not find root %llu\n", tree_id);
1984                 ret = -ENOENT;
1985                 goto out;
1986         }
1987
1988         key.objectid = dirid;
1989         key.type = BTRFS_INODE_REF_KEY;
1990         key.offset = (u64)-1;
1991
1992         while(1) {
1993                 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
1994                 if (ret < 0)
1995                         goto out;
1996
1997                 l = path->nodes[0];
1998                 slot = path->slots[0];
1999                 if (ret > 0 && slot > 0)
2000                         slot--;
2001                 btrfs_item_key_to_cpu(l, &key, slot);
2002
2003                 if (ret > 0 && (key.objectid != dirid ||
2004                                 key.type != BTRFS_INODE_REF_KEY)) {
2005                         ret = -ENOENT;
2006                         goto out;
2007                 }
2008
2009                 iref = btrfs_item_ptr(l, slot, struct btrfs_inode_ref);
2010                 len = btrfs_inode_ref_name_len(l, iref);
2011                 ptr -= len + 1;
2012                 total_len += len + 1;
2013                 if (ptr < name)
2014                         goto out;
2015
2016                 *(ptr + len) = '/';
2017                 read_extent_buffer(l, ptr,(unsigned long)(iref + 1), len);
2018
2019                 if (key.offset == BTRFS_FIRST_FREE_OBJECTID)
2020                         break;
2021
2022                 btrfs_release_path(path);
2023                 key.objectid = key.offset;
2024                 key.offset = (u64)-1;
2025                 dirid = key.objectid;
2026         }
2027         if (ptr < name)
2028                 goto out;
2029         memmove(name, ptr, total_len);
2030         name[total_len]='\0';
2031         ret = 0;
2032 out:
2033         btrfs_free_path(path);
2034         return ret;
2035 }
2036
2037 static noinline int btrfs_ioctl_ino_lookup(struct file *file,
2038                                            void __user *argp)
2039 {
2040          struct btrfs_ioctl_ino_lookup_args *args;
2041          struct inode *inode;
2042          int ret;
2043
2044         if (!capable(CAP_SYS_ADMIN))
2045                 return -EPERM;
2046
2047         args = memdup_user(argp, sizeof(*args));
2048         if (IS_ERR(args))
2049                 return PTR_ERR(args);
2050
2051         inode = file_inode(file);
2052
2053         if (args->treeid == 0)
2054                 args->treeid = BTRFS_I(inode)->root->root_key.objectid;
2055
2056         ret = btrfs_search_path_in_tree(BTRFS_I(inode)->root->fs_info,
2057                                         args->treeid, args->objectid,
2058                                         args->name);
2059
2060         if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
2061                 ret = -EFAULT;
2062
2063         kfree(args);
2064         return ret;
2065 }
2066
2067 static noinline int btrfs_ioctl_snap_destroy(struct file *file,
2068                                              void __user *arg)
2069 {
2070         struct dentry *parent = fdentry(file);
2071         struct dentry *dentry;
2072         struct inode *dir = parent->d_inode;
2073         struct inode *inode;
2074         struct btrfs_root *root = BTRFS_I(dir)->root;
2075         struct btrfs_root *dest = NULL;
2076         struct btrfs_ioctl_vol_args *vol_args;
2077         struct btrfs_trans_handle *trans;
2078         struct btrfs_block_rsv block_rsv;
2079         u64 qgroup_reserved;
2080         int namelen;
2081         int ret;
2082         int err = 0;
2083
2084         vol_args = memdup_user(arg, sizeof(*vol_args));
2085         if (IS_ERR(vol_args))
2086                 return PTR_ERR(vol_args);
2087
2088         vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2089         namelen = strlen(vol_args->name);
2090         if (strchr(vol_args->name, '/') ||
2091             strncmp(vol_args->name, "..", namelen) == 0) {
2092                 err = -EINVAL;
2093                 goto out;
2094         }
2095
2096         err = mnt_want_write_file(file);
2097         if (err)
2098                 goto out;
2099
2100         err = mutex_lock_killable_nested(&dir->i_mutex, I_MUTEX_PARENT);
2101         if (err == -EINTR)
2102                 goto out_drop_write;
2103         dentry = lookup_one_len(vol_args->name, parent, namelen);
2104         if (IS_ERR(dentry)) {
2105                 err = PTR_ERR(dentry);
2106                 goto out_unlock_dir;
2107         }
2108
2109         if (!dentry->d_inode) {
2110                 err = -ENOENT;
2111                 goto out_dput;
2112         }
2113
2114         inode = dentry->d_inode;
2115         dest = BTRFS_I(inode)->root;
2116         if (!capable(CAP_SYS_ADMIN)){
2117                 /*
2118                  * Regular user.  Only allow this with a special mount
2119                  * option, when the user has write+exec access to the
2120                  * subvol root, and when rmdir(2) would have been
2121                  * allowed.
2122                  *
2123                  * Note that this is _not_ check that the subvol is
2124                  * empty or doesn't contain data that we wouldn't
2125                  * otherwise be able to delete.
2126                  *
2127                  * Users who want to delete empty subvols should try
2128                  * rmdir(2).
2129                  */
2130                 err = -EPERM;
2131                 if (!btrfs_test_opt(root, USER_SUBVOL_RM_ALLOWED))
2132                         goto out_dput;
2133
2134                 /*
2135                  * Do not allow deletion if the parent dir is the same
2136                  * as the dir to be deleted.  That means the ioctl
2137                  * must be called on the dentry referencing the root
2138                  * of the subvol, not a random directory contained
2139                  * within it.
2140                  */
2141                 err = -EINVAL;
2142                 if (root == dest)
2143                         goto out_dput;
2144
2145                 err = inode_permission(inode, MAY_WRITE | MAY_EXEC);
2146                 if (err)
2147                         goto out_dput;
2148         }
2149
2150         /* check if subvolume may be deleted by a user */
2151         err = btrfs_may_delete(dir, dentry, 1);
2152         if (err)
2153                 goto out_dput;
2154
2155         if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) {
2156                 err = -EINVAL;
2157                 goto out_dput;
2158         }
2159
2160         mutex_lock(&inode->i_mutex);
2161         err = d_invalidate(dentry);
2162         if (err)
2163                 goto out_unlock;
2164
2165         down_write(&root->fs_info->subvol_sem);
2166
2167         err = may_destroy_subvol(dest);
2168         if (err)
2169                 goto out_up_write;
2170
2171         btrfs_init_block_rsv(&block_rsv, BTRFS_BLOCK_RSV_TEMP);
2172         /*
2173          * One for dir inode, two for dir entries, two for root
2174          * ref/backref.
2175          */
2176         err = btrfs_subvolume_reserve_metadata(root, &block_rsv,
2177                                                5, &qgroup_reserved);
2178         if (err)
2179                 goto out_up_write;
2180
2181         trans = btrfs_start_transaction(root, 0);
2182         if (IS_ERR(trans)) {
2183                 err = PTR_ERR(trans);
2184                 goto out_release;
2185         }
2186         trans->block_rsv = &block_rsv;
2187         trans->bytes_reserved = block_rsv.size;
2188
2189         ret = btrfs_unlink_subvol(trans, root, dir,
2190                                 dest->root_key.objectid,
2191                                 dentry->d_name.name,
2192                                 dentry->d_name.len);
2193         if (ret) {
2194                 err = ret;
2195                 btrfs_abort_transaction(trans, root, ret);
2196                 goto out_end_trans;
2197         }
2198
2199         btrfs_record_root_in_trans(trans, dest);
2200
2201         memset(&dest->root_item.drop_progress, 0,
2202                 sizeof(dest->root_item.drop_progress));
2203         dest->root_item.drop_level = 0;
2204         btrfs_set_root_refs(&dest->root_item, 0);
2205
2206         if (!xchg(&dest->orphan_item_inserted, 1)) {
2207                 ret = btrfs_insert_orphan_item(trans,
2208                                         root->fs_info->tree_root,
2209                                         dest->root_key.objectid);
2210                 if (ret) {
2211                         btrfs_abort_transaction(trans, root, ret);
2212                         err = ret;
2213                         goto out_end_trans;
2214                 }
2215         }
2216 out_end_trans:
2217         trans->block_rsv = NULL;
2218         trans->bytes_reserved = 0;
2219         ret = btrfs_end_transaction(trans, root);
2220         if (ret && !err)
2221                 err = ret;
2222         inode->i_flags |= S_DEAD;
2223 out_release:
2224         btrfs_subvolume_release_metadata(root, &block_rsv, qgroup_reserved);
2225 out_up_write:
2226         up_write(&root->fs_info->subvol_sem);
2227 out_unlock:
2228         mutex_unlock(&inode->i_mutex);
2229         if (!err) {
2230                 shrink_dcache_sb(root->fs_info->sb);
2231                 btrfs_invalidate_inodes(dest);
2232                 d_delete(dentry);
2233
2234                 /* the last ref */
2235                 if (dest->cache_inode) {
2236                         iput(dest->cache_inode);
2237                         dest->cache_inode = NULL;
2238                 }
2239         }
2240 out_dput:
2241         dput(dentry);
2242 out_unlock_dir:
2243         mutex_unlock(&dir->i_mutex);
2244 out_drop_write:
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         if (atomic_xchg(&root->fs_info->mutually_exclusive_operation_running,
2365                         1)) {
2366                 pr_info("btrfs: dev add/delete/balance/replace/resize operation in progress\n");
2367                 mnt_drop_write_file(file);
2368                 return -EINVAL;
2369         }
2370
2371         mutex_lock(&root->fs_info->volume_mutex);
2372         vol_args = memdup_user(arg, sizeof(*vol_args));
2373         if (IS_ERR(vol_args)) {
2374                 ret = PTR_ERR(vol_args);
2375                 goto out;
2376         }
2377
2378         vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2379         ret = btrfs_rm_device(root, vol_args->name);
2380
2381         kfree(vol_args);
2382 out:
2383         mutex_unlock(&root->fs_info->volume_mutex);
2384         atomic_set(&root->fs_info->mutually_exclusive_operation_running, 0);
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
2491         /*
2492          * TODO:
2493          * - split compressed inline extents.  annoying: we need to
2494          *   decompress into destination's address_space (the file offset
2495          *   may change, so source mapping won't do), then recompress (or
2496          *   otherwise reinsert) a subrange.
2497          * - allow ranges within the same file to be cloned (provided
2498          *   they don't overlap)?
2499          */
2500
2501         /* the destination must be opened for writing */
2502         if (!(file->f_mode & FMODE_WRITE) || (file->f_flags & O_APPEND))
2503                 return -EINVAL;
2504
2505         if (btrfs_root_readonly(root))
2506                 return -EROFS;
2507
2508         ret = mnt_want_write_file(file);
2509         if (ret)
2510                 return ret;
2511
2512         src_file = fdget(srcfd);
2513         if (!src_file.file) {
2514                 ret = -EBADF;
2515                 goto out_drop_write;
2516         }
2517
2518         ret = -EXDEV;
2519         if (src_file.file->f_path.mnt != file->f_path.mnt)
2520                 goto out_fput;
2521
2522         src = file_inode(src_file.file);
2523
2524         ret = -EINVAL;
2525         if (src == inode)
2526                 goto out_fput;
2527
2528         /* the src must be open for reading */
2529         if (!(src_file.file->f_mode & FMODE_READ))
2530                 goto out_fput;
2531
2532         /* don't make the dst file partly checksummed */
2533         if ((BTRFS_I(src)->flags & BTRFS_INODE_NODATASUM) !=
2534             (BTRFS_I(inode)->flags & BTRFS_INODE_NODATASUM))
2535                 goto out_fput;
2536
2537         ret = -EISDIR;
2538         if (S_ISDIR(src->i_mode) || S_ISDIR(inode->i_mode))
2539                 goto out_fput;
2540
2541         ret = -EXDEV;
2542         if (src->i_sb != inode->i_sb)
2543                 goto out_fput;
2544
2545         ret = -ENOMEM;
2546         buf = vmalloc(btrfs_level_size(root, 0));
2547         if (!buf)
2548                 goto out_fput;
2549
2550         path = btrfs_alloc_path();
2551         if (!path) {
2552                 vfree(buf);
2553                 goto out_fput;
2554         }
2555         path->reada = 2;
2556
2557         if (inode < src) {
2558                 mutex_lock_nested(&inode->i_mutex, I_MUTEX_PARENT);
2559                 mutex_lock_nested(&src->i_mutex, I_MUTEX_CHILD);
2560         } else {
2561                 mutex_lock_nested(&src->i_mutex, I_MUTEX_PARENT);
2562                 mutex_lock_nested(&inode->i_mutex, I_MUTEX_CHILD);
2563         }
2564
2565         /* determine range to clone */
2566         ret = -EINVAL;
2567         if (off + len > src->i_size || off + len < off)
2568                 goto out_unlock;
2569         if (len == 0)
2570                 olen = len = src->i_size - off;
2571         /* if we extend to eof, continue to block boundary */
2572         if (off + len == src->i_size)
2573                 len = ALIGN(src->i_size, bs) - off;
2574
2575         /* verify the end result is block aligned */
2576         if (!IS_ALIGNED(off, bs) || !IS_ALIGNED(off + len, bs) ||
2577             !IS_ALIGNED(destoff, bs))
2578                 goto out_unlock;
2579
2580         if (destoff > inode->i_size) {
2581                 ret = btrfs_cont_expand(inode, inode->i_size, destoff);
2582                 if (ret)
2583                         goto out_unlock;
2584         }
2585
2586         /* truncate page cache pages from target inode range */
2587         truncate_inode_pages_range(&inode->i_data, destoff,
2588                                    PAGE_CACHE_ALIGN(destoff + len) - 1);
2589
2590         /* do any pending delalloc/csum calc on src, one way or
2591            another, and lock file content */
2592         while (1) {
2593                 struct btrfs_ordered_extent *ordered;
2594                 lock_extent(&BTRFS_I(src)->io_tree, off, off + len - 1);
2595                 ordered = btrfs_lookup_first_ordered_extent(src, off + len - 1);
2596                 if (!ordered &&
2597                     !test_range_bit(&BTRFS_I(src)->io_tree, off, off + len - 1,
2598                                     EXTENT_DELALLOC, 0, NULL))
2599                         break;
2600                 unlock_extent(&BTRFS_I(src)->io_tree, off, off + len - 1);
2601                 if (ordered)
2602                         btrfs_put_ordered_extent(ordered);
2603                 btrfs_wait_ordered_range(src, off, len);
2604         }
2605
2606         /* clone data */
2607         key.objectid = btrfs_ino(src);
2608         key.type = BTRFS_EXTENT_DATA_KEY;
2609         key.offset = 0;
2610
2611         while (1) {
2612                 /*
2613                  * note the key will change type as we walk through the
2614                  * tree.
2615                  */
2616                 ret = btrfs_search_slot(NULL, BTRFS_I(src)->root, &key, path,
2617                                 0, 0);
2618                 if (ret < 0)
2619                         goto out;
2620
2621                 nritems = btrfs_header_nritems(path->nodes[0]);
2622                 if (path->slots[0] >= nritems) {
2623                         ret = btrfs_next_leaf(BTRFS_I(src)->root, path);
2624                         if (ret < 0)
2625                                 goto out;
2626                         if (ret > 0)
2627                                 break;
2628                         nritems = btrfs_header_nritems(path->nodes[0]);
2629                 }
2630                 leaf = path->nodes[0];
2631                 slot = path->slots[0];
2632
2633                 btrfs_item_key_to_cpu(leaf, &key, slot);
2634                 if (btrfs_key_type(&key) > BTRFS_EXTENT_DATA_KEY ||
2635                     key.objectid != btrfs_ino(src))
2636                         break;
2637
2638                 if (btrfs_key_type(&key) == BTRFS_EXTENT_DATA_KEY) {
2639                         struct btrfs_file_extent_item *extent;
2640                         int type;
2641                         u32 size;
2642                         struct btrfs_key new_key;
2643                         u64 disko = 0, diskl = 0;
2644                         u64 datao = 0, datal = 0;
2645                         u8 comp;
2646                         u64 endoff;
2647
2648                         size = btrfs_item_size_nr(leaf, slot);
2649                         read_extent_buffer(leaf, buf,
2650                                            btrfs_item_ptr_offset(leaf, slot),
2651                                            size);
2652
2653                         extent = btrfs_item_ptr(leaf, slot,
2654                                                 struct btrfs_file_extent_item);
2655                         comp = btrfs_file_extent_compression(leaf, extent);
2656                         type = btrfs_file_extent_type(leaf, extent);
2657                         if (type == BTRFS_FILE_EXTENT_REG ||
2658                             type == BTRFS_FILE_EXTENT_PREALLOC) {
2659                                 disko = btrfs_file_extent_disk_bytenr(leaf,
2660                                                                       extent);
2661                                 diskl = btrfs_file_extent_disk_num_bytes(leaf,
2662                                                                  extent);
2663                                 datao = btrfs_file_extent_offset(leaf, extent);
2664                                 datal = btrfs_file_extent_num_bytes(leaf,
2665                                                                     extent);
2666                         } else if (type == BTRFS_FILE_EXTENT_INLINE) {
2667                                 /* take upper bound, may be compressed */
2668                                 datal = btrfs_file_extent_ram_bytes(leaf,
2669                                                                     extent);
2670                         }
2671                         btrfs_release_path(path);
2672
2673                         if (key.offset + datal <= off ||
2674                             key.offset >= off + len - 1)
2675                                 goto next;
2676
2677                         memcpy(&new_key, &key, sizeof(new_key));
2678                         new_key.objectid = btrfs_ino(inode);
2679                         if (off <= key.offset)
2680                                 new_key.offset = key.offset + destoff - off;
2681                         else
2682                                 new_key.offset = destoff;
2683
2684                         /*
2685                          * 1 - adjusting old extent (we may have to split it)
2686                          * 1 - add new extent
2687                          * 1 - inode update
2688                          */
2689                         trans = btrfs_start_transaction(root, 3);
2690                         if (IS_ERR(trans)) {
2691                                 ret = PTR_ERR(trans);
2692                                 goto out;
2693                         }
2694
2695                         if (type == BTRFS_FILE_EXTENT_REG ||
2696                             type == BTRFS_FILE_EXTENT_PREALLOC) {
2697                                 /*
2698                                  *    a  | --- range to clone ---|  b
2699                                  * | ------------- extent ------------- |
2700                                  */
2701
2702                                 /* substract range b */
2703                                 if (key.offset + datal > off + len)
2704                                         datal = off + len - key.offset;
2705
2706                                 /* substract range a */
2707                                 if (off > key.offset) {
2708                                         datao += off - key.offset;
2709                                         datal -= off - key.offset;
2710                                 }
2711
2712                                 ret = btrfs_drop_extents(trans, root, inode,
2713                                                          new_key.offset,
2714                                                          new_key.offset + datal,
2715                                                          1);
2716                                 if (ret) {
2717                                         btrfs_abort_transaction(trans, root,
2718                                                                 ret);
2719                                         btrfs_end_transaction(trans, root);
2720                                         goto out;
2721                                 }
2722
2723                                 ret = btrfs_insert_empty_item(trans, root, path,
2724                                                               &new_key, size);
2725                                 if (ret) {
2726                                         btrfs_abort_transaction(trans, root,
2727                                                                 ret);
2728                                         btrfs_end_transaction(trans, root);
2729                                         goto out;
2730                                 }
2731
2732                                 leaf = path->nodes[0];
2733                                 slot = path->slots[0];
2734                                 write_extent_buffer(leaf, buf,
2735                                             btrfs_item_ptr_offset(leaf, slot),
2736                                             size);
2737
2738                                 extent = btrfs_item_ptr(leaf, slot,
2739                                                 struct btrfs_file_extent_item);
2740
2741                                 /* disko == 0 means it's a hole */
2742                                 if (!disko)
2743                                         datao = 0;
2744
2745                                 btrfs_set_file_extent_offset(leaf, extent,
2746                                                              datao);
2747                                 btrfs_set_file_extent_num_bytes(leaf, extent,
2748                                                                 datal);
2749                                 if (disko) {
2750                                         inode_add_bytes(inode, datal);
2751                                         ret = btrfs_inc_extent_ref(trans, root,
2752                                                         disko, diskl, 0,
2753                                                         root->root_key.objectid,
2754                                                         btrfs_ino(inode),
2755                                                         new_key.offset - datao,
2756                                                         0);
2757                                         if (ret) {
2758                                                 btrfs_abort_transaction(trans,
2759                                                                         root,
2760                                                                         ret);
2761                                                 btrfs_end_transaction(trans,
2762                                                                       root);
2763                                                 goto out;
2764
2765                                         }
2766                                 }
2767                         } else if (type == BTRFS_FILE_EXTENT_INLINE) {
2768                                 u64 skip = 0;
2769                                 u64 trim = 0;
2770                                 if (off > key.offset) {
2771                                         skip = off - key.offset;
2772                                         new_key.offset += skip;
2773                                 }
2774
2775                                 if (key.offset + datal > off + len)
2776                                         trim = key.offset + datal - (off + len);
2777
2778                                 if (comp && (skip || trim)) {
2779                                         ret = -EINVAL;
2780                                         btrfs_end_transaction(trans, root);
2781                                         goto out;
2782                                 }
2783                                 size -= skip + trim;
2784                                 datal -= skip + trim;
2785
2786                                 ret = btrfs_drop_extents(trans, root, inode,
2787                                                          new_key.offset,
2788                                                          new_key.offset + datal,
2789                                                          1);
2790                                 if (ret) {
2791                                         btrfs_abort_transaction(trans, root,
2792                                                                 ret);
2793                                         btrfs_end_transaction(trans, root);
2794                                         goto out;
2795                                 }
2796
2797                                 ret = btrfs_insert_empty_item(trans, root, path,
2798                                                               &new_key, size);
2799                                 if (ret) {
2800                                         btrfs_abort_transaction(trans, root,
2801                                                                 ret);
2802                                         btrfs_end_transaction(trans, root);
2803                                         goto out;
2804                                 }
2805
2806                                 if (skip) {
2807                                         u32 start =
2808                                           btrfs_file_extent_calc_inline_size(0);
2809                                         memmove(buf+start, buf+start+skip,
2810                                                 datal);
2811                                 }
2812
2813                                 leaf = path->nodes[0];
2814                                 slot = path->slots[0];
2815                                 write_extent_buffer(leaf, buf,
2816                                             btrfs_item_ptr_offset(leaf, slot),
2817                                             size);
2818                                 inode_add_bytes(inode, datal);
2819                         }
2820
2821                         btrfs_mark_buffer_dirty(leaf);
2822                         btrfs_release_path(path);
2823
2824                         inode_inc_iversion(inode);
2825                         inode->i_mtime = inode->i_ctime = CURRENT_TIME;
2826
2827                         /*
2828                          * we round up to the block size at eof when
2829                          * determining which extents to clone above,
2830                          * but shouldn't round up the file size
2831                          */
2832                         endoff = new_key.offset + datal;
2833                         if (endoff > destoff+olen)
2834                                 endoff = destoff+olen;
2835                         if (endoff > inode->i_size)
2836                                 btrfs_i_size_write(inode, endoff);
2837
2838                         ret = btrfs_update_inode(trans, root, inode);
2839                         if (ret) {
2840                                 btrfs_abort_transaction(trans, root, ret);
2841                                 btrfs_end_transaction(trans, root);
2842                                 goto out;
2843                         }
2844                         ret = btrfs_end_transaction(trans, root);
2845                 }
2846 next:
2847                 btrfs_release_path(path);
2848                 key.offset++;
2849         }
2850         ret = 0;
2851 out:
2852         btrfs_release_path(path);
2853         unlock_extent(&BTRFS_I(src)->io_tree, off, off + len - 1);
2854 out_unlock:
2855         mutex_unlock(&src->i_mutex);
2856         mutex_unlock(&inode->i_mutex);
2857         vfree(buf);
2858         btrfs_free_path(path);
2859 out_fput:
2860         fdput(src_file);
2861 out_drop_write:
2862         mnt_drop_write_file(file);
2863         return ret;
2864 }
2865
2866 static long btrfs_ioctl_clone_range(struct file *file, void __user *argp)
2867 {
2868         struct btrfs_ioctl_clone_range_args args;
2869
2870         if (copy_from_user(&args, argp, sizeof(args)))
2871                 return -EFAULT;
2872         return btrfs_ioctl_clone(file, args.src_fd, args.src_offset,
2873                                  args.src_length, args.dest_offset);
2874 }
2875
2876 /*
2877  * there are many ways the trans_start and trans_end ioctls can lead
2878  * to deadlocks.  They should only be used by applications that
2879  * basically own the machine, and have a very in depth understanding
2880  * of all the possible deadlocks and enospc problems.
2881  */
2882 static long btrfs_ioctl_trans_start(struct file *file)
2883 {
2884         struct inode *inode = file_inode(file);
2885         struct btrfs_root *root = BTRFS_I(inode)->root;
2886         struct btrfs_trans_handle *trans;
2887         int ret;
2888
2889         ret = -EPERM;
2890         if (!capable(CAP_SYS_ADMIN))
2891                 goto out;
2892
2893         ret = -EINPROGRESS;
2894         if (file->private_data)
2895                 goto out;
2896
2897         ret = -EROFS;
2898         if (btrfs_root_readonly(root))
2899                 goto out;
2900
2901         ret = mnt_want_write_file(file);
2902         if (ret)
2903                 goto out;
2904
2905         atomic_inc(&root->fs_info->open_ioctl_trans);
2906
2907         ret = -ENOMEM;
2908         trans = btrfs_start_ioctl_transaction(root);
2909         if (IS_ERR(trans))
2910                 goto out_drop;
2911
2912         file->private_data = trans;
2913         return 0;
2914
2915 out_drop:
2916         atomic_dec(&root->fs_info->open_ioctl_trans);
2917         mnt_drop_write_file(file);
2918 out:
2919         return ret;
2920 }
2921
2922 static long btrfs_ioctl_default_subvol(struct file *file, void __user *argp)
2923 {
2924         struct inode *inode = file_inode(file);
2925         struct btrfs_root *root = BTRFS_I(inode)->root;
2926         struct btrfs_root *new_root;
2927         struct btrfs_dir_item *di;
2928         struct btrfs_trans_handle *trans;
2929         struct btrfs_path *path;
2930         struct btrfs_key location;
2931         struct btrfs_disk_key disk_key;
2932         u64 objectid = 0;
2933         u64 dir_id;
2934         int ret;
2935
2936         if (!capable(CAP_SYS_ADMIN))
2937                 return -EPERM;
2938
2939         ret = mnt_want_write_file(file);
2940         if (ret)
2941                 return ret;
2942
2943         if (copy_from_user(&objectid, argp, sizeof(objectid))) {
2944                 ret = -EFAULT;
2945                 goto out;
2946         }
2947
2948         if (!objectid)
2949                 objectid = root->root_key.objectid;
2950
2951         location.objectid = objectid;
2952         location.type = BTRFS_ROOT_ITEM_KEY;
2953         location.offset = (u64)-1;
2954
2955         new_root = btrfs_read_fs_root_no_name(root->fs_info, &location);
2956         if (IS_ERR(new_root)) {
2957                 ret = PTR_ERR(new_root);
2958                 goto out;
2959         }
2960
2961         if (btrfs_root_refs(&new_root->root_item) == 0) {
2962                 ret = -ENOENT;
2963                 goto out;
2964         }
2965
2966         path = btrfs_alloc_path();
2967         if (!path) {
2968                 ret = -ENOMEM;
2969                 goto out;
2970         }
2971         path->leave_spinning = 1;
2972
2973         trans = btrfs_start_transaction(root, 1);
2974         if (IS_ERR(trans)) {
2975                 btrfs_free_path(path);
2976                 ret = PTR_ERR(trans);
2977                 goto out;
2978         }
2979
2980         dir_id = btrfs_super_root_dir(root->fs_info->super_copy);
2981         di = btrfs_lookup_dir_item(trans, root->fs_info->tree_root, path,
2982                                    dir_id, "default", 7, 1);
2983         if (IS_ERR_OR_NULL(di)) {
2984                 btrfs_free_path(path);
2985                 btrfs_end_transaction(trans, root);
2986                 printk(KERN_ERR "Umm, you don't have the default dir item, "
2987                        "this isn't going to work\n");
2988                 ret = -ENOENT;
2989                 goto out;
2990         }
2991
2992         btrfs_cpu_key_to_disk(&disk_key, &new_root->root_key);
2993         btrfs_set_dir_item_key(path->nodes[0], di, &disk_key);
2994         btrfs_mark_buffer_dirty(path->nodes[0]);
2995         btrfs_free_path(path);
2996
2997         btrfs_set_fs_incompat(root->fs_info, DEFAULT_SUBVOL);
2998         btrfs_end_transaction(trans, root);
2999 out:
3000         mnt_drop_write_file(file);
3001         return ret;
3002 }
3003
3004 void btrfs_get_block_group_info(struct list_head *groups_list,
3005                                 struct btrfs_ioctl_space_info *space)
3006 {
3007         struct btrfs_block_group_cache *block_group;
3008
3009         space->total_bytes = 0;
3010         space->used_bytes = 0;
3011         space->flags = 0;
3012         list_for_each_entry(block_group, groups_list, list) {
3013                 space->flags = block_group->flags;
3014                 space->total_bytes += block_group->key.offset;
3015                 space->used_bytes +=
3016                         btrfs_block_group_used(&block_group->item);
3017         }
3018 }
3019
3020 static long btrfs_ioctl_space_info(struct btrfs_root *root, void __user *arg)
3021 {
3022         struct btrfs_ioctl_space_args space_args;
3023         struct btrfs_ioctl_space_info space;
3024         struct btrfs_ioctl_space_info *dest;
3025         struct btrfs_ioctl_space_info *dest_orig;
3026         struct btrfs_ioctl_space_info __user *user_dest;
3027         struct btrfs_space_info *info;
3028         u64 types[] = {BTRFS_BLOCK_GROUP_DATA,
3029                        BTRFS_BLOCK_GROUP_SYSTEM,
3030                        BTRFS_BLOCK_GROUP_METADATA,
3031                        BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA};
3032         int num_types = 4;
3033         int alloc_size;
3034         int ret = 0;
3035         u64 slot_count = 0;
3036         int i, c;
3037
3038         if (copy_from_user(&space_args,
3039                            (struct btrfs_ioctl_space_args __user *)arg,
3040                            sizeof(space_args)))
3041                 return -EFAULT;
3042
3043         for (i = 0; i < num_types; i++) {
3044                 struct btrfs_space_info *tmp;
3045
3046                 info = NULL;
3047                 rcu_read_lock();
3048                 list_for_each_entry_rcu(tmp, &root->fs_info->space_info,
3049                                         list) {
3050                         if (tmp->flags == types[i]) {
3051                                 info = tmp;
3052                                 break;
3053                         }
3054                 }
3055                 rcu_read_unlock();
3056
3057                 if (!info)
3058                         continue;
3059
3060                 down_read(&info->groups_sem);
3061                 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
3062                         if (!list_empty(&info->block_groups[c]))
3063                                 slot_count++;
3064                 }
3065                 up_read(&info->groups_sem);
3066         }
3067
3068         /* space_slots == 0 means they are asking for a count */
3069         if (space_args.space_slots == 0) {
3070                 space_args.total_spaces = slot_count;
3071                 goto out;
3072         }
3073
3074         slot_count = min_t(u64, space_args.space_slots, slot_count);
3075
3076         alloc_size = sizeof(*dest) * slot_count;
3077
3078         /* we generally have at most 6 or so space infos, one for each raid
3079          * level.  So, a whole page should be more than enough for everyone
3080          */
3081         if (alloc_size > PAGE_CACHE_SIZE)
3082                 return -ENOMEM;
3083
3084         space_args.total_spaces = 0;
3085         dest = kmalloc(alloc_size, GFP_NOFS);
3086         if (!dest)
3087                 return -ENOMEM;
3088         dest_orig = dest;
3089
3090         /* now we have a buffer to copy into */
3091         for (i = 0; i < num_types; i++) {
3092                 struct btrfs_space_info *tmp;
3093
3094                 if (!slot_count)
3095                         break;
3096
3097                 info = NULL;
3098                 rcu_read_lock();
3099                 list_for_each_entry_rcu(tmp, &root->fs_info->space_info,
3100                                         list) {
3101                         if (tmp->flags == types[i]) {
3102                                 info = tmp;
3103                                 break;
3104                         }
3105                 }
3106                 rcu_read_unlock();
3107
3108                 if (!info)
3109                         continue;
3110                 down_read(&info->groups_sem);
3111                 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
3112                         if (!list_empty(&info->block_groups[c])) {
3113                                 btrfs_get_block_group_info(
3114                                         &info->block_groups[c], &space);
3115                                 memcpy(dest, &space, sizeof(space));
3116                                 dest++;
3117                                 space_args.total_spaces++;
3118                                 slot_count--;
3119                         }
3120                         if (!slot_count)
3121                                 break;
3122                 }
3123                 up_read(&info->groups_sem);
3124         }
3125
3126         user_dest = (struct btrfs_ioctl_space_info __user *)
3127                 (arg + sizeof(struct btrfs_ioctl_space_args));
3128
3129         if (copy_to_user(user_dest, dest_orig, alloc_size))
3130                 ret = -EFAULT;
3131
3132         kfree(dest_orig);
3133 out:
3134         if (ret == 0 && copy_to_user(arg, &space_args, sizeof(space_args)))
3135                 ret = -EFAULT;
3136
3137         return ret;
3138 }
3139
3140 /*
3141  * there are many ways the trans_start and trans_end ioctls can lead
3142  * to deadlocks.  They should only be used by applications that
3143  * basically own the machine, and have a very in depth understanding
3144  * of all the possible deadlocks and enospc problems.
3145  */
3146 long btrfs_ioctl_trans_end(struct file *file)
3147 {
3148         struct inode *inode = file_inode(file);
3149         struct btrfs_root *root = BTRFS_I(inode)->root;
3150         struct btrfs_trans_handle *trans;
3151
3152         trans = file->private_data;
3153         if (!trans)
3154                 return -EINVAL;
3155         file->private_data = NULL;
3156
3157         btrfs_end_transaction(trans, root);
3158
3159         atomic_dec(&root->fs_info->open_ioctl_trans);
3160
3161         mnt_drop_write_file(file);
3162         return 0;
3163 }
3164
3165 static noinline long btrfs_ioctl_start_sync(struct btrfs_root *root,
3166                                             void __user *argp)
3167 {
3168         struct btrfs_trans_handle *trans;
3169         u64 transid;
3170         int ret;
3171
3172         trans = btrfs_attach_transaction_barrier(root);
3173         if (IS_ERR(trans)) {
3174                 if (PTR_ERR(trans) != -ENOENT)
3175                         return PTR_ERR(trans);
3176
3177                 /* No running transaction, don't bother */
3178                 transid = root->fs_info->last_trans_committed;
3179                 goto out;
3180         }
3181         transid = trans->transid;
3182         ret = btrfs_commit_transaction_async(trans, root, 0);
3183         if (ret) {
3184                 btrfs_end_transaction(trans, root);
3185                 return ret;
3186         }
3187 out:
3188         if (argp)
3189                 if (copy_to_user(argp, &transid, sizeof(transid)))
3190                         return -EFAULT;
3191         return 0;
3192 }
3193
3194 static noinline long btrfs_ioctl_wait_sync(struct btrfs_root *root,
3195                                            void __user *argp)
3196 {
3197         u64 transid;
3198
3199         if (argp) {
3200                 if (copy_from_user(&transid, argp, sizeof(transid)))
3201                         return -EFAULT;
3202         } else {
3203                 transid = 0;  /* current trans */
3204         }
3205         return btrfs_wait_for_commit(root, transid);
3206 }
3207
3208 static long btrfs_ioctl_scrub(struct file *file, void __user *arg)
3209 {
3210         struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
3211         struct btrfs_ioctl_scrub_args *sa;
3212         int ret;
3213
3214         if (!capable(CAP_SYS_ADMIN))
3215                 return -EPERM;
3216
3217         sa = memdup_user(arg, sizeof(*sa));
3218         if (IS_ERR(sa))
3219                 return PTR_ERR(sa);
3220
3221         if (!(sa->flags & BTRFS_SCRUB_READONLY)) {
3222                 ret = mnt_want_write_file(file);
3223                 if (ret)
3224                         goto out;
3225         }
3226
3227         ret = btrfs_scrub_dev(root->fs_info, sa->devid, sa->start, sa->end,
3228                               &sa->progress, sa->flags & BTRFS_SCRUB_READONLY,
3229                               0);
3230
3231         if (copy_to_user(arg, sa, sizeof(*sa)))
3232                 ret = -EFAULT;
3233
3234         if (!(sa->flags & BTRFS_SCRUB_READONLY))
3235                 mnt_drop_write_file(file);
3236 out:
3237         kfree(sa);
3238         return ret;
3239 }
3240
3241 static long btrfs_ioctl_scrub_cancel(struct btrfs_root *root, void __user *arg)
3242 {
3243         if (!capable(CAP_SYS_ADMIN))
3244                 return -EPERM;
3245
3246         return btrfs_scrub_cancel(root->fs_info);
3247 }
3248
3249 static long btrfs_ioctl_scrub_progress(struct btrfs_root *root,
3250                                        void __user *arg)
3251 {
3252         struct btrfs_ioctl_scrub_args *sa;
3253         int ret;
3254
3255         if (!capable(CAP_SYS_ADMIN))
3256                 return -EPERM;
3257
3258         sa = memdup_user(arg, sizeof(*sa));
3259         if (IS_ERR(sa))
3260                 return PTR_ERR(sa);
3261
3262         ret = btrfs_scrub_progress(root, sa->devid, &sa->progress);
3263
3264         if (copy_to_user(arg, sa, sizeof(*sa)))
3265                 ret = -EFAULT;
3266
3267         kfree(sa);
3268         return ret;
3269 }
3270
3271 static long btrfs_ioctl_get_dev_stats(struct btrfs_root *root,
3272                                       void __user *arg)
3273 {
3274         struct btrfs_ioctl_get_dev_stats *sa;
3275         int ret;
3276
3277         sa = memdup_user(arg, sizeof(*sa));
3278         if (IS_ERR(sa))
3279                 return PTR_ERR(sa);
3280
3281         if ((sa->flags & BTRFS_DEV_STATS_RESET) && !capable(CAP_SYS_ADMIN)) {
3282                 kfree(sa);
3283                 return -EPERM;
3284         }
3285
3286         ret = btrfs_get_dev_stats(root, sa);
3287
3288         if (copy_to_user(arg, sa, sizeof(*sa)))
3289                 ret = -EFAULT;
3290
3291         kfree(sa);
3292         return ret;
3293 }
3294
3295 static long btrfs_ioctl_dev_replace(struct btrfs_root *root, void __user *arg)
3296 {
3297         struct btrfs_ioctl_dev_replace_args *p;
3298         int ret;
3299
3300         if (!capable(CAP_SYS_ADMIN))
3301                 return -EPERM;
3302
3303         p = memdup_user(arg, sizeof(*p));
3304         if (IS_ERR(p))
3305                 return PTR_ERR(p);
3306
3307         switch (p->cmd) {
3308         case BTRFS_IOCTL_DEV_REPLACE_CMD_START:
3309                 if (root->fs_info->sb->s_flags & MS_RDONLY)
3310                         return -EROFS;
3311
3312                 if (atomic_xchg(
3313                         &root->fs_info->mutually_exclusive_operation_running,
3314                         1)) {
3315                         pr_info("btrfs: dev add/delete/balance/replace/resize operation in progress\n");
3316                         ret = -EINPROGRESS;
3317                 } else {
3318                         ret = btrfs_dev_replace_start(root, p);
3319                         atomic_set(
3320                          &root->fs_info->mutually_exclusive_operation_running,
3321                          0);
3322                 }
3323                 break;
3324         case BTRFS_IOCTL_DEV_REPLACE_CMD_STATUS:
3325                 btrfs_dev_replace_status(root->fs_info, p);
3326                 ret = 0;
3327                 break;
3328         case BTRFS_IOCTL_DEV_REPLACE_CMD_CANCEL:
3329                 ret = btrfs_dev_replace_cancel(root->fs_info, p);
3330                 break;
3331         default:
3332                 ret = -EINVAL;
3333                 break;
3334         }
3335
3336         if (copy_to_user(arg, p, sizeof(*p)))
3337                 ret = -EFAULT;
3338
3339         kfree(p);
3340         return ret;
3341 }
3342
3343 static long btrfs_ioctl_ino_to_path(struct btrfs_root *root, void __user *arg)
3344 {
3345         int ret = 0;
3346         int i;
3347         u64 rel_ptr;
3348         int size;
3349         struct btrfs_ioctl_ino_path_args *ipa = NULL;
3350         struct inode_fs_paths *ipath = NULL;
3351         struct btrfs_path *path;
3352
3353         if (!capable(CAP_DAC_READ_SEARCH))
3354                 return -EPERM;
3355
3356         path = btrfs_alloc_path();
3357         if (!path) {
3358                 ret = -ENOMEM;
3359                 goto out;
3360         }
3361
3362         ipa = memdup_user(arg, sizeof(*ipa));
3363         if (IS_ERR(ipa)) {
3364                 ret = PTR_ERR(ipa);
3365                 ipa = NULL;
3366                 goto out;
3367         }
3368
3369         size = min_t(u32, ipa->size, 4096);
3370         ipath = init_ipath(size, root, path);
3371         if (IS_ERR(ipath)) {
3372                 ret = PTR_ERR(ipath);
3373                 ipath = NULL;
3374                 goto out;
3375         }
3376
3377         ret = paths_from_inode(ipa->inum, ipath);
3378         if (ret < 0)
3379                 goto out;
3380
3381         for (i = 0; i < ipath->fspath->elem_cnt; ++i) {
3382                 rel_ptr = ipath->fspath->val[i] -
3383                           (u64)(unsigned long)ipath->fspath->val;
3384                 ipath->fspath->val[i] = rel_ptr;
3385         }
3386
3387         ret = copy_to_user((void *)(unsigned long)ipa->fspath,
3388                            (void *)(unsigned long)ipath->fspath, size);
3389         if (ret) {
3390                 ret = -EFAULT;
3391                 goto out;
3392         }
3393
3394 out:
3395         btrfs_free_path(path);
3396         free_ipath(ipath);
3397         kfree(ipa);
3398
3399         return ret;
3400 }
3401
3402 static int build_ino_list(u64 inum, u64 offset, u64 root, void *ctx)
3403 {
3404         struct btrfs_data_container *inodes = ctx;
3405         const size_t c = 3 * sizeof(u64);
3406
3407         if (inodes->bytes_left >= c) {
3408                 inodes->bytes_left -= c;
3409                 inodes->val[inodes->elem_cnt] = inum;
3410                 inodes->val[inodes->elem_cnt + 1] = offset;
3411                 inodes->val[inodes->elem_cnt + 2] = root;
3412                 inodes->elem_cnt += 3;
3413         } else {
3414                 inodes->bytes_missing += c - inodes->bytes_left;
3415                 inodes->bytes_left = 0;
3416                 inodes->elem_missed += 3;
3417         }
3418
3419         return 0;
3420 }
3421
3422 static long btrfs_ioctl_logical_to_ino(struct btrfs_root *root,
3423                                         void __user *arg)
3424 {
3425         int ret = 0;
3426         int size;
3427         struct btrfs_ioctl_logical_ino_args *loi;
3428         struct btrfs_data_container *inodes = NULL;
3429         struct btrfs_path *path = NULL;
3430
3431         if (!capable(CAP_SYS_ADMIN))
3432                 return -EPERM;
3433
3434         loi = memdup_user(arg, sizeof(*loi));
3435         if (IS_ERR(loi)) {
3436                 ret = PTR_ERR(loi);
3437                 loi = NULL;
3438                 goto out;
3439         }
3440
3441         path = btrfs_alloc_path();
3442         if (!path) {
3443                 ret = -ENOMEM;
3444                 goto out;
3445         }
3446
3447         size = min_t(u32, loi->size, 64 * 1024);
3448         inodes = init_data_container(size);
3449         if (IS_ERR(inodes)) {
3450                 ret = PTR_ERR(inodes);
3451                 inodes = NULL;
3452                 goto out;
3453         }
3454
3455         ret = iterate_inodes_from_logical(loi->logical, root->fs_info, path,
3456                                           build_ino_list, inodes);
3457         if (ret == -EINVAL)
3458                 ret = -ENOENT;
3459         if (ret < 0)
3460                 goto out;
3461
3462         ret = copy_to_user((void *)(unsigned long)loi->inodes,
3463                            (void *)(unsigned long)inodes, size);
3464         if (ret)
3465                 ret = -EFAULT;
3466
3467 out:
3468         btrfs_free_path(path);
3469         vfree(inodes);
3470         kfree(loi);
3471
3472         return ret;
3473 }
3474
3475 void update_ioctl_balance_args(struct btrfs_fs_info *fs_info, int lock,
3476                                struct btrfs_ioctl_balance_args *bargs)
3477 {
3478         struct btrfs_balance_control *bctl = fs_info->balance_ctl;
3479
3480         bargs->flags = bctl->flags;
3481
3482         if (atomic_read(&fs_info->balance_running))
3483                 bargs->state |= BTRFS_BALANCE_STATE_RUNNING;
3484         if (atomic_read(&fs_info->balance_pause_req))
3485                 bargs->state |= BTRFS_BALANCE_STATE_PAUSE_REQ;
3486         if (atomic_read(&fs_info->balance_cancel_req))
3487                 bargs->state |= BTRFS_BALANCE_STATE_CANCEL_REQ;
3488
3489         memcpy(&bargs->data, &bctl->data, sizeof(bargs->data));
3490         memcpy(&bargs->meta, &bctl->meta, sizeof(bargs->meta));
3491         memcpy(&bargs->sys, &bctl->sys, sizeof(bargs->sys));
3492
3493         if (lock) {
3494                 spin_lock(&fs_info->balance_lock);
3495                 memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat));
3496                 spin_unlock(&fs_info->balance_lock);
3497         } else {
3498                 memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat));
3499         }
3500 }
3501
3502 static long btrfs_ioctl_balance(struct file *file, void __user *arg)
3503 {
3504         struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
3505         struct btrfs_fs_info *fs_info = root->fs_info;
3506         struct btrfs_ioctl_balance_args *bargs;
3507         struct btrfs_balance_control *bctl;
3508         bool need_unlock; /* for mut. excl. ops lock */
3509         int ret;
3510
3511         if (!capable(CAP_SYS_ADMIN))
3512                 return -EPERM;
3513
3514         ret = mnt_want_write_file(file);
3515         if (ret)
3516                 return ret;
3517
3518 again:
3519         if (!atomic_xchg(&fs_info->mutually_exclusive_operation_running, 1)) {
3520                 mutex_lock(&fs_info->volume_mutex);
3521                 mutex_lock(&fs_info->balance_mutex);
3522                 need_unlock = true;
3523                 goto locked;
3524         }
3525
3526         /*
3527          * mut. excl. ops lock is locked.  Three possibilites:
3528          *   (1) some other op is running
3529          *   (2) balance is running
3530          *   (3) balance is paused -- special case (think resume)
3531          */
3532         mutex_lock(&fs_info->balance_mutex);
3533         if (fs_info->balance_ctl) {
3534                 /* this is either (2) or (3) */
3535                 if (!atomic_read(&fs_info->balance_running)) {
3536                         mutex_unlock(&fs_info->balance_mutex);
3537                         if (!mutex_trylock(&fs_info->volume_mutex))
3538                                 goto again;
3539                         mutex_lock(&fs_info->balance_mutex);
3540
3541                         if (fs_info->balance_ctl &&
3542                             !atomic_read(&fs_info->balance_running)) {
3543                                 /* this is (3) */
3544                                 need_unlock = false;
3545                                 goto locked;
3546                         }
3547
3548                         mutex_unlock(&fs_info->balance_mutex);
3549                         mutex_unlock(&fs_info->volume_mutex);
3550                         goto again;
3551                 } else {
3552                         /* this is (2) */
3553                         mutex_unlock(&fs_info->balance_mutex);
3554                         ret = -EINPROGRESS;
3555                         goto out;
3556                 }
3557         } else {
3558                 /* this is (1) */
3559                 mutex_unlock(&fs_info->balance_mutex);
3560                 pr_info("btrfs: dev add/delete/balance/replace/resize operation in progress\n");
3561                 ret = -EINVAL;
3562                 goto out;
3563         }
3564
3565 locked:
3566         BUG_ON(!atomic_read(&fs_info->mutually_exclusive_operation_running));
3567
3568         if (arg) {
3569                 bargs = memdup_user(arg, sizeof(*bargs));
3570                 if (IS_ERR(bargs)) {
3571                         ret = PTR_ERR(bargs);
3572                         goto out_unlock;
3573                 }
3574
3575                 if (bargs->flags & BTRFS_BALANCE_RESUME) {
3576                         if (!fs_info->balance_ctl) {
3577                                 ret = -ENOTCONN;
3578                                 goto out_bargs;
3579                         }
3580
3581                         bctl = fs_info->balance_ctl;
3582                         spin_lock(&fs_info->balance_lock);
3583                         bctl->flags |= BTRFS_BALANCE_RESUME;
3584                         spin_unlock(&fs_info->balance_lock);
3585
3586                         goto do_balance;
3587                 }
3588         } else {
3589                 bargs = NULL;
3590         }
3591
3592         if (fs_info->balance_ctl) {
3593                 ret = -EINPROGRESS;
3594                 goto out_bargs;
3595         }
3596
3597         bctl = kzalloc(sizeof(*bctl), GFP_NOFS);
3598         if (!bctl) {
3599                 ret = -ENOMEM;
3600                 goto out_bargs;
3601         }
3602
3603         bctl->fs_info = fs_info;
3604         if (arg) {
3605                 memcpy(&bctl->data, &bargs->data, sizeof(bctl->data));
3606                 memcpy(&bctl->meta, &bargs->meta, sizeof(bctl->meta));
3607                 memcpy(&bctl->sys, &bargs->sys, sizeof(bctl->sys));
3608
3609                 bctl->flags = bargs->flags;
3610         } else {
3611                 /* balance everything - no filters */
3612                 bctl->flags |= BTRFS_BALANCE_TYPE_MASK;
3613         }
3614
3615 do_balance:
3616         /*
3617          * Ownership of bctl and mutually_exclusive_operation_running
3618          * goes to to btrfs_balance.  bctl is freed in __cancel_balance,
3619          * or, if restriper was paused all the way until unmount, in
3620          * free_fs_info.  mutually_exclusive_operation_running is
3621          * cleared in __cancel_balance.
3622          */
3623         need_unlock = false;
3624
3625         ret = btrfs_balance(bctl, bargs);
3626
3627         if (arg) {
3628                 if (copy_to_user(arg, bargs, sizeof(*bargs)))
3629                         ret = -EFAULT;
3630         }
3631
3632 out_bargs:
3633         kfree(bargs);
3634 out_unlock:
3635         mutex_unlock(&fs_info->balance_mutex);
3636         mutex_unlock(&fs_info->volume_mutex);
3637         if (need_unlock)
3638                 atomic_set(&fs_info->mutually_exclusive_operation_running, 0);
3639 out:
3640         mnt_drop_write_file(file);
3641         return ret;
3642 }
3643
3644 static long btrfs_ioctl_balance_ctl(struct btrfs_root *root, int cmd)
3645 {
3646         if (!capable(CAP_SYS_ADMIN))
3647                 return -EPERM;
3648
3649         switch (cmd) {
3650         case BTRFS_BALANCE_CTL_PAUSE:
3651                 return btrfs_pause_balance(root->fs_info);
3652         case BTRFS_BALANCE_CTL_CANCEL:
3653                 return btrfs_cancel_balance(root->fs_info);
3654         }
3655
3656         return -EINVAL;
3657 }
3658
3659 static long btrfs_ioctl_balance_progress(struct btrfs_root *root,
3660                                          void __user *arg)
3661 {
3662         struct btrfs_fs_info *fs_info = root->fs_info;
3663         struct btrfs_ioctl_balance_args *bargs;
3664         int ret = 0;
3665
3666         if (!capable(CAP_SYS_ADMIN))
3667                 return -EPERM;
3668
3669         mutex_lock(&fs_info->balance_mutex);
3670         if (!fs_info->balance_ctl) {
3671                 ret = -ENOTCONN;
3672                 goto out;
3673         }
3674
3675         bargs = kzalloc(sizeof(*bargs), GFP_NOFS);
3676         if (!bargs) {
3677                 ret = -ENOMEM;
3678                 goto out;
3679         }
3680
3681         update_ioctl_balance_args(fs_info, 1, bargs);
3682
3683         if (copy_to_user(arg, bargs, sizeof(*bargs)))
3684                 ret = -EFAULT;
3685
3686         kfree(bargs);
3687 out:
3688         mutex_unlock(&fs_info->balance_mutex);
3689         return ret;
3690 }
3691
3692 static long btrfs_ioctl_quota_ctl(struct file *file, void __user *arg)
3693 {
3694         struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
3695         struct btrfs_ioctl_quota_ctl_args *sa;
3696         struct btrfs_trans_handle *trans = NULL;
3697         int ret;
3698         int err;
3699
3700         if (!capable(CAP_SYS_ADMIN))
3701                 return -EPERM;
3702
3703         ret = mnt_want_write_file(file);
3704         if (ret)
3705                 return ret;
3706
3707         sa = memdup_user(arg, sizeof(*sa));
3708         if (IS_ERR(sa)) {
3709                 ret = PTR_ERR(sa);
3710                 goto drop_write;
3711         }
3712
3713         down_write(&root->fs_info->subvol_sem);
3714         trans = btrfs_start_transaction(root->fs_info->tree_root, 2);
3715         if (IS_ERR(trans)) {
3716                 ret = PTR_ERR(trans);
3717                 goto out;
3718         }
3719
3720         switch (sa->cmd) {
3721         case BTRFS_QUOTA_CTL_ENABLE:
3722                 ret = btrfs_quota_enable(trans, root->fs_info);
3723                 break;
3724         case BTRFS_QUOTA_CTL_DISABLE:
3725                 ret = btrfs_quota_disable(trans, root->fs_info);
3726                 break;
3727         default:
3728                 ret = -EINVAL;
3729                 break;
3730         }
3731
3732         if (copy_to_user(arg, sa, sizeof(*sa)))
3733                 ret = -EFAULT;
3734
3735         err = btrfs_commit_transaction(trans, root->fs_info->tree_root);
3736         if (err && !ret)
3737                 ret = err;
3738 out:
3739         kfree(sa);
3740         up_write(&root->fs_info->subvol_sem);
3741 drop_write:
3742         mnt_drop_write_file(file);
3743         return ret;
3744 }
3745
3746 static long btrfs_ioctl_qgroup_assign(struct file *file, void __user *arg)
3747 {
3748         struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
3749         struct btrfs_ioctl_qgroup_assign_args *sa;
3750         struct btrfs_trans_handle *trans;
3751         int ret;
3752         int err;
3753
3754         if (!capable(CAP_SYS_ADMIN))
3755                 return -EPERM;
3756
3757         ret = mnt_want_write_file(file);
3758         if (ret)
3759                 return ret;
3760
3761         sa = memdup_user(arg, sizeof(*sa));
3762         if (IS_ERR(sa)) {
3763                 ret = PTR_ERR(sa);
3764                 goto drop_write;
3765         }
3766
3767         trans = btrfs_join_transaction(root);
3768         if (IS_ERR(trans)) {
3769                 ret = PTR_ERR(trans);
3770                 goto out;
3771         }
3772
3773         /* FIXME: check if the IDs really exist */
3774         if (sa->assign) {
3775                 ret = btrfs_add_qgroup_relation(trans, root->fs_info,
3776                                                 sa->src, sa->dst);
3777         } else {
3778                 ret = btrfs_del_qgroup_relation(trans, root->fs_info,
3779                                                 sa->src, sa->dst);
3780         }
3781
3782         err = btrfs_end_transaction(trans, root);
3783         if (err && !ret)
3784                 ret = err;
3785
3786 out:
3787         kfree(sa);
3788 drop_write:
3789         mnt_drop_write_file(file);
3790         return ret;
3791 }
3792
3793 static long btrfs_ioctl_qgroup_create(struct file *file, void __user *arg)
3794 {
3795         struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
3796         struct btrfs_ioctl_qgroup_create_args *sa;
3797         struct btrfs_trans_handle *trans;
3798         int ret;
3799         int err;
3800
3801         if (!capable(CAP_SYS_ADMIN))
3802                 return -EPERM;
3803
3804         ret = mnt_want_write_file(file);
3805         if (ret)
3806                 return ret;
3807
3808         sa = memdup_user(arg, sizeof(*sa));
3809         if (IS_ERR(sa)) {
3810                 ret = PTR_ERR(sa);
3811                 goto drop_write;
3812         }
3813
3814         if (!sa->qgroupid) {
3815                 ret = -EINVAL;
3816                 goto out;
3817         }
3818
3819         trans = btrfs_join_transaction(root);
3820         if (IS_ERR(trans)) {
3821                 ret = PTR_ERR(trans);
3822                 goto out;
3823         }
3824
3825         /* FIXME: check if the IDs really exist */
3826         if (sa->create) {
3827                 ret = btrfs_create_qgroup(trans, root->fs_info, sa->qgroupid,
3828                                           NULL);
3829         } else {
3830                 ret = btrfs_remove_qgroup(trans, root->fs_info, sa->qgroupid);
3831         }
3832
3833         err = btrfs_end_transaction(trans, root);
3834         if (err && !ret)
3835                 ret = err;
3836
3837 out:
3838         kfree(sa);
3839 drop_write:
3840         mnt_drop_write_file(file);
3841         return ret;
3842 }
3843
3844 static long btrfs_ioctl_qgroup_limit(struct file *file, void __user *arg)
3845 {
3846         struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
3847         struct btrfs_ioctl_qgroup_limit_args *sa;
3848         struct btrfs_trans_handle *trans;
3849         int ret;
3850         int err;
3851         u64 qgroupid;
3852
3853         if (!capable(CAP_SYS_ADMIN))
3854                 return -EPERM;
3855
3856         ret = mnt_want_write_file(file);
3857         if (ret)
3858                 return ret;
3859
3860         sa = memdup_user(arg, sizeof(*sa));
3861         if (IS_ERR(sa)) {
3862                 ret = PTR_ERR(sa);
3863                 goto drop_write;
3864         }
3865
3866         trans = btrfs_join_transaction(root);
3867         if (IS_ERR(trans)) {
3868                 ret = PTR_ERR(trans);
3869                 goto out;
3870         }
3871
3872         qgroupid = sa->qgroupid;
3873         if (!qgroupid) {
3874                 /* take the current subvol as qgroup */
3875                 qgroupid = root->root_key.objectid;
3876         }
3877
3878         /* FIXME: check if the IDs really exist */
3879         ret = btrfs_limit_qgroup(trans, root->fs_info, qgroupid, &sa->lim);
3880
3881         err = btrfs_end_transaction(trans, root);
3882         if (err && !ret)
3883                 ret = err;
3884
3885 out:
3886         kfree(sa);
3887 drop_write:
3888         mnt_drop_write_file(file);
3889         return ret;
3890 }
3891
3892 static long btrfs_ioctl_quota_rescan(struct file *file, void __user *arg)
3893 {
3894         struct btrfs_root *root = BTRFS_I(fdentry(file)->d_inode)->root;
3895         struct btrfs_ioctl_quota_rescan_args *qsa;
3896         int ret;
3897
3898         if (!capable(CAP_SYS_ADMIN))
3899                 return -EPERM;
3900
3901         ret = mnt_want_write_file(file);
3902         if (ret)
3903                 return ret;
3904
3905         qsa = memdup_user(arg, sizeof(*qsa));
3906         if (IS_ERR(qsa)) {
3907                 ret = PTR_ERR(qsa);
3908                 goto drop_write;
3909         }
3910
3911         if (qsa->flags) {
3912                 ret = -EINVAL;
3913                 goto out;
3914         }
3915
3916         ret = btrfs_qgroup_rescan(root->fs_info);
3917
3918 out:
3919         kfree(qsa);
3920 drop_write:
3921         mnt_drop_write_file(file);
3922         return ret;
3923 }
3924
3925 static long btrfs_ioctl_quota_rescan_status(struct file *file, void __user *arg)
3926 {
3927         struct btrfs_root *root = BTRFS_I(fdentry(file)->d_inode)->root;
3928         struct btrfs_ioctl_quota_rescan_args *qsa;
3929         int ret = 0;
3930
3931         if (!capable(CAP_SYS_ADMIN))
3932                 return -EPERM;
3933
3934         qsa = kzalloc(sizeof(*qsa), GFP_NOFS);
3935         if (!qsa)
3936                 return -ENOMEM;
3937
3938         if (root->fs_info->qgroup_flags & BTRFS_QGROUP_STATUS_FLAG_RESCAN) {
3939                 qsa->flags = 1;
3940                 qsa->progress = root->fs_info->qgroup_rescan_progress.objectid;
3941         }
3942
3943         if (copy_to_user(arg, qsa, sizeof(*qsa)))
3944                 ret = -EFAULT;
3945
3946         kfree(qsa);
3947         return ret;
3948 }
3949
3950 static long btrfs_ioctl_set_received_subvol(struct file *file,
3951                                             void __user *arg)
3952 {
3953         struct btrfs_ioctl_received_subvol_args *sa = NULL;
3954         struct inode *inode = file_inode(file);
3955         struct btrfs_root *root = BTRFS_I(inode)->root;
3956         struct btrfs_root_item *root_item = &root->root_item;
3957         struct btrfs_trans_handle *trans;
3958         struct timespec ct = CURRENT_TIME;
3959         int ret = 0;
3960
3961         ret = mnt_want_write_file(file);
3962         if (ret < 0)
3963                 return ret;
3964
3965         down_write(&root->fs_info->subvol_sem);
3966
3967         if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) {
3968                 ret = -EINVAL;
3969                 goto out;
3970         }
3971
3972         if (btrfs_root_readonly(root)) {
3973                 ret = -EROFS;
3974                 goto out;
3975         }
3976
3977         if (!inode_owner_or_capable(inode)) {
3978                 ret = -EACCES;
3979                 goto out;
3980         }
3981
3982         sa = memdup_user(arg, sizeof(*sa));
3983         if (IS_ERR(sa)) {
3984                 ret = PTR_ERR(sa);
3985                 sa = NULL;
3986                 goto out;
3987         }
3988
3989         trans = btrfs_start_transaction(root, 1);
3990         if (IS_ERR(trans)) {
3991                 ret = PTR_ERR(trans);
3992                 trans = NULL;
3993                 goto out;
3994         }
3995
3996         sa->rtransid = trans->transid;
3997         sa->rtime.sec = ct.tv_sec;
3998         sa->rtime.nsec = ct.tv_nsec;
3999
4000         memcpy(root_item->received_uuid, sa->uuid, BTRFS_UUID_SIZE);
4001         btrfs_set_root_stransid(root_item, sa->stransid);
4002         btrfs_set_root_rtransid(root_item, sa->rtransid);
4003         root_item->stime.sec = cpu_to_le64(sa->stime.sec);
4004         root_item->stime.nsec = cpu_to_le32(sa->stime.nsec);
4005         root_item->rtime.sec = cpu_to_le64(sa->rtime.sec);
4006         root_item->rtime.nsec = cpu_to_le32(sa->rtime.nsec);
4007
4008         ret = btrfs_update_root(trans, root->fs_info->tree_root,
4009                                 &root->root_key, &root->root_item);
4010         if (ret < 0) {
4011                 btrfs_end_transaction(trans, root);
4012                 trans = NULL;
4013                 goto out;
4014         } else {
4015                 ret = btrfs_commit_transaction(trans, root);
4016                 if (ret < 0)
4017                         goto out;
4018         }
4019
4020         ret = copy_to_user(arg, sa, sizeof(*sa));
4021         if (ret)
4022                 ret = -EFAULT;
4023
4024 out:
4025         kfree(sa);
4026         up_write(&root->fs_info->subvol_sem);
4027         mnt_drop_write_file(file);
4028         return ret;
4029 }
4030
4031 static int btrfs_ioctl_get_fslabel(struct file *file, void __user *arg)
4032 {
4033         struct btrfs_root *root = BTRFS_I(fdentry(file)->d_inode)->root;
4034         const char *label = root->fs_info->super_copy->label;
4035         size_t len = strnlen(label, BTRFS_LABEL_SIZE);
4036         int ret;
4037
4038         if (len == BTRFS_LABEL_SIZE) {
4039                 pr_warn("btrfs: label is too long, return the first %zu bytes\n",
4040                         --len);
4041         }
4042
4043         mutex_lock(&root->fs_info->volume_mutex);
4044         ret = copy_to_user(arg, label, len);
4045         mutex_unlock(&root->fs_info->volume_mutex);
4046
4047         return ret ? -EFAULT : 0;
4048 }
4049
4050 static int btrfs_ioctl_set_fslabel(struct file *file, void __user *arg)
4051 {
4052         struct btrfs_root *root = BTRFS_I(fdentry(file)->d_inode)->root;
4053         struct btrfs_super_block *super_block = root->fs_info->super_copy;
4054         struct btrfs_trans_handle *trans;
4055         char label[BTRFS_LABEL_SIZE];
4056         int ret;
4057
4058         if (!capable(CAP_SYS_ADMIN))
4059                 return -EPERM;
4060
4061         if (copy_from_user(label, arg, sizeof(label)))
4062                 return -EFAULT;
4063
4064         if (strnlen(label, BTRFS_LABEL_SIZE) == BTRFS_LABEL_SIZE) {
4065                 pr_err("btrfs: unable to set label with more than %d bytes\n",
4066                        BTRFS_LABEL_SIZE - 1);
4067                 return -EINVAL;
4068         }
4069
4070         ret = mnt_want_write_file(file);
4071         if (ret)
4072                 return ret;
4073
4074         mutex_lock(&root->fs_info->volume_mutex);
4075         trans = btrfs_start_transaction(root, 0);
4076         if (IS_ERR(trans)) {
4077                 ret = PTR_ERR(trans);
4078                 goto out_unlock;
4079         }
4080
4081         strcpy(super_block->label, label);
4082         ret = btrfs_end_transaction(trans, root);
4083
4084 out_unlock:
4085         mutex_unlock(&root->fs_info->volume_mutex);
4086         mnt_drop_write_file(file);
4087         return ret;
4088 }
4089
4090 long btrfs_ioctl(struct file *file, unsigned int
4091                 cmd, unsigned long arg)
4092 {
4093         struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4094         void __user *argp = (void __user *)arg;
4095
4096         switch (cmd) {
4097         case FS_IOC_GETFLAGS:
4098                 return btrfs_ioctl_getflags(file, argp);
4099         case FS_IOC_SETFLAGS:
4100                 return btrfs_ioctl_setflags(file, argp);
4101         case FS_IOC_GETVERSION:
4102                 return btrfs_ioctl_getversion(file, argp);
4103         case FITRIM:
4104                 return btrfs_ioctl_fitrim(file, argp);
4105         case BTRFS_IOC_SNAP_CREATE:
4106                 return btrfs_ioctl_snap_create(file, argp, 0);
4107         case BTRFS_IOC_SNAP_CREATE_V2:
4108                 return btrfs_ioctl_snap_create_v2(file, argp, 0);
4109         case BTRFS_IOC_SUBVOL_CREATE:
4110                 return btrfs_ioctl_snap_create(file, argp, 1);
4111         case BTRFS_IOC_SUBVOL_CREATE_V2:
4112                 return btrfs_ioctl_snap_create_v2(file, argp, 1);
4113         case BTRFS_IOC_SNAP_DESTROY:
4114                 return btrfs_ioctl_snap_destroy(file, argp);
4115         case BTRFS_IOC_SUBVOL_GETFLAGS:
4116                 return btrfs_ioctl_subvol_getflags(file, argp);
4117         case BTRFS_IOC_SUBVOL_SETFLAGS:
4118                 return btrfs_ioctl_subvol_setflags(file, argp);
4119         case BTRFS_IOC_DEFAULT_SUBVOL:
4120                 return btrfs_ioctl_default_subvol(file, argp);
4121         case BTRFS_IOC_DEFRAG:
4122                 return btrfs_ioctl_defrag(file, NULL);
4123         case BTRFS_IOC_DEFRAG_RANGE:
4124                 return btrfs_ioctl_defrag(file, argp);
4125         case BTRFS_IOC_RESIZE:
4126                 return btrfs_ioctl_resize(file, argp);
4127         case BTRFS_IOC_ADD_DEV:
4128                 return btrfs_ioctl_add_dev(root, argp);
4129         case BTRFS_IOC_RM_DEV:
4130                 return btrfs_ioctl_rm_dev(file, argp);
4131         case BTRFS_IOC_FS_INFO:
4132                 return btrfs_ioctl_fs_info(root, argp);
4133         case BTRFS_IOC_DEV_INFO:
4134                 return btrfs_ioctl_dev_info(root, argp);
4135         case BTRFS_IOC_BALANCE:
4136                 return btrfs_ioctl_balance(file, NULL);
4137         case BTRFS_IOC_CLONE:
4138                 return btrfs_ioctl_clone(file, arg, 0, 0, 0);
4139         case BTRFS_IOC_CLONE_RANGE:
4140                 return btrfs_ioctl_clone_range(file, argp);
4141         case BTRFS_IOC_TRANS_START:
4142                 return btrfs_ioctl_trans_start(file);
4143         case BTRFS_IOC_TRANS_END:
4144                 return btrfs_ioctl_trans_end(file);
4145         case BTRFS_IOC_TREE_SEARCH:
4146                 return btrfs_ioctl_tree_search(file, argp);
4147         case BTRFS_IOC_INO_LOOKUP:
4148                 return btrfs_ioctl_ino_lookup(file, argp);
4149         case BTRFS_IOC_INO_PATHS:
4150                 return btrfs_ioctl_ino_to_path(root, argp);
4151         case BTRFS_IOC_LOGICAL_INO:
4152                 return btrfs_ioctl_logical_to_ino(root, argp);
4153         case BTRFS_IOC_SPACE_INFO:
4154                 return btrfs_ioctl_space_info(root, argp);
4155         case BTRFS_IOC_SYNC:
4156                 btrfs_sync_fs(file->f_dentry->d_sb, 1);
4157                 return 0;
4158         case BTRFS_IOC_START_SYNC:
4159                 return btrfs_ioctl_start_sync(root, argp);
4160         case BTRFS_IOC_WAIT_SYNC:
4161                 return btrfs_ioctl_wait_sync(root, argp);
4162         case BTRFS_IOC_SCRUB:
4163                 return btrfs_ioctl_scrub(file, argp);
4164         case BTRFS_IOC_SCRUB_CANCEL:
4165                 return btrfs_ioctl_scrub_cancel(root, argp);
4166         case BTRFS_IOC_SCRUB_PROGRESS:
4167                 return btrfs_ioctl_scrub_progress(root, argp);
4168         case BTRFS_IOC_BALANCE_V2:
4169                 return btrfs_ioctl_balance(file, argp);
4170         case BTRFS_IOC_BALANCE_CTL:
4171                 return btrfs_ioctl_balance_ctl(root, arg);
4172         case BTRFS_IOC_BALANCE_PROGRESS:
4173                 return btrfs_ioctl_balance_progress(root, argp);
4174         case BTRFS_IOC_SET_RECEIVED_SUBVOL:
4175                 return btrfs_ioctl_set_received_subvol(file, argp);
4176         case BTRFS_IOC_SEND:
4177                 return btrfs_ioctl_send(file, argp);
4178         case BTRFS_IOC_GET_DEV_STATS:
4179                 return btrfs_ioctl_get_dev_stats(root, argp);
4180         case BTRFS_IOC_QUOTA_CTL:
4181                 return btrfs_ioctl_quota_ctl(file, argp);
4182         case BTRFS_IOC_QGROUP_ASSIGN:
4183                 return btrfs_ioctl_qgroup_assign(file, argp);
4184         case BTRFS_IOC_QGROUP_CREATE:
4185                 return btrfs_ioctl_qgroup_create(file, argp);
4186         case BTRFS_IOC_QGROUP_LIMIT:
4187                 return btrfs_ioctl_qgroup_limit(file, argp);
4188         case BTRFS_IOC_QUOTA_RESCAN:
4189                 return btrfs_ioctl_quota_rescan(file, argp);
4190         case BTRFS_IOC_QUOTA_RESCAN_STATUS:
4191                 return btrfs_ioctl_quota_rescan_status(file, argp);
4192         case BTRFS_IOC_DEV_REPLACE:
4193                 return btrfs_ioctl_dev_replace(root, argp);
4194         case BTRFS_IOC_GET_FSLABEL:
4195                 return btrfs_ioctl_get_fslabel(file, argp);
4196         case BTRFS_IOC_SET_FSLABEL:
4197                 return btrfs_ioctl_set_fslabel(file, argp);
4198         }
4199
4200         return -ENOTTY;
4201 }