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