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 "compat.h"
  43 #include "ctree.h"
  44 #include "disk-io.h"
  45 #include "transaction.h"
  46 #include "btrfs_inode.h"
  47 #include "ioctl.h"
  48 #include "print-tree.h"
  49 #include "volumes.h"
  50 #include "locking.h"
  51 #include "ctree.h"
  52
  53 /* Mask out flags that are inappropriate for the given type of inode. */
  54 static inline __u32 btrfs_mask_flags(umode_t mode, __u32 flags)
  55 {
  56         if (S_ISDIR(mode))
  57                 return flags;
  58         else if (S_ISREG(mode))
  59                 return flags & ~FS_DIRSYNC_FL;
  60         else
  61                 return flags & (FS_NODUMP_FL | FS_NOATIME_FL);
  62 }
  63
  64 /*
  65  * Export inode flags to the format expected by the FS_IOC_GETFLAGS ioctl.
  66  */
  67 static unsigned int btrfs_flags_to_ioctl(unsigned int flags)
  68 {
  69         unsigned int iflags = 0;
  70
  71         if (flags & BTRFS_INODE_SYNC)
  72                 iflags |= FS_SYNC_FL;
  73         if (flags & BTRFS_INODE_IMMUTABLE)
  74                 iflags |= FS_IMMUTABLE_FL;
  75         if (flags & BTRFS_INODE_APPEND)
  76                 iflags |= FS_APPEND_FL;
  77         if (flags & BTRFS_INODE_NODUMP)
  78                 iflags |= FS_NODUMP_FL;
  79         if (flags & BTRFS_INODE_NOATIME)
  80                 iflags |= FS_NOATIME_FL;
  81         if (flags & BTRFS_INODE_DIRSYNC)
  82                 iflags |= FS_DIRSYNC_FL;
  83
  84         return iflags;
  85 }
  86
  87 /*
  88  * Update inode->i_flags based on the btrfs internal flags.
  89  */
  90 void btrfs_update_iflags(struct inode *inode)
  91 {
  92         struct btrfs_inode *ip = BTRFS_I(inode);
  93
  94         inode->i_flags &= ~(S_SYNC|S_APPEND|S_IMMUTABLE|S_NOATIME|S_DIRSYNC);
  95
  96         if (ip->flags & BTRFS_INODE_SYNC)
  97                 inode->i_flags |= S_SYNC;
  98         if (ip->flags & BTRFS_INODE_IMMUTABLE)
  99                 inode->i_flags |= S_IMMUTABLE;
 100         if (ip->flags & BTRFS_INODE_APPEND)
 101                 inode->i_flags |= S_APPEND;
 102         if (ip->flags & BTRFS_INODE_NOATIME)
 103                 inode->i_flags |= S_NOATIME;
 104         if (ip->flags & BTRFS_INODE_DIRSYNC)
 105                 inode->i_flags |= S_DIRSYNC;
 106 }
 107
 108 /*
 109  * Inherit flags from the parent inode.
 110  *
 111  * Unlike extN we don't have any flags we don't want to inherit currently.
 112  */
 113 void btrfs_inherit_iflags(struct inode *inode, struct inode *dir)
 114 {
 115         unsigned int flags;
 116
 117         if (!dir)
 118                 return;
 119
 120         flags = BTRFS_I(dir)->flags;
 121
 122         if (S_ISREG(inode->i_mode))
 123                 flags &= ~BTRFS_INODE_DIRSYNC;
 124         else if (!S_ISDIR(inode->i_mode))
 125                 flags &= (BTRFS_INODE_NODUMP | BTRFS_INODE_NOATIME);
 126
 127         BTRFS_I(inode)->flags = flags;
 128         btrfs_update_iflags(inode);
 129 }
 130
 131 static int btrfs_ioctl_getflags(struct file *file, void __user *arg)
 132 {
 133         struct btrfs_inode *ip = BTRFS_I(file->f_path.dentry->d_inode);
 134         unsigned int flags = btrfs_flags_to_ioctl(ip->flags);
 135
 136         if (copy_to_user(arg, &flags, sizeof(flags)))
 137                 return -EFAULT;
 138         return 0;
 139 }
 140
 141 static int btrfs_ioctl_setflags(struct file *file, void __user *arg)
 142 {
 143         struct inode *inode = file->f_path.dentry->d_inode;
 144         struct btrfs_inode *ip = BTRFS_I(inode);
 145         struct btrfs_root *root = ip->root;
 146         struct btrfs_trans_handle *trans;
 147         unsigned int flags, oldflags;
 148         int ret;
 149
 150         if (copy_from_user(&flags, arg, sizeof(flags)))
 151                 return -EFAULT;
 152
 153         if (flags & ~(FS_IMMUTABLE_FL | FS_APPEND_FL | \
 154                       FS_NOATIME_FL | FS_NODUMP_FL | \
 155                       FS_SYNC_FL | FS_DIRSYNC_FL))
 156                 return -EOPNOTSUPP;
 157
 158         if (!is_owner_or_cap(inode))
 159                 return -EACCES;
 160
 161         mutex_lock(&inode->i_mutex);
 162
 163         flags = btrfs_mask_flags(inode->i_mode, flags);
 164         oldflags = btrfs_flags_to_ioctl(ip->flags);
 165         if ((flags ^ oldflags) & (FS_APPEND_FL | FS_IMMUTABLE_FL)) {
 166                 if (!capable(CAP_LINUX_IMMUTABLE)) {
 167                         ret = -EPERM;
 168                         goto out_unlock;
 169                 }
 170         }
 171
 172         ret = mnt_want_write(file->f_path.mnt);
 173         if (ret)
 174                 goto out_unlock;
 175
 176         if (flags & FS_SYNC_FL)
 177                 ip->flags |= BTRFS_INODE_SYNC;
 178         else
 179                 ip->flags &= ~BTRFS_INODE_SYNC;
 180         if (flags & FS_IMMUTABLE_FL)
 181                 ip->flags |= BTRFS_INODE_IMMUTABLE;
 182         else
 183                 ip->flags &= ~BTRFS_INODE_IMMUTABLE;
 184         if (flags & FS_APPEND_FL)
 185                 ip->flags |= BTRFS_INODE_APPEND;
 186         else
 187                 ip->flags &= ~BTRFS_INODE_APPEND;
 188         if (flags & FS_NODUMP_FL)
 189                 ip->flags |= BTRFS_INODE_NODUMP;
 190         else
 191                 ip->flags &= ~BTRFS_INODE_NODUMP;
 192         if (flags & FS_NOATIME_FL)
 193                 ip->flags |= BTRFS_INODE_NOATIME;
 194         else
 195                 ip->flags &= ~BTRFS_INODE_NOATIME;
 196         if (flags & FS_DIRSYNC_FL)
 197                 ip->flags |= BTRFS_INODE_DIRSYNC;
 198         else
 199                 ip->flags &= ~BTRFS_INODE_DIRSYNC;
 200
 201
 202         trans = btrfs_join_transaction(root, 1);
 203         BUG_ON(!trans);
 204
 205         ret = btrfs_update_inode(trans, root, inode);
 206         BUG_ON(ret);
 207
 208         btrfs_update_iflags(inode);
 209         inode->i_ctime = CURRENT_TIME;
 210         btrfs_end_transaction(trans, root);
 211
 212         mnt_drop_write(file->f_path.mnt);
 213  out_unlock:
 214         mutex_unlock(&inode->i_mutex);
 215         return 0;
 216 }
 217
 218 static int btrfs_ioctl_getversion(struct file *file, int __user *arg)
 219 {
 220         struct inode *inode = file->f_path.dentry->d_inode;
 221
 222         return put_user(inode->i_generation, arg);
 223 }
 224
 225 static noinline int create_subvol(struct btrfs_root *root,
 226                                   struct dentry *dentry,
 227                                   char *name, int namelen)
 228 {
 229         struct btrfs_trans_handle *trans;
 230         struct btrfs_key key;
 231         struct btrfs_root_item root_item;
 232         struct btrfs_inode_item *inode_item;
 233         struct extent_buffer *leaf;
 234         struct btrfs_root *new_root;
 235         struct inode *dir = dentry->d_parent->d_inode;
 236         int ret;
 237         int err;
 238         u64 objectid;
 239         u64 new_dirid = BTRFS_FIRST_FREE_OBJECTID;
 240         u64 index = 0;
 241
 242         /*
 243          * 1 - inode item
 244          * 2 - refs
 245          * 1 - root item
 246          * 2 - dir items
 247          */
 248         ret = btrfs_reserve_metadata_space(root, 6);
 249         if (ret)
 250                 return ret;
 251
 252         trans = btrfs_start_transaction(root, 1);
 253         BUG_ON(!trans);
 254
 255         ret = btrfs_find_free_objectid(trans, root->fs_info->tree_root,
 256                                        0, &objectid);
 257         if (ret)
 258                 goto fail;
 259
 260         leaf = btrfs_alloc_free_block(trans, root, root->leafsize,
 261                                       0, objectid, NULL, 0, 0, 0);
 262         if (IS_ERR(leaf)) {
 263                 ret = PTR_ERR(leaf);
 264                 goto fail;
 265         }
 266
 267         memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header));
 268         btrfs_set_header_bytenr(leaf, leaf->start);
 269         btrfs_set_header_generation(leaf, trans->transid);
 270         btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV);
 271         btrfs_set_header_owner(leaf, objectid);
 272
 273         write_extent_buffer(leaf, root->fs_info->fsid,
 274                             (unsigned long)btrfs_header_fsid(leaf),
 275                             BTRFS_FSID_SIZE);
 276         write_extent_buffer(leaf, root->fs_info->chunk_tree_uuid,
 277                             (unsigned long)btrfs_header_chunk_tree_uuid(leaf),
 278                             BTRFS_UUID_SIZE);
 279         btrfs_mark_buffer_dirty(leaf);
 280
 281         inode_item = &root_item.inode;
 282         memset(inode_item, 0, sizeof(*inode_item));
 283         inode_item->generation = cpu_to_le64(1);
 284         inode_item->size = cpu_to_le64(3);
 285         inode_item->nlink = cpu_to_le32(1);
 286         inode_item->nbytes = cpu_to_le64(root->leafsize);
 287         inode_item->mode = cpu_to_le32(S_IFDIR | 0755);
 288
 289         btrfs_set_root_bytenr(&root_item, leaf->start);
 290         btrfs_set_root_generation(&root_item, trans->transid);
 291         btrfs_set_root_level(&root_item, 0);
 292         btrfs_set_root_refs(&root_item, 1);
 293         btrfs_set_root_used(&root_item, leaf->len);
 294         btrfs_set_root_last_snapshot(&root_item, 0);
 295
 296         memset(&root_item.drop_progress, 0, sizeof(root_item.drop_progress));
 297         root_item.drop_level = 0;
 298
 299         btrfs_tree_unlock(leaf);
 300         free_extent_buffer(leaf);
 301         leaf = NULL;
 302
 303         btrfs_set_root_dirid(&root_item, new_dirid);
 304
 305         key.objectid = objectid;
 306         key.offset = 0;
 307         btrfs_set_key_type(&key, BTRFS_ROOT_ITEM_KEY);
 308         ret = btrfs_insert_root(trans, root->fs_info->tree_root, &key,
 309                                 &root_item);
 310         if (ret)
 311                 goto fail;
 312
 313         key.offset = (u64)-1;
 314         new_root = btrfs_read_fs_root_no_name(root->fs_info, &key);
 315         BUG_ON(IS_ERR(new_root));
 316
 317         btrfs_record_root_in_trans(trans, new_root);
 318
 319         ret = btrfs_create_subvol_root(trans, new_root, new_dirid,
 320                                        BTRFS_I(dir)->block_group);
 321         /*
 322          * insert the directory item
 323          */
 324         ret = btrfs_set_inode_index(dir, &index);
 325         BUG_ON(ret);
 326
 327         ret = btrfs_insert_dir_item(trans, root,
 328                                     name, namelen, dir->i_ino, &key,
 329                                     BTRFS_FT_DIR, index);
 330         if (ret)
 331                 goto fail;
 332
 333         btrfs_i_size_write(dir, dir->i_size + namelen * 2);
 334         ret = btrfs_update_inode(trans, root, dir);
 335         BUG_ON(ret);
 336
 337         ret = btrfs_add_root_ref(trans, root->fs_info->tree_root,
 338                                  objectid, root->root_key.objectid,
 339                                  dir->i_ino, index, name, namelen);
 340
 341         BUG_ON(ret);
 342
 343         d_instantiate(dentry, btrfs_lookup_dentry(dir, dentry));
 344 fail:
 345         err = btrfs_commit_transaction(trans, root);
 346         if (err && !ret)
 347                 ret = err;
 348
 349         btrfs_unreserve_metadata_space(root, 6);
 350         return ret;
 351 }
 352
 353 static int create_snapshot(struct btrfs_root *root, struct dentry *dentry,
 354                            char *name, int namelen)
 355 {
 356         struct inode *inode;
 357         struct btrfs_pending_snapshot *pending_snapshot;
 358         struct btrfs_trans_handle *trans;
 359         int ret;
 360
 361         if (!root->ref_cows)
 362                 return -EINVAL;
 363
 364         /*
 365          * 1 - inode item
 366          * 2 - refs
 367          * 1 - root item
 368          * 2 - dir items
 369          */
 370         ret = btrfs_reserve_metadata_space(root, 6);
 371         if (ret)
 372                 goto fail;
 373
 374         pending_snapshot = kzalloc(sizeof(*pending_snapshot), GFP_NOFS);
 375         if (!pending_snapshot) {
 376                 ret = -ENOMEM;
 377                 btrfs_unreserve_metadata_space(root, 6);
 378                 goto fail;
 379         }
 380         pending_snapshot->name = kmalloc(namelen + 1, GFP_NOFS);
 381         if (!pending_snapshot->name) {
 382                 ret = -ENOMEM;
 383                 kfree(pending_snapshot);
 384                 btrfs_unreserve_metadata_space(root, 6);
 385                 goto fail;
 386         }
 387         memcpy(pending_snapshot->name, name, namelen);
 388         pending_snapshot->name[namelen] = '\0';
 389         pending_snapshot->dentry = dentry;
 390         trans = btrfs_start_transaction(root, 1);
 391         BUG_ON(!trans);
 392         pending_snapshot->root = root;
 393         list_add(&pending_snapshot->list,
 394                  &trans->transaction->pending_snapshots);
 395         ret = btrfs_commit_transaction(trans, root);
 396         BUG_ON(ret);
 397         btrfs_unreserve_metadata_space(root, 6);
 398
 399         inode = btrfs_lookup_dentry(dentry->d_parent->d_inode, dentry);
 400         if (IS_ERR(inode)) {
 401                 ret = PTR_ERR(inode);
 402                 goto fail;
 403         }
 404         BUG_ON(!inode);
 405         d_instantiate(dentry, inode);
 406         ret = 0;
 407 fail:
 408         return ret;
 409 }
 410
 411 /* copy of may_create in fs/namei.c() */
 412 static inline int btrfs_may_create(struct inode *dir, struct dentry *child)
 413 {
 414         if (child->d_inode)
 415                 return -EEXIST;
 416         if (IS_DEADDIR(dir))
 417                 return -ENOENT;
 418         return inode_permission(dir, MAY_WRITE | MAY_EXEC);
 419 }
 420
 421 /*
 422  * Create a new subvolume below @parent.  This is largely modeled after
 423  * sys_mkdirat and vfs_mkdir, but we only do a single component lookup
 424  * inside this filesystem so it's quite a bit simpler.
 425  */
 426 static noinline int btrfs_mksubvol(struct path *parent,
 427                                    char *name, int namelen,
 428                                    struct btrfs_root *snap_src)
 429 {
 430         struct inode *dir  = parent->dentry->d_inode;
 431         struct dentry *dentry;
 432         int error;
 433
 434         mutex_lock_nested(&dir->i_mutex, I_MUTEX_PARENT);
 435
 436         dentry = lookup_one_len(name, parent->dentry, namelen);
 437         error = PTR_ERR(dentry);
 438         if (IS_ERR(dentry))
 439                 goto out_unlock;
 440
 441         error = -EEXIST;
 442         if (dentry->d_inode)
 443                 goto out_dput;
 444
 445         error = mnt_want_write(parent->mnt);
 446         if (error)
 447                 goto out_dput;
 448
 449         error = btrfs_may_create(dir, dentry);
 450         if (error)
 451                 goto out_drop_write;
 452
 453         down_read(&BTRFS_I(dir)->root->fs_info->subvol_sem);
 454
 455         if (btrfs_root_refs(&BTRFS_I(dir)->root->root_item) == 0)
 456                 goto out_up_read;
 457
 458         if (snap_src) {
 459                 error = create_snapshot(snap_src, dentry,
 460                                         name, namelen);
 461         } else {
 462                 error = create_subvol(BTRFS_I(dir)->root, dentry,
 463                                       name, namelen);
 464         }
 465         if (!error)
 466                 fsnotify_mkdir(dir, dentry);
 467 out_up_read:
 468         up_read(&BTRFS_I(dir)->root->fs_info->subvol_sem);
 469 out_drop_write:
 470         mnt_drop_write(parent->mnt);
 471 out_dput:
 472         dput(dentry);
 473 out_unlock:
 474         mutex_unlock(&dir->i_mutex);
 475         return error;
 476 }
 477
 478 static int btrfs_defrag_file(struct file *file)
 479 {
 480         struct inode *inode = fdentry(file)->d_inode;
 481         struct btrfs_root *root = BTRFS_I(inode)->root;
 482         struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
 483         struct btrfs_ordered_extent *ordered;
 484         struct page *page;
 485         unsigned long last_index;
 486         unsigned long ra_pages = root->fs_info->bdi.ra_pages;
 487         unsigned long total_read = 0;
 488         u64 page_start;
 489         u64 page_end;
 490         unsigned long i;
 491         int ret;
 492
 493         ret = btrfs_check_data_free_space(root, inode, inode->i_size);
 494         if (ret)
 495                 return -ENOSPC;
 496
 497         mutex_lock(&inode->i_mutex);
 498         last_index = inode->i_size >> PAGE_CACHE_SHIFT;
 499         for (i = 0; i <= last_index; i++) {
 500                 if (total_read % ra_pages == 0) {
 501                         btrfs_force_ra(inode->i_mapping, &file->f_ra, file, i,
 502                                        min(last_index, i + ra_pages - 1));
 503                 }
 504                 total_read++;
 505 again:
 506                 page = grab_cache_page(inode->i_mapping, i);
 507                 if (!page)
 508                         goto out_unlock;
 509                 if (!PageUptodate(page)) {
 510                         btrfs_readpage(NULL, page);
 511                         lock_page(page);
 512                         if (!PageUptodate(page)) {
 513                                 unlock_page(page);
 514                                 page_cache_release(page);
 515                                 goto out_unlock;
 516                         }
 517                 }
 518
 519                 wait_on_page_writeback(page);
 520
 521                 page_start = (u64)page->index << PAGE_CACHE_SHIFT;
 522                 page_end = page_start + PAGE_CACHE_SIZE - 1;
 523                 lock_extent(io_tree, page_start, page_end, GFP_NOFS);
 524
 525                 ordered = btrfs_lookup_ordered_extent(inode, page_start);
 526                 if (ordered) {
 527                         unlock_extent(io_tree, page_start, page_end, GFP_NOFS);
 528                         unlock_page(page);
 529                         page_cache_release(page);
 530                         btrfs_start_ordered_extent(inode, ordered, 1);
 531                         btrfs_put_ordered_extent(ordered);
 532                         goto again;
 533                 }
 534                 set_page_extent_mapped(page);
 535
 536                 /*
 537                  * this makes sure page_mkwrite is called on the
 538                  * page if it is dirtied again later
 539                  */
 540                 clear_page_dirty_for_io(page);
 541
 542                 btrfs_set_extent_delalloc(inode, page_start, page_end);
 543                 set_page_dirty(page);
 544                 unlock_extent(io_tree, page_start, page_end, GFP_NOFS);
 545                 unlock_page(page);
 546                 page_cache_release(page);
 547                 balance_dirty_pages_ratelimited_nr(inode->i_mapping, 1);
 548         }
 549
 550 out_unlock:
 551         mutex_unlock(&inode->i_mutex);
 552         return 0;
 553 }
 554
 555 static noinline int btrfs_ioctl_resize(struct btrfs_root *root,
 556                                         void __user *arg)
 557 {
 558         u64 new_size;
 559         u64 old_size;
 560         u64 devid = 1;
 561         struct btrfs_ioctl_vol_args *vol_args;
 562         struct btrfs_trans_handle *trans;
 563         struct btrfs_device *device = NULL;
 564         char *sizestr;
 565         char *devstr = NULL;
 566         int ret = 0;
 567         int namelen;
 568         int mod = 0;
 569
 570         if (root->fs_info->sb->s_flags & MS_RDONLY)
 571                 return -EROFS;
 572
 573         if (!capable(CAP_SYS_ADMIN))
 574                 return -EPERM;
 575
 576         vol_args = memdup_user(arg, sizeof(*vol_args));
 577         if (IS_ERR(vol_args))
 578                 return PTR_ERR(vol_args);
 579
 580         vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
 581         namelen = strlen(vol_args->name);
 582
 583         mutex_lock(&root->fs_info->volume_mutex);
 584         sizestr = vol_args->name;
 585         devstr = strchr(sizestr, ':');
 586         if (devstr) {
 587                 char *end;
 588                 sizestr = devstr + 1;
 589                 *devstr = '\0';
 590                 devstr = vol_args->name;
 591                 devid = simple_strtoull(devstr, &end, 10);
 592                 printk(KERN_INFO "resizing devid %llu\n",
 593                        (unsigned long long)devid);
 594         }
 595         device = btrfs_find_device(root, devid, NULL, NULL);
 596         if (!device) {
 597                 printk(KERN_INFO "resizer unable to find device %llu\n",
 598                        (unsigned long long)devid);
 599                 ret = -EINVAL;
 600                 goto out_unlock;
 601         }
 602         if (!strcmp(sizestr, "max"))
 603                 new_size = device->bdev->bd_inode->i_size;
 604         else {
 605                 if (sizestr[0] == '-') {
 606                         mod = -1;
 607                         sizestr++;
 608                 } else if (sizestr[0] == '+') {
 609                         mod = 1;
 610                         sizestr++;
 611                 }
 612                 new_size = btrfs_parse_size(sizestr);
 613                 if (new_size == 0) {
 614                         ret = -EINVAL;
 615                         goto out_unlock;
 616                 }
 617         }
 618
 619         old_size = device->total_bytes;
 620
 621         if (mod < 0) {
 622                 if (new_size > old_size) {
 623                         ret = -EINVAL;
 624                         goto out_unlock;
 625                 }
 626                 new_size = old_size - new_size;
 627         } else if (mod > 0) {
 628                 new_size = old_size + new_size;
 629         }
 630
 631         if (new_size < 256 * 1024 * 1024) {
 632                 ret = -EINVAL;
 633                 goto out_unlock;
 634         }
 635         if (new_size > device->bdev->bd_inode->i_size) {
 636                 ret = -EFBIG;
 637                 goto out_unlock;
 638         }
 639
 640         do_div(new_size, root->sectorsize);
 641         new_size *= root->sectorsize;
 642
 643         printk(KERN_INFO "new size for %s is %llu\n",
 644                 device->name, (unsigned long long)new_size);
 645
 646         if (new_size > old_size) {
 647                 trans = btrfs_start_transaction(root, 1);
 648                 ret = btrfs_grow_device(trans, device, new_size);
 649                 btrfs_commit_transaction(trans, root);
 650         } else {
 651                 ret = btrfs_shrink_device(device, new_size);
 652         }
 653
 654 out_unlock:
 655         mutex_unlock(&root->fs_info->volume_mutex);
 656         kfree(vol_args);
 657         return ret;
 658 }
 659
 660 static noinline int btrfs_ioctl_snap_create(struct file *file,
 661                                             void __user *arg, int subvol)
 662 {
 663         struct btrfs_root *root = BTRFS_I(fdentry(file)->d_inode)->root;
 664         struct btrfs_ioctl_vol_args *vol_args;
 665         struct file *src_file;
 666         int namelen;
 667         int ret = 0;
 668
 669         if (root->fs_info->sb->s_flags & MS_RDONLY)
 670                 return -EROFS;
 671
 672         vol_args = memdup_user(arg, sizeof(*vol_args));
 673         if (IS_ERR(vol_args))
 674                 return PTR_ERR(vol_args);
 675
 676         vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
 677         namelen = strlen(vol_args->name);
 678         if (strchr(vol_args->name, '/')) {
 679                 ret = -EINVAL;
 680                 goto out;
 681         }
 682
 683         if (subvol) {
 684                 ret = btrfs_mksubvol(&file->f_path, vol_args->name, namelen,
 685                                      NULL);
 686         } else {
 687                 struct inode *src_inode;
 688                 src_file = fget(vol_args->fd);
 689                 if (!src_file) {
 690                         ret = -EINVAL;
 691                         goto out;
 692                 }
 693
 694                 src_inode = src_file->f_path.dentry->d_inode;
 695                 if (src_inode->i_sb != file->f_path.dentry->d_inode->i_sb) {
 696                         printk(KERN_INFO "btrfs: Snapshot src from "
 697                                "another FS\n");
 698                         ret = -EINVAL;
 699                         fput(src_file);
 700                         goto out;
 701                 }
 702                 ret = btrfs_mksubvol(&file->f_path, vol_args->name, namelen,
 703                                      BTRFS_I(src_inode)->root);
 704                 fput(src_file);
 705         }
 706 out:
 707         kfree(vol_args);
 708         return ret;
 709 }
 710
 711 /*
 712  * helper to check if the subvolume references other subvolumes
 713  */
 714 static noinline int may_destroy_subvol(struct btrfs_root *root)
 715 {
 716         struct btrfs_path *path;
 717         struct btrfs_key key;
 718         int ret;
 719
 720         path = btrfs_alloc_path();
 721         if (!path)
 722                 return -ENOMEM;
 723
 724         key.objectid = root->root_key.objectid;
 725         key.type = BTRFS_ROOT_REF_KEY;
 726         key.offset = (u64)-1;
 727
 728         ret = btrfs_search_slot(NULL, root->fs_info->tree_root,
 729                                 &key, path, 0, 0);
 730         if (ret < 0)
 731                 goto out;
 732         BUG_ON(ret == 0);
 733
 734         ret = 0;
 735         if (path->slots[0] > 0) {
 736                 path->slots[0]--;
 737                 btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
 738                 if (key.objectid == root->root_key.objectid &&
 739                     key.type == BTRFS_ROOT_REF_KEY)
 740                         ret = -ENOTEMPTY;
 741         }
 742 out:
 743         btrfs_free_path(path);
 744         return ret;
 745 }
 746
 747 static noinline int key_in_sk(struct btrfs_key *key,
 748                               struct btrfs_ioctl_search_key *sk)
 749 {
 750         if (key->objectid < sk->min_objectid)
 751                 return 0;
 752         if (key->offset < sk->min_offset)
 753                 return 0;
 754         if (key->type < sk->min_type)
 755                 return 0;
 756         if (key->objectid > sk->max_objectid)
 757                 return 0;
 758         if (key->type > sk->max_type)
 759                 return 0;
 760         if (key->offset > sk->max_offset)
 761                 return 0;
 762         return 1;
 763 }
 764
 765 static noinline int copy_to_sk(struct btrfs_root *root,
 766                                struct btrfs_path *path,
 767                                struct btrfs_key *key,
 768                                struct btrfs_ioctl_search_key *sk,
 769                                char *buf,
 770                                unsigned long *sk_offset,
 771                                int *num_found)
 772 {
 773         u64 found_transid;
 774         struct extent_buffer *leaf;
 775         struct btrfs_ioctl_search_header sh;
 776         unsigned long item_off;
 777         unsigned long item_len;
 778         int nritems;
 779         int i;
 780         int slot;
 781         int found = 0;
 782         int ret = 0;
 783
 784         leaf = path->nodes[0];
 785         slot = path->slots[0];
 786         nritems = btrfs_header_nritems(leaf);
 787
 788         if (btrfs_header_generation(leaf) > sk->max_transid) {
 789                 i = nritems;
 790                 goto advance_key;
 791         }
 792         found_transid = btrfs_header_generation(leaf);
 793
 794         for (i = slot; i < nritems; i++) {
 795                 item_off = btrfs_item_ptr_offset(leaf, i);
 796                 item_len = btrfs_item_size_nr(leaf, i);
 797
 798                 if (item_len > BTRFS_SEARCH_ARGS_BUFSIZE)
 799                         item_len = 0;
 800
 801                 if (sizeof(sh) + item_len + *sk_offset >
 802                     BTRFS_SEARCH_ARGS_BUFSIZE) {
 803                         ret = 1;
 804                         goto overflow;
 805                 }
 806
 807                 btrfs_item_key_to_cpu(leaf, key, i);
 808                 if (!key_in_sk(key, sk))
 809                         continue;
 810
 811                 sh.objectid = key->objectid;
 812                 sh.offset = key->offset;
 813                 sh.type = key->type;
 814                 sh.len = item_len;
 815                 sh.transid = found_transid;
 816
 817                 /* copy search result header */
 818                 memcpy(buf + *sk_offset, &sh, sizeof(sh));
 819                 *sk_offset += sizeof(sh);
 820
 821                 if (item_len) {
 822                         char *p = buf + *sk_offset;
 823                         /* copy the item */
 824                         read_extent_buffer(leaf, p,
 825                                            item_off, item_len);
 826                         *sk_offset += item_len;
 827                         found++;
 828                 }
 829
 830                 if (*num_found >= sk->nr_items)
 831                         break;
 832         }
 833 advance_key:
 834         if (key->offset < (u64)-1)
 835                 key->offset++;
 836         else if (key->type < (u64)-1)
 837                 key->type++;
 838         else if (key->objectid < (u64)-1)
 839                 key->objectid++;
 840         ret = 0;
 841 overflow:
 842         *num_found += found;
 843         return ret;
 844 }
 845
 846 static noinline int search_ioctl(struct inode *inode,
 847                                  struct btrfs_ioctl_search_args *args)
 848 {
 849         struct btrfs_root *root;
 850         struct btrfs_key key;
 851         struct btrfs_key max_key;
 852         struct btrfs_path *path;
 853         struct btrfs_ioctl_search_key *sk = &args->key;
 854         struct btrfs_fs_info *info = BTRFS_I(inode)->root->fs_info;
 855         int ret;
 856         int num_found = 0;
 857         unsigned long sk_offset = 0;
 858
 859         path = btrfs_alloc_path();
 860         if (!path)
 861                 return -ENOMEM;
 862
 863         if (sk->tree_id == 0) {
 864                 /* search the root of the inode that was passed */
 865                 root = BTRFS_I(inode)->root;
 866         } else {
 867                 key.objectid = sk->tree_id;
 868                 key.type = BTRFS_ROOT_ITEM_KEY;
 869                 key.offset = (u64)-1;
 870                 root = btrfs_read_fs_root_no_name(info, &key);
 871                 if (IS_ERR(root)) {
 872                         printk(KERN_ERR "could not find root %llu\n",
 873                                sk->tree_id);
 874                         btrfs_free_path(path);
 875                         return -ENOENT;
 876                 }
 877         }
 878
 879         key.objectid = sk->min_objectid;
 880         key.type = sk->min_type;
 881         key.offset = sk->min_offset;
 882
 883         max_key.objectid = sk->max_objectid;
 884         max_key.type = sk->max_type;
 885         max_key.offset = sk->max_offset;
 886
 887         path->keep_locks = 1;
 888
 889         while(1) {
 890                 ret = btrfs_search_forward(root, &key, &max_key, path, 0,
 891                                            sk->min_transid);
 892                 if (ret != 0) {
 893                         if (ret > 0)
 894                                 ret = 0;
 895                         goto err;
 896                 }
 897                 ret = copy_to_sk(root, path, &key, sk, args->buf,
 898                                  &sk_offset, &num_found);
 899                 btrfs_release_path(root, path);
 900                 if (ret || num_found >= sk->nr_items)
 901                         break;
 902
 903         }
 904         ret = 0;
 905 err:
 906         sk->nr_items = num_found;
 907         btrfs_free_path(path);
 908         return ret;
 909 }
 910
 911 static noinline int btrfs_ioctl_tree_search(struct file *file,
 912                                            void __user *argp)
 913 {
 914          struct btrfs_ioctl_search_args *args;
 915          struct inode *inode;
 916          int ret;
 917
 918         if (!capable(CAP_SYS_ADMIN))
 919                 return -EPERM;
 920
 921         args = kmalloc(sizeof(*args), GFP_KERNEL);
 922         if (!args)
 923                 return -ENOMEM;
 924
 925         if (copy_from_user(args, argp, sizeof(*args))) {
 926                 kfree(args);
 927                 return -EFAULT;
 928         }
 929         inode = fdentry(file)->d_inode;
 930         ret = search_ioctl(inode, args);
 931         if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
 932                 ret = -EFAULT;
 933         kfree(args);
 934         return ret;
 935 }
 936
 937 /*
 938  * Search INODE_REFs to identify path name of 'dirid' directory
 939  * in a 'tree_id' tree. and sets path name to 'name'.
 940  */
 941 static noinline int btrfs_search_path_in_tree(struct btrfs_fs_info *info,
 942                                 u64 tree_id, u64 dirid, char *name)
 943 {
 944         struct btrfs_root *root;
 945         struct btrfs_key key;
 946         char *ptr;
 947         int ret = -1;
 948         int slot;
 949         int len;
 950         int total_len = 0;
 951         struct btrfs_inode_ref *iref;
 952         struct extent_buffer *l;
 953         struct btrfs_path *path;
 954
 955         if (dirid == BTRFS_FIRST_FREE_OBJECTID) {
 956                 name[0]='\0';
 957                 return 0;
 958         }
 959
 960         path = btrfs_alloc_path();
 961         if (!path)
 962                 return -ENOMEM;
 963
 964         ptr = &name[BTRFS_INO_LOOKUP_PATH_MAX];
 965
 966         key.objectid = tree_id;
 967         key.type = BTRFS_ROOT_ITEM_KEY;
 968         key.offset = (u64)-1;
 969         root = btrfs_read_fs_root_no_name(info, &key);
 970         if (IS_ERR(root)) {
 971                 printk(KERN_ERR "could not find root %llu\n", tree_id);
 972                 return -ENOENT;
 973         }
 974
 975         key.objectid = dirid;
 976         key.type = BTRFS_INODE_REF_KEY;
 977         key.offset = 0;
 978
 979         while(1) {
 980                 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
 981                 if (ret < 0)
 982                         goto out;
 983
 984                 l = path->nodes[0];
 985                 slot = path->slots[0];
 986                 btrfs_item_key_to_cpu(l, &key, slot);
 987
 988                 if (ret > 0 && (key.objectid != dirid ||
 989                                 key.type != BTRFS_INODE_REF_KEY)) {
 990                         ret = -ENOENT;
 991                         goto out;
 992                 }
 993
 994                 iref = btrfs_item_ptr(l, slot, struct btrfs_inode_ref);
 995                 len = btrfs_inode_ref_name_len(l, iref);
 996                 ptr -= len + 1;
 997                 total_len += len + 1;
 998                 if (ptr < name)
 999                         goto out;
1000
1001                 *(ptr + len) = '/';
1002                 read_extent_buffer(l, ptr,(unsigned long)(iref + 1), len);
1003
1004                 if (key.offset == BTRFS_FIRST_FREE_OBJECTID)
1005                         break;
1006
1007                 btrfs_release_path(root, path);
1008                 key.objectid = key.offset;
1009                 key.offset = 0;
1010                 dirid = key.objectid;
1011
1012         }
1013         if (ptr < name)
1014                 goto out;
1015         memcpy(name, ptr, total_len);
1016         name[total_len]='\0';
1017         ret = 0;
1018 out:
1019         btrfs_free_path(path);
1020         return ret;
1021 }
1022
1023 static noinline int btrfs_ioctl_ino_lookup(struct file *file,
1024                                            void __user *argp)
1025 {
1026          struct btrfs_ioctl_ino_lookup_args *args;
1027          struct inode *inode;
1028          int ret;
1029
1030         if (!capable(CAP_SYS_ADMIN))
1031                 return -EPERM;
1032
1033         args = kmalloc(sizeof(*args), GFP_KERNEL);
1034         if (copy_from_user(args, argp, sizeof(*args))) {
1035                 kfree(args);
1036                 return -EFAULT;
1037         }
1038         inode = fdentry(file)->d_inode;
1039
1040         ret = btrfs_search_path_in_tree(BTRFS_I(inode)->root->fs_info,
1041                                         args->treeid, args->objectid,
1042                                         args->name);
1043
1044         if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
1045                 ret = -EFAULT;
1046
1047         kfree(args);
1048         return ret;
1049 }
1050
1051 static noinline int btrfs_ioctl_snap_destroy(struct file *file,
1052                                              void __user *arg)
1053 {
1054         struct dentry *parent = fdentry(file);
1055         struct dentry *dentry;
1056         struct inode *dir = parent->d_inode;
1057         struct inode *inode;
1058         struct btrfs_root *root = BTRFS_I(dir)->root;
1059         struct btrfs_root *dest = NULL;
1060         struct btrfs_ioctl_vol_args *vol_args;
1061         struct btrfs_trans_handle *trans;
1062         int namelen;
1063         int ret;
1064         int err = 0;
1065
1066         if (!capable(CAP_SYS_ADMIN))
1067                 return -EPERM;
1068
1069         vol_args = memdup_user(arg, sizeof(*vol_args));
1070         if (IS_ERR(vol_args))
1071                 return PTR_ERR(vol_args);
1072
1073         vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1074         namelen = strlen(vol_args->name);
1075         if (strchr(vol_args->name, '/') ||
1076             strncmp(vol_args->name, "..", namelen) == 0) {
1077                 err = -EINVAL;
1078                 goto out;
1079         }
1080
1081         err = mnt_want_write(file->f_path.mnt);
1082         if (err)
1083                 goto out;
1084
1085         mutex_lock_nested(&dir->i_mutex, I_MUTEX_PARENT);
1086         dentry = lookup_one_len(vol_args->name, parent, namelen);
1087         if (IS_ERR(dentry)) {
1088                 err = PTR_ERR(dentry);
1089                 goto out_unlock_dir;
1090         }
1091
1092         if (!dentry->d_inode) {
1093                 err = -ENOENT;
1094                 goto out_dput;
1095         }
1096
1097         inode = dentry->d_inode;
1098         if (inode->i_ino != BTRFS_FIRST_FREE_OBJECTID) {
1099                 err = -EINVAL;
1100                 goto out_dput;
1101         }
1102
1103         dest = BTRFS_I(inode)->root;
1104
1105         mutex_lock(&inode->i_mutex);
1106         err = d_invalidate(dentry);
1107         if (err)
1108                 goto out_unlock;
1109
1110         down_write(&root->fs_info->subvol_sem);
1111
1112         err = may_destroy_subvol(dest);
1113         if (err)
1114                 goto out_up_write;
1115
1116         trans = btrfs_start_transaction(root, 1);
1117         ret = btrfs_unlink_subvol(trans, root, dir,
1118                                 dest->root_key.objectid,
1119                                 dentry->d_name.name,
1120                                 dentry->d_name.len);
1121         BUG_ON(ret);
1122
1123         btrfs_record_root_in_trans(trans, dest);
1124
1125         memset(&dest->root_item.drop_progress, 0,
1126                 sizeof(dest->root_item.drop_progress));
1127         dest->root_item.drop_level = 0;
1128         btrfs_set_root_refs(&dest->root_item, 0);
1129
1130         ret = btrfs_insert_orphan_item(trans,
1131                                 root->fs_info->tree_root,
1132                                 dest->root_key.objectid);
1133         BUG_ON(ret);
1134
1135         ret = btrfs_commit_transaction(trans, root);
1136         BUG_ON(ret);
1137         inode->i_flags |= S_DEAD;
1138 out_up_write:
1139         up_write(&root->fs_info->subvol_sem);
1140 out_unlock:
1141         mutex_unlock(&inode->i_mutex);
1142         if (!err) {
1143                 shrink_dcache_sb(root->fs_info->sb);
1144                 btrfs_invalidate_inodes(dest);
1145                 d_delete(dentry);
1146         }
1147 out_dput:
1148         dput(dentry);
1149 out_unlock_dir:
1150         mutex_unlock(&dir->i_mutex);
1151         mnt_drop_write(file->f_path.mnt);
1152 out:
1153         kfree(vol_args);
1154         return err;
1155 }
1156
1157 static int btrfs_ioctl_defrag(struct file *file)
1158 {
1159         struct inode *inode = fdentry(file)->d_inode;
1160         struct btrfs_root *root = BTRFS_I(inode)->root;
1161         int ret;
1162
1163         ret = mnt_want_write(file->f_path.mnt);
1164         if (ret)
1165                 return ret;
1166
1167         switch (inode->i_mode & S_IFMT) {
1168         case S_IFDIR:
1169                 if (!capable(CAP_SYS_ADMIN)) {
1170                         ret = -EPERM;
1171                         goto out;
1172                 }
1173                 btrfs_defrag_root(root, 0);
1174                 btrfs_defrag_root(root->fs_info->extent_root, 0);
1175                 break;
1176         case S_IFREG:
1177                 if (!(file->f_mode & FMODE_WRITE)) {
1178                         ret = -EINVAL;
1179                         goto out;
1180                 }
1181                 btrfs_defrag_file(file);
1182                 break;
1183         }
1184 out:
1185         mnt_drop_write(file->f_path.mnt);
1186         return ret;
1187 }
1188
1189 static long btrfs_ioctl_add_dev(struct btrfs_root *root, void __user *arg)
1190 {
1191         struct btrfs_ioctl_vol_args *vol_args;
1192         int ret;
1193
1194         if (!capable(CAP_SYS_ADMIN))
1195                 return -EPERM;
1196
1197         vol_args = memdup_user(arg, sizeof(*vol_args));
1198         if (IS_ERR(vol_args))
1199                 return PTR_ERR(vol_args);
1200
1201         vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1202         ret = btrfs_init_new_device(root, vol_args->name);
1203
1204         kfree(vol_args);
1205         return ret;
1206 }
1207
1208 static long btrfs_ioctl_rm_dev(struct btrfs_root *root, void __user *arg)
1209 {
1210         struct btrfs_ioctl_vol_args *vol_args;
1211         int ret;
1212
1213         if (!capable(CAP_SYS_ADMIN))
1214                 return -EPERM;
1215
1216         if (root->fs_info->sb->s_flags & MS_RDONLY)
1217                 return -EROFS;
1218
1219         vol_args = memdup_user(arg, sizeof(*vol_args));
1220         if (IS_ERR(vol_args))
1221                 return PTR_ERR(vol_args);
1222
1223         vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1224         ret = btrfs_rm_device(root, vol_args->name);
1225
1226         kfree(vol_args);
1227         return ret;
1228 }
1229
1230 static noinline long btrfs_ioctl_clone(struct file *file, unsigned long srcfd,
1231                                        u64 off, u64 olen, u64 destoff)
1232 {
1233         struct inode *inode = fdentry(file)->d_inode;
1234         struct btrfs_root *root = BTRFS_I(inode)->root;
1235         struct file *src_file;
1236         struct inode *src;
1237         struct btrfs_trans_handle *trans;
1238         struct btrfs_path *path;
1239         struct extent_buffer *leaf;
1240         char *buf;
1241         struct btrfs_key key;
1242         u32 nritems;
1243         int slot;
1244         int ret;
1245         u64 len = olen;
1246         u64 bs = root->fs_info->sb->s_blocksize;
1247         u64 hint_byte;
1248
1249         /*
1250          * TODO:
1251          * - split compressed inline extents.  annoying: we need to
1252          *   decompress into destination's address_space (the file offset
1253          *   may change, so source mapping won't do), then recompress (or
1254          *   otherwise reinsert) a subrange.
1255          * - allow ranges within the same file to be cloned (provided
1256          *   they don't overlap)?
1257          */
1258
1259         /* the destination must be opened for writing */
1260         if (!(file->f_mode & FMODE_WRITE))
1261                 return -EINVAL;
1262
1263         ret = mnt_want_write(file->f_path.mnt);
1264         if (ret)
1265                 return ret;
1266
1267         src_file = fget(srcfd);
1268         if (!src_file) {
1269                 ret = -EBADF;
1270                 goto out_drop_write;
1271         }
1272         src = src_file->f_dentry->d_inode;
1273
1274         ret = -EINVAL;
1275         if (src == inode)
1276                 goto out_fput;
1277
1278         ret = -EISDIR;
1279         if (S_ISDIR(src->i_mode) || S_ISDIR(inode->i_mode))
1280                 goto out_fput;
1281
1282         ret = -EXDEV;
1283         if (src->i_sb != inode->i_sb || BTRFS_I(src)->root != root)
1284                 goto out_fput;
1285
1286         ret = -ENOMEM;
1287         buf = vmalloc(btrfs_level_size(root, 0));
1288         if (!buf)
1289                 goto out_fput;
1290
1291         path = btrfs_alloc_path();
1292         if (!path) {
1293                 vfree(buf);
1294                 goto out_fput;
1295         }
1296         path->reada = 2;
1297
1298         if (inode < src) {
1299                 mutex_lock(&inode->i_mutex);
1300                 mutex_lock(&src->i_mutex);
1301         } else {
1302                 mutex_lock(&src->i_mutex);
1303                 mutex_lock(&inode->i_mutex);
1304         }
1305
1306         /* determine range to clone */
1307         ret = -EINVAL;
1308         if (off >= src->i_size || off + len > src->i_size)
1309                 goto out_unlock;
1310         if (len == 0)
1311                 olen = len = src->i_size - off;
1312         /* if we extend to eof, continue to block boundary */
1313         if (off + len == src->i_size)
1314                 len = ((src->i_size + bs-1) & ~(bs-1))
1315                         - off;
1316
1317         /* verify the end result is block aligned */
1318         if ((off & (bs-1)) ||
1319             ((off + len) & (bs-1)))
1320                 goto out_unlock;
1321
1322         /* do any pending delalloc/csum calc on src, one way or
1323            another, and lock file content */
1324         while (1) {
1325                 struct btrfs_ordered_extent *ordered;
1326                 lock_extent(&BTRFS_I(src)->io_tree, off, off+len, GFP_NOFS);
1327                 ordered = btrfs_lookup_first_ordered_extent(inode, off+len);
1328                 if (BTRFS_I(src)->delalloc_bytes == 0 && !ordered)
1329                         break;
1330                 unlock_extent(&BTRFS_I(src)->io_tree, off, off+len, GFP_NOFS);
1331                 if (ordered)
1332                         btrfs_put_ordered_extent(ordered);
1333                 btrfs_wait_ordered_range(src, off, off+len);
1334         }
1335
1336         trans = btrfs_start_transaction(root, 1);
1337         BUG_ON(!trans);
1338
1339         /* punch hole in destination first */
1340         btrfs_drop_extents(trans, inode, off, off + len, &hint_byte, 1);
1341
1342         /* clone data */
1343         key.objectid = src->i_ino;
1344         key.type = BTRFS_EXTENT_DATA_KEY;
1345         key.offset = 0;
1346
1347         while (1) {
1348                 /*
1349                  * note the key will change type as we walk through the
1350                  * tree.
1351                  */
1352                 ret = btrfs_search_slot(trans, root, &key, path, 0, 0);
1353                 if (ret < 0)
1354                         goto out;
1355
1356                 nritems = btrfs_header_nritems(path->nodes[0]);
1357                 if (path->slots[0] >= nritems) {
1358                         ret = btrfs_next_leaf(root, path);
1359                         if (ret < 0)
1360                                 goto out;
1361                         if (ret > 0)
1362                                 break;
1363                         nritems = btrfs_header_nritems(path->nodes[0]);
1364                 }
1365                 leaf = path->nodes[0];
1366                 slot = path->slots[0];
1367
1368                 btrfs_item_key_to_cpu(leaf, &key, slot);
1369                 if (btrfs_key_type(&key) > BTRFS_EXTENT_DATA_KEY ||
1370                     key.objectid != src->i_ino)
1371                         break;
1372
1373                 if (btrfs_key_type(&key) == BTRFS_EXTENT_DATA_KEY) {
1374                         struct btrfs_file_extent_item *extent;
1375                         int type;
1376                         u32 size;
1377                         struct btrfs_key new_key;
1378                         u64 disko = 0, diskl = 0;
1379                         u64 datao = 0, datal = 0;
1380                         u8 comp;
1381
1382                         size = btrfs_item_size_nr(leaf, slot);
1383                         read_extent_buffer(leaf, buf,
1384                                            btrfs_item_ptr_offset(leaf, slot),
1385                                            size);
1386
1387                         extent = btrfs_item_ptr(leaf, slot,
1388                                                 struct btrfs_file_extent_item);
1389                         comp = btrfs_file_extent_compression(leaf, extent);
1390                         type = btrfs_file_extent_type(leaf, extent);
1391                         if (type == BTRFS_FILE_EXTENT_REG ||
1392                             type == BTRFS_FILE_EXTENT_PREALLOC) {
1393                                 disko = btrfs_file_extent_disk_bytenr(leaf,
1394                                                                       extent);
1395                                 diskl = btrfs_file_extent_disk_num_bytes(leaf,
1396                                                                  extent);
1397                                 datao = btrfs_file_extent_offset(leaf, extent);
1398                                 datal = btrfs_file_extent_num_bytes(leaf,
1399                                                                     extent);
1400                         } else if (type == BTRFS_FILE_EXTENT_INLINE) {
1401                                 /* take upper bound, may be compressed */
1402                                 datal = btrfs_file_extent_ram_bytes(leaf,
1403                                                                     extent);
1404                         }
1405                         btrfs_release_path(root, path);
1406
1407                         if (key.offset + datal < off ||
1408                             key.offset >= off+len)
1409                                 goto next;
1410
1411                         memcpy(&new_key, &key, sizeof(new_key));
1412                         new_key.objectid = inode->i_ino;
1413                         new_key.offset = key.offset + destoff - off;
1414
1415                         if (type == BTRFS_FILE_EXTENT_REG ||
1416                             type == BTRFS_FILE_EXTENT_PREALLOC) {
1417                                 ret = btrfs_insert_empty_item(trans, root, path,
1418                                                               &new_key, size);
1419                                 if (ret)
1420                                         goto out;
1421
1422                                 leaf = path->nodes[0];
1423                                 slot = path->slots[0];
1424                                 write_extent_buffer(leaf, buf,
1425                                             btrfs_item_ptr_offset(leaf, slot),
1426                                             size);
1427
1428                                 extent = btrfs_item_ptr(leaf, slot,
1429                                                 struct btrfs_file_extent_item);
1430
1431                                 if (off > key.offset) {
1432                                         datao += off - key.offset;
1433                                         datal -= off - key.offset;
1434                                 }
1435
1436                                 if (key.offset + datal > off + len)
1437                                         datal = off + len - key.offset;
1438
1439                                 /* disko == 0 means it's a hole */
1440                                 if (!disko)
1441                                         datao = 0;
1442
1443                                 btrfs_set_file_extent_offset(leaf, extent,
1444                                                              datao);
1445                                 btrfs_set_file_extent_num_bytes(leaf, extent,
1446                                                                 datal);
1447                                 if (disko) {
1448                                         inode_add_bytes(inode, datal);
1449                                         ret = btrfs_inc_extent_ref(trans, root,
1450                                                         disko, diskl, 0,
1451                                                         root->root_key.objectid,
1452                                                         inode->i_ino,
1453                                                         new_key.offset - datao);
1454                                         BUG_ON(ret);
1455                                 }
1456                         } else if (type == BTRFS_FILE_EXTENT_INLINE) {
1457                                 u64 skip = 0;
1458                                 u64 trim = 0;
1459                                 if (off > key.offset) {
1460                                         skip = off - key.offset;
1461                                         new_key.offset += skip;
1462                                 }
1463
1464                                 if (key.offset + datal > off+len)
1465                                         trim = key.offset + datal - (off+len);
1466
1467                                 if (comp && (skip || trim)) {
1468                                         ret = -EINVAL;
1469                                         goto out;
1470                                 }
1471                                 size -= skip + trim;
1472                                 datal -= skip + trim;
1473                                 ret = btrfs_insert_empty_item(trans, root, path,
1474                                                               &new_key, size);
1475                                 if (ret)
1476                                         goto out;
1477
1478                                 if (skip) {
1479                                         u32 start =
1480                                           btrfs_file_extent_calc_inline_size(0);
1481                                         memmove(buf+start, buf+start+skip,
1482                                                 datal);
1483                                 }
1484
1485                                 leaf = path->nodes[0];
1486                                 slot = path->slots[0];
1487                                 write_extent_buffer(leaf, buf,
1488                                             btrfs_item_ptr_offset(leaf, slot),
1489                                             size);
1490                                 inode_add_bytes(inode, datal);
1491                         }
1492
1493                         btrfs_mark_buffer_dirty(leaf);
1494                 }
1495
1496 next:
1497                 btrfs_release_path(root, path);
1498                 key.offset++;
1499         }
1500         ret = 0;
1501 out:
1502         btrfs_release_path(root, path);
1503         if (ret == 0) {
1504                 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
1505                 if (destoff + olen > inode->i_size)
1506                         btrfs_i_size_write(inode, destoff + olen);
1507                 BTRFS_I(inode)->flags = BTRFS_I(src)->flags;
1508                 ret = btrfs_update_inode(trans, root, inode);
1509         }
1510         btrfs_end_transaction(trans, root);
1511         unlock_extent(&BTRFS_I(src)->io_tree, off, off+len, GFP_NOFS);
1512         if (ret)
1513                 vmtruncate(inode, 0);
1514 out_unlock:
1515         mutex_unlock(&src->i_mutex);
1516         mutex_unlock(&inode->i_mutex);
1517         vfree(buf);
1518         btrfs_free_path(path);
1519 out_fput:
1520         fput(src_file);
1521 out_drop_write:
1522         mnt_drop_write(file->f_path.mnt);
1523         return ret;
1524 }
1525
1526 static long btrfs_ioctl_clone_range(struct file *file, void __user *argp)
1527 {
1528         struct btrfs_ioctl_clone_range_args args;
1529
1530         if (copy_from_user(&args, argp, sizeof(args)))
1531                 return -EFAULT;
1532         return btrfs_ioctl_clone(file, args.src_fd, args.src_offset,
1533                                  args.src_length, args.dest_offset);
1534 }
1535
1536 /*
1537  * there are many ways the trans_start and trans_end ioctls can lead
1538  * to deadlocks.  They should only be used by applications that
1539  * basically own the machine, and have a very in depth understanding
1540  * of all the possible deadlocks and enospc problems.
1541  */
1542 static long btrfs_ioctl_trans_start(struct file *file)
1543 {
1544         struct inode *inode = fdentry(file)->d_inode;
1545         struct btrfs_root *root = BTRFS_I(inode)->root;
1546         struct btrfs_trans_handle *trans;
1547         int ret;
1548
1549         ret = -EPERM;
1550         if (!capable(CAP_SYS_ADMIN))
1551                 goto out;
1552
1553         ret = -EINPROGRESS;
1554         if (file->private_data)
1555                 goto out;
1556
1557         ret = mnt_want_write(file->f_path.mnt);
1558         if (ret)
1559                 goto out;
1560
1561         mutex_lock(&root->fs_info->trans_mutex);
1562         root->fs_info->open_ioctl_trans++;
1563         mutex_unlock(&root->fs_info->trans_mutex);
1564
1565         ret = -ENOMEM;
1566         trans = btrfs_start_ioctl_transaction(root, 0);
1567         if (!trans)
1568                 goto out_drop;
1569
1570         file->private_data = trans;
1571         return 0;
1572
1573 out_drop:
1574         mutex_lock(&root->fs_info->trans_mutex);
1575         root->fs_info->open_ioctl_trans--;
1576         mutex_unlock(&root->fs_info->trans_mutex);
1577         mnt_drop_write(file->f_path.mnt);
1578 out:
1579         return ret;
1580 }
1581
1582 static long btrfs_ioctl_default_subvol(struct file *file, void __user *argp)
1583 {
1584         struct inode *inode = fdentry(file)->d_inode;
1585         struct btrfs_root *root = BTRFS_I(inode)->root;
1586         struct btrfs_root *new_root;
1587         struct btrfs_dir_item *di;
1588         struct btrfs_trans_handle *trans;
1589         struct btrfs_path *path;
1590         struct btrfs_key location;
1591         struct btrfs_disk_key disk_key;
1592         struct btrfs_super_block *disk_super;
1593         u64 features;
1594         u64 objectid = 0;
1595         u64 dir_id;
1596
1597         if (!capable(CAP_SYS_ADMIN))
1598                 return -EPERM;
1599
1600         if (copy_from_user(&objectid, argp, sizeof(objectid)))
1601                 return -EFAULT;
1602
1603         if (!objectid)
1604                 objectid = root->root_key.objectid;
1605
1606         location.objectid = objectid;
1607         location.type = BTRFS_ROOT_ITEM_KEY;
1608         location.offset = (u64)-1;
1609
1610         new_root = btrfs_read_fs_root_no_name(root->fs_info, &location);
1611         if (IS_ERR(new_root))
1612                 return PTR_ERR(new_root);
1613
1614         if (btrfs_root_refs(&new_root->root_item) == 0)
1615                 return -ENOENT;
1616
1617         path = btrfs_alloc_path();
1618         if (!path)
1619                 return -ENOMEM;
1620         path->leave_spinning = 1;
1621
1622         trans = btrfs_start_transaction(root, 1);
1623         if (!trans) {
1624                 btrfs_free_path(path);
1625                 return -ENOMEM;
1626         }
1627
1628         dir_id = btrfs_super_root_dir(&root->fs_info->super_copy);
1629         di = btrfs_lookup_dir_item(trans, root->fs_info->tree_root, path,
1630                                    dir_id, "default", 7, 1);
1631         if (!di) {
1632                 btrfs_free_path(path);
1633                 btrfs_end_transaction(trans, root);
1634                 printk(KERN_ERR "Umm, you don't have the default dir item, "
1635                        "this isn't going to work\n");
1636                 return -ENOENT;
1637         }
1638
1639         btrfs_cpu_key_to_disk(&disk_key, &new_root->root_key);
1640         btrfs_set_dir_item_key(path->nodes[0], di, &disk_key);
1641         btrfs_mark_buffer_dirty(path->nodes[0]);
1642         btrfs_free_path(path);
1643
1644         disk_super = &root->fs_info->super_copy;
1645         features = btrfs_super_incompat_flags(disk_super);
1646         if (!(features & BTRFS_FEATURE_INCOMPAT_DEFAULT_SUBVOL)) {
1647                 features |= BTRFS_FEATURE_INCOMPAT_DEFAULT_SUBVOL;
1648                 btrfs_set_super_incompat_flags(disk_super, features);
1649         }
1650         btrfs_end_transaction(trans, root);
1651
1652         return 0;
1653 }
1654
1655 /*
1656  * there are many ways the trans_start and trans_end ioctls can lead
1657  * to deadlocks.  They should only be used by applications that
1658  * basically own the machine, and have a very in depth understanding
1659  * of all the possible deadlocks and enospc problems.
1660  */
1661 long btrfs_ioctl_trans_end(struct file *file)
1662 {
1663         struct inode *inode = fdentry(file)->d_inode;
1664         struct btrfs_root *root = BTRFS_I(inode)->root;
1665         struct btrfs_trans_handle *trans;
1666
1667         trans = file->private_data;
1668         if (!trans)
1669                 return -EINVAL;
1670         file->private_data = NULL;
1671
1672         btrfs_end_transaction(trans, root);
1673
1674         mutex_lock(&root->fs_info->trans_mutex);
1675         root->fs_info->open_ioctl_trans--;
1676         mutex_unlock(&root->fs_info->trans_mutex);
1677
1678         mnt_drop_write(file->f_path.mnt);
1679         return 0;
1680 }
1681
1682 long btrfs_ioctl(struct file *file, unsigned int
1683                 cmd, unsigned long arg)
1684 {
1685         struct btrfs_root *root = BTRFS_I(fdentry(file)->d_inode)->root;
1686         void __user *argp = (void __user *)arg;
1687
1688         switch (cmd) {
1689         case FS_IOC_GETFLAGS:
1690                 return btrfs_ioctl_getflags(file, argp);
1691         case FS_IOC_SETFLAGS:
1692                 return btrfs_ioctl_setflags(file, argp);
1693         case FS_IOC_GETVERSION:
1694                 return btrfs_ioctl_getversion(file, argp);
1695         case BTRFS_IOC_SNAP_CREATE:
1696                 return btrfs_ioctl_snap_create(file, argp, 0);
1697         case BTRFS_IOC_SUBVOL_CREATE:
1698                 return btrfs_ioctl_snap_create(file, argp, 1);
1699         case BTRFS_IOC_SNAP_DESTROY:
1700                 return btrfs_ioctl_snap_destroy(file, argp);
1701         case BTRFS_IOC_DEFAULT_SUBVOL:
1702                 return btrfs_ioctl_default_subvol(file, argp);
1703         case BTRFS_IOC_DEFRAG:
1704                 return btrfs_ioctl_defrag(file);
1705         case BTRFS_IOC_RESIZE:
1706                 return btrfs_ioctl_resize(root, argp);
1707         case BTRFS_IOC_ADD_DEV:
1708                 return btrfs_ioctl_add_dev(root, argp);
1709         case BTRFS_IOC_RM_DEV:
1710                 return btrfs_ioctl_rm_dev(root, argp);
1711         case BTRFS_IOC_BALANCE:
1712                 return btrfs_balance(root->fs_info->dev_root);
1713         case BTRFS_IOC_CLONE:
1714                 return btrfs_ioctl_clone(file, arg, 0, 0, 0);
1715         case BTRFS_IOC_CLONE_RANGE:
1716                 return btrfs_ioctl_clone_range(file, argp);
1717         case BTRFS_IOC_TRANS_START:
1718                 return btrfs_ioctl_trans_start(file);
1719         case BTRFS_IOC_TRANS_END:
1720                 return btrfs_ioctl_trans_end(file);
1721         case BTRFS_IOC_TREE_SEARCH:
1722                 return btrfs_ioctl_tree_search(file, argp);
1723         case BTRFS_IOC_INO_LOOKUP:
1724                 return btrfs_ioctl_ino_lookup(file, argp);
1725         case BTRFS_IOC_SYNC:
1726                 btrfs_sync_fs(file->f_dentry->d_sb, 1);
1727                 return 0;
1728         }
1729
1730         return -ENOTTY;
1731 }