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 52 /* Mask out flags that are inappropriate for the given type of inode. */ 53 static inline __u32 btrfs_mask_flags(umode_t mode, __u32 flags) 54 { 55 if (S_ISDIR(mode)) 56 return flags; 57 else if (S_ISREG(mode)) 58 return flags & ~FS_DIRSYNC_FL; 59 else 60 return flags & (FS_NODUMP_FL | FS_NOATIME_FL); 61 } 62 63 /* 64 * Export inode flags to the format expected by the FS_IOC_GETFLAGS ioctl. 65 */ 66 static unsigned int btrfs_flags_to_ioctl(unsigned int flags) 67 { 68 unsigned int iflags = 0; 69 70 if (flags & BTRFS_INODE_SYNC) 71 iflags |= FS_SYNC_FL; 72 if (flags & BTRFS_INODE_IMMUTABLE) 73 iflags |= FS_IMMUTABLE_FL; 74 if (flags & BTRFS_INODE_APPEND) 75 iflags |= FS_APPEND_FL; 76 if (flags & BTRFS_INODE_NODUMP) 77 iflags |= FS_NODUMP_FL; 78 if (flags & BTRFS_INODE_NOATIME) 79 iflags |= FS_NOATIME_FL; 80 if (flags & BTRFS_INODE_DIRSYNC) 81 iflags |= FS_DIRSYNC_FL; 82 83 return iflags; 84 } 85 86 /* 87 * Update inode->i_flags based on the btrfs internal flags. 88 */ 89 void btrfs_update_iflags(struct inode *inode) 90 { 91 struct btrfs_inode *ip = BTRFS_I(inode); 92 93 inode->i_flags &= ~(S_SYNC|S_APPEND|S_IMMUTABLE|S_NOATIME|S_DIRSYNC); 94 95 if (ip->flags & BTRFS_INODE_SYNC) 96 inode->i_flags |= S_SYNC; 97 if (ip->flags & BTRFS_INODE_IMMUTABLE) 98 inode->i_flags |= S_IMMUTABLE; 99 if (ip->flags & BTRFS_INODE_APPEND) 100 inode->i_flags |= S_APPEND; 101 if (ip->flags & BTRFS_INODE_NOATIME) 102 inode->i_flags |= S_NOATIME; 103 if (ip->flags & BTRFS_INODE_DIRSYNC) 104 inode->i_flags |= S_DIRSYNC; 105 } 106 107 /* 108 * Inherit flags from the parent inode. 109 * 110 * Unlike extN we don't have any flags we don't want to inherit currently. 111 */ 112 void btrfs_inherit_iflags(struct inode *inode, struct inode *dir) 113 { 114 unsigned int flags; 115 116 if (!dir) 117 return; 118 119 flags = BTRFS_I(dir)->flags; 120 121 if (S_ISREG(inode->i_mode)) 122 flags &= ~BTRFS_INODE_DIRSYNC; 123 else if (!S_ISDIR(inode->i_mode)) 124 flags &= (BTRFS_INODE_NODUMP | BTRFS_INODE_NOATIME); 125 126 BTRFS_I(inode)->flags = flags; 127 btrfs_update_iflags(inode); 128 } 129 130 static int btrfs_ioctl_getflags(struct file *file, void __user *arg) 131 { 132 struct btrfs_inode *ip = BTRFS_I(file->f_path.dentry->d_inode); 133 unsigned int flags = btrfs_flags_to_ioctl(ip->flags); 134 135 if (copy_to_user(arg, &flags, sizeof(flags))) 136 return -EFAULT; 137 return 0; 138 } 139 140 static int btrfs_ioctl_setflags(struct file *file, void __user *arg) 141 { 142 struct inode *inode = file->f_path.dentry->d_inode; 143 struct btrfs_inode *ip = BTRFS_I(inode); 144 struct btrfs_root *root = ip->root; 145 struct btrfs_trans_handle *trans; 146 unsigned int flags, oldflags; 147 int ret; 148 149 if (copy_from_user(&flags, arg, sizeof(flags))) 150 return -EFAULT; 151 152 if (flags & ~(FS_IMMUTABLE_FL | FS_APPEND_FL | \ 153 FS_NOATIME_FL | FS_NODUMP_FL | \ 154 FS_SYNC_FL | FS_DIRSYNC_FL)) 155 return -EOPNOTSUPP; 156 157 if (!is_owner_or_cap(inode)) 158 return -EACCES; 159 160 mutex_lock(&inode->i_mutex); 161 162 flags = btrfs_mask_flags(inode->i_mode, flags); 163 oldflags = btrfs_flags_to_ioctl(ip->flags); 164 if ((flags ^ oldflags) & (FS_APPEND_FL | FS_IMMUTABLE_FL)) { 165 if (!capable(CAP_LINUX_IMMUTABLE)) { 166 ret = -EPERM; 167 goto out_unlock; 168 } 169 } 170 171 ret = mnt_want_write(file->f_path.mnt); 172 if (ret) 173 goto out_unlock; 174 175 if (flags & FS_SYNC_FL) 176 ip->flags |= BTRFS_INODE_SYNC; 177 else 178 ip->flags &= ~BTRFS_INODE_SYNC; 179 if (flags & FS_IMMUTABLE_FL) 180 ip->flags |= BTRFS_INODE_IMMUTABLE; 181 else 182 ip->flags &= ~BTRFS_INODE_IMMUTABLE; 183 if (flags & FS_APPEND_FL) 184 ip->flags |= BTRFS_INODE_APPEND; 185 else 186 ip->flags &= ~BTRFS_INODE_APPEND; 187 if (flags & FS_NODUMP_FL) 188 ip->flags |= BTRFS_INODE_NODUMP; 189 else 190 ip->flags &= ~BTRFS_INODE_NODUMP; 191 if (flags & FS_NOATIME_FL) 192 ip->flags |= BTRFS_INODE_NOATIME; 193 else 194 ip->flags &= ~BTRFS_INODE_NOATIME; 195 if (flags & FS_DIRSYNC_FL) 196 ip->flags |= BTRFS_INODE_DIRSYNC; 197 else 198 ip->flags &= ~BTRFS_INODE_DIRSYNC; 199 200 201 trans = btrfs_join_transaction(root, 1); 202 BUG_ON(!trans); 203 204 ret = btrfs_update_inode(trans, root, inode); 205 BUG_ON(ret); 206 207 btrfs_update_iflags(inode); 208 inode->i_ctime = CURRENT_TIME; 209 btrfs_end_transaction(trans, root); 210 211 mnt_drop_write(file->f_path.mnt); 212 out_unlock: 213 mutex_unlock(&inode->i_mutex); 214 return 0; 215 } 216 217 static int btrfs_ioctl_getversion(struct file *file, int __user *arg) 218 { 219 struct inode *inode = file->f_path.dentry->d_inode; 220 221 return put_user(inode->i_generation, arg); 222 } 223 224 static noinline int create_subvol(struct btrfs_root *root, 225 struct dentry *dentry, 226 char *name, int namelen) 227 { 228 struct btrfs_trans_handle *trans; 229 struct btrfs_key key; 230 struct btrfs_root_item root_item; 231 struct btrfs_inode_item *inode_item; 232 struct extent_buffer *leaf; 233 struct btrfs_root *new_root; 234 struct inode *dir = dentry->d_parent->d_inode; 235 int ret; 236 int err; 237 u64 objectid; 238 u64 new_dirid = BTRFS_FIRST_FREE_OBJECTID; 239 u64 index = 0; 240 unsigned long nr = 1; 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, 0); 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 nr = trans->blocks_used; 346 err = btrfs_commit_transaction(trans, root); 347 if (err && !ret) 348 ret = err; 349 350 btrfs_unreserve_metadata_space(root, 6); 351 btrfs_btree_balance_dirty(root, nr); 352 return ret; 353 } 354 355 static int create_snapshot(struct btrfs_root *root, struct dentry *dentry, 356 char *name, int namelen) 357 { 358 struct btrfs_pending_snapshot *pending_snapshot; 359 struct btrfs_trans_handle *trans; 360 int ret = 0; 361 int err; 362 unsigned long nr = 0; 363 364 if (!root->ref_cows) 365 return -EINVAL; 366 367 /* 368 * 1 - inode item 369 * 2 - refs 370 * 1 - root item 371 * 2 - dir items 372 */ 373 ret = btrfs_reserve_metadata_space(root, 6); 374 if (ret) 375 goto fail_unlock; 376 377 pending_snapshot = kzalloc(sizeof(*pending_snapshot), GFP_NOFS); 378 if (!pending_snapshot) { 379 ret = -ENOMEM; 380 btrfs_unreserve_metadata_space(root, 6); 381 goto fail_unlock; 382 } 383 pending_snapshot->name = kmalloc(namelen + 1, GFP_NOFS); 384 if (!pending_snapshot->name) { 385 ret = -ENOMEM; 386 kfree(pending_snapshot); 387 btrfs_unreserve_metadata_space(root, 6); 388 goto fail_unlock; 389 } 390 memcpy(pending_snapshot->name, name, namelen); 391 pending_snapshot->name[namelen] = '\0'; 392 pending_snapshot->dentry = dentry; 393 trans = btrfs_start_transaction(root, 1); 394 BUG_ON(!trans); 395 pending_snapshot->root = root; 396 list_add(&pending_snapshot->list, 397 &trans->transaction->pending_snapshots); 398 err = btrfs_commit_transaction(trans, root); 399 400 fail_unlock: 401 btrfs_btree_balance_dirty(root, nr); 402 return ret; 403 } 404 405 /* copy of may_create in fs/namei.c() */ 406 static inline int btrfs_may_create(struct inode *dir, struct dentry *child) 407 { 408 if (child->d_inode) 409 return -EEXIST; 410 if (IS_DEADDIR(dir)) 411 return -ENOENT; 412 return inode_permission(dir, MAY_WRITE | MAY_EXEC); 413 } 414 415 /* 416 * Create a new subvolume below @parent. This is largely modeled after 417 * sys_mkdirat and vfs_mkdir, but we only do a single component lookup 418 * inside this filesystem so it's quite a bit simpler. 419 */ 420 static noinline int btrfs_mksubvol(struct path *parent, 421 char *name, int namelen, 422 struct btrfs_root *snap_src) 423 { 424 struct inode *dir = parent->dentry->d_inode; 425 struct dentry *dentry; 426 int error; 427 428 mutex_lock_nested(&dir->i_mutex, I_MUTEX_PARENT); 429 430 dentry = lookup_one_len(name, parent->dentry, namelen); 431 error = PTR_ERR(dentry); 432 if (IS_ERR(dentry)) 433 goto out_unlock; 434 435 error = -EEXIST; 436 if (dentry->d_inode) 437 goto out_dput; 438 439 error = mnt_want_write(parent->mnt); 440 if (error) 441 goto out_dput; 442 443 error = btrfs_may_create(dir, dentry); 444 if (error) 445 goto out_drop_write; 446 447 down_read(&BTRFS_I(dir)->root->fs_info->subvol_sem); 448 449 if (btrfs_root_refs(&BTRFS_I(dir)->root->root_item) == 0) 450 goto out_up_read; 451 452 if (snap_src) { 453 error = create_snapshot(snap_src, dentry, 454 name, namelen); 455 } else { 456 error = create_subvol(BTRFS_I(dir)->root, dentry, 457 name, namelen); 458 } 459 if (!error) 460 fsnotify_mkdir(dir, dentry); 461 out_up_read: 462 up_read(&BTRFS_I(dir)->root->fs_info->subvol_sem); 463 out_drop_write: 464 mnt_drop_write(parent->mnt); 465 out_dput: 466 dput(dentry); 467 out_unlock: 468 mutex_unlock(&dir->i_mutex); 469 return error; 470 } 471 472 static int btrfs_defrag_file(struct file *file) 473 { 474 struct inode *inode = fdentry(file)->d_inode; 475 struct btrfs_root *root = BTRFS_I(inode)->root; 476 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree; 477 struct btrfs_ordered_extent *ordered; 478 struct page *page; 479 unsigned long last_index; 480 unsigned long ra_pages = root->fs_info->bdi.ra_pages; 481 unsigned long total_read = 0; 482 u64 page_start; 483 u64 page_end; 484 unsigned long i; 485 int ret; 486 487 ret = btrfs_check_data_free_space(root, inode, inode->i_size); 488 if (ret) 489 return -ENOSPC; 490 491 mutex_lock(&inode->i_mutex); 492 last_index = inode->i_size >> PAGE_CACHE_SHIFT; 493 for (i = 0; i <= last_index; i++) { 494 if (total_read % ra_pages == 0) { 495 btrfs_force_ra(inode->i_mapping, &file->f_ra, file, i, 496 min(last_index, i + ra_pages - 1)); 497 } 498 total_read++; 499 again: 500 page = grab_cache_page(inode->i_mapping, i); 501 if (!page) 502 goto out_unlock; 503 if (!PageUptodate(page)) { 504 btrfs_readpage(NULL, page); 505 lock_page(page); 506 if (!PageUptodate(page)) { 507 unlock_page(page); 508 page_cache_release(page); 509 goto out_unlock; 510 } 511 } 512 513 wait_on_page_writeback(page); 514 515 page_start = (u64)page->index << PAGE_CACHE_SHIFT; 516 page_end = page_start + PAGE_CACHE_SIZE - 1; 517 lock_extent(io_tree, page_start, page_end, GFP_NOFS); 518 519 ordered = btrfs_lookup_ordered_extent(inode, page_start); 520 if (ordered) { 521 unlock_extent(io_tree, page_start, page_end, GFP_NOFS); 522 unlock_page(page); 523 page_cache_release(page); 524 btrfs_start_ordered_extent(inode, ordered, 1); 525 btrfs_put_ordered_extent(ordered); 526 goto again; 527 } 528 set_page_extent_mapped(page); 529 530 /* 531 * this makes sure page_mkwrite is called on the 532 * page if it is dirtied again later 533 */ 534 clear_page_dirty_for_io(page); 535 536 btrfs_set_extent_delalloc(inode, page_start, page_end); 537 set_page_dirty(page); 538 unlock_extent(io_tree, page_start, page_end, GFP_NOFS); 539 unlock_page(page); 540 page_cache_release(page); 541 balance_dirty_pages_ratelimited_nr(inode->i_mapping, 1); 542 } 543 544 out_unlock: 545 mutex_unlock(&inode->i_mutex); 546 return 0; 547 } 548 549 static noinline int btrfs_ioctl_resize(struct btrfs_root *root, 550 void __user *arg) 551 { 552 u64 new_size; 553 u64 old_size; 554 u64 devid = 1; 555 struct btrfs_ioctl_vol_args *vol_args; 556 struct btrfs_trans_handle *trans; 557 struct btrfs_device *device = NULL; 558 char *sizestr; 559 char *devstr = NULL; 560 int ret = 0; 561 int namelen; 562 int mod = 0; 563 564 if (root->fs_info->sb->s_flags & MS_RDONLY) 565 return -EROFS; 566 567 if (!capable(CAP_SYS_ADMIN)) 568 return -EPERM; 569 570 vol_args = memdup_user(arg, sizeof(*vol_args)); 571 if (IS_ERR(vol_args)) 572 return PTR_ERR(vol_args); 573 574 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; 575 namelen = strlen(vol_args->name); 576 577 mutex_lock(&root->fs_info->volume_mutex); 578 sizestr = vol_args->name; 579 devstr = strchr(sizestr, ':'); 580 if (devstr) { 581 char *end; 582 sizestr = devstr + 1; 583 *devstr = '\0'; 584 devstr = vol_args->name; 585 devid = simple_strtoull(devstr, &end, 10); 586 printk(KERN_INFO "resizing devid %llu\n", 587 (unsigned long long)devid); 588 } 589 device = btrfs_find_device(root, devid, NULL, NULL); 590 if (!device) { 591 printk(KERN_INFO "resizer unable to find device %llu\n", 592 (unsigned long long)devid); 593 ret = -EINVAL; 594 goto out_unlock; 595 } 596 if (!strcmp(sizestr, "max")) 597 new_size = device->bdev->bd_inode->i_size; 598 else { 599 if (sizestr[0] == '-') { 600 mod = -1; 601 sizestr++; 602 } else if (sizestr[0] == '+') { 603 mod = 1; 604 sizestr++; 605 } 606 new_size = btrfs_parse_size(sizestr); 607 if (new_size == 0) { 608 ret = -EINVAL; 609 goto out_unlock; 610 } 611 } 612 613 old_size = device->total_bytes; 614 615 if (mod < 0) { 616 if (new_size > old_size) { 617 ret = -EINVAL; 618 goto out_unlock; 619 } 620 new_size = old_size - new_size; 621 } else if (mod > 0) { 622 new_size = old_size + new_size; 623 } 624 625 if (new_size < 256 * 1024 * 1024) { 626 ret = -EINVAL; 627 goto out_unlock; 628 } 629 if (new_size > device->bdev->bd_inode->i_size) { 630 ret = -EFBIG; 631 goto out_unlock; 632 } 633 634 do_div(new_size, root->sectorsize); 635 new_size *= root->sectorsize; 636 637 printk(KERN_INFO "new size for %s is %llu\n", 638 device->name, (unsigned long long)new_size); 639 640 if (new_size > old_size) { 641 trans = btrfs_start_transaction(root, 1); 642 ret = btrfs_grow_device(trans, device, new_size); 643 btrfs_commit_transaction(trans, root); 644 } else { 645 ret = btrfs_shrink_device(device, new_size); 646 } 647 648 out_unlock: 649 mutex_unlock(&root->fs_info->volume_mutex); 650 kfree(vol_args); 651 return ret; 652 } 653 654 static noinline int btrfs_ioctl_snap_create(struct file *file, 655 void __user *arg, int subvol) 656 { 657 struct btrfs_root *root = BTRFS_I(fdentry(file)->d_inode)->root; 658 struct btrfs_ioctl_vol_args *vol_args; 659 struct file *src_file; 660 int namelen; 661 int ret = 0; 662 663 if (root->fs_info->sb->s_flags & MS_RDONLY) 664 return -EROFS; 665 666 vol_args = memdup_user(arg, sizeof(*vol_args)); 667 if (IS_ERR(vol_args)) 668 return PTR_ERR(vol_args); 669 670 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; 671 namelen = strlen(vol_args->name); 672 if (strchr(vol_args->name, '/')) { 673 ret = -EINVAL; 674 goto out; 675 } 676 677 if (subvol) { 678 ret = btrfs_mksubvol(&file->f_path, vol_args->name, namelen, 679 NULL); 680 } else { 681 struct inode *src_inode; 682 src_file = fget(vol_args->fd); 683 if (!src_file) { 684 ret = -EINVAL; 685 goto out; 686 } 687 688 src_inode = src_file->f_path.dentry->d_inode; 689 if (src_inode->i_sb != file->f_path.dentry->d_inode->i_sb) { 690 printk(KERN_INFO "btrfs: Snapshot src from " 691 "another FS\n"); 692 ret = -EINVAL; 693 fput(src_file); 694 goto out; 695 } 696 ret = btrfs_mksubvol(&file->f_path, vol_args->name, namelen, 697 BTRFS_I(src_inode)->root); 698 fput(src_file); 699 } 700 out: 701 kfree(vol_args); 702 return ret; 703 } 704 705 /* 706 * helper to check if the subvolume references other subvolumes 707 */ 708 static noinline int may_destroy_subvol(struct btrfs_root *root) 709 { 710 struct btrfs_path *path; 711 struct btrfs_key key; 712 int ret; 713 714 path = btrfs_alloc_path(); 715 if (!path) 716 return -ENOMEM; 717 718 key.objectid = root->root_key.objectid; 719 key.type = BTRFS_ROOT_REF_KEY; 720 key.offset = (u64)-1; 721 722 ret = btrfs_search_slot(NULL, root->fs_info->tree_root, 723 &key, path, 0, 0); 724 if (ret < 0) 725 goto out; 726 BUG_ON(ret == 0); 727 728 ret = 0; 729 if (path->slots[0] > 0) { 730 path->slots[0]--; 731 btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]); 732 if (key.objectid == root->root_key.objectid && 733 key.type == BTRFS_ROOT_REF_KEY) 734 ret = -ENOTEMPTY; 735 } 736 out: 737 btrfs_free_path(path); 738 return ret; 739 } 740 741 static noinline int btrfs_ioctl_snap_destroy(struct file *file, 742 void __user *arg) 743 { 744 struct dentry *parent = fdentry(file); 745 struct dentry *dentry; 746 struct inode *dir = parent->d_inode; 747 struct inode *inode; 748 struct btrfs_root *root = BTRFS_I(dir)->root; 749 struct btrfs_root *dest = NULL; 750 struct btrfs_ioctl_vol_args *vol_args; 751 struct btrfs_trans_handle *trans; 752 int namelen; 753 int ret; 754 int err = 0; 755 756 if (!capable(CAP_SYS_ADMIN)) 757 return -EPERM; 758 759 vol_args = memdup_user(arg, sizeof(*vol_args)); 760 if (IS_ERR(vol_args)) 761 return PTR_ERR(vol_args); 762 763 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; 764 namelen = strlen(vol_args->name); 765 if (strchr(vol_args->name, '/') || 766 strncmp(vol_args->name, "..", namelen) == 0) { 767 err = -EINVAL; 768 goto out; 769 } 770 771 err = mnt_want_write(file->f_path.mnt); 772 if (err) 773 goto out; 774 775 mutex_lock_nested(&dir->i_mutex, I_MUTEX_PARENT); 776 dentry = lookup_one_len(vol_args->name, parent, namelen); 777 if (IS_ERR(dentry)) { 778 err = PTR_ERR(dentry); 779 goto out_unlock_dir; 780 } 781 782 if (!dentry->d_inode) { 783 err = -ENOENT; 784 goto out_dput; 785 } 786 787 inode = dentry->d_inode; 788 if (inode->i_ino != BTRFS_FIRST_FREE_OBJECTID) { 789 err = -EINVAL; 790 goto out_dput; 791 } 792 793 dest = BTRFS_I(inode)->root; 794 795 mutex_lock(&inode->i_mutex); 796 err = d_invalidate(dentry); 797 if (err) 798 goto out_unlock; 799 800 down_write(&root->fs_info->subvol_sem); 801 802 err = may_destroy_subvol(dest); 803 if (err) 804 goto out_up_write; 805 806 trans = btrfs_start_transaction(root, 1); 807 ret = btrfs_unlink_subvol(trans, root, dir, 808 dest->root_key.objectid, 809 dentry->d_name.name, 810 dentry->d_name.len); 811 BUG_ON(ret); 812 813 btrfs_record_root_in_trans(trans, dest); 814 815 memset(&dest->root_item.drop_progress, 0, 816 sizeof(dest->root_item.drop_progress)); 817 dest->root_item.drop_level = 0; 818 btrfs_set_root_refs(&dest->root_item, 0); 819 820 ret = btrfs_insert_orphan_item(trans, 821 root->fs_info->tree_root, 822 dest->root_key.objectid); 823 BUG_ON(ret); 824 825 ret = btrfs_commit_transaction(trans, root); 826 BUG_ON(ret); 827 inode->i_flags |= S_DEAD; 828 out_up_write: 829 up_write(&root->fs_info->subvol_sem); 830 out_unlock: 831 mutex_unlock(&inode->i_mutex); 832 if (!err) { 833 shrink_dcache_sb(root->fs_info->sb); 834 btrfs_invalidate_inodes(dest); 835 d_delete(dentry); 836 } 837 out_dput: 838 dput(dentry); 839 out_unlock_dir: 840 mutex_unlock(&dir->i_mutex); 841 mnt_drop_write(file->f_path.mnt); 842 out: 843 kfree(vol_args); 844 return err; 845 } 846 847 static int btrfs_ioctl_defrag(struct file *file) 848 { 849 struct inode *inode = fdentry(file)->d_inode; 850 struct btrfs_root *root = BTRFS_I(inode)->root; 851 int ret; 852 853 ret = mnt_want_write(file->f_path.mnt); 854 if (ret) 855 return ret; 856 857 switch (inode->i_mode & S_IFMT) { 858 case S_IFDIR: 859 if (!capable(CAP_SYS_ADMIN)) { 860 ret = -EPERM; 861 goto out; 862 } 863 btrfs_defrag_root(root, 0); 864 btrfs_defrag_root(root->fs_info->extent_root, 0); 865 break; 866 case S_IFREG: 867 if (!(file->f_mode & FMODE_WRITE)) { 868 ret = -EINVAL; 869 goto out; 870 } 871 btrfs_defrag_file(file); 872 break; 873 } 874 out: 875 mnt_drop_write(file->f_path.mnt); 876 return ret; 877 } 878 879 static long btrfs_ioctl_add_dev(struct btrfs_root *root, void __user *arg) 880 { 881 struct btrfs_ioctl_vol_args *vol_args; 882 int ret; 883 884 if (!capable(CAP_SYS_ADMIN)) 885 return -EPERM; 886 887 vol_args = memdup_user(arg, sizeof(*vol_args)); 888 if (IS_ERR(vol_args)) 889 return PTR_ERR(vol_args); 890 891 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; 892 ret = btrfs_init_new_device(root, vol_args->name); 893 894 kfree(vol_args); 895 return ret; 896 } 897 898 static long btrfs_ioctl_rm_dev(struct btrfs_root *root, void __user *arg) 899 { 900 struct btrfs_ioctl_vol_args *vol_args; 901 int ret; 902 903 if (!capable(CAP_SYS_ADMIN)) 904 return -EPERM; 905 906 if (root->fs_info->sb->s_flags & MS_RDONLY) 907 return -EROFS; 908 909 vol_args = memdup_user(arg, sizeof(*vol_args)); 910 if (IS_ERR(vol_args)) 911 return PTR_ERR(vol_args); 912 913 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; 914 ret = btrfs_rm_device(root, vol_args->name); 915 916 kfree(vol_args); 917 return ret; 918 } 919 920 static noinline long btrfs_ioctl_clone(struct file *file, unsigned long srcfd, 921 u64 off, u64 olen, u64 destoff) 922 { 923 struct inode *inode = fdentry(file)->d_inode; 924 struct btrfs_root *root = BTRFS_I(inode)->root; 925 struct file *src_file; 926 struct inode *src; 927 struct btrfs_trans_handle *trans; 928 struct btrfs_path *path; 929 struct extent_buffer *leaf; 930 char *buf; 931 struct btrfs_key key; 932 u32 nritems; 933 int slot; 934 int ret; 935 u64 len = olen; 936 u64 bs = root->fs_info->sb->s_blocksize; 937 u64 hint_byte; 938 939 /* 940 * TODO: 941 * - split compressed inline extents. annoying: we need to 942 * decompress into destination's address_space (the file offset 943 * may change, so source mapping won't do), then recompress (or 944 * otherwise reinsert) a subrange. 945 * - allow ranges within the same file to be cloned (provided 946 * they don't overlap)? 947 */ 948 949 /* the destination must be opened for writing */ 950 if (!(file->f_mode & FMODE_WRITE)) 951 return -EINVAL; 952 953 ret = mnt_want_write(file->f_path.mnt); 954 if (ret) 955 return ret; 956 957 src_file = fget(srcfd); 958 if (!src_file) { 959 ret = -EBADF; 960 goto out_drop_write; 961 } 962 src = src_file->f_dentry->d_inode; 963 964 ret = -EINVAL; 965 if (src == inode) 966 goto out_fput; 967 968 ret = -EISDIR; 969 if (S_ISDIR(src->i_mode) || S_ISDIR(inode->i_mode)) 970 goto out_fput; 971 972 ret = -EXDEV; 973 if (src->i_sb != inode->i_sb || BTRFS_I(src)->root != root) 974 goto out_fput; 975 976 ret = -ENOMEM; 977 buf = vmalloc(btrfs_level_size(root, 0)); 978 if (!buf) 979 goto out_fput; 980 981 path = btrfs_alloc_path(); 982 if (!path) { 983 vfree(buf); 984 goto out_fput; 985 } 986 path->reada = 2; 987 988 if (inode < src) { 989 mutex_lock(&inode->i_mutex); 990 mutex_lock(&src->i_mutex); 991 } else { 992 mutex_lock(&src->i_mutex); 993 mutex_lock(&inode->i_mutex); 994 } 995 996 /* determine range to clone */ 997 ret = -EINVAL; 998 if (off >= src->i_size || off + len > src->i_size) 999 goto out_unlock; 1000 if (len == 0) 1001 olen = len = src->i_size - off; 1002 /* if we extend to eof, continue to block boundary */ 1003 if (off + len == src->i_size) 1004 len = ((src->i_size + bs-1) & ~(bs-1)) 1005 - off; 1006 1007 /* verify the end result is block aligned */ 1008 if ((off & (bs-1)) || 1009 ((off + len) & (bs-1))) 1010 goto out_unlock; 1011 1012 /* do any pending delalloc/csum calc on src, one way or 1013 another, and lock file content */ 1014 while (1) { 1015 struct btrfs_ordered_extent *ordered; 1016 lock_extent(&BTRFS_I(src)->io_tree, off, off+len, GFP_NOFS); 1017 ordered = btrfs_lookup_first_ordered_extent(inode, off+len); 1018 if (BTRFS_I(src)->delalloc_bytes == 0 && !ordered) 1019 break; 1020 unlock_extent(&BTRFS_I(src)->io_tree, off, off+len, GFP_NOFS); 1021 if (ordered) 1022 btrfs_put_ordered_extent(ordered); 1023 btrfs_wait_ordered_range(src, off, off+len); 1024 } 1025 1026 trans = btrfs_start_transaction(root, 1); 1027 BUG_ON(!trans); 1028 1029 /* punch hole in destination first */ 1030 btrfs_drop_extents(trans, root, inode, off, off + len, 1031 off + len, 0, &hint_byte, 1); 1032 1033 /* clone data */ 1034 key.objectid = src->i_ino; 1035 key.type = BTRFS_EXTENT_DATA_KEY; 1036 key.offset = 0; 1037 1038 while (1) { 1039 /* 1040 * note the key will change type as we walk through the 1041 * tree. 1042 */ 1043 ret = btrfs_search_slot(trans, root, &key, path, 0, 0); 1044 if (ret < 0) 1045 goto out; 1046 1047 nritems = btrfs_header_nritems(path->nodes[0]); 1048 if (path->slots[0] >= nritems) { 1049 ret = btrfs_next_leaf(root, path); 1050 if (ret < 0) 1051 goto out; 1052 if (ret > 0) 1053 break; 1054 nritems = btrfs_header_nritems(path->nodes[0]); 1055 } 1056 leaf = path->nodes[0]; 1057 slot = path->slots[0]; 1058 1059 btrfs_item_key_to_cpu(leaf, &key, slot); 1060 if (btrfs_key_type(&key) > BTRFS_EXTENT_DATA_KEY || 1061 key.objectid != src->i_ino) 1062 break; 1063 1064 if (btrfs_key_type(&key) == BTRFS_EXTENT_DATA_KEY) { 1065 struct btrfs_file_extent_item *extent; 1066 int type; 1067 u32 size; 1068 struct btrfs_key new_key; 1069 u64 disko = 0, diskl = 0; 1070 u64 datao = 0, datal = 0; 1071 u8 comp; 1072 1073 size = btrfs_item_size_nr(leaf, slot); 1074 read_extent_buffer(leaf, buf, 1075 btrfs_item_ptr_offset(leaf, slot), 1076 size); 1077 1078 extent = btrfs_item_ptr(leaf, slot, 1079 struct btrfs_file_extent_item); 1080 comp = btrfs_file_extent_compression(leaf, extent); 1081 type = btrfs_file_extent_type(leaf, extent); 1082 if (type == BTRFS_FILE_EXTENT_REG || 1083 type == BTRFS_FILE_EXTENT_PREALLOC) { 1084 disko = btrfs_file_extent_disk_bytenr(leaf, 1085 extent); 1086 diskl = btrfs_file_extent_disk_num_bytes(leaf, 1087 extent); 1088 datao = btrfs_file_extent_offset(leaf, extent); 1089 datal = btrfs_file_extent_num_bytes(leaf, 1090 extent); 1091 } else if (type == BTRFS_FILE_EXTENT_INLINE) { 1092 /* take upper bound, may be compressed */ 1093 datal = btrfs_file_extent_ram_bytes(leaf, 1094 extent); 1095 } 1096 btrfs_release_path(root, path); 1097 1098 if (key.offset + datal < off || 1099 key.offset >= off+len) 1100 goto next; 1101 1102 memcpy(&new_key, &key, sizeof(new_key)); 1103 new_key.objectid = inode->i_ino; 1104 new_key.offset = key.offset + destoff - off; 1105 1106 if (type == BTRFS_FILE_EXTENT_REG || 1107 type == BTRFS_FILE_EXTENT_PREALLOC) { 1108 ret = btrfs_insert_empty_item(trans, root, path, 1109 &new_key, size); 1110 if (ret) 1111 goto out; 1112 1113 leaf = path->nodes[0]; 1114 slot = path->slots[0]; 1115 write_extent_buffer(leaf, buf, 1116 btrfs_item_ptr_offset(leaf, slot), 1117 size); 1118 1119 extent = btrfs_item_ptr(leaf, slot, 1120 struct btrfs_file_extent_item); 1121 1122 if (off > key.offset) { 1123 datao += off - key.offset; 1124 datal -= off - key.offset; 1125 } 1126 1127 if (key.offset + datal > off + len) 1128 datal = off + len - key.offset; 1129 1130 /* disko == 0 means it's a hole */ 1131 if (!disko) 1132 datao = 0; 1133 1134 btrfs_set_file_extent_offset(leaf, extent, 1135 datao); 1136 btrfs_set_file_extent_num_bytes(leaf, extent, 1137 datal); 1138 if (disko) { 1139 inode_add_bytes(inode, datal); 1140 ret = btrfs_inc_extent_ref(trans, root, 1141 disko, diskl, 0, 1142 root->root_key.objectid, 1143 inode->i_ino, 1144 new_key.offset - datao); 1145 BUG_ON(ret); 1146 } 1147 } else if (type == BTRFS_FILE_EXTENT_INLINE) { 1148 u64 skip = 0; 1149 u64 trim = 0; 1150 if (off > key.offset) { 1151 skip = off - key.offset; 1152 new_key.offset += skip; 1153 } 1154 1155 if (key.offset + datal > off+len) 1156 trim = key.offset + datal - (off+len); 1157 1158 if (comp && (skip || trim)) { 1159 ret = -EINVAL; 1160 goto out; 1161 } 1162 size -= skip + trim; 1163 datal -= skip + trim; 1164 ret = btrfs_insert_empty_item(trans, root, path, 1165 &new_key, size); 1166 if (ret) 1167 goto out; 1168 1169 if (skip) { 1170 u32 start = 1171 btrfs_file_extent_calc_inline_size(0); 1172 memmove(buf+start, buf+start+skip, 1173 datal); 1174 } 1175 1176 leaf = path->nodes[0]; 1177 slot = path->slots[0]; 1178 write_extent_buffer(leaf, buf, 1179 btrfs_item_ptr_offset(leaf, slot), 1180 size); 1181 inode_add_bytes(inode, datal); 1182 } 1183 1184 btrfs_mark_buffer_dirty(leaf); 1185 } 1186 1187 next: 1188 btrfs_release_path(root, path); 1189 key.offset++; 1190 } 1191 ret = 0; 1192 out: 1193 btrfs_release_path(root, path); 1194 if (ret == 0) { 1195 inode->i_mtime = inode->i_ctime = CURRENT_TIME; 1196 if (destoff + olen > inode->i_size) 1197 btrfs_i_size_write(inode, destoff + olen); 1198 BTRFS_I(inode)->flags = BTRFS_I(src)->flags; 1199 ret = btrfs_update_inode(trans, root, inode); 1200 } 1201 btrfs_end_transaction(trans, root); 1202 unlock_extent(&BTRFS_I(src)->io_tree, off, off+len, GFP_NOFS); 1203 if (ret) 1204 vmtruncate(inode, 0); 1205 out_unlock: 1206 mutex_unlock(&src->i_mutex); 1207 mutex_unlock(&inode->i_mutex); 1208 vfree(buf); 1209 btrfs_free_path(path); 1210 out_fput: 1211 fput(src_file); 1212 out_drop_write: 1213 mnt_drop_write(file->f_path.mnt); 1214 return ret; 1215 } 1216 1217 static long btrfs_ioctl_clone_range(struct file *file, void __user *argp) 1218 { 1219 struct btrfs_ioctl_clone_range_args args; 1220 1221 if (copy_from_user(&args, argp, sizeof(args))) 1222 return -EFAULT; 1223 return btrfs_ioctl_clone(file, args.src_fd, args.src_offset, 1224 args.src_length, args.dest_offset); 1225 } 1226 1227 /* 1228 * there are many ways the trans_start and trans_end ioctls can lead 1229 * to deadlocks. They should only be used by applications that 1230 * basically own the machine, and have a very in depth understanding 1231 * of all the possible deadlocks and enospc problems. 1232 */ 1233 static long btrfs_ioctl_trans_start(struct file *file) 1234 { 1235 struct inode *inode = fdentry(file)->d_inode; 1236 struct btrfs_root *root = BTRFS_I(inode)->root; 1237 struct btrfs_trans_handle *trans; 1238 int ret; 1239 1240 ret = -EPERM; 1241 if (!capable(CAP_SYS_ADMIN)) 1242 goto out; 1243 1244 ret = -EINPROGRESS; 1245 if (file->private_data) 1246 goto out; 1247 1248 ret = mnt_want_write(file->f_path.mnt); 1249 if (ret) 1250 goto out; 1251 1252 mutex_lock(&root->fs_info->trans_mutex); 1253 root->fs_info->open_ioctl_trans++; 1254 mutex_unlock(&root->fs_info->trans_mutex); 1255 1256 ret = -ENOMEM; 1257 trans = btrfs_start_ioctl_transaction(root, 0); 1258 if (!trans) 1259 goto out_drop; 1260 1261 file->private_data = trans; 1262 return 0; 1263 1264 out_drop: 1265 mutex_lock(&root->fs_info->trans_mutex); 1266 root->fs_info->open_ioctl_trans--; 1267 mutex_unlock(&root->fs_info->trans_mutex); 1268 mnt_drop_write(file->f_path.mnt); 1269 out: 1270 return ret; 1271 } 1272 1273 /* 1274 * there are many ways the trans_start and trans_end ioctls can lead 1275 * to deadlocks. They should only be used by applications that 1276 * basically own the machine, and have a very in depth understanding 1277 * of all the possible deadlocks and enospc problems. 1278 */ 1279 long btrfs_ioctl_trans_end(struct file *file) 1280 { 1281 struct inode *inode = fdentry(file)->d_inode; 1282 struct btrfs_root *root = BTRFS_I(inode)->root; 1283 struct btrfs_trans_handle *trans; 1284 1285 trans = file->private_data; 1286 if (!trans) 1287 return -EINVAL; 1288 file->private_data = NULL; 1289 1290 btrfs_end_transaction(trans, root); 1291 1292 mutex_lock(&root->fs_info->trans_mutex); 1293 root->fs_info->open_ioctl_trans--; 1294 mutex_unlock(&root->fs_info->trans_mutex); 1295 1296 mnt_drop_write(file->f_path.mnt); 1297 return 0; 1298 } 1299 1300 long btrfs_ioctl(struct file *file, unsigned int 1301 cmd, unsigned long arg) 1302 { 1303 struct btrfs_root *root = BTRFS_I(fdentry(file)->d_inode)->root; 1304 void __user *argp = (void __user *)arg; 1305 1306 switch (cmd) { 1307 case FS_IOC_GETFLAGS: 1308 return btrfs_ioctl_getflags(file, argp); 1309 case FS_IOC_SETFLAGS: 1310 return btrfs_ioctl_setflags(file, argp); 1311 case FS_IOC_GETVERSION: 1312 return btrfs_ioctl_getversion(file, argp); 1313 case BTRFS_IOC_SNAP_CREATE: 1314 return btrfs_ioctl_snap_create(file, argp, 0); 1315 case BTRFS_IOC_SUBVOL_CREATE: 1316 return btrfs_ioctl_snap_create(file, argp, 1); 1317 case BTRFS_IOC_SNAP_DESTROY: 1318 return btrfs_ioctl_snap_destroy(file, argp); 1319 case BTRFS_IOC_DEFRAG: 1320 return btrfs_ioctl_defrag(file); 1321 case BTRFS_IOC_RESIZE: 1322 return btrfs_ioctl_resize(root, argp); 1323 case BTRFS_IOC_ADD_DEV: 1324 return btrfs_ioctl_add_dev(root, argp); 1325 case BTRFS_IOC_RM_DEV: 1326 return btrfs_ioctl_rm_dev(root, argp); 1327 case BTRFS_IOC_BALANCE: 1328 return btrfs_balance(root->fs_info->dev_root); 1329 case BTRFS_IOC_CLONE: 1330 return btrfs_ioctl_clone(file, arg, 0, 0, 0); 1331 case BTRFS_IOC_CLONE_RANGE: 1332 return btrfs_ioctl_clone_range(file, argp); 1333 case BTRFS_IOC_TRANS_START: 1334 return btrfs_ioctl_trans_start(file); 1335 case BTRFS_IOC_TRANS_END: 1336 return btrfs_ioctl_trans_end(file); 1337 case BTRFS_IOC_SYNC: 1338 btrfs_sync_fs(file->f_dentry->d_sb, 1); 1339 return 0; 1340 } 1341 1342 return -ENOTTY; 1343 } 1344