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/smp_lock.h> 31 #include <linux/backing-dev.h> 32 #include <linux/mount.h> 33 #include <linux/mpage.h> 34 #include <linux/namei.h> 35 #include <linux/swap.h> 36 #include <linux/writeback.h> 37 #include <linux/statfs.h> 38 #include <linux/compat.h> 39 #include <linux/bit_spinlock.h> 40 #include <linux/security.h> 41 #include <linux/xattr.h> 42 #include <linux/vmalloc.h> 43 #include "compat.h" 44 #include "ctree.h" 45 #include "disk-io.h" 46 #include "transaction.h" 47 #include "btrfs_inode.h" 48 #include "ioctl.h" 49 #include "print-tree.h" 50 #include "volumes.h" 51 #include "locking.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 = root; 235 struct inode *dir; 236 int ret; 237 int err; 238 u64 objectid; 239 u64 new_dirid = BTRFS_FIRST_FREE_OBJECTID; 240 u64 index = 0; 241 unsigned long nr = 1; 242 243 ret = btrfs_check_metadata_free_space(root); 244 if (ret) 245 goto fail_commit; 246 247 trans = btrfs_start_transaction(root, 1); 248 BUG_ON(!trans); 249 250 ret = btrfs_find_free_objectid(trans, root->fs_info->tree_root, 251 0, &objectid); 252 if (ret) 253 goto fail; 254 255 leaf = btrfs_alloc_free_block(trans, root, root->leafsize, 256 0, objectid, NULL, 0, 0, 0); 257 if (IS_ERR(leaf)) { 258 ret = PTR_ERR(leaf); 259 goto fail; 260 } 261 262 memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header)); 263 btrfs_set_header_bytenr(leaf, leaf->start); 264 btrfs_set_header_generation(leaf, trans->transid); 265 btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV); 266 btrfs_set_header_owner(leaf, objectid); 267 268 write_extent_buffer(leaf, root->fs_info->fsid, 269 (unsigned long)btrfs_header_fsid(leaf), 270 BTRFS_FSID_SIZE); 271 write_extent_buffer(leaf, root->fs_info->chunk_tree_uuid, 272 (unsigned long)btrfs_header_chunk_tree_uuid(leaf), 273 BTRFS_UUID_SIZE); 274 btrfs_mark_buffer_dirty(leaf); 275 276 inode_item = &root_item.inode; 277 memset(inode_item, 0, sizeof(*inode_item)); 278 inode_item->generation = cpu_to_le64(1); 279 inode_item->size = cpu_to_le64(3); 280 inode_item->nlink = cpu_to_le32(1); 281 inode_item->nbytes = cpu_to_le64(root->leafsize); 282 inode_item->mode = cpu_to_le32(S_IFDIR | 0755); 283 284 btrfs_set_root_bytenr(&root_item, leaf->start); 285 btrfs_set_root_generation(&root_item, trans->transid); 286 btrfs_set_root_level(&root_item, 0); 287 btrfs_set_root_refs(&root_item, 1); 288 btrfs_set_root_used(&root_item, 0); 289 btrfs_set_root_last_snapshot(&root_item, 0); 290 291 memset(&root_item.drop_progress, 0, sizeof(root_item.drop_progress)); 292 root_item.drop_level = 0; 293 294 btrfs_tree_unlock(leaf); 295 free_extent_buffer(leaf); 296 leaf = NULL; 297 298 btrfs_set_root_dirid(&root_item, new_dirid); 299 300 key.objectid = objectid; 301 key.offset = 0; 302 btrfs_set_key_type(&key, BTRFS_ROOT_ITEM_KEY); 303 ret = btrfs_insert_root(trans, root->fs_info->tree_root, &key, 304 &root_item); 305 if (ret) 306 goto fail; 307 308 /* 309 * insert the directory item 310 */ 311 key.offset = (u64)-1; 312 dir = dentry->d_parent->d_inode; 313 ret = btrfs_set_inode_index(dir, &index); 314 BUG_ON(ret); 315 316 ret = btrfs_insert_dir_item(trans, root, 317 name, namelen, dir->i_ino, &key, 318 BTRFS_FT_DIR, index); 319 if (ret) 320 goto fail; 321 322 btrfs_i_size_write(dir, dir->i_size + namelen * 2); 323 ret = btrfs_update_inode(trans, root, dir); 324 BUG_ON(ret); 325 326 /* add the backref first */ 327 ret = btrfs_add_root_ref(trans, root->fs_info->tree_root, 328 objectid, BTRFS_ROOT_BACKREF_KEY, 329 root->root_key.objectid, 330 dir->i_ino, index, name, namelen); 331 332 BUG_ON(ret); 333 334 /* now add the forward ref */ 335 ret = btrfs_add_root_ref(trans, root->fs_info->tree_root, 336 root->root_key.objectid, BTRFS_ROOT_REF_KEY, 337 objectid, 338 dir->i_ino, index, name, namelen); 339 340 BUG_ON(ret); 341 342 ret = btrfs_commit_transaction(trans, root); 343 if (ret) 344 goto fail_commit; 345 346 new_root = btrfs_read_fs_root_no_name(root->fs_info, &key); 347 BUG_ON(!new_root); 348 349 trans = btrfs_start_transaction(new_root, 1); 350 BUG_ON(!trans); 351 352 ret = btrfs_create_subvol_root(trans, new_root, dentry, new_dirid, 353 BTRFS_I(dir)->block_group); 354 if (ret) 355 goto fail; 356 357 fail: 358 nr = trans->blocks_used; 359 err = btrfs_commit_transaction(trans, new_root); 360 if (err && !ret) 361 ret = err; 362 fail_commit: 363 btrfs_btree_balance_dirty(root, nr); 364 return ret; 365 } 366 367 static int create_snapshot(struct btrfs_root *root, struct dentry *dentry, 368 char *name, int namelen) 369 { 370 struct btrfs_pending_snapshot *pending_snapshot; 371 struct btrfs_trans_handle *trans; 372 int ret = 0; 373 int err; 374 unsigned long nr = 0; 375 376 if (!root->ref_cows) 377 return -EINVAL; 378 379 ret = btrfs_check_metadata_free_space(root); 380 if (ret) 381 goto fail_unlock; 382 383 pending_snapshot = kzalloc(sizeof(*pending_snapshot), GFP_NOFS); 384 if (!pending_snapshot) { 385 ret = -ENOMEM; 386 goto fail_unlock; 387 } 388 pending_snapshot->name = kmalloc(namelen + 1, GFP_NOFS); 389 if (!pending_snapshot->name) { 390 ret = -ENOMEM; 391 kfree(pending_snapshot); 392 goto fail_unlock; 393 } 394 memcpy(pending_snapshot->name, name, namelen); 395 pending_snapshot->name[namelen] = '\0'; 396 pending_snapshot->dentry = dentry; 397 trans = btrfs_start_transaction(root, 1); 398 BUG_ON(!trans); 399 pending_snapshot->root = root; 400 list_add(&pending_snapshot->list, 401 &trans->transaction->pending_snapshots); 402 err = btrfs_commit_transaction(trans, root); 403 404 fail_unlock: 405 btrfs_btree_balance_dirty(root, nr); 406 return ret; 407 } 408 409 /* copy of may_create in fs/namei.c() */ 410 static inline int btrfs_may_create(struct inode *dir, struct dentry *child) 411 { 412 if (child->d_inode) 413 return -EEXIST; 414 if (IS_DEADDIR(dir)) 415 return -ENOENT; 416 return inode_permission(dir, MAY_WRITE | MAY_EXEC); 417 } 418 419 /* 420 * Create a new subvolume below @parent. This is largely modeled after 421 * sys_mkdirat and vfs_mkdir, but we only do a single component lookup 422 * inside this filesystem so it's quite a bit simpler. 423 */ 424 static noinline int btrfs_mksubvol(struct path *parent, char *name, 425 int mode, int namelen, 426 struct btrfs_root *snap_src) 427 { 428 struct dentry *dentry; 429 int error; 430 431 mutex_lock_nested(&parent->dentry->d_inode->i_mutex, I_MUTEX_PARENT); 432 433 dentry = lookup_one_len(name, parent->dentry, namelen); 434 error = PTR_ERR(dentry); 435 if (IS_ERR(dentry)) 436 goto out_unlock; 437 438 error = -EEXIST; 439 if (dentry->d_inode) 440 goto out_dput; 441 442 if (!IS_POSIXACL(parent->dentry->d_inode)) 443 mode &= ~current_umask(); 444 445 error = mnt_want_write(parent->mnt); 446 if (error) 447 goto out_dput; 448 449 error = btrfs_may_create(parent->dentry->d_inode, dentry); 450 if (error) 451 goto out_drop_write; 452 453 /* 454 * Actually perform the low-level subvolume creation after all 455 * this VFS fuzz. 456 * 457 * Eventually we want to pass in an inode under which we create this 458 * subvolume, but for now all are under the filesystem root. 459 * 460 * Also we should pass on the mode eventually to allow creating new 461 * subvolume with specific mode bits. 462 */ 463 if (snap_src) { 464 struct dentry *dir = dentry->d_parent; 465 struct dentry *test = dir->d_parent; 466 struct btrfs_path *path = btrfs_alloc_path(); 467 int ret; 468 u64 test_oid; 469 u64 parent_oid = BTRFS_I(dir->d_inode)->root->root_key.objectid; 470 471 test_oid = snap_src->root_key.objectid; 472 473 ret = btrfs_find_root_ref(snap_src->fs_info->tree_root, 474 path, parent_oid, test_oid); 475 if (ret == 0) 476 goto create; 477 btrfs_release_path(snap_src->fs_info->tree_root, path); 478 479 /* we need to make sure we aren't creating a directory loop 480 * by taking a snapshot of something that has our current 481 * subvol in its directory tree. So, this loops through 482 * the dentries and checks the forward refs for each subvolume 483 * to see if is references the subvolume where we are 484 * placing this new snapshot. 485 */ 486 while (1) { 487 if (!test || 488 dir == snap_src->fs_info->sb->s_root || 489 test == snap_src->fs_info->sb->s_root || 490 test->d_inode->i_sb != snap_src->fs_info->sb) { 491 break; 492 } 493 if (S_ISLNK(test->d_inode->i_mode)) { 494 printk(KERN_INFO "Btrfs symlink in snapshot " 495 "path, failed\n"); 496 error = -EMLINK; 497 btrfs_free_path(path); 498 goto out_drop_write; 499 } 500 test_oid = 501 BTRFS_I(test->d_inode)->root->root_key.objectid; 502 ret = btrfs_find_root_ref(snap_src->fs_info->tree_root, 503 path, test_oid, parent_oid); 504 if (ret == 0) { 505 printk(KERN_INFO "Btrfs snapshot creation " 506 "failed, looping\n"); 507 error = -EMLINK; 508 btrfs_free_path(path); 509 goto out_drop_write; 510 } 511 btrfs_release_path(snap_src->fs_info->tree_root, path); 512 test = test->d_parent; 513 } 514 create: 515 btrfs_free_path(path); 516 error = create_snapshot(snap_src, dentry, name, namelen); 517 } else { 518 error = create_subvol(BTRFS_I(parent->dentry->d_inode)->root, 519 dentry, name, namelen); 520 } 521 if (error) 522 goto out_drop_write; 523 524 fsnotify_mkdir(parent->dentry->d_inode, dentry); 525 out_drop_write: 526 mnt_drop_write(parent->mnt); 527 out_dput: 528 dput(dentry); 529 out_unlock: 530 mutex_unlock(&parent->dentry->d_inode->i_mutex); 531 return error; 532 } 533 534 535 static int btrfs_defrag_file(struct file *file) 536 { 537 struct inode *inode = fdentry(file)->d_inode; 538 struct btrfs_root *root = BTRFS_I(inode)->root; 539 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree; 540 struct btrfs_ordered_extent *ordered; 541 struct page *page; 542 unsigned long last_index; 543 unsigned long ra_pages = root->fs_info->bdi.ra_pages; 544 unsigned long total_read = 0; 545 u64 page_start; 546 u64 page_end; 547 unsigned long i; 548 int ret; 549 550 ret = btrfs_check_data_free_space(root, inode, inode->i_size); 551 if (ret) 552 return -ENOSPC; 553 554 mutex_lock(&inode->i_mutex); 555 last_index = inode->i_size >> PAGE_CACHE_SHIFT; 556 for (i = 0; i <= last_index; i++) { 557 if (total_read % ra_pages == 0) { 558 btrfs_force_ra(inode->i_mapping, &file->f_ra, file, i, 559 min(last_index, i + ra_pages - 1)); 560 } 561 total_read++; 562 again: 563 page = grab_cache_page(inode->i_mapping, i); 564 if (!page) 565 goto out_unlock; 566 if (!PageUptodate(page)) { 567 btrfs_readpage(NULL, page); 568 lock_page(page); 569 if (!PageUptodate(page)) { 570 unlock_page(page); 571 page_cache_release(page); 572 goto out_unlock; 573 } 574 } 575 576 wait_on_page_writeback(page); 577 578 page_start = (u64)page->index << PAGE_CACHE_SHIFT; 579 page_end = page_start + PAGE_CACHE_SIZE - 1; 580 lock_extent(io_tree, page_start, page_end, GFP_NOFS); 581 582 ordered = btrfs_lookup_ordered_extent(inode, page_start); 583 if (ordered) { 584 unlock_extent(io_tree, page_start, page_end, GFP_NOFS); 585 unlock_page(page); 586 page_cache_release(page); 587 btrfs_start_ordered_extent(inode, ordered, 1); 588 btrfs_put_ordered_extent(ordered); 589 goto again; 590 } 591 set_page_extent_mapped(page); 592 593 /* 594 * this makes sure page_mkwrite is called on the 595 * page if it is dirtied again later 596 */ 597 clear_page_dirty_for_io(page); 598 599 btrfs_set_extent_delalloc(inode, page_start, page_end); 600 601 unlock_extent(io_tree, page_start, page_end, GFP_NOFS); 602 set_page_dirty(page); 603 unlock_page(page); 604 page_cache_release(page); 605 balance_dirty_pages_ratelimited_nr(inode->i_mapping, 1); 606 } 607 608 out_unlock: 609 mutex_unlock(&inode->i_mutex); 610 return 0; 611 } 612 613 static int btrfs_ioctl_resize(struct btrfs_root *root, void __user *arg) 614 { 615 u64 new_size; 616 u64 old_size; 617 u64 devid = 1; 618 struct btrfs_ioctl_vol_args *vol_args; 619 struct btrfs_trans_handle *trans; 620 struct btrfs_device *device = NULL; 621 char *sizestr; 622 char *devstr = NULL; 623 int ret = 0; 624 int namelen; 625 int mod = 0; 626 627 if (root->fs_info->sb->s_flags & MS_RDONLY) 628 return -EROFS; 629 630 if (!capable(CAP_SYS_ADMIN)) 631 return -EPERM; 632 633 vol_args = memdup_user(arg, sizeof(*vol_args)); 634 if (IS_ERR(vol_args)) 635 return PTR_ERR(vol_args); 636 637 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; 638 namelen = strlen(vol_args->name); 639 640 mutex_lock(&root->fs_info->volume_mutex); 641 sizestr = vol_args->name; 642 devstr = strchr(sizestr, ':'); 643 if (devstr) { 644 char *end; 645 sizestr = devstr + 1; 646 *devstr = '\0'; 647 devstr = vol_args->name; 648 devid = simple_strtoull(devstr, &end, 10); 649 printk(KERN_INFO "resizing devid %llu\n", 650 (unsigned long long)devid); 651 } 652 device = btrfs_find_device(root, devid, NULL, NULL); 653 if (!device) { 654 printk(KERN_INFO "resizer unable to find device %llu\n", 655 (unsigned long long)devid); 656 ret = -EINVAL; 657 goto out_unlock; 658 } 659 if (!strcmp(sizestr, "max")) 660 new_size = device->bdev->bd_inode->i_size; 661 else { 662 if (sizestr[0] == '-') { 663 mod = -1; 664 sizestr++; 665 } else if (sizestr[0] == '+') { 666 mod = 1; 667 sizestr++; 668 } 669 new_size = btrfs_parse_size(sizestr); 670 if (new_size == 0) { 671 ret = -EINVAL; 672 goto out_unlock; 673 } 674 } 675 676 old_size = device->total_bytes; 677 678 if (mod < 0) { 679 if (new_size > old_size) { 680 ret = -EINVAL; 681 goto out_unlock; 682 } 683 new_size = old_size - new_size; 684 } else if (mod > 0) { 685 new_size = old_size + new_size; 686 } 687 688 if (new_size < 256 * 1024 * 1024) { 689 ret = -EINVAL; 690 goto out_unlock; 691 } 692 if (new_size > device->bdev->bd_inode->i_size) { 693 ret = -EFBIG; 694 goto out_unlock; 695 } 696 697 do_div(new_size, root->sectorsize); 698 new_size *= root->sectorsize; 699 700 printk(KERN_INFO "new size for %s is %llu\n", 701 device->name, (unsigned long long)new_size); 702 703 if (new_size > old_size) { 704 trans = btrfs_start_transaction(root, 1); 705 ret = btrfs_grow_device(trans, device, new_size); 706 btrfs_commit_transaction(trans, root); 707 } else { 708 ret = btrfs_shrink_device(device, new_size); 709 } 710 711 out_unlock: 712 mutex_unlock(&root->fs_info->volume_mutex); 713 kfree(vol_args); 714 return ret; 715 } 716 717 static noinline int btrfs_ioctl_snap_create(struct file *file, 718 void __user *arg, int subvol) 719 { 720 struct btrfs_root *root = BTRFS_I(fdentry(file)->d_inode)->root; 721 struct btrfs_ioctl_vol_args *vol_args; 722 struct btrfs_dir_item *di; 723 struct btrfs_path *path; 724 struct file *src_file; 725 u64 root_dirid; 726 int namelen; 727 int ret = 0; 728 729 if (root->fs_info->sb->s_flags & MS_RDONLY) 730 return -EROFS; 731 732 vol_args = memdup_user(arg, sizeof(*vol_args)); 733 if (IS_ERR(vol_args)) 734 return PTR_ERR(vol_args); 735 736 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; 737 namelen = strlen(vol_args->name); 738 if (strchr(vol_args->name, '/')) { 739 ret = -EINVAL; 740 goto out; 741 } 742 743 path = btrfs_alloc_path(); 744 if (!path) { 745 ret = -ENOMEM; 746 goto out; 747 } 748 749 root_dirid = root->fs_info->sb->s_root->d_inode->i_ino, 750 di = btrfs_lookup_dir_item(NULL, root->fs_info->tree_root, 751 path, root_dirid, 752 vol_args->name, namelen, 0); 753 btrfs_free_path(path); 754 755 if (di && !IS_ERR(di)) { 756 ret = -EEXIST; 757 goto out; 758 } 759 760 if (IS_ERR(di)) { 761 ret = PTR_ERR(di); 762 goto out; 763 } 764 765 if (subvol) { 766 ret = btrfs_mksubvol(&file->f_path, vol_args->name, 767 file->f_path.dentry->d_inode->i_mode, 768 namelen, NULL); 769 } else { 770 struct inode *src_inode; 771 src_file = fget(vol_args->fd); 772 if (!src_file) { 773 ret = -EINVAL; 774 goto out; 775 } 776 777 src_inode = src_file->f_path.dentry->d_inode; 778 if (src_inode->i_sb != file->f_path.dentry->d_inode->i_sb) { 779 printk(KERN_INFO "btrfs: Snapshot src from " 780 "another FS\n"); 781 ret = -EINVAL; 782 fput(src_file); 783 goto out; 784 } 785 ret = btrfs_mksubvol(&file->f_path, vol_args->name, 786 file->f_path.dentry->d_inode->i_mode, 787 namelen, BTRFS_I(src_inode)->root); 788 fput(src_file); 789 } 790 791 out: 792 kfree(vol_args); 793 return ret; 794 } 795 796 static int btrfs_ioctl_defrag(struct file *file) 797 { 798 struct inode *inode = fdentry(file)->d_inode; 799 struct btrfs_root *root = BTRFS_I(inode)->root; 800 int ret; 801 802 ret = mnt_want_write(file->f_path.mnt); 803 if (ret) 804 return ret; 805 806 switch (inode->i_mode & S_IFMT) { 807 case S_IFDIR: 808 if (!capable(CAP_SYS_ADMIN)) { 809 ret = -EPERM; 810 goto out; 811 } 812 btrfs_defrag_root(root, 0); 813 btrfs_defrag_root(root->fs_info->extent_root, 0); 814 break; 815 case S_IFREG: 816 if (!(file->f_mode & FMODE_WRITE)) { 817 ret = -EINVAL; 818 goto out; 819 } 820 btrfs_defrag_file(file); 821 break; 822 } 823 out: 824 mnt_drop_write(file->f_path.mnt); 825 return ret; 826 } 827 828 static long btrfs_ioctl_add_dev(struct btrfs_root *root, void __user *arg) 829 { 830 struct btrfs_ioctl_vol_args *vol_args; 831 int ret; 832 833 if (!capable(CAP_SYS_ADMIN)) 834 return -EPERM; 835 836 vol_args = memdup_user(arg, sizeof(*vol_args)); 837 if (IS_ERR(vol_args)) 838 return PTR_ERR(vol_args); 839 840 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; 841 ret = btrfs_init_new_device(root, vol_args->name); 842 843 kfree(vol_args); 844 return ret; 845 } 846 847 static long btrfs_ioctl_rm_dev(struct btrfs_root *root, void __user *arg) 848 { 849 struct btrfs_ioctl_vol_args *vol_args; 850 int ret; 851 852 if (!capable(CAP_SYS_ADMIN)) 853 return -EPERM; 854 855 if (root->fs_info->sb->s_flags & MS_RDONLY) 856 return -EROFS; 857 858 vol_args = memdup_user(arg, sizeof(*vol_args)); 859 if (IS_ERR(vol_args)) 860 return PTR_ERR(vol_args); 861 862 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; 863 ret = btrfs_rm_device(root, vol_args->name); 864 865 kfree(vol_args); 866 return ret; 867 } 868 869 static long btrfs_ioctl_clone(struct file *file, unsigned long srcfd, 870 u64 off, u64 olen, u64 destoff) 871 { 872 struct inode *inode = fdentry(file)->d_inode; 873 struct btrfs_root *root = BTRFS_I(inode)->root; 874 struct file *src_file; 875 struct inode *src; 876 struct btrfs_trans_handle *trans; 877 struct btrfs_path *path; 878 struct extent_buffer *leaf; 879 char *buf; 880 struct btrfs_key key; 881 u32 nritems; 882 int slot; 883 int ret; 884 u64 len = olen; 885 u64 bs = root->fs_info->sb->s_blocksize; 886 u64 hint_byte; 887 888 /* 889 * TODO: 890 * - split compressed inline extents. annoying: we need to 891 * decompress into destination's address_space (the file offset 892 * may change, so source mapping won't do), then recompress (or 893 * otherwise reinsert) a subrange. 894 * - allow ranges within the same file to be cloned (provided 895 * they don't overlap)? 896 */ 897 898 /* the destination must be opened for writing */ 899 if (!(file->f_mode & FMODE_WRITE)) 900 return -EINVAL; 901 902 ret = mnt_want_write(file->f_path.mnt); 903 if (ret) 904 return ret; 905 906 src_file = fget(srcfd); 907 if (!src_file) { 908 ret = -EBADF; 909 goto out_drop_write; 910 } 911 src = src_file->f_dentry->d_inode; 912 913 ret = -EINVAL; 914 if (src == inode) 915 goto out_fput; 916 917 ret = -EISDIR; 918 if (S_ISDIR(src->i_mode) || S_ISDIR(inode->i_mode)) 919 goto out_fput; 920 921 ret = -EXDEV; 922 if (src->i_sb != inode->i_sb || BTRFS_I(src)->root != root) 923 goto out_fput; 924 925 ret = -ENOMEM; 926 buf = vmalloc(btrfs_level_size(root, 0)); 927 if (!buf) 928 goto out_fput; 929 930 path = btrfs_alloc_path(); 931 if (!path) { 932 vfree(buf); 933 goto out_fput; 934 } 935 path->reada = 2; 936 937 if (inode < src) { 938 mutex_lock(&inode->i_mutex); 939 mutex_lock(&src->i_mutex); 940 } else { 941 mutex_lock(&src->i_mutex); 942 mutex_lock(&inode->i_mutex); 943 } 944 945 /* determine range to clone */ 946 ret = -EINVAL; 947 if (off >= src->i_size || off + len > src->i_size) 948 goto out_unlock; 949 if (len == 0) 950 olen = len = src->i_size - off; 951 /* if we extend to eof, continue to block boundary */ 952 if (off + len == src->i_size) 953 len = ((src->i_size + bs-1) & ~(bs-1)) 954 - off; 955 956 /* verify the end result is block aligned */ 957 if ((off & (bs-1)) || 958 ((off + len) & (bs-1))) 959 goto out_unlock; 960 961 /* do any pending delalloc/csum calc on src, one way or 962 another, and lock file content */ 963 while (1) { 964 struct btrfs_ordered_extent *ordered; 965 lock_extent(&BTRFS_I(src)->io_tree, off, off+len, GFP_NOFS); 966 ordered = btrfs_lookup_first_ordered_extent(inode, off+len); 967 if (BTRFS_I(src)->delalloc_bytes == 0 && !ordered) 968 break; 969 unlock_extent(&BTRFS_I(src)->io_tree, off, off+len, GFP_NOFS); 970 if (ordered) 971 btrfs_put_ordered_extent(ordered); 972 btrfs_wait_ordered_range(src, off, off+len); 973 } 974 975 trans = btrfs_start_transaction(root, 1); 976 BUG_ON(!trans); 977 978 /* punch hole in destination first */ 979 btrfs_drop_extents(trans, root, inode, off, off + len, 980 off + len, 0, &hint_byte); 981 982 /* clone data */ 983 key.objectid = src->i_ino; 984 key.type = BTRFS_EXTENT_DATA_KEY; 985 key.offset = 0; 986 987 while (1) { 988 /* 989 * note the key will change type as we walk through the 990 * tree. 991 */ 992 ret = btrfs_search_slot(trans, root, &key, path, 0, 0); 993 if (ret < 0) 994 goto out; 995 996 nritems = btrfs_header_nritems(path->nodes[0]); 997 if (path->slots[0] >= nritems) { 998 ret = btrfs_next_leaf(root, path); 999 if (ret < 0) 1000 goto out; 1001 if (ret > 0) 1002 break; 1003 nritems = btrfs_header_nritems(path->nodes[0]); 1004 } 1005 leaf = path->nodes[0]; 1006 slot = path->slots[0]; 1007 1008 btrfs_item_key_to_cpu(leaf, &key, slot); 1009 if (btrfs_key_type(&key) > BTRFS_EXTENT_DATA_KEY || 1010 key.objectid != src->i_ino) 1011 break; 1012 1013 if (btrfs_key_type(&key) == BTRFS_EXTENT_DATA_KEY) { 1014 struct btrfs_file_extent_item *extent; 1015 int type; 1016 u32 size; 1017 struct btrfs_key new_key; 1018 u64 disko = 0, diskl = 0; 1019 u64 datao = 0, datal = 0; 1020 u8 comp; 1021 1022 size = btrfs_item_size_nr(leaf, slot); 1023 read_extent_buffer(leaf, buf, 1024 btrfs_item_ptr_offset(leaf, slot), 1025 size); 1026 1027 extent = btrfs_item_ptr(leaf, slot, 1028 struct btrfs_file_extent_item); 1029 comp = btrfs_file_extent_compression(leaf, extent); 1030 type = btrfs_file_extent_type(leaf, extent); 1031 if (type == BTRFS_FILE_EXTENT_REG) { 1032 disko = btrfs_file_extent_disk_bytenr(leaf, 1033 extent); 1034 diskl = btrfs_file_extent_disk_num_bytes(leaf, 1035 extent); 1036 datao = btrfs_file_extent_offset(leaf, extent); 1037 datal = btrfs_file_extent_num_bytes(leaf, 1038 extent); 1039 } else if (type == BTRFS_FILE_EXTENT_INLINE) { 1040 /* take upper bound, may be compressed */ 1041 datal = btrfs_file_extent_ram_bytes(leaf, 1042 extent); 1043 } 1044 btrfs_release_path(root, path); 1045 1046 if (key.offset + datal < off || 1047 key.offset >= off+len) 1048 goto next; 1049 1050 memcpy(&new_key, &key, sizeof(new_key)); 1051 new_key.objectid = inode->i_ino; 1052 new_key.offset = key.offset + destoff - off; 1053 1054 if (type == BTRFS_FILE_EXTENT_REG) { 1055 ret = btrfs_insert_empty_item(trans, root, path, 1056 &new_key, size); 1057 if (ret) 1058 goto out; 1059 1060 leaf = path->nodes[0]; 1061 slot = path->slots[0]; 1062 write_extent_buffer(leaf, buf, 1063 btrfs_item_ptr_offset(leaf, slot), 1064 size); 1065 1066 extent = btrfs_item_ptr(leaf, slot, 1067 struct btrfs_file_extent_item); 1068 1069 if (off > key.offset) { 1070 datao += off - key.offset; 1071 datal -= off - key.offset; 1072 } 1073 if (key.offset + datao + datal + key.offset > 1074 off + len) 1075 datal = off + len - key.offset - datao; 1076 /* disko == 0 means it's a hole */ 1077 if (!disko) 1078 datao = 0; 1079 1080 btrfs_set_file_extent_offset(leaf, extent, 1081 datao); 1082 btrfs_set_file_extent_num_bytes(leaf, extent, 1083 datal); 1084 if (disko) { 1085 inode_add_bytes(inode, datal); 1086 ret = btrfs_inc_extent_ref(trans, root, 1087 disko, diskl, 0, 1088 root->root_key.objectid, 1089 inode->i_ino, 1090 new_key.offset - datao); 1091 BUG_ON(ret); 1092 } 1093 } else if (type == BTRFS_FILE_EXTENT_INLINE) { 1094 u64 skip = 0; 1095 u64 trim = 0; 1096 if (off > key.offset) { 1097 skip = off - key.offset; 1098 new_key.offset += skip; 1099 } 1100 1101 if (key.offset + datal > off+len) 1102 trim = key.offset + datal - (off+len); 1103 1104 if (comp && (skip || trim)) { 1105 ret = -EINVAL; 1106 goto out; 1107 } 1108 size -= skip + trim; 1109 datal -= skip + trim; 1110 ret = btrfs_insert_empty_item(trans, root, path, 1111 &new_key, size); 1112 if (ret) 1113 goto out; 1114 1115 if (skip) { 1116 u32 start = 1117 btrfs_file_extent_calc_inline_size(0); 1118 memmove(buf+start, buf+start+skip, 1119 datal); 1120 } 1121 1122 leaf = path->nodes[0]; 1123 slot = path->slots[0]; 1124 write_extent_buffer(leaf, buf, 1125 btrfs_item_ptr_offset(leaf, slot), 1126 size); 1127 inode_add_bytes(inode, datal); 1128 } 1129 1130 btrfs_mark_buffer_dirty(leaf); 1131 } 1132 1133 next: 1134 btrfs_release_path(root, path); 1135 key.offset++; 1136 } 1137 ret = 0; 1138 out: 1139 btrfs_release_path(root, path); 1140 if (ret == 0) { 1141 inode->i_mtime = inode->i_ctime = CURRENT_TIME; 1142 if (destoff + olen > inode->i_size) 1143 btrfs_i_size_write(inode, destoff + olen); 1144 BTRFS_I(inode)->flags = BTRFS_I(src)->flags; 1145 ret = btrfs_update_inode(trans, root, inode); 1146 } 1147 btrfs_end_transaction(trans, root); 1148 unlock_extent(&BTRFS_I(src)->io_tree, off, off+len, GFP_NOFS); 1149 if (ret) 1150 vmtruncate(inode, 0); 1151 out_unlock: 1152 mutex_unlock(&src->i_mutex); 1153 mutex_unlock(&inode->i_mutex); 1154 vfree(buf); 1155 btrfs_free_path(path); 1156 out_fput: 1157 fput(src_file); 1158 out_drop_write: 1159 mnt_drop_write(file->f_path.mnt); 1160 return ret; 1161 } 1162 1163 static long btrfs_ioctl_clone_range(struct file *file, void __user *argp) 1164 { 1165 struct btrfs_ioctl_clone_range_args args; 1166 1167 if (copy_from_user(&args, argp, sizeof(args))) 1168 return -EFAULT; 1169 return btrfs_ioctl_clone(file, args.src_fd, args.src_offset, 1170 args.src_length, args.dest_offset); 1171 } 1172 1173 /* 1174 * there are many ways the trans_start and trans_end ioctls can lead 1175 * to deadlocks. They should only be used by applications that 1176 * basically own the machine, and have a very in depth understanding 1177 * of all the possible deadlocks and enospc problems. 1178 */ 1179 static long btrfs_ioctl_trans_start(struct file *file) 1180 { 1181 struct inode *inode = fdentry(file)->d_inode; 1182 struct btrfs_root *root = BTRFS_I(inode)->root; 1183 struct btrfs_trans_handle *trans; 1184 int ret = 0; 1185 1186 if (!capable(CAP_SYS_ADMIN)) 1187 return -EPERM; 1188 1189 if (file->private_data) { 1190 ret = -EINPROGRESS; 1191 goto out; 1192 } 1193 1194 ret = mnt_want_write(file->f_path.mnt); 1195 if (ret) 1196 goto out; 1197 1198 mutex_lock(&root->fs_info->trans_mutex); 1199 root->fs_info->open_ioctl_trans++; 1200 mutex_unlock(&root->fs_info->trans_mutex); 1201 1202 trans = btrfs_start_ioctl_transaction(root, 0); 1203 if (trans) 1204 file->private_data = trans; 1205 else 1206 ret = -ENOMEM; 1207 /*printk(KERN_INFO "btrfs_ioctl_trans_start on %p\n", file);*/ 1208 out: 1209 return ret; 1210 } 1211 1212 /* 1213 * there are many ways the trans_start and trans_end ioctls can lead 1214 * to deadlocks. They should only be used by applications that 1215 * basically own the machine, and have a very in depth understanding 1216 * of all the possible deadlocks and enospc problems. 1217 */ 1218 long btrfs_ioctl_trans_end(struct file *file) 1219 { 1220 struct inode *inode = fdentry(file)->d_inode; 1221 struct btrfs_root *root = BTRFS_I(inode)->root; 1222 struct btrfs_trans_handle *trans; 1223 int ret = 0; 1224 1225 trans = file->private_data; 1226 if (!trans) { 1227 ret = -EINVAL; 1228 goto out; 1229 } 1230 btrfs_end_transaction(trans, root); 1231 file->private_data = NULL; 1232 1233 mutex_lock(&root->fs_info->trans_mutex); 1234 root->fs_info->open_ioctl_trans--; 1235 mutex_unlock(&root->fs_info->trans_mutex); 1236 1237 mnt_drop_write(file->f_path.mnt); 1238 1239 out: 1240 return ret; 1241 } 1242 1243 long btrfs_ioctl(struct file *file, unsigned int 1244 cmd, unsigned long arg) 1245 { 1246 struct btrfs_root *root = BTRFS_I(fdentry(file)->d_inode)->root; 1247 void __user *argp = (void __user *)arg; 1248 1249 switch (cmd) { 1250 case FS_IOC_GETFLAGS: 1251 return btrfs_ioctl_getflags(file, argp); 1252 case FS_IOC_SETFLAGS: 1253 return btrfs_ioctl_setflags(file, argp); 1254 case FS_IOC_GETVERSION: 1255 return btrfs_ioctl_getversion(file, argp); 1256 case BTRFS_IOC_SNAP_CREATE: 1257 return btrfs_ioctl_snap_create(file, argp, 0); 1258 case BTRFS_IOC_SUBVOL_CREATE: 1259 return btrfs_ioctl_snap_create(file, argp, 1); 1260 case BTRFS_IOC_DEFRAG: 1261 return btrfs_ioctl_defrag(file); 1262 case BTRFS_IOC_RESIZE: 1263 return btrfs_ioctl_resize(root, argp); 1264 case BTRFS_IOC_ADD_DEV: 1265 return btrfs_ioctl_add_dev(root, argp); 1266 case BTRFS_IOC_RM_DEV: 1267 return btrfs_ioctl_rm_dev(root, argp); 1268 case BTRFS_IOC_BALANCE: 1269 return btrfs_balance(root->fs_info->dev_root); 1270 case BTRFS_IOC_CLONE: 1271 return btrfs_ioctl_clone(file, arg, 0, 0, 0); 1272 case BTRFS_IOC_CLONE_RANGE: 1273 return btrfs_ioctl_clone_range(file, argp); 1274 case BTRFS_IOC_TRANS_START: 1275 return btrfs_ioctl_trans_start(file); 1276 case BTRFS_IOC_TRANS_END: 1277 return btrfs_ioctl_trans_end(file); 1278 case BTRFS_IOC_SYNC: 1279 btrfs_sync_fs(file->f_dentry->d_sb, 1); 1280 return 0; 1281 } 1282 1283 return -ENOTTY; 1284 } 1285