1 /* 2 * Copyright (C) 2007 Oracle. All rights reserved. 3 * 4 * This program is free software; you can redistribute it and/or 5 * modify it under the terms of the GNU General Public 6 * License v2 as published by the Free Software Foundation. 7 * 8 * This program is distributed in the hope that it will be useful, 9 * but WITHOUT ANY WARRANTY; without even the implied warranty of 10 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU 11 * General Public License for more details. 12 * 13 * You should have received a copy of the GNU General Public 14 * License along with this program; if not, write to the 15 * Free Software Foundation, Inc., 59 Temple Place - Suite 330, 16 * Boston, MA 021110-1307, USA. 17 */ 18 19 #include <linux/kernel.h> 20 #include <linux/bio.h> 21 #include <linux/buffer_head.h> 22 #include <linux/file.h> 23 #include <linux/fs.h> 24 #include <linux/fsnotify.h> 25 #include <linux/pagemap.h> 26 #include <linux/highmem.h> 27 #include <linux/time.h> 28 #include <linux/init.h> 29 #include <linux/string.h> 30 #include <linux/backing-dev.h> 31 #include <linux/mount.h> 32 #include <linux/mpage.h> 33 #include <linux/namei.h> 34 #include <linux/swap.h> 35 #include <linux/writeback.h> 36 #include <linux/statfs.h> 37 #include <linux/compat.h> 38 #include <linux/bit_spinlock.h> 39 #include <linux/security.h> 40 #include <linux/xattr.h> 41 #include <linux/vmalloc.h> 42 #include <linux/slab.h> 43 #include "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; 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 should_defrag_range(struct inode *inode, u64 start, u64 len, 479 int thresh, u64 *last_len, u64 *skip, 480 u64 *defrag_end) 481 { 482 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree; 483 struct extent_map *em = NULL; 484 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree; 485 int ret = 1; 486 487 488 if (thresh == 0) 489 thresh = 256 * 1024; 490 491 /* 492 * make sure that once we start defragging and extent, we keep on 493 * defragging it 494 */ 495 if (start < *defrag_end) 496 return 1; 497 498 *skip = 0; 499 500 /* 501 * hopefully we have this extent in the tree already, try without 502 * the full extent lock 503 */ 504 read_lock(&em_tree->lock); 505 em = lookup_extent_mapping(em_tree, start, len); 506 read_unlock(&em_tree->lock); 507 508 if (!em) { 509 /* get the big lock and read metadata off disk */ 510 lock_extent(io_tree, start, start + len - 1, GFP_NOFS); 511 em = btrfs_get_extent(inode, NULL, 0, start, len, 0); 512 unlock_extent(io_tree, start, start + len - 1, GFP_NOFS); 513 514 if (IS_ERR(em)) 515 return 0; 516 } 517 518 /* this will cover holes, and inline extents */ 519 if (em->block_start >= EXTENT_MAP_LAST_BYTE) 520 ret = 0; 521 522 /* 523 * we hit a real extent, if it is big don't bother defragging it again 524 */ 525 if ((*last_len == 0 || *last_len >= thresh) && em->len >= thresh) 526 ret = 0; 527 528 /* 529 * last_len ends up being a counter of how many bytes we've defragged. 530 * every time we choose not to defrag an extent, we reset *last_len 531 * so that the next tiny extent will force a defrag. 532 * 533 * The end result of this is that tiny extents before a single big 534 * extent will force at least part of that big extent to be defragged. 535 */ 536 if (ret) { 537 *last_len += len; 538 *defrag_end = extent_map_end(em); 539 } else { 540 *last_len = 0; 541 *skip = extent_map_end(em); 542 *defrag_end = 0; 543 } 544 545 free_extent_map(em); 546 return ret; 547 } 548 549 static int btrfs_defrag_file(struct file *file, 550 struct btrfs_ioctl_defrag_range_args *range) 551 { 552 struct inode *inode = fdentry(file)->d_inode; 553 struct btrfs_root *root = BTRFS_I(inode)->root; 554 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree; 555 struct btrfs_ordered_extent *ordered; 556 struct page *page; 557 unsigned long last_index; 558 unsigned long ra_pages = root->fs_info->bdi.ra_pages; 559 unsigned long total_read = 0; 560 u64 page_start; 561 u64 page_end; 562 u64 last_len = 0; 563 u64 skip = 0; 564 u64 defrag_end = 0; 565 unsigned long i; 566 int ret; 567 568 if (inode->i_size == 0) 569 return 0; 570 571 if (range->start + range->len > range->start) { 572 last_index = min_t(u64, inode->i_size - 1, 573 range->start + range->len - 1) >> PAGE_CACHE_SHIFT; 574 } else { 575 last_index = (inode->i_size - 1) >> PAGE_CACHE_SHIFT; 576 } 577 578 i = range->start >> PAGE_CACHE_SHIFT; 579 while (i <= last_index) { 580 if (!should_defrag_range(inode, (u64)i << PAGE_CACHE_SHIFT, 581 PAGE_CACHE_SIZE, 582 range->extent_thresh, 583 &last_len, &skip, 584 &defrag_end)) { 585 unsigned long next; 586 /* 587 * the should_defrag function tells us how much to skip 588 * bump our counter by the suggested amount 589 */ 590 next = (skip + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT; 591 i = max(i + 1, next); 592 continue; 593 } 594 595 if (total_read % ra_pages == 0) { 596 btrfs_force_ra(inode->i_mapping, &file->f_ra, file, i, 597 min(last_index, i + ra_pages - 1)); 598 } 599 total_read++; 600 mutex_lock(&inode->i_mutex); 601 if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS) 602 BTRFS_I(inode)->force_compress = 1; 603 604 ret = btrfs_check_data_free_space(root, inode, PAGE_CACHE_SIZE); 605 if (ret) { 606 ret = -ENOSPC; 607 break; 608 } 609 610 ret = btrfs_reserve_metadata_for_delalloc(root, inode, 1); 611 if (ret) { 612 btrfs_free_reserved_data_space(root, inode, 613 PAGE_CACHE_SIZE); 614 ret = -ENOSPC; 615 break; 616 } 617 again: 618 if (inode->i_size == 0 || 619 i > ((inode->i_size - 1) >> PAGE_CACHE_SHIFT)) { 620 ret = 0; 621 goto err_reservations; 622 } 623 624 page = grab_cache_page(inode->i_mapping, i); 625 if (!page) 626 goto err_reservations; 627 628 if (!PageUptodate(page)) { 629 btrfs_readpage(NULL, page); 630 lock_page(page); 631 if (!PageUptodate(page)) { 632 unlock_page(page); 633 page_cache_release(page); 634 goto err_reservations; 635 } 636 } 637 638 if (page->mapping != inode->i_mapping) { 639 unlock_page(page); 640 page_cache_release(page); 641 goto again; 642 } 643 644 wait_on_page_writeback(page); 645 646 if (PageDirty(page)) { 647 btrfs_free_reserved_data_space(root, inode, 648 PAGE_CACHE_SIZE); 649 goto loop_unlock; 650 } 651 652 page_start = (u64)page->index << PAGE_CACHE_SHIFT; 653 page_end = page_start + PAGE_CACHE_SIZE - 1; 654 lock_extent(io_tree, page_start, page_end, GFP_NOFS); 655 656 ordered = btrfs_lookup_ordered_extent(inode, page_start); 657 if (ordered) { 658 unlock_extent(io_tree, page_start, page_end, GFP_NOFS); 659 unlock_page(page); 660 page_cache_release(page); 661 btrfs_start_ordered_extent(inode, ordered, 1); 662 btrfs_put_ordered_extent(ordered); 663 goto again; 664 } 665 set_page_extent_mapped(page); 666 667 /* 668 * this makes sure page_mkwrite is called on the 669 * page if it is dirtied again later 670 */ 671 clear_page_dirty_for_io(page); 672 clear_extent_bits(&BTRFS_I(inode)->io_tree, page_start, 673 page_end, EXTENT_DIRTY | EXTENT_DELALLOC | 674 EXTENT_DO_ACCOUNTING, GFP_NOFS); 675 676 btrfs_set_extent_delalloc(inode, page_start, page_end, NULL); 677 ClearPageChecked(page); 678 set_page_dirty(page); 679 unlock_extent(io_tree, page_start, page_end, GFP_NOFS); 680 681 loop_unlock: 682 unlock_page(page); 683 page_cache_release(page); 684 mutex_unlock(&inode->i_mutex); 685 686 btrfs_unreserve_metadata_for_delalloc(root, inode, 1); 687 balance_dirty_pages_ratelimited_nr(inode->i_mapping, 1); 688 i++; 689 } 690 691 if ((range->flags & BTRFS_DEFRAG_RANGE_START_IO)) 692 filemap_flush(inode->i_mapping); 693 694 if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) { 695 /* the filemap_flush will queue IO into the worker threads, but 696 * we have to make sure the IO is actually started and that 697 * ordered extents get created before we return 698 */ 699 atomic_inc(&root->fs_info->async_submit_draining); 700 while (atomic_read(&root->fs_info->nr_async_submits) || 701 atomic_read(&root->fs_info->async_delalloc_pages)) { 702 wait_event(root->fs_info->async_submit_wait, 703 (atomic_read(&root->fs_info->nr_async_submits) == 0 && 704 atomic_read(&root->fs_info->async_delalloc_pages) == 0)); 705 } 706 atomic_dec(&root->fs_info->async_submit_draining); 707 708 mutex_lock(&inode->i_mutex); 709 BTRFS_I(inode)->force_compress = 0; 710 mutex_unlock(&inode->i_mutex); 711 } 712 713 return 0; 714 715 err_reservations: 716 mutex_unlock(&inode->i_mutex); 717 btrfs_free_reserved_data_space(root, inode, PAGE_CACHE_SIZE); 718 btrfs_unreserve_metadata_for_delalloc(root, inode, 1); 719 return ret; 720 } 721 722 static noinline int btrfs_ioctl_resize(struct btrfs_root *root, 723 void __user *arg) 724 { 725 u64 new_size; 726 u64 old_size; 727 u64 devid = 1; 728 struct btrfs_ioctl_vol_args *vol_args; 729 struct btrfs_trans_handle *trans; 730 struct btrfs_device *device = NULL; 731 char *sizestr; 732 char *devstr = NULL; 733 int ret = 0; 734 int namelen; 735 int mod = 0; 736 737 if (root->fs_info->sb->s_flags & MS_RDONLY) 738 return -EROFS; 739 740 if (!capable(CAP_SYS_ADMIN)) 741 return -EPERM; 742 743 vol_args = memdup_user(arg, sizeof(*vol_args)); 744 if (IS_ERR(vol_args)) 745 return PTR_ERR(vol_args); 746 747 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; 748 namelen = strlen(vol_args->name); 749 750 mutex_lock(&root->fs_info->volume_mutex); 751 sizestr = vol_args->name; 752 devstr = strchr(sizestr, ':'); 753 if (devstr) { 754 char *end; 755 sizestr = devstr + 1; 756 *devstr = '\0'; 757 devstr = vol_args->name; 758 devid = simple_strtoull(devstr, &end, 10); 759 printk(KERN_INFO "resizing devid %llu\n", 760 (unsigned long long)devid); 761 } 762 device = btrfs_find_device(root, devid, NULL, NULL); 763 if (!device) { 764 printk(KERN_INFO "resizer unable to find device %llu\n", 765 (unsigned long long)devid); 766 ret = -EINVAL; 767 goto out_unlock; 768 } 769 if (!strcmp(sizestr, "max")) 770 new_size = device->bdev->bd_inode->i_size; 771 else { 772 if (sizestr[0] == '-') { 773 mod = -1; 774 sizestr++; 775 } else if (sizestr[0] == '+') { 776 mod = 1; 777 sizestr++; 778 } 779 new_size = memparse(sizestr, NULL); 780 if (new_size == 0) { 781 ret = -EINVAL; 782 goto out_unlock; 783 } 784 } 785 786 old_size = device->total_bytes; 787 788 if (mod < 0) { 789 if (new_size > old_size) { 790 ret = -EINVAL; 791 goto out_unlock; 792 } 793 new_size = old_size - new_size; 794 } else if (mod > 0) { 795 new_size = old_size + new_size; 796 } 797 798 if (new_size < 256 * 1024 * 1024) { 799 ret = -EINVAL; 800 goto out_unlock; 801 } 802 if (new_size > device->bdev->bd_inode->i_size) { 803 ret = -EFBIG; 804 goto out_unlock; 805 } 806 807 do_div(new_size, root->sectorsize); 808 new_size *= root->sectorsize; 809 810 printk(KERN_INFO "new size for %s is %llu\n", 811 device->name, (unsigned long long)new_size); 812 813 if (new_size > old_size) { 814 trans = btrfs_start_transaction(root, 1); 815 ret = btrfs_grow_device(trans, device, new_size); 816 btrfs_commit_transaction(trans, root); 817 } else { 818 ret = btrfs_shrink_device(device, new_size); 819 } 820 821 out_unlock: 822 mutex_unlock(&root->fs_info->volume_mutex); 823 kfree(vol_args); 824 return ret; 825 } 826 827 static noinline int btrfs_ioctl_snap_create(struct file *file, 828 void __user *arg, int subvol) 829 { 830 struct btrfs_root *root = BTRFS_I(fdentry(file)->d_inode)->root; 831 struct btrfs_ioctl_vol_args *vol_args; 832 struct file *src_file; 833 int namelen; 834 int ret = 0; 835 836 if (root->fs_info->sb->s_flags & MS_RDONLY) 837 return -EROFS; 838 839 vol_args = memdup_user(arg, sizeof(*vol_args)); 840 if (IS_ERR(vol_args)) 841 return PTR_ERR(vol_args); 842 843 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; 844 namelen = strlen(vol_args->name); 845 if (strchr(vol_args->name, '/')) { 846 ret = -EINVAL; 847 goto out; 848 } 849 850 if (subvol) { 851 ret = btrfs_mksubvol(&file->f_path, vol_args->name, namelen, 852 NULL); 853 } else { 854 struct inode *src_inode; 855 src_file = fget(vol_args->fd); 856 if (!src_file) { 857 ret = -EINVAL; 858 goto out; 859 } 860 861 src_inode = src_file->f_path.dentry->d_inode; 862 if (src_inode->i_sb != file->f_path.dentry->d_inode->i_sb) { 863 printk(KERN_INFO "btrfs: Snapshot src from " 864 "another FS\n"); 865 ret = -EINVAL; 866 fput(src_file); 867 goto out; 868 } 869 ret = btrfs_mksubvol(&file->f_path, vol_args->name, namelen, 870 BTRFS_I(src_inode)->root); 871 fput(src_file); 872 } 873 out: 874 kfree(vol_args); 875 return ret; 876 } 877 878 /* 879 * helper to check if the subvolume references other subvolumes 880 */ 881 static noinline int may_destroy_subvol(struct btrfs_root *root) 882 { 883 struct btrfs_path *path; 884 struct btrfs_key key; 885 int ret; 886 887 path = btrfs_alloc_path(); 888 if (!path) 889 return -ENOMEM; 890 891 key.objectid = root->root_key.objectid; 892 key.type = BTRFS_ROOT_REF_KEY; 893 key.offset = (u64)-1; 894 895 ret = btrfs_search_slot(NULL, root->fs_info->tree_root, 896 &key, path, 0, 0); 897 if (ret < 0) 898 goto out; 899 BUG_ON(ret == 0); 900 901 ret = 0; 902 if (path->slots[0] > 0) { 903 path->slots[0]--; 904 btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]); 905 if (key.objectid == root->root_key.objectid && 906 key.type == BTRFS_ROOT_REF_KEY) 907 ret = -ENOTEMPTY; 908 } 909 out: 910 btrfs_free_path(path); 911 return ret; 912 } 913 914 static noinline int key_in_sk(struct btrfs_key *key, 915 struct btrfs_ioctl_search_key *sk) 916 { 917 struct btrfs_key test; 918 int ret; 919 920 test.objectid = sk->min_objectid; 921 test.type = sk->min_type; 922 test.offset = sk->min_offset; 923 924 ret = btrfs_comp_cpu_keys(key, &test); 925 if (ret < 0) 926 return 0; 927 928 test.objectid = sk->max_objectid; 929 test.type = sk->max_type; 930 test.offset = sk->max_offset; 931 932 ret = btrfs_comp_cpu_keys(key, &test); 933 if (ret > 0) 934 return 0; 935 return 1; 936 } 937 938 static noinline int copy_to_sk(struct btrfs_root *root, 939 struct btrfs_path *path, 940 struct btrfs_key *key, 941 struct btrfs_ioctl_search_key *sk, 942 char *buf, 943 unsigned long *sk_offset, 944 int *num_found) 945 { 946 u64 found_transid; 947 struct extent_buffer *leaf; 948 struct btrfs_ioctl_search_header sh; 949 unsigned long item_off; 950 unsigned long item_len; 951 int nritems; 952 int i; 953 int slot; 954 int found = 0; 955 int ret = 0; 956 957 leaf = path->nodes[0]; 958 slot = path->slots[0]; 959 nritems = btrfs_header_nritems(leaf); 960 961 if (btrfs_header_generation(leaf) > sk->max_transid) { 962 i = nritems; 963 goto advance_key; 964 } 965 found_transid = btrfs_header_generation(leaf); 966 967 for (i = slot; i < nritems; i++) { 968 item_off = btrfs_item_ptr_offset(leaf, i); 969 item_len = btrfs_item_size_nr(leaf, i); 970 971 if (item_len > BTRFS_SEARCH_ARGS_BUFSIZE) 972 item_len = 0; 973 974 if (sizeof(sh) + item_len + *sk_offset > 975 BTRFS_SEARCH_ARGS_BUFSIZE) { 976 ret = 1; 977 goto overflow; 978 } 979 980 btrfs_item_key_to_cpu(leaf, key, i); 981 if (!key_in_sk(key, sk)) 982 continue; 983 984 sh.objectid = key->objectid; 985 sh.offset = key->offset; 986 sh.type = key->type; 987 sh.len = item_len; 988 sh.transid = found_transid; 989 990 /* copy search result header */ 991 memcpy(buf + *sk_offset, &sh, sizeof(sh)); 992 *sk_offset += sizeof(sh); 993 994 if (item_len) { 995 char *p = buf + *sk_offset; 996 /* copy the item */ 997 read_extent_buffer(leaf, p, 998 item_off, item_len); 999 *sk_offset += item_len; 1000 } 1001 found++; 1002 1003 if (*num_found >= sk->nr_items) 1004 break; 1005 } 1006 advance_key: 1007 ret = 0; 1008 if (key->offset < (u64)-1 && key->offset < sk->max_offset) 1009 key->offset++; 1010 else if (key->type < (u8)-1 && key->type < sk->max_type) { 1011 key->offset = 0; 1012 key->type++; 1013 } else if (key->objectid < (u64)-1 && key->objectid < sk->max_objectid) { 1014 key->offset = 0; 1015 key->type = 0; 1016 key->objectid++; 1017 } else 1018 ret = 1; 1019 overflow: 1020 *num_found += found; 1021 return ret; 1022 } 1023 1024 static noinline int search_ioctl(struct inode *inode, 1025 struct btrfs_ioctl_search_args *args) 1026 { 1027 struct btrfs_root *root; 1028 struct btrfs_key key; 1029 struct btrfs_key max_key; 1030 struct btrfs_path *path; 1031 struct btrfs_ioctl_search_key *sk = &args->key; 1032 struct btrfs_fs_info *info = BTRFS_I(inode)->root->fs_info; 1033 int ret; 1034 int num_found = 0; 1035 unsigned long sk_offset = 0; 1036 1037 path = btrfs_alloc_path(); 1038 if (!path) 1039 return -ENOMEM; 1040 1041 if (sk->tree_id == 0) { 1042 /* search the root of the inode that was passed */ 1043 root = BTRFS_I(inode)->root; 1044 } else { 1045 key.objectid = sk->tree_id; 1046 key.type = BTRFS_ROOT_ITEM_KEY; 1047 key.offset = (u64)-1; 1048 root = btrfs_read_fs_root_no_name(info, &key); 1049 if (IS_ERR(root)) { 1050 printk(KERN_ERR "could not find root %llu\n", 1051 sk->tree_id); 1052 btrfs_free_path(path); 1053 return -ENOENT; 1054 } 1055 } 1056 1057 key.objectid = sk->min_objectid; 1058 key.type = sk->min_type; 1059 key.offset = sk->min_offset; 1060 1061 max_key.objectid = sk->max_objectid; 1062 max_key.type = sk->max_type; 1063 max_key.offset = sk->max_offset; 1064 1065 path->keep_locks = 1; 1066 1067 while(1) { 1068 ret = btrfs_search_forward(root, &key, &max_key, path, 0, 1069 sk->min_transid); 1070 if (ret != 0) { 1071 if (ret > 0) 1072 ret = 0; 1073 goto err; 1074 } 1075 ret = copy_to_sk(root, path, &key, sk, args->buf, 1076 &sk_offset, &num_found); 1077 btrfs_release_path(root, path); 1078 if (ret || num_found >= sk->nr_items) 1079 break; 1080 1081 } 1082 ret = 0; 1083 err: 1084 sk->nr_items = num_found; 1085 btrfs_free_path(path); 1086 return ret; 1087 } 1088 1089 static noinline int btrfs_ioctl_tree_search(struct file *file, 1090 void __user *argp) 1091 { 1092 struct btrfs_ioctl_search_args *args; 1093 struct inode *inode; 1094 int ret; 1095 1096 if (!capable(CAP_SYS_ADMIN)) 1097 return -EPERM; 1098 1099 args = kmalloc(sizeof(*args), GFP_KERNEL); 1100 if (!args) 1101 return -ENOMEM; 1102 1103 if (copy_from_user(args, argp, sizeof(*args))) { 1104 kfree(args); 1105 return -EFAULT; 1106 } 1107 inode = fdentry(file)->d_inode; 1108 ret = search_ioctl(inode, args); 1109 if (ret == 0 && copy_to_user(argp, args, sizeof(*args))) 1110 ret = -EFAULT; 1111 kfree(args); 1112 return ret; 1113 } 1114 1115 /* 1116 * Search INODE_REFs to identify path name of 'dirid' directory 1117 * in a 'tree_id' tree. and sets path name to 'name'. 1118 */ 1119 static noinline int btrfs_search_path_in_tree(struct btrfs_fs_info *info, 1120 u64 tree_id, u64 dirid, char *name) 1121 { 1122 struct btrfs_root *root; 1123 struct btrfs_key key; 1124 char *ptr; 1125 int ret = -1; 1126 int slot; 1127 int len; 1128 int total_len = 0; 1129 struct btrfs_inode_ref *iref; 1130 struct extent_buffer *l; 1131 struct btrfs_path *path; 1132 1133 if (dirid == BTRFS_FIRST_FREE_OBJECTID) { 1134 name[0]='\0'; 1135 return 0; 1136 } 1137 1138 path = btrfs_alloc_path(); 1139 if (!path) 1140 return -ENOMEM; 1141 1142 ptr = &name[BTRFS_INO_LOOKUP_PATH_MAX]; 1143 1144 key.objectid = tree_id; 1145 key.type = BTRFS_ROOT_ITEM_KEY; 1146 key.offset = (u64)-1; 1147 root = btrfs_read_fs_root_no_name(info, &key); 1148 if (IS_ERR(root)) { 1149 printk(KERN_ERR "could not find root %llu\n", tree_id); 1150 ret = -ENOENT; 1151 goto out; 1152 } 1153 1154 key.objectid = dirid; 1155 key.type = BTRFS_INODE_REF_KEY; 1156 key.offset = (u64)-1; 1157 1158 while(1) { 1159 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); 1160 if (ret < 0) 1161 goto out; 1162 1163 l = path->nodes[0]; 1164 slot = path->slots[0]; 1165 if (ret > 0 && slot > 0) 1166 slot--; 1167 btrfs_item_key_to_cpu(l, &key, slot); 1168 1169 if (ret > 0 && (key.objectid != dirid || 1170 key.type != BTRFS_INODE_REF_KEY)) { 1171 ret = -ENOENT; 1172 goto out; 1173 } 1174 1175 iref = btrfs_item_ptr(l, slot, struct btrfs_inode_ref); 1176 len = btrfs_inode_ref_name_len(l, iref); 1177 ptr -= len + 1; 1178 total_len += len + 1; 1179 if (ptr < name) 1180 goto out; 1181 1182 *(ptr + len) = '/'; 1183 read_extent_buffer(l, ptr,(unsigned long)(iref + 1), len); 1184 1185 if (key.offset == BTRFS_FIRST_FREE_OBJECTID) 1186 break; 1187 1188 btrfs_release_path(root, path); 1189 key.objectid = key.offset; 1190 key.offset = (u64)-1; 1191 dirid = key.objectid; 1192 1193 } 1194 if (ptr < name) 1195 goto out; 1196 memcpy(name, ptr, total_len); 1197 name[total_len]='\0'; 1198 ret = 0; 1199 out: 1200 btrfs_free_path(path); 1201 return ret; 1202 } 1203 1204 static noinline int btrfs_ioctl_ino_lookup(struct file *file, 1205 void __user *argp) 1206 { 1207 struct btrfs_ioctl_ino_lookup_args *args; 1208 struct inode *inode; 1209 int ret; 1210 1211 if (!capable(CAP_SYS_ADMIN)) 1212 return -EPERM; 1213 1214 args = kmalloc(sizeof(*args), GFP_KERNEL); 1215 if (!args) 1216 return -ENOMEM; 1217 1218 if (copy_from_user(args, argp, sizeof(*args))) { 1219 kfree(args); 1220 return -EFAULT; 1221 } 1222 inode = fdentry(file)->d_inode; 1223 1224 if (args->treeid == 0) 1225 args->treeid = BTRFS_I(inode)->root->root_key.objectid; 1226 1227 ret = btrfs_search_path_in_tree(BTRFS_I(inode)->root->fs_info, 1228 args->treeid, args->objectid, 1229 args->name); 1230 1231 if (ret == 0 && copy_to_user(argp, args, sizeof(*args))) 1232 ret = -EFAULT; 1233 1234 kfree(args); 1235 return ret; 1236 } 1237 1238 static noinline int btrfs_ioctl_snap_destroy(struct file *file, 1239 void __user *arg) 1240 { 1241 struct dentry *parent = fdentry(file); 1242 struct dentry *dentry; 1243 struct inode *dir = parent->d_inode; 1244 struct inode *inode; 1245 struct btrfs_root *root = BTRFS_I(dir)->root; 1246 struct btrfs_root *dest = NULL; 1247 struct btrfs_ioctl_vol_args *vol_args; 1248 struct btrfs_trans_handle *trans; 1249 int namelen; 1250 int ret; 1251 int err = 0; 1252 1253 if (!capable(CAP_SYS_ADMIN)) 1254 return -EPERM; 1255 1256 vol_args = memdup_user(arg, sizeof(*vol_args)); 1257 if (IS_ERR(vol_args)) 1258 return PTR_ERR(vol_args); 1259 1260 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; 1261 namelen = strlen(vol_args->name); 1262 if (strchr(vol_args->name, '/') || 1263 strncmp(vol_args->name, "..", namelen) == 0) { 1264 err = -EINVAL; 1265 goto out; 1266 } 1267 1268 err = mnt_want_write(file->f_path.mnt); 1269 if (err) 1270 goto out; 1271 1272 mutex_lock_nested(&dir->i_mutex, I_MUTEX_PARENT); 1273 dentry = lookup_one_len(vol_args->name, parent, namelen); 1274 if (IS_ERR(dentry)) { 1275 err = PTR_ERR(dentry); 1276 goto out_unlock_dir; 1277 } 1278 1279 if (!dentry->d_inode) { 1280 err = -ENOENT; 1281 goto out_dput; 1282 } 1283 1284 inode = dentry->d_inode; 1285 if (inode->i_ino != BTRFS_FIRST_FREE_OBJECTID) { 1286 err = -EINVAL; 1287 goto out_dput; 1288 } 1289 1290 dest = BTRFS_I(inode)->root; 1291 1292 mutex_lock(&inode->i_mutex); 1293 err = d_invalidate(dentry); 1294 if (err) 1295 goto out_unlock; 1296 1297 down_write(&root->fs_info->subvol_sem); 1298 1299 err = may_destroy_subvol(dest); 1300 if (err) 1301 goto out_up_write; 1302 1303 trans = btrfs_start_transaction(root, 1); 1304 ret = btrfs_unlink_subvol(trans, root, dir, 1305 dest->root_key.objectid, 1306 dentry->d_name.name, 1307 dentry->d_name.len); 1308 BUG_ON(ret); 1309 1310 btrfs_record_root_in_trans(trans, dest); 1311 1312 memset(&dest->root_item.drop_progress, 0, 1313 sizeof(dest->root_item.drop_progress)); 1314 dest->root_item.drop_level = 0; 1315 btrfs_set_root_refs(&dest->root_item, 0); 1316 1317 ret = btrfs_insert_orphan_item(trans, 1318 root->fs_info->tree_root, 1319 dest->root_key.objectid); 1320 BUG_ON(ret); 1321 1322 ret = btrfs_commit_transaction(trans, root); 1323 BUG_ON(ret); 1324 inode->i_flags |= S_DEAD; 1325 out_up_write: 1326 up_write(&root->fs_info->subvol_sem); 1327 out_unlock: 1328 mutex_unlock(&inode->i_mutex); 1329 if (!err) { 1330 shrink_dcache_sb(root->fs_info->sb); 1331 btrfs_invalidate_inodes(dest); 1332 d_delete(dentry); 1333 } 1334 out_dput: 1335 dput(dentry); 1336 out_unlock_dir: 1337 mutex_unlock(&dir->i_mutex); 1338 mnt_drop_write(file->f_path.mnt); 1339 out: 1340 kfree(vol_args); 1341 return err; 1342 } 1343 1344 static int btrfs_ioctl_defrag(struct file *file, void __user *argp) 1345 { 1346 struct inode *inode = fdentry(file)->d_inode; 1347 struct btrfs_root *root = BTRFS_I(inode)->root; 1348 struct btrfs_ioctl_defrag_range_args *range; 1349 int ret; 1350 1351 ret = mnt_want_write(file->f_path.mnt); 1352 if (ret) 1353 return ret; 1354 1355 switch (inode->i_mode & S_IFMT) { 1356 case S_IFDIR: 1357 if (!capable(CAP_SYS_ADMIN)) { 1358 ret = -EPERM; 1359 goto out; 1360 } 1361 btrfs_defrag_root(root, 0); 1362 btrfs_defrag_root(root->fs_info->extent_root, 0); 1363 break; 1364 case S_IFREG: 1365 if (!(file->f_mode & FMODE_WRITE)) { 1366 ret = -EINVAL; 1367 goto out; 1368 } 1369 1370 range = kzalloc(sizeof(*range), GFP_KERNEL); 1371 if (!range) { 1372 ret = -ENOMEM; 1373 goto out; 1374 } 1375 1376 if (argp) { 1377 if (copy_from_user(range, argp, 1378 sizeof(*range))) { 1379 ret = -EFAULT; 1380 kfree(range); 1381 goto out; 1382 } 1383 /* compression requires us to start the IO */ 1384 if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) { 1385 range->flags |= BTRFS_DEFRAG_RANGE_START_IO; 1386 range->extent_thresh = (u32)-1; 1387 } 1388 } else { 1389 /* the rest are all set to zero by kzalloc */ 1390 range->len = (u64)-1; 1391 } 1392 btrfs_defrag_file(file, range); 1393 kfree(range); 1394 break; 1395 } 1396 out: 1397 mnt_drop_write(file->f_path.mnt); 1398 return ret; 1399 } 1400 1401 static long btrfs_ioctl_add_dev(struct btrfs_root *root, void __user *arg) 1402 { 1403 struct btrfs_ioctl_vol_args *vol_args; 1404 int ret; 1405 1406 if (!capable(CAP_SYS_ADMIN)) 1407 return -EPERM; 1408 1409 vol_args = memdup_user(arg, sizeof(*vol_args)); 1410 if (IS_ERR(vol_args)) 1411 return PTR_ERR(vol_args); 1412 1413 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; 1414 ret = btrfs_init_new_device(root, vol_args->name); 1415 1416 kfree(vol_args); 1417 return ret; 1418 } 1419 1420 static long btrfs_ioctl_rm_dev(struct btrfs_root *root, void __user *arg) 1421 { 1422 struct btrfs_ioctl_vol_args *vol_args; 1423 int ret; 1424 1425 if (!capable(CAP_SYS_ADMIN)) 1426 return -EPERM; 1427 1428 if (root->fs_info->sb->s_flags & MS_RDONLY) 1429 return -EROFS; 1430 1431 vol_args = memdup_user(arg, sizeof(*vol_args)); 1432 if (IS_ERR(vol_args)) 1433 return PTR_ERR(vol_args); 1434 1435 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; 1436 ret = btrfs_rm_device(root, vol_args->name); 1437 1438 kfree(vol_args); 1439 return ret; 1440 } 1441 1442 static noinline long btrfs_ioctl_clone(struct file *file, unsigned long srcfd, 1443 u64 off, u64 olen, u64 destoff) 1444 { 1445 struct inode *inode = fdentry(file)->d_inode; 1446 struct btrfs_root *root = BTRFS_I(inode)->root; 1447 struct file *src_file; 1448 struct inode *src; 1449 struct btrfs_trans_handle *trans; 1450 struct btrfs_path *path; 1451 struct extent_buffer *leaf; 1452 char *buf; 1453 struct btrfs_key key; 1454 u32 nritems; 1455 int slot; 1456 int ret; 1457 u64 len = olen; 1458 u64 bs = root->fs_info->sb->s_blocksize; 1459 u64 hint_byte; 1460 1461 /* 1462 * TODO: 1463 * - split compressed inline extents. annoying: we need to 1464 * decompress into destination's address_space (the file offset 1465 * may change, so source mapping won't do), then recompress (or 1466 * otherwise reinsert) a subrange. 1467 * - allow ranges within the same file to be cloned (provided 1468 * they don't overlap)? 1469 */ 1470 1471 /* the destination must be opened for writing */ 1472 if (!(file->f_mode & FMODE_WRITE)) 1473 return -EINVAL; 1474 1475 ret = mnt_want_write(file->f_path.mnt); 1476 if (ret) 1477 return ret; 1478 1479 src_file = fget(srcfd); 1480 if (!src_file) { 1481 ret = -EBADF; 1482 goto out_drop_write; 1483 } 1484 1485 src = src_file->f_dentry->d_inode; 1486 1487 ret = -EINVAL; 1488 if (src == inode) 1489 goto out_fput; 1490 1491 /* the src must be open for reading */ 1492 if (!(src_file->f_mode & FMODE_READ)) 1493 goto out_fput; 1494 1495 ret = -EISDIR; 1496 if (S_ISDIR(src->i_mode) || S_ISDIR(inode->i_mode)) 1497 goto out_fput; 1498 1499 ret = -EXDEV; 1500 if (src->i_sb != inode->i_sb || BTRFS_I(src)->root != root) 1501 goto out_fput; 1502 1503 ret = -ENOMEM; 1504 buf = vmalloc(btrfs_level_size(root, 0)); 1505 if (!buf) 1506 goto out_fput; 1507 1508 path = btrfs_alloc_path(); 1509 if (!path) { 1510 vfree(buf); 1511 goto out_fput; 1512 } 1513 path->reada = 2; 1514 1515 if (inode < src) { 1516 mutex_lock(&inode->i_mutex); 1517 mutex_lock(&src->i_mutex); 1518 } else { 1519 mutex_lock(&src->i_mutex); 1520 mutex_lock(&inode->i_mutex); 1521 } 1522 1523 /* determine range to clone */ 1524 ret = -EINVAL; 1525 if (off >= src->i_size || off + len > src->i_size) 1526 goto out_unlock; 1527 if (len == 0) 1528 olen = len = src->i_size - off; 1529 /* if we extend to eof, continue to block boundary */ 1530 if (off + len == src->i_size) 1531 len = ((src->i_size + bs-1) & ~(bs-1)) 1532 - off; 1533 1534 /* verify the end result is block aligned */ 1535 if ((off & (bs-1)) || 1536 ((off + len) & (bs-1))) 1537 goto out_unlock; 1538 1539 /* do any pending delalloc/csum calc on src, one way or 1540 another, and lock file content */ 1541 while (1) { 1542 struct btrfs_ordered_extent *ordered; 1543 lock_extent(&BTRFS_I(src)->io_tree, off, off+len, GFP_NOFS); 1544 ordered = btrfs_lookup_first_ordered_extent(inode, off+len); 1545 if (BTRFS_I(src)->delalloc_bytes == 0 && !ordered) 1546 break; 1547 unlock_extent(&BTRFS_I(src)->io_tree, off, off+len, GFP_NOFS); 1548 if (ordered) 1549 btrfs_put_ordered_extent(ordered); 1550 btrfs_wait_ordered_range(src, off, off+len); 1551 } 1552 1553 trans = btrfs_start_transaction(root, 1); 1554 BUG_ON(!trans); 1555 1556 /* punch hole in destination first */ 1557 btrfs_drop_extents(trans, inode, off, off + len, &hint_byte, 1); 1558 1559 /* clone data */ 1560 key.objectid = src->i_ino; 1561 key.type = BTRFS_EXTENT_DATA_KEY; 1562 key.offset = 0; 1563 1564 while (1) { 1565 /* 1566 * note the key will change type as we walk through the 1567 * tree. 1568 */ 1569 ret = btrfs_search_slot(trans, root, &key, path, 0, 0); 1570 if (ret < 0) 1571 goto out; 1572 1573 nritems = btrfs_header_nritems(path->nodes[0]); 1574 if (path->slots[0] >= nritems) { 1575 ret = btrfs_next_leaf(root, path); 1576 if (ret < 0) 1577 goto out; 1578 if (ret > 0) 1579 break; 1580 nritems = btrfs_header_nritems(path->nodes[0]); 1581 } 1582 leaf = path->nodes[0]; 1583 slot = path->slots[0]; 1584 1585 btrfs_item_key_to_cpu(leaf, &key, slot); 1586 if (btrfs_key_type(&key) > BTRFS_EXTENT_DATA_KEY || 1587 key.objectid != src->i_ino) 1588 break; 1589 1590 if (btrfs_key_type(&key) == BTRFS_EXTENT_DATA_KEY) { 1591 struct btrfs_file_extent_item *extent; 1592 int type; 1593 u32 size; 1594 struct btrfs_key new_key; 1595 u64 disko = 0, diskl = 0; 1596 u64 datao = 0, datal = 0; 1597 u8 comp; 1598 1599 size = btrfs_item_size_nr(leaf, slot); 1600 read_extent_buffer(leaf, buf, 1601 btrfs_item_ptr_offset(leaf, slot), 1602 size); 1603 1604 extent = btrfs_item_ptr(leaf, slot, 1605 struct btrfs_file_extent_item); 1606 comp = btrfs_file_extent_compression(leaf, extent); 1607 type = btrfs_file_extent_type(leaf, extent); 1608 if (type == BTRFS_FILE_EXTENT_REG || 1609 type == BTRFS_FILE_EXTENT_PREALLOC) { 1610 disko = btrfs_file_extent_disk_bytenr(leaf, 1611 extent); 1612 diskl = btrfs_file_extent_disk_num_bytes(leaf, 1613 extent); 1614 datao = btrfs_file_extent_offset(leaf, extent); 1615 datal = btrfs_file_extent_num_bytes(leaf, 1616 extent); 1617 } else if (type == BTRFS_FILE_EXTENT_INLINE) { 1618 /* take upper bound, may be compressed */ 1619 datal = btrfs_file_extent_ram_bytes(leaf, 1620 extent); 1621 } 1622 btrfs_release_path(root, path); 1623 1624 if (key.offset + datal < off || 1625 key.offset >= off+len) 1626 goto next; 1627 1628 memcpy(&new_key, &key, sizeof(new_key)); 1629 new_key.objectid = inode->i_ino; 1630 new_key.offset = key.offset + destoff - off; 1631 1632 if (type == BTRFS_FILE_EXTENT_REG || 1633 type == BTRFS_FILE_EXTENT_PREALLOC) { 1634 ret = btrfs_insert_empty_item(trans, root, path, 1635 &new_key, size); 1636 if (ret) 1637 goto out; 1638 1639 leaf = path->nodes[0]; 1640 slot = path->slots[0]; 1641 write_extent_buffer(leaf, buf, 1642 btrfs_item_ptr_offset(leaf, slot), 1643 size); 1644 1645 extent = btrfs_item_ptr(leaf, slot, 1646 struct btrfs_file_extent_item); 1647 1648 if (off > key.offset) { 1649 datao += off - key.offset; 1650 datal -= off - key.offset; 1651 } 1652 1653 if (key.offset + datal > off + len) 1654 datal = off + len - key.offset; 1655 1656 /* disko == 0 means it's a hole */ 1657 if (!disko) 1658 datao = 0; 1659 1660 btrfs_set_file_extent_offset(leaf, extent, 1661 datao); 1662 btrfs_set_file_extent_num_bytes(leaf, extent, 1663 datal); 1664 if (disko) { 1665 inode_add_bytes(inode, datal); 1666 ret = btrfs_inc_extent_ref(trans, root, 1667 disko, diskl, 0, 1668 root->root_key.objectid, 1669 inode->i_ino, 1670 new_key.offset - datao); 1671 BUG_ON(ret); 1672 } 1673 } else if (type == BTRFS_FILE_EXTENT_INLINE) { 1674 u64 skip = 0; 1675 u64 trim = 0; 1676 if (off > key.offset) { 1677 skip = off - key.offset; 1678 new_key.offset += skip; 1679 } 1680 1681 if (key.offset + datal > off+len) 1682 trim = key.offset + datal - (off+len); 1683 1684 if (comp && (skip || trim)) { 1685 ret = -EINVAL; 1686 goto out; 1687 } 1688 size -= skip + trim; 1689 datal -= skip + trim; 1690 ret = btrfs_insert_empty_item(trans, root, path, 1691 &new_key, size); 1692 if (ret) 1693 goto out; 1694 1695 if (skip) { 1696 u32 start = 1697 btrfs_file_extent_calc_inline_size(0); 1698 memmove(buf+start, buf+start+skip, 1699 datal); 1700 } 1701 1702 leaf = path->nodes[0]; 1703 slot = path->slots[0]; 1704 write_extent_buffer(leaf, buf, 1705 btrfs_item_ptr_offset(leaf, slot), 1706 size); 1707 inode_add_bytes(inode, datal); 1708 } 1709 1710 btrfs_mark_buffer_dirty(leaf); 1711 } 1712 1713 next: 1714 btrfs_release_path(root, path); 1715 key.offset++; 1716 } 1717 ret = 0; 1718 out: 1719 btrfs_release_path(root, path); 1720 if (ret == 0) { 1721 inode->i_mtime = inode->i_ctime = CURRENT_TIME; 1722 if (destoff + olen > inode->i_size) 1723 btrfs_i_size_write(inode, destoff + olen); 1724 BTRFS_I(inode)->flags = BTRFS_I(src)->flags; 1725 ret = btrfs_update_inode(trans, root, inode); 1726 } 1727 btrfs_end_transaction(trans, root); 1728 unlock_extent(&BTRFS_I(src)->io_tree, off, off+len, GFP_NOFS); 1729 if (ret) 1730 vmtruncate(inode, 0); 1731 out_unlock: 1732 mutex_unlock(&src->i_mutex); 1733 mutex_unlock(&inode->i_mutex); 1734 vfree(buf); 1735 btrfs_free_path(path); 1736 out_fput: 1737 fput(src_file); 1738 out_drop_write: 1739 mnt_drop_write(file->f_path.mnt); 1740 return ret; 1741 } 1742 1743 static long btrfs_ioctl_clone_range(struct file *file, void __user *argp) 1744 { 1745 struct btrfs_ioctl_clone_range_args args; 1746 1747 if (copy_from_user(&args, argp, sizeof(args))) 1748 return -EFAULT; 1749 return btrfs_ioctl_clone(file, args.src_fd, args.src_offset, 1750 args.src_length, args.dest_offset); 1751 } 1752 1753 /* 1754 * there are many ways the trans_start and trans_end ioctls can lead 1755 * to deadlocks. They should only be used by applications that 1756 * basically own the machine, and have a very in depth understanding 1757 * of all the possible deadlocks and enospc problems. 1758 */ 1759 static long btrfs_ioctl_trans_start(struct file *file) 1760 { 1761 struct inode *inode = fdentry(file)->d_inode; 1762 struct btrfs_root *root = BTRFS_I(inode)->root; 1763 struct btrfs_trans_handle *trans; 1764 int ret; 1765 1766 ret = -EPERM; 1767 if (!capable(CAP_SYS_ADMIN)) 1768 goto out; 1769 1770 ret = -EINPROGRESS; 1771 if (file->private_data) 1772 goto out; 1773 1774 ret = mnt_want_write(file->f_path.mnt); 1775 if (ret) 1776 goto out; 1777 1778 mutex_lock(&root->fs_info->trans_mutex); 1779 root->fs_info->open_ioctl_trans++; 1780 mutex_unlock(&root->fs_info->trans_mutex); 1781 1782 ret = -ENOMEM; 1783 trans = btrfs_start_ioctl_transaction(root, 0); 1784 if (!trans) 1785 goto out_drop; 1786 1787 file->private_data = trans; 1788 return 0; 1789 1790 out_drop: 1791 mutex_lock(&root->fs_info->trans_mutex); 1792 root->fs_info->open_ioctl_trans--; 1793 mutex_unlock(&root->fs_info->trans_mutex); 1794 mnt_drop_write(file->f_path.mnt); 1795 out: 1796 return ret; 1797 } 1798 1799 static long btrfs_ioctl_default_subvol(struct file *file, void __user *argp) 1800 { 1801 struct inode *inode = fdentry(file)->d_inode; 1802 struct btrfs_root *root = BTRFS_I(inode)->root; 1803 struct btrfs_root *new_root; 1804 struct btrfs_dir_item *di; 1805 struct btrfs_trans_handle *trans; 1806 struct btrfs_path *path; 1807 struct btrfs_key location; 1808 struct btrfs_disk_key disk_key; 1809 struct btrfs_super_block *disk_super; 1810 u64 features; 1811 u64 objectid = 0; 1812 u64 dir_id; 1813 1814 if (!capable(CAP_SYS_ADMIN)) 1815 return -EPERM; 1816 1817 if (copy_from_user(&objectid, argp, sizeof(objectid))) 1818 return -EFAULT; 1819 1820 if (!objectid) 1821 objectid = root->root_key.objectid; 1822 1823 location.objectid = objectid; 1824 location.type = BTRFS_ROOT_ITEM_KEY; 1825 location.offset = (u64)-1; 1826 1827 new_root = btrfs_read_fs_root_no_name(root->fs_info, &location); 1828 if (IS_ERR(new_root)) 1829 return PTR_ERR(new_root); 1830 1831 if (btrfs_root_refs(&new_root->root_item) == 0) 1832 return -ENOENT; 1833 1834 path = btrfs_alloc_path(); 1835 if (!path) 1836 return -ENOMEM; 1837 path->leave_spinning = 1; 1838 1839 trans = btrfs_start_transaction(root, 1); 1840 if (!trans) { 1841 btrfs_free_path(path); 1842 return -ENOMEM; 1843 } 1844 1845 dir_id = btrfs_super_root_dir(&root->fs_info->super_copy); 1846 di = btrfs_lookup_dir_item(trans, root->fs_info->tree_root, path, 1847 dir_id, "default", 7, 1); 1848 if (!di) { 1849 btrfs_free_path(path); 1850 btrfs_end_transaction(trans, root); 1851 printk(KERN_ERR "Umm, you don't have the default dir item, " 1852 "this isn't going to work\n"); 1853 return -ENOENT; 1854 } 1855 1856 btrfs_cpu_key_to_disk(&disk_key, &new_root->root_key); 1857 btrfs_set_dir_item_key(path->nodes[0], di, &disk_key); 1858 btrfs_mark_buffer_dirty(path->nodes[0]); 1859 btrfs_free_path(path); 1860 1861 disk_super = &root->fs_info->super_copy; 1862 features = btrfs_super_incompat_flags(disk_super); 1863 if (!(features & BTRFS_FEATURE_INCOMPAT_DEFAULT_SUBVOL)) { 1864 features |= BTRFS_FEATURE_INCOMPAT_DEFAULT_SUBVOL; 1865 btrfs_set_super_incompat_flags(disk_super, features); 1866 } 1867 btrfs_end_transaction(trans, root); 1868 1869 return 0; 1870 } 1871 1872 long btrfs_ioctl_space_info(struct btrfs_root *root, void __user *arg) 1873 { 1874 struct btrfs_ioctl_space_args space_args; 1875 struct btrfs_ioctl_space_info space; 1876 struct btrfs_ioctl_space_info *dest; 1877 struct btrfs_ioctl_space_info *dest_orig; 1878 struct btrfs_ioctl_space_info *user_dest; 1879 struct btrfs_space_info *info; 1880 int alloc_size; 1881 int ret = 0; 1882 int slot_count = 0; 1883 1884 if (copy_from_user(&space_args, 1885 (struct btrfs_ioctl_space_args __user *)arg, 1886 sizeof(space_args))) 1887 return -EFAULT; 1888 1889 /* first we count slots */ 1890 rcu_read_lock(); 1891 list_for_each_entry_rcu(info, &root->fs_info->space_info, list) 1892 slot_count++; 1893 rcu_read_unlock(); 1894 1895 /* space_slots == 0 means they are asking for a count */ 1896 if (space_args.space_slots == 0) { 1897 space_args.total_spaces = slot_count; 1898 goto out; 1899 } 1900 alloc_size = sizeof(*dest) * slot_count; 1901 /* we generally have at most 6 or so space infos, one for each raid 1902 * level. So, a whole page should be more than enough for everyone 1903 */ 1904 if (alloc_size > PAGE_CACHE_SIZE) 1905 return -ENOMEM; 1906 1907 space_args.total_spaces = 0; 1908 dest = kmalloc(alloc_size, GFP_NOFS); 1909 if (!dest) 1910 return -ENOMEM; 1911 dest_orig = dest; 1912 1913 /* now we have a buffer to copy into */ 1914 rcu_read_lock(); 1915 list_for_each_entry_rcu(info, &root->fs_info->space_info, list) { 1916 /* make sure we don't copy more than we allocated 1917 * in our buffer 1918 */ 1919 if (slot_count == 0) 1920 break; 1921 slot_count--; 1922 1923 /* make sure userland has enough room in their buffer */ 1924 if (space_args.total_spaces >= space_args.space_slots) 1925 break; 1926 1927 space.flags = info->flags; 1928 space.total_bytes = info->total_bytes; 1929 space.used_bytes = info->bytes_used; 1930 memcpy(dest, &space, sizeof(space)); 1931 dest++; 1932 space_args.total_spaces++; 1933 } 1934 rcu_read_unlock(); 1935 1936 user_dest = (struct btrfs_ioctl_space_info *) 1937 (arg + sizeof(struct btrfs_ioctl_space_args)); 1938 1939 if (copy_to_user(user_dest, dest_orig, alloc_size)) 1940 ret = -EFAULT; 1941 1942 kfree(dest_orig); 1943 out: 1944 if (ret == 0 && copy_to_user(arg, &space_args, sizeof(space_args))) 1945 ret = -EFAULT; 1946 1947 return ret; 1948 } 1949 1950 /* 1951 * there are many ways the trans_start and trans_end ioctls can lead 1952 * to deadlocks. They should only be used by applications that 1953 * basically own the machine, and have a very in depth understanding 1954 * of all the possible deadlocks and enospc problems. 1955 */ 1956 long btrfs_ioctl_trans_end(struct file *file) 1957 { 1958 struct inode *inode = fdentry(file)->d_inode; 1959 struct btrfs_root *root = BTRFS_I(inode)->root; 1960 struct btrfs_trans_handle *trans; 1961 1962 trans = file->private_data; 1963 if (!trans) 1964 return -EINVAL; 1965 file->private_data = NULL; 1966 1967 btrfs_end_transaction(trans, root); 1968 1969 mutex_lock(&root->fs_info->trans_mutex); 1970 root->fs_info->open_ioctl_trans--; 1971 mutex_unlock(&root->fs_info->trans_mutex); 1972 1973 mnt_drop_write(file->f_path.mnt); 1974 return 0; 1975 } 1976 1977 long btrfs_ioctl(struct file *file, unsigned int 1978 cmd, unsigned long arg) 1979 { 1980 struct btrfs_root *root = BTRFS_I(fdentry(file)->d_inode)->root; 1981 void __user *argp = (void __user *)arg; 1982 1983 switch (cmd) { 1984 case FS_IOC_GETFLAGS: 1985 return btrfs_ioctl_getflags(file, argp); 1986 case FS_IOC_SETFLAGS: 1987 return btrfs_ioctl_setflags(file, argp); 1988 case FS_IOC_GETVERSION: 1989 return btrfs_ioctl_getversion(file, argp); 1990 case BTRFS_IOC_SNAP_CREATE: 1991 return btrfs_ioctl_snap_create(file, argp, 0); 1992 case BTRFS_IOC_SUBVOL_CREATE: 1993 return btrfs_ioctl_snap_create(file, argp, 1); 1994 case BTRFS_IOC_SNAP_DESTROY: 1995 return btrfs_ioctl_snap_destroy(file, argp); 1996 case BTRFS_IOC_DEFAULT_SUBVOL: 1997 return btrfs_ioctl_default_subvol(file, argp); 1998 case BTRFS_IOC_DEFRAG: 1999 return btrfs_ioctl_defrag(file, NULL); 2000 case BTRFS_IOC_DEFRAG_RANGE: 2001 return btrfs_ioctl_defrag(file, argp); 2002 case BTRFS_IOC_RESIZE: 2003 return btrfs_ioctl_resize(root, argp); 2004 case BTRFS_IOC_ADD_DEV: 2005 return btrfs_ioctl_add_dev(root, argp); 2006 case BTRFS_IOC_RM_DEV: 2007 return btrfs_ioctl_rm_dev(root, argp); 2008 case BTRFS_IOC_BALANCE: 2009 return btrfs_balance(root->fs_info->dev_root); 2010 case BTRFS_IOC_CLONE: 2011 return btrfs_ioctl_clone(file, arg, 0, 0, 0); 2012 case BTRFS_IOC_CLONE_RANGE: 2013 return btrfs_ioctl_clone_range(file, argp); 2014 case BTRFS_IOC_TRANS_START: 2015 return btrfs_ioctl_trans_start(file); 2016 case BTRFS_IOC_TRANS_END: 2017 return btrfs_ioctl_trans_end(file); 2018 case BTRFS_IOC_TREE_SEARCH: 2019 return btrfs_ioctl_tree_search(file, argp); 2020 case BTRFS_IOC_INO_LOOKUP: 2021 return btrfs_ioctl_ino_lookup(file, argp); 2022 case BTRFS_IOC_SPACE_INFO: 2023 return btrfs_ioctl_space_info(root, argp); 2024 case BTRFS_IOC_SYNC: 2025 btrfs_sync_fs(file->f_dentry->d_sb, 1); 2026 return 0; 2027 } 2028 2029 return -ENOTTY; 2030 } 2031