1 /* 2 * Copyright (C) 2007 Oracle. All rights reserved. 3 * 4 * This program is free software; you can redistribute it and/or 5 * modify it under the terms of the GNU General Public 6 * License v2 as published by the Free Software Foundation. 7 * 8 * This program is distributed in the hope that it will be useful, 9 * but WITHOUT ANY WARRANTY; without even the implied warranty of 10 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU 11 * General Public License for more details. 12 * 13 * You should have received a copy of the GNU General Public 14 * License along with this program; if not, write to the 15 * Free Software Foundation, Inc., 59 Temple Place - Suite 330, 16 * Boston, MA 021110-1307, USA. 17 */ 18 19 #include <linux/kernel.h> 20 #include <linux/bio.h> 21 #include <linux/buffer_head.h> 22 #include <linux/file.h> 23 #include <linux/fs.h> 24 #include <linux/fsnotify.h> 25 #include <linux/pagemap.h> 26 #include <linux/highmem.h> 27 #include <linux/time.h> 28 #include <linux/init.h> 29 #include <linux/string.h> 30 #include <linux/backing-dev.h> 31 #include <linux/mount.h> 32 #include <linux/mpage.h> 33 #include <linux/namei.h> 34 #include <linux/swap.h> 35 #include <linux/writeback.h> 36 #include <linux/statfs.h> 37 #include <linux/compat.h> 38 #include <linux/bit_spinlock.h> 39 #include <linux/security.h> 40 #include <linux/xattr.h> 41 #include <linux/vmalloc.h> 42 #include <linux/slab.h> 43 #include <linux/blkdev.h> 44 #include <linux/uuid.h> 45 #include <linux/btrfs.h> 46 #include <linux/uaccess.h> 47 #include "ctree.h" 48 #include "disk-io.h" 49 #include "transaction.h" 50 #include "btrfs_inode.h" 51 #include "print-tree.h" 52 #include "volumes.h" 53 #include "locking.h" 54 #include "inode-map.h" 55 #include "backref.h" 56 #include "rcu-string.h" 57 #include "send.h" 58 #include "dev-replace.h" 59 #include "props.h" 60 #include "sysfs.h" 61 #include "qgroup.h" 62 #include "tree-log.h" 63 #include "compression.h" 64 65 #ifdef CONFIG_64BIT 66 /* If we have a 32-bit userspace and 64-bit kernel, then the UAPI 67 * structures are incorrect, as the timespec structure from userspace 68 * is 4 bytes too small. We define these alternatives here to teach 69 * the kernel about the 32-bit struct packing. 70 */ 71 struct btrfs_ioctl_timespec_32 { 72 __u64 sec; 73 __u32 nsec; 74 } __attribute__ ((__packed__)); 75 76 struct btrfs_ioctl_received_subvol_args_32 { 77 char uuid[BTRFS_UUID_SIZE]; /* in */ 78 __u64 stransid; /* in */ 79 __u64 rtransid; /* out */ 80 struct btrfs_ioctl_timespec_32 stime; /* in */ 81 struct btrfs_ioctl_timespec_32 rtime; /* out */ 82 __u64 flags; /* in */ 83 __u64 reserved[16]; /* in */ 84 } __attribute__ ((__packed__)); 85 86 #define BTRFS_IOC_SET_RECEIVED_SUBVOL_32 _IOWR(BTRFS_IOCTL_MAGIC, 37, \ 87 struct btrfs_ioctl_received_subvol_args_32) 88 #endif 89 90 91 static int btrfs_clone(struct inode *src, struct inode *inode, 92 u64 off, u64 olen, u64 olen_aligned, u64 destoff, 93 int no_time_update); 94 95 /* Mask out flags that are inappropriate for the given type of inode. */ 96 static inline __u32 btrfs_mask_flags(umode_t mode, __u32 flags) 97 { 98 if (S_ISDIR(mode)) 99 return flags; 100 else if (S_ISREG(mode)) 101 return flags & ~FS_DIRSYNC_FL; 102 else 103 return flags & (FS_NODUMP_FL | FS_NOATIME_FL); 104 } 105 106 /* 107 * Export inode flags to the format expected by the FS_IOC_GETFLAGS ioctl. 108 */ 109 static unsigned int btrfs_flags_to_ioctl(unsigned int flags) 110 { 111 unsigned int iflags = 0; 112 113 if (flags & BTRFS_INODE_SYNC) 114 iflags |= FS_SYNC_FL; 115 if (flags & BTRFS_INODE_IMMUTABLE) 116 iflags |= FS_IMMUTABLE_FL; 117 if (flags & BTRFS_INODE_APPEND) 118 iflags |= FS_APPEND_FL; 119 if (flags & BTRFS_INODE_NODUMP) 120 iflags |= FS_NODUMP_FL; 121 if (flags & BTRFS_INODE_NOATIME) 122 iflags |= FS_NOATIME_FL; 123 if (flags & BTRFS_INODE_DIRSYNC) 124 iflags |= FS_DIRSYNC_FL; 125 if (flags & BTRFS_INODE_NODATACOW) 126 iflags |= FS_NOCOW_FL; 127 128 if (flags & BTRFS_INODE_NOCOMPRESS) 129 iflags |= FS_NOCOMP_FL; 130 else if (flags & BTRFS_INODE_COMPRESS) 131 iflags |= FS_COMPR_FL; 132 133 return iflags; 134 } 135 136 /* 137 * Update inode->i_flags based on the btrfs internal flags. 138 */ 139 void btrfs_update_iflags(struct inode *inode) 140 { 141 struct btrfs_inode *ip = BTRFS_I(inode); 142 unsigned int new_fl = 0; 143 144 if (ip->flags & BTRFS_INODE_SYNC) 145 new_fl |= S_SYNC; 146 if (ip->flags & BTRFS_INODE_IMMUTABLE) 147 new_fl |= S_IMMUTABLE; 148 if (ip->flags & BTRFS_INODE_APPEND) 149 new_fl |= S_APPEND; 150 if (ip->flags & BTRFS_INODE_NOATIME) 151 new_fl |= S_NOATIME; 152 if (ip->flags & BTRFS_INODE_DIRSYNC) 153 new_fl |= S_DIRSYNC; 154 155 set_mask_bits(&inode->i_flags, 156 S_SYNC | S_APPEND | S_IMMUTABLE | S_NOATIME | S_DIRSYNC, 157 new_fl); 158 } 159 160 /* 161 * Inherit flags from the parent inode. 162 * 163 * Currently only the compression flags and the cow flags are inherited. 164 */ 165 void btrfs_inherit_iflags(struct inode *inode, struct inode *dir) 166 { 167 unsigned int flags; 168 169 if (!dir) 170 return; 171 172 flags = BTRFS_I(dir)->flags; 173 174 if (flags & BTRFS_INODE_NOCOMPRESS) { 175 BTRFS_I(inode)->flags &= ~BTRFS_INODE_COMPRESS; 176 BTRFS_I(inode)->flags |= BTRFS_INODE_NOCOMPRESS; 177 } else if (flags & BTRFS_INODE_COMPRESS) { 178 BTRFS_I(inode)->flags &= ~BTRFS_INODE_NOCOMPRESS; 179 BTRFS_I(inode)->flags |= BTRFS_INODE_COMPRESS; 180 } 181 182 if (flags & BTRFS_INODE_NODATACOW) { 183 BTRFS_I(inode)->flags |= BTRFS_INODE_NODATACOW; 184 if (S_ISREG(inode->i_mode)) 185 BTRFS_I(inode)->flags |= BTRFS_INODE_NODATASUM; 186 } 187 188 btrfs_update_iflags(inode); 189 } 190 191 static int btrfs_ioctl_getflags(struct file *file, void __user *arg) 192 { 193 struct btrfs_inode *ip = BTRFS_I(file_inode(file)); 194 unsigned int flags = btrfs_flags_to_ioctl(ip->flags); 195 196 if (copy_to_user(arg, &flags, sizeof(flags))) 197 return -EFAULT; 198 return 0; 199 } 200 201 static int check_flags(unsigned int flags) 202 { 203 if (flags & ~(FS_IMMUTABLE_FL | FS_APPEND_FL | \ 204 FS_NOATIME_FL | FS_NODUMP_FL | \ 205 FS_SYNC_FL | FS_DIRSYNC_FL | \ 206 FS_NOCOMP_FL | FS_COMPR_FL | 207 FS_NOCOW_FL)) 208 return -EOPNOTSUPP; 209 210 if ((flags & FS_NOCOMP_FL) && (flags & FS_COMPR_FL)) 211 return -EINVAL; 212 213 return 0; 214 } 215 216 static int btrfs_ioctl_setflags(struct file *file, void __user *arg) 217 { 218 struct inode *inode = file_inode(file); 219 struct btrfs_inode *ip = BTRFS_I(inode); 220 struct btrfs_root *root = ip->root; 221 struct btrfs_trans_handle *trans; 222 unsigned int flags, oldflags; 223 int ret; 224 u64 ip_oldflags; 225 unsigned int i_oldflags; 226 umode_t mode; 227 228 if (!inode_owner_or_capable(inode)) 229 return -EPERM; 230 231 if (btrfs_root_readonly(root)) 232 return -EROFS; 233 234 if (copy_from_user(&flags, arg, sizeof(flags))) 235 return -EFAULT; 236 237 ret = check_flags(flags); 238 if (ret) 239 return ret; 240 241 ret = mnt_want_write_file(file); 242 if (ret) 243 return ret; 244 245 inode_lock(inode); 246 247 ip_oldflags = ip->flags; 248 i_oldflags = inode->i_flags; 249 mode = inode->i_mode; 250 251 flags = btrfs_mask_flags(inode->i_mode, flags); 252 oldflags = btrfs_flags_to_ioctl(ip->flags); 253 if ((flags ^ oldflags) & (FS_APPEND_FL | FS_IMMUTABLE_FL)) { 254 if (!capable(CAP_LINUX_IMMUTABLE)) { 255 ret = -EPERM; 256 goto out_unlock; 257 } 258 } 259 260 if (flags & FS_SYNC_FL) 261 ip->flags |= BTRFS_INODE_SYNC; 262 else 263 ip->flags &= ~BTRFS_INODE_SYNC; 264 if (flags & FS_IMMUTABLE_FL) 265 ip->flags |= BTRFS_INODE_IMMUTABLE; 266 else 267 ip->flags &= ~BTRFS_INODE_IMMUTABLE; 268 if (flags & FS_APPEND_FL) 269 ip->flags |= BTRFS_INODE_APPEND; 270 else 271 ip->flags &= ~BTRFS_INODE_APPEND; 272 if (flags & FS_NODUMP_FL) 273 ip->flags |= BTRFS_INODE_NODUMP; 274 else 275 ip->flags &= ~BTRFS_INODE_NODUMP; 276 if (flags & FS_NOATIME_FL) 277 ip->flags |= BTRFS_INODE_NOATIME; 278 else 279 ip->flags &= ~BTRFS_INODE_NOATIME; 280 if (flags & FS_DIRSYNC_FL) 281 ip->flags |= BTRFS_INODE_DIRSYNC; 282 else 283 ip->flags &= ~BTRFS_INODE_DIRSYNC; 284 if (flags & FS_NOCOW_FL) { 285 if (S_ISREG(mode)) { 286 /* 287 * It's safe to turn csums off here, no extents exist. 288 * Otherwise we want the flag to reflect the real COW 289 * status of the file and will not set it. 290 */ 291 if (inode->i_size == 0) 292 ip->flags |= BTRFS_INODE_NODATACOW 293 | BTRFS_INODE_NODATASUM; 294 } else { 295 ip->flags |= BTRFS_INODE_NODATACOW; 296 } 297 } else { 298 /* 299 * Revert back under same assumptions as above 300 */ 301 if (S_ISREG(mode)) { 302 if (inode->i_size == 0) 303 ip->flags &= ~(BTRFS_INODE_NODATACOW 304 | BTRFS_INODE_NODATASUM); 305 } else { 306 ip->flags &= ~BTRFS_INODE_NODATACOW; 307 } 308 } 309 310 /* 311 * The COMPRESS flag can only be changed by users, while the NOCOMPRESS 312 * flag may be changed automatically if compression code won't make 313 * things smaller. 314 */ 315 if (flags & FS_NOCOMP_FL) { 316 ip->flags &= ~BTRFS_INODE_COMPRESS; 317 ip->flags |= BTRFS_INODE_NOCOMPRESS; 318 319 ret = btrfs_set_prop(inode, "btrfs.compression", NULL, 0, 0); 320 if (ret && ret != -ENODATA) 321 goto out_drop; 322 } else if (flags & FS_COMPR_FL) { 323 const char *comp; 324 325 ip->flags |= BTRFS_INODE_COMPRESS; 326 ip->flags &= ~BTRFS_INODE_NOCOMPRESS; 327 328 if (root->fs_info->compress_type == BTRFS_COMPRESS_LZO) 329 comp = "lzo"; 330 else 331 comp = "zlib"; 332 ret = btrfs_set_prop(inode, "btrfs.compression", 333 comp, strlen(comp), 0); 334 if (ret) 335 goto out_drop; 336 337 } else { 338 ret = btrfs_set_prop(inode, "btrfs.compression", NULL, 0, 0); 339 if (ret && ret != -ENODATA) 340 goto out_drop; 341 ip->flags &= ~(BTRFS_INODE_COMPRESS | BTRFS_INODE_NOCOMPRESS); 342 } 343 344 trans = btrfs_start_transaction(root, 1); 345 if (IS_ERR(trans)) { 346 ret = PTR_ERR(trans); 347 goto out_drop; 348 } 349 350 btrfs_update_iflags(inode); 351 inode_inc_iversion(inode); 352 inode->i_ctime = current_time(inode); 353 ret = btrfs_update_inode(trans, root, inode); 354 355 btrfs_end_transaction(trans, root); 356 out_drop: 357 if (ret) { 358 ip->flags = ip_oldflags; 359 inode->i_flags = i_oldflags; 360 } 361 362 out_unlock: 363 inode_unlock(inode); 364 mnt_drop_write_file(file); 365 return ret; 366 } 367 368 static int btrfs_ioctl_getversion(struct file *file, int __user *arg) 369 { 370 struct inode *inode = file_inode(file); 371 372 return put_user(inode->i_generation, arg); 373 } 374 375 static noinline int btrfs_ioctl_fitrim(struct file *file, void __user *arg) 376 { 377 struct btrfs_fs_info *fs_info = btrfs_sb(file_inode(file)->i_sb); 378 struct btrfs_device *device; 379 struct request_queue *q; 380 struct fstrim_range range; 381 u64 minlen = ULLONG_MAX; 382 u64 num_devices = 0; 383 u64 total_bytes = btrfs_super_total_bytes(fs_info->super_copy); 384 int ret; 385 386 if (!capable(CAP_SYS_ADMIN)) 387 return -EPERM; 388 389 rcu_read_lock(); 390 list_for_each_entry_rcu(device, &fs_info->fs_devices->devices, 391 dev_list) { 392 if (!device->bdev) 393 continue; 394 q = bdev_get_queue(device->bdev); 395 if (blk_queue_discard(q)) { 396 num_devices++; 397 minlen = min((u64)q->limits.discard_granularity, 398 minlen); 399 } 400 } 401 rcu_read_unlock(); 402 403 if (!num_devices) 404 return -EOPNOTSUPP; 405 if (copy_from_user(&range, arg, sizeof(range))) 406 return -EFAULT; 407 if (range.start > total_bytes || 408 range.len < fs_info->sb->s_blocksize) 409 return -EINVAL; 410 411 range.len = min(range.len, total_bytes - range.start); 412 range.minlen = max(range.minlen, minlen); 413 ret = btrfs_trim_fs(fs_info->tree_root, &range); 414 if (ret < 0) 415 return ret; 416 417 if (copy_to_user(arg, &range, sizeof(range))) 418 return -EFAULT; 419 420 return 0; 421 } 422 423 int btrfs_is_empty_uuid(u8 *uuid) 424 { 425 int i; 426 427 for (i = 0; i < BTRFS_UUID_SIZE; i++) { 428 if (uuid[i]) 429 return 0; 430 } 431 return 1; 432 } 433 434 static noinline int create_subvol(struct inode *dir, 435 struct dentry *dentry, 436 char *name, int namelen, 437 u64 *async_transid, 438 struct btrfs_qgroup_inherit *inherit) 439 { 440 struct btrfs_trans_handle *trans; 441 struct btrfs_key key; 442 struct btrfs_root_item *root_item; 443 struct btrfs_inode_item *inode_item; 444 struct extent_buffer *leaf; 445 struct btrfs_root *root = BTRFS_I(dir)->root; 446 struct btrfs_root *new_root; 447 struct btrfs_block_rsv block_rsv; 448 struct timespec cur_time = current_time(dir); 449 struct inode *inode; 450 int ret; 451 int err; 452 u64 objectid; 453 u64 new_dirid = BTRFS_FIRST_FREE_OBJECTID; 454 u64 index = 0; 455 u64 qgroup_reserved; 456 uuid_le new_uuid; 457 458 root_item = kzalloc(sizeof(*root_item), GFP_KERNEL); 459 if (!root_item) 460 return -ENOMEM; 461 462 ret = btrfs_find_free_objectid(root->fs_info->tree_root, &objectid); 463 if (ret) 464 goto fail_free; 465 466 /* 467 * Don't create subvolume whose level is not zero. Or qgroup will be 468 * screwed up since it assumes subvolume qgroup's level to be 0. 469 */ 470 if (btrfs_qgroup_level(objectid)) { 471 ret = -ENOSPC; 472 goto fail_free; 473 } 474 475 btrfs_init_block_rsv(&block_rsv, BTRFS_BLOCK_RSV_TEMP); 476 /* 477 * The same as the snapshot creation, please see the comment 478 * of create_snapshot(). 479 */ 480 ret = btrfs_subvolume_reserve_metadata(root, &block_rsv, 481 8, &qgroup_reserved, false); 482 if (ret) 483 goto fail_free; 484 485 trans = btrfs_start_transaction(root, 0); 486 if (IS_ERR(trans)) { 487 ret = PTR_ERR(trans); 488 btrfs_subvolume_release_metadata(root, &block_rsv, 489 qgroup_reserved); 490 goto fail_free; 491 } 492 trans->block_rsv = &block_rsv; 493 trans->bytes_reserved = block_rsv.size; 494 495 ret = btrfs_qgroup_inherit(trans, root->fs_info, 0, objectid, inherit); 496 if (ret) 497 goto fail; 498 499 leaf = btrfs_alloc_tree_block(trans, root, 0, objectid, NULL, 0, 0, 0); 500 if (IS_ERR(leaf)) { 501 ret = PTR_ERR(leaf); 502 goto fail; 503 } 504 505 memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header)); 506 btrfs_set_header_bytenr(leaf, leaf->start); 507 btrfs_set_header_generation(leaf, trans->transid); 508 btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV); 509 btrfs_set_header_owner(leaf, objectid); 510 511 write_extent_buffer(leaf, root->fs_info->fsid, btrfs_header_fsid(), 512 BTRFS_FSID_SIZE); 513 write_extent_buffer(leaf, root->fs_info->chunk_tree_uuid, 514 btrfs_header_chunk_tree_uuid(leaf), 515 BTRFS_UUID_SIZE); 516 btrfs_mark_buffer_dirty(leaf); 517 518 inode_item = &root_item->inode; 519 btrfs_set_stack_inode_generation(inode_item, 1); 520 btrfs_set_stack_inode_size(inode_item, 3); 521 btrfs_set_stack_inode_nlink(inode_item, 1); 522 btrfs_set_stack_inode_nbytes(inode_item, root->nodesize); 523 btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755); 524 525 btrfs_set_root_flags(root_item, 0); 526 btrfs_set_root_limit(root_item, 0); 527 btrfs_set_stack_inode_flags(inode_item, BTRFS_INODE_ROOT_ITEM_INIT); 528 529 btrfs_set_root_bytenr(root_item, leaf->start); 530 btrfs_set_root_generation(root_item, trans->transid); 531 btrfs_set_root_level(root_item, 0); 532 btrfs_set_root_refs(root_item, 1); 533 btrfs_set_root_used(root_item, leaf->len); 534 btrfs_set_root_last_snapshot(root_item, 0); 535 536 btrfs_set_root_generation_v2(root_item, 537 btrfs_root_generation(root_item)); 538 uuid_le_gen(&new_uuid); 539 memcpy(root_item->uuid, new_uuid.b, BTRFS_UUID_SIZE); 540 btrfs_set_stack_timespec_sec(&root_item->otime, cur_time.tv_sec); 541 btrfs_set_stack_timespec_nsec(&root_item->otime, cur_time.tv_nsec); 542 root_item->ctime = root_item->otime; 543 btrfs_set_root_ctransid(root_item, trans->transid); 544 btrfs_set_root_otransid(root_item, trans->transid); 545 546 btrfs_tree_unlock(leaf); 547 free_extent_buffer(leaf); 548 leaf = NULL; 549 550 btrfs_set_root_dirid(root_item, new_dirid); 551 552 key.objectid = objectid; 553 key.offset = 0; 554 key.type = BTRFS_ROOT_ITEM_KEY; 555 ret = btrfs_insert_root(trans, root->fs_info->tree_root, &key, 556 root_item); 557 if (ret) 558 goto fail; 559 560 key.offset = (u64)-1; 561 new_root = btrfs_read_fs_root_no_name(root->fs_info, &key); 562 if (IS_ERR(new_root)) { 563 ret = PTR_ERR(new_root); 564 btrfs_abort_transaction(trans, ret); 565 goto fail; 566 } 567 568 btrfs_record_root_in_trans(trans, new_root); 569 570 ret = btrfs_create_subvol_root(trans, new_root, root, new_dirid); 571 if (ret) { 572 /* We potentially lose an unused inode item here */ 573 btrfs_abort_transaction(trans, ret); 574 goto fail; 575 } 576 577 mutex_lock(&new_root->objectid_mutex); 578 new_root->highest_objectid = new_dirid; 579 mutex_unlock(&new_root->objectid_mutex); 580 581 /* 582 * insert the directory item 583 */ 584 ret = btrfs_set_inode_index(dir, &index); 585 if (ret) { 586 btrfs_abort_transaction(trans, ret); 587 goto fail; 588 } 589 590 ret = btrfs_insert_dir_item(trans, root, 591 name, namelen, dir, &key, 592 BTRFS_FT_DIR, index); 593 if (ret) { 594 btrfs_abort_transaction(trans, ret); 595 goto fail; 596 } 597 598 btrfs_i_size_write(dir, dir->i_size + namelen * 2); 599 ret = btrfs_update_inode(trans, root, dir); 600 BUG_ON(ret); 601 602 ret = btrfs_add_root_ref(trans, root->fs_info->tree_root, 603 objectid, root->root_key.objectid, 604 btrfs_ino(dir), index, name, namelen); 605 BUG_ON(ret); 606 607 ret = btrfs_uuid_tree_add(trans, root->fs_info->uuid_root, 608 root_item->uuid, BTRFS_UUID_KEY_SUBVOL, 609 objectid); 610 if (ret) 611 btrfs_abort_transaction(trans, ret); 612 613 fail: 614 kfree(root_item); 615 trans->block_rsv = NULL; 616 trans->bytes_reserved = 0; 617 btrfs_subvolume_release_metadata(root, &block_rsv, qgroup_reserved); 618 619 if (async_transid) { 620 *async_transid = trans->transid; 621 err = btrfs_commit_transaction_async(trans, root, 1); 622 if (err) 623 err = btrfs_commit_transaction(trans, root); 624 } else { 625 err = btrfs_commit_transaction(trans, root); 626 } 627 if (err && !ret) 628 ret = err; 629 630 if (!ret) { 631 inode = btrfs_lookup_dentry(dir, dentry); 632 if (IS_ERR(inode)) 633 return PTR_ERR(inode); 634 d_instantiate(dentry, inode); 635 } 636 return ret; 637 638 fail_free: 639 kfree(root_item); 640 return ret; 641 } 642 643 static void btrfs_wait_for_no_snapshoting_writes(struct btrfs_root *root) 644 { 645 s64 writers; 646 DEFINE_WAIT(wait); 647 648 do { 649 prepare_to_wait(&root->subv_writers->wait, &wait, 650 TASK_UNINTERRUPTIBLE); 651 652 writers = percpu_counter_sum(&root->subv_writers->counter); 653 if (writers) 654 schedule(); 655 656 finish_wait(&root->subv_writers->wait, &wait); 657 } while (writers); 658 } 659 660 static int create_snapshot(struct btrfs_root *root, struct inode *dir, 661 struct dentry *dentry, char *name, int namelen, 662 u64 *async_transid, bool readonly, 663 struct btrfs_qgroup_inherit *inherit) 664 { 665 struct inode *inode; 666 struct btrfs_pending_snapshot *pending_snapshot; 667 struct btrfs_trans_handle *trans; 668 int ret; 669 670 if (!test_bit(BTRFS_ROOT_REF_COWS, &root->state)) 671 return -EINVAL; 672 673 pending_snapshot = kzalloc(sizeof(*pending_snapshot), GFP_NOFS); 674 if (!pending_snapshot) 675 return -ENOMEM; 676 677 pending_snapshot->root_item = kzalloc(sizeof(struct btrfs_root_item), 678 GFP_NOFS); 679 pending_snapshot->path = btrfs_alloc_path(); 680 if (!pending_snapshot->root_item || !pending_snapshot->path) { 681 ret = -ENOMEM; 682 goto free_pending; 683 } 684 685 atomic_inc(&root->will_be_snapshoted); 686 smp_mb__after_atomic(); 687 btrfs_wait_for_no_snapshoting_writes(root); 688 689 ret = btrfs_start_delalloc_inodes(root, 0); 690 if (ret) 691 goto dec_and_free; 692 693 btrfs_wait_ordered_extents(root, -1, 0, (u64)-1); 694 695 btrfs_init_block_rsv(&pending_snapshot->block_rsv, 696 BTRFS_BLOCK_RSV_TEMP); 697 /* 698 * 1 - parent dir inode 699 * 2 - dir entries 700 * 1 - root item 701 * 2 - root ref/backref 702 * 1 - root of snapshot 703 * 1 - UUID item 704 */ 705 ret = btrfs_subvolume_reserve_metadata(BTRFS_I(dir)->root, 706 &pending_snapshot->block_rsv, 8, 707 &pending_snapshot->qgroup_reserved, 708 false); 709 if (ret) 710 goto dec_and_free; 711 712 pending_snapshot->dentry = dentry; 713 pending_snapshot->root = root; 714 pending_snapshot->readonly = readonly; 715 pending_snapshot->dir = dir; 716 pending_snapshot->inherit = inherit; 717 718 trans = btrfs_start_transaction(root, 0); 719 if (IS_ERR(trans)) { 720 ret = PTR_ERR(trans); 721 goto fail; 722 } 723 724 spin_lock(&root->fs_info->trans_lock); 725 list_add(&pending_snapshot->list, 726 &trans->transaction->pending_snapshots); 727 spin_unlock(&root->fs_info->trans_lock); 728 if (async_transid) { 729 *async_transid = trans->transid; 730 ret = btrfs_commit_transaction_async(trans, 731 root->fs_info->extent_root, 1); 732 if (ret) 733 ret = btrfs_commit_transaction(trans, root); 734 } else { 735 ret = btrfs_commit_transaction(trans, 736 root->fs_info->extent_root); 737 } 738 if (ret) 739 goto fail; 740 741 ret = pending_snapshot->error; 742 if (ret) 743 goto fail; 744 745 ret = btrfs_orphan_cleanup(pending_snapshot->snap); 746 if (ret) 747 goto fail; 748 749 inode = btrfs_lookup_dentry(d_inode(dentry->d_parent), dentry); 750 if (IS_ERR(inode)) { 751 ret = PTR_ERR(inode); 752 goto fail; 753 } 754 755 d_instantiate(dentry, inode); 756 ret = 0; 757 fail: 758 btrfs_subvolume_release_metadata(BTRFS_I(dir)->root, 759 &pending_snapshot->block_rsv, 760 pending_snapshot->qgroup_reserved); 761 dec_and_free: 762 if (atomic_dec_and_test(&root->will_be_snapshoted)) 763 wake_up_atomic_t(&root->will_be_snapshoted); 764 free_pending: 765 kfree(pending_snapshot->root_item); 766 btrfs_free_path(pending_snapshot->path); 767 kfree(pending_snapshot); 768 769 return ret; 770 } 771 772 /* copy of may_delete in fs/namei.c() 773 * Check whether we can remove a link victim from directory dir, check 774 * whether the type of victim is right. 775 * 1. We can't do it if dir is read-only (done in permission()) 776 * 2. We should have write and exec permissions on dir 777 * 3. We can't remove anything from append-only dir 778 * 4. We can't do anything with immutable dir (done in permission()) 779 * 5. If the sticky bit on dir is set we should either 780 * a. be owner of dir, or 781 * b. be owner of victim, or 782 * c. have CAP_FOWNER capability 783 * 6. If the victim is append-only or immutable we can't do anything with 784 * links pointing to it. 785 * 7. If we were asked to remove a directory and victim isn't one - ENOTDIR. 786 * 8. If we were asked to remove a non-directory and victim isn't one - EISDIR. 787 * 9. We can't remove a root or mountpoint. 788 * 10. We don't allow removal of NFS sillyrenamed files; it's handled by 789 * nfs_async_unlink(). 790 */ 791 792 static int btrfs_may_delete(struct inode *dir, struct dentry *victim, int isdir) 793 { 794 int error; 795 796 if (d_really_is_negative(victim)) 797 return -ENOENT; 798 799 BUG_ON(d_inode(victim->d_parent) != dir); 800 audit_inode_child(dir, victim, AUDIT_TYPE_CHILD_DELETE); 801 802 error = inode_permission(dir, MAY_WRITE | MAY_EXEC); 803 if (error) 804 return error; 805 if (IS_APPEND(dir)) 806 return -EPERM; 807 if (check_sticky(dir, d_inode(victim)) || IS_APPEND(d_inode(victim)) || 808 IS_IMMUTABLE(d_inode(victim)) || IS_SWAPFILE(d_inode(victim))) 809 return -EPERM; 810 if (isdir) { 811 if (!d_is_dir(victim)) 812 return -ENOTDIR; 813 if (IS_ROOT(victim)) 814 return -EBUSY; 815 } else if (d_is_dir(victim)) 816 return -EISDIR; 817 if (IS_DEADDIR(dir)) 818 return -ENOENT; 819 if (victim->d_flags & DCACHE_NFSFS_RENAMED) 820 return -EBUSY; 821 return 0; 822 } 823 824 /* copy of may_create in fs/namei.c() */ 825 static inline int btrfs_may_create(struct inode *dir, struct dentry *child) 826 { 827 if (d_really_is_positive(child)) 828 return -EEXIST; 829 if (IS_DEADDIR(dir)) 830 return -ENOENT; 831 return inode_permission(dir, MAY_WRITE | MAY_EXEC); 832 } 833 834 /* 835 * Create a new subvolume below @parent. This is largely modeled after 836 * sys_mkdirat and vfs_mkdir, but we only do a single component lookup 837 * inside this filesystem so it's quite a bit simpler. 838 */ 839 static noinline int btrfs_mksubvol(struct path *parent, 840 char *name, int namelen, 841 struct btrfs_root *snap_src, 842 u64 *async_transid, bool readonly, 843 struct btrfs_qgroup_inherit *inherit) 844 { 845 struct inode *dir = d_inode(parent->dentry); 846 struct dentry *dentry; 847 int error; 848 849 error = down_write_killable_nested(&dir->i_rwsem, I_MUTEX_PARENT); 850 if (error == -EINTR) 851 return error; 852 853 dentry = lookup_one_len(name, parent->dentry, namelen); 854 error = PTR_ERR(dentry); 855 if (IS_ERR(dentry)) 856 goto out_unlock; 857 858 error = btrfs_may_create(dir, dentry); 859 if (error) 860 goto out_dput; 861 862 /* 863 * even if this name doesn't exist, we may get hash collisions. 864 * check for them now when we can safely fail 865 */ 866 error = btrfs_check_dir_item_collision(BTRFS_I(dir)->root, 867 dir->i_ino, name, 868 namelen); 869 if (error) 870 goto out_dput; 871 872 down_read(&BTRFS_I(dir)->root->fs_info->subvol_sem); 873 874 if (btrfs_root_refs(&BTRFS_I(dir)->root->root_item) == 0) 875 goto out_up_read; 876 877 if (snap_src) { 878 error = create_snapshot(snap_src, dir, dentry, name, namelen, 879 async_transid, readonly, inherit); 880 } else { 881 error = create_subvol(dir, dentry, name, namelen, 882 async_transid, inherit); 883 } 884 if (!error) 885 fsnotify_mkdir(dir, dentry); 886 out_up_read: 887 up_read(&BTRFS_I(dir)->root->fs_info->subvol_sem); 888 out_dput: 889 dput(dentry); 890 out_unlock: 891 inode_unlock(dir); 892 return error; 893 } 894 895 /* 896 * When we're defragging a range, we don't want to kick it off again 897 * if it is really just waiting for delalloc to send it down. 898 * If we find a nice big extent or delalloc range for the bytes in the 899 * file you want to defrag, we return 0 to let you know to skip this 900 * part of the file 901 */ 902 static int check_defrag_in_cache(struct inode *inode, u64 offset, u32 thresh) 903 { 904 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree; 905 struct extent_map *em = NULL; 906 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree; 907 u64 end; 908 909 read_lock(&em_tree->lock); 910 em = lookup_extent_mapping(em_tree, offset, PAGE_SIZE); 911 read_unlock(&em_tree->lock); 912 913 if (em) { 914 end = extent_map_end(em); 915 free_extent_map(em); 916 if (end - offset > thresh) 917 return 0; 918 } 919 /* if we already have a nice delalloc here, just stop */ 920 thresh /= 2; 921 end = count_range_bits(io_tree, &offset, offset + thresh, 922 thresh, EXTENT_DELALLOC, 1); 923 if (end >= thresh) 924 return 0; 925 return 1; 926 } 927 928 /* 929 * helper function to walk through a file and find extents 930 * newer than a specific transid, and smaller than thresh. 931 * 932 * This is used by the defragging code to find new and small 933 * extents 934 */ 935 static int find_new_extents(struct btrfs_root *root, 936 struct inode *inode, u64 newer_than, 937 u64 *off, u32 thresh) 938 { 939 struct btrfs_path *path; 940 struct btrfs_key min_key; 941 struct extent_buffer *leaf; 942 struct btrfs_file_extent_item *extent; 943 int type; 944 int ret; 945 u64 ino = btrfs_ino(inode); 946 947 path = btrfs_alloc_path(); 948 if (!path) 949 return -ENOMEM; 950 951 min_key.objectid = ino; 952 min_key.type = BTRFS_EXTENT_DATA_KEY; 953 min_key.offset = *off; 954 955 while (1) { 956 ret = btrfs_search_forward(root, &min_key, path, newer_than); 957 if (ret != 0) 958 goto none; 959 process_slot: 960 if (min_key.objectid != ino) 961 goto none; 962 if (min_key.type != BTRFS_EXTENT_DATA_KEY) 963 goto none; 964 965 leaf = path->nodes[0]; 966 extent = btrfs_item_ptr(leaf, path->slots[0], 967 struct btrfs_file_extent_item); 968 969 type = btrfs_file_extent_type(leaf, extent); 970 if (type == BTRFS_FILE_EXTENT_REG && 971 btrfs_file_extent_num_bytes(leaf, extent) < thresh && 972 check_defrag_in_cache(inode, min_key.offset, thresh)) { 973 *off = min_key.offset; 974 btrfs_free_path(path); 975 return 0; 976 } 977 978 path->slots[0]++; 979 if (path->slots[0] < btrfs_header_nritems(leaf)) { 980 btrfs_item_key_to_cpu(leaf, &min_key, path->slots[0]); 981 goto process_slot; 982 } 983 984 if (min_key.offset == (u64)-1) 985 goto none; 986 987 min_key.offset++; 988 btrfs_release_path(path); 989 } 990 none: 991 btrfs_free_path(path); 992 return -ENOENT; 993 } 994 995 static struct extent_map *defrag_lookup_extent(struct inode *inode, u64 start) 996 { 997 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree; 998 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree; 999 struct extent_map *em; 1000 u64 len = PAGE_SIZE; 1001 1002 /* 1003 * hopefully we have this extent in the tree already, try without 1004 * the full extent lock 1005 */ 1006 read_lock(&em_tree->lock); 1007 em = lookup_extent_mapping(em_tree, start, len); 1008 read_unlock(&em_tree->lock); 1009 1010 if (!em) { 1011 struct extent_state *cached = NULL; 1012 u64 end = start + len - 1; 1013 1014 /* get the big lock and read metadata off disk */ 1015 lock_extent_bits(io_tree, start, end, &cached); 1016 em = btrfs_get_extent(inode, NULL, 0, start, len, 0); 1017 unlock_extent_cached(io_tree, start, end, &cached, GFP_NOFS); 1018 1019 if (IS_ERR(em)) 1020 return NULL; 1021 } 1022 1023 return em; 1024 } 1025 1026 static bool defrag_check_next_extent(struct inode *inode, struct extent_map *em) 1027 { 1028 struct extent_map *next; 1029 bool ret = true; 1030 1031 /* this is the last extent */ 1032 if (em->start + em->len >= i_size_read(inode)) 1033 return false; 1034 1035 next = defrag_lookup_extent(inode, em->start + em->len); 1036 if (!next || next->block_start >= EXTENT_MAP_LAST_BYTE) 1037 ret = false; 1038 else if ((em->block_start + em->block_len == next->block_start) && 1039 (em->block_len > SZ_128K && next->block_len > SZ_128K)) 1040 ret = false; 1041 1042 free_extent_map(next); 1043 return ret; 1044 } 1045 1046 static int should_defrag_range(struct inode *inode, u64 start, u32 thresh, 1047 u64 *last_len, u64 *skip, u64 *defrag_end, 1048 int compress) 1049 { 1050 struct extent_map *em; 1051 int ret = 1; 1052 bool next_mergeable = true; 1053 bool prev_mergeable = true; 1054 1055 /* 1056 * make sure that once we start defragging an extent, we keep on 1057 * defragging it 1058 */ 1059 if (start < *defrag_end) 1060 return 1; 1061 1062 *skip = 0; 1063 1064 em = defrag_lookup_extent(inode, start); 1065 if (!em) 1066 return 0; 1067 1068 /* this will cover holes, and inline extents */ 1069 if (em->block_start >= EXTENT_MAP_LAST_BYTE) { 1070 ret = 0; 1071 goto out; 1072 } 1073 1074 if (!*defrag_end) 1075 prev_mergeable = false; 1076 1077 next_mergeable = defrag_check_next_extent(inode, em); 1078 /* 1079 * we hit a real extent, if it is big or the next extent is not a 1080 * real extent, don't bother defragging it 1081 */ 1082 if (!compress && (*last_len == 0 || *last_len >= thresh) && 1083 (em->len >= thresh || (!next_mergeable && !prev_mergeable))) 1084 ret = 0; 1085 out: 1086 /* 1087 * last_len ends up being a counter of how many bytes we've defragged. 1088 * every time we choose not to defrag an extent, we reset *last_len 1089 * so that the next tiny extent will force a defrag. 1090 * 1091 * The end result of this is that tiny extents before a single big 1092 * extent will force at least part of that big extent to be defragged. 1093 */ 1094 if (ret) { 1095 *defrag_end = extent_map_end(em); 1096 } else { 1097 *last_len = 0; 1098 *skip = extent_map_end(em); 1099 *defrag_end = 0; 1100 } 1101 1102 free_extent_map(em); 1103 return ret; 1104 } 1105 1106 /* 1107 * it doesn't do much good to defrag one or two pages 1108 * at a time. This pulls in a nice chunk of pages 1109 * to COW and defrag. 1110 * 1111 * It also makes sure the delalloc code has enough 1112 * dirty data to avoid making new small extents as part 1113 * of the defrag 1114 * 1115 * It's a good idea to start RA on this range 1116 * before calling this. 1117 */ 1118 static int cluster_pages_for_defrag(struct inode *inode, 1119 struct page **pages, 1120 unsigned long start_index, 1121 unsigned long num_pages) 1122 { 1123 unsigned long file_end; 1124 u64 isize = i_size_read(inode); 1125 u64 page_start; 1126 u64 page_end; 1127 u64 page_cnt; 1128 int ret; 1129 int i; 1130 int i_done; 1131 struct btrfs_ordered_extent *ordered; 1132 struct extent_state *cached_state = NULL; 1133 struct extent_io_tree *tree; 1134 gfp_t mask = btrfs_alloc_write_mask(inode->i_mapping); 1135 1136 file_end = (isize - 1) >> PAGE_SHIFT; 1137 if (!isize || start_index > file_end) 1138 return 0; 1139 1140 page_cnt = min_t(u64, (u64)num_pages, (u64)file_end - start_index + 1); 1141 1142 ret = btrfs_delalloc_reserve_space(inode, 1143 start_index << PAGE_SHIFT, 1144 page_cnt << PAGE_SHIFT); 1145 if (ret) 1146 return ret; 1147 i_done = 0; 1148 tree = &BTRFS_I(inode)->io_tree; 1149 1150 /* step one, lock all the pages */ 1151 for (i = 0; i < page_cnt; i++) { 1152 struct page *page; 1153 again: 1154 page = find_or_create_page(inode->i_mapping, 1155 start_index + i, mask); 1156 if (!page) 1157 break; 1158 1159 page_start = page_offset(page); 1160 page_end = page_start + PAGE_SIZE - 1; 1161 while (1) { 1162 lock_extent_bits(tree, page_start, page_end, 1163 &cached_state); 1164 ordered = btrfs_lookup_ordered_extent(inode, 1165 page_start); 1166 unlock_extent_cached(tree, page_start, page_end, 1167 &cached_state, GFP_NOFS); 1168 if (!ordered) 1169 break; 1170 1171 unlock_page(page); 1172 btrfs_start_ordered_extent(inode, ordered, 1); 1173 btrfs_put_ordered_extent(ordered); 1174 lock_page(page); 1175 /* 1176 * we unlocked the page above, so we need check if 1177 * it was released or not. 1178 */ 1179 if (page->mapping != inode->i_mapping) { 1180 unlock_page(page); 1181 put_page(page); 1182 goto again; 1183 } 1184 } 1185 1186 if (!PageUptodate(page)) { 1187 btrfs_readpage(NULL, page); 1188 lock_page(page); 1189 if (!PageUptodate(page)) { 1190 unlock_page(page); 1191 put_page(page); 1192 ret = -EIO; 1193 break; 1194 } 1195 } 1196 1197 if (page->mapping != inode->i_mapping) { 1198 unlock_page(page); 1199 put_page(page); 1200 goto again; 1201 } 1202 1203 pages[i] = page; 1204 i_done++; 1205 } 1206 if (!i_done || ret) 1207 goto out; 1208 1209 if (!(inode->i_sb->s_flags & MS_ACTIVE)) 1210 goto out; 1211 1212 /* 1213 * so now we have a nice long stream of locked 1214 * and up to date pages, lets wait on them 1215 */ 1216 for (i = 0; i < i_done; i++) 1217 wait_on_page_writeback(pages[i]); 1218 1219 page_start = page_offset(pages[0]); 1220 page_end = page_offset(pages[i_done - 1]) + PAGE_SIZE; 1221 1222 lock_extent_bits(&BTRFS_I(inode)->io_tree, 1223 page_start, page_end - 1, &cached_state); 1224 clear_extent_bit(&BTRFS_I(inode)->io_tree, page_start, 1225 page_end - 1, EXTENT_DIRTY | EXTENT_DELALLOC | 1226 EXTENT_DO_ACCOUNTING | EXTENT_DEFRAG, 0, 0, 1227 &cached_state, GFP_NOFS); 1228 1229 if (i_done != page_cnt) { 1230 spin_lock(&BTRFS_I(inode)->lock); 1231 BTRFS_I(inode)->outstanding_extents++; 1232 spin_unlock(&BTRFS_I(inode)->lock); 1233 btrfs_delalloc_release_space(inode, 1234 start_index << PAGE_SHIFT, 1235 (page_cnt - i_done) << PAGE_SHIFT); 1236 } 1237 1238 1239 set_extent_defrag(&BTRFS_I(inode)->io_tree, page_start, page_end - 1, 1240 &cached_state); 1241 1242 unlock_extent_cached(&BTRFS_I(inode)->io_tree, 1243 page_start, page_end - 1, &cached_state, 1244 GFP_NOFS); 1245 1246 for (i = 0; i < i_done; i++) { 1247 clear_page_dirty_for_io(pages[i]); 1248 ClearPageChecked(pages[i]); 1249 set_page_extent_mapped(pages[i]); 1250 set_page_dirty(pages[i]); 1251 unlock_page(pages[i]); 1252 put_page(pages[i]); 1253 } 1254 return i_done; 1255 out: 1256 for (i = 0; i < i_done; i++) { 1257 unlock_page(pages[i]); 1258 put_page(pages[i]); 1259 } 1260 btrfs_delalloc_release_space(inode, 1261 start_index << PAGE_SHIFT, 1262 page_cnt << PAGE_SHIFT); 1263 return ret; 1264 1265 } 1266 1267 int btrfs_defrag_file(struct inode *inode, struct file *file, 1268 struct btrfs_ioctl_defrag_range_args *range, 1269 u64 newer_than, unsigned long max_to_defrag) 1270 { 1271 struct btrfs_root *root = BTRFS_I(inode)->root; 1272 struct file_ra_state *ra = NULL; 1273 unsigned long last_index; 1274 u64 isize = i_size_read(inode); 1275 u64 last_len = 0; 1276 u64 skip = 0; 1277 u64 defrag_end = 0; 1278 u64 newer_off = range->start; 1279 unsigned long i; 1280 unsigned long ra_index = 0; 1281 int ret; 1282 int defrag_count = 0; 1283 int compress_type = BTRFS_COMPRESS_ZLIB; 1284 u32 extent_thresh = range->extent_thresh; 1285 unsigned long max_cluster = SZ_256K >> PAGE_SHIFT; 1286 unsigned long cluster = max_cluster; 1287 u64 new_align = ~((u64)SZ_128K - 1); 1288 struct page **pages = NULL; 1289 1290 if (isize == 0) 1291 return 0; 1292 1293 if (range->start >= isize) 1294 return -EINVAL; 1295 1296 if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS) { 1297 if (range->compress_type > BTRFS_COMPRESS_TYPES) 1298 return -EINVAL; 1299 if (range->compress_type) 1300 compress_type = range->compress_type; 1301 } 1302 1303 if (extent_thresh == 0) 1304 extent_thresh = SZ_256K; 1305 1306 /* 1307 * if we were not given a file, allocate a readahead 1308 * context 1309 */ 1310 if (!file) { 1311 ra = kzalloc(sizeof(*ra), GFP_NOFS); 1312 if (!ra) 1313 return -ENOMEM; 1314 file_ra_state_init(ra, inode->i_mapping); 1315 } else { 1316 ra = &file->f_ra; 1317 } 1318 1319 pages = kmalloc_array(max_cluster, sizeof(struct page *), 1320 GFP_NOFS); 1321 if (!pages) { 1322 ret = -ENOMEM; 1323 goto out_ra; 1324 } 1325 1326 /* find the last page to defrag */ 1327 if (range->start + range->len > range->start) { 1328 last_index = min_t(u64, isize - 1, 1329 range->start + range->len - 1) >> PAGE_SHIFT; 1330 } else { 1331 last_index = (isize - 1) >> PAGE_SHIFT; 1332 } 1333 1334 if (newer_than) { 1335 ret = find_new_extents(root, inode, newer_than, 1336 &newer_off, SZ_64K); 1337 if (!ret) { 1338 range->start = newer_off; 1339 /* 1340 * we always align our defrag to help keep 1341 * the extents in the file evenly spaced 1342 */ 1343 i = (newer_off & new_align) >> PAGE_SHIFT; 1344 } else 1345 goto out_ra; 1346 } else { 1347 i = range->start >> PAGE_SHIFT; 1348 } 1349 if (!max_to_defrag) 1350 max_to_defrag = last_index - i + 1; 1351 1352 /* 1353 * make writeback starts from i, so the defrag range can be 1354 * written sequentially. 1355 */ 1356 if (i < inode->i_mapping->writeback_index) 1357 inode->i_mapping->writeback_index = i; 1358 1359 while (i <= last_index && defrag_count < max_to_defrag && 1360 (i < DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE))) { 1361 /* 1362 * make sure we stop running if someone unmounts 1363 * the FS 1364 */ 1365 if (!(inode->i_sb->s_flags & MS_ACTIVE)) 1366 break; 1367 1368 if (btrfs_defrag_cancelled(root->fs_info)) { 1369 btrfs_debug(root->fs_info, "defrag_file cancelled"); 1370 ret = -EAGAIN; 1371 break; 1372 } 1373 1374 if (!should_defrag_range(inode, (u64)i << PAGE_SHIFT, 1375 extent_thresh, &last_len, &skip, 1376 &defrag_end, range->flags & 1377 BTRFS_DEFRAG_RANGE_COMPRESS)) { 1378 unsigned long next; 1379 /* 1380 * the should_defrag function tells us how much to skip 1381 * bump our counter by the suggested amount 1382 */ 1383 next = DIV_ROUND_UP(skip, PAGE_SIZE); 1384 i = max(i + 1, next); 1385 continue; 1386 } 1387 1388 if (!newer_than) { 1389 cluster = (PAGE_ALIGN(defrag_end) >> 1390 PAGE_SHIFT) - i; 1391 cluster = min(cluster, max_cluster); 1392 } else { 1393 cluster = max_cluster; 1394 } 1395 1396 if (i + cluster > ra_index) { 1397 ra_index = max(i, ra_index); 1398 btrfs_force_ra(inode->i_mapping, ra, file, ra_index, 1399 cluster); 1400 ra_index += cluster; 1401 } 1402 1403 inode_lock(inode); 1404 if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS) 1405 BTRFS_I(inode)->force_compress = compress_type; 1406 ret = cluster_pages_for_defrag(inode, pages, i, cluster); 1407 if (ret < 0) { 1408 inode_unlock(inode); 1409 goto out_ra; 1410 } 1411 1412 defrag_count += ret; 1413 balance_dirty_pages_ratelimited(inode->i_mapping); 1414 inode_unlock(inode); 1415 1416 if (newer_than) { 1417 if (newer_off == (u64)-1) 1418 break; 1419 1420 if (ret > 0) 1421 i += ret; 1422 1423 newer_off = max(newer_off + 1, 1424 (u64)i << PAGE_SHIFT); 1425 1426 ret = find_new_extents(root, inode, newer_than, 1427 &newer_off, SZ_64K); 1428 if (!ret) { 1429 range->start = newer_off; 1430 i = (newer_off & new_align) >> PAGE_SHIFT; 1431 } else { 1432 break; 1433 } 1434 } else { 1435 if (ret > 0) { 1436 i += ret; 1437 last_len += ret << PAGE_SHIFT; 1438 } else { 1439 i++; 1440 last_len = 0; 1441 } 1442 } 1443 } 1444 1445 if ((range->flags & BTRFS_DEFRAG_RANGE_START_IO)) { 1446 filemap_flush(inode->i_mapping); 1447 if (test_bit(BTRFS_INODE_HAS_ASYNC_EXTENT, 1448 &BTRFS_I(inode)->runtime_flags)) 1449 filemap_flush(inode->i_mapping); 1450 } 1451 1452 if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) { 1453 /* the filemap_flush will queue IO into the worker threads, but 1454 * we have to make sure the IO is actually started and that 1455 * ordered extents get created before we return 1456 */ 1457 atomic_inc(&root->fs_info->async_submit_draining); 1458 while (atomic_read(&root->fs_info->nr_async_submits) || 1459 atomic_read(&root->fs_info->async_delalloc_pages)) { 1460 wait_event(root->fs_info->async_submit_wait, 1461 (atomic_read(&root->fs_info->nr_async_submits) == 0 && 1462 atomic_read(&root->fs_info->async_delalloc_pages) == 0)); 1463 } 1464 atomic_dec(&root->fs_info->async_submit_draining); 1465 } 1466 1467 if (range->compress_type == BTRFS_COMPRESS_LZO) { 1468 btrfs_set_fs_incompat(root->fs_info, COMPRESS_LZO); 1469 } 1470 1471 ret = defrag_count; 1472 1473 out_ra: 1474 if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS) { 1475 inode_lock(inode); 1476 BTRFS_I(inode)->force_compress = BTRFS_COMPRESS_NONE; 1477 inode_unlock(inode); 1478 } 1479 if (!file) 1480 kfree(ra); 1481 kfree(pages); 1482 return ret; 1483 } 1484 1485 static noinline int btrfs_ioctl_resize(struct file *file, 1486 void __user *arg) 1487 { 1488 u64 new_size; 1489 u64 old_size; 1490 u64 devid = 1; 1491 struct btrfs_root *root = BTRFS_I(file_inode(file))->root; 1492 struct btrfs_ioctl_vol_args *vol_args; 1493 struct btrfs_trans_handle *trans; 1494 struct btrfs_device *device = NULL; 1495 char *sizestr; 1496 char *retptr; 1497 char *devstr = NULL; 1498 int ret = 0; 1499 int mod = 0; 1500 1501 if (!capable(CAP_SYS_ADMIN)) 1502 return -EPERM; 1503 1504 ret = mnt_want_write_file(file); 1505 if (ret) 1506 return ret; 1507 1508 if (atomic_xchg(&root->fs_info->mutually_exclusive_operation_running, 1509 1)) { 1510 mnt_drop_write_file(file); 1511 return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS; 1512 } 1513 1514 mutex_lock(&root->fs_info->volume_mutex); 1515 vol_args = memdup_user(arg, sizeof(*vol_args)); 1516 if (IS_ERR(vol_args)) { 1517 ret = PTR_ERR(vol_args); 1518 goto out; 1519 } 1520 1521 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; 1522 1523 sizestr = vol_args->name; 1524 devstr = strchr(sizestr, ':'); 1525 if (devstr) { 1526 sizestr = devstr + 1; 1527 *devstr = '\0'; 1528 devstr = vol_args->name; 1529 ret = kstrtoull(devstr, 10, &devid); 1530 if (ret) 1531 goto out_free; 1532 if (!devid) { 1533 ret = -EINVAL; 1534 goto out_free; 1535 } 1536 btrfs_info(root->fs_info, "resizing devid %llu", devid); 1537 } 1538 1539 device = btrfs_find_device(root->fs_info, devid, NULL, NULL); 1540 if (!device) { 1541 btrfs_info(root->fs_info, "resizer unable to find device %llu", 1542 devid); 1543 ret = -ENODEV; 1544 goto out_free; 1545 } 1546 1547 if (!device->writeable) { 1548 btrfs_info(root->fs_info, 1549 "resizer unable to apply on readonly device %llu", 1550 devid); 1551 ret = -EPERM; 1552 goto out_free; 1553 } 1554 1555 if (!strcmp(sizestr, "max")) 1556 new_size = device->bdev->bd_inode->i_size; 1557 else { 1558 if (sizestr[0] == '-') { 1559 mod = -1; 1560 sizestr++; 1561 } else if (sizestr[0] == '+') { 1562 mod = 1; 1563 sizestr++; 1564 } 1565 new_size = memparse(sizestr, &retptr); 1566 if (*retptr != '\0' || new_size == 0) { 1567 ret = -EINVAL; 1568 goto out_free; 1569 } 1570 } 1571 1572 if (device->is_tgtdev_for_dev_replace) { 1573 ret = -EPERM; 1574 goto out_free; 1575 } 1576 1577 old_size = btrfs_device_get_total_bytes(device); 1578 1579 if (mod < 0) { 1580 if (new_size > old_size) { 1581 ret = -EINVAL; 1582 goto out_free; 1583 } 1584 new_size = old_size - new_size; 1585 } else if (mod > 0) { 1586 if (new_size > ULLONG_MAX - old_size) { 1587 ret = -ERANGE; 1588 goto out_free; 1589 } 1590 new_size = old_size + new_size; 1591 } 1592 1593 if (new_size < SZ_256M) { 1594 ret = -EINVAL; 1595 goto out_free; 1596 } 1597 if (new_size > device->bdev->bd_inode->i_size) { 1598 ret = -EFBIG; 1599 goto out_free; 1600 } 1601 1602 new_size = div_u64(new_size, root->sectorsize); 1603 new_size *= root->sectorsize; 1604 1605 btrfs_info_in_rcu(root->fs_info, "new size for %s is %llu", 1606 rcu_str_deref(device->name), new_size); 1607 1608 if (new_size > old_size) { 1609 trans = btrfs_start_transaction(root, 0); 1610 if (IS_ERR(trans)) { 1611 ret = PTR_ERR(trans); 1612 goto out_free; 1613 } 1614 ret = btrfs_grow_device(trans, device, new_size); 1615 btrfs_commit_transaction(trans, root); 1616 } else if (new_size < old_size) { 1617 ret = btrfs_shrink_device(device, new_size); 1618 } /* equal, nothing need to do */ 1619 1620 out_free: 1621 kfree(vol_args); 1622 out: 1623 mutex_unlock(&root->fs_info->volume_mutex); 1624 atomic_set(&root->fs_info->mutually_exclusive_operation_running, 0); 1625 mnt_drop_write_file(file); 1626 return ret; 1627 } 1628 1629 static noinline int btrfs_ioctl_snap_create_transid(struct file *file, 1630 char *name, unsigned long fd, int subvol, 1631 u64 *transid, bool readonly, 1632 struct btrfs_qgroup_inherit *inherit) 1633 { 1634 int namelen; 1635 int ret = 0; 1636 1637 if (!S_ISDIR(file_inode(file)->i_mode)) 1638 return -ENOTDIR; 1639 1640 ret = mnt_want_write_file(file); 1641 if (ret) 1642 goto out; 1643 1644 namelen = strlen(name); 1645 if (strchr(name, '/')) { 1646 ret = -EINVAL; 1647 goto out_drop_write; 1648 } 1649 1650 if (name[0] == '.' && 1651 (namelen == 1 || (name[1] == '.' && namelen == 2))) { 1652 ret = -EEXIST; 1653 goto out_drop_write; 1654 } 1655 1656 if (subvol) { 1657 ret = btrfs_mksubvol(&file->f_path, name, namelen, 1658 NULL, transid, readonly, inherit); 1659 } else { 1660 struct fd src = fdget(fd); 1661 struct inode *src_inode; 1662 if (!src.file) { 1663 ret = -EINVAL; 1664 goto out_drop_write; 1665 } 1666 1667 src_inode = file_inode(src.file); 1668 if (src_inode->i_sb != file_inode(file)->i_sb) { 1669 btrfs_info(BTRFS_I(file_inode(file))->root->fs_info, 1670 "Snapshot src from another FS"); 1671 ret = -EXDEV; 1672 } else if (!inode_owner_or_capable(src_inode)) { 1673 /* 1674 * Subvolume creation is not restricted, but snapshots 1675 * are limited to own subvolumes only 1676 */ 1677 ret = -EPERM; 1678 } else { 1679 ret = btrfs_mksubvol(&file->f_path, name, namelen, 1680 BTRFS_I(src_inode)->root, 1681 transid, readonly, inherit); 1682 } 1683 fdput(src); 1684 } 1685 out_drop_write: 1686 mnt_drop_write_file(file); 1687 out: 1688 return ret; 1689 } 1690 1691 static noinline int btrfs_ioctl_snap_create(struct file *file, 1692 void __user *arg, int subvol) 1693 { 1694 struct btrfs_ioctl_vol_args *vol_args; 1695 int ret; 1696 1697 if (!S_ISDIR(file_inode(file)->i_mode)) 1698 return -ENOTDIR; 1699 1700 vol_args = memdup_user(arg, sizeof(*vol_args)); 1701 if (IS_ERR(vol_args)) 1702 return PTR_ERR(vol_args); 1703 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; 1704 1705 ret = btrfs_ioctl_snap_create_transid(file, vol_args->name, 1706 vol_args->fd, subvol, 1707 NULL, false, NULL); 1708 1709 kfree(vol_args); 1710 return ret; 1711 } 1712 1713 static noinline int btrfs_ioctl_snap_create_v2(struct file *file, 1714 void __user *arg, int subvol) 1715 { 1716 struct btrfs_ioctl_vol_args_v2 *vol_args; 1717 int ret; 1718 u64 transid = 0; 1719 u64 *ptr = NULL; 1720 bool readonly = false; 1721 struct btrfs_qgroup_inherit *inherit = NULL; 1722 1723 if (!S_ISDIR(file_inode(file)->i_mode)) 1724 return -ENOTDIR; 1725 1726 vol_args = memdup_user(arg, sizeof(*vol_args)); 1727 if (IS_ERR(vol_args)) 1728 return PTR_ERR(vol_args); 1729 vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0'; 1730 1731 if (vol_args->flags & 1732 ~(BTRFS_SUBVOL_CREATE_ASYNC | BTRFS_SUBVOL_RDONLY | 1733 BTRFS_SUBVOL_QGROUP_INHERIT)) { 1734 ret = -EOPNOTSUPP; 1735 goto free_args; 1736 } 1737 1738 if (vol_args->flags & BTRFS_SUBVOL_CREATE_ASYNC) 1739 ptr = &transid; 1740 if (vol_args->flags & BTRFS_SUBVOL_RDONLY) 1741 readonly = true; 1742 if (vol_args->flags & BTRFS_SUBVOL_QGROUP_INHERIT) { 1743 if (vol_args->size > PAGE_SIZE) { 1744 ret = -EINVAL; 1745 goto free_args; 1746 } 1747 inherit = memdup_user(vol_args->qgroup_inherit, vol_args->size); 1748 if (IS_ERR(inherit)) { 1749 ret = PTR_ERR(inherit); 1750 goto free_args; 1751 } 1752 } 1753 1754 ret = btrfs_ioctl_snap_create_transid(file, vol_args->name, 1755 vol_args->fd, subvol, ptr, 1756 readonly, inherit); 1757 if (ret) 1758 goto free_inherit; 1759 1760 if (ptr && copy_to_user(arg + 1761 offsetof(struct btrfs_ioctl_vol_args_v2, 1762 transid), 1763 ptr, sizeof(*ptr))) 1764 ret = -EFAULT; 1765 1766 free_inherit: 1767 kfree(inherit); 1768 free_args: 1769 kfree(vol_args); 1770 return ret; 1771 } 1772 1773 static noinline int btrfs_ioctl_subvol_getflags(struct file *file, 1774 void __user *arg) 1775 { 1776 struct inode *inode = file_inode(file); 1777 struct btrfs_root *root = BTRFS_I(inode)->root; 1778 int ret = 0; 1779 u64 flags = 0; 1780 1781 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) 1782 return -EINVAL; 1783 1784 down_read(&root->fs_info->subvol_sem); 1785 if (btrfs_root_readonly(root)) 1786 flags |= BTRFS_SUBVOL_RDONLY; 1787 up_read(&root->fs_info->subvol_sem); 1788 1789 if (copy_to_user(arg, &flags, sizeof(flags))) 1790 ret = -EFAULT; 1791 1792 return ret; 1793 } 1794 1795 static noinline int btrfs_ioctl_subvol_setflags(struct file *file, 1796 void __user *arg) 1797 { 1798 struct inode *inode = file_inode(file); 1799 struct btrfs_root *root = BTRFS_I(inode)->root; 1800 struct btrfs_trans_handle *trans; 1801 u64 root_flags; 1802 u64 flags; 1803 int ret = 0; 1804 1805 if (!inode_owner_or_capable(inode)) 1806 return -EPERM; 1807 1808 ret = mnt_want_write_file(file); 1809 if (ret) 1810 goto out; 1811 1812 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) { 1813 ret = -EINVAL; 1814 goto out_drop_write; 1815 } 1816 1817 if (copy_from_user(&flags, arg, sizeof(flags))) { 1818 ret = -EFAULT; 1819 goto out_drop_write; 1820 } 1821 1822 if (flags & BTRFS_SUBVOL_CREATE_ASYNC) { 1823 ret = -EINVAL; 1824 goto out_drop_write; 1825 } 1826 1827 if (flags & ~BTRFS_SUBVOL_RDONLY) { 1828 ret = -EOPNOTSUPP; 1829 goto out_drop_write; 1830 } 1831 1832 down_write(&root->fs_info->subvol_sem); 1833 1834 /* nothing to do */ 1835 if (!!(flags & BTRFS_SUBVOL_RDONLY) == btrfs_root_readonly(root)) 1836 goto out_drop_sem; 1837 1838 root_flags = btrfs_root_flags(&root->root_item); 1839 if (flags & BTRFS_SUBVOL_RDONLY) { 1840 btrfs_set_root_flags(&root->root_item, 1841 root_flags | BTRFS_ROOT_SUBVOL_RDONLY); 1842 } else { 1843 /* 1844 * Block RO -> RW transition if this subvolume is involved in 1845 * send 1846 */ 1847 spin_lock(&root->root_item_lock); 1848 if (root->send_in_progress == 0) { 1849 btrfs_set_root_flags(&root->root_item, 1850 root_flags & ~BTRFS_ROOT_SUBVOL_RDONLY); 1851 spin_unlock(&root->root_item_lock); 1852 } else { 1853 spin_unlock(&root->root_item_lock); 1854 btrfs_warn(root->fs_info, 1855 "Attempt to set subvolume %llu read-write during send", 1856 root->root_key.objectid); 1857 ret = -EPERM; 1858 goto out_drop_sem; 1859 } 1860 } 1861 1862 trans = btrfs_start_transaction(root, 1); 1863 if (IS_ERR(trans)) { 1864 ret = PTR_ERR(trans); 1865 goto out_reset; 1866 } 1867 1868 ret = btrfs_update_root(trans, root->fs_info->tree_root, 1869 &root->root_key, &root->root_item); 1870 1871 btrfs_commit_transaction(trans, root); 1872 out_reset: 1873 if (ret) 1874 btrfs_set_root_flags(&root->root_item, root_flags); 1875 out_drop_sem: 1876 up_write(&root->fs_info->subvol_sem); 1877 out_drop_write: 1878 mnt_drop_write_file(file); 1879 out: 1880 return ret; 1881 } 1882 1883 /* 1884 * helper to check if the subvolume references other subvolumes 1885 */ 1886 static noinline int may_destroy_subvol(struct btrfs_root *root) 1887 { 1888 struct btrfs_path *path; 1889 struct btrfs_dir_item *di; 1890 struct btrfs_key key; 1891 u64 dir_id; 1892 int ret; 1893 1894 path = btrfs_alloc_path(); 1895 if (!path) 1896 return -ENOMEM; 1897 1898 /* Make sure this root isn't set as the default subvol */ 1899 dir_id = btrfs_super_root_dir(root->fs_info->super_copy); 1900 di = btrfs_lookup_dir_item(NULL, root->fs_info->tree_root, path, 1901 dir_id, "default", 7, 0); 1902 if (di && !IS_ERR(di)) { 1903 btrfs_dir_item_key_to_cpu(path->nodes[0], di, &key); 1904 if (key.objectid == root->root_key.objectid) { 1905 ret = -EPERM; 1906 btrfs_err(root->fs_info, 1907 "deleting default subvolume %llu is not allowed", 1908 key.objectid); 1909 goto out; 1910 } 1911 btrfs_release_path(path); 1912 } 1913 1914 key.objectid = root->root_key.objectid; 1915 key.type = BTRFS_ROOT_REF_KEY; 1916 key.offset = (u64)-1; 1917 1918 ret = btrfs_search_slot(NULL, root->fs_info->tree_root, 1919 &key, path, 0, 0); 1920 if (ret < 0) 1921 goto out; 1922 BUG_ON(ret == 0); 1923 1924 ret = 0; 1925 if (path->slots[0] > 0) { 1926 path->slots[0]--; 1927 btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]); 1928 if (key.objectid == root->root_key.objectid && 1929 key.type == BTRFS_ROOT_REF_KEY) 1930 ret = -ENOTEMPTY; 1931 } 1932 out: 1933 btrfs_free_path(path); 1934 return ret; 1935 } 1936 1937 static noinline int key_in_sk(struct btrfs_key *key, 1938 struct btrfs_ioctl_search_key *sk) 1939 { 1940 struct btrfs_key test; 1941 int ret; 1942 1943 test.objectid = sk->min_objectid; 1944 test.type = sk->min_type; 1945 test.offset = sk->min_offset; 1946 1947 ret = btrfs_comp_cpu_keys(key, &test); 1948 if (ret < 0) 1949 return 0; 1950 1951 test.objectid = sk->max_objectid; 1952 test.type = sk->max_type; 1953 test.offset = sk->max_offset; 1954 1955 ret = btrfs_comp_cpu_keys(key, &test); 1956 if (ret > 0) 1957 return 0; 1958 return 1; 1959 } 1960 1961 static noinline int copy_to_sk(struct btrfs_path *path, 1962 struct btrfs_key *key, 1963 struct btrfs_ioctl_search_key *sk, 1964 size_t *buf_size, 1965 char __user *ubuf, 1966 unsigned long *sk_offset, 1967 int *num_found) 1968 { 1969 u64 found_transid; 1970 struct extent_buffer *leaf; 1971 struct btrfs_ioctl_search_header sh; 1972 struct btrfs_key test; 1973 unsigned long item_off; 1974 unsigned long item_len; 1975 int nritems; 1976 int i; 1977 int slot; 1978 int ret = 0; 1979 1980 leaf = path->nodes[0]; 1981 slot = path->slots[0]; 1982 nritems = btrfs_header_nritems(leaf); 1983 1984 if (btrfs_header_generation(leaf) > sk->max_transid) { 1985 i = nritems; 1986 goto advance_key; 1987 } 1988 found_transid = btrfs_header_generation(leaf); 1989 1990 for (i = slot; i < nritems; i++) { 1991 item_off = btrfs_item_ptr_offset(leaf, i); 1992 item_len = btrfs_item_size_nr(leaf, i); 1993 1994 btrfs_item_key_to_cpu(leaf, key, i); 1995 if (!key_in_sk(key, sk)) 1996 continue; 1997 1998 if (sizeof(sh) + item_len > *buf_size) { 1999 if (*num_found) { 2000 ret = 1; 2001 goto out; 2002 } 2003 2004 /* 2005 * return one empty item back for v1, which does not 2006 * handle -EOVERFLOW 2007 */ 2008 2009 *buf_size = sizeof(sh) + item_len; 2010 item_len = 0; 2011 ret = -EOVERFLOW; 2012 } 2013 2014 if (sizeof(sh) + item_len + *sk_offset > *buf_size) { 2015 ret = 1; 2016 goto out; 2017 } 2018 2019 sh.objectid = key->objectid; 2020 sh.offset = key->offset; 2021 sh.type = key->type; 2022 sh.len = item_len; 2023 sh.transid = found_transid; 2024 2025 /* copy search result header */ 2026 if (copy_to_user(ubuf + *sk_offset, &sh, sizeof(sh))) { 2027 ret = -EFAULT; 2028 goto out; 2029 } 2030 2031 *sk_offset += sizeof(sh); 2032 2033 if (item_len) { 2034 char __user *up = ubuf + *sk_offset; 2035 /* copy the item */ 2036 if (read_extent_buffer_to_user(leaf, up, 2037 item_off, item_len)) { 2038 ret = -EFAULT; 2039 goto out; 2040 } 2041 2042 *sk_offset += item_len; 2043 } 2044 (*num_found)++; 2045 2046 if (ret) /* -EOVERFLOW from above */ 2047 goto out; 2048 2049 if (*num_found >= sk->nr_items) { 2050 ret = 1; 2051 goto out; 2052 } 2053 } 2054 advance_key: 2055 ret = 0; 2056 test.objectid = sk->max_objectid; 2057 test.type = sk->max_type; 2058 test.offset = sk->max_offset; 2059 if (btrfs_comp_cpu_keys(key, &test) >= 0) 2060 ret = 1; 2061 else if (key->offset < (u64)-1) 2062 key->offset++; 2063 else if (key->type < (u8)-1) { 2064 key->offset = 0; 2065 key->type++; 2066 } else if (key->objectid < (u64)-1) { 2067 key->offset = 0; 2068 key->type = 0; 2069 key->objectid++; 2070 } else 2071 ret = 1; 2072 out: 2073 /* 2074 * 0: all items from this leaf copied, continue with next 2075 * 1: * more items can be copied, but unused buffer is too small 2076 * * all items were found 2077 * Either way, it will stops the loop which iterates to the next 2078 * leaf 2079 * -EOVERFLOW: item was to large for buffer 2080 * -EFAULT: could not copy extent buffer back to userspace 2081 */ 2082 return ret; 2083 } 2084 2085 static noinline int search_ioctl(struct inode *inode, 2086 struct btrfs_ioctl_search_key *sk, 2087 size_t *buf_size, 2088 char __user *ubuf) 2089 { 2090 struct btrfs_root *root; 2091 struct btrfs_key key; 2092 struct btrfs_path *path; 2093 struct btrfs_fs_info *info = BTRFS_I(inode)->root->fs_info; 2094 int ret; 2095 int num_found = 0; 2096 unsigned long sk_offset = 0; 2097 2098 if (*buf_size < sizeof(struct btrfs_ioctl_search_header)) { 2099 *buf_size = sizeof(struct btrfs_ioctl_search_header); 2100 return -EOVERFLOW; 2101 } 2102 2103 path = btrfs_alloc_path(); 2104 if (!path) 2105 return -ENOMEM; 2106 2107 if (sk->tree_id == 0) { 2108 /* search the root of the inode that was passed */ 2109 root = BTRFS_I(inode)->root; 2110 } else { 2111 key.objectid = sk->tree_id; 2112 key.type = BTRFS_ROOT_ITEM_KEY; 2113 key.offset = (u64)-1; 2114 root = btrfs_read_fs_root_no_name(info, &key); 2115 if (IS_ERR(root)) { 2116 btrfs_free_path(path); 2117 return -ENOENT; 2118 } 2119 } 2120 2121 key.objectid = sk->min_objectid; 2122 key.type = sk->min_type; 2123 key.offset = sk->min_offset; 2124 2125 while (1) { 2126 ret = btrfs_search_forward(root, &key, path, sk->min_transid); 2127 if (ret != 0) { 2128 if (ret > 0) 2129 ret = 0; 2130 goto err; 2131 } 2132 ret = copy_to_sk(path, &key, sk, buf_size, ubuf, 2133 &sk_offset, &num_found); 2134 btrfs_release_path(path); 2135 if (ret) 2136 break; 2137 2138 } 2139 if (ret > 0) 2140 ret = 0; 2141 err: 2142 sk->nr_items = num_found; 2143 btrfs_free_path(path); 2144 return ret; 2145 } 2146 2147 static noinline int btrfs_ioctl_tree_search(struct file *file, 2148 void __user *argp) 2149 { 2150 struct btrfs_ioctl_search_args __user *uargs; 2151 struct btrfs_ioctl_search_key sk; 2152 struct inode *inode; 2153 int ret; 2154 size_t buf_size; 2155 2156 if (!capable(CAP_SYS_ADMIN)) 2157 return -EPERM; 2158 2159 uargs = (struct btrfs_ioctl_search_args __user *)argp; 2160 2161 if (copy_from_user(&sk, &uargs->key, sizeof(sk))) 2162 return -EFAULT; 2163 2164 buf_size = sizeof(uargs->buf); 2165 2166 inode = file_inode(file); 2167 ret = search_ioctl(inode, &sk, &buf_size, uargs->buf); 2168 2169 /* 2170 * In the origin implementation an overflow is handled by returning a 2171 * search header with a len of zero, so reset ret. 2172 */ 2173 if (ret == -EOVERFLOW) 2174 ret = 0; 2175 2176 if (ret == 0 && copy_to_user(&uargs->key, &sk, sizeof(sk))) 2177 ret = -EFAULT; 2178 return ret; 2179 } 2180 2181 static noinline int btrfs_ioctl_tree_search_v2(struct file *file, 2182 void __user *argp) 2183 { 2184 struct btrfs_ioctl_search_args_v2 __user *uarg; 2185 struct btrfs_ioctl_search_args_v2 args; 2186 struct inode *inode; 2187 int ret; 2188 size_t buf_size; 2189 const size_t buf_limit = SZ_16M; 2190 2191 if (!capable(CAP_SYS_ADMIN)) 2192 return -EPERM; 2193 2194 /* copy search header and buffer size */ 2195 uarg = (struct btrfs_ioctl_search_args_v2 __user *)argp; 2196 if (copy_from_user(&args, uarg, sizeof(args))) 2197 return -EFAULT; 2198 2199 buf_size = args.buf_size; 2200 2201 if (buf_size < sizeof(struct btrfs_ioctl_search_header)) 2202 return -EOVERFLOW; 2203 2204 /* limit result size to 16MB */ 2205 if (buf_size > buf_limit) 2206 buf_size = buf_limit; 2207 2208 inode = file_inode(file); 2209 ret = search_ioctl(inode, &args.key, &buf_size, 2210 (char *)(&uarg->buf[0])); 2211 if (ret == 0 && copy_to_user(&uarg->key, &args.key, sizeof(args.key))) 2212 ret = -EFAULT; 2213 else if (ret == -EOVERFLOW && 2214 copy_to_user(&uarg->buf_size, &buf_size, sizeof(buf_size))) 2215 ret = -EFAULT; 2216 2217 return ret; 2218 } 2219 2220 /* 2221 * Search INODE_REFs to identify path name of 'dirid' directory 2222 * in a 'tree_id' tree. and sets path name to 'name'. 2223 */ 2224 static noinline int btrfs_search_path_in_tree(struct btrfs_fs_info *info, 2225 u64 tree_id, u64 dirid, char *name) 2226 { 2227 struct btrfs_root *root; 2228 struct btrfs_key key; 2229 char *ptr; 2230 int ret = -1; 2231 int slot; 2232 int len; 2233 int total_len = 0; 2234 struct btrfs_inode_ref *iref; 2235 struct extent_buffer *l; 2236 struct btrfs_path *path; 2237 2238 if (dirid == BTRFS_FIRST_FREE_OBJECTID) { 2239 name[0]='\0'; 2240 return 0; 2241 } 2242 2243 path = btrfs_alloc_path(); 2244 if (!path) 2245 return -ENOMEM; 2246 2247 ptr = &name[BTRFS_INO_LOOKUP_PATH_MAX]; 2248 2249 key.objectid = tree_id; 2250 key.type = BTRFS_ROOT_ITEM_KEY; 2251 key.offset = (u64)-1; 2252 root = btrfs_read_fs_root_no_name(info, &key); 2253 if (IS_ERR(root)) { 2254 btrfs_err(info, "could not find root %llu", tree_id); 2255 ret = -ENOENT; 2256 goto out; 2257 } 2258 2259 key.objectid = dirid; 2260 key.type = BTRFS_INODE_REF_KEY; 2261 key.offset = (u64)-1; 2262 2263 while (1) { 2264 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); 2265 if (ret < 0) 2266 goto out; 2267 else if (ret > 0) { 2268 ret = btrfs_previous_item(root, path, dirid, 2269 BTRFS_INODE_REF_KEY); 2270 if (ret < 0) 2271 goto out; 2272 else if (ret > 0) { 2273 ret = -ENOENT; 2274 goto out; 2275 } 2276 } 2277 2278 l = path->nodes[0]; 2279 slot = path->slots[0]; 2280 btrfs_item_key_to_cpu(l, &key, slot); 2281 2282 iref = btrfs_item_ptr(l, slot, struct btrfs_inode_ref); 2283 len = btrfs_inode_ref_name_len(l, iref); 2284 ptr -= len + 1; 2285 total_len += len + 1; 2286 if (ptr < name) { 2287 ret = -ENAMETOOLONG; 2288 goto out; 2289 } 2290 2291 *(ptr + len) = '/'; 2292 read_extent_buffer(l, ptr, (unsigned long)(iref + 1), len); 2293 2294 if (key.offset == BTRFS_FIRST_FREE_OBJECTID) 2295 break; 2296 2297 btrfs_release_path(path); 2298 key.objectid = key.offset; 2299 key.offset = (u64)-1; 2300 dirid = key.objectid; 2301 } 2302 memmove(name, ptr, total_len); 2303 name[total_len] = '\0'; 2304 ret = 0; 2305 out: 2306 btrfs_free_path(path); 2307 return ret; 2308 } 2309 2310 static noinline int btrfs_ioctl_ino_lookup(struct file *file, 2311 void __user *argp) 2312 { 2313 struct btrfs_ioctl_ino_lookup_args *args; 2314 struct inode *inode; 2315 int ret = 0; 2316 2317 args = memdup_user(argp, sizeof(*args)); 2318 if (IS_ERR(args)) 2319 return PTR_ERR(args); 2320 2321 inode = file_inode(file); 2322 2323 /* 2324 * Unprivileged query to obtain the containing subvolume root id. The 2325 * path is reset so it's consistent with btrfs_search_path_in_tree. 2326 */ 2327 if (args->treeid == 0) 2328 args->treeid = BTRFS_I(inode)->root->root_key.objectid; 2329 2330 if (args->objectid == BTRFS_FIRST_FREE_OBJECTID) { 2331 args->name[0] = 0; 2332 goto out; 2333 } 2334 2335 if (!capable(CAP_SYS_ADMIN)) { 2336 ret = -EPERM; 2337 goto out; 2338 } 2339 2340 ret = btrfs_search_path_in_tree(BTRFS_I(inode)->root->fs_info, 2341 args->treeid, args->objectid, 2342 args->name); 2343 2344 out: 2345 if (ret == 0 && copy_to_user(argp, args, sizeof(*args))) 2346 ret = -EFAULT; 2347 2348 kfree(args); 2349 return ret; 2350 } 2351 2352 static noinline int btrfs_ioctl_snap_destroy(struct file *file, 2353 void __user *arg) 2354 { 2355 struct dentry *parent = file->f_path.dentry; 2356 struct dentry *dentry; 2357 struct inode *dir = d_inode(parent); 2358 struct inode *inode; 2359 struct btrfs_root *root = BTRFS_I(dir)->root; 2360 struct btrfs_root *dest = NULL; 2361 struct btrfs_ioctl_vol_args *vol_args; 2362 struct btrfs_trans_handle *trans; 2363 struct btrfs_block_rsv block_rsv; 2364 u64 root_flags; 2365 u64 qgroup_reserved; 2366 int namelen; 2367 int ret; 2368 int err = 0; 2369 2370 if (!S_ISDIR(dir->i_mode)) 2371 return -ENOTDIR; 2372 2373 vol_args = memdup_user(arg, sizeof(*vol_args)); 2374 if (IS_ERR(vol_args)) 2375 return PTR_ERR(vol_args); 2376 2377 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; 2378 namelen = strlen(vol_args->name); 2379 if (strchr(vol_args->name, '/') || 2380 strncmp(vol_args->name, "..", namelen) == 0) { 2381 err = -EINVAL; 2382 goto out; 2383 } 2384 2385 err = mnt_want_write_file(file); 2386 if (err) 2387 goto out; 2388 2389 2390 err = down_write_killable_nested(&dir->i_rwsem, I_MUTEX_PARENT); 2391 if (err == -EINTR) 2392 goto out_drop_write; 2393 dentry = lookup_one_len(vol_args->name, parent, namelen); 2394 if (IS_ERR(dentry)) { 2395 err = PTR_ERR(dentry); 2396 goto out_unlock_dir; 2397 } 2398 2399 if (d_really_is_negative(dentry)) { 2400 err = -ENOENT; 2401 goto out_dput; 2402 } 2403 2404 inode = d_inode(dentry); 2405 dest = BTRFS_I(inode)->root; 2406 if (!capable(CAP_SYS_ADMIN)) { 2407 /* 2408 * Regular user. Only allow this with a special mount 2409 * option, when the user has write+exec access to the 2410 * subvol root, and when rmdir(2) would have been 2411 * allowed. 2412 * 2413 * Note that this is _not_ check that the subvol is 2414 * empty or doesn't contain data that we wouldn't 2415 * otherwise be able to delete. 2416 * 2417 * Users who want to delete empty subvols should try 2418 * rmdir(2). 2419 */ 2420 err = -EPERM; 2421 if (!btrfs_test_opt(root->fs_info, USER_SUBVOL_RM_ALLOWED)) 2422 goto out_dput; 2423 2424 /* 2425 * Do not allow deletion if the parent dir is the same 2426 * as the dir to be deleted. That means the ioctl 2427 * must be called on the dentry referencing the root 2428 * of the subvol, not a random directory contained 2429 * within it. 2430 */ 2431 err = -EINVAL; 2432 if (root == dest) 2433 goto out_dput; 2434 2435 err = inode_permission(inode, MAY_WRITE | MAY_EXEC); 2436 if (err) 2437 goto out_dput; 2438 } 2439 2440 /* check if subvolume may be deleted by a user */ 2441 err = btrfs_may_delete(dir, dentry, 1); 2442 if (err) 2443 goto out_dput; 2444 2445 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) { 2446 err = -EINVAL; 2447 goto out_dput; 2448 } 2449 2450 inode_lock(inode); 2451 2452 /* 2453 * Don't allow to delete a subvolume with send in progress. This is 2454 * inside the i_mutex so the error handling that has to drop the bit 2455 * again is not run concurrently. 2456 */ 2457 spin_lock(&dest->root_item_lock); 2458 root_flags = btrfs_root_flags(&dest->root_item); 2459 if (dest->send_in_progress == 0) { 2460 btrfs_set_root_flags(&dest->root_item, 2461 root_flags | BTRFS_ROOT_SUBVOL_DEAD); 2462 spin_unlock(&dest->root_item_lock); 2463 } else { 2464 spin_unlock(&dest->root_item_lock); 2465 btrfs_warn(root->fs_info, 2466 "Attempt to delete subvolume %llu during send", 2467 dest->root_key.objectid); 2468 err = -EPERM; 2469 goto out_unlock_inode; 2470 } 2471 2472 down_write(&root->fs_info->subvol_sem); 2473 2474 err = may_destroy_subvol(dest); 2475 if (err) 2476 goto out_up_write; 2477 2478 btrfs_init_block_rsv(&block_rsv, BTRFS_BLOCK_RSV_TEMP); 2479 /* 2480 * One for dir inode, two for dir entries, two for root 2481 * ref/backref. 2482 */ 2483 err = btrfs_subvolume_reserve_metadata(root, &block_rsv, 2484 5, &qgroup_reserved, true); 2485 if (err) 2486 goto out_up_write; 2487 2488 trans = btrfs_start_transaction(root, 0); 2489 if (IS_ERR(trans)) { 2490 err = PTR_ERR(trans); 2491 goto out_release; 2492 } 2493 trans->block_rsv = &block_rsv; 2494 trans->bytes_reserved = block_rsv.size; 2495 2496 btrfs_record_snapshot_destroy(trans, dir); 2497 2498 ret = btrfs_unlink_subvol(trans, root, dir, 2499 dest->root_key.objectid, 2500 dentry->d_name.name, 2501 dentry->d_name.len); 2502 if (ret) { 2503 err = ret; 2504 btrfs_abort_transaction(trans, ret); 2505 goto out_end_trans; 2506 } 2507 2508 btrfs_record_root_in_trans(trans, dest); 2509 2510 memset(&dest->root_item.drop_progress, 0, 2511 sizeof(dest->root_item.drop_progress)); 2512 dest->root_item.drop_level = 0; 2513 btrfs_set_root_refs(&dest->root_item, 0); 2514 2515 if (!test_and_set_bit(BTRFS_ROOT_ORPHAN_ITEM_INSERTED, &dest->state)) { 2516 ret = btrfs_insert_orphan_item(trans, 2517 root->fs_info->tree_root, 2518 dest->root_key.objectid); 2519 if (ret) { 2520 btrfs_abort_transaction(trans, ret); 2521 err = ret; 2522 goto out_end_trans; 2523 } 2524 } 2525 2526 ret = btrfs_uuid_tree_rem(trans, root->fs_info->uuid_root, 2527 dest->root_item.uuid, BTRFS_UUID_KEY_SUBVOL, 2528 dest->root_key.objectid); 2529 if (ret && ret != -ENOENT) { 2530 btrfs_abort_transaction(trans, ret); 2531 err = ret; 2532 goto out_end_trans; 2533 } 2534 if (!btrfs_is_empty_uuid(dest->root_item.received_uuid)) { 2535 ret = btrfs_uuid_tree_rem(trans, root->fs_info->uuid_root, 2536 dest->root_item.received_uuid, 2537 BTRFS_UUID_KEY_RECEIVED_SUBVOL, 2538 dest->root_key.objectid); 2539 if (ret && ret != -ENOENT) { 2540 btrfs_abort_transaction(trans, ret); 2541 err = ret; 2542 goto out_end_trans; 2543 } 2544 } 2545 2546 out_end_trans: 2547 trans->block_rsv = NULL; 2548 trans->bytes_reserved = 0; 2549 ret = btrfs_end_transaction(trans, root); 2550 if (ret && !err) 2551 err = ret; 2552 inode->i_flags |= S_DEAD; 2553 out_release: 2554 btrfs_subvolume_release_metadata(root, &block_rsv, qgroup_reserved); 2555 out_up_write: 2556 up_write(&root->fs_info->subvol_sem); 2557 if (err) { 2558 spin_lock(&dest->root_item_lock); 2559 root_flags = btrfs_root_flags(&dest->root_item); 2560 btrfs_set_root_flags(&dest->root_item, 2561 root_flags & ~BTRFS_ROOT_SUBVOL_DEAD); 2562 spin_unlock(&dest->root_item_lock); 2563 } 2564 out_unlock_inode: 2565 inode_unlock(inode); 2566 if (!err) { 2567 d_invalidate(dentry); 2568 btrfs_invalidate_inodes(dest); 2569 d_delete(dentry); 2570 ASSERT(dest->send_in_progress == 0); 2571 2572 /* the last ref */ 2573 if (dest->ino_cache_inode) { 2574 iput(dest->ino_cache_inode); 2575 dest->ino_cache_inode = NULL; 2576 } 2577 } 2578 out_dput: 2579 dput(dentry); 2580 out_unlock_dir: 2581 inode_unlock(dir); 2582 out_drop_write: 2583 mnt_drop_write_file(file); 2584 out: 2585 kfree(vol_args); 2586 return err; 2587 } 2588 2589 static int btrfs_ioctl_defrag(struct file *file, void __user *argp) 2590 { 2591 struct inode *inode = file_inode(file); 2592 struct btrfs_root *root = BTRFS_I(inode)->root; 2593 struct btrfs_ioctl_defrag_range_args *range; 2594 int ret; 2595 2596 ret = mnt_want_write_file(file); 2597 if (ret) 2598 return ret; 2599 2600 if (btrfs_root_readonly(root)) { 2601 ret = -EROFS; 2602 goto out; 2603 } 2604 2605 switch (inode->i_mode & S_IFMT) { 2606 case S_IFDIR: 2607 if (!capable(CAP_SYS_ADMIN)) { 2608 ret = -EPERM; 2609 goto out; 2610 } 2611 ret = btrfs_defrag_root(root); 2612 if (ret) 2613 goto out; 2614 ret = btrfs_defrag_root(root->fs_info->extent_root); 2615 break; 2616 case S_IFREG: 2617 if (!(file->f_mode & FMODE_WRITE)) { 2618 ret = -EINVAL; 2619 goto out; 2620 } 2621 2622 range = kzalloc(sizeof(*range), GFP_KERNEL); 2623 if (!range) { 2624 ret = -ENOMEM; 2625 goto out; 2626 } 2627 2628 if (argp) { 2629 if (copy_from_user(range, argp, 2630 sizeof(*range))) { 2631 ret = -EFAULT; 2632 kfree(range); 2633 goto out; 2634 } 2635 /* compression requires us to start the IO */ 2636 if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) { 2637 range->flags |= BTRFS_DEFRAG_RANGE_START_IO; 2638 range->extent_thresh = (u32)-1; 2639 } 2640 } else { 2641 /* the rest are all set to zero by kzalloc */ 2642 range->len = (u64)-1; 2643 } 2644 ret = btrfs_defrag_file(file_inode(file), file, 2645 range, 0, 0); 2646 if (ret > 0) 2647 ret = 0; 2648 kfree(range); 2649 break; 2650 default: 2651 ret = -EINVAL; 2652 } 2653 out: 2654 mnt_drop_write_file(file); 2655 return ret; 2656 } 2657 2658 static long btrfs_ioctl_add_dev(struct btrfs_root *root, void __user *arg) 2659 { 2660 struct btrfs_ioctl_vol_args *vol_args; 2661 int ret; 2662 2663 if (!capable(CAP_SYS_ADMIN)) 2664 return -EPERM; 2665 2666 if (atomic_xchg(&root->fs_info->mutually_exclusive_operation_running, 2667 1)) { 2668 return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS; 2669 } 2670 2671 mutex_lock(&root->fs_info->volume_mutex); 2672 vol_args = memdup_user(arg, sizeof(*vol_args)); 2673 if (IS_ERR(vol_args)) { 2674 ret = PTR_ERR(vol_args); 2675 goto out; 2676 } 2677 2678 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; 2679 ret = btrfs_init_new_device(root, vol_args->name); 2680 2681 if (!ret) 2682 btrfs_info(root->fs_info, "disk added %s",vol_args->name); 2683 2684 kfree(vol_args); 2685 out: 2686 mutex_unlock(&root->fs_info->volume_mutex); 2687 atomic_set(&root->fs_info->mutually_exclusive_operation_running, 0); 2688 return ret; 2689 } 2690 2691 static long btrfs_ioctl_rm_dev_v2(struct file *file, void __user *arg) 2692 { 2693 struct btrfs_root *root = BTRFS_I(file_inode(file))->root; 2694 struct btrfs_ioctl_vol_args_v2 *vol_args; 2695 int ret; 2696 2697 if (!capable(CAP_SYS_ADMIN)) 2698 return -EPERM; 2699 2700 ret = mnt_want_write_file(file); 2701 if (ret) 2702 return ret; 2703 2704 vol_args = memdup_user(arg, sizeof(*vol_args)); 2705 if (IS_ERR(vol_args)) { 2706 ret = PTR_ERR(vol_args); 2707 goto err_drop; 2708 } 2709 2710 /* Check for compatibility reject unknown flags */ 2711 if (vol_args->flags & ~BTRFS_VOL_ARG_V2_FLAGS_SUPPORTED) 2712 return -EOPNOTSUPP; 2713 2714 if (atomic_xchg(&root->fs_info->mutually_exclusive_operation_running, 2715 1)) { 2716 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS; 2717 goto out; 2718 } 2719 2720 mutex_lock(&root->fs_info->volume_mutex); 2721 if (vol_args->flags & BTRFS_DEVICE_SPEC_BY_ID) { 2722 ret = btrfs_rm_device(root, NULL, vol_args->devid); 2723 } else { 2724 vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0'; 2725 ret = btrfs_rm_device(root, vol_args->name, 0); 2726 } 2727 mutex_unlock(&root->fs_info->volume_mutex); 2728 atomic_set(&root->fs_info->mutually_exclusive_operation_running, 0); 2729 2730 if (!ret) { 2731 if (vol_args->flags & BTRFS_DEVICE_SPEC_BY_ID) 2732 btrfs_info(root->fs_info, "device deleted: id %llu", 2733 vol_args->devid); 2734 else 2735 btrfs_info(root->fs_info, "device deleted: %s", 2736 vol_args->name); 2737 } 2738 out: 2739 kfree(vol_args); 2740 err_drop: 2741 mnt_drop_write_file(file); 2742 return ret; 2743 } 2744 2745 static long btrfs_ioctl_rm_dev(struct file *file, void __user *arg) 2746 { 2747 struct btrfs_root *root = BTRFS_I(file_inode(file))->root; 2748 struct btrfs_ioctl_vol_args *vol_args; 2749 int ret; 2750 2751 if (!capable(CAP_SYS_ADMIN)) 2752 return -EPERM; 2753 2754 ret = mnt_want_write_file(file); 2755 if (ret) 2756 return ret; 2757 2758 if (atomic_xchg(&root->fs_info->mutually_exclusive_operation_running, 2759 1)) { 2760 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS; 2761 goto out_drop_write; 2762 } 2763 2764 vol_args = memdup_user(arg, sizeof(*vol_args)); 2765 if (IS_ERR(vol_args)) { 2766 ret = PTR_ERR(vol_args); 2767 goto out; 2768 } 2769 2770 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; 2771 mutex_lock(&root->fs_info->volume_mutex); 2772 ret = btrfs_rm_device(root, vol_args->name, 0); 2773 mutex_unlock(&root->fs_info->volume_mutex); 2774 2775 if (!ret) 2776 btrfs_info(root->fs_info, "disk deleted %s",vol_args->name); 2777 kfree(vol_args); 2778 out: 2779 atomic_set(&root->fs_info->mutually_exclusive_operation_running, 0); 2780 out_drop_write: 2781 mnt_drop_write_file(file); 2782 2783 return ret; 2784 } 2785 2786 static long btrfs_ioctl_fs_info(struct btrfs_root *root, void __user *arg) 2787 { 2788 struct btrfs_ioctl_fs_info_args *fi_args; 2789 struct btrfs_device *device; 2790 struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices; 2791 int ret = 0; 2792 2793 fi_args = kzalloc(sizeof(*fi_args), GFP_KERNEL); 2794 if (!fi_args) 2795 return -ENOMEM; 2796 2797 mutex_lock(&fs_devices->device_list_mutex); 2798 fi_args->num_devices = fs_devices->num_devices; 2799 memcpy(&fi_args->fsid, root->fs_info->fsid, sizeof(fi_args->fsid)); 2800 2801 list_for_each_entry(device, &fs_devices->devices, dev_list) { 2802 if (device->devid > fi_args->max_id) 2803 fi_args->max_id = device->devid; 2804 } 2805 mutex_unlock(&fs_devices->device_list_mutex); 2806 2807 fi_args->nodesize = root->fs_info->super_copy->nodesize; 2808 fi_args->sectorsize = root->fs_info->super_copy->sectorsize; 2809 fi_args->clone_alignment = root->fs_info->super_copy->sectorsize; 2810 2811 if (copy_to_user(arg, fi_args, sizeof(*fi_args))) 2812 ret = -EFAULT; 2813 2814 kfree(fi_args); 2815 return ret; 2816 } 2817 2818 static long btrfs_ioctl_dev_info(struct btrfs_root *root, void __user *arg) 2819 { 2820 struct btrfs_ioctl_dev_info_args *di_args; 2821 struct btrfs_device *dev; 2822 struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices; 2823 int ret = 0; 2824 char *s_uuid = NULL; 2825 2826 di_args = memdup_user(arg, sizeof(*di_args)); 2827 if (IS_ERR(di_args)) 2828 return PTR_ERR(di_args); 2829 2830 if (!btrfs_is_empty_uuid(di_args->uuid)) 2831 s_uuid = di_args->uuid; 2832 2833 mutex_lock(&fs_devices->device_list_mutex); 2834 dev = btrfs_find_device(root->fs_info, di_args->devid, s_uuid, NULL); 2835 2836 if (!dev) { 2837 ret = -ENODEV; 2838 goto out; 2839 } 2840 2841 di_args->devid = dev->devid; 2842 di_args->bytes_used = btrfs_device_get_bytes_used(dev); 2843 di_args->total_bytes = btrfs_device_get_total_bytes(dev); 2844 memcpy(di_args->uuid, dev->uuid, sizeof(di_args->uuid)); 2845 if (dev->name) { 2846 struct rcu_string *name; 2847 2848 rcu_read_lock(); 2849 name = rcu_dereference(dev->name); 2850 strncpy(di_args->path, name->str, sizeof(di_args->path)); 2851 rcu_read_unlock(); 2852 di_args->path[sizeof(di_args->path) - 1] = 0; 2853 } else { 2854 di_args->path[0] = '\0'; 2855 } 2856 2857 out: 2858 mutex_unlock(&fs_devices->device_list_mutex); 2859 if (ret == 0 && copy_to_user(arg, di_args, sizeof(*di_args))) 2860 ret = -EFAULT; 2861 2862 kfree(di_args); 2863 return ret; 2864 } 2865 2866 static struct page *extent_same_get_page(struct inode *inode, pgoff_t index) 2867 { 2868 struct page *page; 2869 2870 page = grab_cache_page(inode->i_mapping, index); 2871 if (!page) 2872 return ERR_PTR(-ENOMEM); 2873 2874 if (!PageUptodate(page)) { 2875 int ret; 2876 2877 ret = btrfs_readpage(NULL, page); 2878 if (ret) 2879 return ERR_PTR(ret); 2880 lock_page(page); 2881 if (!PageUptodate(page)) { 2882 unlock_page(page); 2883 put_page(page); 2884 return ERR_PTR(-EIO); 2885 } 2886 if (page->mapping != inode->i_mapping) { 2887 unlock_page(page); 2888 put_page(page); 2889 return ERR_PTR(-EAGAIN); 2890 } 2891 } 2892 2893 return page; 2894 } 2895 2896 static int gather_extent_pages(struct inode *inode, struct page **pages, 2897 int num_pages, u64 off) 2898 { 2899 int i; 2900 pgoff_t index = off >> PAGE_SHIFT; 2901 2902 for (i = 0; i < num_pages; i++) { 2903 again: 2904 pages[i] = extent_same_get_page(inode, index + i); 2905 if (IS_ERR(pages[i])) { 2906 int err = PTR_ERR(pages[i]); 2907 2908 if (err == -EAGAIN) 2909 goto again; 2910 pages[i] = NULL; 2911 return err; 2912 } 2913 } 2914 return 0; 2915 } 2916 2917 static int lock_extent_range(struct inode *inode, u64 off, u64 len, 2918 bool retry_range_locking) 2919 { 2920 /* 2921 * Do any pending delalloc/csum calculations on inode, one way or 2922 * another, and lock file content. 2923 * The locking order is: 2924 * 2925 * 1) pages 2926 * 2) range in the inode's io tree 2927 */ 2928 while (1) { 2929 struct btrfs_ordered_extent *ordered; 2930 lock_extent(&BTRFS_I(inode)->io_tree, off, off + len - 1); 2931 ordered = btrfs_lookup_first_ordered_extent(inode, 2932 off + len - 1); 2933 if ((!ordered || 2934 ordered->file_offset + ordered->len <= off || 2935 ordered->file_offset >= off + len) && 2936 !test_range_bit(&BTRFS_I(inode)->io_tree, off, 2937 off + len - 1, EXTENT_DELALLOC, 0, NULL)) { 2938 if (ordered) 2939 btrfs_put_ordered_extent(ordered); 2940 break; 2941 } 2942 unlock_extent(&BTRFS_I(inode)->io_tree, off, off + len - 1); 2943 if (ordered) 2944 btrfs_put_ordered_extent(ordered); 2945 if (!retry_range_locking) 2946 return -EAGAIN; 2947 btrfs_wait_ordered_range(inode, off, len); 2948 } 2949 return 0; 2950 } 2951 2952 static void btrfs_double_inode_unlock(struct inode *inode1, struct inode *inode2) 2953 { 2954 inode_unlock(inode1); 2955 inode_unlock(inode2); 2956 } 2957 2958 static void btrfs_double_inode_lock(struct inode *inode1, struct inode *inode2) 2959 { 2960 if (inode1 < inode2) 2961 swap(inode1, inode2); 2962 2963 inode_lock_nested(inode1, I_MUTEX_PARENT); 2964 inode_lock_nested(inode2, I_MUTEX_CHILD); 2965 } 2966 2967 static void btrfs_double_extent_unlock(struct inode *inode1, u64 loff1, 2968 struct inode *inode2, u64 loff2, u64 len) 2969 { 2970 unlock_extent(&BTRFS_I(inode1)->io_tree, loff1, loff1 + len - 1); 2971 unlock_extent(&BTRFS_I(inode2)->io_tree, loff2, loff2 + len - 1); 2972 } 2973 2974 static int btrfs_double_extent_lock(struct inode *inode1, u64 loff1, 2975 struct inode *inode2, u64 loff2, u64 len, 2976 bool retry_range_locking) 2977 { 2978 int ret; 2979 2980 if (inode1 < inode2) { 2981 swap(inode1, inode2); 2982 swap(loff1, loff2); 2983 } 2984 ret = lock_extent_range(inode1, loff1, len, retry_range_locking); 2985 if (ret) 2986 return ret; 2987 ret = lock_extent_range(inode2, loff2, len, retry_range_locking); 2988 if (ret) 2989 unlock_extent(&BTRFS_I(inode1)->io_tree, loff1, 2990 loff1 + len - 1); 2991 return ret; 2992 } 2993 2994 struct cmp_pages { 2995 int num_pages; 2996 struct page **src_pages; 2997 struct page **dst_pages; 2998 }; 2999 3000 static void btrfs_cmp_data_free(struct cmp_pages *cmp) 3001 { 3002 int i; 3003 struct page *pg; 3004 3005 for (i = 0; i < cmp->num_pages; i++) { 3006 pg = cmp->src_pages[i]; 3007 if (pg) { 3008 unlock_page(pg); 3009 put_page(pg); 3010 } 3011 pg = cmp->dst_pages[i]; 3012 if (pg) { 3013 unlock_page(pg); 3014 put_page(pg); 3015 } 3016 } 3017 kfree(cmp->src_pages); 3018 kfree(cmp->dst_pages); 3019 } 3020 3021 static int btrfs_cmp_data_prepare(struct inode *src, u64 loff, 3022 struct inode *dst, u64 dst_loff, 3023 u64 len, struct cmp_pages *cmp) 3024 { 3025 int ret; 3026 int num_pages = PAGE_ALIGN(len) >> PAGE_SHIFT; 3027 struct page **src_pgarr, **dst_pgarr; 3028 3029 /* 3030 * We must gather up all the pages before we initiate our 3031 * extent locking. We use an array for the page pointers. Size 3032 * of the array is bounded by len, which is in turn bounded by 3033 * BTRFS_MAX_DEDUPE_LEN. 3034 */ 3035 src_pgarr = kcalloc(num_pages, sizeof(struct page *), GFP_KERNEL); 3036 dst_pgarr = kcalloc(num_pages, sizeof(struct page *), GFP_KERNEL); 3037 if (!src_pgarr || !dst_pgarr) { 3038 kfree(src_pgarr); 3039 kfree(dst_pgarr); 3040 return -ENOMEM; 3041 } 3042 cmp->num_pages = num_pages; 3043 cmp->src_pages = src_pgarr; 3044 cmp->dst_pages = dst_pgarr; 3045 3046 ret = gather_extent_pages(src, cmp->src_pages, cmp->num_pages, loff); 3047 if (ret) 3048 goto out; 3049 3050 ret = gather_extent_pages(dst, cmp->dst_pages, cmp->num_pages, dst_loff); 3051 3052 out: 3053 if (ret) 3054 btrfs_cmp_data_free(cmp); 3055 return 0; 3056 } 3057 3058 static int btrfs_cmp_data(struct inode *src, u64 loff, struct inode *dst, 3059 u64 dst_loff, u64 len, struct cmp_pages *cmp) 3060 { 3061 int ret = 0; 3062 int i; 3063 struct page *src_page, *dst_page; 3064 unsigned int cmp_len = PAGE_SIZE; 3065 void *addr, *dst_addr; 3066 3067 i = 0; 3068 while (len) { 3069 if (len < PAGE_SIZE) 3070 cmp_len = len; 3071 3072 BUG_ON(i >= cmp->num_pages); 3073 3074 src_page = cmp->src_pages[i]; 3075 dst_page = cmp->dst_pages[i]; 3076 ASSERT(PageLocked(src_page)); 3077 ASSERT(PageLocked(dst_page)); 3078 3079 addr = kmap_atomic(src_page); 3080 dst_addr = kmap_atomic(dst_page); 3081 3082 flush_dcache_page(src_page); 3083 flush_dcache_page(dst_page); 3084 3085 if (memcmp(addr, dst_addr, cmp_len)) 3086 ret = -EBADE; 3087 3088 kunmap_atomic(addr); 3089 kunmap_atomic(dst_addr); 3090 3091 if (ret) 3092 break; 3093 3094 len -= cmp_len; 3095 i++; 3096 } 3097 3098 return ret; 3099 } 3100 3101 static int extent_same_check_offsets(struct inode *inode, u64 off, u64 *plen, 3102 u64 olen) 3103 { 3104 u64 len = *plen; 3105 u64 bs = BTRFS_I(inode)->root->fs_info->sb->s_blocksize; 3106 3107 if (off + olen > inode->i_size || off + olen < off) 3108 return -EINVAL; 3109 3110 /* if we extend to eof, continue to block boundary */ 3111 if (off + len == inode->i_size) 3112 *plen = len = ALIGN(inode->i_size, bs) - off; 3113 3114 /* Check that we are block aligned - btrfs_clone() requires this */ 3115 if (!IS_ALIGNED(off, bs) || !IS_ALIGNED(off + len, bs)) 3116 return -EINVAL; 3117 3118 return 0; 3119 } 3120 3121 static int btrfs_extent_same(struct inode *src, u64 loff, u64 olen, 3122 struct inode *dst, u64 dst_loff) 3123 { 3124 int ret; 3125 u64 len = olen; 3126 struct cmp_pages cmp; 3127 int same_inode = 0; 3128 u64 same_lock_start = 0; 3129 u64 same_lock_len = 0; 3130 3131 if (src == dst) 3132 same_inode = 1; 3133 3134 if (len == 0) 3135 return 0; 3136 3137 if (same_inode) { 3138 inode_lock(src); 3139 3140 ret = extent_same_check_offsets(src, loff, &len, olen); 3141 if (ret) 3142 goto out_unlock; 3143 ret = extent_same_check_offsets(src, dst_loff, &len, olen); 3144 if (ret) 3145 goto out_unlock; 3146 3147 /* 3148 * Single inode case wants the same checks, except we 3149 * don't want our length pushed out past i_size as 3150 * comparing that data range makes no sense. 3151 * 3152 * extent_same_check_offsets() will do this for an 3153 * unaligned length at i_size, so catch it here and 3154 * reject the request. 3155 * 3156 * This effectively means we require aligned extents 3157 * for the single-inode case, whereas the other cases 3158 * allow an unaligned length so long as it ends at 3159 * i_size. 3160 */ 3161 if (len != olen) { 3162 ret = -EINVAL; 3163 goto out_unlock; 3164 } 3165 3166 /* Check for overlapping ranges */ 3167 if (dst_loff + len > loff && dst_loff < loff + len) { 3168 ret = -EINVAL; 3169 goto out_unlock; 3170 } 3171 3172 same_lock_start = min_t(u64, loff, dst_loff); 3173 same_lock_len = max_t(u64, loff, dst_loff) + len - same_lock_start; 3174 } else { 3175 btrfs_double_inode_lock(src, dst); 3176 3177 ret = extent_same_check_offsets(src, loff, &len, olen); 3178 if (ret) 3179 goto out_unlock; 3180 3181 ret = extent_same_check_offsets(dst, dst_loff, &len, olen); 3182 if (ret) 3183 goto out_unlock; 3184 } 3185 3186 /* don't make the dst file partly checksummed */ 3187 if ((BTRFS_I(src)->flags & BTRFS_INODE_NODATASUM) != 3188 (BTRFS_I(dst)->flags & BTRFS_INODE_NODATASUM)) { 3189 ret = -EINVAL; 3190 goto out_unlock; 3191 } 3192 3193 again: 3194 ret = btrfs_cmp_data_prepare(src, loff, dst, dst_loff, olen, &cmp); 3195 if (ret) 3196 goto out_unlock; 3197 3198 if (same_inode) 3199 ret = lock_extent_range(src, same_lock_start, same_lock_len, 3200 false); 3201 else 3202 ret = btrfs_double_extent_lock(src, loff, dst, dst_loff, len, 3203 false); 3204 /* 3205 * If one of the inodes has dirty pages in the respective range or 3206 * ordered extents, we need to flush dellaloc and wait for all ordered 3207 * extents in the range. We must unlock the pages and the ranges in the 3208 * io trees to avoid deadlocks when flushing delalloc (requires locking 3209 * pages) and when waiting for ordered extents to complete (they require 3210 * range locking). 3211 */ 3212 if (ret == -EAGAIN) { 3213 /* 3214 * Ranges in the io trees already unlocked. Now unlock all 3215 * pages before waiting for all IO to complete. 3216 */ 3217 btrfs_cmp_data_free(&cmp); 3218 if (same_inode) { 3219 btrfs_wait_ordered_range(src, same_lock_start, 3220 same_lock_len); 3221 } else { 3222 btrfs_wait_ordered_range(src, loff, len); 3223 btrfs_wait_ordered_range(dst, dst_loff, len); 3224 } 3225 goto again; 3226 } 3227 ASSERT(ret == 0); 3228 if (WARN_ON(ret)) { 3229 /* ranges in the io trees already unlocked */ 3230 btrfs_cmp_data_free(&cmp); 3231 return ret; 3232 } 3233 3234 /* pass original length for comparison so we stay within i_size */ 3235 ret = btrfs_cmp_data(src, loff, dst, dst_loff, olen, &cmp); 3236 if (ret == 0) 3237 ret = btrfs_clone(src, dst, loff, olen, len, dst_loff, 1); 3238 3239 if (same_inode) 3240 unlock_extent(&BTRFS_I(src)->io_tree, same_lock_start, 3241 same_lock_start + same_lock_len - 1); 3242 else 3243 btrfs_double_extent_unlock(src, loff, dst, dst_loff, len); 3244 3245 btrfs_cmp_data_free(&cmp); 3246 out_unlock: 3247 if (same_inode) 3248 inode_unlock(src); 3249 else 3250 btrfs_double_inode_unlock(src, dst); 3251 3252 return ret; 3253 } 3254 3255 #define BTRFS_MAX_DEDUPE_LEN SZ_16M 3256 3257 ssize_t btrfs_dedupe_file_range(struct file *src_file, u64 loff, u64 olen, 3258 struct file *dst_file, u64 dst_loff) 3259 { 3260 struct inode *src = file_inode(src_file); 3261 struct inode *dst = file_inode(dst_file); 3262 u64 bs = BTRFS_I(src)->root->fs_info->sb->s_blocksize; 3263 ssize_t res; 3264 3265 if (olen > BTRFS_MAX_DEDUPE_LEN) 3266 olen = BTRFS_MAX_DEDUPE_LEN; 3267 3268 if (WARN_ON_ONCE(bs < PAGE_SIZE)) { 3269 /* 3270 * Btrfs does not support blocksize < page_size. As a 3271 * result, btrfs_cmp_data() won't correctly handle 3272 * this situation without an update. 3273 */ 3274 return -EINVAL; 3275 } 3276 3277 res = btrfs_extent_same(src, loff, olen, dst, dst_loff); 3278 if (res) 3279 return res; 3280 return olen; 3281 } 3282 3283 static int clone_finish_inode_update(struct btrfs_trans_handle *trans, 3284 struct inode *inode, 3285 u64 endoff, 3286 const u64 destoff, 3287 const u64 olen, 3288 int no_time_update) 3289 { 3290 struct btrfs_root *root = BTRFS_I(inode)->root; 3291 int ret; 3292 3293 inode_inc_iversion(inode); 3294 if (!no_time_update) 3295 inode->i_mtime = inode->i_ctime = current_time(inode); 3296 /* 3297 * We round up to the block size at eof when determining which 3298 * extents to clone above, but shouldn't round up the file size. 3299 */ 3300 if (endoff > destoff + olen) 3301 endoff = destoff + olen; 3302 if (endoff > inode->i_size) 3303 btrfs_i_size_write(inode, endoff); 3304 3305 ret = btrfs_update_inode(trans, root, inode); 3306 if (ret) { 3307 btrfs_abort_transaction(trans, ret); 3308 btrfs_end_transaction(trans, root); 3309 goto out; 3310 } 3311 ret = btrfs_end_transaction(trans, root); 3312 out: 3313 return ret; 3314 } 3315 3316 static void clone_update_extent_map(struct inode *inode, 3317 const struct btrfs_trans_handle *trans, 3318 const struct btrfs_path *path, 3319 const u64 hole_offset, 3320 const u64 hole_len) 3321 { 3322 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree; 3323 struct extent_map *em; 3324 int ret; 3325 3326 em = alloc_extent_map(); 3327 if (!em) { 3328 set_bit(BTRFS_INODE_NEEDS_FULL_SYNC, 3329 &BTRFS_I(inode)->runtime_flags); 3330 return; 3331 } 3332 3333 if (path) { 3334 struct btrfs_file_extent_item *fi; 3335 3336 fi = btrfs_item_ptr(path->nodes[0], path->slots[0], 3337 struct btrfs_file_extent_item); 3338 btrfs_extent_item_to_extent_map(inode, path, fi, false, em); 3339 em->generation = -1; 3340 if (btrfs_file_extent_type(path->nodes[0], fi) == 3341 BTRFS_FILE_EXTENT_INLINE) 3342 set_bit(BTRFS_INODE_NEEDS_FULL_SYNC, 3343 &BTRFS_I(inode)->runtime_flags); 3344 } else { 3345 em->start = hole_offset; 3346 em->len = hole_len; 3347 em->ram_bytes = em->len; 3348 em->orig_start = hole_offset; 3349 em->block_start = EXTENT_MAP_HOLE; 3350 em->block_len = 0; 3351 em->orig_block_len = 0; 3352 em->compress_type = BTRFS_COMPRESS_NONE; 3353 em->generation = trans->transid; 3354 } 3355 3356 while (1) { 3357 write_lock(&em_tree->lock); 3358 ret = add_extent_mapping(em_tree, em, 1); 3359 write_unlock(&em_tree->lock); 3360 if (ret != -EEXIST) { 3361 free_extent_map(em); 3362 break; 3363 } 3364 btrfs_drop_extent_cache(inode, em->start, 3365 em->start + em->len - 1, 0); 3366 } 3367 3368 if (ret) 3369 set_bit(BTRFS_INODE_NEEDS_FULL_SYNC, 3370 &BTRFS_I(inode)->runtime_flags); 3371 } 3372 3373 /* 3374 * Make sure we do not end up inserting an inline extent into a file that has 3375 * already other (non-inline) extents. If a file has an inline extent it can 3376 * not have any other extents and the (single) inline extent must start at the 3377 * file offset 0. Failing to respect these rules will lead to file corruption, 3378 * resulting in EIO errors on read/write operations, hitting BUG_ON's in mm, etc 3379 * 3380 * We can have extents that have been already written to disk or we can have 3381 * dirty ranges still in delalloc, in which case the extent maps and items are 3382 * created only when we run delalloc, and the delalloc ranges might fall outside 3383 * the range we are currently locking in the inode's io tree. So we check the 3384 * inode's i_size because of that (i_size updates are done while holding the 3385 * i_mutex, which we are holding here). 3386 * We also check to see if the inode has a size not greater than "datal" but has 3387 * extents beyond it, due to an fallocate with FALLOC_FL_KEEP_SIZE (and we are 3388 * protected against such concurrent fallocate calls by the i_mutex). 3389 * 3390 * If the file has no extents but a size greater than datal, do not allow the 3391 * copy because we would need turn the inline extent into a non-inline one (even 3392 * with NO_HOLES enabled). If we find our destination inode only has one inline 3393 * extent, just overwrite it with the source inline extent if its size is less 3394 * than the source extent's size, or we could copy the source inline extent's 3395 * data into the destination inode's inline extent if the later is greater then 3396 * the former. 3397 */ 3398 static int clone_copy_inline_extent(struct inode *src, 3399 struct inode *dst, 3400 struct btrfs_trans_handle *trans, 3401 struct btrfs_path *path, 3402 struct btrfs_key *new_key, 3403 const u64 drop_start, 3404 const u64 datal, 3405 const u64 skip, 3406 const u64 size, 3407 char *inline_data) 3408 { 3409 struct btrfs_root *root = BTRFS_I(dst)->root; 3410 const u64 aligned_end = ALIGN(new_key->offset + datal, 3411 root->sectorsize); 3412 int ret; 3413 struct btrfs_key key; 3414 3415 if (new_key->offset > 0) 3416 return -EOPNOTSUPP; 3417 3418 key.objectid = btrfs_ino(dst); 3419 key.type = BTRFS_EXTENT_DATA_KEY; 3420 key.offset = 0; 3421 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); 3422 if (ret < 0) { 3423 return ret; 3424 } else if (ret > 0) { 3425 if (path->slots[0] >= btrfs_header_nritems(path->nodes[0])) { 3426 ret = btrfs_next_leaf(root, path); 3427 if (ret < 0) 3428 return ret; 3429 else if (ret > 0) 3430 goto copy_inline_extent; 3431 } 3432 btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]); 3433 if (key.objectid == btrfs_ino(dst) && 3434 key.type == BTRFS_EXTENT_DATA_KEY) { 3435 ASSERT(key.offset > 0); 3436 return -EOPNOTSUPP; 3437 } 3438 } else if (i_size_read(dst) <= datal) { 3439 struct btrfs_file_extent_item *ei; 3440 u64 ext_len; 3441 3442 /* 3443 * If the file size is <= datal, make sure there are no other 3444 * extents following (can happen do to an fallocate call with 3445 * the flag FALLOC_FL_KEEP_SIZE). 3446 */ 3447 ei = btrfs_item_ptr(path->nodes[0], path->slots[0], 3448 struct btrfs_file_extent_item); 3449 /* 3450 * If it's an inline extent, it can not have other extents 3451 * following it. 3452 */ 3453 if (btrfs_file_extent_type(path->nodes[0], ei) == 3454 BTRFS_FILE_EXTENT_INLINE) 3455 goto copy_inline_extent; 3456 3457 ext_len = btrfs_file_extent_num_bytes(path->nodes[0], ei); 3458 if (ext_len > aligned_end) 3459 return -EOPNOTSUPP; 3460 3461 ret = btrfs_next_item(root, path); 3462 if (ret < 0) { 3463 return ret; 3464 } else if (ret == 0) { 3465 btrfs_item_key_to_cpu(path->nodes[0], &key, 3466 path->slots[0]); 3467 if (key.objectid == btrfs_ino(dst) && 3468 key.type == BTRFS_EXTENT_DATA_KEY) 3469 return -EOPNOTSUPP; 3470 } 3471 } 3472 3473 copy_inline_extent: 3474 /* 3475 * We have no extent items, or we have an extent at offset 0 which may 3476 * or may not be inlined. All these cases are dealt the same way. 3477 */ 3478 if (i_size_read(dst) > datal) { 3479 /* 3480 * If the destination inode has an inline extent... 3481 * This would require copying the data from the source inline 3482 * extent into the beginning of the destination's inline extent. 3483 * But this is really complex, both extents can be compressed 3484 * or just one of them, which would require decompressing and 3485 * re-compressing data (which could increase the new compressed 3486 * size, not allowing the compressed data to fit anymore in an 3487 * inline extent). 3488 * So just don't support this case for now (it should be rare, 3489 * we are not really saving space when cloning inline extents). 3490 */ 3491 return -EOPNOTSUPP; 3492 } 3493 3494 btrfs_release_path(path); 3495 ret = btrfs_drop_extents(trans, root, dst, drop_start, aligned_end, 1); 3496 if (ret) 3497 return ret; 3498 ret = btrfs_insert_empty_item(trans, root, path, new_key, size); 3499 if (ret) 3500 return ret; 3501 3502 if (skip) { 3503 const u32 start = btrfs_file_extent_calc_inline_size(0); 3504 3505 memmove(inline_data + start, inline_data + start + skip, datal); 3506 } 3507 3508 write_extent_buffer(path->nodes[0], inline_data, 3509 btrfs_item_ptr_offset(path->nodes[0], 3510 path->slots[0]), 3511 size); 3512 inode_add_bytes(dst, datal); 3513 3514 return 0; 3515 } 3516 3517 /** 3518 * btrfs_clone() - clone a range from inode file to another 3519 * 3520 * @src: Inode to clone from 3521 * @inode: Inode to clone to 3522 * @off: Offset within source to start clone from 3523 * @olen: Original length, passed by user, of range to clone 3524 * @olen_aligned: Block-aligned value of olen 3525 * @destoff: Offset within @inode to start clone 3526 * @no_time_update: Whether to update mtime/ctime on the target inode 3527 */ 3528 static int btrfs_clone(struct inode *src, struct inode *inode, 3529 const u64 off, const u64 olen, const u64 olen_aligned, 3530 const u64 destoff, int no_time_update) 3531 { 3532 struct btrfs_root *root = BTRFS_I(inode)->root; 3533 struct btrfs_path *path = NULL; 3534 struct extent_buffer *leaf; 3535 struct btrfs_trans_handle *trans; 3536 char *buf = NULL; 3537 struct btrfs_key key; 3538 u32 nritems; 3539 int slot; 3540 int ret; 3541 const u64 len = olen_aligned; 3542 u64 last_dest_end = destoff; 3543 3544 ret = -ENOMEM; 3545 buf = kmalloc(root->nodesize, GFP_KERNEL | __GFP_NOWARN); 3546 if (!buf) { 3547 buf = vmalloc(root->nodesize); 3548 if (!buf) 3549 return ret; 3550 } 3551 3552 path = btrfs_alloc_path(); 3553 if (!path) { 3554 kvfree(buf); 3555 return ret; 3556 } 3557 3558 path->reada = READA_FORWARD; 3559 /* clone data */ 3560 key.objectid = btrfs_ino(src); 3561 key.type = BTRFS_EXTENT_DATA_KEY; 3562 key.offset = off; 3563 3564 while (1) { 3565 u64 next_key_min_offset = key.offset + 1; 3566 3567 /* 3568 * note the key will change type as we walk through the 3569 * tree. 3570 */ 3571 path->leave_spinning = 1; 3572 ret = btrfs_search_slot(NULL, BTRFS_I(src)->root, &key, path, 3573 0, 0); 3574 if (ret < 0) 3575 goto out; 3576 /* 3577 * First search, if no extent item that starts at offset off was 3578 * found but the previous item is an extent item, it's possible 3579 * it might overlap our target range, therefore process it. 3580 */ 3581 if (key.offset == off && ret > 0 && path->slots[0] > 0) { 3582 btrfs_item_key_to_cpu(path->nodes[0], &key, 3583 path->slots[0] - 1); 3584 if (key.type == BTRFS_EXTENT_DATA_KEY) 3585 path->slots[0]--; 3586 } 3587 3588 nritems = btrfs_header_nritems(path->nodes[0]); 3589 process_slot: 3590 if (path->slots[0] >= nritems) { 3591 ret = btrfs_next_leaf(BTRFS_I(src)->root, path); 3592 if (ret < 0) 3593 goto out; 3594 if (ret > 0) 3595 break; 3596 nritems = btrfs_header_nritems(path->nodes[0]); 3597 } 3598 leaf = path->nodes[0]; 3599 slot = path->slots[0]; 3600 3601 btrfs_item_key_to_cpu(leaf, &key, slot); 3602 if (key.type > BTRFS_EXTENT_DATA_KEY || 3603 key.objectid != btrfs_ino(src)) 3604 break; 3605 3606 if (key.type == BTRFS_EXTENT_DATA_KEY) { 3607 struct btrfs_file_extent_item *extent; 3608 int type; 3609 u32 size; 3610 struct btrfs_key new_key; 3611 u64 disko = 0, diskl = 0; 3612 u64 datao = 0, datal = 0; 3613 u8 comp; 3614 u64 drop_start; 3615 3616 extent = btrfs_item_ptr(leaf, slot, 3617 struct btrfs_file_extent_item); 3618 comp = btrfs_file_extent_compression(leaf, extent); 3619 type = btrfs_file_extent_type(leaf, extent); 3620 if (type == BTRFS_FILE_EXTENT_REG || 3621 type == BTRFS_FILE_EXTENT_PREALLOC) { 3622 disko = btrfs_file_extent_disk_bytenr(leaf, 3623 extent); 3624 diskl = btrfs_file_extent_disk_num_bytes(leaf, 3625 extent); 3626 datao = btrfs_file_extent_offset(leaf, extent); 3627 datal = btrfs_file_extent_num_bytes(leaf, 3628 extent); 3629 } else if (type == BTRFS_FILE_EXTENT_INLINE) { 3630 /* take upper bound, may be compressed */ 3631 datal = btrfs_file_extent_ram_bytes(leaf, 3632 extent); 3633 } 3634 3635 /* 3636 * The first search might have left us at an extent 3637 * item that ends before our target range's start, can 3638 * happen if we have holes and NO_HOLES feature enabled. 3639 */ 3640 if (key.offset + datal <= off) { 3641 path->slots[0]++; 3642 goto process_slot; 3643 } else if (key.offset >= off + len) { 3644 break; 3645 } 3646 next_key_min_offset = key.offset + datal; 3647 size = btrfs_item_size_nr(leaf, slot); 3648 read_extent_buffer(leaf, buf, 3649 btrfs_item_ptr_offset(leaf, slot), 3650 size); 3651 3652 btrfs_release_path(path); 3653 path->leave_spinning = 0; 3654 3655 memcpy(&new_key, &key, sizeof(new_key)); 3656 new_key.objectid = btrfs_ino(inode); 3657 if (off <= key.offset) 3658 new_key.offset = key.offset + destoff - off; 3659 else 3660 new_key.offset = destoff; 3661 3662 /* 3663 * Deal with a hole that doesn't have an extent item 3664 * that represents it (NO_HOLES feature enabled). 3665 * This hole is either in the middle of the cloning 3666 * range or at the beginning (fully overlaps it or 3667 * partially overlaps it). 3668 */ 3669 if (new_key.offset != last_dest_end) 3670 drop_start = last_dest_end; 3671 else 3672 drop_start = new_key.offset; 3673 3674 /* 3675 * 1 - adjusting old extent (we may have to split it) 3676 * 1 - add new extent 3677 * 1 - inode update 3678 */ 3679 trans = btrfs_start_transaction(root, 3); 3680 if (IS_ERR(trans)) { 3681 ret = PTR_ERR(trans); 3682 goto out; 3683 } 3684 3685 if (type == BTRFS_FILE_EXTENT_REG || 3686 type == BTRFS_FILE_EXTENT_PREALLOC) { 3687 /* 3688 * a | --- range to clone ---| b 3689 * | ------------- extent ------------- | 3690 */ 3691 3692 /* subtract range b */ 3693 if (key.offset + datal > off + len) 3694 datal = off + len - key.offset; 3695 3696 /* subtract range a */ 3697 if (off > key.offset) { 3698 datao += off - key.offset; 3699 datal -= off - key.offset; 3700 } 3701 3702 ret = btrfs_drop_extents(trans, root, inode, 3703 drop_start, 3704 new_key.offset + datal, 3705 1); 3706 if (ret) { 3707 if (ret != -EOPNOTSUPP) 3708 btrfs_abort_transaction(trans, 3709 ret); 3710 btrfs_end_transaction(trans, root); 3711 goto out; 3712 } 3713 3714 ret = btrfs_insert_empty_item(trans, root, path, 3715 &new_key, size); 3716 if (ret) { 3717 btrfs_abort_transaction(trans, ret); 3718 btrfs_end_transaction(trans, root); 3719 goto out; 3720 } 3721 3722 leaf = path->nodes[0]; 3723 slot = path->slots[0]; 3724 write_extent_buffer(leaf, buf, 3725 btrfs_item_ptr_offset(leaf, slot), 3726 size); 3727 3728 extent = btrfs_item_ptr(leaf, slot, 3729 struct btrfs_file_extent_item); 3730 3731 /* disko == 0 means it's a hole */ 3732 if (!disko) 3733 datao = 0; 3734 3735 btrfs_set_file_extent_offset(leaf, extent, 3736 datao); 3737 btrfs_set_file_extent_num_bytes(leaf, extent, 3738 datal); 3739 3740 if (disko) { 3741 inode_add_bytes(inode, datal); 3742 ret = btrfs_inc_extent_ref(trans, root, 3743 disko, diskl, 0, 3744 root->root_key.objectid, 3745 btrfs_ino(inode), 3746 new_key.offset - datao); 3747 if (ret) { 3748 btrfs_abort_transaction(trans, 3749 ret); 3750 btrfs_end_transaction(trans, 3751 root); 3752 goto out; 3753 3754 } 3755 } 3756 } else if (type == BTRFS_FILE_EXTENT_INLINE) { 3757 u64 skip = 0; 3758 u64 trim = 0; 3759 3760 if (off > key.offset) { 3761 skip = off - key.offset; 3762 new_key.offset += skip; 3763 } 3764 3765 if (key.offset + datal > off + len) 3766 trim = key.offset + datal - (off + len); 3767 3768 if (comp && (skip || trim)) { 3769 ret = -EINVAL; 3770 btrfs_end_transaction(trans, root); 3771 goto out; 3772 } 3773 size -= skip + trim; 3774 datal -= skip + trim; 3775 3776 ret = clone_copy_inline_extent(src, inode, 3777 trans, path, 3778 &new_key, 3779 drop_start, 3780 datal, 3781 skip, size, buf); 3782 if (ret) { 3783 if (ret != -EOPNOTSUPP) 3784 btrfs_abort_transaction(trans, 3785 ret); 3786 btrfs_end_transaction(trans, root); 3787 goto out; 3788 } 3789 leaf = path->nodes[0]; 3790 slot = path->slots[0]; 3791 } 3792 3793 /* If we have an implicit hole (NO_HOLES feature). */ 3794 if (drop_start < new_key.offset) 3795 clone_update_extent_map(inode, trans, 3796 NULL, drop_start, 3797 new_key.offset - drop_start); 3798 3799 clone_update_extent_map(inode, trans, path, 0, 0); 3800 3801 btrfs_mark_buffer_dirty(leaf); 3802 btrfs_release_path(path); 3803 3804 last_dest_end = ALIGN(new_key.offset + datal, 3805 root->sectorsize); 3806 ret = clone_finish_inode_update(trans, inode, 3807 last_dest_end, 3808 destoff, olen, 3809 no_time_update); 3810 if (ret) 3811 goto out; 3812 if (new_key.offset + datal >= destoff + len) 3813 break; 3814 } 3815 btrfs_release_path(path); 3816 key.offset = next_key_min_offset; 3817 } 3818 ret = 0; 3819 3820 if (last_dest_end < destoff + len) { 3821 /* 3822 * We have an implicit hole (NO_HOLES feature is enabled) that 3823 * fully or partially overlaps our cloning range at its end. 3824 */ 3825 btrfs_release_path(path); 3826 3827 /* 3828 * 1 - remove extent(s) 3829 * 1 - inode update 3830 */ 3831 trans = btrfs_start_transaction(root, 2); 3832 if (IS_ERR(trans)) { 3833 ret = PTR_ERR(trans); 3834 goto out; 3835 } 3836 ret = btrfs_drop_extents(trans, root, inode, 3837 last_dest_end, destoff + len, 1); 3838 if (ret) { 3839 if (ret != -EOPNOTSUPP) 3840 btrfs_abort_transaction(trans, ret); 3841 btrfs_end_transaction(trans, root); 3842 goto out; 3843 } 3844 clone_update_extent_map(inode, trans, NULL, last_dest_end, 3845 destoff + len - last_dest_end); 3846 ret = clone_finish_inode_update(trans, inode, destoff + len, 3847 destoff, olen, no_time_update); 3848 } 3849 3850 out: 3851 btrfs_free_path(path); 3852 kvfree(buf); 3853 return ret; 3854 } 3855 3856 static noinline int btrfs_clone_files(struct file *file, struct file *file_src, 3857 u64 off, u64 olen, u64 destoff) 3858 { 3859 struct inode *inode = file_inode(file); 3860 struct inode *src = file_inode(file_src); 3861 struct btrfs_root *root = BTRFS_I(inode)->root; 3862 int ret; 3863 u64 len = olen; 3864 u64 bs = root->fs_info->sb->s_blocksize; 3865 int same_inode = src == inode; 3866 3867 /* 3868 * TODO: 3869 * - split compressed inline extents. annoying: we need to 3870 * decompress into destination's address_space (the file offset 3871 * may change, so source mapping won't do), then recompress (or 3872 * otherwise reinsert) a subrange. 3873 * 3874 * - split destination inode's inline extents. The inline extents can 3875 * be either compressed or non-compressed. 3876 */ 3877 3878 if (btrfs_root_readonly(root)) 3879 return -EROFS; 3880 3881 if (file_src->f_path.mnt != file->f_path.mnt || 3882 src->i_sb != inode->i_sb) 3883 return -EXDEV; 3884 3885 /* don't make the dst file partly checksummed */ 3886 if ((BTRFS_I(src)->flags & BTRFS_INODE_NODATASUM) != 3887 (BTRFS_I(inode)->flags & BTRFS_INODE_NODATASUM)) 3888 return -EINVAL; 3889 3890 if (S_ISDIR(src->i_mode) || S_ISDIR(inode->i_mode)) 3891 return -EISDIR; 3892 3893 if (!same_inode) { 3894 btrfs_double_inode_lock(src, inode); 3895 } else { 3896 inode_lock(src); 3897 } 3898 3899 /* determine range to clone */ 3900 ret = -EINVAL; 3901 if (off + len > src->i_size || off + len < off) 3902 goto out_unlock; 3903 if (len == 0) 3904 olen = len = src->i_size - off; 3905 /* if we extend to eof, continue to block boundary */ 3906 if (off + len == src->i_size) 3907 len = ALIGN(src->i_size, bs) - off; 3908 3909 if (len == 0) { 3910 ret = 0; 3911 goto out_unlock; 3912 } 3913 3914 /* verify the end result is block aligned */ 3915 if (!IS_ALIGNED(off, bs) || !IS_ALIGNED(off + len, bs) || 3916 !IS_ALIGNED(destoff, bs)) 3917 goto out_unlock; 3918 3919 /* verify if ranges are overlapped within the same file */ 3920 if (same_inode) { 3921 if (destoff + len > off && destoff < off + len) 3922 goto out_unlock; 3923 } 3924 3925 if (destoff > inode->i_size) { 3926 ret = btrfs_cont_expand(inode, inode->i_size, destoff); 3927 if (ret) 3928 goto out_unlock; 3929 } 3930 3931 /* 3932 * Lock the target range too. Right after we replace the file extent 3933 * items in the fs tree (which now point to the cloned data), we might 3934 * have a worker replace them with extent items relative to a write 3935 * operation that was issued before this clone operation (i.e. confront 3936 * with inode.c:btrfs_finish_ordered_io). 3937 */ 3938 if (same_inode) { 3939 u64 lock_start = min_t(u64, off, destoff); 3940 u64 lock_len = max_t(u64, off, destoff) + len - lock_start; 3941 3942 ret = lock_extent_range(src, lock_start, lock_len, true); 3943 } else { 3944 ret = btrfs_double_extent_lock(src, off, inode, destoff, len, 3945 true); 3946 } 3947 ASSERT(ret == 0); 3948 if (WARN_ON(ret)) { 3949 /* ranges in the io trees already unlocked */ 3950 goto out_unlock; 3951 } 3952 3953 ret = btrfs_clone(src, inode, off, olen, len, destoff, 0); 3954 3955 if (same_inode) { 3956 u64 lock_start = min_t(u64, off, destoff); 3957 u64 lock_end = max_t(u64, off, destoff) + len - 1; 3958 3959 unlock_extent(&BTRFS_I(src)->io_tree, lock_start, lock_end); 3960 } else { 3961 btrfs_double_extent_unlock(src, off, inode, destoff, len); 3962 } 3963 /* 3964 * Truncate page cache pages so that future reads will see the cloned 3965 * data immediately and not the previous data. 3966 */ 3967 truncate_inode_pages_range(&inode->i_data, 3968 round_down(destoff, PAGE_SIZE), 3969 round_up(destoff + len, PAGE_SIZE) - 1); 3970 out_unlock: 3971 if (!same_inode) 3972 btrfs_double_inode_unlock(src, inode); 3973 else 3974 inode_unlock(src); 3975 return ret; 3976 } 3977 3978 ssize_t btrfs_copy_file_range(struct file *file_in, loff_t pos_in, 3979 struct file *file_out, loff_t pos_out, 3980 size_t len, unsigned int flags) 3981 { 3982 ssize_t ret; 3983 3984 ret = btrfs_clone_files(file_out, file_in, pos_in, len, pos_out); 3985 if (ret == 0) 3986 ret = len; 3987 return ret; 3988 } 3989 3990 int btrfs_clone_file_range(struct file *src_file, loff_t off, 3991 struct file *dst_file, loff_t destoff, u64 len) 3992 { 3993 return btrfs_clone_files(dst_file, src_file, off, len, destoff); 3994 } 3995 3996 /* 3997 * there are many ways the trans_start and trans_end ioctls can lead 3998 * to deadlocks. They should only be used by applications that 3999 * basically own the machine, and have a very in depth understanding 4000 * of all the possible deadlocks and enospc problems. 4001 */ 4002 static long btrfs_ioctl_trans_start(struct file *file) 4003 { 4004 struct inode *inode = file_inode(file); 4005 struct btrfs_root *root = BTRFS_I(inode)->root; 4006 struct btrfs_trans_handle *trans; 4007 int ret; 4008 4009 ret = -EPERM; 4010 if (!capable(CAP_SYS_ADMIN)) 4011 goto out; 4012 4013 ret = -EINPROGRESS; 4014 if (file->private_data) 4015 goto out; 4016 4017 ret = -EROFS; 4018 if (btrfs_root_readonly(root)) 4019 goto out; 4020 4021 ret = mnt_want_write_file(file); 4022 if (ret) 4023 goto out; 4024 4025 atomic_inc(&root->fs_info->open_ioctl_trans); 4026 4027 ret = -ENOMEM; 4028 trans = btrfs_start_ioctl_transaction(root); 4029 if (IS_ERR(trans)) 4030 goto out_drop; 4031 4032 file->private_data = trans; 4033 return 0; 4034 4035 out_drop: 4036 atomic_dec(&root->fs_info->open_ioctl_trans); 4037 mnt_drop_write_file(file); 4038 out: 4039 return ret; 4040 } 4041 4042 static long btrfs_ioctl_default_subvol(struct file *file, void __user *argp) 4043 { 4044 struct inode *inode = file_inode(file); 4045 struct btrfs_root *root = BTRFS_I(inode)->root; 4046 struct btrfs_root *new_root; 4047 struct btrfs_dir_item *di; 4048 struct btrfs_trans_handle *trans; 4049 struct btrfs_path *path; 4050 struct btrfs_key location; 4051 struct btrfs_disk_key disk_key; 4052 u64 objectid = 0; 4053 u64 dir_id; 4054 int ret; 4055 4056 if (!capable(CAP_SYS_ADMIN)) 4057 return -EPERM; 4058 4059 ret = mnt_want_write_file(file); 4060 if (ret) 4061 return ret; 4062 4063 if (copy_from_user(&objectid, argp, sizeof(objectid))) { 4064 ret = -EFAULT; 4065 goto out; 4066 } 4067 4068 if (!objectid) 4069 objectid = BTRFS_FS_TREE_OBJECTID; 4070 4071 location.objectid = objectid; 4072 location.type = BTRFS_ROOT_ITEM_KEY; 4073 location.offset = (u64)-1; 4074 4075 new_root = btrfs_read_fs_root_no_name(root->fs_info, &location); 4076 if (IS_ERR(new_root)) { 4077 ret = PTR_ERR(new_root); 4078 goto out; 4079 } 4080 4081 path = btrfs_alloc_path(); 4082 if (!path) { 4083 ret = -ENOMEM; 4084 goto out; 4085 } 4086 path->leave_spinning = 1; 4087 4088 trans = btrfs_start_transaction(root, 1); 4089 if (IS_ERR(trans)) { 4090 btrfs_free_path(path); 4091 ret = PTR_ERR(trans); 4092 goto out; 4093 } 4094 4095 dir_id = btrfs_super_root_dir(root->fs_info->super_copy); 4096 di = btrfs_lookup_dir_item(trans, root->fs_info->tree_root, path, 4097 dir_id, "default", 7, 1); 4098 if (IS_ERR_OR_NULL(di)) { 4099 btrfs_free_path(path); 4100 btrfs_end_transaction(trans, root); 4101 btrfs_err(new_root->fs_info, 4102 "Umm, you don't have the default diritem, this isn't going to work"); 4103 ret = -ENOENT; 4104 goto out; 4105 } 4106 4107 btrfs_cpu_key_to_disk(&disk_key, &new_root->root_key); 4108 btrfs_set_dir_item_key(path->nodes[0], di, &disk_key); 4109 btrfs_mark_buffer_dirty(path->nodes[0]); 4110 btrfs_free_path(path); 4111 4112 btrfs_set_fs_incompat(root->fs_info, DEFAULT_SUBVOL); 4113 btrfs_end_transaction(trans, root); 4114 out: 4115 mnt_drop_write_file(file); 4116 return ret; 4117 } 4118 4119 void btrfs_get_block_group_info(struct list_head *groups_list, 4120 struct btrfs_ioctl_space_info *space) 4121 { 4122 struct btrfs_block_group_cache *block_group; 4123 4124 space->total_bytes = 0; 4125 space->used_bytes = 0; 4126 space->flags = 0; 4127 list_for_each_entry(block_group, groups_list, list) { 4128 space->flags = block_group->flags; 4129 space->total_bytes += block_group->key.offset; 4130 space->used_bytes += 4131 btrfs_block_group_used(&block_group->item); 4132 } 4133 } 4134 4135 static long btrfs_ioctl_space_info(struct btrfs_root *root, void __user *arg) 4136 { 4137 struct btrfs_ioctl_space_args space_args; 4138 struct btrfs_ioctl_space_info space; 4139 struct btrfs_ioctl_space_info *dest; 4140 struct btrfs_ioctl_space_info *dest_orig; 4141 struct btrfs_ioctl_space_info __user *user_dest; 4142 struct btrfs_space_info *info; 4143 u64 types[] = {BTRFS_BLOCK_GROUP_DATA, 4144 BTRFS_BLOCK_GROUP_SYSTEM, 4145 BTRFS_BLOCK_GROUP_METADATA, 4146 BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA}; 4147 int num_types = 4; 4148 int alloc_size; 4149 int ret = 0; 4150 u64 slot_count = 0; 4151 int i, c; 4152 4153 if (copy_from_user(&space_args, 4154 (struct btrfs_ioctl_space_args __user *)arg, 4155 sizeof(space_args))) 4156 return -EFAULT; 4157 4158 for (i = 0; i < num_types; i++) { 4159 struct btrfs_space_info *tmp; 4160 4161 info = NULL; 4162 rcu_read_lock(); 4163 list_for_each_entry_rcu(tmp, &root->fs_info->space_info, 4164 list) { 4165 if (tmp->flags == types[i]) { 4166 info = tmp; 4167 break; 4168 } 4169 } 4170 rcu_read_unlock(); 4171 4172 if (!info) 4173 continue; 4174 4175 down_read(&info->groups_sem); 4176 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) { 4177 if (!list_empty(&info->block_groups[c])) 4178 slot_count++; 4179 } 4180 up_read(&info->groups_sem); 4181 } 4182 4183 /* 4184 * Global block reserve, exported as a space_info 4185 */ 4186 slot_count++; 4187 4188 /* space_slots == 0 means they are asking for a count */ 4189 if (space_args.space_slots == 0) { 4190 space_args.total_spaces = slot_count; 4191 goto out; 4192 } 4193 4194 slot_count = min_t(u64, space_args.space_slots, slot_count); 4195 4196 alloc_size = sizeof(*dest) * slot_count; 4197 4198 /* we generally have at most 6 or so space infos, one for each raid 4199 * level. So, a whole page should be more than enough for everyone 4200 */ 4201 if (alloc_size > PAGE_SIZE) 4202 return -ENOMEM; 4203 4204 space_args.total_spaces = 0; 4205 dest = kmalloc(alloc_size, GFP_KERNEL); 4206 if (!dest) 4207 return -ENOMEM; 4208 dest_orig = dest; 4209 4210 /* now we have a buffer to copy into */ 4211 for (i = 0; i < num_types; i++) { 4212 struct btrfs_space_info *tmp; 4213 4214 if (!slot_count) 4215 break; 4216 4217 info = NULL; 4218 rcu_read_lock(); 4219 list_for_each_entry_rcu(tmp, &root->fs_info->space_info, 4220 list) { 4221 if (tmp->flags == types[i]) { 4222 info = tmp; 4223 break; 4224 } 4225 } 4226 rcu_read_unlock(); 4227 4228 if (!info) 4229 continue; 4230 down_read(&info->groups_sem); 4231 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) { 4232 if (!list_empty(&info->block_groups[c])) { 4233 btrfs_get_block_group_info( 4234 &info->block_groups[c], &space); 4235 memcpy(dest, &space, sizeof(space)); 4236 dest++; 4237 space_args.total_spaces++; 4238 slot_count--; 4239 } 4240 if (!slot_count) 4241 break; 4242 } 4243 up_read(&info->groups_sem); 4244 } 4245 4246 /* 4247 * Add global block reserve 4248 */ 4249 if (slot_count) { 4250 struct btrfs_block_rsv *block_rsv = &root->fs_info->global_block_rsv; 4251 4252 spin_lock(&block_rsv->lock); 4253 space.total_bytes = block_rsv->size; 4254 space.used_bytes = block_rsv->size - block_rsv->reserved; 4255 spin_unlock(&block_rsv->lock); 4256 space.flags = BTRFS_SPACE_INFO_GLOBAL_RSV; 4257 memcpy(dest, &space, sizeof(space)); 4258 space_args.total_spaces++; 4259 } 4260 4261 user_dest = (struct btrfs_ioctl_space_info __user *) 4262 (arg + sizeof(struct btrfs_ioctl_space_args)); 4263 4264 if (copy_to_user(user_dest, dest_orig, alloc_size)) 4265 ret = -EFAULT; 4266 4267 kfree(dest_orig); 4268 out: 4269 if (ret == 0 && copy_to_user(arg, &space_args, sizeof(space_args))) 4270 ret = -EFAULT; 4271 4272 return ret; 4273 } 4274 4275 /* 4276 * there are many ways the trans_start and trans_end ioctls can lead 4277 * to deadlocks. They should only be used by applications that 4278 * basically own the machine, and have a very in depth understanding 4279 * of all the possible deadlocks and enospc problems. 4280 */ 4281 long btrfs_ioctl_trans_end(struct file *file) 4282 { 4283 struct inode *inode = file_inode(file); 4284 struct btrfs_root *root = BTRFS_I(inode)->root; 4285 struct btrfs_trans_handle *trans; 4286 4287 trans = file->private_data; 4288 if (!trans) 4289 return -EINVAL; 4290 file->private_data = NULL; 4291 4292 btrfs_end_transaction(trans, root); 4293 4294 atomic_dec(&root->fs_info->open_ioctl_trans); 4295 4296 mnt_drop_write_file(file); 4297 return 0; 4298 } 4299 4300 static noinline long btrfs_ioctl_start_sync(struct btrfs_root *root, 4301 void __user *argp) 4302 { 4303 struct btrfs_trans_handle *trans; 4304 u64 transid; 4305 int ret; 4306 4307 trans = btrfs_attach_transaction_barrier(root); 4308 if (IS_ERR(trans)) { 4309 if (PTR_ERR(trans) != -ENOENT) 4310 return PTR_ERR(trans); 4311 4312 /* No running transaction, don't bother */ 4313 transid = root->fs_info->last_trans_committed; 4314 goto out; 4315 } 4316 transid = trans->transid; 4317 ret = btrfs_commit_transaction_async(trans, root, 0); 4318 if (ret) { 4319 btrfs_end_transaction(trans, root); 4320 return ret; 4321 } 4322 out: 4323 if (argp) 4324 if (copy_to_user(argp, &transid, sizeof(transid))) 4325 return -EFAULT; 4326 return 0; 4327 } 4328 4329 static noinline long btrfs_ioctl_wait_sync(struct btrfs_root *root, 4330 void __user *argp) 4331 { 4332 u64 transid; 4333 4334 if (argp) { 4335 if (copy_from_user(&transid, argp, sizeof(transid))) 4336 return -EFAULT; 4337 } else { 4338 transid = 0; /* current trans */ 4339 } 4340 return btrfs_wait_for_commit(root, transid); 4341 } 4342 4343 static long btrfs_ioctl_scrub(struct file *file, void __user *arg) 4344 { 4345 struct btrfs_root *root = BTRFS_I(file_inode(file))->root; 4346 struct btrfs_ioctl_scrub_args *sa; 4347 int ret; 4348 4349 if (!capable(CAP_SYS_ADMIN)) 4350 return -EPERM; 4351 4352 sa = memdup_user(arg, sizeof(*sa)); 4353 if (IS_ERR(sa)) 4354 return PTR_ERR(sa); 4355 4356 if (!(sa->flags & BTRFS_SCRUB_READONLY)) { 4357 ret = mnt_want_write_file(file); 4358 if (ret) 4359 goto out; 4360 } 4361 4362 ret = btrfs_scrub_dev(root->fs_info, sa->devid, sa->start, sa->end, 4363 &sa->progress, sa->flags & BTRFS_SCRUB_READONLY, 4364 0); 4365 4366 if (copy_to_user(arg, sa, sizeof(*sa))) 4367 ret = -EFAULT; 4368 4369 if (!(sa->flags & BTRFS_SCRUB_READONLY)) 4370 mnt_drop_write_file(file); 4371 out: 4372 kfree(sa); 4373 return ret; 4374 } 4375 4376 static long btrfs_ioctl_scrub_cancel(struct btrfs_root *root, void __user *arg) 4377 { 4378 if (!capable(CAP_SYS_ADMIN)) 4379 return -EPERM; 4380 4381 return btrfs_scrub_cancel(root->fs_info); 4382 } 4383 4384 static long btrfs_ioctl_scrub_progress(struct btrfs_root *root, 4385 void __user *arg) 4386 { 4387 struct btrfs_ioctl_scrub_args *sa; 4388 int ret; 4389 4390 if (!capable(CAP_SYS_ADMIN)) 4391 return -EPERM; 4392 4393 sa = memdup_user(arg, sizeof(*sa)); 4394 if (IS_ERR(sa)) 4395 return PTR_ERR(sa); 4396 4397 ret = btrfs_scrub_progress(root, sa->devid, &sa->progress); 4398 4399 if (copy_to_user(arg, sa, sizeof(*sa))) 4400 ret = -EFAULT; 4401 4402 kfree(sa); 4403 return ret; 4404 } 4405 4406 static long btrfs_ioctl_get_dev_stats(struct btrfs_root *root, 4407 void __user *arg) 4408 { 4409 struct btrfs_ioctl_get_dev_stats *sa; 4410 int ret; 4411 4412 sa = memdup_user(arg, sizeof(*sa)); 4413 if (IS_ERR(sa)) 4414 return PTR_ERR(sa); 4415 4416 if ((sa->flags & BTRFS_DEV_STATS_RESET) && !capable(CAP_SYS_ADMIN)) { 4417 kfree(sa); 4418 return -EPERM; 4419 } 4420 4421 ret = btrfs_get_dev_stats(root, sa); 4422 4423 if (copy_to_user(arg, sa, sizeof(*sa))) 4424 ret = -EFAULT; 4425 4426 kfree(sa); 4427 return ret; 4428 } 4429 4430 static long btrfs_ioctl_dev_replace(struct btrfs_root *root, void __user *arg) 4431 { 4432 struct btrfs_ioctl_dev_replace_args *p; 4433 int ret; 4434 4435 if (!capable(CAP_SYS_ADMIN)) 4436 return -EPERM; 4437 4438 p = memdup_user(arg, sizeof(*p)); 4439 if (IS_ERR(p)) 4440 return PTR_ERR(p); 4441 4442 switch (p->cmd) { 4443 case BTRFS_IOCTL_DEV_REPLACE_CMD_START: 4444 if (root->fs_info->sb->s_flags & MS_RDONLY) { 4445 ret = -EROFS; 4446 goto out; 4447 } 4448 if (atomic_xchg( 4449 &root->fs_info->mutually_exclusive_operation_running, 4450 1)) { 4451 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS; 4452 } else { 4453 ret = btrfs_dev_replace_by_ioctl(root, p); 4454 atomic_set( 4455 &root->fs_info->mutually_exclusive_operation_running, 4456 0); 4457 } 4458 break; 4459 case BTRFS_IOCTL_DEV_REPLACE_CMD_STATUS: 4460 btrfs_dev_replace_status(root->fs_info, p); 4461 ret = 0; 4462 break; 4463 case BTRFS_IOCTL_DEV_REPLACE_CMD_CANCEL: 4464 ret = btrfs_dev_replace_cancel(root->fs_info, p); 4465 break; 4466 default: 4467 ret = -EINVAL; 4468 break; 4469 } 4470 4471 if (copy_to_user(arg, p, sizeof(*p))) 4472 ret = -EFAULT; 4473 out: 4474 kfree(p); 4475 return ret; 4476 } 4477 4478 static long btrfs_ioctl_ino_to_path(struct btrfs_root *root, void __user *arg) 4479 { 4480 int ret = 0; 4481 int i; 4482 u64 rel_ptr; 4483 int size; 4484 struct btrfs_ioctl_ino_path_args *ipa = NULL; 4485 struct inode_fs_paths *ipath = NULL; 4486 struct btrfs_path *path; 4487 4488 if (!capable(CAP_DAC_READ_SEARCH)) 4489 return -EPERM; 4490 4491 path = btrfs_alloc_path(); 4492 if (!path) { 4493 ret = -ENOMEM; 4494 goto out; 4495 } 4496 4497 ipa = memdup_user(arg, sizeof(*ipa)); 4498 if (IS_ERR(ipa)) { 4499 ret = PTR_ERR(ipa); 4500 ipa = NULL; 4501 goto out; 4502 } 4503 4504 size = min_t(u32, ipa->size, 4096); 4505 ipath = init_ipath(size, root, path); 4506 if (IS_ERR(ipath)) { 4507 ret = PTR_ERR(ipath); 4508 ipath = NULL; 4509 goto out; 4510 } 4511 4512 ret = paths_from_inode(ipa->inum, ipath); 4513 if (ret < 0) 4514 goto out; 4515 4516 for (i = 0; i < ipath->fspath->elem_cnt; ++i) { 4517 rel_ptr = ipath->fspath->val[i] - 4518 (u64)(unsigned long)ipath->fspath->val; 4519 ipath->fspath->val[i] = rel_ptr; 4520 } 4521 4522 ret = copy_to_user((void *)(unsigned long)ipa->fspath, 4523 (void *)(unsigned long)ipath->fspath, size); 4524 if (ret) { 4525 ret = -EFAULT; 4526 goto out; 4527 } 4528 4529 out: 4530 btrfs_free_path(path); 4531 free_ipath(ipath); 4532 kfree(ipa); 4533 4534 return ret; 4535 } 4536 4537 static int build_ino_list(u64 inum, u64 offset, u64 root, void *ctx) 4538 { 4539 struct btrfs_data_container *inodes = ctx; 4540 const size_t c = 3 * sizeof(u64); 4541 4542 if (inodes->bytes_left >= c) { 4543 inodes->bytes_left -= c; 4544 inodes->val[inodes->elem_cnt] = inum; 4545 inodes->val[inodes->elem_cnt + 1] = offset; 4546 inodes->val[inodes->elem_cnt + 2] = root; 4547 inodes->elem_cnt += 3; 4548 } else { 4549 inodes->bytes_missing += c - inodes->bytes_left; 4550 inodes->bytes_left = 0; 4551 inodes->elem_missed += 3; 4552 } 4553 4554 return 0; 4555 } 4556 4557 static long btrfs_ioctl_logical_to_ino(struct btrfs_root *root, 4558 void __user *arg) 4559 { 4560 int ret = 0; 4561 int size; 4562 struct btrfs_ioctl_logical_ino_args *loi; 4563 struct btrfs_data_container *inodes = NULL; 4564 struct btrfs_path *path = NULL; 4565 4566 if (!capable(CAP_SYS_ADMIN)) 4567 return -EPERM; 4568 4569 loi = memdup_user(arg, sizeof(*loi)); 4570 if (IS_ERR(loi)) { 4571 ret = PTR_ERR(loi); 4572 loi = NULL; 4573 goto out; 4574 } 4575 4576 path = btrfs_alloc_path(); 4577 if (!path) { 4578 ret = -ENOMEM; 4579 goto out; 4580 } 4581 4582 size = min_t(u32, loi->size, SZ_64K); 4583 inodes = init_data_container(size); 4584 if (IS_ERR(inodes)) { 4585 ret = PTR_ERR(inodes); 4586 inodes = NULL; 4587 goto out; 4588 } 4589 4590 ret = iterate_inodes_from_logical(loi->logical, root->fs_info, path, 4591 build_ino_list, inodes); 4592 if (ret == -EINVAL) 4593 ret = -ENOENT; 4594 if (ret < 0) 4595 goto out; 4596 4597 ret = copy_to_user((void *)(unsigned long)loi->inodes, 4598 (void *)(unsigned long)inodes, size); 4599 if (ret) 4600 ret = -EFAULT; 4601 4602 out: 4603 btrfs_free_path(path); 4604 vfree(inodes); 4605 kfree(loi); 4606 4607 return ret; 4608 } 4609 4610 void update_ioctl_balance_args(struct btrfs_fs_info *fs_info, int lock, 4611 struct btrfs_ioctl_balance_args *bargs) 4612 { 4613 struct btrfs_balance_control *bctl = fs_info->balance_ctl; 4614 4615 bargs->flags = bctl->flags; 4616 4617 if (atomic_read(&fs_info->balance_running)) 4618 bargs->state |= BTRFS_BALANCE_STATE_RUNNING; 4619 if (atomic_read(&fs_info->balance_pause_req)) 4620 bargs->state |= BTRFS_BALANCE_STATE_PAUSE_REQ; 4621 if (atomic_read(&fs_info->balance_cancel_req)) 4622 bargs->state |= BTRFS_BALANCE_STATE_CANCEL_REQ; 4623 4624 memcpy(&bargs->data, &bctl->data, sizeof(bargs->data)); 4625 memcpy(&bargs->meta, &bctl->meta, sizeof(bargs->meta)); 4626 memcpy(&bargs->sys, &bctl->sys, sizeof(bargs->sys)); 4627 4628 if (lock) { 4629 spin_lock(&fs_info->balance_lock); 4630 memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat)); 4631 spin_unlock(&fs_info->balance_lock); 4632 } else { 4633 memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat)); 4634 } 4635 } 4636 4637 static long btrfs_ioctl_balance(struct file *file, void __user *arg) 4638 { 4639 struct btrfs_root *root = BTRFS_I(file_inode(file))->root; 4640 struct btrfs_fs_info *fs_info = root->fs_info; 4641 struct btrfs_ioctl_balance_args *bargs; 4642 struct btrfs_balance_control *bctl; 4643 bool need_unlock; /* for mut. excl. ops lock */ 4644 int ret; 4645 4646 if (!capable(CAP_SYS_ADMIN)) 4647 return -EPERM; 4648 4649 ret = mnt_want_write_file(file); 4650 if (ret) 4651 return ret; 4652 4653 again: 4654 if (!atomic_xchg(&fs_info->mutually_exclusive_operation_running, 1)) { 4655 mutex_lock(&fs_info->volume_mutex); 4656 mutex_lock(&fs_info->balance_mutex); 4657 need_unlock = true; 4658 goto locked; 4659 } 4660 4661 /* 4662 * mut. excl. ops lock is locked. Three possibilities: 4663 * (1) some other op is running 4664 * (2) balance is running 4665 * (3) balance is paused -- special case (think resume) 4666 */ 4667 mutex_lock(&fs_info->balance_mutex); 4668 if (fs_info->balance_ctl) { 4669 /* this is either (2) or (3) */ 4670 if (!atomic_read(&fs_info->balance_running)) { 4671 mutex_unlock(&fs_info->balance_mutex); 4672 if (!mutex_trylock(&fs_info->volume_mutex)) 4673 goto again; 4674 mutex_lock(&fs_info->balance_mutex); 4675 4676 if (fs_info->balance_ctl && 4677 !atomic_read(&fs_info->balance_running)) { 4678 /* this is (3) */ 4679 need_unlock = false; 4680 goto locked; 4681 } 4682 4683 mutex_unlock(&fs_info->balance_mutex); 4684 mutex_unlock(&fs_info->volume_mutex); 4685 goto again; 4686 } else { 4687 /* this is (2) */ 4688 mutex_unlock(&fs_info->balance_mutex); 4689 ret = -EINPROGRESS; 4690 goto out; 4691 } 4692 } else { 4693 /* this is (1) */ 4694 mutex_unlock(&fs_info->balance_mutex); 4695 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS; 4696 goto out; 4697 } 4698 4699 locked: 4700 BUG_ON(!atomic_read(&fs_info->mutually_exclusive_operation_running)); 4701 4702 if (arg) { 4703 bargs = memdup_user(arg, sizeof(*bargs)); 4704 if (IS_ERR(bargs)) { 4705 ret = PTR_ERR(bargs); 4706 goto out_unlock; 4707 } 4708 4709 if (bargs->flags & BTRFS_BALANCE_RESUME) { 4710 if (!fs_info->balance_ctl) { 4711 ret = -ENOTCONN; 4712 goto out_bargs; 4713 } 4714 4715 bctl = fs_info->balance_ctl; 4716 spin_lock(&fs_info->balance_lock); 4717 bctl->flags |= BTRFS_BALANCE_RESUME; 4718 spin_unlock(&fs_info->balance_lock); 4719 4720 goto do_balance; 4721 } 4722 } else { 4723 bargs = NULL; 4724 } 4725 4726 if (fs_info->balance_ctl) { 4727 ret = -EINPROGRESS; 4728 goto out_bargs; 4729 } 4730 4731 bctl = kzalloc(sizeof(*bctl), GFP_KERNEL); 4732 if (!bctl) { 4733 ret = -ENOMEM; 4734 goto out_bargs; 4735 } 4736 4737 bctl->fs_info = fs_info; 4738 if (arg) { 4739 memcpy(&bctl->data, &bargs->data, sizeof(bctl->data)); 4740 memcpy(&bctl->meta, &bargs->meta, sizeof(bctl->meta)); 4741 memcpy(&bctl->sys, &bargs->sys, sizeof(bctl->sys)); 4742 4743 bctl->flags = bargs->flags; 4744 } else { 4745 /* balance everything - no filters */ 4746 bctl->flags |= BTRFS_BALANCE_TYPE_MASK; 4747 } 4748 4749 if (bctl->flags & ~(BTRFS_BALANCE_ARGS_MASK | BTRFS_BALANCE_TYPE_MASK)) { 4750 ret = -EINVAL; 4751 goto out_bctl; 4752 } 4753 4754 do_balance: 4755 /* 4756 * Ownership of bctl and mutually_exclusive_operation_running 4757 * goes to to btrfs_balance. bctl is freed in __cancel_balance, 4758 * or, if restriper was paused all the way until unmount, in 4759 * free_fs_info. mutually_exclusive_operation_running is 4760 * cleared in __cancel_balance. 4761 */ 4762 need_unlock = false; 4763 4764 ret = btrfs_balance(bctl, bargs); 4765 bctl = NULL; 4766 4767 if (arg) { 4768 if (copy_to_user(arg, bargs, sizeof(*bargs))) 4769 ret = -EFAULT; 4770 } 4771 4772 out_bctl: 4773 kfree(bctl); 4774 out_bargs: 4775 kfree(bargs); 4776 out_unlock: 4777 mutex_unlock(&fs_info->balance_mutex); 4778 mutex_unlock(&fs_info->volume_mutex); 4779 if (need_unlock) 4780 atomic_set(&fs_info->mutually_exclusive_operation_running, 0); 4781 out: 4782 mnt_drop_write_file(file); 4783 return ret; 4784 } 4785 4786 static long btrfs_ioctl_balance_ctl(struct btrfs_root *root, int cmd) 4787 { 4788 if (!capable(CAP_SYS_ADMIN)) 4789 return -EPERM; 4790 4791 switch (cmd) { 4792 case BTRFS_BALANCE_CTL_PAUSE: 4793 return btrfs_pause_balance(root->fs_info); 4794 case BTRFS_BALANCE_CTL_CANCEL: 4795 return btrfs_cancel_balance(root->fs_info); 4796 } 4797 4798 return -EINVAL; 4799 } 4800 4801 static long btrfs_ioctl_balance_progress(struct btrfs_root *root, 4802 void __user *arg) 4803 { 4804 struct btrfs_fs_info *fs_info = root->fs_info; 4805 struct btrfs_ioctl_balance_args *bargs; 4806 int ret = 0; 4807 4808 if (!capable(CAP_SYS_ADMIN)) 4809 return -EPERM; 4810 4811 mutex_lock(&fs_info->balance_mutex); 4812 if (!fs_info->balance_ctl) { 4813 ret = -ENOTCONN; 4814 goto out; 4815 } 4816 4817 bargs = kzalloc(sizeof(*bargs), GFP_KERNEL); 4818 if (!bargs) { 4819 ret = -ENOMEM; 4820 goto out; 4821 } 4822 4823 update_ioctl_balance_args(fs_info, 1, bargs); 4824 4825 if (copy_to_user(arg, bargs, sizeof(*bargs))) 4826 ret = -EFAULT; 4827 4828 kfree(bargs); 4829 out: 4830 mutex_unlock(&fs_info->balance_mutex); 4831 return ret; 4832 } 4833 4834 static long btrfs_ioctl_quota_ctl(struct file *file, void __user *arg) 4835 { 4836 struct btrfs_root *root = BTRFS_I(file_inode(file))->root; 4837 struct btrfs_ioctl_quota_ctl_args *sa; 4838 struct btrfs_trans_handle *trans = NULL; 4839 int ret; 4840 int err; 4841 4842 if (!capable(CAP_SYS_ADMIN)) 4843 return -EPERM; 4844 4845 ret = mnt_want_write_file(file); 4846 if (ret) 4847 return ret; 4848 4849 sa = memdup_user(arg, sizeof(*sa)); 4850 if (IS_ERR(sa)) { 4851 ret = PTR_ERR(sa); 4852 goto drop_write; 4853 } 4854 4855 down_write(&root->fs_info->subvol_sem); 4856 trans = btrfs_start_transaction(root->fs_info->tree_root, 2); 4857 if (IS_ERR(trans)) { 4858 ret = PTR_ERR(trans); 4859 goto out; 4860 } 4861 4862 switch (sa->cmd) { 4863 case BTRFS_QUOTA_CTL_ENABLE: 4864 ret = btrfs_quota_enable(trans, root->fs_info); 4865 break; 4866 case BTRFS_QUOTA_CTL_DISABLE: 4867 ret = btrfs_quota_disable(trans, root->fs_info); 4868 break; 4869 default: 4870 ret = -EINVAL; 4871 break; 4872 } 4873 4874 err = btrfs_commit_transaction(trans, root->fs_info->tree_root); 4875 if (err && !ret) 4876 ret = err; 4877 out: 4878 kfree(sa); 4879 up_write(&root->fs_info->subvol_sem); 4880 drop_write: 4881 mnt_drop_write_file(file); 4882 return ret; 4883 } 4884 4885 static long btrfs_ioctl_qgroup_assign(struct file *file, void __user *arg) 4886 { 4887 struct btrfs_root *root = BTRFS_I(file_inode(file))->root; 4888 struct btrfs_ioctl_qgroup_assign_args *sa; 4889 struct btrfs_trans_handle *trans; 4890 int ret; 4891 int err; 4892 4893 if (!capable(CAP_SYS_ADMIN)) 4894 return -EPERM; 4895 4896 ret = mnt_want_write_file(file); 4897 if (ret) 4898 return ret; 4899 4900 sa = memdup_user(arg, sizeof(*sa)); 4901 if (IS_ERR(sa)) { 4902 ret = PTR_ERR(sa); 4903 goto drop_write; 4904 } 4905 4906 trans = btrfs_join_transaction(root); 4907 if (IS_ERR(trans)) { 4908 ret = PTR_ERR(trans); 4909 goto out; 4910 } 4911 4912 /* FIXME: check if the IDs really exist */ 4913 if (sa->assign) { 4914 ret = btrfs_add_qgroup_relation(trans, root->fs_info, 4915 sa->src, sa->dst); 4916 } else { 4917 ret = btrfs_del_qgroup_relation(trans, root->fs_info, 4918 sa->src, sa->dst); 4919 } 4920 4921 /* update qgroup status and info */ 4922 err = btrfs_run_qgroups(trans, root->fs_info); 4923 if (err < 0) 4924 btrfs_handle_fs_error(root->fs_info, err, 4925 "failed to update qgroup status and info"); 4926 err = btrfs_end_transaction(trans, root); 4927 if (err && !ret) 4928 ret = err; 4929 4930 out: 4931 kfree(sa); 4932 drop_write: 4933 mnt_drop_write_file(file); 4934 return ret; 4935 } 4936 4937 static long btrfs_ioctl_qgroup_create(struct file *file, void __user *arg) 4938 { 4939 struct btrfs_root *root = BTRFS_I(file_inode(file))->root; 4940 struct btrfs_ioctl_qgroup_create_args *sa; 4941 struct btrfs_trans_handle *trans; 4942 int ret; 4943 int err; 4944 4945 if (!capable(CAP_SYS_ADMIN)) 4946 return -EPERM; 4947 4948 ret = mnt_want_write_file(file); 4949 if (ret) 4950 return ret; 4951 4952 sa = memdup_user(arg, sizeof(*sa)); 4953 if (IS_ERR(sa)) { 4954 ret = PTR_ERR(sa); 4955 goto drop_write; 4956 } 4957 4958 if (!sa->qgroupid) { 4959 ret = -EINVAL; 4960 goto out; 4961 } 4962 4963 trans = btrfs_join_transaction(root); 4964 if (IS_ERR(trans)) { 4965 ret = PTR_ERR(trans); 4966 goto out; 4967 } 4968 4969 /* FIXME: check if the IDs really exist */ 4970 if (sa->create) { 4971 ret = btrfs_create_qgroup(trans, root->fs_info, sa->qgroupid); 4972 } else { 4973 ret = btrfs_remove_qgroup(trans, root->fs_info, sa->qgroupid); 4974 } 4975 4976 err = btrfs_end_transaction(trans, root); 4977 if (err && !ret) 4978 ret = err; 4979 4980 out: 4981 kfree(sa); 4982 drop_write: 4983 mnt_drop_write_file(file); 4984 return ret; 4985 } 4986 4987 static long btrfs_ioctl_qgroup_limit(struct file *file, void __user *arg) 4988 { 4989 struct btrfs_root *root = BTRFS_I(file_inode(file))->root; 4990 struct btrfs_ioctl_qgroup_limit_args *sa; 4991 struct btrfs_trans_handle *trans; 4992 int ret; 4993 int err; 4994 u64 qgroupid; 4995 4996 if (!capable(CAP_SYS_ADMIN)) 4997 return -EPERM; 4998 4999 ret = mnt_want_write_file(file); 5000 if (ret) 5001 return ret; 5002 5003 sa = memdup_user(arg, sizeof(*sa)); 5004 if (IS_ERR(sa)) { 5005 ret = PTR_ERR(sa); 5006 goto drop_write; 5007 } 5008 5009 trans = btrfs_join_transaction(root); 5010 if (IS_ERR(trans)) { 5011 ret = PTR_ERR(trans); 5012 goto out; 5013 } 5014 5015 qgroupid = sa->qgroupid; 5016 if (!qgroupid) { 5017 /* take the current subvol as qgroup */ 5018 qgroupid = root->root_key.objectid; 5019 } 5020 5021 /* FIXME: check if the IDs really exist */ 5022 ret = btrfs_limit_qgroup(trans, root->fs_info, qgroupid, &sa->lim); 5023 5024 err = btrfs_end_transaction(trans, root); 5025 if (err && !ret) 5026 ret = err; 5027 5028 out: 5029 kfree(sa); 5030 drop_write: 5031 mnt_drop_write_file(file); 5032 return ret; 5033 } 5034 5035 static long btrfs_ioctl_quota_rescan(struct file *file, void __user *arg) 5036 { 5037 struct btrfs_root *root = BTRFS_I(file_inode(file))->root; 5038 struct btrfs_ioctl_quota_rescan_args *qsa; 5039 int ret; 5040 5041 if (!capable(CAP_SYS_ADMIN)) 5042 return -EPERM; 5043 5044 ret = mnt_want_write_file(file); 5045 if (ret) 5046 return ret; 5047 5048 qsa = memdup_user(arg, sizeof(*qsa)); 5049 if (IS_ERR(qsa)) { 5050 ret = PTR_ERR(qsa); 5051 goto drop_write; 5052 } 5053 5054 if (qsa->flags) { 5055 ret = -EINVAL; 5056 goto out; 5057 } 5058 5059 ret = btrfs_qgroup_rescan(root->fs_info); 5060 5061 out: 5062 kfree(qsa); 5063 drop_write: 5064 mnt_drop_write_file(file); 5065 return ret; 5066 } 5067 5068 static long btrfs_ioctl_quota_rescan_status(struct file *file, void __user *arg) 5069 { 5070 struct btrfs_root *root = BTRFS_I(file_inode(file))->root; 5071 struct btrfs_ioctl_quota_rescan_args *qsa; 5072 int ret = 0; 5073 5074 if (!capable(CAP_SYS_ADMIN)) 5075 return -EPERM; 5076 5077 qsa = kzalloc(sizeof(*qsa), GFP_KERNEL); 5078 if (!qsa) 5079 return -ENOMEM; 5080 5081 if (root->fs_info->qgroup_flags & BTRFS_QGROUP_STATUS_FLAG_RESCAN) { 5082 qsa->flags = 1; 5083 qsa->progress = root->fs_info->qgroup_rescan_progress.objectid; 5084 } 5085 5086 if (copy_to_user(arg, qsa, sizeof(*qsa))) 5087 ret = -EFAULT; 5088 5089 kfree(qsa); 5090 return ret; 5091 } 5092 5093 static long btrfs_ioctl_quota_rescan_wait(struct file *file, void __user *arg) 5094 { 5095 struct btrfs_root *root = BTRFS_I(file_inode(file))->root; 5096 5097 if (!capable(CAP_SYS_ADMIN)) 5098 return -EPERM; 5099 5100 return btrfs_qgroup_wait_for_completion(root->fs_info, true); 5101 } 5102 5103 static long _btrfs_ioctl_set_received_subvol(struct file *file, 5104 struct btrfs_ioctl_received_subvol_args *sa) 5105 { 5106 struct inode *inode = file_inode(file); 5107 struct btrfs_root *root = BTRFS_I(inode)->root; 5108 struct btrfs_root_item *root_item = &root->root_item; 5109 struct btrfs_trans_handle *trans; 5110 struct timespec ct = current_time(inode); 5111 int ret = 0; 5112 int received_uuid_changed; 5113 5114 if (!inode_owner_or_capable(inode)) 5115 return -EPERM; 5116 5117 ret = mnt_want_write_file(file); 5118 if (ret < 0) 5119 return ret; 5120 5121 down_write(&root->fs_info->subvol_sem); 5122 5123 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) { 5124 ret = -EINVAL; 5125 goto out; 5126 } 5127 5128 if (btrfs_root_readonly(root)) { 5129 ret = -EROFS; 5130 goto out; 5131 } 5132 5133 /* 5134 * 1 - root item 5135 * 2 - uuid items (received uuid + subvol uuid) 5136 */ 5137 trans = btrfs_start_transaction(root, 3); 5138 if (IS_ERR(trans)) { 5139 ret = PTR_ERR(trans); 5140 trans = NULL; 5141 goto out; 5142 } 5143 5144 sa->rtransid = trans->transid; 5145 sa->rtime.sec = ct.tv_sec; 5146 sa->rtime.nsec = ct.tv_nsec; 5147 5148 received_uuid_changed = memcmp(root_item->received_uuid, sa->uuid, 5149 BTRFS_UUID_SIZE); 5150 if (received_uuid_changed && 5151 !btrfs_is_empty_uuid(root_item->received_uuid)) 5152 btrfs_uuid_tree_rem(trans, root->fs_info->uuid_root, 5153 root_item->received_uuid, 5154 BTRFS_UUID_KEY_RECEIVED_SUBVOL, 5155 root->root_key.objectid); 5156 memcpy(root_item->received_uuid, sa->uuid, BTRFS_UUID_SIZE); 5157 btrfs_set_root_stransid(root_item, sa->stransid); 5158 btrfs_set_root_rtransid(root_item, sa->rtransid); 5159 btrfs_set_stack_timespec_sec(&root_item->stime, sa->stime.sec); 5160 btrfs_set_stack_timespec_nsec(&root_item->stime, sa->stime.nsec); 5161 btrfs_set_stack_timespec_sec(&root_item->rtime, sa->rtime.sec); 5162 btrfs_set_stack_timespec_nsec(&root_item->rtime, sa->rtime.nsec); 5163 5164 ret = btrfs_update_root(trans, root->fs_info->tree_root, 5165 &root->root_key, &root->root_item); 5166 if (ret < 0) { 5167 btrfs_end_transaction(trans, root); 5168 goto out; 5169 } 5170 if (received_uuid_changed && !btrfs_is_empty_uuid(sa->uuid)) { 5171 ret = btrfs_uuid_tree_add(trans, root->fs_info->uuid_root, 5172 sa->uuid, 5173 BTRFS_UUID_KEY_RECEIVED_SUBVOL, 5174 root->root_key.objectid); 5175 if (ret < 0 && ret != -EEXIST) { 5176 btrfs_abort_transaction(trans, ret); 5177 goto out; 5178 } 5179 } 5180 ret = btrfs_commit_transaction(trans, root); 5181 if (ret < 0) { 5182 btrfs_abort_transaction(trans, ret); 5183 goto out; 5184 } 5185 5186 out: 5187 up_write(&root->fs_info->subvol_sem); 5188 mnt_drop_write_file(file); 5189 return ret; 5190 } 5191 5192 #ifdef CONFIG_64BIT 5193 static long btrfs_ioctl_set_received_subvol_32(struct file *file, 5194 void __user *arg) 5195 { 5196 struct btrfs_ioctl_received_subvol_args_32 *args32 = NULL; 5197 struct btrfs_ioctl_received_subvol_args *args64 = NULL; 5198 int ret = 0; 5199 5200 args32 = memdup_user(arg, sizeof(*args32)); 5201 if (IS_ERR(args32)) { 5202 ret = PTR_ERR(args32); 5203 args32 = NULL; 5204 goto out; 5205 } 5206 5207 args64 = kmalloc(sizeof(*args64), GFP_KERNEL); 5208 if (!args64) { 5209 ret = -ENOMEM; 5210 goto out; 5211 } 5212 5213 memcpy(args64->uuid, args32->uuid, BTRFS_UUID_SIZE); 5214 args64->stransid = args32->stransid; 5215 args64->rtransid = args32->rtransid; 5216 args64->stime.sec = args32->stime.sec; 5217 args64->stime.nsec = args32->stime.nsec; 5218 args64->rtime.sec = args32->rtime.sec; 5219 args64->rtime.nsec = args32->rtime.nsec; 5220 args64->flags = args32->flags; 5221 5222 ret = _btrfs_ioctl_set_received_subvol(file, args64); 5223 if (ret) 5224 goto out; 5225 5226 memcpy(args32->uuid, args64->uuid, BTRFS_UUID_SIZE); 5227 args32->stransid = args64->stransid; 5228 args32->rtransid = args64->rtransid; 5229 args32->stime.sec = args64->stime.sec; 5230 args32->stime.nsec = args64->stime.nsec; 5231 args32->rtime.sec = args64->rtime.sec; 5232 args32->rtime.nsec = args64->rtime.nsec; 5233 args32->flags = args64->flags; 5234 5235 ret = copy_to_user(arg, args32, sizeof(*args32)); 5236 if (ret) 5237 ret = -EFAULT; 5238 5239 out: 5240 kfree(args32); 5241 kfree(args64); 5242 return ret; 5243 } 5244 #endif 5245 5246 static long btrfs_ioctl_set_received_subvol(struct file *file, 5247 void __user *arg) 5248 { 5249 struct btrfs_ioctl_received_subvol_args *sa = NULL; 5250 int ret = 0; 5251 5252 sa = memdup_user(arg, sizeof(*sa)); 5253 if (IS_ERR(sa)) { 5254 ret = PTR_ERR(sa); 5255 sa = NULL; 5256 goto out; 5257 } 5258 5259 ret = _btrfs_ioctl_set_received_subvol(file, sa); 5260 5261 if (ret) 5262 goto out; 5263 5264 ret = copy_to_user(arg, sa, sizeof(*sa)); 5265 if (ret) 5266 ret = -EFAULT; 5267 5268 out: 5269 kfree(sa); 5270 return ret; 5271 } 5272 5273 static int btrfs_ioctl_get_fslabel(struct file *file, void __user *arg) 5274 { 5275 struct btrfs_root *root = BTRFS_I(file_inode(file))->root; 5276 size_t len; 5277 int ret; 5278 char label[BTRFS_LABEL_SIZE]; 5279 5280 spin_lock(&root->fs_info->super_lock); 5281 memcpy(label, root->fs_info->super_copy->label, BTRFS_LABEL_SIZE); 5282 spin_unlock(&root->fs_info->super_lock); 5283 5284 len = strnlen(label, BTRFS_LABEL_SIZE); 5285 5286 if (len == BTRFS_LABEL_SIZE) { 5287 btrfs_warn(root->fs_info, 5288 "label is too long, return the first %zu bytes", --len); 5289 } 5290 5291 ret = copy_to_user(arg, label, len); 5292 5293 return ret ? -EFAULT : 0; 5294 } 5295 5296 static int btrfs_ioctl_set_fslabel(struct file *file, void __user *arg) 5297 { 5298 struct btrfs_root *root = BTRFS_I(file_inode(file))->root; 5299 struct btrfs_super_block *super_block = root->fs_info->super_copy; 5300 struct btrfs_trans_handle *trans; 5301 char label[BTRFS_LABEL_SIZE]; 5302 int ret; 5303 5304 if (!capable(CAP_SYS_ADMIN)) 5305 return -EPERM; 5306 5307 if (copy_from_user(label, arg, sizeof(label))) 5308 return -EFAULT; 5309 5310 if (strnlen(label, BTRFS_LABEL_SIZE) == BTRFS_LABEL_SIZE) { 5311 btrfs_err(root->fs_info, 5312 "unable to set label with more than %d bytes", 5313 BTRFS_LABEL_SIZE - 1); 5314 return -EINVAL; 5315 } 5316 5317 ret = mnt_want_write_file(file); 5318 if (ret) 5319 return ret; 5320 5321 trans = btrfs_start_transaction(root, 0); 5322 if (IS_ERR(trans)) { 5323 ret = PTR_ERR(trans); 5324 goto out_unlock; 5325 } 5326 5327 spin_lock(&root->fs_info->super_lock); 5328 strcpy(super_block->label, label); 5329 spin_unlock(&root->fs_info->super_lock); 5330 ret = btrfs_commit_transaction(trans, root); 5331 5332 out_unlock: 5333 mnt_drop_write_file(file); 5334 return ret; 5335 } 5336 5337 #define INIT_FEATURE_FLAGS(suffix) \ 5338 { .compat_flags = BTRFS_FEATURE_COMPAT_##suffix, \ 5339 .compat_ro_flags = BTRFS_FEATURE_COMPAT_RO_##suffix, \ 5340 .incompat_flags = BTRFS_FEATURE_INCOMPAT_##suffix } 5341 5342 int btrfs_ioctl_get_supported_features(void __user *arg) 5343 { 5344 static const struct btrfs_ioctl_feature_flags features[3] = { 5345 INIT_FEATURE_FLAGS(SUPP), 5346 INIT_FEATURE_FLAGS(SAFE_SET), 5347 INIT_FEATURE_FLAGS(SAFE_CLEAR) 5348 }; 5349 5350 if (copy_to_user(arg, &features, sizeof(features))) 5351 return -EFAULT; 5352 5353 return 0; 5354 } 5355 5356 static int btrfs_ioctl_get_features(struct file *file, void __user *arg) 5357 { 5358 struct btrfs_root *root = BTRFS_I(file_inode(file))->root; 5359 struct btrfs_super_block *super_block = root->fs_info->super_copy; 5360 struct btrfs_ioctl_feature_flags features; 5361 5362 features.compat_flags = btrfs_super_compat_flags(super_block); 5363 features.compat_ro_flags = btrfs_super_compat_ro_flags(super_block); 5364 features.incompat_flags = btrfs_super_incompat_flags(super_block); 5365 5366 if (copy_to_user(arg, &features, sizeof(features))) 5367 return -EFAULT; 5368 5369 return 0; 5370 } 5371 5372 static int check_feature_bits(struct btrfs_root *root, 5373 enum btrfs_feature_set set, 5374 u64 change_mask, u64 flags, u64 supported_flags, 5375 u64 safe_set, u64 safe_clear) 5376 { 5377 const char *type = btrfs_feature_set_names[set]; 5378 char *names; 5379 u64 disallowed, unsupported; 5380 u64 set_mask = flags & change_mask; 5381 u64 clear_mask = ~flags & change_mask; 5382 5383 unsupported = set_mask & ~supported_flags; 5384 if (unsupported) { 5385 names = btrfs_printable_features(set, unsupported); 5386 if (names) { 5387 btrfs_warn(root->fs_info, 5388 "this kernel does not support the %s feature bit%s", 5389 names, strchr(names, ',') ? "s" : ""); 5390 kfree(names); 5391 } else 5392 btrfs_warn(root->fs_info, 5393 "this kernel does not support %s bits 0x%llx", 5394 type, unsupported); 5395 return -EOPNOTSUPP; 5396 } 5397 5398 disallowed = set_mask & ~safe_set; 5399 if (disallowed) { 5400 names = btrfs_printable_features(set, disallowed); 5401 if (names) { 5402 btrfs_warn(root->fs_info, 5403 "can't set the %s feature bit%s while mounted", 5404 names, strchr(names, ',') ? "s" : ""); 5405 kfree(names); 5406 } else 5407 btrfs_warn(root->fs_info, 5408 "can't set %s bits 0x%llx while mounted", 5409 type, disallowed); 5410 return -EPERM; 5411 } 5412 5413 disallowed = clear_mask & ~safe_clear; 5414 if (disallowed) { 5415 names = btrfs_printable_features(set, disallowed); 5416 if (names) { 5417 btrfs_warn(root->fs_info, 5418 "can't clear the %s feature bit%s while mounted", 5419 names, strchr(names, ',') ? "s" : ""); 5420 kfree(names); 5421 } else 5422 btrfs_warn(root->fs_info, 5423 "can't clear %s bits 0x%llx while mounted", 5424 type, disallowed); 5425 return -EPERM; 5426 } 5427 5428 return 0; 5429 } 5430 5431 #define check_feature(root, change_mask, flags, mask_base) \ 5432 check_feature_bits(root, FEAT_##mask_base, change_mask, flags, \ 5433 BTRFS_FEATURE_ ## mask_base ## _SUPP, \ 5434 BTRFS_FEATURE_ ## mask_base ## _SAFE_SET, \ 5435 BTRFS_FEATURE_ ## mask_base ## _SAFE_CLEAR) 5436 5437 static int btrfs_ioctl_set_features(struct file *file, void __user *arg) 5438 { 5439 struct btrfs_root *root = BTRFS_I(file_inode(file))->root; 5440 struct btrfs_super_block *super_block = root->fs_info->super_copy; 5441 struct btrfs_ioctl_feature_flags flags[2]; 5442 struct btrfs_trans_handle *trans; 5443 u64 newflags; 5444 int ret; 5445 5446 if (!capable(CAP_SYS_ADMIN)) 5447 return -EPERM; 5448 5449 if (copy_from_user(flags, arg, sizeof(flags))) 5450 return -EFAULT; 5451 5452 /* Nothing to do */ 5453 if (!flags[0].compat_flags && !flags[0].compat_ro_flags && 5454 !flags[0].incompat_flags) 5455 return 0; 5456 5457 ret = check_feature(root, flags[0].compat_flags, 5458 flags[1].compat_flags, COMPAT); 5459 if (ret) 5460 return ret; 5461 5462 ret = check_feature(root, flags[0].compat_ro_flags, 5463 flags[1].compat_ro_flags, COMPAT_RO); 5464 if (ret) 5465 return ret; 5466 5467 ret = check_feature(root, flags[0].incompat_flags, 5468 flags[1].incompat_flags, INCOMPAT); 5469 if (ret) 5470 return ret; 5471 5472 ret = mnt_want_write_file(file); 5473 if (ret) 5474 return ret; 5475 5476 trans = btrfs_start_transaction(root, 0); 5477 if (IS_ERR(trans)) { 5478 ret = PTR_ERR(trans); 5479 goto out_drop_write; 5480 } 5481 5482 spin_lock(&root->fs_info->super_lock); 5483 newflags = btrfs_super_compat_flags(super_block); 5484 newflags |= flags[0].compat_flags & flags[1].compat_flags; 5485 newflags &= ~(flags[0].compat_flags & ~flags[1].compat_flags); 5486 btrfs_set_super_compat_flags(super_block, newflags); 5487 5488 newflags = btrfs_super_compat_ro_flags(super_block); 5489 newflags |= flags[0].compat_ro_flags & flags[1].compat_ro_flags; 5490 newflags &= ~(flags[0].compat_ro_flags & ~flags[1].compat_ro_flags); 5491 btrfs_set_super_compat_ro_flags(super_block, newflags); 5492 5493 newflags = btrfs_super_incompat_flags(super_block); 5494 newflags |= flags[0].incompat_flags & flags[1].incompat_flags; 5495 newflags &= ~(flags[0].incompat_flags & ~flags[1].incompat_flags); 5496 btrfs_set_super_incompat_flags(super_block, newflags); 5497 spin_unlock(&root->fs_info->super_lock); 5498 5499 ret = btrfs_commit_transaction(trans, root); 5500 out_drop_write: 5501 mnt_drop_write_file(file); 5502 5503 return ret; 5504 } 5505 5506 long btrfs_ioctl(struct file *file, unsigned int 5507 cmd, unsigned long arg) 5508 { 5509 struct btrfs_root *root = BTRFS_I(file_inode(file))->root; 5510 void __user *argp = (void __user *)arg; 5511 5512 switch (cmd) { 5513 case FS_IOC_GETFLAGS: 5514 return btrfs_ioctl_getflags(file, argp); 5515 case FS_IOC_SETFLAGS: 5516 return btrfs_ioctl_setflags(file, argp); 5517 case FS_IOC_GETVERSION: 5518 return btrfs_ioctl_getversion(file, argp); 5519 case FITRIM: 5520 return btrfs_ioctl_fitrim(file, argp); 5521 case BTRFS_IOC_SNAP_CREATE: 5522 return btrfs_ioctl_snap_create(file, argp, 0); 5523 case BTRFS_IOC_SNAP_CREATE_V2: 5524 return btrfs_ioctl_snap_create_v2(file, argp, 0); 5525 case BTRFS_IOC_SUBVOL_CREATE: 5526 return btrfs_ioctl_snap_create(file, argp, 1); 5527 case BTRFS_IOC_SUBVOL_CREATE_V2: 5528 return btrfs_ioctl_snap_create_v2(file, argp, 1); 5529 case BTRFS_IOC_SNAP_DESTROY: 5530 return btrfs_ioctl_snap_destroy(file, argp); 5531 case BTRFS_IOC_SUBVOL_GETFLAGS: 5532 return btrfs_ioctl_subvol_getflags(file, argp); 5533 case BTRFS_IOC_SUBVOL_SETFLAGS: 5534 return btrfs_ioctl_subvol_setflags(file, argp); 5535 case BTRFS_IOC_DEFAULT_SUBVOL: 5536 return btrfs_ioctl_default_subvol(file, argp); 5537 case BTRFS_IOC_DEFRAG: 5538 return btrfs_ioctl_defrag(file, NULL); 5539 case BTRFS_IOC_DEFRAG_RANGE: 5540 return btrfs_ioctl_defrag(file, argp); 5541 case BTRFS_IOC_RESIZE: 5542 return btrfs_ioctl_resize(file, argp); 5543 case BTRFS_IOC_ADD_DEV: 5544 return btrfs_ioctl_add_dev(root, argp); 5545 case BTRFS_IOC_RM_DEV: 5546 return btrfs_ioctl_rm_dev(file, argp); 5547 case BTRFS_IOC_RM_DEV_V2: 5548 return btrfs_ioctl_rm_dev_v2(file, argp); 5549 case BTRFS_IOC_FS_INFO: 5550 return btrfs_ioctl_fs_info(root, argp); 5551 case BTRFS_IOC_DEV_INFO: 5552 return btrfs_ioctl_dev_info(root, argp); 5553 case BTRFS_IOC_BALANCE: 5554 return btrfs_ioctl_balance(file, NULL); 5555 case BTRFS_IOC_TRANS_START: 5556 return btrfs_ioctl_trans_start(file); 5557 case BTRFS_IOC_TRANS_END: 5558 return btrfs_ioctl_trans_end(file); 5559 case BTRFS_IOC_TREE_SEARCH: 5560 return btrfs_ioctl_tree_search(file, argp); 5561 case BTRFS_IOC_TREE_SEARCH_V2: 5562 return btrfs_ioctl_tree_search_v2(file, argp); 5563 case BTRFS_IOC_INO_LOOKUP: 5564 return btrfs_ioctl_ino_lookup(file, argp); 5565 case BTRFS_IOC_INO_PATHS: 5566 return btrfs_ioctl_ino_to_path(root, argp); 5567 case BTRFS_IOC_LOGICAL_INO: 5568 return btrfs_ioctl_logical_to_ino(root, argp); 5569 case BTRFS_IOC_SPACE_INFO: 5570 return btrfs_ioctl_space_info(root, argp); 5571 case BTRFS_IOC_SYNC: { 5572 int ret; 5573 5574 ret = btrfs_start_delalloc_roots(root->fs_info, 0, -1); 5575 if (ret) 5576 return ret; 5577 ret = btrfs_sync_fs(file_inode(file)->i_sb, 1); 5578 /* 5579 * The transaction thread may want to do more work, 5580 * namely it pokes the cleaner kthread that will start 5581 * processing uncleaned subvols. 5582 */ 5583 wake_up_process(root->fs_info->transaction_kthread); 5584 return ret; 5585 } 5586 case BTRFS_IOC_START_SYNC: 5587 return btrfs_ioctl_start_sync(root, argp); 5588 case BTRFS_IOC_WAIT_SYNC: 5589 return btrfs_ioctl_wait_sync(root, argp); 5590 case BTRFS_IOC_SCRUB: 5591 return btrfs_ioctl_scrub(file, argp); 5592 case BTRFS_IOC_SCRUB_CANCEL: 5593 return btrfs_ioctl_scrub_cancel(root, argp); 5594 case BTRFS_IOC_SCRUB_PROGRESS: 5595 return btrfs_ioctl_scrub_progress(root, argp); 5596 case BTRFS_IOC_BALANCE_V2: 5597 return btrfs_ioctl_balance(file, argp); 5598 case BTRFS_IOC_BALANCE_CTL: 5599 return btrfs_ioctl_balance_ctl(root, arg); 5600 case BTRFS_IOC_BALANCE_PROGRESS: 5601 return btrfs_ioctl_balance_progress(root, argp); 5602 case BTRFS_IOC_SET_RECEIVED_SUBVOL: 5603 return btrfs_ioctl_set_received_subvol(file, argp); 5604 #ifdef CONFIG_64BIT 5605 case BTRFS_IOC_SET_RECEIVED_SUBVOL_32: 5606 return btrfs_ioctl_set_received_subvol_32(file, argp); 5607 #endif 5608 case BTRFS_IOC_SEND: 5609 return btrfs_ioctl_send(file, argp); 5610 case BTRFS_IOC_GET_DEV_STATS: 5611 return btrfs_ioctl_get_dev_stats(root, argp); 5612 case BTRFS_IOC_QUOTA_CTL: 5613 return btrfs_ioctl_quota_ctl(file, argp); 5614 case BTRFS_IOC_QGROUP_ASSIGN: 5615 return btrfs_ioctl_qgroup_assign(file, argp); 5616 case BTRFS_IOC_QGROUP_CREATE: 5617 return btrfs_ioctl_qgroup_create(file, argp); 5618 case BTRFS_IOC_QGROUP_LIMIT: 5619 return btrfs_ioctl_qgroup_limit(file, argp); 5620 case BTRFS_IOC_QUOTA_RESCAN: 5621 return btrfs_ioctl_quota_rescan(file, argp); 5622 case BTRFS_IOC_QUOTA_RESCAN_STATUS: 5623 return btrfs_ioctl_quota_rescan_status(file, argp); 5624 case BTRFS_IOC_QUOTA_RESCAN_WAIT: 5625 return btrfs_ioctl_quota_rescan_wait(file, argp); 5626 case BTRFS_IOC_DEV_REPLACE: 5627 return btrfs_ioctl_dev_replace(root, argp); 5628 case BTRFS_IOC_GET_FSLABEL: 5629 return btrfs_ioctl_get_fslabel(file, argp); 5630 case BTRFS_IOC_SET_FSLABEL: 5631 return btrfs_ioctl_set_fslabel(file, argp); 5632 case BTRFS_IOC_GET_SUPPORTED_FEATURES: 5633 return btrfs_ioctl_get_supported_features(argp); 5634 case BTRFS_IOC_GET_FEATURES: 5635 return btrfs_ioctl_get_features(file, argp); 5636 case BTRFS_IOC_SET_FEATURES: 5637 return btrfs_ioctl_set_features(file, argp); 5638 } 5639 5640 return -ENOTTY; 5641 } 5642 5643 #ifdef CONFIG_COMPAT 5644 long btrfs_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg) 5645 { 5646 switch (cmd) { 5647 case FS_IOC32_GETFLAGS: 5648 cmd = FS_IOC_GETFLAGS; 5649 break; 5650 case FS_IOC32_SETFLAGS: 5651 cmd = FS_IOC_SETFLAGS; 5652 break; 5653 case FS_IOC32_GETVERSION: 5654 cmd = FS_IOC_GETVERSION; 5655 break; 5656 default: 5657 return -ENOIOCTLCMD; 5658 } 5659 5660 return btrfs_ioctl(file, cmd, (unsigned long) compat_ptr(arg)); 5661 } 5662 #endif 5663