1 // SPDX-License-Identifier: GPL-2.0 2 /* 3 * Copyright (C) 2007 Oracle. All rights reserved. 4 */ 5 6 #include <linux/kernel.h> 7 #include <linux/bio.h> 8 #include <linux/file.h> 9 #include <linux/fs.h> 10 #include <linux/fsnotify.h> 11 #include <linux/pagemap.h> 12 #include <linux/highmem.h> 13 #include <linux/time.h> 14 #include <linux/string.h> 15 #include <linux/backing-dev.h> 16 #include <linux/mount.h> 17 #include <linux/namei.h> 18 #include <linux/writeback.h> 19 #include <linux/compat.h> 20 #include <linux/security.h> 21 #include <linux/xattr.h> 22 #include <linux/mm.h> 23 #include <linux/slab.h> 24 #include <linux/blkdev.h> 25 #include <linux/uuid.h> 26 #include <linux/btrfs.h> 27 #include <linux/uaccess.h> 28 #include <linux/iversion.h> 29 #include <linux/fileattr.h> 30 #include <linux/fsverity.h> 31 #include <linux/sched/xacct.h> 32 #include <linux/io_uring/cmd.h> 33 #include "ctree.h" 34 #include "disk-io.h" 35 #include "export.h" 36 #include "transaction.h" 37 #include "btrfs_inode.h" 38 #include "volumes.h" 39 #include "locking.h" 40 #include "backref.h" 41 #include "send.h" 42 #include "dev-replace.h" 43 #include "props.h" 44 #include "sysfs.h" 45 #include "qgroup.h" 46 #include "tree-log.h" 47 #include "compression.h" 48 #include "space-info.h" 49 #include "block-group.h" 50 #include "fs.h" 51 #include "accessors.h" 52 #include "extent-tree.h" 53 #include "root-tree.h" 54 #include "defrag.h" 55 #include "dir-item.h" 56 #include "uuid-tree.h" 57 #include "ioctl.h" 58 #include "file.h" 59 #include "file-item.h" 60 #include "scrub.h" 61 #include "super.h" 62 63 #ifdef CONFIG_64BIT 64 /* If we have a 32-bit userspace and 64-bit kernel, then the UAPI 65 * structures are incorrect, as the timespec structure from userspace 66 * is 4 bytes too small. We define these alternatives here to teach 67 * the kernel about the 32-bit struct packing. 68 */ 69 struct btrfs_ioctl_timespec_32 { 70 __u64 sec; 71 __u32 nsec; 72 } __attribute__ ((__packed__)); 73 74 struct btrfs_ioctl_received_subvol_args_32 { 75 char uuid[BTRFS_UUID_SIZE]; /* in */ 76 __u64 stransid; /* in */ 77 __u64 rtransid; /* out */ 78 struct btrfs_ioctl_timespec_32 stime; /* in */ 79 struct btrfs_ioctl_timespec_32 rtime; /* out */ 80 __u64 flags; /* in */ 81 __u64 reserved[16]; /* in */ 82 } __attribute__ ((__packed__)); 83 84 #define BTRFS_IOC_SET_RECEIVED_SUBVOL_32 _IOWR(BTRFS_IOCTL_MAGIC, 37, \ 85 struct btrfs_ioctl_received_subvol_args_32) 86 87 struct btrfs_ioctl_get_subvol_info_args_32 { 88 __u64 treeid; 89 char name[BTRFS_VOL_NAME_MAX + 1]; 90 __u64 parent_id; 91 __u64 dirid; 92 __u64 generation; 93 __u64 flags; 94 __u8 uuid[BTRFS_UUID_SIZE]; 95 __u8 parent_uuid[BTRFS_UUID_SIZE]; 96 __u8 received_uuid[BTRFS_UUID_SIZE]; 97 __u64 ctransid; 98 __u64 otransid; 99 __u64 stransid; 100 __u64 rtransid; 101 struct btrfs_ioctl_timespec_32 ctime; 102 struct btrfs_ioctl_timespec_32 otime; 103 struct btrfs_ioctl_timespec_32 stime; 104 struct btrfs_ioctl_timespec_32 rtime; 105 __u64 reserved[8]; 106 } __attribute__ ((__packed__)); 107 108 #define BTRFS_IOC_GET_SUBVOL_INFO_32 _IOR(BTRFS_IOCTL_MAGIC, 60, \ 109 struct btrfs_ioctl_get_subvol_info_args_32) 110 #endif 111 112 #if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT) 113 struct btrfs_ioctl_send_args_32 { 114 __s64 send_fd; /* in */ 115 __u64 clone_sources_count; /* in */ 116 compat_uptr_t clone_sources; /* in */ 117 __u64 parent_root; /* in */ 118 __u64 flags; /* in */ 119 __u32 version; /* in */ 120 __u8 reserved[28]; /* in */ 121 } __attribute__ ((__packed__)); 122 123 #define BTRFS_IOC_SEND_32 _IOW(BTRFS_IOCTL_MAGIC, 38, \ 124 struct btrfs_ioctl_send_args_32) 125 126 struct btrfs_ioctl_encoded_io_args_32 { 127 compat_uptr_t iov; 128 compat_ulong_t iovcnt; 129 __s64 offset; 130 __u64 flags; 131 __u64 len; 132 __u64 unencoded_len; 133 __u64 unencoded_offset; 134 __u32 compression; 135 __u32 encryption; 136 __u8 reserved[64]; 137 }; 138 139 #define BTRFS_IOC_ENCODED_READ_32 _IOR(BTRFS_IOCTL_MAGIC, 64, \ 140 struct btrfs_ioctl_encoded_io_args_32) 141 #define BTRFS_IOC_ENCODED_WRITE_32 _IOW(BTRFS_IOCTL_MAGIC, 64, \ 142 struct btrfs_ioctl_encoded_io_args_32) 143 #endif 144 145 /* Mask out flags that are inappropriate for the given type of inode. */ 146 static unsigned int btrfs_mask_fsflags_for_type(const struct inode *inode, 147 unsigned int flags) 148 { 149 if (S_ISDIR(inode->i_mode)) 150 return flags; 151 else if (S_ISREG(inode->i_mode)) 152 return flags & ~FS_DIRSYNC_FL; 153 else 154 return flags & (FS_NODUMP_FL | FS_NOATIME_FL); 155 } 156 157 /* 158 * Export internal inode flags to the format expected by the FS_IOC_GETFLAGS 159 * ioctl. 160 */ 161 static unsigned int btrfs_inode_flags_to_fsflags(const struct btrfs_inode *inode) 162 { 163 unsigned int iflags = 0; 164 u32 flags = inode->flags; 165 u32 ro_flags = inode->ro_flags; 166 167 if (flags & BTRFS_INODE_SYNC) 168 iflags |= FS_SYNC_FL; 169 if (flags & BTRFS_INODE_IMMUTABLE) 170 iflags |= FS_IMMUTABLE_FL; 171 if (flags & BTRFS_INODE_APPEND) 172 iflags |= FS_APPEND_FL; 173 if (flags & BTRFS_INODE_NODUMP) 174 iflags |= FS_NODUMP_FL; 175 if (flags & BTRFS_INODE_NOATIME) 176 iflags |= FS_NOATIME_FL; 177 if (flags & BTRFS_INODE_DIRSYNC) 178 iflags |= FS_DIRSYNC_FL; 179 if (flags & BTRFS_INODE_NODATACOW) 180 iflags |= FS_NOCOW_FL; 181 if (ro_flags & BTRFS_INODE_RO_VERITY) 182 iflags |= FS_VERITY_FL; 183 184 if (flags & BTRFS_INODE_NOCOMPRESS) 185 iflags |= FS_NOCOMP_FL; 186 else if (flags & BTRFS_INODE_COMPRESS) 187 iflags |= FS_COMPR_FL; 188 189 return iflags; 190 } 191 192 /* 193 * Update inode->i_flags based on the btrfs internal flags. 194 */ 195 void btrfs_sync_inode_flags_to_i_flags(struct btrfs_inode *inode) 196 { 197 unsigned int new_fl = 0; 198 199 if (inode->flags & BTRFS_INODE_SYNC) 200 new_fl |= S_SYNC; 201 if (inode->flags & BTRFS_INODE_IMMUTABLE) 202 new_fl |= S_IMMUTABLE; 203 if (inode->flags & BTRFS_INODE_APPEND) 204 new_fl |= S_APPEND; 205 if (inode->flags & BTRFS_INODE_NOATIME) 206 new_fl |= S_NOATIME; 207 if (inode->flags & BTRFS_INODE_DIRSYNC) 208 new_fl |= S_DIRSYNC; 209 if (inode->ro_flags & BTRFS_INODE_RO_VERITY) 210 new_fl |= S_VERITY; 211 212 set_mask_bits(&inode->vfs_inode.i_flags, 213 S_SYNC | S_APPEND | S_IMMUTABLE | S_NOATIME | S_DIRSYNC | 214 S_VERITY, new_fl); 215 } 216 217 /* 218 * Check if @flags are a supported and valid set of FS_*_FL flags and that 219 * the old and new flags are not conflicting 220 */ 221 static int check_fsflags(unsigned int old_flags, unsigned int flags) 222 { 223 if (flags & ~(FS_IMMUTABLE_FL | FS_APPEND_FL | \ 224 FS_NOATIME_FL | FS_NODUMP_FL | \ 225 FS_SYNC_FL | FS_DIRSYNC_FL | \ 226 FS_NOCOMP_FL | FS_COMPR_FL | 227 FS_NOCOW_FL)) 228 return -EOPNOTSUPP; 229 230 /* COMPR and NOCOMP on new/old are valid */ 231 if ((flags & FS_NOCOMP_FL) && (flags & FS_COMPR_FL)) 232 return -EINVAL; 233 234 if ((flags & FS_COMPR_FL) && (flags & FS_NOCOW_FL)) 235 return -EINVAL; 236 237 /* NOCOW and compression options are mutually exclusive */ 238 if ((old_flags & FS_NOCOW_FL) && (flags & (FS_COMPR_FL | FS_NOCOMP_FL))) 239 return -EINVAL; 240 if ((flags & FS_NOCOW_FL) && (old_flags & (FS_COMPR_FL | FS_NOCOMP_FL))) 241 return -EINVAL; 242 243 return 0; 244 } 245 246 static int check_fsflags_compatible(const struct btrfs_fs_info *fs_info, 247 unsigned int flags) 248 { 249 if (btrfs_is_zoned(fs_info) && (flags & FS_NOCOW_FL)) 250 return -EPERM; 251 252 return 0; 253 } 254 255 int btrfs_check_ioctl_vol_args_path(const struct btrfs_ioctl_vol_args *vol_args) 256 { 257 if (memchr(vol_args->name, 0, sizeof(vol_args->name)) == NULL) 258 return -ENAMETOOLONG; 259 return 0; 260 } 261 262 static int btrfs_check_ioctl_vol_args2_subvol_name(const struct btrfs_ioctl_vol_args_v2 *vol_args2) 263 { 264 if (memchr(vol_args2->name, 0, sizeof(vol_args2->name)) == NULL) 265 return -ENAMETOOLONG; 266 return 0; 267 } 268 269 /* 270 * Set flags/xflags from the internal inode flags. The remaining items of 271 * fsxattr are zeroed. 272 */ 273 int btrfs_fileattr_get(struct dentry *dentry, struct file_kattr *fa) 274 { 275 const struct btrfs_inode *inode = BTRFS_I(d_inode(dentry)); 276 277 fileattr_fill_flags(fa, btrfs_inode_flags_to_fsflags(inode)); 278 return 0; 279 } 280 281 int btrfs_fileattr_set(struct mnt_idmap *idmap, 282 struct dentry *dentry, struct file_kattr *fa) 283 { 284 struct btrfs_inode *inode = BTRFS_I(d_inode(dentry)); 285 struct btrfs_root *root = inode->root; 286 struct btrfs_fs_info *fs_info = root->fs_info; 287 struct btrfs_trans_handle *trans; 288 unsigned int fsflags, old_fsflags; 289 int ret; 290 const char *comp = NULL; 291 u32 inode_flags; 292 bool prop_set = false; 293 294 if (btrfs_root_readonly(root)) 295 return -EROFS; 296 297 if (fileattr_has_fsx(fa)) 298 return -EOPNOTSUPP; 299 300 fsflags = btrfs_mask_fsflags_for_type(&inode->vfs_inode, fa->flags); 301 old_fsflags = btrfs_inode_flags_to_fsflags(inode); 302 ret = check_fsflags(old_fsflags, fsflags); 303 if (ret) 304 return ret; 305 306 ret = check_fsflags_compatible(fs_info, fsflags); 307 if (ret) 308 return ret; 309 310 inode_flags = inode->flags; 311 if (fsflags & FS_SYNC_FL) 312 inode_flags |= BTRFS_INODE_SYNC; 313 else 314 inode_flags &= ~BTRFS_INODE_SYNC; 315 if (fsflags & FS_IMMUTABLE_FL) 316 inode_flags |= BTRFS_INODE_IMMUTABLE; 317 else 318 inode_flags &= ~BTRFS_INODE_IMMUTABLE; 319 if (fsflags & FS_APPEND_FL) 320 inode_flags |= BTRFS_INODE_APPEND; 321 else 322 inode_flags &= ~BTRFS_INODE_APPEND; 323 if (fsflags & FS_NODUMP_FL) 324 inode_flags |= BTRFS_INODE_NODUMP; 325 else 326 inode_flags &= ~BTRFS_INODE_NODUMP; 327 if (fsflags & FS_NOATIME_FL) 328 inode_flags |= BTRFS_INODE_NOATIME; 329 else 330 inode_flags &= ~BTRFS_INODE_NOATIME; 331 332 /* If coming from FS_IOC_FSSETXATTR then skip unconverted flags */ 333 if (!fa->flags_valid) { 334 /* 1 item for the inode */ 335 trans = btrfs_start_transaction(root, 1); 336 if (IS_ERR(trans)) 337 return PTR_ERR(trans); 338 goto update_flags; 339 } 340 341 if (fsflags & FS_DIRSYNC_FL) 342 inode_flags |= BTRFS_INODE_DIRSYNC; 343 else 344 inode_flags &= ~BTRFS_INODE_DIRSYNC; 345 if (fsflags & FS_NOCOW_FL) { 346 if (S_ISREG(inode->vfs_inode.i_mode)) { 347 /* 348 * It's safe to turn csums off here, no extents exist. 349 * Otherwise we want the flag to reflect the real COW 350 * status of the file and will not set it. 351 */ 352 if (inode->vfs_inode.i_size == 0) 353 inode_flags |= BTRFS_INODE_NODATACOW | 354 BTRFS_INODE_NODATASUM; 355 } else { 356 inode_flags |= BTRFS_INODE_NODATACOW; 357 } 358 } else { 359 /* 360 * Revert back under same assumptions as above 361 */ 362 if (S_ISREG(inode->vfs_inode.i_mode)) { 363 if (inode->vfs_inode.i_size == 0) 364 inode_flags &= ~(BTRFS_INODE_NODATACOW | 365 BTRFS_INODE_NODATASUM); 366 } else { 367 inode_flags &= ~BTRFS_INODE_NODATACOW; 368 } 369 } 370 371 /* 372 * The COMPRESS flag can only be changed by users, while the NOCOMPRESS 373 * flag may be changed automatically if compression code won't make 374 * things smaller. 375 */ 376 if (fsflags & FS_NOCOMP_FL) { 377 inode_flags &= ~BTRFS_INODE_COMPRESS; 378 inode_flags |= BTRFS_INODE_NOCOMPRESS; 379 } else if (fsflags & FS_COMPR_FL) { 380 381 if (IS_SWAPFILE(&inode->vfs_inode)) 382 return -ETXTBSY; 383 384 inode_flags |= BTRFS_INODE_COMPRESS; 385 inode_flags &= ~BTRFS_INODE_NOCOMPRESS; 386 387 comp = btrfs_compress_type2str(fs_info->compress_type); 388 if (!comp || comp[0] == 0) 389 comp = btrfs_compress_type2str(BTRFS_COMPRESS_ZLIB); 390 } else { 391 inode_flags &= ~(BTRFS_INODE_COMPRESS | BTRFS_INODE_NOCOMPRESS); 392 } 393 394 /* 395 * 1 for inode item 396 * 2 for properties 397 */ 398 trans = btrfs_start_transaction(root, 3); 399 if (IS_ERR(trans)) 400 return PTR_ERR(trans); 401 402 if (comp) { 403 ret = btrfs_set_prop(trans, inode, "btrfs.compression", 404 comp, strlen(comp), 0); 405 if (ret) 406 goto out_end_trans; 407 prop_set = true; 408 } else { 409 ret = btrfs_set_prop(trans, inode, "btrfs.compression", NULL, 0, 0); 410 prop_set = (ret == 0); 411 /* If ret == -ENODATA ignore and proceed to update inode item. */ 412 if (ret && ret != -ENODATA) 413 goto out_end_trans; 414 } 415 416 update_flags: 417 inode->flags = inode_flags; 418 btrfs_update_inode_mapping_flags(inode); 419 btrfs_sync_inode_flags_to_i_flags(inode); 420 inode_inc_iversion(&inode->vfs_inode); 421 inode_set_ctime_current(&inode->vfs_inode); 422 ret = btrfs_update_inode(trans, inode); 423 /* 424 * If we set a property or deleted one, we must abort if we fail to 425 * update the inode, to avoid persisting an inconsistent state. 426 */ 427 if (unlikely(ret && prop_set)) 428 btrfs_abort_transaction(trans, ret); 429 430 out_end_trans: 431 btrfs_end_transaction(trans); 432 return ret; 433 } 434 435 static int btrfs_ioctl_getversion(const struct inode *inode, int __user *arg) 436 { 437 return put_user(inode->i_generation, arg); 438 } 439 440 static noinline int btrfs_ioctl_fitrim(struct btrfs_fs_info *fs_info, 441 void __user *arg) 442 { 443 struct btrfs_device *device; 444 struct fstrim_range range; 445 u64 minlen = ULLONG_MAX; 446 u64 num_devices = 0; 447 int ret; 448 449 if (!capable(CAP_SYS_ADMIN)) 450 return -EPERM; 451 452 /* 453 * btrfs_trim_block_group() depends on space cache, which is not 454 * available in zoned filesystem. So, disallow fitrim on a zoned 455 * filesystem for now. 456 */ 457 if (btrfs_is_zoned(fs_info)) 458 return -EOPNOTSUPP; 459 460 /* 461 * If the fs is mounted with nologreplay, which requires it to be 462 * mounted in RO mode as well, we can not allow discard on free space 463 * inside block groups, because log trees refer to extents that are not 464 * pinned in a block group's free space cache (pinning the extents is 465 * precisely the first phase of replaying a log tree). 466 */ 467 if (btrfs_test_opt(fs_info, NOLOGREPLAY)) 468 return -EROFS; 469 470 rcu_read_lock(); 471 list_for_each_entry_rcu(device, &fs_info->fs_devices->devices, 472 dev_list) { 473 if (!device->bdev || !bdev_max_discard_sectors(device->bdev)) 474 continue; 475 num_devices++; 476 minlen = min_t(u64, bdev_discard_granularity(device->bdev), 477 minlen); 478 } 479 rcu_read_unlock(); 480 481 if (!num_devices) 482 return -EOPNOTSUPP; 483 if (copy_from_user(&range, arg, sizeof(range))) 484 return -EFAULT; 485 486 /* 487 * NOTE: Don't truncate the range using super->total_bytes. Bytenr of 488 * block group is in the logical address space, which can be any 489 * sectorsize aligned bytenr in the range [0, U64_MAX]. 490 */ 491 if (range.len < fs_info->sectorsize) 492 return -EINVAL; 493 494 range.minlen = max(range.minlen, minlen); 495 ret = btrfs_trim_fs(fs_info, &range); 496 497 if (copy_to_user(arg, &range, sizeof(range))) 498 return -EFAULT; 499 500 return ret; 501 } 502 503 /* 504 * Calculate the number of transaction items to reserve for creating a subvolume 505 * or snapshot, not including the inode, directory entries, or parent directory. 506 */ 507 static unsigned int create_subvol_num_items(const struct btrfs_qgroup_inherit *inherit) 508 { 509 /* 510 * 1 to add root block 511 * 1 to add root item 512 * 1 to add root ref 513 * 1 to add root backref 514 * 1 to add UUID item 515 * 1 to add qgroup info 516 * 1 to add qgroup limit 517 * 518 * Ideally the last two would only be accounted if qgroups are enabled, 519 * but that can change between now and the time we would insert them. 520 */ 521 unsigned int num_items = 7; 522 523 if (inherit) { 524 /* 2 to add qgroup relations for each inherited qgroup */ 525 num_items += 2 * inherit->num_qgroups; 526 } 527 return num_items; 528 } 529 530 static noinline int create_subvol(struct mnt_idmap *idmap, 531 struct inode *dir, struct dentry *dentry, 532 struct btrfs_qgroup_inherit *inherit) 533 { 534 struct btrfs_fs_info *fs_info = inode_to_fs_info(dir); 535 struct btrfs_trans_handle *trans; 536 struct btrfs_key key; 537 struct btrfs_root_item AUTO_KFREE(root_item); 538 struct btrfs_inode_item *inode_item; 539 struct extent_buffer *leaf; 540 struct btrfs_root *root = BTRFS_I(dir)->root; 541 struct btrfs_root *new_root; 542 struct btrfs_block_rsv block_rsv; 543 struct timespec64 cur_time = current_time(dir); 544 struct btrfs_new_inode_args new_inode_args = { 545 .dir = dir, 546 .dentry = dentry, 547 .subvol = true, 548 }; 549 unsigned int trans_num_items; 550 int ret; 551 dev_t anon_dev; 552 u64 objectid; 553 u64 qgroup_reserved = 0; 554 555 root_item = kzalloc(sizeof(*root_item), GFP_KERNEL); 556 if (!root_item) 557 return -ENOMEM; 558 559 ret = btrfs_get_free_objectid(fs_info->tree_root, &objectid); 560 if (ret) 561 return ret; 562 563 /* 564 * Don't create subvolume whose level is not zero. Or qgroup will be 565 * screwed up since it assumes subvolume qgroup's level to be 0. 566 */ 567 if (btrfs_qgroup_level(objectid)) 568 return -ENOSPC; 569 570 ret = get_anon_bdev(&anon_dev); 571 if (ret < 0) 572 return ret; 573 574 new_inode_args.inode = btrfs_new_subvol_inode(idmap, dir); 575 if (!new_inode_args.inode) { 576 ret = -ENOMEM; 577 goto out_anon_dev; 578 } 579 ret = btrfs_new_inode_prepare(&new_inode_args, &trans_num_items); 580 if (ret) 581 goto out_inode; 582 trans_num_items += create_subvol_num_items(inherit); 583 584 btrfs_init_block_rsv(&block_rsv, BTRFS_BLOCK_RSV_TEMP); 585 ret = btrfs_subvolume_reserve_metadata(root, &block_rsv, 586 trans_num_items, false); 587 if (ret) 588 goto out_new_inode_args; 589 qgroup_reserved = block_rsv.qgroup_rsv_reserved; 590 591 trans = btrfs_start_transaction(root, 0); 592 if (IS_ERR(trans)) { 593 ret = PTR_ERR(trans); 594 goto out_release_rsv; 595 } 596 btrfs_qgroup_convert_reserved_meta(root, qgroup_reserved); 597 qgroup_reserved = 0; 598 trans->block_rsv = &block_rsv; 599 trans->bytes_reserved = block_rsv.size; 600 601 ret = btrfs_qgroup_inherit(trans, 0, objectid, btrfs_root_id(root), inherit); 602 if (ret) 603 goto out; 604 605 leaf = btrfs_alloc_tree_block(trans, root, 0, objectid, NULL, 0, 0, 0, 606 0, BTRFS_NESTING_NORMAL); 607 if (IS_ERR(leaf)) { 608 ret = PTR_ERR(leaf); 609 goto out; 610 } 611 612 btrfs_mark_buffer_dirty(trans, leaf); 613 614 inode_item = &root_item->inode; 615 btrfs_set_stack_inode_generation(inode_item, 1); 616 btrfs_set_stack_inode_size(inode_item, 3); 617 btrfs_set_stack_inode_nlink(inode_item, 1); 618 btrfs_set_stack_inode_nbytes(inode_item, 619 fs_info->nodesize); 620 btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755); 621 622 btrfs_set_root_flags(root_item, 0); 623 btrfs_set_root_limit(root_item, 0); 624 btrfs_set_stack_inode_flags(inode_item, BTRFS_INODE_ROOT_ITEM_INIT); 625 626 btrfs_set_root_bytenr(root_item, leaf->start); 627 btrfs_set_root_generation(root_item, trans->transid); 628 btrfs_set_root_level(root_item, 0); 629 btrfs_set_root_refs(root_item, 1); 630 btrfs_set_root_used(root_item, leaf->len); 631 btrfs_set_root_last_snapshot(root_item, 0); 632 633 btrfs_set_root_generation_v2(root_item, 634 btrfs_root_generation(root_item)); 635 generate_random_guid(root_item->uuid); 636 btrfs_set_stack_timespec_sec(&root_item->otime, cur_time.tv_sec); 637 btrfs_set_stack_timespec_nsec(&root_item->otime, cur_time.tv_nsec); 638 root_item->ctime = root_item->otime; 639 btrfs_set_root_ctransid(root_item, trans->transid); 640 btrfs_set_root_otransid(root_item, trans->transid); 641 642 btrfs_tree_unlock(leaf); 643 644 btrfs_set_root_dirid(root_item, BTRFS_FIRST_FREE_OBJECTID); 645 646 key.objectid = objectid; 647 key.type = BTRFS_ROOT_ITEM_KEY; 648 key.offset = 0; 649 ret = btrfs_insert_root(trans, fs_info->tree_root, &key, 650 root_item); 651 if (ret) { 652 int ret2; 653 654 /* 655 * Since we don't abort the transaction in this case, free the 656 * tree block so that we don't leak space and leave the 657 * filesystem in an inconsistent state (an extent item in the 658 * extent tree with a backreference for a root that does not 659 * exists). 660 */ 661 btrfs_tree_lock(leaf); 662 btrfs_clear_buffer_dirty(trans, leaf); 663 btrfs_tree_unlock(leaf); 664 ret2 = btrfs_free_tree_block(trans, objectid, leaf, 0, 1); 665 if (unlikely(ret2 < 0)) 666 btrfs_abort_transaction(trans, ret2); 667 free_extent_buffer(leaf); 668 goto out; 669 } 670 671 free_extent_buffer(leaf); 672 leaf = NULL; 673 674 new_root = btrfs_get_new_fs_root(fs_info, objectid, &anon_dev); 675 if (IS_ERR(new_root)) { 676 ret = PTR_ERR(new_root); 677 btrfs_abort_transaction(trans, ret); 678 goto out; 679 } 680 /* anon_dev is owned by new_root now. */ 681 anon_dev = 0; 682 BTRFS_I(new_inode_args.inode)->root = new_root; 683 /* ... and new_root is owned by new_inode_args.inode now. */ 684 685 ret = btrfs_record_root_in_trans(trans, new_root); 686 if (unlikely(ret)) { 687 btrfs_abort_transaction(trans, ret); 688 goto out; 689 } 690 691 ret = btrfs_uuid_tree_add(trans, root_item->uuid, 692 BTRFS_UUID_KEY_SUBVOL, objectid); 693 if (unlikely(ret)) { 694 btrfs_abort_transaction(trans, ret); 695 goto out; 696 } 697 698 btrfs_record_new_subvolume(trans, BTRFS_I(dir)); 699 700 ret = btrfs_create_new_inode(trans, &new_inode_args); 701 if (unlikely(ret)) { 702 btrfs_abort_transaction(trans, ret); 703 goto out; 704 } 705 706 /* 707 * Subvolumes have orphans cleaned on first dentry lookup. A new 708 * subvolume cannot have any orphans, so we should set the bit before we 709 * add the subvolume dentry to the dentry cache, so that it is in the 710 * same state as a subvolume after first lookup. 711 */ 712 set_bit(BTRFS_ROOT_ORPHAN_CLEANUP, &new_root->state); 713 d_instantiate_new(dentry, new_inode_args.inode); 714 new_inode_args.inode = NULL; 715 716 out: 717 trans->block_rsv = NULL; 718 trans->bytes_reserved = 0; 719 btrfs_end_transaction(trans); 720 out_release_rsv: 721 btrfs_block_rsv_release(fs_info, &block_rsv, (u64)-1, NULL); 722 if (qgroup_reserved) 723 btrfs_qgroup_free_meta_prealloc(root, qgroup_reserved); 724 out_new_inode_args: 725 btrfs_new_inode_args_destroy(&new_inode_args); 726 out_inode: 727 iput(new_inode_args.inode); 728 out_anon_dev: 729 if (anon_dev) 730 free_anon_bdev(anon_dev); 731 732 return ret; 733 } 734 735 static int create_snapshot(struct btrfs_root *root, struct inode *dir, 736 struct dentry *dentry, bool readonly, 737 struct btrfs_qgroup_inherit *inherit) 738 { 739 struct btrfs_fs_info *fs_info = inode_to_fs_info(dir); 740 struct inode *inode; 741 struct btrfs_pending_snapshot AUTO_KFREE(pending_snapshot); 742 unsigned int trans_num_items; 743 struct btrfs_trans_handle *trans; 744 struct btrfs_block_rsv *block_rsv; 745 u64 qgroup_reserved = 0; 746 int ret; 747 748 /* We do not support snapshotting right now. */ 749 if (btrfs_fs_incompat(fs_info, EXTENT_TREE_V2)) { 750 btrfs_warn(fs_info, 751 "extent tree v2 doesn't support snapshotting yet"); 752 return -EOPNOTSUPP; 753 } 754 755 if (btrfs_root_refs(&root->root_item) == 0) 756 return -ENOENT; 757 758 if (!test_bit(BTRFS_ROOT_SHAREABLE, &root->state)) 759 return -EINVAL; 760 761 if (atomic_read(&root->nr_swapfiles)) { 762 btrfs_warn(fs_info, 763 "cannot snapshot subvolume with active swapfile"); 764 return -ETXTBSY; 765 } 766 767 pending_snapshot = kzalloc_obj(*pending_snapshot); 768 if (!pending_snapshot) 769 return -ENOMEM; 770 771 ret = get_anon_bdev(&pending_snapshot->anon_dev); 772 if (ret < 0) 773 goto free_pending; 774 pending_snapshot->root_item = kzalloc_obj(struct btrfs_root_item); 775 pending_snapshot->path = btrfs_alloc_path(); 776 if (!pending_snapshot->root_item || !pending_snapshot->path) { 777 ret = -ENOMEM; 778 goto free_pending; 779 } 780 781 block_rsv = &pending_snapshot->block_rsv; 782 btrfs_init_block_rsv(block_rsv, BTRFS_BLOCK_RSV_TEMP); 783 /* 784 * 1 to add dir item 785 * 1 to add dir index 786 * 1 to update parent inode item 787 */ 788 trans_num_items = create_subvol_num_items(inherit) + 3; 789 ret = btrfs_subvolume_reserve_metadata(BTRFS_I(dir)->root, block_rsv, 790 trans_num_items, false); 791 if (ret) 792 goto free_pending; 793 qgroup_reserved = block_rsv->qgroup_rsv_reserved; 794 795 pending_snapshot->dentry = dentry; 796 pending_snapshot->root = root; 797 pending_snapshot->readonly = readonly; 798 pending_snapshot->dir = BTRFS_I(dir); 799 pending_snapshot->inherit = inherit; 800 801 trans = btrfs_start_transaction(root, 0); 802 if (IS_ERR(trans)) { 803 ret = PTR_ERR(trans); 804 goto fail; 805 } 806 ret = btrfs_record_root_in_trans(trans, BTRFS_I(dir)->root); 807 if (ret) { 808 btrfs_end_transaction(trans); 809 goto fail; 810 } 811 btrfs_qgroup_convert_reserved_meta(root, qgroup_reserved); 812 qgroup_reserved = 0; 813 814 trans->pending_snapshot = pending_snapshot; 815 816 ret = btrfs_commit_transaction(trans); 817 if (ret) 818 goto fail; 819 820 ret = pending_snapshot->error; 821 if (ret) 822 goto fail; 823 824 ret = btrfs_orphan_cleanup(pending_snapshot->snap); 825 if (ret) 826 goto fail; 827 828 inode = btrfs_lookup_dentry(d_inode(dentry->d_parent), dentry); 829 if (IS_ERR(inode)) { 830 ret = PTR_ERR(inode); 831 goto fail; 832 } 833 834 d_instantiate(dentry, inode); 835 ret = 0; 836 pending_snapshot->anon_dev = 0; 837 fail: 838 /* Prevent double freeing of anon_dev */ 839 if (ret && pending_snapshot->snap) 840 pending_snapshot->snap->anon_dev = 0; 841 btrfs_put_root(pending_snapshot->snap); 842 btrfs_block_rsv_release(fs_info, block_rsv, (u64)-1, NULL); 843 if (qgroup_reserved) 844 btrfs_qgroup_free_meta_prealloc(root, qgroup_reserved); 845 free_pending: 846 if (pending_snapshot->anon_dev) 847 free_anon_bdev(pending_snapshot->anon_dev); 848 kfree(pending_snapshot->root_item); 849 btrfs_free_path(pending_snapshot->path); 850 851 return ret; 852 } 853 854 /* 855 * Create a new subvolume below @parent. This is largely modeled after 856 * sys_mkdirat and vfs_mkdir, but we only do a single component lookup 857 * inside this filesystem so it's quite a bit simpler. 858 */ 859 static noinline int btrfs_mksubvol(struct dentry *parent, 860 struct mnt_idmap *idmap, 861 struct qstr *qname, struct btrfs_root *snap_src, 862 bool readonly, 863 struct btrfs_qgroup_inherit *inherit) 864 { 865 struct inode *dir = d_inode(parent); 866 struct btrfs_fs_info *fs_info = inode_to_fs_info(dir); 867 struct dentry *dentry; 868 struct fscrypt_str name_str = FSTR_INIT((char *)qname->name, qname->len); 869 int ret; 870 871 dentry = start_creating_killable(idmap, parent, qname); 872 if (IS_ERR(dentry)) 873 return PTR_ERR(dentry); 874 875 ret = may_create_dentry(idmap, dir, dentry); 876 if (ret) 877 goto out_dput; 878 879 /* 880 * even if this name doesn't exist, we may get hash collisions. 881 * check for them now when we can safely fail 882 */ 883 ret = btrfs_check_dir_item_collision(BTRFS_I(dir)->root, dir->i_ino, &name_str); 884 if (ret) 885 goto out_dput; 886 887 down_read(&fs_info->subvol_sem); 888 889 if (btrfs_root_refs(&BTRFS_I(dir)->root->root_item) == 0) 890 goto out_up_read; 891 892 if (snap_src) 893 ret = create_snapshot(snap_src, dir, dentry, readonly, inherit); 894 else 895 ret = create_subvol(idmap, dir, dentry, inherit); 896 897 if (!ret) 898 fsnotify_mkdir(dir, dentry); 899 out_up_read: 900 up_read(&fs_info->subvol_sem); 901 out_dput: 902 end_creating(dentry); 903 return ret; 904 } 905 906 static noinline int btrfs_mksnapshot(struct dentry *parent, 907 struct mnt_idmap *idmap, 908 struct qstr *qname, 909 struct btrfs_root *root, 910 bool readonly, 911 struct btrfs_qgroup_inherit *inherit) 912 { 913 int ret; 914 915 /* 916 * Force new buffered writes to reserve space even when NOCOW is 917 * possible. This is to avoid later writeback (running delalloc) to 918 * fallback to COW mode and unexpectedly fail with ENOSPC. 919 */ 920 btrfs_drew_read_lock(&root->snapshot_lock); 921 922 ret = btrfs_start_delalloc_snapshot(root, false); 923 if (ret) 924 goto out; 925 926 /* 927 * All previous writes have started writeback in NOCOW mode, so now 928 * we force future writes to fallback to COW mode during snapshot 929 * creation. 930 */ 931 atomic_inc(&root->snapshot_force_cow); 932 933 btrfs_wait_ordered_extents(root, U64_MAX, NULL); 934 935 ret = btrfs_mksubvol(parent, idmap, qname, root, readonly, inherit); 936 937 atomic_dec(&root->snapshot_force_cow); 938 out: 939 btrfs_drew_read_unlock(&root->snapshot_lock); 940 return ret; 941 } 942 943 /* 944 * Try to start exclusive operation @type or cancel it if it's running. 945 * 946 * Return: 947 * 0 - normal mode, newly claimed op started 948 * >0 - normal mode, something else is running, 949 * return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS to user space 950 * ECANCELED - cancel mode, successful cancel 951 * ENOTCONN - cancel mode, operation not running anymore 952 */ 953 static int exclop_start_or_cancel_reloc(struct btrfs_fs_info *fs_info, 954 enum btrfs_exclusive_operation type, bool cancel) 955 { 956 if (!cancel) { 957 /* Start normal op */ 958 if (!btrfs_exclop_start(fs_info, type)) 959 return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS; 960 /* Exclusive operation is now claimed */ 961 return 0; 962 } 963 964 /* Cancel running op */ 965 if (btrfs_exclop_start_try_lock(fs_info, type)) { 966 /* 967 * This blocks any exclop finish from setting it to NONE, so we 968 * request cancellation. Either it runs and we will wait for it, 969 * or it has finished and no waiting will happen. 970 */ 971 atomic_inc(&fs_info->reloc_cancel_req); 972 btrfs_exclop_start_unlock(fs_info); 973 974 if (test_bit(BTRFS_FS_RELOC_RUNNING, &fs_info->flags)) 975 wait_on_bit(&fs_info->flags, BTRFS_FS_RELOC_RUNNING, 976 TASK_INTERRUPTIBLE); 977 978 return -ECANCELED; 979 } 980 981 /* Something else is running or none */ 982 return -ENOTCONN; 983 } 984 985 static noinline int btrfs_ioctl_resize(struct file *file, 986 void __user *arg) 987 { 988 BTRFS_DEV_LOOKUP_ARGS(args); 989 struct btrfs_root *root = BTRFS_I(file_inode(file))->root; 990 struct btrfs_fs_info *fs_info = root->fs_info; 991 u64 new_size; 992 u64 old_size; 993 u64 devid = 1; 994 struct btrfs_ioctl_vol_args AUTO_KFREE(vol_args); 995 struct btrfs_device *device = NULL; 996 char *sizestr; 997 char *devstr = NULL; 998 int ret = 0; 999 int mod = 0; 1000 bool cancel; 1001 1002 if (!capable(CAP_SYS_ADMIN)) 1003 return -EPERM; 1004 1005 ret = mnt_want_write_file(file); 1006 if (ret) 1007 return ret; 1008 1009 /* 1010 * Read the arguments before checking exclusivity to be able to 1011 * distinguish regular resize and cancel 1012 */ 1013 vol_args = memdup_user(arg, sizeof(*vol_args)); 1014 if (IS_ERR(vol_args)) { 1015 ret = PTR_ERR(vol_args); 1016 goto out_drop; 1017 } 1018 ret = btrfs_check_ioctl_vol_args_path(vol_args); 1019 if (ret < 0) 1020 goto out_drop; 1021 1022 sizestr = vol_args->name; 1023 cancel = (strcmp("cancel", sizestr) == 0); 1024 ret = exclop_start_or_cancel_reloc(fs_info, BTRFS_EXCLOP_RESIZE, cancel); 1025 if (ret) 1026 goto out_drop; 1027 /* Exclusive operation is now claimed */ 1028 1029 devstr = strchr(sizestr, ':'); 1030 if (devstr) { 1031 sizestr = devstr + 1; 1032 *devstr = '\0'; 1033 devstr = vol_args->name; 1034 ret = kstrtoull(devstr, 10, &devid); 1035 if (ret) 1036 goto out_finish; 1037 if (!devid) { 1038 ret = -EINVAL; 1039 goto out_finish; 1040 } 1041 btrfs_info(fs_info, "resizing devid %llu", devid); 1042 } 1043 1044 args.devid = devid; 1045 device = btrfs_find_device(fs_info->fs_devices, &args); 1046 if (!device) { 1047 btrfs_info(fs_info, "resizer unable to find device %llu", 1048 devid); 1049 ret = -ENODEV; 1050 goto out_finish; 1051 } 1052 1053 if (!test_bit(BTRFS_DEV_STATE_WRITEABLE, &device->dev_state)) { 1054 btrfs_info(fs_info, 1055 "resizer unable to apply on readonly device %llu", 1056 devid); 1057 ret = -EPERM; 1058 goto out_finish; 1059 } 1060 1061 if (!strcmp(sizestr, "max")) 1062 new_size = bdev_nr_bytes(device->bdev); 1063 else { 1064 char *retptr; 1065 1066 if (sizestr[0] == '-') { 1067 mod = -1; 1068 sizestr++; 1069 } else if (sizestr[0] == '+') { 1070 mod = 1; 1071 sizestr++; 1072 } 1073 new_size = memparse(sizestr, &retptr); 1074 if (*retptr != '\0' || new_size == 0) { 1075 ret = -EINVAL; 1076 goto out_finish; 1077 } 1078 } 1079 1080 if (test_bit(BTRFS_DEV_STATE_REPLACE_TGT, &device->dev_state)) { 1081 ret = -EPERM; 1082 goto out_finish; 1083 } 1084 1085 old_size = btrfs_device_get_total_bytes(device); 1086 1087 if (mod < 0) { 1088 if (new_size > old_size) { 1089 ret = -EINVAL; 1090 goto out_finish; 1091 } 1092 new_size = old_size - new_size; 1093 } else if (mod > 0) { 1094 if (new_size > ULLONG_MAX - old_size) { 1095 ret = -ERANGE; 1096 goto out_finish; 1097 } 1098 new_size = old_size + new_size; 1099 } 1100 1101 if (new_size < SZ_256M) { 1102 ret = -EINVAL; 1103 goto out_finish; 1104 } 1105 if (new_size > bdev_nr_bytes(device->bdev)) { 1106 ret = -EFBIG; 1107 goto out_finish; 1108 } 1109 1110 new_size = round_down(new_size, fs_info->sectorsize); 1111 1112 if (new_size > old_size) { 1113 struct btrfs_trans_handle *trans; 1114 1115 trans = btrfs_start_transaction(root, 0); 1116 if (IS_ERR(trans)) { 1117 ret = PTR_ERR(trans); 1118 goto out_finish; 1119 } 1120 ret = btrfs_grow_device(trans, device, new_size); 1121 btrfs_commit_transaction(trans); 1122 } else if (new_size < old_size) { 1123 ret = btrfs_shrink_device(device, new_size); 1124 } /* equal, nothing need to do */ 1125 1126 if (ret == 0 && new_size != old_size) 1127 btrfs_info(fs_info, 1128 "resize device %s (devid %llu) from %llu to %llu", 1129 btrfs_dev_name(device), device->devid, 1130 old_size, new_size); 1131 out_finish: 1132 btrfs_exclop_finish(fs_info); 1133 out_drop: 1134 mnt_drop_write_file(file); 1135 return ret; 1136 } 1137 1138 static noinline int __btrfs_ioctl_snap_create(struct file *file, 1139 struct mnt_idmap *idmap, 1140 const char *name, unsigned long fd, bool subvol, 1141 bool readonly, 1142 struct btrfs_qgroup_inherit *inherit) 1143 { 1144 int ret; 1145 struct qstr qname = QSTR(name); 1146 1147 if (!S_ISDIR(file_inode(file)->i_mode)) 1148 return -ENOTDIR; 1149 1150 ret = mnt_want_write_file(file); 1151 if (ret) 1152 return ret; 1153 1154 if (strchr(name, '/')) { 1155 ret = -EINVAL; 1156 goto out_drop_write; 1157 } 1158 1159 if (qname.name[0] == '.' && 1160 (qname.len == 1 || (qname.name[1] == '.' && qname.len == 2))) { 1161 ret = -EEXIST; 1162 goto out_drop_write; 1163 } 1164 1165 if (subvol) { 1166 ret = btrfs_mksubvol(file_dentry(file), idmap, &qname, NULL, 1167 readonly, inherit); 1168 } else { 1169 CLASS(fd, src)(fd); 1170 struct inode *src_inode; 1171 if (fd_empty(src)) { 1172 ret = -EINVAL; 1173 goto out_drop_write; 1174 } 1175 1176 src_inode = file_inode(fd_file(src)); 1177 if (src_inode->i_sb != file_inode(file)->i_sb) { 1178 btrfs_info(BTRFS_I(file_inode(file))->root->fs_info, 1179 "Snapshot src from another FS"); 1180 ret = -EXDEV; 1181 } else if (!inode_owner_or_capable(idmap, src_inode)) { 1182 /* 1183 * Subvolume creation is not restricted, but snapshots 1184 * are limited to own subvolumes only 1185 */ 1186 ret = -EPERM; 1187 } else if (btrfs_ino(BTRFS_I(src_inode)) != BTRFS_FIRST_FREE_OBJECTID) { 1188 /* 1189 * Snapshots must be made with the src_inode referring 1190 * to the subvolume inode, otherwise the permission 1191 * checking above is useless because we may have 1192 * permission on a lower directory but not the subvol 1193 * itself. 1194 */ 1195 ret = -EINVAL; 1196 } else { 1197 ret = btrfs_mksnapshot(file_dentry(file), idmap, &qname, 1198 BTRFS_I(src_inode)->root, 1199 readonly, inherit); 1200 } 1201 } 1202 out_drop_write: 1203 mnt_drop_write_file(file); 1204 return ret; 1205 } 1206 1207 static noinline int btrfs_ioctl_snap_create(struct file *file, 1208 void __user *arg, bool subvol) 1209 { 1210 struct btrfs_ioctl_vol_args AUTO_KFREE(vol_args); 1211 int ret; 1212 1213 if (!S_ISDIR(file_inode(file)->i_mode)) 1214 return -ENOTDIR; 1215 1216 vol_args = memdup_user(arg, sizeof(*vol_args)); 1217 if (IS_ERR(vol_args)) 1218 return PTR_ERR(vol_args); 1219 ret = btrfs_check_ioctl_vol_args_path(vol_args); 1220 if (ret < 0) 1221 return ret; 1222 1223 return __btrfs_ioctl_snap_create(file, file_mnt_idmap(file), 1224 vol_args->name, vol_args->fd, subvol, 1225 false, NULL); 1226 } 1227 1228 static noinline int btrfs_ioctl_snap_create_v2(struct file *file, 1229 void __user *arg, bool subvol) 1230 { 1231 struct btrfs_ioctl_vol_args_v2 AUTO_KFREE(vol_args); 1232 struct btrfs_qgroup_inherit AUTO_KFREE(inherit); 1233 int ret; 1234 bool readonly = false; 1235 1236 if (!S_ISDIR(file_inode(file)->i_mode)) 1237 return -ENOTDIR; 1238 1239 vol_args = memdup_user(arg, sizeof(*vol_args)); 1240 if (IS_ERR(vol_args)) 1241 return PTR_ERR(vol_args); 1242 ret = btrfs_check_ioctl_vol_args2_subvol_name(vol_args); 1243 if (ret < 0) 1244 return ret; 1245 1246 if (vol_args->flags & ~BTRFS_SUBVOL_CREATE_ARGS_MASK) 1247 return -EOPNOTSUPP; 1248 1249 if (vol_args->flags & BTRFS_SUBVOL_RDONLY) 1250 readonly = true; 1251 if (vol_args->flags & BTRFS_SUBVOL_QGROUP_INHERIT) { 1252 struct btrfs_fs_info *fs_info = inode_to_fs_info(file_inode(file)); 1253 1254 if (vol_args->size < sizeof(*inherit) || vol_args->size > PAGE_SIZE) 1255 return -EINVAL; 1256 1257 inherit = memdup_user(vol_args->qgroup_inherit, vol_args->size); 1258 if (IS_ERR(inherit)) { 1259 return PTR_ERR(inherit); 1260 } 1261 1262 ret = btrfs_qgroup_check_inherit(fs_info, inherit, vol_args->size); 1263 if (ret < 0) 1264 return ret; 1265 } 1266 1267 return __btrfs_ioctl_snap_create(file, file_mnt_idmap(file), 1268 vol_args->name, vol_args->fd, subvol, 1269 readonly, inherit); 1270 } 1271 1272 static noinline int btrfs_ioctl_subvol_getflags(struct btrfs_inode *inode, 1273 void __user *arg) 1274 { 1275 struct btrfs_root *root = inode->root; 1276 struct btrfs_fs_info *fs_info = root->fs_info; 1277 int ret = 0; 1278 u64 flags = 0; 1279 1280 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) 1281 return -EINVAL; 1282 1283 down_read(&fs_info->subvol_sem); 1284 if (btrfs_root_readonly(root)) 1285 flags |= BTRFS_SUBVOL_RDONLY; 1286 up_read(&fs_info->subvol_sem); 1287 1288 if (copy_to_user(arg, &flags, sizeof(flags))) 1289 ret = -EFAULT; 1290 1291 return ret; 1292 } 1293 1294 static noinline int btrfs_ioctl_subvol_setflags(struct file *file, 1295 void __user *arg) 1296 { 1297 struct inode *inode = file_inode(file); 1298 struct btrfs_fs_info *fs_info = inode_to_fs_info(inode); 1299 struct btrfs_root *root = BTRFS_I(inode)->root; 1300 struct btrfs_trans_handle *trans; 1301 u64 root_flags; 1302 u64 flags; 1303 int ret; 1304 1305 if (!inode_owner_or_capable(file_mnt_idmap(file), inode)) 1306 return -EPERM; 1307 1308 ret = mnt_want_write_file(file); 1309 if (ret) 1310 return ret; 1311 1312 if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) { 1313 ret = -EINVAL; 1314 goto out_drop_write; 1315 } 1316 1317 if (copy_from_user(&flags, arg, sizeof(flags))) { 1318 ret = -EFAULT; 1319 goto out_drop_write; 1320 } 1321 1322 if (flags & ~BTRFS_SUBVOL_RDONLY) { 1323 ret = -EOPNOTSUPP; 1324 goto out_drop_write; 1325 } 1326 1327 down_write(&fs_info->subvol_sem); 1328 1329 /* nothing to do */ 1330 if (!!(flags & BTRFS_SUBVOL_RDONLY) == btrfs_root_readonly(root)) 1331 goto out_drop_sem; 1332 1333 root_flags = btrfs_root_flags(&root->root_item); 1334 if (flags & BTRFS_SUBVOL_RDONLY) { 1335 btrfs_set_root_flags(&root->root_item, 1336 root_flags | BTRFS_ROOT_SUBVOL_RDONLY); 1337 } else { 1338 /* 1339 * Block RO -> RW transition if this subvolume is involved in 1340 * send 1341 */ 1342 spin_lock(&root->root_item_lock); 1343 if (root->send_in_progress == 0) { 1344 btrfs_set_root_flags(&root->root_item, 1345 root_flags & ~BTRFS_ROOT_SUBVOL_RDONLY); 1346 spin_unlock(&root->root_item_lock); 1347 } else { 1348 spin_unlock(&root->root_item_lock); 1349 btrfs_warn(fs_info, 1350 "Attempt to set subvolume %llu read-write during send", 1351 btrfs_root_id(root)); 1352 ret = -EPERM; 1353 goto out_drop_sem; 1354 } 1355 } 1356 1357 trans = btrfs_start_transaction(root, 1); 1358 if (IS_ERR(trans)) { 1359 ret = PTR_ERR(trans); 1360 goto out_reset; 1361 } 1362 1363 ret = btrfs_update_root(trans, fs_info->tree_root, 1364 &root->root_key, &root->root_item); 1365 if (ret < 0) { 1366 btrfs_end_transaction(trans); 1367 goto out_reset; 1368 } 1369 1370 ret = btrfs_commit_transaction(trans); 1371 1372 out_reset: 1373 if (ret) 1374 btrfs_set_root_flags(&root->root_item, root_flags); 1375 out_drop_sem: 1376 up_write(&fs_info->subvol_sem); 1377 out_drop_write: 1378 mnt_drop_write_file(file); 1379 return ret; 1380 } 1381 1382 static noinline bool key_in_sk(const struct btrfs_key *key, 1383 const struct btrfs_ioctl_search_key *sk) 1384 { 1385 struct btrfs_key test; 1386 int ret; 1387 1388 test.objectid = sk->min_objectid; 1389 test.type = sk->min_type; 1390 test.offset = sk->min_offset; 1391 1392 ret = btrfs_comp_cpu_keys(key, &test); 1393 if (ret < 0) 1394 return false; 1395 1396 test.objectid = sk->max_objectid; 1397 test.type = sk->max_type; 1398 test.offset = sk->max_offset; 1399 1400 ret = btrfs_comp_cpu_keys(key, &test); 1401 if (ret > 0) 1402 return false; 1403 return true; 1404 } 1405 1406 static noinline int copy_to_sk(struct btrfs_path *path, 1407 struct btrfs_key *key, 1408 const struct btrfs_ioctl_search_key *sk, 1409 u64 *buf_size, 1410 char __user *ubuf, 1411 unsigned long *sk_offset, 1412 int *num_found) 1413 { 1414 u64 found_transid; 1415 struct extent_buffer *leaf; 1416 struct btrfs_ioctl_search_header sh; 1417 struct btrfs_key test; 1418 unsigned long item_off; 1419 unsigned long item_len; 1420 int nritems; 1421 int i; 1422 int slot; 1423 int ret = 0; 1424 1425 leaf = path->nodes[0]; 1426 slot = path->slots[0]; 1427 nritems = btrfs_header_nritems(leaf); 1428 1429 if (btrfs_header_generation(leaf) > sk->max_transid) { 1430 i = nritems; 1431 goto advance_key; 1432 } 1433 found_transid = btrfs_header_generation(leaf); 1434 1435 for (i = slot; i < nritems; i++) { 1436 item_off = btrfs_item_ptr_offset(leaf, i); 1437 item_len = btrfs_item_size(leaf, i); 1438 1439 btrfs_item_key_to_cpu(leaf, key, i); 1440 if (!key_in_sk(key, sk)) 1441 continue; 1442 1443 if (sizeof(sh) + item_len > *buf_size) { 1444 if (*num_found) 1445 return 1; 1446 1447 /* 1448 * return one empty item back for v1, which does not 1449 * handle -EOVERFLOW 1450 */ 1451 1452 *buf_size = sizeof(sh) + item_len; 1453 item_len = 0; 1454 ret = -EOVERFLOW; 1455 } 1456 1457 if (sizeof(sh) + item_len + *sk_offset > *buf_size) 1458 return 1; 1459 1460 sh.objectid = key->objectid; 1461 sh.type = key->type; 1462 sh.offset = key->offset; 1463 sh.len = item_len; 1464 sh.transid = found_transid; 1465 1466 /* 1467 * Copy search result header. If we fault then loop again so we 1468 * can fault in the pages and -EFAULT there if there's a 1469 * problem. Otherwise we'll fault and then copy the buffer in 1470 * properly this next time through 1471 */ 1472 if (copy_to_user_nofault(ubuf + *sk_offset, &sh, sizeof(sh))) 1473 return 0; 1474 1475 *sk_offset += sizeof(sh); 1476 1477 if (item_len) { 1478 char __user *up = ubuf + *sk_offset; 1479 /* 1480 * Copy the item, same behavior as above, but reset the 1481 * * sk_offset so we copy the full thing again. 1482 */ 1483 if (read_extent_buffer_to_user_nofault(leaf, up, 1484 item_off, item_len)) { 1485 *sk_offset -= sizeof(sh); 1486 return 0; 1487 } 1488 1489 *sk_offset += item_len; 1490 } 1491 (*num_found)++; 1492 1493 /* -EOVERFLOW from above. */ 1494 if (ret) 1495 return ret; 1496 1497 if (*num_found >= sk->nr_items) 1498 return 1; 1499 } 1500 advance_key: 1501 ret = 0; 1502 test.objectid = sk->max_objectid; 1503 test.type = sk->max_type; 1504 test.offset = sk->max_offset; 1505 if (btrfs_comp_cpu_keys(key, &test) >= 0) 1506 ret = 1; 1507 else if (key->offset < (u64)-1) 1508 key->offset++; 1509 else if (key->type < (u8)-1) { 1510 key->offset = 0; 1511 key->type++; 1512 } else if (key->objectid < (u64)-1) { 1513 key->offset = 0; 1514 key->type = 0; 1515 key->objectid++; 1516 } else 1517 ret = 1; 1518 1519 /* 1520 * 0: all items from this leaf copied, continue with next 1521 * 1: * more items can be copied, but unused buffer is too small 1522 * * all items were found 1523 * Either way, it will stops the loop which iterates to the next 1524 * leaf 1525 * -EOVERFLOW: item was to large for buffer 1526 * -EFAULT: could not copy extent buffer back to userspace 1527 */ 1528 return ret; 1529 } 1530 1531 static noinline int search_ioctl(struct btrfs_root *root, 1532 struct btrfs_ioctl_search_key *sk, 1533 u64 *buf_size, 1534 char __user *ubuf) 1535 { 1536 struct btrfs_fs_info *info = root->fs_info; 1537 struct btrfs_key key; 1538 BTRFS_PATH_AUTO_FREE(path); 1539 int ret; 1540 int num_found = 0; 1541 unsigned long sk_offset = 0; 1542 1543 if (*buf_size < sizeof(struct btrfs_ioctl_search_header)) { 1544 *buf_size = sizeof(struct btrfs_ioctl_search_header); 1545 return -EOVERFLOW; 1546 } 1547 1548 path = btrfs_alloc_path(); 1549 if (!path) 1550 return -ENOMEM; 1551 1552 if (sk->tree_id == 0) { 1553 /* Search the root that we got passed. */ 1554 root = btrfs_grab_root(root); 1555 } else { 1556 /* Look up the root from the arguments. */ 1557 root = btrfs_get_fs_root(info, sk->tree_id, true); 1558 if (IS_ERR(root)) 1559 return PTR_ERR(root); 1560 } 1561 1562 key.objectid = sk->min_objectid; 1563 key.type = sk->min_type; 1564 key.offset = sk->min_offset; 1565 1566 while (1) { 1567 /* 1568 * Ensure that the whole user buffer is faulted in at sub-page 1569 * granularity, otherwise the loop may live-lock. 1570 */ 1571 if (fault_in_subpage_writeable(ubuf + sk_offset, *buf_size - sk_offset)) { 1572 ret = -EFAULT; 1573 break; 1574 } 1575 1576 ret = btrfs_search_forward(root, &key, path, sk->min_transid); 1577 if (ret) 1578 break; 1579 1580 ret = copy_to_sk(path, &key, sk, buf_size, ubuf, 1581 &sk_offset, &num_found); 1582 btrfs_release_path(path); 1583 if (ret) 1584 break; 1585 1586 } 1587 /* Normalize return values from btrfs_search_forward() and copy_to_sk(). */ 1588 if (ret > 0) 1589 ret = 0; 1590 1591 sk->nr_items = num_found; 1592 btrfs_put_root(root); 1593 return ret; 1594 } 1595 1596 static noinline int btrfs_ioctl_tree_search(struct btrfs_root *root, 1597 void __user *argp) 1598 { 1599 struct btrfs_ioctl_search_args __user *uargs = argp; 1600 struct btrfs_ioctl_search_key sk; 1601 int ret; 1602 u64 buf_size; 1603 1604 if (!capable(CAP_SYS_ADMIN)) 1605 return -EPERM; 1606 1607 if (copy_from_user(&sk, &uargs->key, sizeof(sk))) 1608 return -EFAULT; 1609 1610 buf_size = sizeof(uargs->buf); 1611 1612 ret = search_ioctl(root, &sk, &buf_size, uargs->buf); 1613 1614 /* 1615 * In the origin implementation an overflow is handled by returning a 1616 * search header with a len of zero, so reset ret. 1617 */ 1618 if (ret == -EOVERFLOW) 1619 ret = 0; 1620 1621 if (ret == 0 && copy_to_user(&uargs->key, &sk, sizeof(sk))) 1622 ret = -EFAULT; 1623 return ret; 1624 } 1625 1626 static noinline int btrfs_ioctl_tree_search_v2(struct btrfs_root *root, 1627 void __user *argp) 1628 { 1629 struct btrfs_ioctl_search_args_v2 __user *uarg = argp; 1630 struct btrfs_ioctl_search_args_v2 args; 1631 int ret; 1632 u64 buf_size; 1633 const u64 buf_limit = SZ_16M; 1634 1635 if (!capable(CAP_SYS_ADMIN)) 1636 return -EPERM; 1637 1638 /* copy search header and buffer size */ 1639 if (copy_from_user(&args, uarg, sizeof(args))) 1640 return -EFAULT; 1641 1642 buf_size = args.buf_size; 1643 1644 /* limit result size to 16MB */ 1645 if (buf_size > buf_limit) 1646 buf_size = buf_limit; 1647 1648 ret = search_ioctl(root, &args.key, &buf_size, 1649 (char __user *)(&uarg->buf[0])); 1650 if (ret == 0 && copy_to_user(&uarg->key, &args.key, sizeof(args.key))) 1651 ret = -EFAULT; 1652 else if (ret == -EOVERFLOW && 1653 copy_to_user(&uarg->buf_size, &buf_size, sizeof(buf_size))) 1654 ret = -EFAULT; 1655 1656 return ret; 1657 } 1658 1659 /* 1660 * Search INODE_REFs to identify path name of 'dirid' directory 1661 * in a 'tree_id' tree. and sets path name to 'name'. 1662 */ 1663 static noinline int btrfs_search_path_in_tree(struct btrfs_fs_info *info, 1664 u64 tree_id, u64 dirid, char *name) 1665 { 1666 struct btrfs_root *root; 1667 struct btrfs_key key; 1668 char *ptr; 1669 int ret = -1; 1670 int slot; 1671 int len; 1672 int total_len = 0; 1673 struct btrfs_inode_ref *iref; 1674 struct extent_buffer *l; 1675 BTRFS_PATH_AUTO_FREE(path); 1676 1677 if (dirid == BTRFS_FIRST_FREE_OBJECTID) { 1678 name[0]='\0'; 1679 return 0; 1680 } 1681 1682 path = btrfs_alloc_path(); 1683 if (!path) 1684 return -ENOMEM; 1685 1686 ptr = &name[BTRFS_INO_LOOKUP_PATH_MAX - 1]; 1687 1688 root = btrfs_get_fs_root(info, tree_id, true); 1689 if (IS_ERR(root)) { 1690 ret = PTR_ERR(root); 1691 root = NULL; 1692 goto out; 1693 } 1694 1695 key.objectid = dirid; 1696 key.type = BTRFS_INODE_REF_KEY; 1697 key.offset = (u64)-1; 1698 1699 while (1) { 1700 ret = btrfs_search_backwards(root, &key, path); 1701 if (ret < 0) 1702 goto out; 1703 else if (ret > 0) { 1704 ret = -ENOENT; 1705 goto out; 1706 } 1707 1708 l = path->nodes[0]; 1709 slot = path->slots[0]; 1710 1711 iref = btrfs_item_ptr(l, slot, struct btrfs_inode_ref); 1712 len = btrfs_inode_ref_name_len(l, iref); 1713 ptr -= len + 1; 1714 total_len += len + 1; 1715 if (ptr < name) { 1716 ret = -ENAMETOOLONG; 1717 goto out; 1718 } 1719 1720 *(ptr + len) = '/'; 1721 read_extent_buffer(l, ptr, (unsigned long)(iref + 1), len); 1722 1723 if (key.offset == BTRFS_FIRST_FREE_OBJECTID) 1724 break; 1725 1726 btrfs_release_path(path); 1727 key.objectid = key.offset; 1728 key.offset = (u64)-1; 1729 dirid = key.objectid; 1730 } 1731 memmove(name, ptr, total_len); 1732 name[total_len] = '\0'; 1733 ret = 0; 1734 out: 1735 btrfs_put_root(root); 1736 return ret; 1737 } 1738 1739 static int btrfs_search_path_in_tree_user(struct mnt_idmap *idmap, 1740 struct inode *inode, 1741 struct btrfs_ioctl_ino_lookup_user_args *args) 1742 { 1743 struct btrfs_fs_info *fs_info = BTRFS_I(inode)->root->fs_info; 1744 u64 upper_limit = btrfs_ino(BTRFS_I(inode)); 1745 u64 treeid = btrfs_root_id(BTRFS_I(inode)->root); 1746 u64 dirid = args->dirid; 1747 unsigned long item_off; 1748 unsigned long item_len; 1749 struct btrfs_inode_ref *iref; 1750 struct btrfs_root_ref *rref; 1751 struct btrfs_root *root = NULL; 1752 BTRFS_PATH_AUTO_FREE(path); 1753 struct btrfs_key key; 1754 struct extent_buffer *leaf; 1755 char *ptr; 1756 int slot; 1757 int len; 1758 int total_len = 0; 1759 int ret; 1760 1761 path = btrfs_alloc_path(); 1762 if (!path) 1763 return -ENOMEM; 1764 1765 /* 1766 * If the bottom subvolume does not exist directly under upper_limit, 1767 * construct the path in from the bottom up. 1768 */ 1769 if (dirid != upper_limit) { 1770 ptr = &args->path[BTRFS_INO_LOOKUP_USER_PATH_MAX - 1]; 1771 1772 root = btrfs_get_fs_root(fs_info, treeid, true); 1773 if (IS_ERR(root)) 1774 return PTR_ERR(root); 1775 1776 key.objectid = dirid; 1777 key.type = BTRFS_INODE_REF_KEY; 1778 key.offset = (u64)-1; 1779 while (1) { 1780 struct btrfs_inode *temp_inode; 1781 1782 ret = btrfs_search_backwards(root, &key, path); 1783 if (ret < 0) 1784 goto out_put; 1785 else if (ret > 0) { 1786 ret = -ENOENT; 1787 goto out_put; 1788 } 1789 1790 leaf = path->nodes[0]; 1791 slot = path->slots[0]; 1792 1793 iref = btrfs_item_ptr(leaf, slot, struct btrfs_inode_ref); 1794 len = btrfs_inode_ref_name_len(leaf, iref); 1795 ptr -= len + 1; 1796 total_len += len + 1; 1797 if (ptr < args->path) { 1798 ret = -ENAMETOOLONG; 1799 goto out_put; 1800 } 1801 1802 *(ptr + len) = '/'; 1803 read_extent_buffer(leaf, ptr, 1804 (unsigned long)(iref + 1), len); 1805 1806 /* 1807 * We don't need the path anymore, so release it and 1808 * avoid deadlocks and lockdep warnings in case 1809 * btrfs_iget() needs to lookup the inode from its root 1810 * btree and lock the same leaf. 1811 */ 1812 btrfs_release_path(path); 1813 temp_inode = btrfs_iget(key.offset, root); 1814 if (IS_ERR(temp_inode)) { 1815 ret = PTR_ERR(temp_inode); 1816 goto out_put; 1817 } 1818 /* Check the read+exec permission of this directory. */ 1819 ret = inode_permission(idmap, &temp_inode->vfs_inode, 1820 MAY_READ | MAY_EXEC); 1821 iput(&temp_inode->vfs_inode); 1822 if (ret) 1823 goto out_put; 1824 1825 if (key.offset == upper_limit) 1826 break; 1827 if (key.objectid == BTRFS_FIRST_FREE_OBJECTID) { 1828 ret = -EACCES; 1829 goto out_put; 1830 } 1831 1832 key.objectid = key.offset; 1833 key.offset = (u64)-1; 1834 dirid = key.objectid; 1835 } 1836 1837 memmove(args->path, ptr, total_len); 1838 args->path[total_len] = '\0'; 1839 btrfs_put_root(root); 1840 root = NULL; 1841 btrfs_release_path(path); 1842 } 1843 1844 /* Get the bottom subvolume's name from ROOT_REF */ 1845 key.objectid = treeid; 1846 key.type = BTRFS_ROOT_REF_KEY; 1847 key.offset = args->treeid; 1848 ret = btrfs_search_slot(NULL, fs_info->tree_root, &key, path, 0, 0); 1849 if (ret < 0) 1850 return ret; 1851 else if (ret > 0) 1852 return -ENOENT; 1853 1854 leaf = path->nodes[0]; 1855 slot = path->slots[0]; 1856 btrfs_item_key_to_cpu(leaf, &key, slot); 1857 1858 item_off = btrfs_item_ptr_offset(leaf, slot); 1859 item_len = btrfs_item_size(leaf, slot); 1860 /* Check if dirid in ROOT_REF corresponds to passed dirid */ 1861 rref = btrfs_item_ptr(leaf, slot, struct btrfs_root_ref); 1862 if (args->dirid != btrfs_root_ref_dirid(leaf, rref)) 1863 return -EINVAL; 1864 1865 /* Copy subvolume's name */ 1866 item_off += sizeof(struct btrfs_root_ref); 1867 item_len -= sizeof(struct btrfs_root_ref); 1868 read_extent_buffer(leaf, args->name, item_off, item_len); 1869 args->name[item_len] = 0; 1870 1871 out_put: 1872 btrfs_put_root(root); 1873 1874 return ret; 1875 } 1876 1877 static noinline int btrfs_ioctl_ino_lookup(struct btrfs_root *root, 1878 void __user *argp) 1879 { 1880 struct btrfs_ioctl_ino_lookup_args AUTO_KFREE(args); 1881 int ret = 0; 1882 1883 args = memdup_user(argp, sizeof(*args)); 1884 if (IS_ERR(args)) 1885 return PTR_ERR(args); 1886 1887 /* 1888 * Unprivileged query to obtain the containing subvolume root id. The 1889 * path is reset so it's consistent with btrfs_search_path_in_tree. 1890 */ 1891 if (args->treeid == 0) 1892 args->treeid = btrfs_root_id(root); 1893 1894 if (args->objectid == BTRFS_FIRST_FREE_OBJECTID) { 1895 args->name[0] = 0; 1896 goto out; 1897 } 1898 1899 if (!capable(CAP_SYS_ADMIN)) { 1900 ret = -EPERM; 1901 goto out; 1902 } 1903 1904 ret = btrfs_search_path_in_tree(root->fs_info, 1905 args->treeid, args->objectid, 1906 args->name); 1907 1908 out: 1909 if (ret == 0 && copy_to_user(argp, args, sizeof(*args))) 1910 return -EFAULT; 1911 1912 return ret; 1913 } 1914 1915 /* 1916 * Version of ino_lookup ioctl (unprivileged) 1917 * 1918 * The main differences from ino_lookup ioctl are: 1919 * 1920 * 1. Read + Exec permission will be checked using inode_permission() during 1921 * path construction. -EACCES will be returned in case of failure. 1922 * 2. Path construction will be stopped at the inode number which corresponds 1923 * to the fd with which this ioctl is called. If constructed path does not 1924 * exist under fd's inode, -EACCES will be returned. 1925 * 3. The name of bottom subvolume is also searched and filled. 1926 */ 1927 static int btrfs_ioctl_ino_lookup_user(struct file *file, void __user *argp) 1928 { 1929 struct btrfs_ioctl_ino_lookup_user_args AUTO_KFREE(args); 1930 struct inode *inode; 1931 int ret; 1932 1933 args = memdup_user(argp, sizeof(*args)); 1934 if (IS_ERR(args)) 1935 return PTR_ERR(args); 1936 1937 inode = file_inode(file); 1938 1939 if (args->dirid == BTRFS_FIRST_FREE_OBJECTID && 1940 btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) { 1941 /* 1942 * The subvolume does not exist under fd with which this is 1943 * called 1944 */ 1945 return -EACCES; 1946 } 1947 1948 ret = btrfs_search_path_in_tree_user(file_mnt_idmap(file), inode, args); 1949 1950 if (ret == 0 && copy_to_user(argp, args, sizeof(*args))) 1951 ret = -EFAULT; 1952 1953 return ret; 1954 } 1955 1956 /* Get the subvolume information in BTRFS_ROOT_ITEM and BTRFS_ROOT_BACKREF */ 1957 static int _btrfs_ioctl_get_subvol_info(struct inode *inode, 1958 struct btrfs_ioctl_get_subvol_info_args *subvol_info) 1959 { 1960 struct btrfs_fs_info *fs_info; 1961 struct btrfs_root *root; 1962 struct btrfs_path *path; 1963 struct btrfs_key key; 1964 struct btrfs_root_item *root_item; 1965 struct btrfs_root_ref *rref; 1966 struct extent_buffer *leaf; 1967 unsigned long item_off; 1968 int slot; 1969 int ret = 0; 1970 1971 path = btrfs_alloc_path(); 1972 if (!path) 1973 return -ENOMEM; 1974 1975 fs_info = BTRFS_I(inode)->root->fs_info; 1976 1977 /* Get root_item of inode's subvolume */ 1978 key.objectid = btrfs_root_id(BTRFS_I(inode)->root); 1979 root = btrfs_get_fs_root(fs_info, key.objectid, true); 1980 if (IS_ERR(root)) { 1981 ret = PTR_ERR(root); 1982 goto out_free; 1983 } 1984 root_item = &root->root_item; 1985 1986 subvol_info->treeid = key.objectid; 1987 1988 subvol_info->generation = btrfs_root_generation(root_item); 1989 subvol_info->flags = btrfs_root_flags(root_item); 1990 1991 memcpy(subvol_info->uuid, root_item->uuid, BTRFS_UUID_SIZE); 1992 memcpy(subvol_info->parent_uuid, root_item->parent_uuid, 1993 BTRFS_UUID_SIZE); 1994 memcpy(subvol_info->received_uuid, root_item->received_uuid, 1995 BTRFS_UUID_SIZE); 1996 1997 subvol_info->ctransid = btrfs_root_ctransid(root_item); 1998 subvol_info->ctime.sec = btrfs_stack_timespec_sec(&root_item->ctime); 1999 subvol_info->ctime.nsec = btrfs_stack_timespec_nsec(&root_item->ctime); 2000 2001 subvol_info->otransid = btrfs_root_otransid(root_item); 2002 subvol_info->otime.sec = btrfs_stack_timespec_sec(&root_item->otime); 2003 subvol_info->otime.nsec = btrfs_stack_timespec_nsec(&root_item->otime); 2004 2005 subvol_info->stransid = btrfs_root_stransid(root_item); 2006 subvol_info->stime.sec = btrfs_stack_timespec_sec(&root_item->stime); 2007 subvol_info->stime.nsec = btrfs_stack_timespec_nsec(&root_item->stime); 2008 2009 subvol_info->rtransid = btrfs_root_rtransid(root_item); 2010 subvol_info->rtime.sec = btrfs_stack_timespec_sec(&root_item->rtime); 2011 subvol_info->rtime.nsec = btrfs_stack_timespec_nsec(&root_item->rtime); 2012 2013 if (key.objectid != BTRFS_FS_TREE_OBJECTID) { 2014 /* Search root tree for ROOT_BACKREF of this subvolume */ 2015 key.type = BTRFS_ROOT_BACKREF_KEY; 2016 key.offset = 0; 2017 ret = btrfs_search_slot(NULL, fs_info->tree_root, &key, path, 0, 0); 2018 if (ret < 0) { 2019 goto out; 2020 } else if (path->slots[0] >= 2021 btrfs_header_nritems(path->nodes[0])) { 2022 ret = btrfs_next_leaf(fs_info->tree_root, path); 2023 if (ret < 0) { 2024 goto out; 2025 } else if (unlikely(ret > 0)) { 2026 ret = -EUCLEAN; 2027 goto out; 2028 } 2029 } 2030 2031 leaf = path->nodes[0]; 2032 slot = path->slots[0]; 2033 btrfs_item_key_to_cpu(leaf, &key, slot); 2034 if (key.objectid == subvol_info->treeid && 2035 key.type == BTRFS_ROOT_BACKREF_KEY) { 2036 u16 name_len; 2037 2038 subvol_info->parent_id = key.offset; 2039 2040 rref = btrfs_item_ptr(leaf, slot, struct btrfs_root_ref); 2041 name_len = btrfs_root_ref_name_len(leaf, rref); 2042 subvol_info->dirid = btrfs_root_ref_dirid(leaf, rref); 2043 2044 item_off = btrfs_item_ptr_offset(leaf, slot) + sizeof(*rref); 2045 read_extent_buffer(leaf, subvol_info->name, 2046 item_off, name_len); 2047 } else { 2048 ret = -ENOENT; 2049 goto out; 2050 } 2051 ret = 0; 2052 } 2053 2054 out: 2055 btrfs_put_root(root); 2056 out_free: 2057 btrfs_free_path(path); 2058 return ret; 2059 } 2060 2061 #ifdef CONFIG_64BIT 2062 static int btrfs_ioctl_get_subvol_info_32(struct inode *inode, void __user *argp) 2063 { 2064 struct btrfs_ioctl_get_subvol_info_args AUTO_KFREE(subvol_info); 2065 struct btrfs_ioctl_get_subvol_info_args_32 AUTO_KFREE(subvol_info_32); 2066 int ret; 2067 2068 subvol_info = kzalloc_obj(*subvol_info); 2069 if (!subvol_info) 2070 return -ENOMEM; 2071 2072 subvol_info_32 = kzalloc_obj(*subvol_info_32); 2073 if (!subvol_info_32) 2074 return -ENOMEM; 2075 2076 ret = _btrfs_ioctl_get_subvol_info(inode, subvol_info); 2077 if (ret) 2078 return ret; 2079 2080 subvol_info_32->treeid = subvol_info->treeid; 2081 memcpy(subvol_info_32->name, subvol_info->name, sizeof(subvol_info_32->name)); 2082 subvol_info_32->parent_id = subvol_info->parent_id; 2083 subvol_info_32->dirid = subvol_info->dirid; 2084 subvol_info_32->generation = subvol_info->generation; 2085 subvol_info_32->flags = subvol_info->flags; 2086 memcpy(subvol_info_32->uuid, subvol_info->uuid, BTRFS_UUID_SIZE); 2087 memcpy(subvol_info_32->parent_uuid, subvol_info->parent_uuid, BTRFS_UUID_SIZE); 2088 memcpy(subvol_info_32->received_uuid, subvol_info->received_uuid, BTRFS_UUID_SIZE); 2089 subvol_info_32->ctransid = subvol_info->ctransid; 2090 subvol_info_32->otransid = subvol_info->otransid; 2091 subvol_info_32->stransid = subvol_info->stransid; 2092 subvol_info_32->rtransid = subvol_info->rtransid; 2093 subvol_info_32->ctime.sec = subvol_info->ctime.sec; 2094 subvol_info_32->ctime.nsec = subvol_info->ctime.nsec; 2095 subvol_info_32->otime.sec = subvol_info->otime.sec; 2096 subvol_info_32->otime.nsec = subvol_info->otime.nsec; 2097 subvol_info_32->stime.sec = subvol_info->stime.sec; 2098 subvol_info_32->stime.nsec = subvol_info->stime.nsec; 2099 subvol_info_32->rtime.sec = subvol_info->rtime.sec; 2100 subvol_info_32->rtime.nsec = subvol_info->rtime.nsec; 2101 2102 if (copy_to_user(argp, subvol_info_32, sizeof(*subvol_info_32))) 2103 ret = -EFAULT; 2104 2105 return ret; 2106 } 2107 #endif 2108 2109 static int btrfs_ioctl_get_subvol_info(struct inode *inode, void __user *argp) 2110 { 2111 struct btrfs_ioctl_get_subvol_info_args AUTO_KFREE(subvol_info); 2112 int ret; 2113 2114 subvol_info = kzalloc_obj(*subvol_info); 2115 if (!subvol_info) 2116 return -ENOMEM; 2117 2118 ret = _btrfs_ioctl_get_subvol_info(inode, subvol_info); 2119 if (!ret && copy_to_user(argp, subvol_info, sizeof(*subvol_info))) 2120 ret = -EFAULT; 2121 2122 return ret; 2123 } 2124 2125 /* 2126 * Return ROOT_REF information of the subvolume containing this inode 2127 * except the subvolume name. 2128 */ 2129 static int btrfs_ioctl_get_subvol_rootref(struct btrfs_root *root, 2130 void __user *argp) 2131 { 2132 struct btrfs_ioctl_get_subvol_rootref_args AUTO_KFREE(rootrefs); 2133 struct btrfs_root_ref *rref; 2134 struct btrfs_path *path; 2135 struct btrfs_key key; 2136 struct extent_buffer *leaf; 2137 u64 objectid; 2138 int slot; 2139 int ret; 2140 u8 found; 2141 2142 path = btrfs_alloc_path(); 2143 if (!path) 2144 return -ENOMEM; 2145 2146 rootrefs = memdup_user(argp, sizeof(*rootrefs)); 2147 if (IS_ERR(rootrefs)) { 2148 btrfs_free_path(path); 2149 return PTR_ERR(rootrefs); 2150 } 2151 2152 objectid = btrfs_root_id(root); 2153 key.objectid = objectid; 2154 key.type = BTRFS_ROOT_REF_KEY; 2155 key.offset = rootrefs->min_treeid; 2156 found = 0; 2157 2158 root = root->fs_info->tree_root; 2159 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); 2160 if (ret < 0) { 2161 goto out; 2162 } else if (path->slots[0] >= 2163 btrfs_header_nritems(path->nodes[0])) { 2164 ret = btrfs_next_leaf(root, path); 2165 if (ret < 0) { 2166 goto out; 2167 } else if (unlikely(ret > 0)) { 2168 ret = -EUCLEAN; 2169 goto out; 2170 } 2171 } 2172 while (1) { 2173 leaf = path->nodes[0]; 2174 slot = path->slots[0]; 2175 2176 btrfs_item_key_to_cpu(leaf, &key, slot); 2177 if (key.objectid != objectid || key.type != BTRFS_ROOT_REF_KEY) { 2178 ret = 0; 2179 goto out; 2180 } 2181 2182 if (found == BTRFS_MAX_ROOTREF_BUFFER_NUM) { 2183 ret = -EOVERFLOW; 2184 goto out; 2185 } 2186 2187 rref = btrfs_item_ptr(leaf, slot, struct btrfs_root_ref); 2188 rootrefs->rootref[found].treeid = key.offset; 2189 rootrefs->rootref[found].dirid = 2190 btrfs_root_ref_dirid(leaf, rref); 2191 found++; 2192 2193 ret = btrfs_next_item(root, path); 2194 if (ret < 0) { 2195 goto out; 2196 } else if (unlikely(ret > 0)) { 2197 ret = -EUCLEAN; 2198 goto out; 2199 } 2200 } 2201 2202 out: 2203 btrfs_free_path(path); 2204 2205 if (!ret || ret == -EOVERFLOW) { 2206 rootrefs->num_items = found; 2207 /* update min_treeid for next search */ 2208 if (found) 2209 rootrefs->min_treeid = 2210 rootrefs->rootref[found - 1].treeid + 1; 2211 if (copy_to_user(argp, rootrefs, sizeof(*rootrefs))) 2212 ret = -EFAULT; 2213 } 2214 2215 return ret; 2216 } 2217 2218 static noinline int btrfs_ioctl_snap_destroy(struct file *file, 2219 void __user *arg, 2220 bool destroy_v2) 2221 { 2222 struct dentry *parent = file->f_path.dentry; 2223 struct dentry *dentry; 2224 struct inode *dir = d_inode(parent); 2225 struct btrfs_fs_info *fs_info = inode_to_fs_info(dir); 2226 struct inode *inode; 2227 struct btrfs_root *root = BTRFS_I(dir)->root; 2228 struct btrfs_root *dest = NULL; 2229 struct btrfs_ioctl_vol_args AUTO_KFREE(vol_args); 2230 struct btrfs_ioctl_vol_args_v2 AUTO_KFREE(vol_args2); 2231 struct mnt_idmap *idmap = file_mnt_idmap(file); 2232 char *subvol_name, *subvol_name_ptr = NULL; 2233 int ret = 0; 2234 bool destroy_parent = false; 2235 2236 /* We don't support snapshots with extent tree v2 yet. */ 2237 if (btrfs_fs_incompat(fs_info, EXTENT_TREE_V2)) { 2238 btrfs_err(fs_info, 2239 "extent tree v2 doesn't support snapshot deletion yet"); 2240 return -EOPNOTSUPP; 2241 } 2242 2243 if (destroy_v2) { 2244 vol_args2 = memdup_user(arg, sizeof(*vol_args2)); 2245 if (IS_ERR(vol_args2)) 2246 return PTR_ERR(vol_args2); 2247 2248 if (vol_args2->flags & ~BTRFS_SUBVOL_DELETE_ARGS_MASK) 2249 return -EOPNOTSUPP; 2250 2251 /* 2252 * If SPEC_BY_ID is not set, we are looking for the subvolume by 2253 * name, same as v1 currently does. 2254 */ 2255 if (!(vol_args2->flags & BTRFS_SUBVOL_SPEC_BY_ID)) { 2256 ret = btrfs_check_ioctl_vol_args2_subvol_name(vol_args2); 2257 if (ret < 0) 2258 return ret; 2259 subvol_name = vol_args2->name; 2260 2261 ret = mnt_want_write_file(file); 2262 if (ret) 2263 return ret; 2264 } else { 2265 struct inode *old_dir; 2266 2267 if (vol_args2->subvolid < BTRFS_FIRST_FREE_OBJECTID) 2268 return -EINVAL; 2269 2270 ret = mnt_want_write_file(file); 2271 if (ret) 2272 return ret; 2273 2274 dentry = btrfs_get_dentry(fs_info->sb, 2275 BTRFS_FIRST_FREE_OBJECTID, 2276 vol_args2->subvolid, 0); 2277 if (IS_ERR(dentry)) { 2278 ret = PTR_ERR(dentry); 2279 goto out_drop_write; 2280 } 2281 2282 /* 2283 * Change the default parent since the subvolume being 2284 * deleted can be outside of the current mount point. 2285 */ 2286 parent = btrfs_get_parent(dentry); 2287 2288 /* 2289 * At this point dentry->d_name can point to '/' if the 2290 * subvolume we want to destroy is outsite of the 2291 * current mount point, so we need to release the 2292 * current dentry and execute the lookup to return a new 2293 * one with ->d_name pointing to the 2294 * <mount point>/subvol_name. 2295 */ 2296 dput(dentry); 2297 if (IS_ERR(parent)) { 2298 ret = PTR_ERR(parent); 2299 goto out_drop_write; 2300 } 2301 old_dir = dir; 2302 dir = d_inode(parent); 2303 2304 /* 2305 * If v2 was used with SPEC_BY_ID, a new parent was 2306 * allocated since the subvolume can be outside of the 2307 * current mount point. Later on we need to release this 2308 * new parent dentry. 2309 */ 2310 destroy_parent = true; 2311 2312 /* 2313 * On idmapped mounts, deletion via subvolid is 2314 * restricted to subvolumes that are immediate 2315 * ancestors of the inode referenced by the file 2316 * descriptor in the ioctl. Otherwise the idmapping 2317 * could potentially be abused to delete subvolumes 2318 * anywhere in the filesystem the user wouldn't be able 2319 * to delete without an idmapped mount. 2320 */ 2321 if (old_dir != dir && idmap != &nop_mnt_idmap) { 2322 ret = -EOPNOTSUPP; 2323 goto free_parent; 2324 } 2325 2326 subvol_name_ptr = btrfs_get_subvol_name_from_objectid( 2327 fs_info, vol_args2->subvolid); 2328 if (IS_ERR(subvol_name_ptr)) { 2329 ret = PTR_ERR(subvol_name_ptr); 2330 goto free_parent; 2331 } 2332 /* subvol_name_ptr is already nul terminated */ 2333 subvol_name = (char *)kbasename(subvol_name_ptr); 2334 } 2335 } else { 2336 vol_args = memdup_user(arg, sizeof(*vol_args)); 2337 if (IS_ERR(vol_args)) 2338 return PTR_ERR(vol_args); 2339 2340 ret = btrfs_check_ioctl_vol_args_path(vol_args); 2341 if (ret < 0) 2342 return ret; 2343 2344 subvol_name = vol_args->name; 2345 2346 ret = mnt_want_write_file(file); 2347 if (ret) 2348 return ret; 2349 } 2350 2351 if (strchr(subvol_name, '/') || 2352 strcmp(subvol_name, "..") == 0) { 2353 ret = -EINVAL; 2354 goto free_subvol_name; 2355 } 2356 2357 if (!S_ISDIR(dir->i_mode)) { 2358 ret = -ENOTDIR; 2359 goto free_subvol_name; 2360 } 2361 2362 dentry = start_removing_killable(idmap, parent, &QSTR(subvol_name)); 2363 if (IS_ERR(dentry)) { 2364 ret = PTR_ERR(dentry); 2365 goto out_end_removing; 2366 } 2367 2368 inode = d_inode(dentry); 2369 dest = BTRFS_I(inode)->root; 2370 if (!capable(CAP_SYS_ADMIN)) { 2371 /* 2372 * Regular user. Only allow this with a special mount 2373 * option, when the user has write+exec access to the 2374 * subvol root, and when rmdir(2) would have been 2375 * allowed. 2376 * 2377 * Note that this is _not_ check that the subvol is 2378 * empty or doesn't contain data that we wouldn't 2379 * otherwise be able to delete. 2380 * 2381 * Users who want to delete empty subvols should try 2382 * rmdir(2). 2383 */ 2384 ret = -EPERM; 2385 if (!btrfs_test_opt(fs_info, USER_SUBVOL_RM_ALLOWED)) 2386 goto out_end_removing; 2387 2388 /* 2389 * Do not allow deletion if the parent dir is the same 2390 * as the dir to be deleted. That means the ioctl 2391 * must be called on the dentry referencing the root 2392 * of the subvol, not a random directory contained 2393 * within it. 2394 */ 2395 ret = -EINVAL; 2396 if (root == dest) 2397 goto out_end_removing; 2398 2399 ret = inode_permission(idmap, inode, MAY_WRITE | MAY_EXEC); 2400 if (ret) 2401 goto out_end_removing; 2402 } 2403 2404 /* check if subvolume may be deleted by a user */ 2405 ret = may_delete_dentry(idmap, dir, dentry, true); 2406 if (ret) 2407 goto out_end_removing; 2408 2409 if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) { 2410 ret = -EINVAL; 2411 goto out_end_removing; 2412 } 2413 2414 btrfs_inode_lock(BTRFS_I(inode), 0); 2415 ret = btrfs_delete_subvolume(BTRFS_I(dir), dentry); 2416 btrfs_inode_unlock(BTRFS_I(inode), 0); 2417 if (!ret) 2418 d_delete_notify(dir, dentry); 2419 2420 out_end_removing: 2421 end_removing(dentry); 2422 free_subvol_name: 2423 kfree(subvol_name_ptr); 2424 free_parent: 2425 if (destroy_parent) 2426 dput(parent); 2427 out_drop_write: 2428 mnt_drop_write_file(file); 2429 return ret; 2430 } 2431 2432 static int btrfs_ioctl_defrag(struct file *file, void __user *argp) 2433 { 2434 struct inode *inode = file_inode(file); 2435 struct btrfs_root *root = BTRFS_I(inode)->root; 2436 struct btrfs_ioctl_defrag_range_args range = {0}; 2437 int ret; 2438 2439 ret = mnt_want_write_file(file); 2440 if (ret) 2441 return ret; 2442 2443 if (btrfs_root_readonly(root)) { 2444 ret = -EROFS; 2445 goto out; 2446 } 2447 2448 switch (inode->i_mode & S_IFMT) { 2449 case S_IFDIR: 2450 if (!capable(CAP_SYS_ADMIN)) { 2451 ret = -EPERM; 2452 goto out; 2453 } 2454 ret = btrfs_defrag_root(root); 2455 break; 2456 case S_IFREG: 2457 /* 2458 * Note that this does not check the file descriptor for write 2459 * access. This prevents defragmenting executables that are 2460 * running and allows defrag on files open in read-only mode. 2461 */ 2462 if (!capable(CAP_SYS_ADMIN) && 2463 inode_permission(&nop_mnt_idmap, inode, MAY_WRITE)) { 2464 ret = -EPERM; 2465 goto out; 2466 } 2467 2468 /* 2469 * Don't allow defrag on pre-content watched files, as it could 2470 * populate the page cache with 0's via readahead. 2471 */ 2472 if (unlikely(FMODE_FSNOTIFY_HSM(file->f_mode))) { 2473 ret = -EINVAL; 2474 goto out; 2475 } 2476 2477 if (argp) { 2478 if (copy_from_user(&range, argp, sizeof(range))) { 2479 ret = -EFAULT; 2480 goto out; 2481 } 2482 if (range.flags & ~BTRFS_DEFRAG_RANGE_FLAGS_SUPP) { 2483 ret = -EOPNOTSUPP; 2484 goto out; 2485 } 2486 if ((range.flags & BTRFS_DEFRAG_RANGE_COMPRESS) && 2487 (range.flags & BTRFS_DEFRAG_RANGE_NOCOMPRESS)) { 2488 ret = -EINVAL; 2489 goto out; 2490 } 2491 /* Compression or no-compression require to start the IO. */ 2492 if ((range.flags & BTRFS_DEFRAG_RANGE_COMPRESS) || 2493 (range.flags & BTRFS_DEFRAG_RANGE_NOCOMPRESS)) { 2494 range.flags |= BTRFS_DEFRAG_RANGE_START_IO; 2495 range.extent_thresh = (u32)-1; 2496 } 2497 } else { 2498 /* the rest are all set to zero by kzalloc */ 2499 range.len = (u64)-1; 2500 } 2501 ret = btrfs_defrag_file(BTRFS_I(file_inode(file)), &file->f_ra, 2502 &range, BTRFS_OLDEST_GENERATION, 0); 2503 if (ret > 0) 2504 ret = 0; 2505 break; 2506 default: 2507 ret = -EINVAL; 2508 } 2509 out: 2510 mnt_drop_write_file(file); 2511 return ret; 2512 } 2513 2514 static long btrfs_ioctl_add_dev(struct btrfs_fs_info *fs_info, void __user *arg) 2515 { 2516 struct btrfs_ioctl_vol_args AUTO_KFREE(vol_args); 2517 bool restore_op = false; 2518 int ret; 2519 2520 if (!capable(CAP_SYS_ADMIN)) 2521 return -EPERM; 2522 2523 if (btrfs_fs_incompat(fs_info, EXTENT_TREE_V2)) { 2524 btrfs_err(fs_info, "device add not supported on extent tree v2 yet"); 2525 return -EINVAL; 2526 } 2527 2528 if (fs_info->fs_devices->temp_fsid) { 2529 btrfs_err(fs_info, 2530 "device add not supported on cloned temp-fsid mount"); 2531 return -EINVAL; 2532 } 2533 2534 if (!btrfs_exclop_start(fs_info, BTRFS_EXCLOP_DEV_ADD)) { 2535 if (!btrfs_exclop_start_try_lock(fs_info, BTRFS_EXCLOP_DEV_ADD)) 2536 return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS; 2537 2538 /* 2539 * We can do the device add because we have a paused balanced, 2540 * change the exclusive op type and remember we should bring 2541 * back the paused balance 2542 */ 2543 fs_info->exclusive_operation = BTRFS_EXCLOP_DEV_ADD; 2544 btrfs_exclop_start_unlock(fs_info); 2545 restore_op = true; 2546 } 2547 2548 vol_args = memdup_user(arg, sizeof(*vol_args)); 2549 if (IS_ERR(vol_args)) { 2550 ret = PTR_ERR(vol_args); 2551 goto out; 2552 } 2553 2554 ret = btrfs_check_ioctl_vol_args_path(vol_args); 2555 if (ret < 0) 2556 goto out; 2557 2558 ret = btrfs_init_new_device(fs_info, vol_args->name); 2559 2560 if (!ret) 2561 btrfs_info(fs_info, "disk added %s", vol_args->name); 2562 2563 out: 2564 if (restore_op) 2565 btrfs_exclop_balance(fs_info, BTRFS_EXCLOP_BALANCE_PAUSED); 2566 else 2567 btrfs_exclop_finish(fs_info); 2568 return ret; 2569 } 2570 2571 static long btrfs_ioctl_rm_dev_v2(struct file *file, void __user *arg) 2572 { 2573 BTRFS_DEV_LOOKUP_ARGS(args); 2574 struct inode *inode = file_inode(file); 2575 struct btrfs_fs_info *fs_info = inode_to_fs_info(inode); 2576 struct btrfs_ioctl_vol_args_v2 AUTO_KFREE(vol_args); 2577 struct file *bdev_file = NULL; 2578 int ret; 2579 bool cancel = false; 2580 2581 if (!capable(CAP_SYS_ADMIN)) 2582 return -EPERM; 2583 2584 vol_args = memdup_user(arg, sizeof(*vol_args)); 2585 if (IS_ERR(vol_args)) 2586 return PTR_ERR(vol_args); 2587 2588 if (vol_args->flags & ~BTRFS_DEVICE_REMOVE_ARGS_MASK) { 2589 ret = -EOPNOTSUPP; 2590 goto out; 2591 } 2592 2593 ret = btrfs_check_ioctl_vol_args2_subvol_name(vol_args); 2594 if (ret < 0) 2595 goto out; 2596 2597 if (vol_args->flags & BTRFS_DEVICE_SPEC_BY_ID) { 2598 args.devid = vol_args->devid; 2599 } else if (!strcmp("cancel", vol_args->name)) { 2600 cancel = true; 2601 } else { 2602 ret = btrfs_get_dev_args_from_path(fs_info, &args, vol_args->name); 2603 if (ret) 2604 goto out; 2605 } 2606 2607 ret = mnt_want_write_file(file); 2608 if (ret) 2609 goto out; 2610 2611 ret = exclop_start_or_cancel_reloc(fs_info, BTRFS_EXCLOP_DEV_REMOVE, 2612 cancel); 2613 if (ret) 2614 goto err_drop; 2615 2616 /* Exclusive operation is now claimed */ 2617 ret = btrfs_rm_device(fs_info, &args, &bdev_file); 2618 2619 btrfs_exclop_finish(fs_info); 2620 2621 if (!ret) { 2622 if (vol_args->flags & BTRFS_DEVICE_SPEC_BY_ID) 2623 btrfs_info(fs_info, "device deleted: id %llu", 2624 vol_args->devid); 2625 else 2626 btrfs_info(fs_info, "device deleted: %s", 2627 vol_args->name); 2628 } 2629 err_drop: 2630 mnt_drop_write_file(file); 2631 if (bdev_file) 2632 bdev_fput(bdev_file); 2633 out: 2634 btrfs_put_dev_args_from_path(&args); 2635 return ret; 2636 } 2637 2638 static long btrfs_ioctl_rm_dev(struct file *file, void __user *arg) 2639 { 2640 BTRFS_DEV_LOOKUP_ARGS(args); 2641 struct inode *inode = file_inode(file); 2642 struct btrfs_fs_info *fs_info = inode_to_fs_info(inode); 2643 struct btrfs_ioctl_vol_args AUTO_KFREE(vol_args); 2644 struct file *bdev_file = NULL; 2645 int ret; 2646 bool cancel = false; 2647 2648 if (!capable(CAP_SYS_ADMIN)) 2649 return -EPERM; 2650 2651 vol_args = memdup_user(arg, sizeof(*vol_args)); 2652 if (IS_ERR(vol_args)) 2653 return PTR_ERR(vol_args); 2654 2655 ret = btrfs_check_ioctl_vol_args_path(vol_args); 2656 if (ret < 0) 2657 return ret; 2658 2659 if (!strcmp("cancel", vol_args->name)) { 2660 cancel = true; 2661 } else { 2662 ret = btrfs_get_dev_args_from_path(fs_info, &args, vol_args->name); 2663 if (ret) 2664 goto out; 2665 } 2666 2667 ret = mnt_want_write_file(file); 2668 if (ret) 2669 goto out; 2670 2671 ret = exclop_start_or_cancel_reloc(fs_info, BTRFS_EXCLOP_DEV_REMOVE, 2672 cancel); 2673 if (ret == 0) { 2674 ret = btrfs_rm_device(fs_info, &args, &bdev_file); 2675 if (!ret) 2676 btrfs_info(fs_info, "disk deleted %s", vol_args->name); 2677 btrfs_exclop_finish(fs_info); 2678 } 2679 2680 mnt_drop_write_file(file); 2681 if (bdev_file) 2682 bdev_fput(bdev_file); 2683 out: 2684 btrfs_put_dev_args_from_path(&args); 2685 return ret; 2686 } 2687 2688 static long btrfs_ioctl_fs_info(const struct btrfs_fs_info *fs_info, 2689 void __user *arg) 2690 { 2691 struct btrfs_ioctl_fs_info_args AUTO_KFREE(fi_args); 2692 struct btrfs_device *device; 2693 struct btrfs_fs_devices *fs_devices = fs_info->fs_devices; 2694 u64 flags_in; 2695 2696 fi_args = memdup_user(arg, sizeof(*fi_args)); 2697 if (IS_ERR(fi_args)) 2698 return PTR_ERR(fi_args); 2699 2700 flags_in = fi_args->flags; 2701 memset(fi_args, 0, sizeof(*fi_args)); 2702 2703 rcu_read_lock(); 2704 fi_args->num_devices = fs_devices->num_devices; 2705 2706 list_for_each_entry_rcu(device, &fs_devices->devices, dev_list) { 2707 if (device->devid > fi_args->max_id) 2708 fi_args->max_id = device->devid; 2709 } 2710 rcu_read_unlock(); 2711 2712 memcpy(&fi_args->fsid, fs_devices->fsid, sizeof(fi_args->fsid)); 2713 fi_args->nodesize = fs_info->nodesize; 2714 fi_args->sectorsize = fs_info->sectorsize; 2715 fi_args->clone_alignment = fs_info->sectorsize; 2716 2717 if (flags_in & BTRFS_FS_INFO_FLAG_CSUM_INFO) { 2718 fi_args->csum_type = btrfs_super_csum_type(fs_info->super_copy); 2719 fi_args->csum_size = btrfs_super_csum_size(fs_info->super_copy); 2720 fi_args->flags |= BTRFS_FS_INFO_FLAG_CSUM_INFO; 2721 } 2722 2723 if (flags_in & BTRFS_FS_INFO_FLAG_GENERATION) { 2724 fi_args->generation = btrfs_get_fs_generation(fs_info); 2725 fi_args->flags |= BTRFS_FS_INFO_FLAG_GENERATION; 2726 } 2727 2728 if (flags_in & BTRFS_FS_INFO_FLAG_METADATA_UUID) { 2729 memcpy(&fi_args->metadata_uuid, fs_devices->metadata_uuid, 2730 sizeof(fi_args->metadata_uuid)); 2731 fi_args->flags |= BTRFS_FS_INFO_FLAG_METADATA_UUID; 2732 } 2733 2734 if (copy_to_user(arg, fi_args, sizeof(*fi_args))) 2735 return -EFAULT; 2736 2737 return 0; 2738 } 2739 2740 static long btrfs_ioctl_dev_info(const struct btrfs_fs_info *fs_info, 2741 void __user *arg) 2742 { 2743 BTRFS_DEV_LOOKUP_ARGS(args); 2744 struct btrfs_ioctl_dev_info_args AUTO_KFREE(di_args); 2745 struct btrfs_device *dev; 2746 int ret = 0; 2747 2748 di_args = memdup_user(arg, sizeof(*di_args)); 2749 if (IS_ERR(di_args)) 2750 return PTR_ERR(di_args); 2751 2752 args.devid = di_args->devid; 2753 if (!btrfs_is_empty_uuid(di_args->uuid)) 2754 args.uuid = di_args->uuid; 2755 2756 rcu_read_lock(); 2757 dev = btrfs_find_device(fs_info->fs_devices, &args); 2758 if (!dev) { 2759 ret = -ENODEV; 2760 goto out; 2761 } 2762 2763 di_args->devid = dev->devid; 2764 di_args->bytes_used = btrfs_device_get_bytes_used(dev); 2765 di_args->total_bytes = btrfs_device_get_total_bytes(dev); 2766 memcpy(di_args->uuid, dev->uuid, sizeof(di_args->uuid)); 2767 memcpy(di_args->fsid, dev->fs_devices->fsid, BTRFS_UUID_SIZE); 2768 if (dev->name) 2769 strscpy(di_args->path, btrfs_dev_name(dev), sizeof(di_args->path)); 2770 else 2771 di_args->path[0] = '\0'; 2772 2773 out: 2774 rcu_read_unlock(); 2775 if (ret == 0 && copy_to_user(arg, di_args, sizeof(*di_args))) 2776 ret = -EFAULT; 2777 2778 return ret; 2779 } 2780 2781 static long btrfs_ioctl_default_subvol(struct file *file, void __user *argp) 2782 { 2783 struct inode *inode = file_inode(file); 2784 struct btrfs_fs_info *fs_info = inode_to_fs_info(inode); 2785 struct btrfs_root *root = BTRFS_I(inode)->root; 2786 struct btrfs_root *new_root; 2787 struct btrfs_dir_item *di; 2788 struct btrfs_trans_handle *trans; 2789 struct btrfs_path *path = NULL; 2790 struct btrfs_disk_key disk_key; 2791 struct fscrypt_str name = FSTR_INIT("default", 7); 2792 u64 objectid = 0; 2793 u64 dir_id; 2794 int ret; 2795 2796 if (!capable(CAP_SYS_ADMIN)) 2797 return -EPERM; 2798 2799 ret = mnt_want_write_file(file); 2800 if (ret) 2801 return ret; 2802 2803 if (copy_from_user(&objectid, argp, sizeof(objectid))) { 2804 ret = -EFAULT; 2805 goto out; 2806 } 2807 2808 if (!objectid) 2809 objectid = BTRFS_FS_TREE_OBJECTID; 2810 2811 new_root = btrfs_get_fs_root(fs_info, objectid, true); 2812 if (IS_ERR(new_root)) { 2813 ret = PTR_ERR(new_root); 2814 goto out; 2815 } 2816 if (!btrfs_is_fstree(btrfs_root_id(new_root))) { 2817 ret = -ENOENT; 2818 goto out_free; 2819 } 2820 2821 path = btrfs_alloc_path(); 2822 if (!path) { 2823 ret = -ENOMEM; 2824 goto out_free; 2825 } 2826 2827 trans = btrfs_start_transaction(root, 1); 2828 if (IS_ERR(trans)) { 2829 ret = PTR_ERR(trans); 2830 goto out_free; 2831 } 2832 2833 dir_id = btrfs_super_root_dir(fs_info->super_copy); 2834 di = btrfs_lookup_dir_item(trans, fs_info->tree_root, path, 2835 dir_id, &name, 1); 2836 if (IS_ERR_OR_NULL(di)) { 2837 btrfs_release_path(path); 2838 btrfs_end_transaction(trans); 2839 if (di) 2840 ret = PTR_ERR(di); 2841 else 2842 ret = -ENOENT; 2843 btrfs_err(fs_info, 2844 "could not find default diritem for dir %llu: %d", 2845 dir_id, ret); 2846 goto out_free; 2847 } 2848 2849 btrfs_cpu_key_to_disk(&disk_key, &new_root->root_key); 2850 btrfs_set_dir_item_key(path->nodes[0], di, &disk_key); 2851 btrfs_release_path(path); 2852 2853 btrfs_set_fs_incompat(fs_info, DEFAULT_SUBVOL); 2854 btrfs_end_transaction(trans); 2855 out_free: 2856 btrfs_put_root(new_root); 2857 btrfs_free_path(path); 2858 out: 2859 mnt_drop_write_file(file); 2860 return ret; 2861 } 2862 2863 static void get_block_group_info(struct list_head *groups_list, 2864 struct btrfs_ioctl_space_info *space) 2865 { 2866 struct btrfs_block_group *block_group; 2867 2868 space->total_bytes = 0; 2869 space->used_bytes = 0; 2870 space->flags = 0; 2871 list_for_each_entry(block_group, groups_list, list) { 2872 space->flags = block_group->flags; 2873 space->total_bytes += block_group->length; 2874 space->used_bytes += block_group->used; 2875 } 2876 } 2877 2878 static long btrfs_ioctl_space_info(struct btrfs_fs_info *fs_info, 2879 void __user *arg) 2880 { 2881 struct btrfs_ioctl_space_args space_args = { 0 }; 2882 struct btrfs_ioctl_space_info space; 2883 struct btrfs_ioctl_space_info *dest; 2884 struct btrfs_ioctl_space_info AUTO_KFREE(dest_orig); 2885 struct btrfs_ioctl_space_info __user *user_dest; 2886 struct btrfs_space_info *info; 2887 static const u64 types[] = { 2888 BTRFS_BLOCK_GROUP_DATA, 2889 BTRFS_BLOCK_GROUP_SYSTEM, 2890 BTRFS_BLOCK_GROUP_METADATA, 2891 BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA 2892 }; 2893 int num_types = 4; 2894 int alloc_size; 2895 int ret = 0; 2896 u64 slot_count = 0; 2897 int i, c; 2898 2899 if (copy_from_user(&space_args, 2900 (struct btrfs_ioctl_space_args __user *)arg, 2901 sizeof(space_args))) 2902 return -EFAULT; 2903 2904 for (i = 0; i < num_types; i++) { 2905 struct btrfs_space_info *tmp; 2906 2907 info = NULL; 2908 list_for_each_entry(tmp, &fs_info->space_info, list) { 2909 if (tmp->flags == types[i]) { 2910 info = tmp; 2911 break; 2912 } 2913 } 2914 2915 if (!info) 2916 continue; 2917 2918 down_read(&info->groups_sem); 2919 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) { 2920 if (!list_empty(&info->block_groups[c])) 2921 slot_count++; 2922 } 2923 up_read(&info->groups_sem); 2924 } 2925 2926 /* 2927 * Global block reserve, exported as a space_info 2928 */ 2929 slot_count++; 2930 2931 /* space_slots == 0 means they are asking for a count */ 2932 if (space_args.space_slots == 0) { 2933 space_args.total_spaces = slot_count; 2934 goto out; 2935 } 2936 2937 slot_count = min_t(u64, space_args.space_slots, slot_count); 2938 2939 alloc_size = sizeof(*dest) * slot_count; 2940 2941 /* we generally have at most 6 or so space infos, one for each raid 2942 * level. So, a whole page should be more than enough for everyone 2943 */ 2944 if (alloc_size > PAGE_SIZE) 2945 return -ENOMEM; 2946 2947 space_args.total_spaces = 0; 2948 dest = kzalloc(alloc_size, GFP_KERNEL); 2949 if (!dest) 2950 return -ENOMEM; 2951 dest_orig = dest; 2952 2953 /* now we have a buffer to copy into */ 2954 for (i = 0; i < num_types; i++) { 2955 struct btrfs_space_info *tmp; 2956 2957 if (!slot_count) 2958 break; 2959 2960 info = NULL; 2961 list_for_each_entry(tmp, &fs_info->space_info, list) { 2962 if (tmp->flags == types[i]) { 2963 info = tmp; 2964 break; 2965 } 2966 } 2967 2968 if (!info) 2969 continue; 2970 down_read(&info->groups_sem); 2971 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) { 2972 if (!list_empty(&info->block_groups[c])) { 2973 get_block_group_info(&info->block_groups[c], 2974 &space); 2975 memcpy(dest, &space, sizeof(space)); 2976 dest++; 2977 space_args.total_spaces++; 2978 slot_count--; 2979 } 2980 if (!slot_count) 2981 break; 2982 } 2983 up_read(&info->groups_sem); 2984 } 2985 2986 /* 2987 * Add global block reserve 2988 */ 2989 if (slot_count) { 2990 struct btrfs_block_rsv *block_rsv = &fs_info->global_block_rsv; 2991 2992 spin_lock(&block_rsv->lock); 2993 space.total_bytes = block_rsv->size; 2994 space.used_bytes = block_rsv->size - block_rsv->reserved; 2995 spin_unlock(&block_rsv->lock); 2996 space.flags = BTRFS_SPACE_INFO_GLOBAL_RSV; 2997 memcpy(dest, &space, sizeof(space)); 2998 space_args.total_spaces++; 2999 } 3000 3001 user_dest = (struct btrfs_ioctl_space_info __user *) 3002 (arg + sizeof(struct btrfs_ioctl_space_args)); 3003 3004 if (copy_to_user(user_dest, dest_orig, 3005 space_args.total_spaces * sizeof(*dest_orig))) 3006 return -EFAULT; 3007 3008 out: 3009 if (ret == 0 && copy_to_user(arg, &space_args, sizeof(space_args))) 3010 ret = -EFAULT; 3011 3012 return ret; 3013 } 3014 3015 static noinline long btrfs_ioctl_start_sync(struct btrfs_root *root, 3016 void __user *argp) 3017 { 3018 struct btrfs_trans_handle *trans; 3019 u64 transid; 3020 3021 /* 3022 * Start orphan cleanup here for the given root in case it hasn't been 3023 * started already by other means. Errors are handled in the other 3024 * functions during transaction commit. 3025 */ 3026 btrfs_orphan_cleanup(root); 3027 3028 trans = btrfs_attach_transaction_barrier(root); 3029 if (IS_ERR(trans)) { 3030 if (PTR_ERR(trans) != -ENOENT) 3031 return PTR_ERR(trans); 3032 3033 /* No running transaction, don't bother */ 3034 transid = btrfs_get_last_trans_committed(root->fs_info); 3035 goto out; 3036 } 3037 transid = trans->transid; 3038 btrfs_commit_transaction_async(trans); 3039 out: 3040 if (argp) 3041 if (copy_to_user(argp, &transid, sizeof(transid))) 3042 return -EFAULT; 3043 return 0; 3044 } 3045 3046 static noinline long btrfs_ioctl_wait_sync(struct btrfs_fs_info *fs_info, 3047 void __user *argp) 3048 { 3049 /* By default wait for the current transaction. */ 3050 u64 transid = 0; 3051 3052 if (argp) 3053 if (copy_from_user(&transid, argp, sizeof(transid))) 3054 return -EFAULT; 3055 3056 return btrfs_wait_for_commit(fs_info, transid); 3057 } 3058 3059 static long btrfs_ioctl_scrub(struct file *file, void __user *arg) 3060 { 3061 struct btrfs_fs_info *fs_info = inode_to_fs_info(file_inode(file)); 3062 struct btrfs_ioctl_scrub_args AUTO_KFREE(sa); 3063 int ret; 3064 3065 if (!capable(CAP_SYS_ADMIN)) 3066 return -EPERM; 3067 3068 if (btrfs_fs_incompat(fs_info, EXTENT_TREE_V2)) { 3069 btrfs_err(fs_info, "scrub: extent tree v2 not yet supported"); 3070 return -EINVAL; 3071 } 3072 3073 sa = memdup_user(arg, sizeof(*sa)); 3074 if (IS_ERR(sa)) 3075 return PTR_ERR(sa); 3076 3077 if (sa->flags & ~BTRFS_SCRUB_SUPPORTED_FLAGS) 3078 return -EOPNOTSUPP; 3079 3080 if (!(sa->flags & BTRFS_SCRUB_READONLY)) { 3081 ret = mnt_want_write_file(file); 3082 if (ret) 3083 return ret; 3084 } 3085 3086 ret = btrfs_scrub_dev(fs_info, sa->devid, sa->start, sa->end, 3087 &sa->progress, sa->flags & BTRFS_SCRUB_READONLY, 3088 false); 3089 3090 /* 3091 * Copy scrub args to user space even if btrfs_scrub_dev() returned an 3092 * error. This is important as it allows user space to know how much 3093 * progress scrub has done. For example, if scrub is canceled we get 3094 * -ECANCELED from btrfs_scrub_dev() and return that error back to user 3095 * space. Later user space can inspect the progress from the structure 3096 * btrfs_ioctl_scrub_args and resume scrub from where it left off 3097 * previously (btrfs-progs does this). 3098 * If we fail to copy the btrfs_ioctl_scrub_args structure to user space 3099 * then return -EFAULT to signal the structure was not copied or it may 3100 * be corrupt and unreliable due to a partial copy. 3101 */ 3102 if (copy_to_user(arg, sa, sizeof(*sa))) 3103 ret = -EFAULT; 3104 3105 if (!(sa->flags & BTRFS_SCRUB_READONLY)) 3106 mnt_drop_write_file(file); 3107 3108 return ret; 3109 } 3110 3111 static long btrfs_ioctl_scrub_cancel(struct btrfs_fs_info *fs_info) 3112 { 3113 if (!capable(CAP_SYS_ADMIN)) 3114 return -EPERM; 3115 3116 return btrfs_scrub_cancel(fs_info); 3117 } 3118 3119 static long btrfs_ioctl_scrub_progress(struct btrfs_fs_info *fs_info, 3120 void __user *arg) 3121 { 3122 struct btrfs_ioctl_scrub_args AUTO_KFREE(sa); 3123 int ret; 3124 3125 if (!capable(CAP_SYS_ADMIN)) 3126 return -EPERM; 3127 3128 sa = memdup_user(arg, sizeof(*sa)); 3129 if (IS_ERR(sa)) 3130 return PTR_ERR(sa); 3131 3132 ret = btrfs_scrub_progress(fs_info, sa->devid, &sa->progress); 3133 3134 if (ret == 0 && copy_to_user(arg, sa, sizeof(*sa))) 3135 return -EFAULT; 3136 3137 return ret; 3138 } 3139 3140 static long btrfs_ioctl_get_dev_stats(struct btrfs_fs_info *fs_info, 3141 void __user *arg) 3142 { 3143 struct btrfs_ioctl_get_dev_stats AUTO_KFREE(sa); 3144 int ret; 3145 3146 sa = memdup_user(arg, sizeof(*sa)); 3147 if (IS_ERR(sa)) 3148 return PTR_ERR(sa); 3149 3150 if ((sa->flags & BTRFS_DEV_STATS_RESET) && !capable(CAP_SYS_ADMIN)) 3151 return -EPERM; 3152 3153 ret = btrfs_get_dev_stats(fs_info, sa); 3154 3155 if (ret == 0 && copy_to_user(arg, sa, sizeof(*sa))) 3156 return -EFAULT; 3157 3158 return ret; 3159 } 3160 3161 static long btrfs_ioctl_dev_replace(struct btrfs_fs_info *fs_info, 3162 void __user *arg) 3163 { 3164 struct btrfs_ioctl_dev_replace_args AUTO_KFREE(p); 3165 int ret; 3166 3167 if (!capable(CAP_SYS_ADMIN)) 3168 return -EPERM; 3169 3170 if (btrfs_fs_incompat(fs_info, EXTENT_TREE_V2)) { 3171 btrfs_err(fs_info, "device replace not supported on extent tree v2 yet"); 3172 return -EINVAL; 3173 } 3174 3175 p = memdup_user(arg, sizeof(*p)); 3176 if (IS_ERR(p)) 3177 return PTR_ERR(p); 3178 3179 switch (p->cmd) { 3180 case BTRFS_IOCTL_DEV_REPLACE_CMD_START: 3181 if (sb_rdonly(fs_info->sb)) 3182 return -EROFS; 3183 if (!btrfs_exclop_start(fs_info, BTRFS_EXCLOP_DEV_REPLACE)) { 3184 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS; 3185 } else { 3186 ret = btrfs_dev_replace_by_ioctl(fs_info, p); 3187 btrfs_exclop_finish(fs_info); 3188 } 3189 break; 3190 case BTRFS_IOCTL_DEV_REPLACE_CMD_STATUS: 3191 btrfs_dev_replace_status(fs_info, p); 3192 ret = 0; 3193 break; 3194 case BTRFS_IOCTL_DEV_REPLACE_CMD_CANCEL: 3195 p->result = btrfs_dev_replace_cancel(fs_info); 3196 ret = 0; 3197 break; 3198 default: 3199 ret = -EINVAL; 3200 break; 3201 } 3202 3203 if ((ret == 0 || ret == -ECANCELED) && copy_to_user(arg, p, sizeof(*p))) 3204 return -EFAULT; 3205 3206 return ret; 3207 } 3208 3209 static long btrfs_ioctl_ino_to_path(struct btrfs_root *root, void __user *arg) 3210 { 3211 int ret = 0; 3212 int i; 3213 u64 rel_ptr; 3214 int size; 3215 struct btrfs_ioctl_ino_path_args AUTO_KFREE(ipa); 3216 struct inode_fs_paths *ipath __free(inode_fs_paths) = NULL; 3217 struct btrfs_path *path; 3218 3219 if (!capable(CAP_DAC_READ_SEARCH)) 3220 return -EPERM; 3221 3222 path = btrfs_alloc_path(); 3223 if (!path) { 3224 ret = -ENOMEM; 3225 goto out; 3226 } 3227 3228 ipa = memdup_user(arg, sizeof(*ipa)); 3229 if (IS_ERR(ipa)) { 3230 ret = PTR_ERR(ipa); 3231 ipa = NULL; 3232 goto out; 3233 } 3234 3235 size = min_t(u32, ipa->size, 4096); 3236 ipath = init_ipath(size, root, path); 3237 if (IS_ERR(ipath)) { 3238 ret = PTR_ERR(ipath); 3239 ipath = NULL; 3240 goto out; 3241 } 3242 3243 ret = paths_from_inode(ipa->inum, ipath); 3244 if (ret < 0) 3245 goto out; 3246 3247 for (i = 0; i < ipath->fspath->elem_cnt; ++i) { 3248 rel_ptr = ipath->fspath->val[i] - 3249 (u64)(unsigned long)ipath->fspath->val; 3250 ipath->fspath->val[i] = rel_ptr; 3251 } 3252 3253 btrfs_free_path(path); 3254 path = NULL; 3255 ret = copy_to_user((void __user *)(unsigned long)ipa->fspath, 3256 ipath->fspath, size); 3257 if (ret) { 3258 ret = -EFAULT; 3259 goto out; 3260 } 3261 3262 out: 3263 btrfs_free_path(path); 3264 3265 return ret; 3266 } 3267 3268 static long btrfs_ioctl_logical_to_ino(struct btrfs_fs_info *fs_info, 3269 void __user *arg, int version) 3270 { 3271 int ret = 0; 3272 int size; 3273 struct btrfs_ioctl_logical_ino_args AUTO_KFREE(loi); 3274 struct btrfs_data_container AUTO_KVFREE(inodes); 3275 bool ignore_offset; 3276 3277 if (!capable(CAP_SYS_ADMIN)) 3278 return -EPERM; 3279 3280 loi = memdup_user(arg, sizeof(*loi)); 3281 if (IS_ERR(loi)) 3282 return PTR_ERR(loi); 3283 3284 if (version == 1) { 3285 ignore_offset = false; 3286 size = min_t(u32, loi->size, SZ_64K); 3287 } else { 3288 /* All reserved bits must be 0 for now */ 3289 if (memchr_inv(loi->reserved, 0, sizeof(loi->reserved))) 3290 return -EINVAL; 3291 3292 /* Only accept flags we have defined so far */ 3293 if (loi->flags & ~(BTRFS_LOGICAL_INO_ARGS_IGNORE_OFFSET)) 3294 return -EINVAL; 3295 3296 ignore_offset = loi->flags & BTRFS_LOGICAL_INO_ARGS_IGNORE_OFFSET; 3297 size = min_t(u32, loi->size, SZ_16M); 3298 } 3299 3300 inodes = init_data_container(size); 3301 if (IS_ERR(inodes)) 3302 return PTR_ERR(inodes); 3303 3304 ret = iterate_inodes_from_logical(loi->logical, fs_info, inodes, ignore_offset); 3305 if (ret == -EINVAL) 3306 return -ENOENT; 3307 if (ret < 0) 3308 return ret; 3309 3310 ret = copy_to_user((void __user *)(unsigned long)loi->inodes, inodes, 3311 size); 3312 if (ret) 3313 ret = -EFAULT; 3314 3315 return ret; 3316 } 3317 3318 void btrfs_update_ioctl_balance_args(struct btrfs_fs_info *fs_info, 3319 struct btrfs_ioctl_balance_args *bargs) 3320 { 3321 struct btrfs_balance_control *bctl = fs_info->balance_ctl; 3322 3323 bargs->flags = bctl->flags; 3324 3325 if (test_bit(BTRFS_FS_BALANCE_RUNNING, &fs_info->flags)) 3326 bargs->state |= BTRFS_BALANCE_STATE_RUNNING; 3327 if (atomic_read(&fs_info->balance_pause_req)) 3328 bargs->state |= BTRFS_BALANCE_STATE_PAUSE_REQ; 3329 if (atomic_read(&fs_info->balance_cancel_req)) 3330 bargs->state |= BTRFS_BALANCE_STATE_CANCEL_REQ; 3331 3332 memcpy(&bargs->data, &bctl->data, sizeof(bargs->data)); 3333 memcpy(&bargs->meta, &bctl->meta, sizeof(bargs->meta)); 3334 memcpy(&bargs->sys, &bctl->sys, sizeof(bargs->sys)); 3335 3336 spin_lock(&fs_info->balance_lock); 3337 memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat)); 3338 spin_unlock(&fs_info->balance_lock); 3339 } 3340 3341 /* 3342 * Try to acquire fs_info::balance_mutex as well as set BTRFS_EXLCOP_BALANCE as 3343 * required. 3344 * 3345 * @fs_info: the filesystem 3346 * @excl_acquired: ptr to boolean value which is set to false in case balance 3347 * is being resumed 3348 * 3349 * Return 0 on success in which case both fs_info::balance is acquired as well 3350 * as exclusive ops are blocked. In case of failure return an error code. 3351 */ 3352 static int btrfs_try_lock_balance(struct btrfs_fs_info *fs_info, bool *excl_acquired) 3353 { 3354 int ret; 3355 3356 /* 3357 * Exclusive operation is locked. Three possibilities: 3358 * (1) some other op is running 3359 * (2) balance is running 3360 * (3) balance is paused -- special case (think resume) 3361 */ 3362 while (1) { 3363 if (btrfs_exclop_start(fs_info, BTRFS_EXCLOP_BALANCE)) { 3364 *excl_acquired = true; 3365 mutex_lock(&fs_info->balance_mutex); 3366 return 0; 3367 } 3368 3369 mutex_lock(&fs_info->balance_mutex); 3370 if (fs_info->balance_ctl) { 3371 /* This is either (2) or (3) */ 3372 if (test_bit(BTRFS_FS_BALANCE_RUNNING, &fs_info->flags)) { 3373 /* This is (2) */ 3374 ret = -EINPROGRESS; 3375 goto out_failure; 3376 3377 } else { 3378 mutex_unlock(&fs_info->balance_mutex); 3379 /* 3380 * Lock released to allow other waiters to 3381 * continue, we'll reexamine the status again. 3382 */ 3383 mutex_lock(&fs_info->balance_mutex); 3384 3385 if (fs_info->balance_ctl && 3386 !test_bit(BTRFS_FS_BALANCE_RUNNING, &fs_info->flags)) { 3387 /* This is (3) */ 3388 *excl_acquired = false; 3389 return 0; 3390 } 3391 } 3392 } else { 3393 /* This is (1) */ 3394 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS; 3395 goto out_failure; 3396 } 3397 3398 mutex_unlock(&fs_info->balance_mutex); 3399 } 3400 3401 out_failure: 3402 mutex_unlock(&fs_info->balance_mutex); 3403 *excl_acquired = false; 3404 return ret; 3405 } 3406 3407 static long btrfs_ioctl_balance(struct file *file, void __user *arg) 3408 { 3409 struct btrfs_root *root = BTRFS_I(file_inode(file))->root; 3410 struct btrfs_fs_info *fs_info = root->fs_info; 3411 struct btrfs_ioctl_balance_args AUTO_KFREE(bargs); 3412 struct btrfs_balance_control *bctl; 3413 bool need_unlock = true; 3414 int ret; 3415 3416 if (!capable(CAP_SYS_ADMIN)) 3417 return -EPERM; 3418 3419 ret = mnt_want_write_file(file); 3420 if (ret) 3421 return ret; 3422 3423 bargs = memdup_user(arg, sizeof(*bargs)); 3424 if (IS_ERR(bargs)) { 3425 ret = PTR_ERR(bargs); 3426 bargs = NULL; 3427 goto out; 3428 } 3429 3430 ret = btrfs_try_lock_balance(fs_info, &need_unlock); 3431 if (ret) 3432 goto out; 3433 3434 lockdep_assert_held(&fs_info->balance_mutex); 3435 3436 if (bargs->flags & BTRFS_BALANCE_RESUME) { 3437 if (!fs_info->balance_ctl) { 3438 ret = -ENOTCONN; 3439 goto out_unlock; 3440 } 3441 3442 bctl = fs_info->balance_ctl; 3443 spin_lock(&fs_info->balance_lock); 3444 bctl->flags |= BTRFS_BALANCE_RESUME; 3445 spin_unlock(&fs_info->balance_lock); 3446 btrfs_exclop_balance(fs_info, BTRFS_EXCLOP_BALANCE); 3447 3448 goto do_balance; 3449 } 3450 3451 if (bargs->flags & ~(BTRFS_BALANCE_ARGS_MASK | BTRFS_BALANCE_TYPE_MASK)) { 3452 ret = -EINVAL; 3453 goto out_unlock; 3454 } 3455 3456 if (fs_info->balance_ctl) { 3457 ret = -EINPROGRESS; 3458 goto out_unlock; 3459 } 3460 3461 bctl = kzalloc_obj(*bctl); 3462 if (!bctl) { 3463 ret = -ENOMEM; 3464 goto out_unlock; 3465 } 3466 3467 memcpy(&bctl->data, &bargs->data, sizeof(bctl->data)); 3468 memcpy(&bctl->meta, &bargs->meta, sizeof(bctl->meta)); 3469 memcpy(&bctl->sys, &bargs->sys, sizeof(bctl->sys)); 3470 3471 bctl->flags = bargs->flags; 3472 do_balance: 3473 /* 3474 * Ownership of bctl and exclusive operation goes to btrfs_balance. 3475 * bctl is freed in reset_balance_state, or, if restriper was paused 3476 * all the way until unmount, in free_fs_info. The flag should be 3477 * cleared after reset_balance_state. 3478 */ 3479 need_unlock = false; 3480 3481 ret = btrfs_balance(fs_info, bctl, bargs); 3482 bctl = NULL; 3483 3484 if (ret == 0 || ret == -ECANCELED) { 3485 if (copy_to_user(arg, bargs, sizeof(*bargs))) 3486 ret = -EFAULT; 3487 } 3488 3489 kfree(bctl); 3490 out_unlock: 3491 mutex_unlock(&fs_info->balance_mutex); 3492 if (need_unlock) 3493 btrfs_exclop_finish(fs_info); 3494 out: 3495 mnt_drop_write_file(file); 3496 return ret; 3497 } 3498 3499 static long btrfs_ioctl_balance_ctl(struct btrfs_fs_info *fs_info, int cmd) 3500 { 3501 if (!capable(CAP_SYS_ADMIN)) 3502 return -EPERM; 3503 3504 switch (cmd) { 3505 case BTRFS_BALANCE_CTL_PAUSE: 3506 return btrfs_pause_balance(fs_info); 3507 case BTRFS_BALANCE_CTL_CANCEL: 3508 return btrfs_cancel_balance(fs_info); 3509 } 3510 3511 return -EINVAL; 3512 } 3513 3514 static long btrfs_ioctl_balance_progress(struct btrfs_fs_info *fs_info, 3515 void __user *arg) 3516 { 3517 struct btrfs_ioctl_balance_args AUTO_KFREE(bargs); 3518 int ret = 0; 3519 3520 if (!capable(CAP_SYS_ADMIN)) 3521 return -EPERM; 3522 3523 mutex_lock(&fs_info->balance_mutex); 3524 if (!fs_info->balance_ctl) { 3525 ret = -ENOTCONN; 3526 goto out; 3527 } 3528 3529 bargs = kzalloc(sizeof(*bargs), GFP_KERNEL); 3530 if (!bargs) { 3531 ret = -ENOMEM; 3532 goto out; 3533 } 3534 3535 btrfs_update_ioctl_balance_args(fs_info, bargs); 3536 3537 if (copy_to_user(arg, bargs, sizeof(*bargs))) 3538 ret = -EFAULT; 3539 out: 3540 mutex_unlock(&fs_info->balance_mutex); 3541 return ret; 3542 } 3543 3544 static long btrfs_ioctl_quota_ctl(struct file *file, void __user *arg) 3545 { 3546 struct inode *inode = file_inode(file); 3547 struct btrfs_fs_info *fs_info = inode_to_fs_info(inode); 3548 struct btrfs_ioctl_quota_ctl_args AUTO_KFREE(sa); 3549 int ret; 3550 3551 if (!capable(CAP_SYS_ADMIN)) 3552 return -EPERM; 3553 3554 ret = mnt_want_write_file(file); 3555 if (ret) 3556 return ret; 3557 3558 sa = memdup_user(arg, sizeof(*sa)); 3559 if (IS_ERR(sa)) { 3560 ret = PTR_ERR(sa); 3561 goto drop_write; 3562 } 3563 3564 switch (sa->cmd) { 3565 case BTRFS_QUOTA_CTL_ENABLE: 3566 case BTRFS_QUOTA_CTL_ENABLE_SIMPLE_QUOTA: 3567 down_write(&fs_info->subvol_sem); 3568 ret = btrfs_quota_enable(fs_info, sa); 3569 up_write(&fs_info->subvol_sem); 3570 break; 3571 case BTRFS_QUOTA_CTL_DISABLE: 3572 /* 3573 * Lock the cleaner mutex to prevent races with concurrent 3574 * relocation, because relocation may be building backrefs for 3575 * blocks of the quota root while we are deleting the root. This 3576 * is like dropping fs roots of deleted snapshots/subvolumes, we 3577 * need the same protection. 3578 * 3579 * This also prevents races between concurrent tasks trying to 3580 * disable quotas, because we will unlock and relock 3581 * qgroup_ioctl_lock across BTRFS_FS_QUOTA_ENABLED changes. 3582 * 3583 * We take this here because we have the dependency of 3584 * 3585 * inode_lock -> subvol_sem 3586 * 3587 * because of rename. With relocation we can prealloc extents, 3588 * so that makes the dependency chain 3589 * 3590 * cleaner_mutex -> inode_lock -> subvol_sem 3591 * 3592 * so we must take the cleaner_mutex here before we take the 3593 * subvol_sem. The deadlock can't actually happen, but this 3594 * quiets lockdep. 3595 */ 3596 mutex_lock(&fs_info->cleaner_mutex); 3597 down_write(&fs_info->subvol_sem); 3598 ret = btrfs_quota_disable(fs_info); 3599 up_write(&fs_info->subvol_sem); 3600 mutex_unlock(&fs_info->cleaner_mutex); 3601 break; 3602 default: 3603 ret = -EINVAL; 3604 break; 3605 } 3606 3607 drop_write: 3608 mnt_drop_write_file(file); 3609 return ret; 3610 } 3611 3612 static long btrfs_ioctl_qgroup_assign(struct file *file, void __user *arg) 3613 { 3614 struct inode *inode = file_inode(file); 3615 struct btrfs_fs_info *fs_info = inode_to_fs_info(inode); 3616 struct btrfs_root *root = BTRFS_I(inode)->root; 3617 struct btrfs_ioctl_qgroup_assign_args AUTO_KFREE(sa); 3618 struct btrfs_qgroup_list AUTO_KFREE(prealloc); 3619 struct btrfs_trans_handle *trans; 3620 int ret; 3621 int err; 3622 3623 if (!capable(CAP_SYS_ADMIN)) 3624 return -EPERM; 3625 3626 if (!btrfs_qgroup_enabled(fs_info)) 3627 return -ENOTCONN; 3628 3629 ret = mnt_want_write_file(file); 3630 if (ret) 3631 return ret; 3632 3633 sa = memdup_user(arg, sizeof(*sa)); 3634 if (IS_ERR(sa)) { 3635 ret = PTR_ERR(sa); 3636 goto drop_write; 3637 } 3638 3639 if (sa->assign) { 3640 prealloc = kzalloc_obj(*prealloc); 3641 if (!prealloc) { 3642 ret = -ENOMEM; 3643 goto drop_write; 3644 } 3645 } 3646 3647 /* 2 BTRFS_QGROUP_RELATION_KEY items. */ 3648 trans = btrfs_start_transaction(root, 2); 3649 if (IS_ERR(trans)) { 3650 ret = PTR_ERR(trans); 3651 goto drop_write; 3652 } 3653 3654 /* 3655 * Prealloc ownership is moved to the relation handler, there it's used 3656 * or freed on error. 3657 */ 3658 if (sa->assign) { 3659 ret = btrfs_add_qgroup_relation(trans, sa->src, sa->dst, prealloc); 3660 prealloc = NULL; 3661 } else { 3662 ret = btrfs_del_qgroup_relation(trans, sa->src, sa->dst); 3663 } 3664 3665 /* update qgroup status and info */ 3666 mutex_lock(&fs_info->qgroup_ioctl_lock); 3667 err = btrfs_run_qgroups(trans); 3668 mutex_unlock(&fs_info->qgroup_ioctl_lock); 3669 if (err < 0) 3670 btrfs_warn(fs_info, 3671 "qgroup status update failed after %s relation, marked as inconsistent", 3672 sa->assign ? "adding" : "deleting"); 3673 err = btrfs_end_transaction(trans); 3674 if (err && !ret) 3675 ret = err; 3676 3677 drop_write: 3678 mnt_drop_write_file(file); 3679 return ret; 3680 } 3681 3682 static long btrfs_ioctl_qgroup_create(struct file *file, void __user *arg) 3683 { 3684 struct inode *inode = file_inode(file); 3685 struct btrfs_root *root = BTRFS_I(inode)->root; 3686 struct btrfs_ioctl_qgroup_create_args AUTO_KFREE(sa); 3687 struct btrfs_trans_handle *trans; 3688 int ret; 3689 int err; 3690 3691 if (!capable(CAP_SYS_ADMIN)) 3692 return -EPERM; 3693 3694 if (!btrfs_qgroup_enabled(root->fs_info)) 3695 return -ENOTCONN; 3696 3697 ret = mnt_want_write_file(file); 3698 if (ret) 3699 return ret; 3700 3701 sa = memdup_user(arg, sizeof(*sa)); 3702 if (IS_ERR(sa)) { 3703 ret = PTR_ERR(sa); 3704 goto drop_write; 3705 } 3706 3707 if (!sa->qgroupid) { 3708 ret = -EINVAL; 3709 goto drop_write; 3710 } 3711 3712 if (sa->create && btrfs_is_fstree(sa->qgroupid)) { 3713 ret = -EINVAL; 3714 goto drop_write; 3715 } 3716 3717 /* 3718 * 1 BTRFS_QGROUP_INFO_KEY item. 3719 * 1 BTRFS_QGROUP_LIMIT_KEY item. 3720 */ 3721 trans = btrfs_start_transaction(root, 2); 3722 if (IS_ERR(trans)) { 3723 ret = PTR_ERR(trans); 3724 goto drop_write; 3725 } 3726 3727 if (sa->create) { 3728 ret = btrfs_create_qgroup(trans, sa->qgroupid); 3729 } else { 3730 ret = btrfs_remove_qgroup(trans, sa->qgroupid); 3731 } 3732 3733 err = btrfs_end_transaction(trans); 3734 if (err && !ret) 3735 ret = err; 3736 3737 drop_write: 3738 mnt_drop_write_file(file); 3739 return ret; 3740 } 3741 3742 static long btrfs_ioctl_qgroup_limit(struct file *file, void __user *arg) 3743 { 3744 struct inode *inode = file_inode(file); 3745 struct btrfs_root *root = BTRFS_I(inode)->root; 3746 struct btrfs_ioctl_qgroup_limit_args AUTO_KFREE(sa); 3747 struct btrfs_trans_handle *trans; 3748 int ret; 3749 int err; 3750 u64 qgroupid; 3751 3752 if (!capable(CAP_SYS_ADMIN)) 3753 return -EPERM; 3754 3755 if (!btrfs_qgroup_enabled(root->fs_info)) 3756 return -ENOTCONN; 3757 3758 ret = mnt_want_write_file(file); 3759 if (ret) 3760 return ret; 3761 3762 sa = memdup_user(arg, sizeof(*sa)); 3763 if (IS_ERR(sa)) { 3764 ret = PTR_ERR(sa); 3765 goto drop_write; 3766 } 3767 3768 /* 1 BTRFS_QGROUP_LIMIT_KEY item. */ 3769 trans = btrfs_start_transaction(root, 1); 3770 if (IS_ERR(trans)) { 3771 ret = PTR_ERR(trans); 3772 goto drop_write; 3773 } 3774 3775 qgroupid = sa->qgroupid; 3776 if (!qgroupid) { 3777 /* take the current subvol as qgroup */ 3778 qgroupid = btrfs_root_id(root); 3779 } 3780 3781 ret = btrfs_limit_qgroup(trans, qgroupid, &sa->lim); 3782 3783 err = btrfs_end_transaction(trans); 3784 if (err && !ret) 3785 ret = err; 3786 3787 drop_write: 3788 mnt_drop_write_file(file); 3789 return ret; 3790 } 3791 3792 static long btrfs_ioctl_quota_rescan(struct file *file, void __user *arg) 3793 { 3794 struct inode *inode = file_inode(file); 3795 struct btrfs_fs_info *fs_info = inode_to_fs_info(inode); 3796 struct btrfs_ioctl_quota_rescan_args AUTO_KFREE(qsa); 3797 int ret; 3798 3799 if (!capable(CAP_SYS_ADMIN)) 3800 return -EPERM; 3801 3802 if (!btrfs_qgroup_enabled(fs_info)) 3803 return -ENOTCONN; 3804 3805 ret = mnt_want_write_file(file); 3806 if (ret) 3807 return ret; 3808 3809 qsa = memdup_user(arg, sizeof(*qsa)); 3810 if (IS_ERR(qsa)) { 3811 ret = PTR_ERR(qsa); 3812 goto drop_write; 3813 } 3814 3815 if (qsa->flags) { 3816 ret = -EINVAL; 3817 goto drop_write; 3818 } 3819 3820 ret = btrfs_qgroup_rescan(fs_info); 3821 3822 drop_write: 3823 mnt_drop_write_file(file); 3824 return ret; 3825 } 3826 3827 static long btrfs_ioctl_quota_rescan_status(struct btrfs_fs_info *fs_info, 3828 void __user *arg) 3829 { 3830 struct btrfs_ioctl_quota_rescan_args qsa = {0}; 3831 3832 if (!capable(CAP_SYS_ADMIN)) 3833 return -EPERM; 3834 3835 if (fs_info->qgroup_flags & BTRFS_QGROUP_STATUS_FLAG_RESCAN) { 3836 qsa.flags = 1; 3837 qsa.progress = fs_info->qgroup_rescan_progress.objectid; 3838 } 3839 3840 if (copy_to_user(arg, &qsa, sizeof(qsa))) 3841 return -EFAULT; 3842 3843 return 0; 3844 } 3845 3846 static long btrfs_ioctl_quota_rescan_wait(struct btrfs_fs_info *fs_info) 3847 { 3848 if (!capable(CAP_SYS_ADMIN)) 3849 return -EPERM; 3850 3851 return btrfs_qgroup_wait_for_completion(fs_info, true); 3852 } 3853 3854 static long _btrfs_ioctl_set_received_subvol(struct file *file, 3855 struct mnt_idmap *idmap, 3856 struct btrfs_ioctl_received_subvol_args *sa) 3857 { 3858 struct inode *inode = file_inode(file); 3859 struct btrfs_fs_info *fs_info = inode_to_fs_info(inode); 3860 struct btrfs_root *root = BTRFS_I(inode)->root; 3861 struct btrfs_root_item *root_item = &root->root_item; 3862 struct btrfs_trans_handle *trans; 3863 struct timespec64 ct = current_time(inode); 3864 int ret = 0; 3865 int received_uuid_changed; 3866 3867 if (!inode_owner_or_capable(idmap, inode)) 3868 return -EPERM; 3869 3870 ret = mnt_want_write_file(file); 3871 if (ret < 0) 3872 return ret; 3873 3874 down_write(&fs_info->subvol_sem); 3875 3876 if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) { 3877 ret = -EINVAL; 3878 goto out; 3879 } 3880 3881 if (btrfs_root_readonly(root)) { 3882 ret = -EROFS; 3883 goto out; 3884 } 3885 3886 received_uuid_changed = memcmp(root_item->received_uuid, sa->uuid, 3887 BTRFS_UUID_SIZE); 3888 3889 /* 3890 * Before we attempt to add the new received uuid, check if we have room 3891 * for it in case there's already an item. If the size of the existing 3892 * item plus this root's ID (u64) exceeds the maximum item size, we can 3893 * return here without the need to abort a transaction. If we don't do 3894 * this check, the btrfs_uuid_tree_add() call below would fail with 3895 * -EOVERFLOW and result in a transaction abort. Malicious users could 3896 * exploit this to turn the fs into RO mode. 3897 */ 3898 if (received_uuid_changed && !btrfs_is_empty_uuid(sa->uuid)) { 3899 ret = btrfs_uuid_tree_check_overflow(fs_info, sa->uuid, 3900 BTRFS_UUID_KEY_RECEIVED_SUBVOL); 3901 if (ret < 0) 3902 goto out; 3903 } 3904 3905 /* 3906 * 1 - root item 3907 * 2 - uuid items (received uuid + subvol uuid) 3908 */ 3909 trans = btrfs_start_transaction(root, 3); 3910 if (IS_ERR(trans)) { 3911 ret = PTR_ERR(trans); 3912 trans = NULL; 3913 goto out; 3914 } 3915 3916 sa->rtransid = trans->transid; 3917 sa->rtime.sec = ct.tv_sec; 3918 sa->rtime.nsec = ct.tv_nsec; 3919 3920 if (received_uuid_changed && 3921 !btrfs_is_empty_uuid(root_item->received_uuid)) { 3922 ret = btrfs_uuid_tree_remove(trans, root_item->received_uuid, 3923 BTRFS_UUID_KEY_RECEIVED_SUBVOL, 3924 btrfs_root_id(root)); 3925 if (unlikely(ret && ret != -ENOENT)) { 3926 btrfs_end_transaction(trans); 3927 goto out; 3928 } 3929 } 3930 memcpy(root_item->received_uuid, sa->uuid, BTRFS_UUID_SIZE); 3931 btrfs_set_root_stransid(root_item, sa->stransid); 3932 btrfs_set_root_rtransid(root_item, sa->rtransid); 3933 btrfs_set_stack_timespec_sec(&root_item->stime, sa->stime.sec); 3934 btrfs_set_stack_timespec_nsec(&root_item->stime, sa->stime.nsec); 3935 btrfs_set_stack_timespec_sec(&root_item->rtime, sa->rtime.sec); 3936 btrfs_set_stack_timespec_nsec(&root_item->rtime, sa->rtime.nsec); 3937 3938 ret = btrfs_update_root(trans, fs_info->tree_root, 3939 &root->root_key, &root->root_item); 3940 if (unlikely(ret < 0)) { 3941 btrfs_abort_transaction(trans, ret); 3942 btrfs_end_transaction(trans); 3943 goto out; 3944 } 3945 if (received_uuid_changed && !btrfs_is_empty_uuid(sa->uuid)) { 3946 ret = btrfs_uuid_tree_add(trans, sa->uuid, 3947 BTRFS_UUID_KEY_RECEIVED_SUBVOL, 3948 btrfs_root_id(root)); 3949 if (unlikely(ret < 0)) { 3950 btrfs_abort_transaction(trans, ret); 3951 btrfs_end_transaction(trans); 3952 goto out; 3953 } 3954 } 3955 ret = btrfs_commit_transaction(trans); 3956 out: 3957 up_write(&fs_info->subvol_sem); 3958 mnt_drop_write_file(file); 3959 return ret; 3960 } 3961 3962 #ifdef CONFIG_64BIT 3963 static long btrfs_ioctl_set_received_subvol_32(struct file *file, 3964 void __user *arg) 3965 { 3966 struct btrfs_ioctl_received_subvol_args_32 AUTO_KFREE(args32); 3967 struct btrfs_ioctl_received_subvol_args AUTO_KFREE(args64); 3968 int ret = 0; 3969 3970 args32 = memdup_user(arg, sizeof(*args32)); 3971 if (IS_ERR(args32)) 3972 return PTR_ERR(args32); 3973 3974 args64 = kmalloc_obj(*args64); 3975 if (!args64) 3976 return -ENOMEM; 3977 3978 memcpy(args64->uuid, args32->uuid, BTRFS_UUID_SIZE); 3979 args64->stransid = args32->stransid; 3980 args64->rtransid = args32->rtransid; 3981 args64->stime.sec = args32->stime.sec; 3982 args64->stime.nsec = args32->stime.nsec; 3983 args64->rtime.sec = args32->rtime.sec; 3984 args64->rtime.nsec = args32->rtime.nsec; 3985 args64->flags = args32->flags; 3986 3987 ret = _btrfs_ioctl_set_received_subvol(file, file_mnt_idmap(file), args64); 3988 if (ret) 3989 return ret; 3990 3991 memcpy(args32->uuid, args64->uuid, BTRFS_UUID_SIZE); 3992 args32->stransid = args64->stransid; 3993 args32->rtransid = args64->rtransid; 3994 args32->stime.sec = args64->stime.sec; 3995 args32->stime.nsec = args64->stime.nsec; 3996 args32->rtime.sec = args64->rtime.sec; 3997 args32->rtime.nsec = args64->rtime.nsec; 3998 args32->flags = args64->flags; 3999 4000 ret = copy_to_user(arg, args32, sizeof(*args32)); 4001 if (ret) 4002 return -EFAULT; 4003 4004 return 0; 4005 } 4006 #endif 4007 4008 static long btrfs_ioctl_set_received_subvol(struct file *file, 4009 void __user *arg) 4010 { 4011 struct btrfs_ioctl_received_subvol_args AUTO_KFREE(sa); 4012 int ret = 0; 4013 4014 sa = memdup_user(arg, sizeof(*sa)); 4015 if (IS_ERR(sa)) 4016 return PTR_ERR(sa); 4017 4018 ret = _btrfs_ioctl_set_received_subvol(file, file_mnt_idmap(file), sa); 4019 if (ret) 4020 return ret; 4021 4022 ret = copy_to_user(arg, sa, sizeof(*sa)); 4023 if (ret) 4024 return -EFAULT; 4025 4026 return 0; 4027 } 4028 4029 static int btrfs_ioctl_get_fslabel(struct btrfs_fs_info *fs_info, 4030 void __user *arg) 4031 { 4032 size_t len; 4033 int ret; 4034 char label[BTRFS_LABEL_SIZE]; 4035 4036 spin_lock(&fs_info->super_lock); 4037 memcpy(label, fs_info->super_copy->label, BTRFS_LABEL_SIZE); 4038 spin_unlock(&fs_info->super_lock); 4039 4040 len = strnlen(label, BTRFS_LABEL_SIZE); 4041 4042 if (len == BTRFS_LABEL_SIZE) { 4043 btrfs_warn(fs_info, 4044 "label is too long, return the first %zu bytes", 4045 --len); 4046 } 4047 4048 ret = copy_to_user(arg, label, len); 4049 4050 return ret ? -EFAULT : 0; 4051 } 4052 4053 static int btrfs_ioctl_set_fslabel(struct file *file, void __user *arg) 4054 { 4055 struct inode *inode = file_inode(file); 4056 struct btrfs_fs_info *fs_info = inode_to_fs_info(inode); 4057 struct btrfs_root *root = BTRFS_I(inode)->root; 4058 struct btrfs_super_block *super_block = fs_info->super_copy; 4059 struct btrfs_trans_handle *trans; 4060 char label[BTRFS_LABEL_SIZE]; 4061 int ret; 4062 4063 if (!capable(CAP_SYS_ADMIN)) 4064 return -EPERM; 4065 4066 if (copy_from_user(label, arg, sizeof(label))) 4067 return -EFAULT; 4068 4069 if (strnlen(label, BTRFS_LABEL_SIZE) == BTRFS_LABEL_SIZE) { 4070 btrfs_err(fs_info, 4071 "unable to set label with more than %d bytes", 4072 BTRFS_LABEL_SIZE - 1); 4073 return -EINVAL; 4074 } 4075 4076 ret = mnt_want_write_file(file); 4077 if (ret) 4078 return ret; 4079 4080 trans = btrfs_start_transaction(root, 0); 4081 if (IS_ERR(trans)) { 4082 ret = PTR_ERR(trans); 4083 goto out_unlock; 4084 } 4085 4086 spin_lock(&fs_info->super_lock); 4087 strscpy(super_block->label, label); 4088 spin_unlock(&fs_info->super_lock); 4089 ret = btrfs_commit_transaction(trans); 4090 4091 out_unlock: 4092 mnt_drop_write_file(file); 4093 return ret; 4094 } 4095 4096 #define INIT_FEATURE_FLAGS(suffix) \ 4097 { .compat_flags = BTRFS_FEATURE_COMPAT_##suffix, \ 4098 .compat_ro_flags = BTRFS_FEATURE_COMPAT_RO_##suffix, \ 4099 .incompat_flags = BTRFS_FEATURE_INCOMPAT_##suffix } 4100 4101 int btrfs_ioctl_get_supported_features(void __user *arg) 4102 { 4103 static const struct btrfs_ioctl_feature_flags features[3] = { 4104 INIT_FEATURE_FLAGS(SUPP), 4105 INIT_FEATURE_FLAGS(SAFE_SET), 4106 INIT_FEATURE_FLAGS(SAFE_CLEAR) 4107 }; 4108 4109 if (copy_to_user(arg, &features, sizeof(features))) 4110 return -EFAULT; 4111 4112 return 0; 4113 } 4114 4115 static int btrfs_ioctl_get_features(struct btrfs_fs_info *fs_info, 4116 void __user *arg) 4117 { 4118 struct btrfs_super_block *super_block = fs_info->super_copy; 4119 struct btrfs_ioctl_feature_flags features; 4120 4121 features.compat_flags = btrfs_super_compat_flags(super_block); 4122 features.compat_ro_flags = btrfs_super_compat_ro_flags(super_block); 4123 features.incompat_flags = btrfs_super_incompat_flags(super_block); 4124 4125 if (copy_to_user(arg, &features, sizeof(features))) 4126 return -EFAULT; 4127 4128 return 0; 4129 } 4130 4131 static int check_feature_bits(const struct btrfs_fs_info *fs_info, 4132 enum btrfs_feature_set set, 4133 u64 change_mask, u64 flags, u64 supported_flags, 4134 u64 safe_set, u64 safe_clear) 4135 { 4136 const char *type = btrfs_feature_set_name(set); 4137 const char AUTO_KFREE(names); 4138 u64 disallowed, unsupported; 4139 u64 set_mask = flags & change_mask; 4140 u64 clear_mask = ~flags & change_mask; 4141 4142 unsupported = set_mask & ~supported_flags; 4143 if (unsupported) { 4144 names = btrfs_printable_features(set, unsupported); 4145 if (names) 4146 btrfs_warn(fs_info, 4147 "this kernel does not support the %s feature bit%s", 4148 names, strchr(names, ',') ? "s" : ""); 4149 else 4150 btrfs_warn(fs_info, 4151 "this kernel does not support %s bits 0x%llx", 4152 type, unsupported); 4153 return -EOPNOTSUPP; 4154 } 4155 4156 disallowed = set_mask & ~safe_set; 4157 if (disallowed) { 4158 names = btrfs_printable_features(set, disallowed); 4159 if (names) 4160 btrfs_warn(fs_info, 4161 "can't set the %s feature bit%s while mounted", 4162 names, strchr(names, ',') ? "s" : ""); 4163 else 4164 btrfs_warn(fs_info, 4165 "can't set %s bits 0x%llx while mounted", 4166 type, disallowed); 4167 return -EPERM; 4168 } 4169 4170 disallowed = clear_mask & ~safe_clear; 4171 if (disallowed) { 4172 names = btrfs_printable_features(set, disallowed); 4173 if (names) 4174 btrfs_warn(fs_info, 4175 "can't clear the %s feature bit%s while mounted", 4176 names, strchr(names, ',') ? "s" : ""); 4177 else 4178 btrfs_warn(fs_info, 4179 "can't clear %s bits 0x%llx while mounted", 4180 type, disallowed); 4181 return -EPERM; 4182 } 4183 4184 return 0; 4185 } 4186 4187 #define check_feature(fs_info, change_mask, flags, mask_base) \ 4188 check_feature_bits(fs_info, FEAT_##mask_base, change_mask, flags, \ 4189 BTRFS_FEATURE_ ## mask_base ## _SUPP, \ 4190 BTRFS_FEATURE_ ## mask_base ## _SAFE_SET, \ 4191 BTRFS_FEATURE_ ## mask_base ## _SAFE_CLEAR) 4192 4193 static int btrfs_ioctl_set_features(struct file *file, void __user *arg) 4194 { 4195 struct inode *inode = file_inode(file); 4196 struct btrfs_fs_info *fs_info = inode_to_fs_info(inode); 4197 struct btrfs_root *root = BTRFS_I(inode)->root; 4198 struct btrfs_super_block *super_block = fs_info->super_copy; 4199 struct btrfs_ioctl_feature_flags flags[2]; 4200 struct btrfs_trans_handle *trans; 4201 u64 newflags; 4202 int ret; 4203 4204 if (!capable(CAP_SYS_ADMIN)) 4205 return -EPERM; 4206 4207 if (copy_from_user(flags, arg, sizeof(flags))) 4208 return -EFAULT; 4209 4210 /* Nothing to do */ 4211 if (!flags[0].compat_flags && !flags[0].compat_ro_flags && 4212 !flags[0].incompat_flags) 4213 return 0; 4214 4215 ret = check_feature(fs_info, flags[0].compat_flags, 4216 flags[1].compat_flags, COMPAT); 4217 if (ret) 4218 return ret; 4219 4220 ret = check_feature(fs_info, flags[0].compat_ro_flags, 4221 flags[1].compat_ro_flags, COMPAT_RO); 4222 if (ret) 4223 return ret; 4224 4225 ret = check_feature(fs_info, flags[0].incompat_flags, 4226 flags[1].incompat_flags, INCOMPAT); 4227 if (ret) 4228 return ret; 4229 4230 ret = mnt_want_write_file(file); 4231 if (ret) 4232 return ret; 4233 4234 trans = btrfs_start_transaction(root, 0); 4235 if (IS_ERR(trans)) { 4236 ret = PTR_ERR(trans); 4237 goto out_drop_write; 4238 } 4239 4240 spin_lock(&fs_info->super_lock); 4241 newflags = btrfs_super_compat_flags(super_block); 4242 newflags |= flags[0].compat_flags & flags[1].compat_flags; 4243 newflags &= ~(flags[0].compat_flags & ~flags[1].compat_flags); 4244 btrfs_set_super_compat_flags(super_block, newflags); 4245 4246 newflags = btrfs_super_compat_ro_flags(super_block); 4247 newflags |= flags[0].compat_ro_flags & flags[1].compat_ro_flags; 4248 newflags &= ~(flags[0].compat_ro_flags & ~flags[1].compat_ro_flags); 4249 btrfs_set_super_compat_ro_flags(super_block, newflags); 4250 4251 newflags = btrfs_super_incompat_flags(super_block); 4252 newflags |= flags[0].incompat_flags & flags[1].incompat_flags; 4253 newflags &= ~(flags[0].incompat_flags & ~flags[1].incompat_flags); 4254 btrfs_set_super_incompat_flags(super_block, newflags); 4255 spin_unlock(&fs_info->super_lock); 4256 4257 ret = btrfs_commit_transaction(trans); 4258 out_drop_write: 4259 mnt_drop_write_file(file); 4260 4261 return ret; 4262 } 4263 4264 static int _btrfs_ioctl_send(struct btrfs_root *root, void __user *argp, bool compat) 4265 { 4266 struct btrfs_ioctl_send_args AUTO_KFREE(arg); 4267 4268 if (compat) { 4269 #if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT) 4270 int ret; 4271 struct btrfs_ioctl_send_args_32 args32 = { 0 }; 4272 4273 ret = copy_from_user(&args32, argp, sizeof(args32)); 4274 if (ret) 4275 return -EFAULT; 4276 arg = kzalloc_obj(*arg); 4277 if (!arg) 4278 return -ENOMEM; 4279 arg->send_fd = args32.send_fd; 4280 arg->clone_sources_count = args32.clone_sources_count; 4281 arg->clone_sources = compat_ptr(args32.clone_sources); 4282 arg->parent_root = args32.parent_root; 4283 arg->flags = args32.flags; 4284 arg->version = args32.version; 4285 memcpy(arg->reserved, args32.reserved, 4286 sizeof(args32.reserved)); 4287 #else 4288 return -ENOTTY; 4289 #endif 4290 } else { 4291 arg = memdup_user(argp, sizeof(*arg)); 4292 if (IS_ERR(arg)) 4293 return PTR_ERR(arg); 4294 } 4295 return btrfs_ioctl_send(root, arg); 4296 } 4297 4298 static int btrfs_ioctl_encoded_read(struct file *file, void __user *argp, 4299 bool compat) 4300 { 4301 struct btrfs_ioctl_encoded_io_args args = { 0 }; 4302 size_t copy_end_kernel = offsetofend(struct btrfs_ioctl_encoded_io_args, 4303 flags); 4304 size_t copy_end; 4305 struct btrfs_inode *inode = BTRFS_I(file_inode(file)); 4306 struct btrfs_fs_info *fs_info = inode->root->fs_info; 4307 struct extent_io_tree *io_tree = &inode->io_tree; 4308 struct iovec iovstack[UIO_FASTIOV]; 4309 struct iovec *iov = iovstack; 4310 struct iov_iter iter; 4311 loff_t pos; 4312 struct kiocb kiocb; 4313 ssize_t ret; 4314 u64 disk_bytenr, disk_io_size; 4315 struct extent_state *cached_state = NULL; 4316 4317 if (!capable(CAP_SYS_ADMIN)) { 4318 ret = -EPERM; 4319 goto out_acct; 4320 } 4321 4322 if (compat) { 4323 #if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT) 4324 struct btrfs_ioctl_encoded_io_args_32 args32; 4325 4326 copy_end = offsetofend(struct btrfs_ioctl_encoded_io_args_32, 4327 flags); 4328 if (copy_from_user(&args32, argp, copy_end)) { 4329 ret = -EFAULT; 4330 goto out_acct; 4331 } 4332 args.iov = compat_ptr(args32.iov); 4333 args.iovcnt = args32.iovcnt; 4334 args.offset = args32.offset; 4335 args.flags = args32.flags; 4336 #else 4337 return -ENOTTY; 4338 #endif 4339 } else { 4340 copy_end = copy_end_kernel; 4341 if (copy_from_user(&args, argp, copy_end)) { 4342 ret = -EFAULT; 4343 goto out_acct; 4344 } 4345 } 4346 if (args.flags != 0) { 4347 ret = -EINVAL; 4348 goto out_acct; 4349 } 4350 4351 ret = import_iovec(ITER_DEST, args.iov, args.iovcnt, ARRAY_SIZE(iovstack), 4352 &iov, &iter); 4353 if (ret < 0) 4354 goto out_acct; 4355 4356 if (iov_iter_count(&iter) == 0) { 4357 ret = 0; 4358 goto out_iov; 4359 } 4360 pos = args.offset; 4361 ret = rw_verify_area(READ, file, &pos, args.len); 4362 if (ret < 0) 4363 goto out_iov; 4364 4365 init_sync_kiocb(&kiocb, file); 4366 kiocb.ki_pos = pos; 4367 4368 ret = btrfs_encoded_read(&kiocb, &iter, &args, &cached_state, 4369 &disk_bytenr, &disk_io_size); 4370 4371 if (ret == -EIOCBQUEUED) { 4372 bool unlocked = false; 4373 u64 start, lockend, count; 4374 4375 start = ALIGN_DOWN(kiocb.ki_pos, fs_info->sectorsize); 4376 lockend = start + BTRFS_MAX_UNCOMPRESSED - 1; 4377 4378 if (args.compression) 4379 count = disk_io_size; 4380 else 4381 count = args.len; 4382 4383 ret = btrfs_encoded_read_regular(&kiocb, &iter, start, lockend, 4384 &cached_state, disk_bytenr, 4385 disk_io_size, count, 4386 args.compression, &unlocked); 4387 4388 if (!unlocked) { 4389 btrfs_unlock_extent(io_tree, start, lockend, &cached_state); 4390 btrfs_inode_unlock(inode, BTRFS_ILOCK_SHARED); 4391 } 4392 } 4393 4394 if (ret >= 0) { 4395 fsnotify_access(file); 4396 if (copy_to_user(argp + copy_end, 4397 (char *)&args + copy_end_kernel, 4398 sizeof(args) - copy_end_kernel)) 4399 ret = -EFAULT; 4400 } 4401 4402 out_iov: 4403 kfree(iov); 4404 out_acct: 4405 if (ret > 0) 4406 add_rchar(current, ret); 4407 inc_syscr(current); 4408 return ret; 4409 } 4410 4411 static int btrfs_ioctl_encoded_write(struct file *file, void __user *argp, bool compat) 4412 { 4413 struct btrfs_ioctl_encoded_io_args args; 4414 struct iovec iovstack[UIO_FASTIOV]; 4415 struct iovec *iov = iovstack; 4416 struct iov_iter iter; 4417 loff_t pos; 4418 struct kiocb kiocb; 4419 ssize_t ret; 4420 4421 if (!capable(CAP_SYS_ADMIN)) { 4422 ret = -EPERM; 4423 goto out_acct; 4424 } 4425 4426 if (!(file->f_mode & FMODE_WRITE)) { 4427 ret = -EBADF; 4428 goto out_acct; 4429 } 4430 4431 if (compat) { 4432 #if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT) 4433 struct btrfs_ioctl_encoded_io_args_32 args32; 4434 4435 if (copy_from_user(&args32, argp, sizeof(args32))) { 4436 ret = -EFAULT; 4437 goto out_acct; 4438 } 4439 args.iov = compat_ptr(args32.iov); 4440 args.iovcnt = args32.iovcnt; 4441 args.offset = args32.offset; 4442 args.flags = args32.flags; 4443 args.len = args32.len; 4444 args.unencoded_len = args32.unencoded_len; 4445 args.unencoded_offset = args32.unencoded_offset; 4446 args.compression = args32.compression; 4447 args.encryption = args32.encryption; 4448 memcpy(args.reserved, args32.reserved, sizeof(args.reserved)); 4449 #else 4450 return -ENOTTY; 4451 #endif 4452 } else { 4453 if (copy_from_user(&args, argp, sizeof(args))) { 4454 ret = -EFAULT; 4455 goto out_acct; 4456 } 4457 } 4458 4459 ret = -EINVAL; 4460 if (args.flags != 0) 4461 goto out_acct; 4462 if (memchr_inv(args.reserved, 0, sizeof(args.reserved))) 4463 goto out_acct; 4464 if (args.compression == BTRFS_ENCODED_IO_COMPRESSION_NONE && 4465 args.encryption == BTRFS_ENCODED_IO_ENCRYPTION_NONE) 4466 goto out_acct; 4467 if (args.compression >= BTRFS_ENCODED_IO_COMPRESSION_TYPES || 4468 args.encryption >= BTRFS_ENCODED_IO_ENCRYPTION_TYPES) 4469 goto out_acct; 4470 if (args.unencoded_offset > args.unencoded_len) 4471 goto out_acct; 4472 if (args.len > args.unencoded_len - args.unencoded_offset) 4473 goto out_acct; 4474 4475 ret = import_iovec(ITER_SOURCE, args.iov, args.iovcnt, ARRAY_SIZE(iovstack), 4476 &iov, &iter); 4477 if (ret < 0) 4478 goto out_acct; 4479 4480 if (iov_iter_count(&iter) == 0) { 4481 ret = 0; 4482 goto out_iov; 4483 } 4484 pos = args.offset; 4485 ret = rw_verify_area(WRITE, file, &pos, args.len); 4486 if (ret < 0) 4487 goto out_iov; 4488 4489 init_sync_kiocb(&kiocb, file); 4490 ret = kiocb_set_rw_flags(&kiocb, 0, WRITE); 4491 if (ret) 4492 goto out_iov; 4493 kiocb.ki_pos = pos; 4494 4495 file_start_write(file); 4496 4497 ret = btrfs_do_write_iter(&kiocb, &iter, &args); 4498 if (ret > 0) 4499 fsnotify_modify(file); 4500 4501 file_end_write(file); 4502 out_iov: 4503 kfree(iov); 4504 out_acct: 4505 if (ret > 0) 4506 add_wchar(current, ret); 4507 inc_syscw(current); 4508 return ret; 4509 } 4510 4511 struct btrfs_uring_encoded_data { 4512 struct btrfs_ioctl_encoded_io_args args; 4513 struct iovec iovstack[UIO_FASTIOV]; 4514 struct iovec *iov; 4515 struct iov_iter iter; 4516 }; 4517 4518 /* 4519 * Context that's attached to an encoded read io_uring command, in cmd->pdu. It 4520 * contains the fields in btrfs_uring_read_extent that are necessary to finish 4521 * off and cleanup the I/O in btrfs_uring_read_finished. 4522 */ 4523 struct btrfs_uring_priv { 4524 struct io_uring_cmd *cmd; 4525 struct page **pages; 4526 unsigned long nr_pages; 4527 struct kiocb iocb; 4528 struct iovec *iov; 4529 struct iov_iter iter; 4530 struct extent_state *cached_state; 4531 u64 count; 4532 u64 start; 4533 u64 lockend; 4534 int err; 4535 bool compressed; 4536 }; 4537 4538 struct io_btrfs_cmd { 4539 struct btrfs_uring_encoded_data *data; 4540 struct btrfs_uring_priv *priv; 4541 }; 4542 4543 static void btrfs_uring_read_finished(struct io_tw_req tw_req, io_tw_token_t tw) 4544 { 4545 struct io_uring_cmd *cmd = io_uring_cmd_from_tw(tw_req); 4546 struct io_btrfs_cmd *bc = io_uring_cmd_to_pdu(cmd, struct io_btrfs_cmd); 4547 struct btrfs_uring_priv *priv = bc->priv; 4548 struct btrfs_inode *inode = BTRFS_I(file_inode(priv->iocb.ki_filp)); 4549 struct extent_io_tree *io_tree = &inode->io_tree; 4550 pgoff_t index; 4551 u64 cur; 4552 size_t page_offset; 4553 ssize_t ret; 4554 4555 /* The inode lock has already been acquired in btrfs_uring_read_extent. */ 4556 btrfs_lockdep_inode_acquire(inode, i_rwsem); 4557 4558 if (priv->err) { 4559 ret = priv->err; 4560 goto out; 4561 } 4562 4563 if (priv->compressed) { 4564 index = 0; 4565 page_offset = 0; 4566 } else { 4567 index = (priv->iocb.ki_pos - priv->start) >> PAGE_SHIFT; 4568 page_offset = offset_in_page(priv->iocb.ki_pos - priv->start); 4569 } 4570 cur = 0; 4571 while (cur < priv->count) { 4572 size_t bytes = min_t(size_t, priv->count - cur, PAGE_SIZE - page_offset); 4573 4574 if (copy_page_to_iter(priv->pages[index], page_offset, bytes, 4575 &priv->iter) != bytes) { 4576 ret = -EFAULT; 4577 goto out; 4578 } 4579 4580 index++; 4581 cur += bytes; 4582 page_offset = 0; 4583 } 4584 ret = priv->count; 4585 4586 out: 4587 btrfs_unlock_extent(io_tree, priv->start, priv->lockend, &priv->cached_state); 4588 btrfs_inode_unlock(inode, BTRFS_ILOCK_SHARED); 4589 4590 io_uring_cmd_done(cmd, ret, IO_URING_CMD_TASK_WORK_ISSUE_FLAGS); 4591 add_rchar(current, ret); 4592 4593 for (index = 0; index < priv->nr_pages; index++) 4594 __free_page(priv->pages[index]); 4595 4596 kfree(priv->pages); 4597 kfree(priv->iov); 4598 kfree(priv); 4599 kfree(bc->data); 4600 } 4601 4602 void btrfs_uring_read_extent_endio(void *ctx, int err) 4603 { 4604 struct btrfs_uring_priv *priv = ctx; 4605 struct io_btrfs_cmd *bc = io_uring_cmd_to_pdu(priv->cmd, struct io_btrfs_cmd); 4606 4607 priv->err = err; 4608 bc->priv = priv; 4609 4610 io_uring_cmd_complete_in_task(priv->cmd, btrfs_uring_read_finished); 4611 } 4612 4613 static int btrfs_uring_read_extent(struct kiocb *iocb, struct iov_iter *iter, 4614 u64 start, u64 lockend, 4615 struct extent_state *cached_state, 4616 u64 disk_bytenr, u64 disk_io_size, 4617 size_t count, bool compressed, 4618 struct iovec *iov, struct io_uring_cmd *cmd) 4619 { 4620 struct btrfs_inode *inode = BTRFS_I(file_inode(iocb->ki_filp)); 4621 struct extent_io_tree *io_tree = &inode->io_tree; 4622 struct page **pages = NULL; 4623 struct btrfs_uring_priv *priv = NULL; 4624 unsigned long nr_pages; 4625 int ret; 4626 4627 nr_pages = DIV_ROUND_UP(disk_io_size, PAGE_SIZE); 4628 pages = kzalloc_objs(struct page *, nr_pages, GFP_NOFS); 4629 if (!pages) 4630 return -ENOMEM; 4631 ret = btrfs_alloc_page_array(nr_pages, pages, GFP_NOFS); 4632 if (ret) { 4633 ret = -ENOMEM; 4634 goto out_fail; 4635 } 4636 4637 priv = kmalloc_obj(*priv, GFP_NOFS); 4638 if (!priv) { 4639 ret = -ENOMEM; 4640 goto out_fail; 4641 } 4642 4643 priv->iocb = *iocb; 4644 priv->iov = iov; 4645 priv->iter = *iter; 4646 priv->count = count; 4647 priv->cmd = cmd; 4648 priv->cached_state = cached_state; 4649 priv->compressed = compressed; 4650 priv->nr_pages = nr_pages; 4651 priv->pages = pages; 4652 priv->start = start; 4653 priv->lockend = lockend; 4654 priv->err = 0; 4655 4656 ret = btrfs_encoded_read_regular_fill_pages(inode, disk_bytenr, 4657 disk_io_size, pages, priv); 4658 if (ret && ret != -EIOCBQUEUED) 4659 goto out_fail; 4660 4661 /* 4662 * If we return -EIOCBQUEUED, we're deferring the cleanup to 4663 * btrfs_uring_read_finished(), which will handle unlocking the extent 4664 * and inode and freeing the allocations. 4665 */ 4666 4667 /* 4668 * We're returning to userspace with the inode lock held, and that's 4669 * okay - it'll get unlocked in a worker thread. Call 4670 * btrfs_lockdep_inode_release() to avoid confusing lockdep. 4671 */ 4672 btrfs_lockdep_inode_release(inode, i_rwsem); 4673 4674 return -EIOCBQUEUED; 4675 4676 out_fail: 4677 btrfs_unlock_extent(io_tree, start, lockend, &cached_state); 4678 btrfs_inode_unlock(inode, BTRFS_ILOCK_SHARED); 4679 kfree(priv); 4680 for (int i = 0; i < nr_pages; i++) { 4681 if (pages[i]) 4682 __free_page(pages[i]); 4683 } 4684 kfree(pages); 4685 return ret; 4686 } 4687 4688 static int btrfs_uring_encoded_read(struct io_uring_cmd *cmd, unsigned int issue_flags) 4689 { 4690 struct file *file = cmd->file; 4691 struct btrfs_inode *inode = BTRFS_I(file->f_inode); 4692 struct extent_io_tree *io_tree = &inode->io_tree; 4693 struct btrfs_fs_info *fs_info = inode->root->fs_info; 4694 size_t copy_end_kernel = offsetofend(struct btrfs_ioctl_encoded_io_args, flags); 4695 size_t copy_end; 4696 int ret; 4697 u64 disk_bytenr, disk_io_size; 4698 loff_t pos; 4699 struct kiocb kiocb; 4700 struct extent_state *cached_state = NULL; 4701 u64 start, lockend; 4702 void __user *sqe_addr; 4703 struct io_btrfs_cmd *bc = io_uring_cmd_to_pdu(cmd, struct io_btrfs_cmd); 4704 struct btrfs_uring_encoded_data *data = NULL; 4705 4706 if (cmd->flags & IORING_URING_CMD_REISSUE) 4707 data = bc->data; 4708 4709 if (!capable(CAP_SYS_ADMIN)) { 4710 ret = -EPERM; 4711 goto out_acct; 4712 } 4713 sqe_addr = u64_to_user_ptr(READ_ONCE(cmd->sqe->addr)); 4714 4715 if (issue_flags & IO_URING_F_COMPAT) { 4716 #if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT) 4717 copy_end = offsetofend(struct btrfs_ioctl_encoded_io_args_32, flags); 4718 #else 4719 ret = -ENOTTY; 4720 goto out_acct; 4721 #endif 4722 } else { 4723 copy_end = copy_end_kernel; 4724 } 4725 4726 if (!data) { 4727 data = kzalloc_obj(*data, GFP_NOFS); 4728 if (!data) { 4729 ret = -ENOMEM; 4730 goto out_acct; 4731 } 4732 4733 bc->data = data; 4734 4735 if (issue_flags & IO_URING_F_COMPAT) { 4736 #if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT) 4737 struct btrfs_ioctl_encoded_io_args_32 args32; 4738 4739 if (copy_from_user(&args32, sqe_addr, copy_end)) { 4740 ret = -EFAULT; 4741 goto out_acct; 4742 } 4743 4744 data->args.iov = compat_ptr(args32.iov); 4745 data->args.iovcnt = args32.iovcnt; 4746 data->args.offset = args32.offset; 4747 data->args.flags = args32.flags; 4748 #endif 4749 } else { 4750 if (copy_from_user(&data->args, sqe_addr, copy_end)) { 4751 ret = -EFAULT; 4752 goto out_acct; 4753 } 4754 } 4755 4756 if (data->args.flags != 0) { 4757 ret = -EINVAL; 4758 goto out_acct; 4759 } 4760 4761 data->iov = data->iovstack; 4762 ret = import_iovec(ITER_DEST, data->args.iov, data->args.iovcnt, 4763 ARRAY_SIZE(data->iovstack), &data->iov, 4764 &data->iter); 4765 if (ret < 0) 4766 goto out_acct; 4767 4768 if (iov_iter_count(&data->iter) == 0) { 4769 ret = 0; 4770 goto out_free; 4771 } 4772 } 4773 4774 pos = data->args.offset; 4775 ret = rw_verify_area(READ, file, &pos, data->args.len); 4776 if (ret < 0) 4777 goto out_free; 4778 4779 init_sync_kiocb(&kiocb, file); 4780 kiocb.ki_pos = pos; 4781 4782 if (issue_flags & IO_URING_F_NONBLOCK) 4783 kiocb.ki_flags |= IOCB_NOWAIT; 4784 4785 start = ALIGN_DOWN(pos, fs_info->sectorsize); 4786 lockend = start + BTRFS_MAX_UNCOMPRESSED - 1; 4787 4788 ret = btrfs_encoded_read(&kiocb, &data->iter, &data->args, &cached_state, 4789 &disk_bytenr, &disk_io_size); 4790 if (ret == -EAGAIN) 4791 goto out_acct; 4792 if (ret < 0 && ret != -EIOCBQUEUED) 4793 goto out_free; 4794 4795 file_accessed(file); 4796 4797 if (copy_to_user(sqe_addr + copy_end, 4798 (const char *)&data->args + copy_end_kernel, 4799 sizeof(data->args) - copy_end_kernel)) { 4800 if (ret == -EIOCBQUEUED) { 4801 btrfs_unlock_extent(io_tree, start, lockend, &cached_state); 4802 btrfs_inode_unlock(inode, BTRFS_ILOCK_SHARED); 4803 } 4804 ret = -EFAULT; 4805 goto out_free; 4806 } 4807 4808 if (ret == -EIOCBQUEUED) { 4809 u64 count = min_t(u64, iov_iter_count(&data->iter), disk_io_size); 4810 4811 /* Match ioctl by not returning past EOF if uncompressed. */ 4812 if (!data->args.compression) 4813 count = min_t(u64, count, data->args.len); 4814 4815 ret = btrfs_uring_read_extent(&kiocb, &data->iter, start, lockend, 4816 cached_state, disk_bytenr, disk_io_size, 4817 count, data->args.compression, 4818 data->iov, cmd); 4819 4820 goto out_acct; 4821 } 4822 4823 out_free: 4824 kfree(data->iov); 4825 4826 out_acct: 4827 if (ret > 0) 4828 add_rchar(current, ret); 4829 inc_syscr(current); 4830 4831 if (ret != -EIOCBQUEUED && ret != -EAGAIN) 4832 kfree(data); 4833 4834 return ret; 4835 } 4836 4837 static int btrfs_uring_encoded_write(struct io_uring_cmd *cmd, unsigned int issue_flags) 4838 { 4839 struct file *file = cmd->file; 4840 loff_t pos; 4841 struct kiocb kiocb; 4842 ssize_t ret; 4843 void __user *sqe_addr; 4844 struct io_btrfs_cmd *bc = io_uring_cmd_to_pdu(cmd, struct io_btrfs_cmd); 4845 struct btrfs_uring_encoded_data *data = NULL; 4846 4847 if (cmd->flags & IORING_URING_CMD_REISSUE) 4848 data = bc->data; 4849 4850 if (!capable(CAP_SYS_ADMIN)) { 4851 ret = -EPERM; 4852 goto out_acct; 4853 } 4854 sqe_addr = u64_to_user_ptr(READ_ONCE(cmd->sqe->addr)); 4855 4856 if (!(file->f_mode & FMODE_WRITE)) { 4857 ret = -EBADF; 4858 goto out_acct; 4859 } 4860 4861 if (!data) { 4862 data = kzalloc_obj(*data, GFP_NOFS); 4863 if (!data) { 4864 ret = -ENOMEM; 4865 goto out_acct; 4866 } 4867 4868 bc->data = data; 4869 4870 if (issue_flags & IO_URING_F_COMPAT) { 4871 #if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT) 4872 struct btrfs_ioctl_encoded_io_args_32 args32; 4873 4874 if (copy_from_user(&args32, sqe_addr, sizeof(args32))) { 4875 ret = -EFAULT; 4876 goto out_acct; 4877 } 4878 data->args.iov = compat_ptr(args32.iov); 4879 data->args.iovcnt = args32.iovcnt; 4880 data->args.offset = args32.offset; 4881 data->args.flags = args32.flags; 4882 data->args.len = args32.len; 4883 data->args.unencoded_len = args32.unencoded_len; 4884 data->args.unencoded_offset = args32.unencoded_offset; 4885 data->args.compression = args32.compression; 4886 data->args.encryption = args32.encryption; 4887 memcpy(data->args.reserved, args32.reserved, 4888 sizeof(data->args.reserved)); 4889 #else 4890 ret = -ENOTTY; 4891 goto out_acct; 4892 #endif 4893 } else { 4894 if (copy_from_user(&data->args, sqe_addr, sizeof(data->args))) { 4895 ret = -EFAULT; 4896 goto out_acct; 4897 } 4898 } 4899 4900 ret = -EINVAL; 4901 if (data->args.flags != 0) 4902 goto out_acct; 4903 if (memchr_inv(data->args.reserved, 0, sizeof(data->args.reserved))) 4904 goto out_acct; 4905 if (data->args.compression == BTRFS_ENCODED_IO_COMPRESSION_NONE && 4906 data->args.encryption == BTRFS_ENCODED_IO_ENCRYPTION_NONE) 4907 goto out_acct; 4908 if (data->args.compression >= BTRFS_ENCODED_IO_COMPRESSION_TYPES || 4909 data->args.encryption >= BTRFS_ENCODED_IO_ENCRYPTION_TYPES) 4910 goto out_acct; 4911 if (data->args.unencoded_offset > data->args.unencoded_len) 4912 goto out_acct; 4913 if (data->args.len > data->args.unencoded_len - data->args.unencoded_offset) 4914 goto out_acct; 4915 4916 data->iov = data->iovstack; 4917 ret = import_iovec(ITER_SOURCE, data->args.iov, data->args.iovcnt, 4918 ARRAY_SIZE(data->iovstack), &data->iov, 4919 &data->iter); 4920 if (ret < 0) 4921 goto out_acct; 4922 4923 if (iov_iter_count(&data->iter) == 0) { 4924 ret = 0; 4925 goto out_iov; 4926 } 4927 } 4928 4929 if (issue_flags & IO_URING_F_NONBLOCK) { 4930 ret = -EAGAIN; 4931 goto out_acct; 4932 } 4933 4934 pos = data->args.offset; 4935 ret = rw_verify_area(WRITE, file, &pos, data->args.len); 4936 if (ret < 0) 4937 goto out_iov; 4938 4939 init_sync_kiocb(&kiocb, file); 4940 ret = kiocb_set_rw_flags(&kiocb, 0, WRITE); 4941 if (ret) 4942 goto out_iov; 4943 kiocb.ki_pos = pos; 4944 4945 file_start_write(file); 4946 4947 ret = btrfs_do_write_iter(&kiocb, &data->iter, &data->args); 4948 if (ret > 0) 4949 fsnotify_modify(file); 4950 4951 file_end_write(file); 4952 out_iov: 4953 kfree(data->iov); 4954 out_acct: 4955 if (ret > 0) 4956 add_wchar(current, ret); 4957 inc_syscw(current); 4958 4959 if (ret != -EAGAIN) 4960 kfree(data); 4961 return ret; 4962 } 4963 4964 int btrfs_uring_cmd(struct io_uring_cmd *cmd, unsigned int issue_flags) 4965 { 4966 if (btrfs_is_shutdown(inode_to_fs_info(file_inode(cmd->file)))) 4967 return -EIO; 4968 4969 switch (cmd->cmd_op) { 4970 case BTRFS_IOC_ENCODED_READ: 4971 #if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT) 4972 case BTRFS_IOC_ENCODED_READ_32: 4973 #endif 4974 return btrfs_uring_encoded_read(cmd, issue_flags); 4975 4976 case BTRFS_IOC_ENCODED_WRITE: 4977 #if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT) 4978 case BTRFS_IOC_ENCODED_WRITE_32: 4979 #endif 4980 return btrfs_uring_encoded_write(cmd, issue_flags); 4981 } 4982 4983 return -EINVAL; 4984 } 4985 4986 static int btrfs_ioctl_subvol_sync(struct btrfs_fs_info *fs_info, void __user *argp) 4987 { 4988 struct btrfs_root *root; 4989 struct btrfs_ioctl_subvol_wait args = { 0 }; 4990 signed long sched_ret; 4991 int refs; 4992 u64 root_flags; 4993 bool wait_for_deletion = false; 4994 bool found = false; 4995 4996 if (copy_from_user(&args, argp, sizeof(args))) 4997 return -EFAULT; 4998 4999 switch (args.mode) { 5000 case BTRFS_SUBVOL_SYNC_WAIT_FOR_QUEUED: 5001 /* 5002 * Wait for the first one deleted that waits until all previous 5003 * are cleaned. 5004 */ 5005 spin_lock(&fs_info->trans_lock); 5006 if (!list_empty(&fs_info->dead_roots)) { 5007 root = list_last_entry(&fs_info->dead_roots, 5008 struct btrfs_root, root_list); 5009 args.subvolid = btrfs_root_id(root); 5010 found = true; 5011 } 5012 spin_unlock(&fs_info->trans_lock); 5013 if (!found) 5014 return -ENOENT; 5015 5016 fallthrough; 5017 case BTRFS_SUBVOL_SYNC_WAIT_FOR_ONE: 5018 if ((0 < args.subvolid && args.subvolid < BTRFS_FIRST_FREE_OBJECTID) || 5019 BTRFS_LAST_FREE_OBJECTID < args.subvolid) 5020 return -EINVAL; 5021 break; 5022 case BTRFS_SUBVOL_SYNC_COUNT: 5023 spin_lock(&fs_info->trans_lock); 5024 args.count = list_count_nodes(&fs_info->dead_roots); 5025 spin_unlock(&fs_info->trans_lock); 5026 if (copy_to_user(argp, &args, sizeof(args))) 5027 return -EFAULT; 5028 return 0; 5029 case BTRFS_SUBVOL_SYNC_PEEK_FIRST: 5030 spin_lock(&fs_info->trans_lock); 5031 /* Last in the list was deleted first. */ 5032 if (!list_empty(&fs_info->dead_roots)) { 5033 root = list_last_entry(&fs_info->dead_roots, 5034 struct btrfs_root, root_list); 5035 args.subvolid = btrfs_root_id(root); 5036 } else { 5037 args.subvolid = 0; 5038 } 5039 spin_unlock(&fs_info->trans_lock); 5040 if (copy_to_user(argp, &args, sizeof(args))) 5041 return -EFAULT; 5042 return 0; 5043 case BTRFS_SUBVOL_SYNC_PEEK_LAST: 5044 spin_lock(&fs_info->trans_lock); 5045 /* First in the list was deleted last. */ 5046 if (!list_empty(&fs_info->dead_roots)) { 5047 root = list_first_entry(&fs_info->dead_roots, 5048 struct btrfs_root, root_list); 5049 args.subvolid = btrfs_root_id(root); 5050 } else { 5051 args.subvolid = 0; 5052 } 5053 spin_unlock(&fs_info->trans_lock); 5054 if (copy_to_user(argp, &args, sizeof(args))) 5055 return -EFAULT; 5056 return 0; 5057 default: 5058 return -EINVAL; 5059 } 5060 5061 /* 32bit limitation: fs_roots_radix key is not wide enough. */ 5062 if (sizeof(unsigned long) != sizeof(u64) && args.subvolid > U32_MAX) 5063 return -EOVERFLOW; 5064 5065 while (1) { 5066 /* Wait for the specific one. */ 5067 if (down_read_interruptible(&fs_info->subvol_sem) == -EINTR) 5068 return -EINTR; 5069 refs = -1; 5070 spin_lock(&fs_info->fs_roots_radix_lock); 5071 root = radix_tree_lookup(&fs_info->fs_roots_radix, 5072 (unsigned long)args.subvolid); 5073 if (root) { 5074 spin_lock(&root->root_item_lock); 5075 refs = btrfs_root_refs(&root->root_item); 5076 root_flags = btrfs_root_flags(&root->root_item); 5077 spin_unlock(&root->root_item_lock); 5078 } 5079 spin_unlock(&fs_info->fs_roots_radix_lock); 5080 up_read(&fs_info->subvol_sem); 5081 5082 /* Subvolume does not exist. */ 5083 if (!root) 5084 return -ENOENT; 5085 5086 /* Subvolume not deleted at all. */ 5087 if (refs > 0) 5088 return -EEXIST; 5089 /* We've waited and now the subvolume is gone. */ 5090 if (wait_for_deletion && refs == -1) { 5091 /* Return the one we waited for as the last one. */ 5092 if (copy_to_user(argp, &args, sizeof(args))) 5093 return -EFAULT; 5094 return 0; 5095 } 5096 5097 /* Subvolume not found on the first try (deleted or never existed). */ 5098 if (refs == -1) 5099 return -ENOENT; 5100 5101 wait_for_deletion = true; 5102 ASSERT(root_flags & BTRFS_ROOT_SUBVOL_DEAD, "root_flags=0x%llx", 5103 root_flags); 5104 sched_ret = schedule_timeout_interruptible(HZ); 5105 /* Early wake up or error. */ 5106 if (sched_ret != 0) 5107 return -EINTR; 5108 } 5109 5110 return 0; 5111 } 5112 5113 static int btrfs_ioctl_shutdown(struct btrfs_fs_info *fs_info, unsigned long arg) 5114 { 5115 int ret = 0; 5116 u32 flags; 5117 5118 if (!capable(CAP_SYS_ADMIN)) 5119 return -EPERM; 5120 5121 if (get_user(flags, (u32 __user *)arg)) 5122 return -EFAULT; 5123 5124 if (flags >= BTRFS_SHUTDOWN_FLAGS_LAST) 5125 return -EINVAL; 5126 5127 if (btrfs_is_shutdown(fs_info)) 5128 return 0; 5129 5130 switch (flags) { 5131 case BTRFS_SHUTDOWN_FLAGS_LOGFLUSH: 5132 case BTRFS_SHUTDOWN_FLAGS_DEFAULT: 5133 ret = freeze_super(fs_info->sb, FREEZE_HOLDER_KERNEL, NULL); 5134 if (ret) 5135 return ret; 5136 btrfs_force_shutdown(fs_info); 5137 ret = thaw_super(fs_info->sb, FREEZE_HOLDER_KERNEL, NULL); 5138 if (ret) 5139 return ret; 5140 break; 5141 case BTRFS_SHUTDOWN_FLAGS_NOLOGFLUSH: 5142 btrfs_force_shutdown(fs_info); 5143 break; 5144 default: 5145 ret = -EINVAL; 5146 break; 5147 } 5148 return ret; 5149 } 5150 5151 #define GET_CSUMS_BUF_MAX SZ_16M 5152 5153 static int copy_csums_to_user(struct btrfs_fs_info *fs_info, u64 disk_bytenr, 5154 u64 len, u8 __user *buf) 5155 { 5156 struct btrfs_root *csum_root; 5157 struct btrfs_ordered_sum *sums; 5158 LIST_HEAD(list); 5159 const u32 csum_size = fs_info->csum_size; 5160 int ret; 5161 5162 csum_root = btrfs_csum_root(fs_info, disk_bytenr); 5163 if (unlikely(!csum_root)) { 5164 btrfs_err(fs_info, "missing csum root for extent at bytenr %llu", disk_bytenr); 5165 return -EUCLEAN; 5166 } 5167 5168 ret = btrfs_lookup_csums_list(csum_root, disk_bytenr, 5169 disk_bytenr + len - 1, &list, false); 5170 if (ret < 0) 5171 return ret; 5172 5173 ret = 0; 5174 while (!list_empty(&list)) { 5175 u64 offset; 5176 size_t copy_size; 5177 5178 sums = list_first_entry(&list, struct btrfs_ordered_sum, list); 5179 list_del(&sums->list); 5180 5181 offset = ((sums->logical - disk_bytenr) >> fs_info->sectorsize_bits) * csum_size; 5182 copy_size = (sums->len >> fs_info->sectorsize_bits) * csum_size; 5183 5184 if (copy_to_user(buf + offset, sums->sums, copy_size)) { 5185 kfree(sums); 5186 ret = -EFAULT; 5187 goto out; 5188 } 5189 5190 kfree(sums); 5191 } 5192 5193 out: 5194 while (!list_empty(&list)) { 5195 sums = list_first_entry(&list, struct btrfs_ordered_sum, list); 5196 list_del(&sums->list); 5197 kfree(sums); 5198 } 5199 return ret; 5200 } 5201 5202 static int btrfs_ioctl_get_csums(struct file *file, void __user *argp) 5203 { 5204 struct inode *vfs_inode = file_inode(file); 5205 struct btrfs_inode *inode = BTRFS_I(vfs_inode); 5206 struct btrfs_fs_info *fs_info = inode->root->fs_info; 5207 struct btrfs_root *root = inode->root; 5208 struct btrfs_ioctl_get_csums_args args = { 0 }; 5209 BTRFS_PATH_AUTO_FREE(path); 5210 const u64 ino = btrfs_ino(inode); 5211 const u32 csum_size = fs_info->csum_size; 5212 u8 __user *ubuf; 5213 u64 buf_limit; 5214 u64 buf_used = 0; 5215 u64 cur_offset; 5216 u64 end_offset; 5217 u64 prev_extent_end; 5218 struct btrfs_key key; 5219 int ret; 5220 5221 if (!(file->f_mode & FMODE_READ)) 5222 return -EBADF; 5223 5224 if (!S_ISREG(vfs_inode->i_mode)) 5225 return -EINVAL; 5226 5227 if (copy_from_user(&args, argp, sizeof(args))) 5228 return -EFAULT; 5229 5230 if (!IS_ALIGNED(args.offset, fs_info->sectorsize) || 5231 !IS_ALIGNED(args.length, fs_info->sectorsize)) 5232 return -EINVAL; 5233 if (args.length == 0) 5234 return -EINVAL; 5235 if (args.offset + args.length < args.offset) 5236 return -EOVERFLOW; 5237 if (args.flags != 0) 5238 return -EINVAL; 5239 if (args.buf_size < sizeof(struct btrfs_ioctl_get_csums_entry)) 5240 return -EINVAL; 5241 5242 buf_limit = min_t(u64, args.buf_size, GET_CSUMS_BUF_MAX); 5243 ubuf = (u8 __user *)(argp + offsetof(struct btrfs_ioctl_get_csums_args, buf)); 5244 5245 if (clear_user(ubuf, buf_limit)) 5246 return -EFAULT; 5247 5248 cur_offset = args.offset; 5249 end_offset = args.offset + args.length; 5250 5251 path = btrfs_alloc_path(); 5252 if (!path) 5253 return -ENOMEM; 5254 5255 ret = btrfs_wait_ordered_range(inode, cur_offset, args.length); 5256 if (ret) 5257 return ret; 5258 5259 ret = down_read_interruptible(&vfs_inode->i_rwsem); 5260 if (ret) 5261 return ret; 5262 5263 ret = btrfs_wait_ordered_range(inode, cur_offset, args.length); 5264 if (ret) 5265 goto out_unlock; 5266 5267 /* NODATASUM early exit. */ 5268 if (inode->flags & BTRFS_INODE_NODATASUM) { 5269 struct btrfs_ioctl_get_csums_entry entry = { 5270 .offset = cur_offset, 5271 .length = end_offset - cur_offset, 5272 .type = BTRFS_GET_CSUMS_NODATASUM, 5273 }; 5274 5275 if (copy_to_user(ubuf, &entry, sizeof(entry))) { 5276 ret = -EFAULT; 5277 goto out_unlock; 5278 } 5279 5280 buf_used = sizeof(entry); 5281 cur_offset = end_offset; 5282 goto done; 5283 } 5284 5285 prev_extent_end = cur_offset; 5286 5287 while (cur_offset < end_offset) { 5288 struct btrfs_file_extent_item *ei; 5289 struct extent_buffer *leaf; 5290 struct btrfs_ioctl_get_csums_entry entry = { 0 }; 5291 u64 extent_end; 5292 u64 disk_bytenr = 0; 5293 u64 extent_offset = 0; 5294 u64 range_start, range_len; 5295 u64 entry_csum_size; 5296 u64 key_offset; 5297 int extent_type; 5298 u8 compression; 5299 u8 encryption; 5300 5301 /* Search for the extent at or before cur_offset. */ 5302 key.objectid = ino; 5303 key.type = BTRFS_EXTENT_DATA_KEY; 5304 key.offset = cur_offset; 5305 5306 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); 5307 if (ret < 0) 5308 goto out_unlock; 5309 5310 if (ret > 0 && path->slots[0] > 0) { 5311 btrfs_item_key_to_cpu(path->nodes[0], &key, 5312 path->slots[0] - 1); 5313 if (key.objectid == ino && key.type == BTRFS_EXTENT_DATA_KEY) { 5314 path->slots[0]--; 5315 if (btrfs_file_extent_end(path) <= cur_offset) 5316 path->slots[0]++; 5317 } 5318 } 5319 5320 if (path->slots[0] >= btrfs_header_nritems(path->nodes[0])) { 5321 ret = btrfs_next_leaf(root, path); 5322 if (ret < 0) 5323 goto out_unlock; 5324 if (ret > 0) { 5325 ret = 0; 5326 btrfs_release_path(path); 5327 break; 5328 } 5329 } 5330 5331 leaf = path->nodes[0]; 5332 5333 btrfs_item_key_to_cpu(leaf, &key, path->slots[0]); 5334 if (key.objectid != ino || key.type != BTRFS_EXTENT_DATA_KEY) { 5335 btrfs_release_path(path); 5336 break; 5337 } 5338 5339 extent_end = btrfs_file_extent_end(path); 5340 key_offset = key.offset; 5341 5342 /* Read extent fields before releasing the path. */ 5343 ei = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_file_extent_item); 5344 extent_type = btrfs_file_extent_type(leaf, ei); 5345 compression = btrfs_file_extent_compression(leaf, ei); 5346 encryption = btrfs_file_extent_encryption(leaf, ei); 5347 5348 if (extent_type != BTRFS_FILE_EXTENT_INLINE) { 5349 disk_bytenr = btrfs_file_extent_disk_bytenr(leaf, ei); 5350 if (disk_bytenr && compression == BTRFS_COMPRESS_NONE) 5351 extent_offset = btrfs_file_extent_offset(leaf, ei); 5352 } 5353 5354 btrfs_release_path(path); 5355 5356 /* Implicit hole (NO_HOLES feature). */ 5357 if (prev_extent_end < key_offset) { 5358 u64 hole_end = min(key_offset, end_offset); 5359 u64 hole_len = hole_end - prev_extent_end; 5360 5361 if (prev_extent_end >= cur_offset) { 5362 entry.offset = prev_extent_end; 5363 entry.length = hole_len; 5364 entry.type = BTRFS_GET_CSUMS_ZEROED; 5365 5366 if (buf_used + sizeof(entry) > buf_limit) 5367 goto done; 5368 if (copy_to_user(ubuf + buf_used, &entry, sizeof(entry))) { 5369 ret = -EFAULT; 5370 goto out_unlock; 5371 } 5372 buf_used += sizeof(entry); 5373 cur_offset = hole_end; 5374 } 5375 5376 if (key_offset >= end_offset) { 5377 cur_offset = end_offset; 5378 break; 5379 } 5380 } 5381 5382 /* Clamp to our query range. */ 5383 range_start = max(cur_offset, key_offset); 5384 range_len = min(extent_end, end_offset) - range_start; 5385 5386 entry.offset = range_start; 5387 entry.length = range_len; 5388 5389 if (extent_type == BTRFS_FILE_EXTENT_INLINE) { 5390 entry.type = BTRFS_GET_CSUMS_INLINE; 5391 if (compression != BTRFS_COMPRESS_NONE) 5392 entry.type |= BTRFS_GET_CSUMS_COMPRESSED; 5393 if (encryption != 0) 5394 entry.type |= BTRFS_GET_CSUMS_ENCRYPTED; 5395 entry_csum_size = 0; 5396 } else if (extent_type == BTRFS_FILE_EXTENT_PREALLOC) { 5397 entry.type = BTRFS_GET_CSUMS_ZEROED; 5398 entry_csum_size = 0; 5399 } else { 5400 /* BTRFS_FILE_EXTENT_REG */ 5401 if (disk_bytenr == 0) { 5402 /* Explicit hole. */ 5403 entry.type = BTRFS_GET_CSUMS_ZEROED; 5404 entry_csum_size = 0; 5405 } else if (encryption != 0 || compression != BTRFS_COMPRESS_NONE) { 5406 entry.type = 0; 5407 if (encryption != 0) 5408 entry.type |= BTRFS_GET_CSUMS_ENCRYPTED; 5409 if (compression != BTRFS_COMPRESS_NONE) 5410 entry.type |= BTRFS_GET_CSUMS_COMPRESSED; 5411 entry_csum_size = 0; 5412 } else { 5413 entry.type = BTRFS_GET_CSUMS_HAS_CSUMS; 5414 entry_csum_size = (range_len >> fs_info->sectorsize_bits) * csum_size; 5415 } 5416 } 5417 5418 /* Check if this entry (+ csum data) fits in the buffer. */ 5419 if (buf_used + sizeof(entry) + entry_csum_size > buf_limit) { 5420 if (buf_used == 0) { 5421 ret = -EOVERFLOW; 5422 goto out_unlock; 5423 } 5424 goto done; 5425 } 5426 5427 if (copy_to_user(ubuf + buf_used, &entry, sizeof(entry))) { 5428 ret = -EFAULT; 5429 goto out_unlock; 5430 } 5431 buf_used += sizeof(entry); 5432 5433 if (entry.type == BTRFS_GET_CSUMS_HAS_CSUMS) { 5434 ret = copy_csums_to_user(fs_info, 5435 disk_bytenr + extent_offset + (range_start - key_offset), 5436 range_len, ubuf + buf_used); 5437 if (ret) 5438 goto out_unlock; 5439 buf_used += entry_csum_size; 5440 } 5441 5442 cur_offset = range_start + range_len; 5443 prev_extent_end = extent_end; 5444 5445 if (fatal_signal_pending(current)) { 5446 if (buf_used == 0) { 5447 ret = -EINTR; 5448 goto out_unlock; 5449 } 5450 goto done; 5451 } 5452 5453 cond_resched(); 5454 } 5455 5456 /* Handle trailing implicit hole. */ 5457 if (cur_offset < end_offset) { 5458 struct btrfs_ioctl_get_csums_entry entry = { 5459 .offset = prev_extent_end, 5460 .length = end_offset - prev_extent_end, 5461 .type = BTRFS_GET_CSUMS_ZEROED, 5462 }; 5463 5464 if (buf_used + sizeof(entry) <= buf_limit) { 5465 if (copy_to_user(ubuf + buf_used, &entry, sizeof(entry))) { 5466 ret = -EFAULT; 5467 goto out_unlock; 5468 } 5469 buf_used += sizeof(entry); 5470 cur_offset = end_offset; 5471 } 5472 } 5473 5474 done: 5475 args.offset = cur_offset; 5476 args.length = (cur_offset < end_offset) ? end_offset - cur_offset : 0; 5477 args.buf_size = buf_used; 5478 5479 if (copy_to_user(argp, &args, sizeof(args))) 5480 ret = -EFAULT; 5481 5482 out_unlock: 5483 up_read(&vfs_inode->i_rwsem); 5484 return ret; 5485 } 5486 5487 long btrfs_ioctl(struct file *file, unsigned int 5488 cmd, unsigned long arg) 5489 { 5490 struct inode *inode = file_inode(file); 5491 struct btrfs_fs_info *fs_info = inode_to_fs_info(inode); 5492 struct btrfs_root *root = BTRFS_I(inode)->root; 5493 void __user *argp = (void __user *)arg; 5494 5495 switch (cmd) { 5496 case FS_IOC_GETVERSION: 5497 return btrfs_ioctl_getversion(inode, argp); 5498 case FS_IOC_GETFSLABEL: 5499 return btrfs_ioctl_get_fslabel(fs_info, argp); 5500 case FS_IOC_SETFSLABEL: 5501 return btrfs_ioctl_set_fslabel(file, argp); 5502 case FITRIM: 5503 return btrfs_ioctl_fitrim(fs_info, argp); 5504 case BTRFS_IOC_SNAP_CREATE: 5505 return btrfs_ioctl_snap_create(file, argp, false); 5506 case BTRFS_IOC_SNAP_CREATE_V2: 5507 return btrfs_ioctl_snap_create_v2(file, argp, false); 5508 case BTRFS_IOC_SUBVOL_CREATE: 5509 return btrfs_ioctl_snap_create(file, argp, true); 5510 case BTRFS_IOC_SUBVOL_CREATE_V2: 5511 return btrfs_ioctl_snap_create_v2(file, argp, true); 5512 case BTRFS_IOC_SNAP_DESTROY: 5513 return btrfs_ioctl_snap_destroy(file, argp, false); 5514 case BTRFS_IOC_SNAP_DESTROY_V2: 5515 return btrfs_ioctl_snap_destroy(file, argp, true); 5516 case BTRFS_IOC_SUBVOL_GETFLAGS: 5517 return btrfs_ioctl_subvol_getflags(BTRFS_I(inode), argp); 5518 case BTRFS_IOC_SUBVOL_SETFLAGS: 5519 return btrfs_ioctl_subvol_setflags(file, argp); 5520 case BTRFS_IOC_DEFAULT_SUBVOL: 5521 return btrfs_ioctl_default_subvol(file, argp); 5522 case BTRFS_IOC_DEFRAG: 5523 return btrfs_ioctl_defrag(file, NULL); 5524 case BTRFS_IOC_DEFRAG_RANGE: 5525 return btrfs_ioctl_defrag(file, argp); 5526 case BTRFS_IOC_RESIZE: 5527 return btrfs_ioctl_resize(file, argp); 5528 case BTRFS_IOC_ADD_DEV: 5529 return btrfs_ioctl_add_dev(fs_info, argp); 5530 case BTRFS_IOC_RM_DEV: 5531 return btrfs_ioctl_rm_dev(file, argp); 5532 case BTRFS_IOC_RM_DEV_V2: 5533 return btrfs_ioctl_rm_dev_v2(file, argp); 5534 case BTRFS_IOC_FS_INFO: 5535 return btrfs_ioctl_fs_info(fs_info, argp); 5536 case BTRFS_IOC_DEV_INFO: 5537 return btrfs_ioctl_dev_info(fs_info, argp); 5538 case BTRFS_IOC_TREE_SEARCH: 5539 return btrfs_ioctl_tree_search(root, argp); 5540 case BTRFS_IOC_TREE_SEARCH_V2: 5541 return btrfs_ioctl_tree_search_v2(root, argp); 5542 case BTRFS_IOC_INO_LOOKUP: 5543 return btrfs_ioctl_ino_lookup(root, argp); 5544 case BTRFS_IOC_INO_PATHS: 5545 return btrfs_ioctl_ino_to_path(root, argp); 5546 case BTRFS_IOC_LOGICAL_INO: 5547 return btrfs_ioctl_logical_to_ino(fs_info, argp, 1); 5548 case BTRFS_IOC_LOGICAL_INO_V2: 5549 return btrfs_ioctl_logical_to_ino(fs_info, argp, 2); 5550 case BTRFS_IOC_SPACE_INFO: 5551 return btrfs_ioctl_space_info(fs_info, argp); 5552 case BTRFS_IOC_SYNC: { 5553 int ret; 5554 5555 ret = btrfs_start_delalloc_roots(fs_info, LONG_MAX, false); 5556 if (ret) 5557 return ret; 5558 ret = btrfs_sync_fs(inode->i_sb, 1); 5559 /* 5560 * There may be work for the cleaner kthread to do (subvolume 5561 * deletion, delayed iputs, defrag inodes, etc), so wake it up. 5562 */ 5563 wake_up_process(fs_info->cleaner_kthread); 5564 return ret; 5565 } 5566 case BTRFS_IOC_START_SYNC: 5567 return btrfs_ioctl_start_sync(root, argp); 5568 case BTRFS_IOC_WAIT_SYNC: 5569 return btrfs_ioctl_wait_sync(fs_info, argp); 5570 case BTRFS_IOC_SCRUB: 5571 return btrfs_ioctl_scrub(file, argp); 5572 case BTRFS_IOC_SCRUB_CANCEL: 5573 return btrfs_ioctl_scrub_cancel(fs_info); 5574 case BTRFS_IOC_SCRUB_PROGRESS: 5575 return btrfs_ioctl_scrub_progress(fs_info, argp); 5576 case BTRFS_IOC_BALANCE_V2: 5577 return btrfs_ioctl_balance(file, argp); 5578 case BTRFS_IOC_BALANCE_CTL: 5579 return btrfs_ioctl_balance_ctl(fs_info, arg); 5580 case BTRFS_IOC_BALANCE_PROGRESS: 5581 return btrfs_ioctl_balance_progress(fs_info, argp); 5582 case BTRFS_IOC_SET_RECEIVED_SUBVOL: 5583 return btrfs_ioctl_set_received_subvol(file, argp); 5584 #ifdef CONFIG_64BIT 5585 case BTRFS_IOC_SET_RECEIVED_SUBVOL_32: 5586 return btrfs_ioctl_set_received_subvol_32(file, argp); 5587 #endif 5588 case BTRFS_IOC_SEND: 5589 return _btrfs_ioctl_send(root, argp, false); 5590 #if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT) 5591 case BTRFS_IOC_SEND_32: 5592 return _btrfs_ioctl_send(root, argp, true); 5593 #endif 5594 case BTRFS_IOC_GET_DEV_STATS: 5595 return btrfs_ioctl_get_dev_stats(fs_info, argp); 5596 case BTRFS_IOC_QUOTA_CTL: 5597 return btrfs_ioctl_quota_ctl(file, argp); 5598 case BTRFS_IOC_QGROUP_ASSIGN: 5599 return btrfs_ioctl_qgroup_assign(file, argp); 5600 case BTRFS_IOC_QGROUP_CREATE: 5601 return btrfs_ioctl_qgroup_create(file, argp); 5602 case BTRFS_IOC_QGROUP_LIMIT: 5603 return btrfs_ioctl_qgroup_limit(file, argp); 5604 case BTRFS_IOC_QUOTA_RESCAN: 5605 return btrfs_ioctl_quota_rescan(file, argp); 5606 case BTRFS_IOC_QUOTA_RESCAN_STATUS: 5607 return btrfs_ioctl_quota_rescan_status(fs_info, argp); 5608 case BTRFS_IOC_QUOTA_RESCAN_WAIT: 5609 return btrfs_ioctl_quota_rescan_wait(fs_info); 5610 case BTRFS_IOC_DEV_REPLACE: 5611 return btrfs_ioctl_dev_replace(fs_info, argp); 5612 case BTRFS_IOC_GET_SUPPORTED_FEATURES: 5613 return btrfs_ioctl_get_supported_features(argp); 5614 case BTRFS_IOC_GET_FEATURES: 5615 return btrfs_ioctl_get_features(fs_info, argp); 5616 case BTRFS_IOC_SET_FEATURES: 5617 return btrfs_ioctl_set_features(file, argp); 5618 case BTRFS_IOC_GET_SUBVOL_INFO: 5619 return btrfs_ioctl_get_subvol_info(inode, argp); 5620 #ifdef CONFIG_64BIT 5621 case BTRFS_IOC_GET_SUBVOL_INFO_32: 5622 return btrfs_ioctl_get_subvol_info_32(inode, argp); 5623 #endif 5624 case BTRFS_IOC_GET_SUBVOL_ROOTREF: 5625 return btrfs_ioctl_get_subvol_rootref(root, argp); 5626 case BTRFS_IOC_INO_LOOKUP_USER: 5627 return btrfs_ioctl_ino_lookup_user(file, argp); 5628 case FS_IOC_ENABLE_VERITY: 5629 return fsverity_ioctl_enable(file, (const void __user *)argp); 5630 case FS_IOC_MEASURE_VERITY: 5631 return fsverity_ioctl_measure(file, argp); 5632 case FS_IOC_READ_VERITY_METADATA: 5633 return fsverity_ioctl_read_metadata(file, argp); 5634 case BTRFS_IOC_ENCODED_READ: 5635 return btrfs_ioctl_encoded_read(file, argp, false); 5636 case BTRFS_IOC_ENCODED_WRITE: 5637 return btrfs_ioctl_encoded_write(file, argp, false); 5638 #if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT) 5639 case BTRFS_IOC_ENCODED_READ_32: 5640 return btrfs_ioctl_encoded_read(file, argp, true); 5641 case BTRFS_IOC_ENCODED_WRITE_32: 5642 return btrfs_ioctl_encoded_write(file, argp, true); 5643 #endif 5644 case BTRFS_IOC_SUBVOL_SYNC_WAIT: 5645 return btrfs_ioctl_subvol_sync(fs_info, argp); 5646 case BTRFS_IOC_SHUTDOWN: 5647 return btrfs_ioctl_shutdown(fs_info, arg); 5648 case BTRFS_IOC_GET_CSUMS: 5649 return btrfs_ioctl_get_csums(file, argp); 5650 } 5651 5652 return -ENOTTY; 5653 } 5654 5655 #ifdef CONFIG_COMPAT 5656 long btrfs_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg) 5657 { 5658 /* 5659 * These all access 32-bit values anyway so no further 5660 * handling is necessary. 5661 */ 5662 switch (cmd) { 5663 case FS_IOC32_GETVERSION: 5664 cmd = FS_IOC_GETVERSION; 5665 break; 5666 } 5667 5668 return btrfs_ioctl(file, cmd, (unsigned long) compat_ptr(arg)); 5669 } 5670 #endif 5671