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 /* We can only change NODATACOW for zero-sized regular file. */ 360 if (S_ISREG(inode->vfs_inode.i_mode) && (inode->vfs_inode.i_size == 0)) { 361 inode_flags &= ~BTRFS_INODE_NODATACOW; 362 /* 363 * There is currently no way to change NODATASUM flag 364 * through fileattr API. If we unconditionally keep the 365 * current NODATASUM flag, chattr +C then chattr -C will 366 * keep the NODATASUM flag, and no way to remove that 367 * flag. 368 * 369 * So respect the current mount option for NODATASUM flag. 370 */ 371 if (!btrfs_test_opt(fs_info, NODATASUM)) 372 inode_flags &= ~BTRFS_INODE_NODATASUM; 373 } else if (!S_ISREG(inode->vfs_inode.i_mode)) { 374 inode_flags &= ~BTRFS_INODE_NODATACOW; 375 } 376 } 377 378 /* 379 * The COMPRESS flag can only be changed by users, while the NOCOMPRESS 380 * flag may be changed automatically if compression code won't make 381 * things smaller. 382 */ 383 if (fsflags & FS_NOCOMP_FL) { 384 inode_flags &= ~BTRFS_INODE_COMPRESS; 385 inode_flags |= BTRFS_INODE_NOCOMPRESS; 386 } else if (fsflags & FS_COMPR_FL) { 387 388 if (IS_SWAPFILE(&inode->vfs_inode)) 389 return -ETXTBSY; 390 391 inode_flags |= BTRFS_INODE_COMPRESS; 392 inode_flags &= ~BTRFS_INODE_NOCOMPRESS; 393 394 comp = btrfs_compress_type2str(fs_info->compress_type); 395 if (!comp || comp[0] == 0) 396 comp = btrfs_compress_type2str(BTRFS_COMPRESS_ZLIB); 397 } else { 398 inode_flags &= ~(BTRFS_INODE_COMPRESS | BTRFS_INODE_NOCOMPRESS); 399 } 400 401 /* 402 * 1 for inode item 403 * 1 for property 404 */ 405 trans = btrfs_start_transaction(root, 2); 406 if (IS_ERR(trans)) 407 return PTR_ERR(trans); 408 409 if (comp) { 410 ret = btrfs_set_prop(trans, inode, "btrfs.compression", 411 comp, strlen(comp), 0); 412 if (ret) 413 goto out_end_trans; 414 prop_set = true; 415 } else { 416 ret = btrfs_set_prop(trans, inode, "btrfs.compression", NULL, 0, 0); 417 prop_set = (ret == 0); 418 /* If ret == -ENODATA ignore and proceed to update inode item. */ 419 if (ret && ret != -ENODATA) 420 goto out_end_trans; 421 } 422 423 update_flags: 424 inode->flags = inode_flags; 425 btrfs_update_inode_mapping_flags(inode); 426 btrfs_sync_inode_flags_to_i_flags(inode); 427 inode_inc_iversion(&inode->vfs_inode); 428 inode_set_ctime_current(&inode->vfs_inode); 429 ret = btrfs_update_inode(trans, inode); 430 /* 431 * If we set a property or deleted one, we must abort if we fail to 432 * update the inode, to avoid persisting an inconsistent state. 433 */ 434 if (unlikely(ret && prop_set)) 435 btrfs_abort_transaction(trans, ret); 436 437 out_end_trans: 438 btrfs_end_transaction(trans); 439 return ret; 440 } 441 442 static int btrfs_ioctl_getversion(const struct inode *inode, int __user *arg) 443 { 444 return put_user(inode->i_generation, arg); 445 } 446 447 static noinline int btrfs_ioctl_fitrim(struct btrfs_fs_info *fs_info, 448 void __user *arg) 449 { 450 struct btrfs_device *device; 451 struct fstrim_range range; 452 u64 minlen = ULLONG_MAX; 453 u64 num_devices = 0; 454 int ret; 455 456 if (!capable(CAP_SYS_ADMIN)) 457 return -EPERM; 458 459 /* 460 * btrfs_trim_block_group() depends on space cache, which is not 461 * available in zoned filesystem. So, disallow fitrim on a zoned 462 * filesystem for now. 463 */ 464 if (btrfs_is_zoned(fs_info)) 465 return -EOPNOTSUPP; 466 467 /* 468 * If the fs is mounted with nologreplay, which requires it to be 469 * mounted in RO mode as well, we can not allow discard on free space 470 * inside block groups, because log trees refer to extents that are not 471 * pinned in a block group's free space cache (pinning the extents is 472 * precisely the first phase of replaying a log tree). 473 */ 474 if (btrfs_test_opt(fs_info, NOLOGREPLAY)) 475 return -EROFS; 476 477 rcu_read_lock(); 478 list_for_each_entry_rcu(device, &fs_info->fs_devices->devices, 479 dev_list) { 480 if (!device->bdev || !bdev_max_discard_sectors(device->bdev)) 481 continue; 482 num_devices++; 483 minlen = min_t(u64, bdev_discard_granularity(device->bdev), 484 minlen); 485 } 486 rcu_read_unlock(); 487 488 if (!num_devices) 489 return -EOPNOTSUPP; 490 if (copy_from_user(&range, arg, sizeof(range))) 491 return -EFAULT; 492 493 /* 494 * NOTE: Don't truncate the range using super->total_bytes. Bytenr of 495 * block group is in the logical address space, which can be any 496 * sectorsize aligned bytenr in the range [0, U64_MAX]. 497 */ 498 if (range.len < fs_info->sectorsize) 499 return -EINVAL; 500 501 range.minlen = max(range.minlen, minlen); 502 ret = btrfs_trim_fs(fs_info, &range); 503 504 if (copy_to_user(arg, &range, sizeof(range))) 505 return -EFAULT; 506 507 return ret; 508 } 509 510 /* 511 * Calculate the number of transaction items to reserve for creating a subvolume 512 * or snapshot, not including the inode, directory entries, or parent directory. 513 */ 514 static unsigned int create_subvol_num_items(const struct btrfs_qgroup_inherit *inherit) 515 { 516 /* 517 * 1 to add root block 518 * 1 to add root item 519 * 1 to add root ref 520 * 1 to add root backref 521 * 1 to add UUID item 522 * 1 to add qgroup info 523 * 1 to add qgroup limit 524 * 525 * Ideally the last two would only be accounted if qgroups are enabled, 526 * but that can change between now and the time we would insert them. 527 */ 528 unsigned int num_items = 7; 529 530 if (inherit) { 531 /* 2 to add qgroup relations for each inherited qgroup */ 532 num_items += 2 * inherit->num_qgroups; 533 } 534 return num_items; 535 } 536 537 static noinline int create_subvol(struct mnt_idmap *idmap, 538 struct inode *dir, struct dentry *dentry, 539 struct btrfs_qgroup_inherit *inherit) 540 { 541 struct btrfs_fs_info *fs_info = inode_to_fs_info(dir); 542 struct btrfs_trans_handle *trans; 543 struct btrfs_key key; 544 struct btrfs_root_item AUTO_KFREE(root_item); 545 struct btrfs_inode_item *inode_item; 546 struct extent_buffer *leaf; 547 struct btrfs_root *root = BTRFS_I(dir)->root; 548 struct btrfs_root *new_root; 549 struct btrfs_block_rsv block_rsv; 550 struct timespec64 cur_time = current_time(dir); 551 struct btrfs_new_inode_args new_inode_args = { 552 .dir = dir, 553 .dentry = dentry, 554 .subvol = true, 555 }; 556 unsigned int trans_num_items; 557 int ret; 558 dev_t anon_dev; 559 u64 objectid; 560 u64 qgroup_reserved = 0; 561 562 root_item = kzalloc(sizeof(*root_item), GFP_KERNEL); 563 if (!root_item) 564 return -ENOMEM; 565 566 ret = btrfs_get_free_objectid(fs_info->tree_root, &objectid); 567 if (ret) 568 return ret; 569 570 /* 571 * Don't create subvolume whose level is not zero. Or qgroup will be 572 * screwed up since it assumes subvolume qgroup's level to be 0. 573 */ 574 if (btrfs_qgroup_level(objectid)) 575 return -ENOSPC; 576 577 ret = get_anon_bdev(&anon_dev); 578 if (ret < 0) 579 return ret; 580 581 new_inode_args.inode = btrfs_new_subvol_inode(idmap, dir); 582 if (!new_inode_args.inode) { 583 ret = -ENOMEM; 584 goto out_anon_dev; 585 } 586 ret = btrfs_new_inode_prepare(&new_inode_args, &trans_num_items); 587 if (ret) 588 goto out_inode; 589 trans_num_items += create_subvol_num_items(inherit); 590 591 btrfs_init_block_rsv(&block_rsv, BTRFS_BLOCK_RSV_TEMP); 592 ret = btrfs_subvolume_reserve_metadata(root, &block_rsv, 593 trans_num_items, false); 594 if (ret) 595 goto out_new_inode_args; 596 qgroup_reserved = block_rsv.qgroup_rsv_reserved; 597 598 trans = btrfs_start_transaction(root, 0); 599 if (IS_ERR(trans)) { 600 ret = PTR_ERR(trans); 601 goto out_release_rsv; 602 } 603 btrfs_qgroup_convert_reserved_meta(root, qgroup_reserved); 604 qgroup_reserved = 0; 605 trans->block_rsv = &block_rsv; 606 trans->bytes_reserved = block_rsv.size; 607 608 ret = btrfs_qgroup_inherit(trans, 0, objectid, btrfs_root_id(root), inherit); 609 if (ret) 610 goto out; 611 612 leaf = btrfs_alloc_tree_block(trans, root, 0, objectid, NULL, 0, 0, 0, 613 0, BTRFS_NESTING_NORMAL); 614 if (IS_ERR(leaf)) { 615 ret = PTR_ERR(leaf); 616 goto out; 617 } 618 619 btrfs_mark_buffer_dirty(trans, leaf); 620 621 inode_item = &root_item->inode; 622 btrfs_set_stack_inode_generation(inode_item, 1); 623 btrfs_set_stack_inode_size(inode_item, 3); 624 btrfs_set_stack_inode_nlink(inode_item, 1); 625 btrfs_set_stack_inode_nbytes(inode_item, 626 fs_info->nodesize); 627 btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755); 628 629 btrfs_set_root_flags(root_item, 0); 630 btrfs_set_root_limit(root_item, 0); 631 btrfs_set_stack_inode_flags(inode_item, BTRFS_INODE_ROOT_ITEM_INIT); 632 633 btrfs_set_root_bytenr(root_item, leaf->start); 634 btrfs_set_root_generation(root_item, trans->transid); 635 btrfs_set_root_level(root_item, 0); 636 btrfs_set_root_refs(root_item, 1); 637 btrfs_set_root_used(root_item, leaf->len); 638 btrfs_set_root_last_snapshot(root_item, 0); 639 640 btrfs_set_root_generation_v2(root_item, 641 btrfs_root_generation(root_item)); 642 generate_random_guid(root_item->uuid); 643 btrfs_set_stack_timespec_sec(&root_item->otime, cur_time.tv_sec); 644 btrfs_set_stack_timespec_nsec(&root_item->otime, cur_time.tv_nsec); 645 root_item->ctime = root_item->otime; 646 btrfs_set_root_ctransid(root_item, trans->transid); 647 btrfs_set_root_otransid(root_item, trans->transid); 648 649 btrfs_tree_unlock(leaf); 650 651 btrfs_set_root_dirid(root_item, BTRFS_FIRST_FREE_OBJECTID); 652 653 key.objectid = objectid; 654 key.type = BTRFS_ROOT_ITEM_KEY; 655 key.offset = 0; 656 ret = btrfs_insert_root(trans, fs_info->tree_root, &key, 657 root_item); 658 if (ret) { 659 int ret2; 660 661 /* 662 * Since we don't abort the transaction in this case, free the 663 * tree block so that we don't leak space and leave the 664 * filesystem in an inconsistent state (an extent item in the 665 * extent tree with a backreference for a root that does not 666 * exists). 667 */ 668 btrfs_tree_lock(leaf); 669 btrfs_clear_buffer_dirty(trans, leaf); 670 btrfs_tree_unlock(leaf); 671 ret2 = btrfs_free_tree_block(trans, objectid, leaf, 0, 1); 672 if (unlikely(ret2 < 0)) 673 btrfs_abort_transaction(trans, ret2); 674 free_extent_buffer(leaf); 675 goto out; 676 } 677 678 free_extent_buffer(leaf); 679 leaf = NULL; 680 681 new_root = btrfs_get_new_fs_root(fs_info, objectid, &anon_dev); 682 if (IS_ERR(new_root)) { 683 ret = PTR_ERR(new_root); 684 btrfs_abort_transaction(trans, ret); 685 goto out; 686 } 687 /* anon_dev is owned by new_root now. */ 688 anon_dev = 0; 689 BTRFS_I(new_inode_args.inode)->root = new_root; 690 /* ... and new_root is owned by new_inode_args.inode now. */ 691 692 ret = btrfs_record_root_in_trans(trans, new_root); 693 if (unlikely(ret)) { 694 btrfs_abort_transaction(trans, ret); 695 goto out; 696 } 697 698 ret = btrfs_uuid_tree_add(trans, root_item->uuid, 699 BTRFS_UUID_KEY_SUBVOL, objectid); 700 if (unlikely(ret)) { 701 btrfs_abort_transaction(trans, ret); 702 goto out; 703 } 704 705 btrfs_record_new_subvolume(trans, BTRFS_I(dir)); 706 707 ret = btrfs_create_new_inode(trans, &new_inode_args); 708 if (unlikely(ret)) { 709 btrfs_abort_transaction(trans, ret); 710 goto out; 711 } 712 713 /* 714 * Subvolumes have orphans cleaned on first dentry lookup. A new 715 * subvolume cannot have any orphans, so we should set the bit before we 716 * add the subvolume dentry to the dentry cache, so that it is in the 717 * same state as a subvolume after first lookup. 718 */ 719 set_bit(BTRFS_ROOT_ORPHAN_CLEANUP, &new_root->state); 720 d_instantiate_new(dentry, new_inode_args.inode); 721 new_inode_args.inode = NULL; 722 723 out: 724 trans->block_rsv = NULL; 725 trans->bytes_reserved = 0; 726 btrfs_end_transaction(trans); 727 out_release_rsv: 728 btrfs_block_rsv_release(fs_info, &block_rsv, (u64)-1, NULL); 729 if (qgroup_reserved) 730 btrfs_qgroup_free_meta_prealloc(root, qgroup_reserved); 731 out_new_inode_args: 732 btrfs_new_inode_args_destroy(&new_inode_args); 733 out_inode: 734 iput(new_inode_args.inode); 735 out_anon_dev: 736 if (anon_dev) 737 free_anon_bdev(anon_dev); 738 739 return ret; 740 } 741 742 static int create_snapshot(struct btrfs_root *root, struct inode *dir, 743 struct dentry *dentry, bool readonly, 744 struct btrfs_qgroup_inherit *inherit) 745 { 746 struct btrfs_fs_info *fs_info = inode_to_fs_info(dir); 747 struct inode *inode; 748 struct btrfs_pending_snapshot AUTO_KFREE(pending_snapshot); 749 unsigned int trans_num_items; 750 struct btrfs_trans_handle *trans; 751 struct btrfs_block_rsv *block_rsv; 752 u64 qgroup_reserved = 0; 753 int ret; 754 755 /* We do not support snapshotting right now. */ 756 if (btrfs_fs_incompat(fs_info, EXTENT_TREE_V2)) { 757 btrfs_warn(fs_info, 758 "extent tree v2 doesn't support snapshotting yet"); 759 return -EOPNOTSUPP; 760 } 761 762 if (btrfs_root_refs(&root->root_item) == 0) 763 return -ENOENT; 764 765 if (!test_bit(BTRFS_ROOT_SHAREABLE, &root->state)) 766 return -EINVAL; 767 768 if (atomic_read(&root->nr_swapfiles)) { 769 btrfs_warn(fs_info, 770 "cannot snapshot subvolume with active swapfile"); 771 return -ETXTBSY; 772 } 773 774 pending_snapshot = kzalloc_obj(*pending_snapshot); 775 if (!pending_snapshot) 776 return -ENOMEM; 777 778 ret = get_anon_bdev(&pending_snapshot->anon_dev); 779 if (ret < 0) 780 goto free_pending; 781 pending_snapshot->root_item = kzalloc_obj(struct btrfs_root_item); 782 pending_snapshot->path = btrfs_alloc_path(); 783 if (!pending_snapshot->root_item || !pending_snapshot->path) { 784 ret = -ENOMEM; 785 goto free_pending; 786 } 787 788 block_rsv = &pending_snapshot->block_rsv; 789 btrfs_init_block_rsv(block_rsv, BTRFS_BLOCK_RSV_TEMP); 790 /* 791 * 1 to add dir item 792 * 1 to add dir index 793 * 1 to update parent inode item 794 */ 795 trans_num_items = create_subvol_num_items(inherit) + 3; 796 ret = btrfs_subvolume_reserve_metadata(BTRFS_I(dir)->root, block_rsv, 797 trans_num_items, false); 798 if (ret) 799 goto free_pending; 800 qgroup_reserved = block_rsv->qgroup_rsv_reserved; 801 802 pending_snapshot->dentry = dentry; 803 pending_snapshot->root = root; 804 pending_snapshot->readonly = readonly; 805 pending_snapshot->dir = BTRFS_I(dir); 806 pending_snapshot->inherit = inherit; 807 808 trans = btrfs_start_transaction(root, 0); 809 if (IS_ERR(trans)) { 810 ret = PTR_ERR(trans); 811 goto fail; 812 } 813 ret = btrfs_record_root_in_trans(trans, BTRFS_I(dir)->root); 814 if (ret) { 815 btrfs_end_transaction(trans); 816 goto fail; 817 } 818 btrfs_qgroup_convert_reserved_meta(root, qgroup_reserved); 819 qgroup_reserved = 0; 820 821 trans->pending_snapshot = pending_snapshot; 822 823 ret = btrfs_commit_transaction(trans); 824 if (ret) 825 goto fail; 826 827 ret = pending_snapshot->error; 828 if (ret) 829 goto fail; 830 831 ret = btrfs_orphan_cleanup(pending_snapshot->snap); 832 if (ret) 833 goto fail; 834 835 inode = btrfs_lookup_dentry(d_inode(dentry->d_parent), dentry); 836 if (IS_ERR(inode)) { 837 ret = PTR_ERR(inode); 838 goto fail; 839 } 840 841 d_instantiate(dentry, inode); 842 ret = 0; 843 pending_snapshot->anon_dev = 0; 844 fail: 845 /* Prevent double freeing of anon_dev */ 846 if (ret && pending_snapshot->snap) 847 pending_snapshot->snap->anon_dev = 0; 848 btrfs_put_root(pending_snapshot->snap); 849 btrfs_block_rsv_release(fs_info, block_rsv, (u64)-1, NULL); 850 if (qgroup_reserved) 851 btrfs_qgroup_free_meta_prealloc(root, qgroup_reserved); 852 free_pending: 853 if (pending_snapshot->anon_dev) 854 free_anon_bdev(pending_snapshot->anon_dev); 855 kfree(pending_snapshot->root_item); 856 btrfs_free_path(pending_snapshot->path); 857 858 return ret; 859 } 860 861 /* 862 * Create a new subvolume below @parent. This is largely modeled after 863 * sys_mkdirat and vfs_mkdir, but we only do a single component lookup 864 * inside this filesystem so it's quite a bit simpler. 865 */ 866 static noinline int btrfs_mksubvol(struct dentry *parent, 867 struct mnt_idmap *idmap, 868 struct qstr *qname, struct btrfs_root *snap_src, 869 bool readonly, 870 struct btrfs_qgroup_inherit *inherit) 871 { 872 struct inode *dir = d_inode(parent); 873 struct btrfs_fs_info *fs_info = inode_to_fs_info(dir); 874 struct dentry *dentry; 875 struct fscrypt_str name_str = FSTR_INIT((char *)qname->name, qname->len); 876 int ret; 877 878 dentry = start_creating_killable(idmap, parent, qname); 879 if (IS_ERR(dentry)) 880 return PTR_ERR(dentry); 881 882 ret = may_create_dentry(idmap, dir, dentry); 883 if (ret) 884 goto out_dput; 885 886 /* 887 * even if this name doesn't exist, we may get hash collisions. 888 * check for them now when we can safely fail 889 */ 890 ret = btrfs_check_dir_item_collision(BTRFS_I(dir)->root, dir->i_ino, &name_str); 891 if (ret) 892 goto out_dput; 893 894 down_read(&fs_info->subvol_sem); 895 896 if (btrfs_root_refs(&BTRFS_I(dir)->root->root_item) == 0) 897 goto out_up_read; 898 899 if (snap_src) 900 ret = create_snapshot(snap_src, dir, dentry, readonly, inherit); 901 else 902 ret = create_subvol(idmap, dir, dentry, inherit); 903 904 if (!ret) 905 fsnotify_mkdir(dir, dentry); 906 out_up_read: 907 up_read(&fs_info->subvol_sem); 908 out_dput: 909 end_creating(dentry); 910 return ret; 911 } 912 913 static noinline int btrfs_mksnapshot(struct dentry *parent, 914 struct mnt_idmap *idmap, 915 struct qstr *qname, 916 struct btrfs_root *root, 917 bool readonly, 918 struct btrfs_qgroup_inherit *inherit) 919 { 920 int ret; 921 922 /* 923 * Force new buffered writes to reserve space even when NOCOW is 924 * possible. This is to avoid later writeback (running delalloc) to 925 * fallback to COW mode and unexpectedly fail with ENOSPC. 926 */ 927 btrfs_drew_read_lock(&root->snapshot_lock); 928 929 ret = btrfs_start_delalloc_snapshot(root, false); 930 if (ret) 931 goto out; 932 933 /* 934 * All previous writes have started writeback in NOCOW mode, so now 935 * we force future writes to fallback to COW mode during snapshot 936 * creation. 937 */ 938 atomic_inc(&root->snapshot_force_cow); 939 940 btrfs_wait_ordered_extents(root, U64_MAX, NULL); 941 942 ret = btrfs_mksubvol(parent, idmap, qname, root, readonly, inherit); 943 944 atomic_dec(&root->snapshot_force_cow); 945 out: 946 btrfs_drew_read_unlock(&root->snapshot_lock); 947 return ret; 948 } 949 950 /* 951 * Try to start exclusive operation @type or cancel it if it's running. 952 * 953 * Return: 954 * 0 - normal mode, newly claimed op started 955 * >0 - normal mode, something else is running, 956 * return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS to user space 957 * ECANCELED - cancel mode, successful cancel 958 * ENOTCONN - cancel mode, operation not running anymore 959 */ 960 static int exclop_start_or_cancel_reloc(struct btrfs_fs_info *fs_info, 961 enum btrfs_exclusive_operation type, bool cancel) 962 { 963 if (!cancel) { 964 /* Start normal op */ 965 if (!btrfs_exclop_start(fs_info, type)) 966 return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS; 967 /* Exclusive operation is now claimed */ 968 return 0; 969 } 970 971 /* Cancel running op */ 972 if (btrfs_exclop_start_try_lock(fs_info, type)) { 973 /* 974 * This blocks any exclop finish from setting it to NONE, so we 975 * request cancellation. Either it runs and we will wait for it, 976 * or it has finished and no waiting will happen. 977 */ 978 atomic_inc(&fs_info->reloc_cancel_req); 979 btrfs_exclop_start_unlock(fs_info); 980 981 if (test_bit(BTRFS_FS_RELOC_RUNNING, &fs_info->flags)) 982 wait_on_bit(&fs_info->flags, BTRFS_FS_RELOC_RUNNING, 983 TASK_INTERRUPTIBLE); 984 985 return -ECANCELED; 986 } 987 988 /* Something else is running or none */ 989 return -ENOTCONN; 990 } 991 992 static noinline int btrfs_ioctl_resize(struct file *file, 993 void __user *arg) 994 { 995 BTRFS_DEV_LOOKUP_ARGS(args); 996 struct btrfs_root *root = BTRFS_I(file_inode(file))->root; 997 struct btrfs_fs_info *fs_info = root->fs_info; 998 u64 new_size; 999 u64 old_size; 1000 u64 devid = 1; 1001 struct btrfs_ioctl_vol_args AUTO_KFREE(vol_args); 1002 struct btrfs_device *device = NULL; 1003 char *sizestr; 1004 char *devstr = NULL; 1005 int ret = 0; 1006 int mod = 0; 1007 bool cancel; 1008 1009 if (!capable(CAP_SYS_ADMIN)) 1010 return -EPERM; 1011 1012 ret = mnt_want_write_file(file); 1013 if (ret) 1014 return ret; 1015 1016 /* 1017 * Read the arguments before checking exclusivity to be able to 1018 * distinguish regular resize and cancel 1019 */ 1020 vol_args = memdup_user(arg, sizeof(*vol_args)); 1021 if (IS_ERR(vol_args)) { 1022 ret = PTR_ERR(vol_args); 1023 goto out_drop; 1024 } 1025 ret = btrfs_check_ioctl_vol_args_path(vol_args); 1026 if (ret < 0) 1027 goto out_drop; 1028 1029 sizestr = vol_args->name; 1030 cancel = (strcmp("cancel", sizestr) == 0); 1031 ret = exclop_start_or_cancel_reloc(fs_info, BTRFS_EXCLOP_RESIZE, cancel); 1032 if (ret) 1033 goto out_drop; 1034 /* Exclusive operation is now claimed */ 1035 1036 devstr = strchr(sizestr, ':'); 1037 if (devstr) { 1038 sizestr = devstr + 1; 1039 *devstr = '\0'; 1040 devstr = vol_args->name; 1041 ret = kstrtoull(devstr, 10, &devid); 1042 if (ret) 1043 goto out_finish; 1044 if (!devid) { 1045 ret = -EINVAL; 1046 goto out_finish; 1047 } 1048 btrfs_info(fs_info, "resizing devid %llu", devid); 1049 } 1050 1051 args.devid = devid; 1052 device = btrfs_find_device(fs_info->fs_devices, &args); 1053 if (!device) { 1054 btrfs_info(fs_info, "resizer unable to find device %llu", 1055 devid); 1056 ret = -ENODEV; 1057 goto out_finish; 1058 } 1059 1060 if (!test_bit(BTRFS_DEV_STATE_WRITEABLE, &device->dev_state)) { 1061 btrfs_info(fs_info, 1062 "resizer unable to apply on readonly device %llu", 1063 devid); 1064 ret = -EPERM; 1065 goto out_finish; 1066 } 1067 1068 if (!strcmp(sizestr, "max")) 1069 new_size = bdev_nr_bytes(device->bdev); 1070 else { 1071 char *retptr; 1072 1073 if (sizestr[0] == '-') { 1074 mod = -1; 1075 sizestr++; 1076 } else if (sizestr[0] == '+') { 1077 mod = 1; 1078 sizestr++; 1079 } 1080 new_size = memparse(sizestr, &retptr); 1081 if (*retptr != '\0' || new_size == 0) { 1082 ret = -EINVAL; 1083 goto out_finish; 1084 } 1085 } 1086 1087 if (test_bit(BTRFS_DEV_STATE_REPLACE_TGT, &device->dev_state)) { 1088 ret = -EPERM; 1089 goto out_finish; 1090 } 1091 1092 old_size = btrfs_device_get_total_bytes(device); 1093 1094 if (mod < 0) { 1095 if (new_size > old_size) { 1096 ret = -EINVAL; 1097 goto out_finish; 1098 } 1099 new_size = old_size - new_size; 1100 } else if (mod > 0) { 1101 if (new_size > ULLONG_MAX - old_size) { 1102 ret = -ERANGE; 1103 goto out_finish; 1104 } 1105 new_size = old_size + new_size; 1106 } 1107 1108 if (new_size < SZ_256M) { 1109 ret = -EINVAL; 1110 goto out_finish; 1111 } 1112 if (new_size > bdev_nr_bytes(device->bdev)) { 1113 ret = -EFBIG; 1114 goto out_finish; 1115 } 1116 1117 new_size = round_down(new_size, fs_info->sectorsize); 1118 1119 if (new_size > old_size) { 1120 struct btrfs_trans_handle *trans; 1121 1122 trans = btrfs_start_transaction(root, 0); 1123 if (IS_ERR(trans)) { 1124 ret = PTR_ERR(trans); 1125 goto out_finish; 1126 } 1127 ret = btrfs_grow_device(trans, device, new_size); 1128 btrfs_commit_transaction(trans); 1129 } else if (new_size < old_size) { 1130 ret = btrfs_shrink_device(device, new_size); 1131 } /* equal, nothing need to do */ 1132 1133 if (ret == 0 && new_size != old_size) 1134 btrfs_info(fs_info, 1135 "resize device %s (devid %llu) from %llu to %llu", 1136 btrfs_dev_name(device), device->devid, 1137 old_size, new_size); 1138 out_finish: 1139 btrfs_exclop_finish(fs_info); 1140 out_drop: 1141 mnt_drop_write_file(file); 1142 return ret; 1143 } 1144 1145 static noinline int __btrfs_ioctl_snap_create(struct file *file, 1146 const char *name, unsigned long fd, bool subvol, 1147 bool readonly, 1148 struct btrfs_qgroup_inherit *inherit) 1149 { 1150 int ret; 1151 struct qstr qname = QSTR(name); 1152 struct mnt_idmap *idmap = file_mnt_idmap(file); 1153 1154 if (!S_ISDIR(file_inode(file)->i_mode)) 1155 return -ENOTDIR; 1156 1157 ret = mnt_want_write_file(file); 1158 if (ret) 1159 return ret; 1160 1161 if (strchr(name, '/')) { 1162 ret = -EINVAL; 1163 goto out_drop_write; 1164 } 1165 1166 if (qname.name[0] == '.' && 1167 (qname.len == 1 || (qname.name[1] == '.' && qname.len == 2))) { 1168 ret = -EEXIST; 1169 goto out_drop_write; 1170 } 1171 1172 if (subvol) { 1173 ret = btrfs_mksubvol(file_dentry(file), idmap, &qname, NULL, 1174 readonly, inherit); 1175 } else { 1176 CLASS(fd, src)(fd); 1177 struct inode *src_inode; 1178 if (fd_empty(src)) { 1179 ret = -EINVAL; 1180 goto out_drop_write; 1181 } 1182 1183 src_inode = file_inode(fd_file(src)); 1184 if (src_inode->i_sb != file_inode(file)->i_sb) { 1185 btrfs_info(BTRFS_I(file_inode(file))->root->fs_info, 1186 "Snapshot src from another FS"); 1187 ret = -EXDEV; 1188 } else if (!inode_owner_or_capable(idmap, src_inode)) { 1189 /* 1190 * Subvolume creation is not restricted, but snapshots 1191 * are limited to own subvolumes only 1192 */ 1193 ret = -EPERM; 1194 } else if (btrfs_ino(BTRFS_I(src_inode)) != BTRFS_FIRST_FREE_OBJECTID) { 1195 /* 1196 * Snapshots must be made with the src_inode referring 1197 * to the subvolume inode, otherwise the permission 1198 * checking above is useless because we may have 1199 * permission on a lower directory but not the subvol 1200 * itself. 1201 */ 1202 ret = -EINVAL; 1203 } else { 1204 ret = btrfs_mksnapshot(file_dentry(file), idmap, &qname, 1205 BTRFS_I(src_inode)->root, 1206 readonly, inherit); 1207 } 1208 } 1209 out_drop_write: 1210 mnt_drop_write_file(file); 1211 return ret; 1212 } 1213 1214 static noinline int btrfs_ioctl_snap_create(struct file *file, 1215 void __user *arg, bool subvol) 1216 { 1217 struct btrfs_ioctl_vol_args AUTO_KFREE(vol_args); 1218 int ret; 1219 1220 if (!S_ISDIR(file_inode(file)->i_mode)) 1221 return -ENOTDIR; 1222 1223 vol_args = memdup_user(arg, sizeof(*vol_args)); 1224 if (IS_ERR(vol_args)) 1225 return PTR_ERR(vol_args); 1226 ret = btrfs_check_ioctl_vol_args_path(vol_args); 1227 if (ret < 0) 1228 return ret; 1229 1230 return __btrfs_ioctl_snap_create(file, vol_args->name, vol_args->fd, subvol, 1231 false, NULL); 1232 } 1233 1234 static noinline int btrfs_ioctl_snap_create_v2(struct file *file, 1235 void __user *arg, bool subvol) 1236 { 1237 struct btrfs_ioctl_vol_args_v2 AUTO_KFREE(vol_args); 1238 struct btrfs_qgroup_inherit AUTO_KFREE(inherit); 1239 int ret; 1240 bool readonly = false; 1241 1242 if (!S_ISDIR(file_inode(file)->i_mode)) 1243 return -ENOTDIR; 1244 1245 vol_args = memdup_user(arg, sizeof(*vol_args)); 1246 if (IS_ERR(vol_args)) 1247 return PTR_ERR(vol_args); 1248 ret = btrfs_check_ioctl_vol_args2_subvol_name(vol_args); 1249 if (ret < 0) 1250 return ret; 1251 1252 if (vol_args->flags & ~BTRFS_SUBVOL_CREATE_ARGS_MASK) 1253 return -EOPNOTSUPP; 1254 1255 if (vol_args->flags & BTRFS_SUBVOL_RDONLY) 1256 readonly = true; 1257 if (vol_args->flags & BTRFS_SUBVOL_QGROUP_INHERIT) { 1258 struct btrfs_fs_info *fs_info = inode_to_fs_info(file_inode(file)); 1259 1260 if (vol_args->size < sizeof(*inherit) || vol_args->size > PAGE_SIZE) 1261 return -EINVAL; 1262 1263 inherit = memdup_user(vol_args->qgroup_inherit, vol_args->size); 1264 if (IS_ERR(inherit)) { 1265 return PTR_ERR(inherit); 1266 } 1267 1268 ret = btrfs_qgroup_check_inherit(fs_info, inherit, vol_args->size); 1269 if (ret < 0) 1270 return ret; 1271 } 1272 1273 return __btrfs_ioctl_snap_create(file, vol_args->name, vol_args->fd, subvol, 1274 readonly, inherit); 1275 } 1276 1277 static noinline int btrfs_ioctl_subvol_getflags(struct btrfs_inode *inode, 1278 void __user *arg) 1279 { 1280 struct btrfs_root *root = inode->root; 1281 struct btrfs_fs_info *fs_info = root->fs_info; 1282 int ret = 0; 1283 u64 flags = 0; 1284 1285 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) 1286 return -EINVAL; 1287 1288 down_read(&fs_info->subvol_sem); 1289 if (btrfs_root_readonly(root)) 1290 flags |= BTRFS_SUBVOL_RDONLY; 1291 up_read(&fs_info->subvol_sem); 1292 1293 if (copy_to_user(arg, &flags, sizeof(flags))) 1294 ret = -EFAULT; 1295 1296 return ret; 1297 } 1298 1299 static noinline int btrfs_ioctl_subvol_setflags(struct file *file, 1300 void __user *arg) 1301 { 1302 struct inode *inode = file_inode(file); 1303 struct btrfs_fs_info *fs_info = inode_to_fs_info(inode); 1304 struct btrfs_root *root = BTRFS_I(inode)->root; 1305 struct btrfs_trans_handle *trans; 1306 u64 root_flags; 1307 u64 flags; 1308 int ret; 1309 1310 if (!inode_owner_or_capable(file_mnt_idmap(file), inode)) 1311 return -EPERM; 1312 1313 ret = mnt_want_write_file(file); 1314 if (ret) 1315 return ret; 1316 1317 if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) { 1318 ret = -EINVAL; 1319 goto out_drop_write; 1320 } 1321 1322 if (copy_from_user(&flags, arg, sizeof(flags))) { 1323 ret = -EFAULT; 1324 goto out_drop_write; 1325 } 1326 1327 if (flags & ~BTRFS_SUBVOL_RDONLY) { 1328 ret = -EOPNOTSUPP; 1329 goto out_drop_write; 1330 } 1331 1332 down_write(&fs_info->subvol_sem); 1333 1334 /* nothing to do */ 1335 if (!!(flags & BTRFS_SUBVOL_RDONLY) == btrfs_root_readonly(root)) 1336 goto out_drop_sem; 1337 1338 root_flags = btrfs_root_flags(&root->root_item); 1339 if (flags & BTRFS_SUBVOL_RDONLY) { 1340 btrfs_set_root_flags(&root->root_item, 1341 root_flags | BTRFS_ROOT_SUBVOL_RDONLY); 1342 } else { 1343 /* 1344 * Block RO -> RW transition if this subvolume is involved in 1345 * send 1346 */ 1347 spin_lock(&root->root_item_lock); 1348 if (root->send_in_progress == 0) { 1349 btrfs_set_root_flags(&root->root_item, 1350 root_flags & ~BTRFS_ROOT_SUBVOL_RDONLY); 1351 spin_unlock(&root->root_item_lock); 1352 } else { 1353 spin_unlock(&root->root_item_lock); 1354 btrfs_warn(fs_info, 1355 "Attempt to set subvolume %llu read-write during send", 1356 btrfs_root_id(root)); 1357 ret = -EPERM; 1358 goto out_drop_sem; 1359 } 1360 } 1361 1362 trans = btrfs_start_transaction(root, 1); 1363 if (IS_ERR(trans)) { 1364 ret = PTR_ERR(trans); 1365 goto out_reset; 1366 } 1367 1368 ret = btrfs_update_root(trans, fs_info->tree_root, 1369 &root->root_key, &root->root_item); 1370 if (ret < 0) { 1371 btrfs_end_transaction(trans); 1372 goto out_reset; 1373 } 1374 1375 ret = btrfs_commit_transaction(trans); 1376 1377 out_reset: 1378 if (ret) 1379 btrfs_set_root_flags(&root->root_item, root_flags); 1380 out_drop_sem: 1381 up_write(&fs_info->subvol_sem); 1382 out_drop_write: 1383 mnt_drop_write_file(file); 1384 return ret; 1385 } 1386 1387 static noinline bool key_in_sk(const struct btrfs_key *key, 1388 const struct btrfs_ioctl_search_key *sk) 1389 { 1390 struct btrfs_key test; 1391 int ret; 1392 1393 test.objectid = sk->min_objectid; 1394 test.type = sk->min_type; 1395 test.offset = sk->min_offset; 1396 1397 ret = btrfs_comp_cpu_keys(key, &test); 1398 if (ret < 0) 1399 return false; 1400 1401 test.objectid = sk->max_objectid; 1402 test.type = sk->max_type; 1403 test.offset = sk->max_offset; 1404 1405 ret = btrfs_comp_cpu_keys(key, &test); 1406 if (ret > 0) 1407 return false; 1408 return true; 1409 } 1410 1411 static noinline int copy_to_sk(struct btrfs_path *path, 1412 struct btrfs_key *key, 1413 const struct btrfs_ioctl_search_key *sk, 1414 u64 *buf_size, 1415 char __user *ubuf, 1416 unsigned long *sk_offset, 1417 int *num_found) 1418 { 1419 u64 found_transid; 1420 struct extent_buffer *leaf; 1421 struct btrfs_ioctl_search_header sh; 1422 struct btrfs_key test; 1423 unsigned long item_off; 1424 unsigned long item_len; 1425 int nritems; 1426 int i; 1427 int slot; 1428 int ret = 0; 1429 1430 leaf = path->nodes[0]; 1431 slot = path->slots[0]; 1432 nritems = btrfs_header_nritems(leaf); 1433 1434 if (btrfs_header_generation(leaf) > sk->max_transid) { 1435 i = nritems; 1436 goto advance_key; 1437 } 1438 found_transid = btrfs_header_generation(leaf); 1439 1440 for (i = slot; i < nritems; i++) { 1441 item_off = btrfs_item_ptr_offset(leaf, i); 1442 item_len = btrfs_item_size(leaf, i); 1443 1444 btrfs_item_key_to_cpu(leaf, key, i); 1445 if (!key_in_sk(key, sk)) 1446 continue; 1447 1448 if (sizeof(sh) + item_len > *buf_size) { 1449 if (*num_found) 1450 return 1; 1451 1452 /* 1453 * return one empty item back for v1, which does not 1454 * handle -EOVERFLOW 1455 */ 1456 1457 *buf_size = sizeof(sh) + item_len; 1458 item_len = 0; 1459 ret = -EOVERFLOW; 1460 } 1461 1462 if (sizeof(sh) + item_len + *sk_offset > *buf_size) 1463 return 1; 1464 1465 sh.objectid = key->objectid; 1466 sh.type = key->type; 1467 sh.offset = key->offset; 1468 sh.len = item_len; 1469 sh.transid = found_transid; 1470 1471 /* 1472 * Copy search result header. If we fault then loop again so we 1473 * can fault in the pages and -EFAULT there if there's a 1474 * problem. Otherwise we'll fault and then copy the buffer in 1475 * properly this next time through 1476 */ 1477 if (copy_to_user_nofault(ubuf + *sk_offset, &sh, sizeof(sh))) 1478 return 0; 1479 1480 *sk_offset += sizeof(sh); 1481 1482 if (item_len) { 1483 char __user *up = ubuf + *sk_offset; 1484 /* 1485 * Copy the item, same behavior as above, but reset the 1486 * * sk_offset so we copy the full thing again. 1487 */ 1488 if (read_extent_buffer_to_user_nofault(leaf, up, 1489 item_off, item_len)) { 1490 *sk_offset -= sizeof(sh); 1491 return 0; 1492 } 1493 1494 *sk_offset += item_len; 1495 } 1496 (*num_found)++; 1497 1498 /* -EOVERFLOW from above. */ 1499 if (ret) 1500 return ret; 1501 1502 if (*num_found >= sk->nr_items) 1503 return 1; 1504 } 1505 advance_key: 1506 ret = 0; 1507 test.objectid = sk->max_objectid; 1508 test.type = sk->max_type; 1509 test.offset = sk->max_offset; 1510 if (btrfs_comp_cpu_keys(key, &test) >= 0) 1511 ret = 1; 1512 else if (key->offset < (u64)-1) 1513 key->offset++; 1514 else if (key->type < (u8)-1) { 1515 key->offset = 0; 1516 key->type++; 1517 } else if (key->objectid < (u64)-1) { 1518 key->offset = 0; 1519 key->type = 0; 1520 key->objectid++; 1521 } else 1522 ret = 1; 1523 1524 /* 1525 * 0: all items from this leaf copied, continue with next 1526 * 1: * more items can be copied, but unused buffer is too small 1527 * * all items were found 1528 * Either way, it will stops the loop which iterates to the next 1529 * leaf 1530 * -EOVERFLOW: item was to large for buffer 1531 * -EFAULT: could not copy extent buffer back to userspace 1532 */ 1533 return ret; 1534 } 1535 1536 static noinline int search_ioctl(struct btrfs_root *root, 1537 struct btrfs_ioctl_search_key *sk, 1538 u64 *buf_size, 1539 char __user *ubuf) 1540 { 1541 struct btrfs_fs_info *info = root->fs_info; 1542 struct btrfs_key key; 1543 BTRFS_PATH_AUTO_FREE(path); 1544 int ret; 1545 int num_found = 0; 1546 unsigned long sk_offset = 0; 1547 1548 if (*buf_size < sizeof(struct btrfs_ioctl_search_header)) { 1549 *buf_size = sizeof(struct btrfs_ioctl_search_header); 1550 return -EOVERFLOW; 1551 } 1552 1553 path = btrfs_alloc_path(); 1554 if (!path) 1555 return -ENOMEM; 1556 1557 if (sk->tree_id == 0) { 1558 /* Search the root that we got passed. */ 1559 root = btrfs_grab_root(root); 1560 } else { 1561 /* Look up the root from the arguments. */ 1562 root = btrfs_get_fs_root(info, sk->tree_id, true); 1563 if (IS_ERR(root)) 1564 return PTR_ERR(root); 1565 } 1566 1567 key.objectid = sk->min_objectid; 1568 key.type = sk->min_type; 1569 key.offset = sk->min_offset; 1570 1571 while (1) { 1572 /* 1573 * Ensure that the whole user buffer is faulted in at sub-page 1574 * granularity, otherwise the loop may live-lock. 1575 */ 1576 if (fault_in_subpage_writeable(ubuf + sk_offset, *buf_size - sk_offset)) { 1577 ret = -EFAULT; 1578 break; 1579 } 1580 1581 ret = btrfs_search_forward(root, &key, path, sk->min_transid); 1582 if (ret) 1583 break; 1584 1585 ret = copy_to_sk(path, &key, sk, buf_size, ubuf, 1586 &sk_offset, &num_found); 1587 btrfs_release_path(path); 1588 if (ret) 1589 break; 1590 1591 } 1592 /* Normalize return values from btrfs_search_forward() and copy_to_sk(). */ 1593 if (ret > 0) 1594 ret = 0; 1595 1596 sk->nr_items = num_found; 1597 btrfs_put_root(root); 1598 return ret; 1599 } 1600 1601 static noinline int btrfs_ioctl_tree_search(struct btrfs_root *root, 1602 void __user *argp) 1603 { 1604 struct btrfs_ioctl_search_args __user *uargs = argp; 1605 struct btrfs_ioctl_search_key sk; 1606 int ret; 1607 u64 buf_size; 1608 1609 if (!capable(CAP_SYS_ADMIN)) 1610 return -EPERM; 1611 1612 if (copy_from_user(&sk, &uargs->key, sizeof(sk))) 1613 return -EFAULT; 1614 1615 buf_size = sizeof(uargs->buf); 1616 1617 ret = search_ioctl(root, &sk, &buf_size, uargs->buf); 1618 1619 /* 1620 * In the origin implementation an overflow is handled by returning a 1621 * search header with a len of zero, so reset ret. 1622 */ 1623 if (ret == -EOVERFLOW) 1624 ret = 0; 1625 1626 if (ret == 0 && copy_to_user(&uargs->key, &sk, sizeof(sk))) 1627 ret = -EFAULT; 1628 return ret; 1629 } 1630 1631 static noinline int btrfs_ioctl_tree_search_v2(struct btrfs_root *root, 1632 void __user *argp) 1633 { 1634 struct btrfs_ioctl_search_args_v2 __user *uarg = argp; 1635 struct btrfs_ioctl_search_args_v2 args; 1636 int ret; 1637 u64 buf_size; 1638 const u64 buf_limit = SZ_16M; 1639 1640 if (!capable(CAP_SYS_ADMIN)) 1641 return -EPERM; 1642 1643 /* copy search header and buffer size */ 1644 if (copy_from_user(&args, uarg, sizeof(args))) 1645 return -EFAULT; 1646 1647 buf_size = args.buf_size; 1648 1649 /* limit result size to 16MB */ 1650 if (buf_size > buf_limit) 1651 buf_size = buf_limit; 1652 1653 ret = search_ioctl(root, &args.key, &buf_size, 1654 (char __user *)(&uarg->buf[0])); 1655 if (ret == 0 && copy_to_user(&uarg->key, &args.key, sizeof(args.key))) 1656 ret = -EFAULT; 1657 else if (ret == -EOVERFLOW && 1658 copy_to_user(&uarg->buf_size, &buf_size, sizeof(buf_size))) 1659 ret = -EFAULT; 1660 1661 return ret; 1662 } 1663 1664 /* 1665 * Search for an INODE_REF in a 'root' tree which identifies the path name of 1666 * 'dirid'. When found, it sets 'name' with the path name. 1667 */ 1668 static noinline int btrfs_search_path_in_tree(struct btrfs_root *root, u64 dirid, char *name) 1669 { 1670 struct btrfs_key key; 1671 char *ptr; 1672 int ret = -1; 1673 int slot; 1674 int len; 1675 int total_len = 0; 1676 struct btrfs_inode_ref *iref; 1677 struct extent_buffer *l; 1678 BTRFS_PATH_AUTO_FREE(path); 1679 1680 if (dirid == BTRFS_FIRST_FREE_OBJECTID) { 1681 name[0]='\0'; 1682 return 0; 1683 } 1684 1685 path = btrfs_alloc_path(); 1686 if (!path) 1687 return -ENOMEM; 1688 1689 ptr = &name[BTRFS_INO_LOOKUP_PATH_MAX - 1]; 1690 1691 key.objectid = dirid; 1692 key.type = BTRFS_INODE_REF_KEY; 1693 key.offset = (u64)-1; 1694 1695 while (1) { 1696 ret = btrfs_search_backwards(root, &key, path); 1697 if (ret < 0) 1698 return ret; 1699 else if (ret > 0) 1700 return -ENOENT; 1701 1702 l = path->nodes[0]; 1703 slot = path->slots[0]; 1704 1705 iref = btrfs_item_ptr(l, slot, struct btrfs_inode_ref); 1706 len = btrfs_inode_ref_name_len(l, iref); 1707 ptr -= len + 1; 1708 total_len += len + 1; 1709 if (ptr < name) 1710 return -ENAMETOOLONG; 1711 1712 *(ptr + len) = '/'; 1713 read_extent_buffer(l, ptr, (unsigned long)(iref + 1), len); 1714 1715 if (key.offset == BTRFS_FIRST_FREE_OBJECTID) 1716 break; 1717 1718 btrfs_release_path(path); 1719 key.objectid = key.offset; 1720 key.offset = (u64)-1; 1721 dirid = key.objectid; 1722 } 1723 memmove(name, ptr, total_len); 1724 name[total_len] = '\0'; 1725 1726 return 0; 1727 } 1728 1729 static int btrfs_search_path_in_tree_user(struct mnt_idmap *idmap, 1730 struct inode *inode, 1731 struct btrfs_ioctl_ino_lookup_user_args *args) 1732 { 1733 struct btrfs_fs_info *fs_info = BTRFS_I(inode)->root->fs_info; 1734 u64 upper_limit = btrfs_ino(BTRFS_I(inode)); 1735 u64 treeid = btrfs_root_id(BTRFS_I(inode)->root); 1736 u64 dirid = args->dirid; 1737 unsigned long item_off; 1738 unsigned long item_len; 1739 struct btrfs_inode_ref *iref; 1740 struct btrfs_root_ref *rref; 1741 struct btrfs_root *root = NULL; 1742 BTRFS_PATH_AUTO_FREE(path); 1743 struct btrfs_key key; 1744 struct extent_buffer *leaf; 1745 char *ptr; 1746 int slot; 1747 int len; 1748 int total_len = 0; 1749 int ret; 1750 1751 path = btrfs_alloc_path(); 1752 if (!path) 1753 return -ENOMEM; 1754 1755 /* 1756 * If the bottom subvolume does not exist directly under upper_limit, 1757 * construct the path in from the bottom up. 1758 */ 1759 if (dirid != upper_limit) { 1760 ptr = &args->path[BTRFS_INO_LOOKUP_USER_PATH_MAX - 1]; 1761 1762 root = btrfs_get_fs_root(fs_info, treeid, true); 1763 if (IS_ERR(root)) 1764 return PTR_ERR(root); 1765 1766 key.objectid = dirid; 1767 key.type = BTRFS_INODE_REF_KEY; 1768 key.offset = (u64)-1; 1769 while (1) { 1770 struct btrfs_inode *temp_inode; 1771 1772 ret = btrfs_search_backwards(root, &key, path); 1773 if (ret < 0) 1774 goto out_put; 1775 else if (ret > 0) { 1776 ret = -ENOENT; 1777 goto out_put; 1778 } 1779 1780 leaf = path->nodes[0]; 1781 slot = path->slots[0]; 1782 1783 iref = btrfs_item_ptr(leaf, slot, struct btrfs_inode_ref); 1784 len = btrfs_inode_ref_name_len(leaf, iref); 1785 ptr -= len + 1; 1786 total_len += len + 1; 1787 if (ptr < args->path) { 1788 ret = -ENAMETOOLONG; 1789 goto out_put; 1790 } 1791 1792 *(ptr + len) = '/'; 1793 read_extent_buffer(leaf, ptr, 1794 (unsigned long)(iref + 1), len); 1795 1796 /* 1797 * We don't need the path anymore, so release it and 1798 * avoid deadlocks and lockdep warnings in case 1799 * btrfs_iget() needs to lookup the inode from its root 1800 * btree and lock the same leaf. 1801 */ 1802 btrfs_release_path(path); 1803 temp_inode = btrfs_iget(key.offset, root); 1804 if (IS_ERR(temp_inode)) { 1805 ret = PTR_ERR(temp_inode); 1806 goto out_put; 1807 } 1808 /* Check the read+exec permission of this directory. */ 1809 ret = inode_permission(idmap, &temp_inode->vfs_inode, 1810 MAY_READ | MAY_EXEC); 1811 iput(&temp_inode->vfs_inode); 1812 if (ret) 1813 goto out_put; 1814 1815 if (key.offset == upper_limit) 1816 break; 1817 if (key.objectid == BTRFS_FIRST_FREE_OBJECTID) { 1818 ret = -EACCES; 1819 goto out_put; 1820 } 1821 1822 key.objectid = key.offset; 1823 key.offset = (u64)-1; 1824 dirid = key.objectid; 1825 } 1826 1827 memmove(args->path, ptr, total_len); 1828 args->path[total_len] = '\0'; 1829 btrfs_put_root(root); 1830 root = NULL; 1831 btrfs_release_path(path); 1832 } 1833 1834 /* Get the bottom subvolume's name from ROOT_REF */ 1835 key.objectid = treeid; 1836 key.type = BTRFS_ROOT_REF_KEY; 1837 key.offset = args->treeid; 1838 ret = btrfs_search_slot(NULL, fs_info->tree_root, &key, path, 0, 0); 1839 if (ret < 0) 1840 return ret; 1841 else if (ret > 0) 1842 return -ENOENT; 1843 1844 leaf = path->nodes[0]; 1845 slot = path->slots[0]; 1846 btrfs_item_key_to_cpu(leaf, &key, slot); 1847 1848 item_off = btrfs_item_ptr_offset(leaf, slot); 1849 item_len = btrfs_item_size(leaf, slot); 1850 /* Check if dirid in ROOT_REF corresponds to passed dirid */ 1851 rref = btrfs_item_ptr(leaf, slot, struct btrfs_root_ref); 1852 if (args->dirid != btrfs_root_ref_dirid(leaf, rref)) 1853 return -EINVAL; 1854 1855 /* Copy subvolume's name */ 1856 item_off += sizeof(struct btrfs_root_ref); 1857 item_len -= sizeof(struct btrfs_root_ref); 1858 read_extent_buffer(leaf, args->name, item_off, item_len); 1859 args->name[item_len] = 0; 1860 1861 out_put: 1862 btrfs_put_root(root); 1863 1864 return ret; 1865 } 1866 1867 static noinline int btrfs_ioctl_ino_lookup(struct btrfs_root *root, 1868 void __user *argp) 1869 { 1870 bool new_root = false; 1871 struct btrfs_ioctl_ino_lookup_args AUTO_KFREE(args); 1872 int ret = 0; 1873 1874 args = memdup_user(argp, sizeof(*args)); 1875 if (IS_ERR(args)) 1876 return PTR_ERR(args); 1877 1878 /* 1879 * Unprivileged query to obtain the containing subvolume root id. The 1880 * path is reset so it's consistent with btrfs_search_path_in_tree. 1881 */ 1882 if (args->treeid == 0) 1883 args->treeid = btrfs_root_id(root); 1884 else 1885 new_root = true; 1886 1887 if (args->objectid == BTRFS_FIRST_FREE_OBJECTID) { 1888 args->name[0] = 0; 1889 goto out; 1890 } 1891 1892 if (!capable(CAP_SYS_ADMIN)) { 1893 ret = -EPERM; 1894 goto out; 1895 } 1896 1897 if (new_root) { 1898 root = btrfs_get_fs_root(root->fs_info, args->treeid, true); 1899 if (IS_ERR(root)) 1900 return PTR_ERR(root); 1901 } 1902 ret = btrfs_search_path_in_tree(root, args->objectid, args->name); 1903 if (new_root) 1904 btrfs_put_root(root); 1905 1906 out: 1907 if (ret == 0 && copy_to_user(argp, args, sizeof(*args))) 1908 return -EFAULT; 1909 1910 return ret; 1911 } 1912 1913 /* 1914 * Version of ino_lookup ioctl (unprivileged) 1915 * 1916 * The main differences from ino_lookup ioctl are: 1917 * 1918 * 1. Read + Exec permission will be checked using inode_permission() during 1919 * path construction. -EACCES will be returned in case of failure. 1920 * 2. Path construction will be stopped at the inode number which corresponds 1921 * to the fd with which this ioctl is called. If constructed path does not 1922 * exist under fd's inode, -EACCES will be returned. 1923 * 3. The name of bottom subvolume is also searched and filled. 1924 */ 1925 static int btrfs_ioctl_ino_lookup_user(struct file *file, void __user *argp) 1926 { 1927 struct btrfs_ioctl_ino_lookup_user_args AUTO_KFREE(args); 1928 struct inode *inode; 1929 int ret; 1930 1931 args = memdup_user(argp, sizeof(*args)); 1932 if (IS_ERR(args)) 1933 return PTR_ERR(args); 1934 1935 inode = file_inode(file); 1936 1937 if (args->dirid == BTRFS_FIRST_FREE_OBJECTID && 1938 btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) { 1939 /* 1940 * The subvolume does not exist under fd with which this is 1941 * called 1942 */ 1943 return -EACCES; 1944 } 1945 1946 ret = btrfs_search_path_in_tree_user(file_mnt_idmap(file), inode, args); 1947 1948 if (ret == 0 && copy_to_user(argp, args, sizeof(*args))) 1949 ret = -EFAULT; 1950 1951 return ret; 1952 } 1953 1954 /* Get the subvolume information in BTRFS_ROOT_ITEM and BTRFS_ROOT_BACKREF */ 1955 static int _btrfs_ioctl_get_subvol_info(struct inode *inode, 1956 struct btrfs_ioctl_get_subvol_info_args *subvol_info) 1957 { 1958 struct btrfs_fs_info *fs_info; 1959 struct btrfs_root *root; 1960 struct btrfs_path *path; 1961 struct btrfs_key key; 1962 struct btrfs_root_item *root_item; 1963 struct btrfs_root_ref *rref; 1964 struct extent_buffer *leaf; 1965 unsigned long item_off; 1966 int slot; 1967 int ret = 0; 1968 1969 path = btrfs_alloc_path(); 1970 if (!path) 1971 return -ENOMEM; 1972 1973 fs_info = BTRFS_I(inode)->root->fs_info; 1974 1975 /* Get root_item of inode's subvolume */ 1976 key.objectid = btrfs_root_id(BTRFS_I(inode)->root); 1977 root = btrfs_get_fs_root(fs_info, key.objectid, true); 1978 if (IS_ERR(root)) { 1979 ret = PTR_ERR(root); 1980 goto out_free; 1981 } 1982 root_item = &root->root_item; 1983 1984 subvol_info->treeid = key.objectid; 1985 1986 subvol_info->generation = btrfs_root_generation(root_item); 1987 subvol_info->flags = btrfs_root_flags(root_item); 1988 1989 memcpy(subvol_info->uuid, root_item->uuid, BTRFS_UUID_SIZE); 1990 memcpy(subvol_info->parent_uuid, root_item->parent_uuid, 1991 BTRFS_UUID_SIZE); 1992 memcpy(subvol_info->received_uuid, root_item->received_uuid, 1993 BTRFS_UUID_SIZE); 1994 1995 subvol_info->ctransid = btrfs_root_ctransid(root_item); 1996 subvol_info->ctime.sec = btrfs_stack_timespec_sec(&root_item->ctime); 1997 subvol_info->ctime.nsec = btrfs_stack_timespec_nsec(&root_item->ctime); 1998 1999 subvol_info->otransid = btrfs_root_otransid(root_item); 2000 subvol_info->otime.sec = btrfs_stack_timespec_sec(&root_item->otime); 2001 subvol_info->otime.nsec = btrfs_stack_timespec_nsec(&root_item->otime); 2002 2003 subvol_info->stransid = btrfs_root_stransid(root_item); 2004 subvol_info->stime.sec = btrfs_stack_timespec_sec(&root_item->stime); 2005 subvol_info->stime.nsec = btrfs_stack_timespec_nsec(&root_item->stime); 2006 2007 subvol_info->rtransid = btrfs_root_rtransid(root_item); 2008 subvol_info->rtime.sec = btrfs_stack_timespec_sec(&root_item->rtime); 2009 subvol_info->rtime.nsec = btrfs_stack_timespec_nsec(&root_item->rtime); 2010 2011 if (key.objectid != BTRFS_FS_TREE_OBJECTID) { 2012 /* Search root tree for ROOT_BACKREF of this subvolume */ 2013 key.type = BTRFS_ROOT_BACKREF_KEY; 2014 key.offset = 0; 2015 ret = btrfs_search_slot(NULL, fs_info->tree_root, &key, path, 0, 0); 2016 if (ret < 0) { 2017 goto out; 2018 } else if (path->slots[0] >= 2019 btrfs_header_nritems(path->nodes[0])) { 2020 ret = btrfs_next_leaf(fs_info->tree_root, path); 2021 if (ret < 0) { 2022 goto out; 2023 } else if (unlikely(ret > 0)) { 2024 ret = -EUCLEAN; 2025 goto out; 2026 } 2027 } 2028 2029 leaf = path->nodes[0]; 2030 slot = path->slots[0]; 2031 btrfs_item_key_to_cpu(leaf, &key, slot); 2032 if (key.objectid == subvol_info->treeid && 2033 key.type == BTRFS_ROOT_BACKREF_KEY) { 2034 u16 name_len; 2035 2036 subvol_info->parent_id = key.offset; 2037 2038 rref = btrfs_item_ptr(leaf, slot, struct btrfs_root_ref); 2039 name_len = btrfs_root_ref_name_len(leaf, rref); 2040 subvol_info->dirid = btrfs_root_ref_dirid(leaf, rref); 2041 2042 item_off = btrfs_item_ptr_offset(leaf, slot) + sizeof(*rref); 2043 read_extent_buffer(leaf, subvol_info->name, 2044 item_off, name_len); 2045 } else { 2046 ret = -ENOENT; 2047 goto out; 2048 } 2049 ret = 0; 2050 } 2051 2052 out: 2053 btrfs_put_root(root); 2054 out_free: 2055 btrfs_free_path(path); 2056 return ret; 2057 } 2058 2059 #ifdef CONFIG_64BIT 2060 static int btrfs_ioctl_get_subvol_info_32(struct inode *inode, void __user *argp) 2061 { 2062 struct btrfs_ioctl_get_subvol_info_args AUTO_KFREE(subvol_info); 2063 struct btrfs_ioctl_get_subvol_info_args_32 AUTO_KFREE(subvol_info_32); 2064 int ret; 2065 2066 subvol_info = kzalloc_obj(*subvol_info); 2067 if (!subvol_info) 2068 return -ENOMEM; 2069 2070 subvol_info_32 = kzalloc_obj(*subvol_info_32); 2071 if (!subvol_info_32) 2072 return -ENOMEM; 2073 2074 ret = _btrfs_ioctl_get_subvol_info(inode, subvol_info); 2075 if (ret) 2076 return ret; 2077 2078 subvol_info_32->treeid = subvol_info->treeid; 2079 memcpy(subvol_info_32->name, subvol_info->name, sizeof(subvol_info_32->name)); 2080 subvol_info_32->parent_id = subvol_info->parent_id; 2081 subvol_info_32->dirid = subvol_info->dirid; 2082 subvol_info_32->generation = subvol_info->generation; 2083 subvol_info_32->flags = subvol_info->flags; 2084 memcpy(subvol_info_32->uuid, subvol_info->uuid, BTRFS_UUID_SIZE); 2085 memcpy(subvol_info_32->parent_uuid, subvol_info->parent_uuid, BTRFS_UUID_SIZE); 2086 memcpy(subvol_info_32->received_uuid, subvol_info->received_uuid, BTRFS_UUID_SIZE); 2087 subvol_info_32->ctransid = subvol_info->ctransid; 2088 subvol_info_32->otransid = subvol_info->otransid; 2089 subvol_info_32->stransid = subvol_info->stransid; 2090 subvol_info_32->rtransid = subvol_info->rtransid; 2091 subvol_info_32->ctime.sec = subvol_info->ctime.sec; 2092 subvol_info_32->ctime.nsec = subvol_info->ctime.nsec; 2093 subvol_info_32->otime.sec = subvol_info->otime.sec; 2094 subvol_info_32->otime.nsec = subvol_info->otime.nsec; 2095 subvol_info_32->stime.sec = subvol_info->stime.sec; 2096 subvol_info_32->stime.nsec = subvol_info->stime.nsec; 2097 subvol_info_32->rtime.sec = subvol_info->rtime.sec; 2098 subvol_info_32->rtime.nsec = subvol_info->rtime.nsec; 2099 2100 if (copy_to_user(argp, subvol_info_32, sizeof(*subvol_info_32))) 2101 ret = -EFAULT; 2102 2103 return ret; 2104 } 2105 #endif 2106 2107 static int btrfs_ioctl_get_subvol_info(struct inode *inode, void __user *argp) 2108 { 2109 struct btrfs_ioctl_get_subvol_info_args AUTO_KFREE(subvol_info); 2110 int ret; 2111 2112 subvol_info = kzalloc_obj(*subvol_info); 2113 if (!subvol_info) 2114 return -ENOMEM; 2115 2116 ret = _btrfs_ioctl_get_subvol_info(inode, subvol_info); 2117 if (!ret && copy_to_user(argp, subvol_info, sizeof(*subvol_info))) 2118 ret = -EFAULT; 2119 2120 return ret; 2121 } 2122 2123 /* 2124 * Return ROOT_REF information of the subvolume containing this inode 2125 * except the subvolume name. 2126 */ 2127 static int btrfs_ioctl_get_subvol_rootref(struct btrfs_root *root, 2128 void __user *argp) 2129 { 2130 struct btrfs_ioctl_get_subvol_rootref_args AUTO_KFREE(rootrefs); 2131 struct btrfs_root_ref *rref; 2132 struct btrfs_path *path; 2133 struct btrfs_key key; 2134 struct extent_buffer *leaf; 2135 u64 objectid; 2136 int slot; 2137 int ret; 2138 u8 found; 2139 2140 path = btrfs_alloc_path(); 2141 if (!path) 2142 return -ENOMEM; 2143 2144 rootrefs = memdup_user(argp, sizeof(*rootrefs)); 2145 if (IS_ERR(rootrefs)) { 2146 btrfs_free_path(path); 2147 return PTR_ERR(rootrefs); 2148 } 2149 2150 objectid = btrfs_root_id(root); 2151 key.objectid = objectid; 2152 key.type = BTRFS_ROOT_REF_KEY; 2153 key.offset = rootrefs->min_treeid; 2154 found = 0; 2155 2156 root = root->fs_info->tree_root; 2157 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); 2158 if (ret < 0) { 2159 goto out; 2160 } else if (path->slots[0] >= 2161 btrfs_header_nritems(path->nodes[0])) { 2162 ret = btrfs_next_leaf(root, path); 2163 if (ret < 0) { 2164 goto out; 2165 } else if (unlikely(ret > 0)) { 2166 ret = -EUCLEAN; 2167 goto out; 2168 } 2169 } 2170 while (1) { 2171 leaf = path->nodes[0]; 2172 slot = path->slots[0]; 2173 2174 btrfs_item_key_to_cpu(leaf, &key, slot); 2175 if (key.objectid != objectid || key.type != BTRFS_ROOT_REF_KEY) { 2176 ret = 0; 2177 goto out; 2178 } 2179 2180 if (found == BTRFS_MAX_ROOTREF_BUFFER_NUM) { 2181 ret = -EOVERFLOW; 2182 goto out; 2183 } 2184 2185 rref = btrfs_item_ptr(leaf, slot, struct btrfs_root_ref); 2186 rootrefs->rootref[found].treeid = key.offset; 2187 rootrefs->rootref[found].dirid = 2188 btrfs_root_ref_dirid(leaf, rref); 2189 found++; 2190 2191 ret = btrfs_next_item(root, path); 2192 if (ret < 0) { 2193 goto out; 2194 } else if (unlikely(ret > 0)) { 2195 ret = -EUCLEAN; 2196 goto out; 2197 } 2198 } 2199 2200 out: 2201 btrfs_free_path(path); 2202 2203 if (!ret || ret == -EOVERFLOW) { 2204 rootrefs->num_items = found; 2205 /* update min_treeid for next search */ 2206 if (found) 2207 rootrefs->min_treeid = 2208 rootrefs->rootref[found - 1].treeid + 1; 2209 if (copy_to_user(argp, rootrefs, sizeof(*rootrefs))) 2210 ret = -EFAULT; 2211 } 2212 2213 return ret; 2214 } 2215 2216 static noinline int btrfs_ioctl_snap_destroy(struct file *file, 2217 void __user *arg, 2218 bool destroy_v2) 2219 { 2220 struct dentry *parent = file->f_path.dentry; 2221 struct dentry *dentry; 2222 struct inode *dir = d_inode(parent); 2223 struct btrfs_fs_info *fs_info = inode_to_fs_info(dir); 2224 struct inode *inode; 2225 struct btrfs_root *root = BTRFS_I(dir)->root; 2226 struct btrfs_root *dest = NULL; 2227 struct btrfs_ioctl_vol_args AUTO_KFREE(vol_args); 2228 struct btrfs_ioctl_vol_args_v2 AUTO_KFREE(vol_args2); 2229 struct mnt_idmap *idmap = file_mnt_idmap(file); 2230 char *subvol_name, *subvol_name_ptr = NULL; 2231 int ret = 0; 2232 bool destroy_parent = false; 2233 2234 /* We don't support snapshots with extent tree v2 yet. */ 2235 if (btrfs_fs_incompat(fs_info, EXTENT_TREE_V2)) { 2236 btrfs_err(fs_info, 2237 "extent tree v2 doesn't support snapshot deletion yet"); 2238 return -EOPNOTSUPP; 2239 } 2240 2241 if (destroy_v2) { 2242 vol_args2 = memdup_user(arg, sizeof(*vol_args2)); 2243 if (IS_ERR(vol_args2)) 2244 return PTR_ERR(vol_args2); 2245 2246 if (vol_args2->flags & ~BTRFS_SUBVOL_DELETE_ARGS_MASK) 2247 return -EOPNOTSUPP; 2248 2249 /* 2250 * If SPEC_BY_ID is not set, we are looking for the subvolume by 2251 * name, same as v1 currently does. 2252 */ 2253 if (!(vol_args2->flags & BTRFS_SUBVOL_SPEC_BY_ID)) { 2254 ret = btrfs_check_ioctl_vol_args2_subvol_name(vol_args2); 2255 if (ret < 0) 2256 return ret; 2257 subvol_name = vol_args2->name; 2258 2259 ret = mnt_want_write_file(file); 2260 if (ret) 2261 return ret; 2262 } else { 2263 struct inode *old_dir; 2264 2265 if (vol_args2->subvolid < BTRFS_FIRST_FREE_OBJECTID) 2266 return -EINVAL; 2267 2268 ret = mnt_want_write_file(file); 2269 if (ret) 2270 return ret; 2271 2272 dentry = btrfs_get_dentry(fs_info->sb, 2273 BTRFS_FIRST_FREE_OBJECTID, 2274 vol_args2->subvolid, 0); 2275 if (IS_ERR(dentry)) { 2276 ret = PTR_ERR(dentry); 2277 goto out_drop_write; 2278 } 2279 2280 /* 2281 * Change the default parent since the subvolume being 2282 * deleted can be outside of the current mount point. 2283 */ 2284 parent = btrfs_get_parent(dentry); 2285 2286 /* 2287 * At this point dentry->d_name can point to '/' if the 2288 * subvolume we want to destroy is outsite of the 2289 * current mount point, so we need to release the 2290 * current dentry and execute the lookup to return a new 2291 * one with ->d_name pointing to the 2292 * <mount point>/subvol_name. 2293 */ 2294 dput(dentry); 2295 if (IS_ERR(parent)) { 2296 ret = PTR_ERR(parent); 2297 goto out_drop_write; 2298 } 2299 old_dir = dir; 2300 dir = d_inode(parent); 2301 2302 /* 2303 * If v2 was used with SPEC_BY_ID, a new parent was 2304 * allocated since the subvolume can be outside of the 2305 * current mount point. Later on we need to release this 2306 * new parent dentry. 2307 */ 2308 destroy_parent = true; 2309 2310 /* 2311 * On idmapped mounts, deletion via subvolid is 2312 * restricted to subvolumes that are immediate 2313 * ancestors of the inode referenced by the file 2314 * descriptor in the ioctl. Otherwise the idmapping 2315 * could potentially be abused to delete subvolumes 2316 * anywhere in the filesystem the user wouldn't be able 2317 * to delete without an idmapped mount. 2318 */ 2319 if (old_dir != dir && idmap != &nop_mnt_idmap) { 2320 ret = -EOPNOTSUPP; 2321 goto free_parent; 2322 } 2323 2324 subvol_name_ptr = btrfs_get_subvol_name_from_objectid( 2325 fs_info, vol_args2->subvolid); 2326 if (IS_ERR(subvol_name_ptr)) { 2327 ret = PTR_ERR(subvol_name_ptr); 2328 goto free_parent; 2329 } 2330 /* subvol_name_ptr is already nul terminated */ 2331 subvol_name = (char *)kbasename(subvol_name_ptr); 2332 } 2333 } else { 2334 vol_args = memdup_user(arg, sizeof(*vol_args)); 2335 if (IS_ERR(vol_args)) 2336 return PTR_ERR(vol_args); 2337 2338 ret = btrfs_check_ioctl_vol_args_path(vol_args); 2339 if (ret < 0) 2340 return ret; 2341 2342 subvol_name = vol_args->name; 2343 2344 ret = mnt_want_write_file(file); 2345 if (ret) 2346 return ret; 2347 } 2348 2349 if (strchr(subvol_name, '/') || 2350 strcmp(subvol_name, "..") == 0) { 2351 ret = -EINVAL; 2352 goto free_subvol_name; 2353 } 2354 2355 if (!S_ISDIR(dir->i_mode)) { 2356 ret = -ENOTDIR; 2357 goto free_subvol_name; 2358 } 2359 2360 dentry = start_removing_killable(idmap, parent, &QSTR(subvol_name)); 2361 if (IS_ERR(dentry)) { 2362 ret = PTR_ERR(dentry); 2363 goto out_end_removing; 2364 } 2365 2366 inode = d_inode(dentry); 2367 dest = BTRFS_I(inode)->root; 2368 if (!capable(CAP_SYS_ADMIN)) { 2369 /* 2370 * Regular user. Only allow this with a special mount 2371 * option, when the user has write+exec access to the 2372 * subvol root, and when rmdir(2) would have been 2373 * allowed. 2374 * 2375 * Note that this is _not_ check that the subvol is 2376 * empty or doesn't contain data that we wouldn't 2377 * otherwise be able to delete. 2378 * 2379 * Users who want to delete empty subvols should try 2380 * rmdir(2). 2381 */ 2382 ret = -EPERM; 2383 if (!btrfs_test_opt(fs_info, USER_SUBVOL_RM_ALLOWED)) 2384 goto out_end_removing; 2385 2386 /* 2387 * Do not allow deletion if the parent dir is the same 2388 * as the dir to be deleted. That means the ioctl 2389 * must be called on the dentry referencing the root 2390 * of the subvol, not a random directory contained 2391 * within it. 2392 */ 2393 ret = -EINVAL; 2394 if (root == dest) 2395 goto out_end_removing; 2396 2397 ret = inode_permission(idmap, inode, MAY_WRITE | MAY_EXEC); 2398 if (ret) 2399 goto out_end_removing; 2400 } 2401 2402 /* check if subvolume may be deleted by a user */ 2403 ret = may_delete_dentry(idmap, dir, dentry, true); 2404 if (ret) 2405 goto out_end_removing; 2406 2407 if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) { 2408 ret = -EINVAL; 2409 goto out_end_removing; 2410 } 2411 2412 btrfs_inode_lock(BTRFS_I(inode), 0); 2413 ret = btrfs_delete_subvolume(BTRFS_I(dir), dentry); 2414 btrfs_inode_unlock(BTRFS_I(inode), 0); 2415 if (!ret) 2416 d_delete_notify(dir, dentry); 2417 2418 out_end_removing: 2419 end_removing(dentry); 2420 free_subvol_name: 2421 kfree(subvol_name_ptr); 2422 free_parent: 2423 if (destroy_parent) 2424 dput(parent); 2425 out_drop_write: 2426 mnt_drop_write_file(file); 2427 return ret; 2428 } 2429 2430 static int btrfs_ioctl_defrag(struct file *file, void __user *argp) 2431 { 2432 struct inode *inode = file_inode(file); 2433 struct btrfs_root *root = BTRFS_I(inode)->root; 2434 struct btrfs_ioctl_defrag_range_args range = {0}; 2435 int ret; 2436 2437 ret = mnt_want_write_file(file); 2438 if (ret) 2439 return ret; 2440 2441 if (btrfs_root_readonly(root)) { 2442 ret = -EROFS; 2443 goto out; 2444 } 2445 2446 switch (inode->i_mode & S_IFMT) { 2447 case S_IFDIR: 2448 if (!capable(CAP_SYS_ADMIN)) { 2449 ret = -EPERM; 2450 goto out; 2451 } 2452 ret = btrfs_defrag_root(root); 2453 break; 2454 case S_IFREG: 2455 /* 2456 * Note that this does not check the file descriptor for write 2457 * access. This prevents defragmenting executables that are 2458 * running and allows defrag on files open in read-only mode. 2459 */ 2460 if (!capable(CAP_SYS_ADMIN) && 2461 inode_permission(&nop_mnt_idmap, inode, MAY_WRITE)) { 2462 ret = -EPERM; 2463 goto out; 2464 } 2465 2466 /* 2467 * Don't allow defrag on pre-content watched files, as it could 2468 * populate the page cache with 0's via readahead. 2469 */ 2470 if (unlikely(FMODE_FSNOTIFY_HSM(file->f_mode))) { 2471 ret = -EINVAL; 2472 goto out; 2473 } 2474 2475 if (argp) { 2476 if (copy_from_user(&range, argp, sizeof(range))) { 2477 ret = -EFAULT; 2478 goto out; 2479 } 2480 if (range.flags & ~BTRFS_DEFRAG_RANGE_FLAGS_SUPP) { 2481 ret = -EOPNOTSUPP; 2482 goto out; 2483 } 2484 if ((range.flags & BTRFS_DEFRAG_RANGE_COMPRESS) && 2485 (range.flags & BTRFS_DEFRAG_RANGE_NOCOMPRESS)) { 2486 ret = -EINVAL; 2487 goto out; 2488 } 2489 /* Compression or no-compression require to start the IO. */ 2490 if ((range.flags & BTRFS_DEFRAG_RANGE_COMPRESS) || 2491 (range.flags & BTRFS_DEFRAG_RANGE_NOCOMPRESS)) { 2492 range.flags |= BTRFS_DEFRAG_RANGE_START_IO; 2493 range.extent_thresh = (u32)-1; 2494 } 2495 } else { 2496 /* the rest are all set to zero by kzalloc */ 2497 range.len = (u64)-1; 2498 } 2499 ret = btrfs_defrag_file(BTRFS_I(file_inode(file)), &file->f_ra, 2500 &range, BTRFS_OLDEST_GENERATION, 0); 2501 if (ret > 0) 2502 ret = 0; 2503 break; 2504 default: 2505 ret = -EINVAL; 2506 } 2507 out: 2508 mnt_drop_write_file(file); 2509 return ret; 2510 } 2511 2512 static long btrfs_ioctl_add_dev(struct btrfs_fs_info *fs_info, void __user *arg) 2513 { 2514 struct btrfs_ioctl_vol_args AUTO_KFREE(vol_args); 2515 bool restore_op = false; 2516 int ret; 2517 2518 if (!capable(CAP_SYS_ADMIN)) 2519 return -EPERM; 2520 2521 if (btrfs_fs_incompat(fs_info, EXTENT_TREE_V2)) { 2522 btrfs_err(fs_info, "device add not supported on extent tree v2 yet"); 2523 return -EINVAL; 2524 } 2525 2526 if (fs_info->fs_devices->temp_fsid) { 2527 btrfs_err(fs_info, 2528 "device add not supported on cloned temp-fsid mount"); 2529 return -EINVAL; 2530 } 2531 2532 if (!btrfs_exclop_start(fs_info, BTRFS_EXCLOP_DEV_ADD)) { 2533 if (!btrfs_exclop_start_try_lock(fs_info, BTRFS_EXCLOP_DEV_ADD)) 2534 return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS; 2535 2536 /* 2537 * We can do the device add because we have a paused balanced, 2538 * change the exclusive op type and remember we should bring 2539 * back the paused balance 2540 */ 2541 fs_info->exclusive_operation = BTRFS_EXCLOP_DEV_ADD; 2542 btrfs_exclop_start_unlock(fs_info); 2543 restore_op = true; 2544 } 2545 2546 vol_args = memdup_user(arg, sizeof(*vol_args)); 2547 if (IS_ERR(vol_args)) { 2548 ret = PTR_ERR(vol_args); 2549 goto out; 2550 } 2551 2552 ret = btrfs_check_ioctl_vol_args_path(vol_args); 2553 if (ret < 0) 2554 goto out; 2555 2556 ret = btrfs_init_new_device(fs_info, vol_args->name); 2557 2558 if (!ret) 2559 btrfs_info(fs_info, "disk added %s", vol_args->name); 2560 2561 out: 2562 if (restore_op) 2563 btrfs_exclop_balance(fs_info, BTRFS_EXCLOP_BALANCE_PAUSED); 2564 else 2565 btrfs_exclop_finish(fs_info); 2566 return ret; 2567 } 2568 2569 static long btrfs_ioctl_rm_dev_v2(struct file *file, void __user *arg) 2570 { 2571 BTRFS_DEV_LOOKUP_ARGS(args); 2572 struct inode *inode = file_inode(file); 2573 struct btrfs_fs_info *fs_info = inode_to_fs_info(inode); 2574 struct btrfs_ioctl_vol_args_v2 AUTO_KFREE(vol_args); 2575 struct file *bdev_file = NULL; 2576 int ret; 2577 bool cancel = false; 2578 2579 if (!capable(CAP_SYS_ADMIN)) 2580 return -EPERM; 2581 2582 vol_args = memdup_user(arg, sizeof(*vol_args)); 2583 if (IS_ERR(vol_args)) 2584 return PTR_ERR(vol_args); 2585 2586 if (vol_args->flags & ~BTRFS_DEVICE_REMOVE_ARGS_MASK) { 2587 ret = -EOPNOTSUPP; 2588 goto out; 2589 } 2590 2591 ret = btrfs_check_ioctl_vol_args2_subvol_name(vol_args); 2592 if (ret < 0) 2593 goto out; 2594 2595 if (vol_args->flags & BTRFS_DEVICE_SPEC_BY_ID) { 2596 args.devid = vol_args->devid; 2597 } else if (!strcmp("cancel", vol_args->name)) { 2598 cancel = true; 2599 } else { 2600 ret = btrfs_get_dev_args_from_path(fs_info, &args, vol_args->name); 2601 if (ret) 2602 goto out; 2603 } 2604 2605 ret = mnt_want_write_file(file); 2606 if (ret) 2607 goto out; 2608 2609 ret = exclop_start_or_cancel_reloc(fs_info, BTRFS_EXCLOP_DEV_REMOVE, 2610 cancel); 2611 if (ret) 2612 goto err_drop; 2613 2614 /* Exclusive operation is now claimed */ 2615 ret = btrfs_rm_device(fs_info, &args, &bdev_file); 2616 2617 btrfs_exclop_finish(fs_info); 2618 2619 if (!ret) { 2620 if (vol_args->flags & BTRFS_DEVICE_SPEC_BY_ID) 2621 btrfs_info(fs_info, "device deleted: id %llu", 2622 vol_args->devid); 2623 else 2624 btrfs_info(fs_info, "device deleted: %s", 2625 vol_args->name); 2626 } 2627 err_drop: 2628 mnt_drop_write_file(file); 2629 if (bdev_file) 2630 btrfs_release_device_allow_freeze(bdev_file); 2631 out: 2632 btrfs_put_dev_args_from_path(&args); 2633 return ret; 2634 } 2635 2636 static long btrfs_ioctl_rm_dev(struct file *file, void __user *arg) 2637 { 2638 BTRFS_DEV_LOOKUP_ARGS(args); 2639 struct inode *inode = file_inode(file); 2640 struct btrfs_fs_info *fs_info = inode_to_fs_info(inode); 2641 struct btrfs_ioctl_vol_args AUTO_KFREE(vol_args); 2642 struct file *bdev_file = NULL; 2643 int ret; 2644 bool cancel = false; 2645 2646 if (!capable(CAP_SYS_ADMIN)) 2647 return -EPERM; 2648 2649 vol_args = memdup_user(arg, sizeof(*vol_args)); 2650 if (IS_ERR(vol_args)) 2651 return PTR_ERR(vol_args); 2652 2653 ret = btrfs_check_ioctl_vol_args_path(vol_args); 2654 if (ret < 0) 2655 return ret; 2656 2657 if (!strcmp("cancel", vol_args->name)) { 2658 cancel = true; 2659 } else { 2660 ret = btrfs_get_dev_args_from_path(fs_info, &args, vol_args->name); 2661 if (ret) 2662 goto out; 2663 } 2664 2665 ret = mnt_want_write_file(file); 2666 if (ret) 2667 goto out; 2668 2669 ret = exclop_start_or_cancel_reloc(fs_info, BTRFS_EXCLOP_DEV_REMOVE, 2670 cancel); 2671 if (ret == 0) { 2672 ret = btrfs_rm_device(fs_info, &args, &bdev_file); 2673 if (!ret) 2674 btrfs_info(fs_info, "disk deleted %s", vol_args->name); 2675 btrfs_exclop_finish(fs_info); 2676 } 2677 2678 mnt_drop_write_file(file); 2679 if (bdev_file) 2680 btrfs_release_device_allow_freeze(bdev_file); 2681 out: 2682 btrfs_put_dev_args_from_path(&args); 2683 return ret; 2684 } 2685 2686 static long btrfs_ioctl_fs_info(const struct btrfs_fs_info *fs_info, 2687 void __user *arg) 2688 { 2689 struct btrfs_ioctl_fs_info_args AUTO_KFREE(fi_args); 2690 struct btrfs_device *device; 2691 struct btrfs_fs_devices *fs_devices = fs_info->fs_devices; 2692 u64 flags_in; 2693 2694 fi_args = memdup_user(arg, sizeof(*fi_args)); 2695 if (IS_ERR(fi_args)) 2696 return PTR_ERR(fi_args); 2697 2698 flags_in = fi_args->flags; 2699 memset(fi_args, 0, sizeof(*fi_args)); 2700 2701 rcu_read_lock(); 2702 fi_args->num_devices = fs_devices->num_devices; 2703 2704 list_for_each_entry_rcu(device, &fs_devices->devices, dev_list) { 2705 if (device->devid > fi_args->max_id) 2706 fi_args->max_id = device->devid; 2707 } 2708 rcu_read_unlock(); 2709 2710 memcpy(&fi_args->fsid, fs_devices->fsid, sizeof(fi_args->fsid)); 2711 fi_args->nodesize = fs_info->nodesize; 2712 fi_args->sectorsize = fs_info->sectorsize; 2713 fi_args->clone_alignment = fs_info->sectorsize; 2714 2715 if (flags_in & BTRFS_FS_INFO_FLAG_CSUM_INFO) { 2716 fi_args->csum_type = btrfs_super_csum_type(fs_info->super_copy); 2717 fi_args->csum_size = btrfs_super_csum_size(fs_info->super_copy); 2718 fi_args->flags |= BTRFS_FS_INFO_FLAG_CSUM_INFO; 2719 } 2720 2721 if (flags_in & BTRFS_FS_INFO_FLAG_GENERATION) { 2722 fi_args->generation = btrfs_get_fs_generation(fs_info); 2723 fi_args->flags |= BTRFS_FS_INFO_FLAG_GENERATION; 2724 } 2725 2726 if (flags_in & BTRFS_FS_INFO_FLAG_METADATA_UUID) { 2727 memcpy(&fi_args->metadata_uuid, fs_devices->metadata_uuid, 2728 sizeof(fi_args->metadata_uuid)); 2729 fi_args->flags |= BTRFS_FS_INFO_FLAG_METADATA_UUID; 2730 } 2731 2732 if (copy_to_user(arg, fi_args, sizeof(*fi_args))) 2733 return -EFAULT; 2734 2735 return 0; 2736 } 2737 2738 static long btrfs_ioctl_dev_info(const struct btrfs_fs_info *fs_info, 2739 void __user *arg) 2740 { 2741 BTRFS_DEV_LOOKUP_ARGS(args); 2742 struct btrfs_ioctl_dev_info_args AUTO_KFREE(di_args); 2743 struct btrfs_device *dev; 2744 int ret = 0; 2745 2746 di_args = memdup_user(arg, sizeof(*di_args)); 2747 if (IS_ERR(di_args)) 2748 return PTR_ERR(di_args); 2749 2750 args.devid = di_args->devid; 2751 if (!btrfs_is_empty_uuid(di_args->uuid)) 2752 args.uuid = di_args->uuid; 2753 2754 rcu_read_lock(); 2755 dev = btrfs_find_device(fs_info->fs_devices, &args); 2756 if (!dev) { 2757 ret = -ENODEV; 2758 goto out; 2759 } 2760 2761 di_args->devid = dev->devid; 2762 di_args->bytes_used = btrfs_device_get_bytes_used(dev); 2763 di_args->total_bytes = btrfs_device_get_total_bytes(dev); 2764 memcpy(di_args->uuid, dev->uuid, sizeof(di_args->uuid)); 2765 memcpy(di_args->fsid, dev->fs_devices->fsid, BTRFS_UUID_SIZE); 2766 if (dev->name) 2767 strscpy(di_args->path, btrfs_dev_name(dev), sizeof(di_args->path)); 2768 else 2769 di_args->path[0] = '\0'; 2770 2771 out: 2772 rcu_read_unlock(); 2773 if (ret == 0 && copy_to_user(arg, di_args, sizeof(*di_args))) 2774 ret = -EFAULT; 2775 2776 return ret; 2777 } 2778 2779 static long btrfs_ioctl_default_subvol(struct file *file, void __user *argp) 2780 { 2781 struct inode *inode = file_inode(file); 2782 struct btrfs_fs_info *fs_info = inode_to_fs_info(inode); 2783 struct btrfs_root *root = BTRFS_I(inode)->root; 2784 struct btrfs_root *new_root; 2785 struct btrfs_dir_item *di; 2786 struct btrfs_trans_handle *trans; 2787 struct btrfs_path *path = NULL; 2788 struct btrfs_disk_key disk_key; 2789 struct fscrypt_str name = FSTR_INIT("default", 7); 2790 u64 objectid = 0; 2791 u64 dir_id; 2792 int ret; 2793 2794 if (!capable(CAP_SYS_ADMIN)) 2795 return -EPERM; 2796 2797 ret = mnt_want_write_file(file); 2798 if (ret) 2799 return ret; 2800 2801 if (copy_from_user(&objectid, argp, sizeof(objectid))) { 2802 ret = -EFAULT; 2803 goto out; 2804 } 2805 2806 if (!objectid) 2807 objectid = BTRFS_FS_TREE_OBJECTID; 2808 2809 new_root = btrfs_get_fs_root(fs_info, objectid, true); 2810 if (IS_ERR(new_root)) { 2811 ret = PTR_ERR(new_root); 2812 goto out; 2813 } 2814 if (!btrfs_is_fstree(btrfs_root_id(new_root))) { 2815 ret = -ENOENT; 2816 goto out_free; 2817 } 2818 2819 path = btrfs_alloc_path(); 2820 if (!path) { 2821 ret = -ENOMEM; 2822 goto out_free; 2823 } 2824 2825 trans = btrfs_start_transaction(root, 1); 2826 if (IS_ERR(trans)) { 2827 ret = PTR_ERR(trans); 2828 goto out_free; 2829 } 2830 2831 dir_id = btrfs_super_root_dir(fs_info->super_copy); 2832 di = btrfs_lookup_dir_item(trans, fs_info->tree_root, path, 2833 dir_id, &name, 1); 2834 if (IS_ERR_OR_NULL(di)) { 2835 btrfs_release_path(path); 2836 btrfs_end_transaction(trans); 2837 if (di) 2838 ret = PTR_ERR(di); 2839 else 2840 ret = -ENOENT; 2841 btrfs_err(fs_info, 2842 "could not find default diritem for dir %llu: %pe", 2843 dir_id, ERR_PTR(ret)); 2844 goto out_free; 2845 } 2846 2847 btrfs_cpu_key_to_disk(&disk_key, &new_root->root_key); 2848 btrfs_set_dir_item_key(path->nodes[0], di, &disk_key); 2849 btrfs_release_path(path); 2850 2851 btrfs_set_fs_incompat(fs_info, DEFAULT_SUBVOL); 2852 btrfs_end_transaction(trans); 2853 out_free: 2854 btrfs_put_root(new_root); 2855 btrfs_free_path(path); 2856 out: 2857 mnt_drop_write_file(file); 2858 return ret; 2859 } 2860 2861 static void get_block_group_info(struct list_head *groups_list, 2862 struct btrfs_ioctl_space_info *space) 2863 { 2864 struct btrfs_block_group *block_group; 2865 2866 space->total_bytes = 0; 2867 space->used_bytes = 0; 2868 space->flags = 0; 2869 list_for_each_entry(block_group, groups_list, list) { 2870 space->flags = block_group->flags; 2871 space->total_bytes += block_group->length; 2872 space->used_bytes += block_group->used; 2873 } 2874 } 2875 2876 static long btrfs_ioctl_space_info(struct btrfs_fs_info *fs_info, 2877 void __user *arg) 2878 { 2879 struct btrfs_ioctl_space_args space_args = { 0 }; 2880 struct btrfs_ioctl_space_info space; 2881 struct btrfs_ioctl_space_info *dest; 2882 struct btrfs_ioctl_space_info AUTO_KFREE(dest_orig); 2883 struct btrfs_ioctl_space_info __user *user_dest; 2884 struct btrfs_space_info *info; 2885 static const u64 types[] = { 2886 BTRFS_BLOCK_GROUP_DATA, 2887 BTRFS_BLOCK_GROUP_SYSTEM, 2888 BTRFS_BLOCK_GROUP_METADATA, 2889 BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA 2890 }; 2891 int num_types = 4; 2892 int alloc_size; 2893 int ret = 0; 2894 u64 slot_count = 0; 2895 int i, c; 2896 2897 if (copy_from_user(&space_args, 2898 (struct btrfs_ioctl_space_args __user *)arg, 2899 sizeof(space_args))) 2900 return -EFAULT; 2901 2902 for (i = 0; i < num_types; i++) { 2903 struct btrfs_space_info *tmp; 2904 2905 info = NULL; 2906 list_for_each_entry(tmp, &fs_info->space_info, list) { 2907 if (tmp->flags == types[i]) { 2908 info = tmp; 2909 break; 2910 } 2911 } 2912 2913 if (!info) 2914 continue; 2915 2916 down_read(&info->groups_sem); 2917 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) { 2918 if (!list_empty(&info->block_groups[c])) 2919 slot_count++; 2920 } 2921 up_read(&info->groups_sem); 2922 } 2923 2924 /* 2925 * Global block reserve, exported as a space_info 2926 */ 2927 slot_count++; 2928 2929 /* space_slots == 0 means they are asking for a count */ 2930 if (space_args.space_slots == 0) { 2931 space_args.total_spaces = slot_count; 2932 goto out; 2933 } 2934 2935 slot_count = min_t(u64, space_args.space_slots, slot_count); 2936 2937 alloc_size = sizeof(*dest) * slot_count; 2938 2939 /* we generally have at most 6 or so space infos, one for each raid 2940 * level. So, a whole page should be more than enough for everyone 2941 */ 2942 if (alloc_size > PAGE_SIZE) 2943 return -ENOMEM; 2944 2945 space_args.total_spaces = 0; 2946 dest = kzalloc(alloc_size, GFP_KERNEL); 2947 if (!dest) 2948 return -ENOMEM; 2949 dest_orig = dest; 2950 2951 /* now we have a buffer to copy into */ 2952 for (i = 0; i < num_types; i++) { 2953 struct btrfs_space_info *tmp; 2954 2955 if (!slot_count) 2956 break; 2957 2958 info = NULL; 2959 list_for_each_entry(tmp, &fs_info->space_info, list) { 2960 if (tmp->flags == types[i]) { 2961 info = tmp; 2962 break; 2963 } 2964 } 2965 2966 if (!info) 2967 continue; 2968 down_read(&info->groups_sem); 2969 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) { 2970 if (!list_empty(&info->block_groups[c])) { 2971 get_block_group_info(&info->block_groups[c], 2972 &space); 2973 memcpy(dest, &space, sizeof(space)); 2974 dest++; 2975 space_args.total_spaces++; 2976 slot_count--; 2977 } 2978 if (!slot_count) 2979 break; 2980 } 2981 up_read(&info->groups_sem); 2982 } 2983 2984 /* 2985 * Add global block reserve 2986 */ 2987 if (slot_count) { 2988 struct btrfs_block_rsv *block_rsv = &fs_info->global_block_rsv; 2989 2990 spin_lock(&block_rsv->lock); 2991 space.total_bytes = block_rsv->size; 2992 space.used_bytes = block_rsv->size - block_rsv->reserved; 2993 spin_unlock(&block_rsv->lock); 2994 space.flags = BTRFS_SPACE_INFO_GLOBAL_RSV; 2995 memcpy(dest, &space, sizeof(space)); 2996 space_args.total_spaces++; 2997 } 2998 2999 user_dest = (struct btrfs_ioctl_space_info __user *) 3000 (arg + sizeof(struct btrfs_ioctl_space_args)); 3001 3002 if (copy_to_user(user_dest, dest_orig, 3003 space_args.total_spaces * sizeof(*dest_orig))) 3004 return -EFAULT; 3005 3006 out: 3007 if (ret == 0 && copy_to_user(arg, &space_args, sizeof(space_args))) 3008 ret = -EFAULT; 3009 3010 return ret; 3011 } 3012 3013 static noinline long btrfs_ioctl_start_sync(struct btrfs_root *root, 3014 void __user *argp) 3015 { 3016 struct btrfs_trans_handle *trans; 3017 u64 transid; 3018 3019 /* 3020 * Start orphan cleanup here for the given root in case it hasn't been 3021 * started already by other means. Errors are handled in the other 3022 * functions during transaction commit. 3023 */ 3024 btrfs_orphan_cleanup(root); 3025 3026 trans = btrfs_attach_transaction_barrier(root); 3027 if (IS_ERR(trans)) { 3028 if (PTR_ERR(trans) != -ENOENT) 3029 return PTR_ERR(trans); 3030 3031 /* No running transaction, don't bother */ 3032 transid = btrfs_get_last_trans_committed(root->fs_info); 3033 goto out; 3034 } 3035 transid = trans->transid; 3036 btrfs_commit_transaction_async(trans); 3037 out: 3038 if (argp) 3039 if (copy_to_user(argp, &transid, sizeof(transid))) 3040 return -EFAULT; 3041 return 0; 3042 } 3043 3044 static noinline long btrfs_ioctl_wait_sync(struct btrfs_fs_info *fs_info, 3045 void __user *argp) 3046 { 3047 /* By default wait for the current transaction. */ 3048 u64 transid = 0; 3049 3050 if (argp) 3051 if (copy_from_user(&transid, argp, sizeof(transid))) 3052 return -EFAULT; 3053 3054 return btrfs_wait_for_commit(fs_info, transid); 3055 } 3056 3057 static long btrfs_ioctl_scrub(struct file *file, void __user *arg) 3058 { 3059 struct btrfs_fs_info *fs_info = inode_to_fs_info(file_inode(file)); 3060 struct btrfs_ioctl_scrub_args AUTO_KFREE(sa); 3061 int ret; 3062 3063 if (!capable(CAP_SYS_ADMIN)) 3064 return -EPERM; 3065 3066 if (btrfs_fs_incompat(fs_info, EXTENT_TREE_V2)) { 3067 btrfs_err(fs_info, "scrub: extent tree v2 not yet supported"); 3068 return -EINVAL; 3069 } 3070 3071 sa = memdup_user(arg, sizeof(*sa)); 3072 if (IS_ERR(sa)) 3073 return PTR_ERR(sa); 3074 3075 if (sa->flags & ~BTRFS_SCRUB_SUPPORTED_FLAGS) 3076 return -EOPNOTSUPP; 3077 3078 if (!(sa->flags & BTRFS_SCRUB_READONLY)) { 3079 ret = mnt_want_write_file(file); 3080 if (ret) 3081 return ret; 3082 } 3083 3084 ret = btrfs_scrub_dev(fs_info, sa->devid, sa->start, sa->end, 3085 &sa->progress, sa->flags & BTRFS_SCRUB_READONLY, 3086 false); 3087 3088 /* 3089 * Copy scrub args to user space even if btrfs_scrub_dev() returned an 3090 * error. This is important as it allows user space to know how much 3091 * progress scrub has done. For example, if scrub is canceled we get 3092 * -ECANCELED from btrfs_scrub_dev() and return that error back to user 3093 * space. Later user space can inspect the progress from the structure 3094 * btrfs_ioctl_scrub_args and resume scrub from where it left off 3095 * previously (btrfs-progs does this). 3096 * If we fail to copy the btrfs_ioctl_scrub_args structure to user space 3097 * then return -EFAULT to signal the structure was not copied or it may 3098 * be corrupt and unreliable due to a partial copy. 3099 */ 3100 if (copy_to_user(arg, sa, sizeof(*sa))) 3101 ret = -EFAULT; 3102 3103 if (!(sa->flags & BTRFS_SCRUB_READONLY)) 3104 mnt_drop_write_file(file); 3105 3106 return ret; 3107 } 3108 3109 static long btrfs_ioctl_scrub_cancel(struct btrfs_fs_info *fs_info) 3110 { 3111 if (!capable(CAP_SYS_ADMIN)) 3112 return -EPERM; 3113 3114 return btrfs_scrub_cancel(fs_info); 3115 } 3116 3117 static long btrfs_ioctl_scrub_progress(struct btrfs_fs_info *fs_info, 3118 void __user *arg) 3119 { 3120 struct btrfs_ioctl_scrub_args AUTO_KFREE(sa); 3121 int ret; 3122 3123 if (!capable(CAP_SYS_ADMIN)) 3124 return -EPERM; 3125 3126 sa = memdup_user(arg, sizeof(*sa)); 3127 if (IS_ERR(sa)) 3128 return PTR_ERR(sa); 3129 3130 ret = btrfs_scrub_progress(fs_info, sa->devid, &sa->progress); 3131 3132 if (ret == 0 && copy_to_user(arg, sa, sizeof(*sa))) 3133 return -EFAULT; 3134 3135 return ret; 3136 } 3137 3138 static long btrfs_ioctl_get_dev_stats(struct btrfs_fs_info *fs_info, 3139 void __user *arg) 3140 { 3141 struct btrfs_ioctl_get_dev_stats AUTO_KFREE(sa); 3142 int ret; 3143 3144 sa = memdup_user(arg, sizeof(*sa)); 3145 if (IS_ERR(sa)) 3146 return PTR_ERR(sa); 3147 3148 if ((sa->flags & BTRFS_DEV_STATS_RESET) && !capable(CAP_SYS_ADMIN)) 3149 return -EPERM; 3150 3151 ret = btrfs_get_dev_stats(fs_info, sa); 3152 3153 if (ret == 0 && copy_to_user(arg, sa, sizeof(*sa))) 3154 return -EFAULT; 3155 3156 return ret; 3157 } 3158 3159 static long btrfs_ioctl_dev_replace(struct btrfs_fs_info *fs_info, 3160 void __user *arg) 3161 { 3162 struct btrfs_ioctl_dev_replace_args AUTO_KFREE(p); 3163 int ret; 3164 3165 if (!capable(CAP_SYS_ADMIN)) 3166 return -EPERM; 3167 3168 if (btrfs_fs_incompat(fs_info, EXTENT_TREE_V2)) { 3169 btrfs_err(fs_info, "device replace not supported on extent tree v2 yet"); 3170 return -EINVAL; 3171 } 3172 3173 p = memdup_user(arg, sizeof(*p)); 3174 if (IS_ERR(p)) 3175 return PTR_ERR(p); 3176 3177 switch (p->cmd) { 3178 case BTRFS_IOCTL_DEV_REPLACE_CMD_START: 3179 if (sb_rdonly(fs_info->sb)) 3180 return -EROFS; 3181 if (!btrfs_exclop_start(fs_info, BTRFS_EXCLOP_DEV_REPLACE)) { 3182 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS; 3183 } else { 3184 ret = btrfs_dev_replace_by_ioctl(fs_info, p); 3185 btrfs_exclop_finish(fs_info); 3186 } 3187 break; 3188 case BTRFS_IOCTL_DEV_REPLACE_CMD_STATUS: 3189 btrfs_dev_replace_status(fs_info, p); 3190 ret = 0; 3191 break; 3192 case BTRFS_IOCTL_DEV_REPLACE_CMD_CANCEL: 3193 p->result = btrfs_dev_replace_cancel(fs_info); 3194 ret = 0; 3195 break; 3196 default: 3197 ret = -EINVAL; 3198 break; 3199 } 3200 3201 if ((ret == 0 || ret == -ECANCELED) && copy_to_user(arg, p, sizeof(*p))) 3202 return -EFAULT; 3203 3204 return ret; 3205 } 3206 3207 static long btrfs_ioctl_ino_to_path(struct btrfs_root *root, void __user *arg) 3208 { 3209 int ret = 0; 3210 int i; 3211 u64 rel_ptr; 3212 int size; 3213 struct btrfs_ioctl_ino_path_args AUTO_KFREE(ipa); 3214 struct inode_fs_paths *ipath __free(inode_fs_paths) = NULL; 3215 struct btrfs_path *path; 3216 3217 if (!capable(CAP_DAC_READ_SEARCH)) 3218 return -EPERM; 3219 3220 path = btrfs_alloc_path(); 3221 if (!path) { 3222 ret = -ENOMEM; 3223 goto out; 3224 } 3225 3226 ipa = memdup_user(arg, sizeof(*ipa)); 3227 if (IS_ERR(ipa)) { 3228 ret = PTR_ERR(ipa); 3229 ipa = NULL; 3230 goto out; 3231 } 3232 3233 size = min_t(u32, ipa->size, 4096); 3234 ipath = init_ipath(size, root, path); 3235 if (IS_ERR(ipath)) { 3236 ret = PTR_ERR(ipath); 3237 ipath = NULL; 3238 goto out; 3239 } 3240 3241 ret = paths_from_inode(ipa->inum, ipath); 3242 if (ret < 0) 3243 goto out; 3244 3245 for (i = 0; i < ipath->fspath->elem_cnt; ++i) { 3246 rel_ptr = ipath->fspath->val[i] - 3247 (u64)(unsigned long)ipath->fspath->val; 3248 ipath->fspath->val[i] = rel_ptr; 3249 } 3250 3251 btrfs_free_path(path); 3252 path = NULL; 3253 ret = copy_to_user((void __user *)(unsigned long)ipa->fspath, 3254 ipath->fspath, size); 3255 if (ret) { 3256 ret = -EFAULT; 3257 goto out; 3258 } 3259 3260 out: 3261 btrfs_free_path(path); 3262 3263 return ret; 3264 } 3265 3266 static long btrfs_ioctl_logical_to_ino(struct btrfs_fs_info *fs_info, 3267 void __user *arg, int version) 3268 { 3269 int ret = 0; 3270 int size; 3271 struct btrfs_ioctl_logical_ino_args AUTO_KFREE(loi); 3272 struct btrfs_data_container AUTO_KVFREE(inodes); 3273 bool ignore_offset; 3274 3275 if (!capable(CAP_SYS_ADMIN)) 3276 return -EPERM; 3277 3278 loi = memdup_user(arg, sizeof(*loi)); 3279 if (IS_ERR(loi)) 3280 return PTR_ERR(loi); 3281 3282 if (version == 1) { 3283 ignore_offset = false; 3284 size = min_t(u32, loi->size, SZ_64K); 3285 } else { 3286 /* All reserved bits must be 0 for now */ 3287 if (memchr_inv(loi->reserved, 0, sizeof(loi->reserved))) 3288 return -EINVAL; 3289 3290 /* Only accept flags we have defined so far */ 3291 if (loi->flags & ~(BTRFS_LOGICAL_INO_ARGS_IGNORE_OFFSET)) 3292 return -EINVAL; 3293 3294 ignore_offset = loi->flags & BTRFS_LOGICAL_INO_ARGS_IGNORE_OFFSET; 3295 size = min_t(u32, loi->size, SZ_16M); 3296 } 3297 3298 inodes = init_data_container(size); 3299 if (IS_ERR(inodes)) 3300 return PTR_ERR(inodes); 3301 3302 ret = iterate_inodes_from_logical(loi->logical, fs_info, inodes, ignore_offset); 3303 if (ret == -EINVAL) 3304 return -ENOENT; 3305 if (ret < 0) 3306 return ret; 3307 3308 ret = copy_to_user((void __user *)(unsigned long)loi->inodes, inodes, 3309 size); 3310 if (ret) 3311 ret = -EFAULT; 3312 3313 return ret; 3314 } 3315 3316 void btrfs_update_ioctl_balance_args(struct btrfs_fs_info *fs_info, 3317 struct btrfs_ioctl_balance_args *bargs) 3318 { 3319 struct btrfs_balance_control *bctl = fs_info->balance_ctl; 3320 3321 bargs->flags = bctl->flags; 3322 3323 if (test_bit(BTRFS_FS_BALANCE_RUNNING, &fs_info->flags)) 3324 bargs->state |= BTRFS_BALANCE_STATE_RUNNING; 3325 if (atomic_read(&fs_info->balance_pause_req)) 3326 bargs->state |= BTRFS_BALANCE_STATE_PAUSE_REQ; 3327 if (atomic_read(&fs_info->balance_cancel_req)) 3328 bargs->state |= BTRFS_BALANCE_STATE_CANCEL_REQ; 3329 3330 memcpy(&bargs->data, &bctl->data, sizeof(bargs->data)); 3331 memcpy(&bargs->meta, &bctl->meta, sizeof(bargs->meta)); 3332 memcpy(&bargs->sys, &bctl->sys, sizeof(bargs->sys)); 3333 3334 spin_lock(&fs_info->balance_lock); 3335 memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat)); 3336 spin_unlock(&fs_info->balance_lock); 3337 } 3338 3339 /* 3340 * Try to acquire fs_info::balance_mutex as well as set BTRFS_EXLCOP_BALANCE as 3341 * required. 3342 * 3343 * @fs_info: the filesystem 3344 * @excl_acquired: ptr to boolean value which is set to false in case balance 3345 * is being resumed 3346 * 3347 * Return 0 on success in which case both fs_info::balance is acquired as well 3348 * as exclusive ops are blocked. In case of failure return an error code. 3349 */ 3350 static int btrfs_try_lock_balance(struct btrfs_fs_info *fs_info, bool *excl_acquired) 3351 { 3352 int ret; 3353 3354 /* 3355 * Exclusive operation is locked. Three possibilities: 3356 * (1) some other op is running 3357 * (2) balance is running 3358 * (3) balance is paused -- special case (think resume) 3359 */ 3360 while (1) { 3361 if (btrfs_exclop_start(fs_info, BTRFS_EXCLOP_BALANCE)) { 3362 *excl_acquired = true; 3363 mutex_lock(&fs_info->balance_mutex); 3364 return 0; 3365 } 3366 3367 mutex_lock(&fs_info->balance_mutex); 3368 if (fs_info->balance_ctl) { 3369 /* This is either (2) or (3) */ 3370 if (test_bit(BTRFS_FS_BALANCE_RUNNING, &fs_info->flags)) { 3371 /* This is (2) */ 3372 ret = -EINPROGRESS; 3373 goto out_failure; 3374 3375 } else { 3376 mutex_unlock(&fs_info->balance_mutex); 3377 /* 3378 * Lock released to allow other waiters to 3379 * continue, we'll reexamine the status again. 3380 */ 3381 mutex_lock(&fs_info->balance_mutex); 3382 3383 if (fs_info->balance_ctl && 3384 !test_bit(BTRFS_FS_BALANCE_RUNNING, &fs_info->flags)) { 3385 /* This is (3) */ 3386 *excl_acquired = false; 3387 return 0; 3388 } 3389 } 3390 } else { 3391 /* This is (1) */ 3392 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS; 3393 goto out_failure; 3394 } 3395 3396 mutex_unlock(&fs_info->balance_mutex); 3397 } 3398 3399 out_failure: 3400 mutex_unlock(&fs_info->balance_mutex); 3401 *excl_acquired = false; 3402 return ret; 3403 } 3404 3405 static long btrfs_ioctl_balance(struct file *file, void __user *arg) 3406 { 3407 struct btrfs_root *root = BTRFS_I(file_inode(file))->root; 3408 struct btrfs_fs_info *fs_info = root->fs_info; 3409 struct btrfs_ioctl_balance_args AUTO_KFREE(bargs); 3410 struct btrfs_balance_control *bctl; 3411 bool need_unlock = true; 3412 int ret; 3413 3414 if (!capable(CAP_SYS_ADMIN)) 3415 return -EPERM; 3416 3417 ret = mnt_want_write_file(file); 3418 if (ret) 3419 return ret; 3420 3421 bargs = memdup_user(arg, sizeof(*bargs)); 3422 if (IS_ERR(bargs)) { 3423 ret = PTR_ERR(bargs); 3424 bargs = NULL; 3425 goto out; 3426 } 3427 3428 ret = btrfs_try_lock_balance(fs_info, &need_unlock); 3429 if (ret) 3430 goto out; 3431 3432 lockdep_assert_held(&fs_info->balance_mutex); 3433 3434 if (bargs->flags & BTRFS_BALANCE_RESUME) { 3435 if (!fs_info->balance_ctl) { 3436 ret = -ENOTCONN; 3437 goto out_unlock; 3438 } 3439 3440 bctl = fs_info->balance_ctl; 3441 spin_lock(&fs_info->balance_lock); 3442 bctl->flags |= BTRFS_BALANCE_RESUME; 3443 spin_unlock(&fs_info->balance_lock); 3444 btrfs_exclop_balance(fs_info, BTRFS_EXCLOP_BALANCE); 3445 3446 goto do_balance; 3447 } 3448 3449 if (bargs->flags & ~(BTRFS_BALANCE_ARGS_MASK | BTRFS_BALANCE_TYPE_MASK)) { 3450 ret = -EINVAL; 3451 goto out_unlock; 3452 } 3453 3454 if (fs_info->balance_ctl) { 3455 ret = -EINPROGRESS; 3456 goto out_unlock; 3457 } 3458 3459 bctl = kzalloc_obj(*bctl); 3460 if (!bctl) { 3461 ret = -ENOMEM; 3462 goto out_unlock; 3463 } 3464 3465 memcpy(&bctl->data, &bargs->data, sizeof(bctl->data)); 3466 memcpy(&bctl->meta, &bargs->meta, sizeof(bctl->meta)); 3467 memcpy(&bctl->sys, &bargs->sys, sizeof(bctl->sys)); 3468 3469 bctl->flags = bargs->flags; 3470 do_balance: 3471 /* 3472 * Ownership of bctl and exclusive operation goes to btrfs_balance. 3473 * bctl is freed in reset_balance_state, or, if restriper was paused 3474 * all the way until unmount, in free_fs_info. The flag should be 3475 * cleared after reset_balance_state. 3476 */ 3477 need_unlock = false; 3478 3479 ret = btrfs_balance(fs_info, bctl, bargs); 3480 bctl = NULL; 3481 3482 if (ret == 0 || ret == -ECANCELED) { 3483 if (copy_to_user(arg, bargs, sizeof(*bargs))) 3484 ret = -EFAULT; 3485 } 3486 3487 kfree(bctl); 3488 out_unlock: 3489 mutex_unlock(&fs_info->balance_mutex); 3490 if (need_unlock) 3491 btrfs_exclop_finish(fs_info); 3492 out: 3493 mnt_drop_write_file(file); 3494 return ret; 3495 } 3496 3497 static long btrfs_ioctl_balance_ctl(struct btrfs_fs_info *fs_info, int cmd) 3498 { 3499 if (!capable(CAP_SYS_ADMIN)) 3500 return -EPERM; 3501 3502 switch (cmd) { 3503 case BTRFS_BALANCE_CTL_PAUSE: 3504 return btrfs_pause_balance(fs_info); 3505 case BTRFS_BALANCE_CTL_CANCEL: 3506 return btrfs_cancel_balance(fs_info); 3507 } 3508 3509 return -EINVAL; 3510 } 3511 3512 static long btrfs_ioctl_balance_progress(struct btrfs_fs_info *fs_info, 3513 void __user *arg) 3514 { 3515 struct btrfs_ioctl_balance_args AUTO_KFREE(bargs); 3516 int ret = 0; 3517 3518 if (!capable(CAP_SYS_ADMIN)) 3519 return -EPERM; 3520 3521 mutex_lock(&fs_info->balance_mutex); 3522 if (!fs_info->balance_ctl) { 3523 ret = -ENOTCONN; 3524 goto out; 3525 } 3526 3527 bargs = kzalloc(sizeof(*bargs), GFP_KERNEL); 3528 if (!bargs) { 3529 ret = -ENOMEM; 3530 goto out; 3531 } 3532 3533 btrfs_update_ioctl_balance_args(fs_info, bargs); 3534 3535 if (copy_to_user(arg, bargs, sizeof(*bargs))) 3536 ret = -EFAULT; 3537 out: 3538 mutex_unlock(&fs_info->balance_mutex); 3539 return ret; 3540 } 3541 3542 static long btrfs_ioctl_quota_ctl(struct file *file, void __user *arg) 3543 { 3544 struct inode *inode = file_inode(file); 3545 struct btrfs_fs_info *fs_info = inode_to_fs_info(inode); 3546 struct btrfs_ioctl_quota_ctl_args AUTO_KFREE(sa); 3547 int ret; 3548 3549 if (!capable(CAP_SYS_ADMIN)) 3550 return -EPERM; 3551 3552 ret = mnt_want_write_file(file); 3553 if (ret) 3554 return ret; 3555 3556 sa = memdup_user(arg, sizeof(*sa)); 3557 if (IS_ERR(sa)) { 3558 ret = PTR_ERR(sa); 3559 goto drop_write; 3560 } 3561 3562 switch (sa->cmd) { 3563 case BTRFS_QUOTA_CTL_ENABLE: 3564 case BTRFS_QUOTA_CTL_ENABLE_SIMPLE_QUOTA: 3565 down_write(&fs_info->subvol_sem); 3566 ret = btrfs_quota_enable(fs_info, sa); 3567 up_write(&fs_info->subvol_sem); 3568 break; 3569 case BTRFS_QUOTA_CTL_DISABLE: 3570 /* 3571 * Lock the cleaner mutex to prevent races with concurrent 3572 * relocation, because relocation may be building backrefs for 3573 * blocks of the quota root while we are deleting the root. This 3574 * is like dropping fs roots of deleted snapshots/subvolumes, we 3575 * need the same protection. 3576 * 3577 * This also prevents races between concurrent tasks trying to 3578 * disable quotas, because we will unlock and relock 3579 * qgroup_ioctl_lock across BTRFS_FS_QUOTA_ENABLED changes. 3580 * 3581 * We take this here because we have the dependency of 3582 * 3583 * inode_lock -> subvol_sem 3584 * 3585 * because of rename. With relocation we can prealloc extents, 3586 * so that makes the dependency chain 3587 * 3588 * cleaner_mutex -> inode_lock -> subvol_sem 3589 * 3590 * so we must take the cleaner_mutex here before we take the 3591 * subvol_sem. The deadlock can't actually happen, but this 3592 * quiets lockdep. 3593 */ 3594 mutex_lock(&fs_info->cleaner_mutex); 3595 down_write(&fs_info->subvol_sem); 3596 ret = btrfs_quota_disable(fs_info); 3597 up_write(&fs_info->subvol_sem); 3598 mutex_unlock(&fs_info->cleaner_mutex); 3599 break; 3600 default: 3601 ret = -EINVAL; 3602 break; 3603 } 3604 3605 drop_write: 3606 mnt_drop_write_file(file); 3607 return ret; 3608 } 3609 3610 static long btrfs_ioctl_qgroup_assign(struct file *file, void __user *arg) 3611 { 3612 struct inode *inode = file_inode(file); 3613 struct btrfs_fs_info *fs_info = inode_to_fs_info(inode); 3614 struct btrfs_root *quota_root; 3615 struct btrfs_ioctl_qgroup_assign_args AUTO_KFREE(sa); 3616 struct btrfs_qgroup_list AUTO_KFREE(prealloc); 3617 struct btrfs_trans_handle *trans; 3618 int ret; 3619 int err; 3620 3621 if (!capable(CAP_SYS_ADMIN)) 3622 return -EPERM; 3623 3624 if (!btrfs_qgroup_enabled(fs_info)) 3625 return -ENOTCONN; 3626 3627 ret = mnt_want_write_file(file); 3628 if (ret) 3629 return ret; 3630 3631 sa = memdup_user(arg, sizeof(*sa)); 3632 if (IS_ERR(sa)) { 3633 ret = PTR_ERR(sa); 3634 goto drop_write; 3635 } 3636 3637 if (sa->assign) { 3638 prealloc = kzalloc_obj(*prealloc); 3639 if (!prealloc) { 3640 ret = -ENOMEM; 3641 goto drop_write; 3642 } 3643 } 3644 3645 mutex_lock(&fs_info->qgroup_ioctl_lock); 3646 quota_root = btrfs_grab_root(fs_info->quota_root); 3647 mutex_unlock(&fs_info->qgroup_ioctl_lock); 3648 3649 if (!quota_root) { 3650 ret = -ENOTCONN; 3651 goto drop_write; 3652 } 3653 3654 /* 2 BTRFS_QGROUP_RELATION_KEY items. */ 3655 trans = btrfs_start_transaction(quota_root, 2); 3656 if (IS_ERR(trans)) { 3657 ret = PTR_ERR(trans); 3658 btrfs_put_root(quota_root); 3659 goto drop_write; 3660 } 3661 3662 /* 3663 * Prealloc ownership is moved to the relation handler, there it's used 3664 * or freed on error. 3665 */ 3666 if (sa->assign) { 3667 ret = btrfs_add_qgroup_relation(trans, sa->src, sa->dst, prealloc); 3668 prealloc = NULL; 3669 } else { 3670 ret = btrfs_del_qgroup_relation(trans, sa->src, sa->dst); 3671 } 3672 3673 /* update qgroup status and info */ 3674 mutex_lock(&fs_info->qgroup_ioctl_lock); 3675 err = btrfs_run_qgroups(trans); 3676 mutex_unlock(&fs_info->qgroup_ioctl_lock); 3677 if (err < 0) 3678 btrfs_warn(fs_info, 3679 "qgroup status update failed after %s relation, marked as inconsistent", 3680 sa->assign ? "adding" : "deleting"); 3681 err = btrfs_end_transaction(trans); 3682 btrfs_put_root(quota_root); 3683 if (err && !ret) 3684 ret = err; 3685 3686 drop_write: 3687 mnt_drop_write_file(file); 3688 return ret; 3689 } 3690 3691 static long btrfs_ioctl_qgroup_create(struct file *file, void __user *arg) 3692 { 3693 struct inode *inode = file_inode(file); 3694 struct btrfs_fs_info *fs_info = inode_to_fs_info(inode); 3695 struct btrfs_root *quota_root; 3696 struct btrfs_ioctl_qgroup_create_args AUTO_KFREE(sa); 3697 struct btrfs_trans_handle *trans; 3698 int ret; 3699 int err; 3700 3701 if (!capable(CAP_SYS_ADMIN)) 3702 return -EPERM; 3703 3704 if (!btrfs_qgroup_enabled(fs_info)) 3705 return -ENOTCONN; 3706 3707 ret = mnt_want_write_file(file); 3708 if (ret) 3709 return ret; 3710 3711 sa = memdup_user(arg, sizeof(*sa)); 3712 if (IS_ERR(sa)) { 3713 ret = PTR_ERR(sa); 3714 goto drop_write; 3715 } 3716 3717 if (!sa->qgroupid) { 3718 ret = -EINVAL; 3719 goto drop_write; 3720 } 3721 3722 if (sa->create && btrfs_is_fstree(sa->qgroupid)) { 3723 ret = -EINVAL; 3724 goto drop_write; 3725 } 3726 3727 mutex_lock(&fs_info->qgroup_ioctl_lock); 3728 quota_root = btrfs_grab_root(fs_info->quota_root); 3729 mutex_unlock(&fs_info->qgroup_ioctl_lock); 3730 3731 if (!quota_root) { 3732 ret = -ENOTCONN; 3733 goto drop_write; 3734 } 3735 3736 /* 3737 * 1 BTRFS_QGROUP_INFO_KEY item. 3738 * 1 BTRFS_QGROUP_LIMIT_KEY item. 3739 */ 3740 trans = btrfs_start_transaction(quota_root, 2); 3741 if (IS_ERR(trans)) { 3742 ret = PTR_ERR(trans); 3743 btrfs_put_root(quota_root); 3744 goto drop_write; 3745 } 3746 3747 if (sa->create) { 3748 ret = btrfs_create_qgroup(trans, sa->qgroupid); 3749 } else { 3750 ret = btrfs_remove_qgroup(trans, sa->qgroupid); 3751 } 3752 3753 err = btrfs_end_transaction(trans); 3754 btrfs_put_root(quota_root); 3755 if (err && !ret) 3756 ret = err; 3757 3758 drop_write: 3759 mnt_drop_write_file(file); 3760 return ret; 3761 } 3762 3763 static long btrfs_ioctl_qgroup_limit(struct file *file, void __user *arg) 3764 { 3765 struct inode *inode = file_inode(file); 3766 struct btrfs_root *root = BTRFS_I(inode)->root; 3767 struct btrfs_root *quota_root; 3768 struct btrfs_fs_info *fs_info = root->fs_info; 3769 struct btrfs_ioctl_qgroup_limit_args AUTO_KFREE(sa); 3770 struct btrfs_trans_handle *trans; 3771 int ret; 3772 int err; 3773 u64 qgroupid; 3774 3775 if (!capable(CAP_SYS_ADMIN)) 3776 return -EPERM; 3777 3778 if (!btrfs_qgroup_enabled(fs_info)) 3779 return -ENOTCONN; 3780 3781 ret = mnt_want_write_file(file); 3782 if (ret) 3783 return ret; 3784 3785 sa = memdup_user(arg, sizeof(*sa)); 3786 if (IS_ERR(sa)) { 3787 ret = PTR_ERR(sa); 3788 goto drop_write; 3789 } 3790 3791 mutex_lock(&fs_info->qgroup_ioctl_lock); 3792 quota_root = btrfs_grab_root(fs_info->quota_root); 3793 mutex_unlock(&fs_info->qgroup_ioctl_lock); 3794 3795 if (!quota_root) { 3796 ret = -ENOTCONN; 3797 goto drop_write; 3798 } 3799 3800 /* 1 BTRFS_QGROUP_LIMIT_KEY item. */ 3801 trans = btrfs_start_transaction(quota_root, 1); 3802 if (IS_ERR(trans)) { 3803 ret = PTR_ERR(trans); 3804 btrfs_put_root(quota_root); 3805 goto drop_write; 3806 } 3807 3808 qgroupid = sa->qgroupid; 3809 if (!qgroupid) { 3810 /* take the current subvol as qgroup */ 3811 qgroupid = btrfs_root_id(root); 3812 } 3813 3814 ret = btrfs_limit_qgroup(trans, qgroupid, &sa->lim); 3815 3816 err = btrfs_end_transaction(trans); 3817 btrfs_put_root(quota_root); 3818 if (err && !ret) 3819 ret = err; 3820 3821 drop_write: 3822 mnt_drop_write_file(file); 3823 return ret; 3824 } 3825 3826 static long btrfs_ioctl_quota_rescan(struct file *file, void __user *arg) 3827 { 3828 struct inode *inode = file_inode(file); 3829 struct btrfs_fs_info *fs_info = inode_to_fs_info(inode); 3830 struct btrfs_ioctl_quota_rescan_args AUTO_KFREE(qsa); 3831 int ret; 3832 3833 if (!capable(CAP_SYS_ADMIN)) 3834 return -EPERM; 3835 3836 if (!btrfs_qgroup_enabled(fs_info)) 3837 return -ENOTCONN; 3838 3839 ret = mnt_want_write_file(file); 3840 if (ret) 3841 return ret; 3842 3843 qsa = memdup_user(arg, sizeof(*qsa)); 3844 if (IS_ERR(qsa)) { 3845 ret = PTR_ERR(qsa); 3846 goto drop_write; 3847 } 3848 3849 if (qsa->flags) { 3850 ret = -EINVAL; 3851 goto drop_write; 3852 } 3853 3854 ret = btrfs_qgroup_rescan(fs_info); 3855 3856 drop_write: 3857 mnt_drop_write_file(file); 3858 return ret; 3859 } 3860 3861 static long btrfs_ioctl_quota_rescan_status(struct btrfs_fs_info *fs_info, 3862 void __user *arg) 3863 { 3864 struct btrfs_ioctl_quota_rescan_args qsa = {0}; 3865 3866 if (!capable(CAP_SYS_ADMIN)) 3867 return -EPERM; 3868 3869 if (fs_info->qgroup_flags & BTRFS_QGROUP_STATUS_FLAG_RESCAN) { 3870 qsa.flags = 1; 3871 qsa.progress = fs_info->qgroup_rescan_progress.objectid; 3872 } 3873 3874 if (copy_to_user(arg, &qsa, sizeof(qsa))) 3875 return -EFAULT; 3876 3877 return 0; 3878 } 3879 3880 static long btrfs_ioctl_quota_rescan_wait(struct btrfs_fs_info *fs_info) 3881 { 3882 if (!capable(CAP_SYS_ADMIN)) 3883 return -EPERM; 3884 3885 return btrfs_qgroup_wait_for_completion(fs_info, true); 3886 } 3887 3888 static long _btrfs_ioctl_set_received_subvol(struct file *file, 3889 struct mnt_idmap *idmap, 3890 struct btrfs_ioctl_received_subvol_args *sa) 3891 { 3892 struct inode *inode = file_inode(file); 3893 struct btrfs_fs_info *fs_info = inode_to_fs_info(inode); 3894 struct btrfs_root *root = BTRFS_I(inode)->root; 3895 struct btrfs_root_item *root_item = &root->root_item; 3896 struct btrfs_trans_handle *trans; 3897 struct timespec64 ct = current_time(inode); 3898 int ret = 0; 3899 int received_uuid_changed; 3900 3901 if (!inode_owner_or_capable(idmap, inode)) 3902 return -EPERM; 3903 3904 ret = mnt_want_write_file(file); 3905 if (ret < 0) 3906 return ret; 3907 3908 down_write(&fs_info->subvol_sem); 3909 3910 if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) { 3911 ret = -EINVAL; 3912 goto out; 3913 } 3914 3915 if (btrfs_root_readonly(root)) { 3916 ret = -EROFS; 3917 goto out; 3918 } 3919 3920 received_uuid_changed = memcmp(root_item->received_uuid, sa->uuid, 3921 BTRFS_UUID_SIZE); 3922 3923 /* 3924 * Before we attempt to add the new received uuid, check if we have room 3925 * for it in case there's already an item. If the size of the existing 3926 * item plus this root's ID (u64) exceeds the maximum item size, we can 3927 * return here without the need to abort a transaction. If we don't do 3928 * this check, the btrfs_uuid_tree_add() call below would fail with 3929 * -EOVERFLOW and result in a transaction abort. Malicious users could 3930 * exploit this to turn the fs into RO mode. 3931 */ 3932 if (received_uuid_changed && !btrfs_is_empty_uuid(sa->uuid)) { 3933 ret = btrfs_uuid_tree_check_overflow(fs_info, sa->uuid, 3934 BTRFS_UUID_KEY_RECEIVED_SUBVOL); 3935 if (ret < 0) 3936 goto out; 3937 } 3938 3939 /* 3940 * 1 - root item 3941 * 2 - uuid items (received uuid + subvol uuid) 3942 */ 3943 trans = btrfs_start_transaction(root, 3); 3944 if (IS_ERR(trans)) { 3945 ret = PTR_ERR(trans); 3946 trans = NULL; 3947 goto out; 3948 } 3949 3950 sa->rtransid = trans->transid; 3951 sa->rtime.sec = ct.tv_sec; 3952 sa->rtime.nsec = ct.tv_nsec; 3953 3954 if (received_uuid_changed && 3955 !btrfs_is_empty_uuid(root_item->received_uuid)) { 3956 ret = btrfs_uuid_tree_remove(trans, root_item->received_uuid, 3957 BTRFS_UUID_KEY_RECEIVED_SUBVOL, 3958 btrfs_root_id(root)); 3959 if (unlikely(ret && ret != -ENOENT)) { 3960 btrfs_end_transaction(trans); 3961 goto out; 3962 } 3963 } 3964 memcpy(root_item->received_uuid, sa->uuid, BTRFS_UUID_SIZE); 3965 btrfs_set_root_stransid(root_item, sa->stransid); 3966 btrfs_set_root_rtransid(root_item, sa->rtransid); 3967 btrfs_set_stack_timespec_sec(&root_item->stime, sa->stime.sec); 3968 btrfs_set_stack_timespec_nsec(&root_item->stime, sa->stime.nsec); 3969 btrfs_set_stack_timespec_sec(&root_item->rtime, sa->rtime.sec); 3970 btrfs_set_stack_timespec_nsec(&root_item->rtime, sa->rtime.nsec); 3971 3972 ret = btrfs_update_root(trans, fs_info->tree_root, 3973 &root->root_key, &root->root_item); 3974 if (unlikely(ret < 0)) { 3975 btrfs_abort_transaction(trans, ret); 3976 btrfs_end_transaction(trans); 3977 goto out; 3978 } 3979 if (received_uuid_changed && !btrfs_is_empty_uuid(sa->uuid)) { 3980 ret = btrfs_uuid_tree_add(trans, sa->uuid, 3981 BTRFS_UUID_KEY_RECEIVED_SUBVOL, 3982 btrfs_root_id(root)); 3983 if (unlikely(ret < 0)) { 3984 btrfs_abort_transaction(trans, ret); 3985 btrfs_end_transaction(trans); 3986 goto out; 3987 } 3988 } 3989 ret = btrfs_commit_transaction(trans); 3990 out: 3991 up_write(&fs_info->subvol_sem); 3992 mnt_drop_write_file(file); 3993 return ret; 3994 } 3995 3996 #ifdef CONFIG_64BIT 3997 static long btrfs_ioctl_set_received_subvol_32(struct file *file, 3998 void __user *arg) 3999 { 4000 struct btrfs_ioctl_received_subvol_args_32 AUTO_KFREE(args32); 4001 struct btrfs_ioctl_received_subvol_args AUTO_KFREE(args64); 4002 int ret = 0; 4003 4004 args32 = memdup_user(arg, sizeof(*args32)); 4005 if (IS_ERR(args32)) 4006 return PTR_ERR(args32); 4007 4008 args64 = kmalloc_obj(*args64); 4009 if (!args64) 4010 return -ENOMEM; 4011 4012 memcpy(args64->uuid, args32->uuid, BTRFS_UUID_SIZE); 4013 args64->stransid = args32->stransid; 4014 args64->rtransid = args32->rtransid; 4015 args64->stime.sec = args32->stime.sec; 4016 args64->stime.nsec = args32->stime.nsec; 4017 args64->rtime.sec = args32->rtime.sec; 4018 args64->rtime.nsec = args32->rtime.nsec; 4019 args64->flags = args32->flags; 4020 4021 ret = _btrfs_ioctl_set_received_subvol(file, file_mnt_idmap(file), args64); 4022 if (ret) 4023 return ret; 4024 4025 memcpy(args32->uuid, args64->uuid, BTRFS_UUID_SIZE); 4026 args32->stransid = args64->stransid; 4027 args32->rtransid = args64->rtransid; 4028 args32->stime.sec = args64->stime.sec; 4029 args32->stime.nsec = args64->stime.nsec; 4030 args32->rtime.sec = args64->rtime.sec; 4031 args32->rtime.nsec = args64->rtime.nsec; 4032 args32->flags = args64->flags; 4033 4034 ret = copy_to_user(arg, args32, sizeof(*args32)); 4035 if (ret) 4036 return -EFAULT; 4037 4038 return 0; 4039 } 4040 #endif 4041 4042 static long btrfs_ioctl_set_received_subvol(struct file *file, 4043 void __user *arg) 4044 { 4045 struct btrfs_ioctl_received_subvol_args AUTO_KFREE(sa); 4046 int ret = 0; 4047 4048 sa = memdup_user(arg, sizeof(*sa)); 4049 if (IS_ERR(sa)) 4050 return PTR_ERR(sa); 4051 4052 ret = _btrfs_ioctl_set_received_subvol(file, file_mnt_idmap(file), sa); 4053 if (ret) 4054 return ret; 4055 4056 ret = copy_to_user(arg, sa, sizeof(*sa)); 4057 if (ret) 4058 return -EFAULT; 4059 4060 return 0; 4061 } 4062 4063 static int btrfs_ioctl_get_fslabel(struct btrfs_fs_info *fs_info, 4064 void __user *arg) 4065 { 4066 size_t len; 4067 int ret; 4068 char label[BTRFS_LABEL_SIZE]; 4069 4070 spin_lock(&fs_info->super_lock); 4071 memcpy(label, fs_info->super_copy->label, BTRFS_LABEL_SIZE); 4072 spin_unlock(&fs_info->super_lock); 4073 4074 len = strnlen(label, BTRFS_LABEL_SIZE); 4075 4076 if (len == BTRFS_LABEL_SIZE) { 4077 btrfs_warn(fs_info, 4078 "label is too long, return the first %zu bytes", 4079 --len); 4080 } 4081 4082 ret = copy_to_user(arg, label, len); 4083 4084 return ret ? -EFAULT : 0; 4085 } 4086 4087 static int btrfs_ioctl_set_fslabel(struct file *file, void __user *arg) 4088 { 4089 struct inode *inode = file_inode(file); 4090 struct btrfs_fs_info *fs_info = inode_to_fs_info(inode); 4091 struct btrfs_root *root = BTRFS_I(inode)->root; 4092 struct btrfs_super_block *super_block = fs_info->super_copy; 4093 struct btrfs_trans_handle *trans; 4094 char label[BTRFS_LABEL_SIZE]; 4095 int ret; 4096 4097 if (!capable(CAP_SYS_ADMIN)) 4098 return -EPERM; 4099 4100 if (copy_from_user(label, arg, sizeof(label))) 4101 return -EFAULT; 4102 4103 if (strnlen(label, BTRFS_LABEL_SIZE) == BTRFS_LABEL_SIZE) { 4104 btrfs_err(fs_info, 4105 "unable to set label with more than %d bytes", 4106 BTRFS_LABEL_SIZE - 1); 4107 return -EINVAL; 4108 } 4109 4110 ret = mnt_want_write_file(file); 4111 if (ret) 4112 return ret; 4113 4114 trans = btrfs_start_transaction(root, 0); 4115 if (IS_ERR(trans)) { 4116 ret = PTR_ERR(trans); 4117 goto out_unlock; 4118 } 4119 4120 spin_lock(&fs_info->super_lock); 4121 strscpy(super_block->label, label); 4122 spin_unlock(&fs_info->super_lock); 4123 ret = btrfs_commit_transaction(trans); 4124 4125 out_unlock: 4126 mnt_drop_write_file(file); 4127 return ret; 4128 } 4129 4130 #define INIT_FEATURE_FLAGS(suffix) \ 4131 { .compat_flags = BTRFS_FEATURE_COMPAT_##suffix, \ 4132 .compat_ro_flags = BTRFS_FEATURE_COMPAT_RO_##suffix, \ 4133 .incompat_flags = BTRFS_FEATURE_INCOMPAT_##suffix } 4134 4135 int btrfs_ioctl_get_supported_features(void __user *arg) 4136 { 4137 static const struct btrfs_ioctl_feature_flags features[3] = { 4138 INIT_FEATURE_FLAGS(SUPP), 4139 INIT_FEATURE_FLAGS(SAFE_SET), 4140 INIT_FEATURE_FLAGS(SAFE_CLEAR) 4141 }; 4142 4143 if (copy_to_user(arg, &features, sizeof(features))) 4144 return -EFAULT; 4145 4146 return 0; 4147 } 4148 4149 static int btrfs_ioctl_get_features(struct btrfs_fs_info *fs_info, 4150 void __user *arg) 4151 { 4152 struct btrfs_super_block *super_block = fs_info->super_copy; 4153 struct btrfs_ioctl_feature_flags features; 4154 4155 features.compat_flags = btrfs_super_compat_flags(super_block); 4156 features.compat_ro_flags = btrfs_super_compat_ro_flags(super_block); 4157 features.incompat_flags = btrfs_super_incompat_flags(super_block); 4158 4159 if (copy_to_user(arg, &features, sizeof(features))) 4160 return -EFAULT; 4161 4162 return 0; 4163 } 4164 4165 static int check_feature_bits(const struct btrfs_fs_info *fs_info, 4166 enum btrfs_feature_set set, 4167 u64 change_mask, u64 flags, u64 supported_flags, 4168 u64 safe_set, u64 safe_clear) 4169 { 4170 const char *type = btrfs_feature_set_name(set); 4171 const char AUTO_KFREE(names); 4172 u64 disallowed, unsupported; 4173 u64 set_mask = flags & change_mask; 4174 u64 clear_mask = ~flags & change_mask; 4175 4176 unsupported = set_mask & ~supported_flags; 4177 if (unsupported) { 4178 names = btrfs_printable_features(set, unsupported); 4179 if (names) 4180 btrfs_warn(fs_info, 4181 "this kernel does not support the %s feature bit%s", 4182 names, strchr(names, ',') ? "s" : ""); 4183 else 4184 btrfs_warn(fs_info, 4185 "this kernel does not support %s bits 0x%llx", 4186 type, unsupported); 4187 return -EOPNOTSUPP; 4188 } 4189 4190 disallowed = set_mask & ~safe_set; 4191 if (disallowed) { 4192 names = btrfs_printable_features(set, disallowed); 4193 if (names) 4194 btrfs_warn(fs_info, 4195 "can't set the %s feature bit%s while mounted", 4196 names, strchr(names, ',') ? "s" : ""); 4197 else 4198 btrfs_warn(fs_info, 4199 "can't set %s bits 0x%llx while mounted", 4200 type, disallowed); 4201 return -EPERM; 4202 } 4203 4204 disallowed = clear_mask & ~safe_clear; 4205 if (disallowed) { 4206 names = btrfs_printable_features(set, disallowed); 4207 if (names) 4208 btrfs_warn(fs_info, 4209 "can't clear the %s feature bit%s while mounted", 4210 names, strchr(names, ',') ? "s" : ""); 4211 else 4212 btrfs_warn(fs_info, 4213 "can't clear %s bits 0x%llx while mounted", 4214 type, disallowed); 4215 return -EPERM; 4216 } 4217 4218 return 0; 4219 } 4220 4221 #define check_feature(fs_info, change_mask, flags, mask_base) \ 4222 check_feature_bits(fs_info, FEAT_##mask_base, change_mask, flags, \ 4223 BTRFS_FEATURE_ ## mask_base ## _SUPP, \ 4224 BTRFS_FEATURE_ ## mask_base ## _SAFE_SET, \ 4225 BTRFS_FEATURE_ ## mask_base ## _SAFE_CLEAR) 4226 4227 static int btrfs_ioctl_set_features(struct file *file, void __user *arg) 4228 { 4229 struct inode *inode = file_inode(file); 4230 struct btrfs_fs_info *fs_info = inode_to_fs_info(inode); 4231 struct btrfs_root *root = BTRFS_I(inode)->root; 4232 struct btrfs_super_block *super_block = fs_info->super_copy; 4233 struct btrfs_ioctl_feature_flags flags[2]; 4234 struct btrfs_trans_handle *trans; 4235 u64 newflags; 4236 int ret; 4237 4238 if (!capable(CAP_SYS_ADMIN)) 4239 return -EPERM; 4240 4241 if (copy_from_user(flags, arg, sizeof(flags))) 4242 return -EFAULT; 4243 4244 /* Nothing to do */ 4245 if (!flags[0].compat_flags && !flags[0].compat_ro_flags && 4246 !flags[0].incompat_flags) 4247 return 0; 4248 4249 ret = check_feature(fs_info, flags[0].compat_flags, 4250 flags[1].compat_flags, COMPAT); 4251 if (ret) 4252 return ret; 4253 4254 ret = check_feature(fs_info, flags[0].compat_ro_flags, 4255 flags[1].compat_ro_flags, COMPAT_RO); 4256 if (ret) 4257 return ret; 4258 4259 ret = check_feature(fs_info, flags[0].incompat_flags, 4260 flags[1].incompat_flags, INCOMPAT); 4261 if (ret) 4262 return ret; 4263 4264 ret = mnt_want_write_file(file); 4265 if (ret) 4266 return ret; 4267 4268 trans = btrfs_start_transaction(root, 0); 4269 if (IS_ERR(trans)) { 4270 ret = PTR_ERR(trans); 4271 goto out_drop_write; 4272 } 4273 4274 spin_lock(&fs_info->super_lock); 4275 newflags = btrfs_super_compat_flags(super_block); 4276 newflags |= flags[0].compat_flags & flags[1].compat_flags; 4277 newflags &= ~(flags[0].compat_flags & ~flags[1].compat_flags); 4278 btrfs_set_super_compat_flags(super_block, newflags); 4279 4280 newflags = btrfs_super_compat_ro_flags(super_block); 4281 newflags |= flags[0].compat_ro_flags & flags[1].compat_ro_flags; 4282 newflags &= ~(flags[0].compat_ro_flags & ~flags[1].compat_ro_flags); 4283 btrfs_set_super_compat_ro_flags(super_block, newflags); 4284 4285 newflags = btrfs_super_incompat_flags(super_block); 4286 newflags |= flags[0].incompat_flags & flags[1].incompat_flags; 4287 newflags &= ~(flags[0].incompat_flags & ~flags[1].incompat_flags); 4288 btrfs_set_super_incompat_flags(super_block, newflags); 4289 spin_unlock(&fs_info->super_lock); 4290 4291 ret = btrfs_commit_transaction(trans); 4292 out_drop_write: 4293 mnt_drop_write_file(file); 4294 4295 return ret; 4296 } 4297 4298 static int _btrfs_ioctl_send(struct btrfs_root *root, void __user *argp, bool compat) 4299 { 4300 struct btrfs_ioctl_send_args AUTO_KFREE(arg); 4301 4302 if (compat) { 4303 #if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT) 4304 int ret; 4305 struct btrfs_ioctl_send_args_32 args32 = { 0 }; 4306 4307 ret = copy_from_user(&args32, argp, sizeof(args32)); 4308 if (ret) 4309 return -EFAULT; 4310 arg = kzalloc_obj(*arg); 4311 if (!arg) 4312 return -ENOMEM; 4313 arg->send_fd = args32.send_fd; 4314 arg->clone_sources_count = args32.clone_sources_count; 4315 arg->clone_sources = compat_ptr(args32.clone_sources); 4316 arg->parent_root = args32.parent_root; 4317 arg->flags = args32.flags; 4318 arg->version = args32.version; 4319 memcpy(arg->reserved, args32.reserved, 4320 sizeof(args32.reserved)); 4321 #else 4322 return -ENOTTY; 4323 #endif 4324 } else { 4325 arg = memdup_user(argp, sizeof(*arg)); 4326 if (IS_ERR(arg)) 4327 return PTR_ERR(arg); 4328 } 4329 return btrfs_ioctl_send(root, arg); 4330 } 4331 4332 static int btrfs_ioctl_encoded_read(struct file *file, void __user *argp, 4333 bool compat) 4334 { 4335 struct btrfs_ioctl_encoded_io_args args = { 0 }; 4336 size_t copy_end_kernel = offsetofend(struct btrfs_ioctl_encoded_io_args, 4337 flags); 4338 size_t copy_end; 4339 struct btrfs_inode *inode = BTRFS_I(file_inode(file)); 4340 struct btrfs_fs_info *fs_info = inode->root->fs_info; 4341 struct extent_io_tree *io_tree = &inode->io_tree; 4342 struct iovec iovstack[UIO_FASTIOV]; 4343 struct iovec *iov = iovstack; 4344 struct iov_iter iter; 4345 loff_t pos; 4346 struct kiocb kiocb; 4347 ssize_t ret; 4348 u64 disk_bytenr, disk_io_size; 4349 struct extent_state *cached_state = NULL; 4350 4351 if (!capable(CAP_SYS_ADMIN)) { 4352 ret = -EPERM; 4353 goto out_acct; 4354 } 4355 4356 if (compat) { 4357 #if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT) 4358 struct btrfs_ioctl_encoded_io_args_32 args32; 4359 4360 copy_end = offsetofend(struct btrfs_ioctl_encoded_io_args_32, 4361 flags); 4362 if (copy_from_user(&args32, argp, copy_end)) { 4363 ret = -EFAULT; 4364 goto out_acct; 4365 } 4366 args.iov = compat_ptr(args32.iov); 4367 args.iovcnt = args32.iovcnt; 4368 args.offset = args32.offset; 4369 args.flags = args32.flags; 4370 #else 4371 return -ENOTTY; 4372 #endif 4373 } else { 4374 copy_end = copy_end_kernel; 4375 if (copy_from_user(&args, argp, copy_end)) { 4376 ret = -EFAULT; 4377 goto out_acct; 4378 } 4379 } 4380 if (args.flags != 0) { 4381 ret = -EINVAL; 4382 goto out_acct; 4383 } 4384 4385 ret = import_iovec(ITER_DEST, args.iov, args.iovcnt, ARRAY_SIZE(iovstack), 4386 &iov, &iter); 4387 if (ret < 0) 4388 goto out_acct; 4389 4390 if (iov_iter_count(&iter) == 0) { 4391 ret = 0; 4392 goto out_iov; 4393 } 4394 pos = args.offset; 4395 ret = rw_verify_area(READ, file, &pos, args.len); 4396 if (ret < 0) 4397 goto out_iov; 4398 4399 init_sync_kiocb(&kiocb, file); 4400 kiocb.ki_pos = pos; 4401 4402 ret = btrfs_encoded_read(&kiocb, &iter, &args, &cached_state, 4403 &disk_bytenr, &disk_io_size); 4404 4405 if (ret == -EIOCBQUEUED) { 4406 bool unlocked = false; 4407 u64 start, lockend, count; 4408 4409 start = ALIGN_DOWN(kiocb.ki_pos, fs_info->sectorsize); 4410 lockend = start + BTRFS_MAX_UNCOMPRESSED - 1; 4411 4412 if (args.compression) 4413 count = disk_io_size; 4414 else 4415 count = args.len; 4416 4417 ret = btrfs_encoded_read_regular(&kiocb, &iter, start, lockend, 4418 &cached_state, disk_bytenr, 4419 disk_io_size, count, 4420 args.compression, &unlocked); 4421 4422 if (!unlocked) { 4423 btrfs_unlock_extent(io_tree, start, lockend, &cached_state); 4424 btrfs_inode_unlock(inode, BTRFS_ILOCK_SHARED); 4425 } 4426 } 4427 4428 if (ret >= 0) { 4429 fsnotify_access(file); 4430 if (copy_to_user(argp + copy_end, 4431 (char *)&args + copy_end_kernel, 4432 sizeof(args) - copy_end_kernel)) 4433 ret = -EFAULT; 4434 } 4435 4436 out_iov: 4437 kfree(iov); 4438 out_acct: 4439 if (ret > 0) 4440 add_rchar(current, ret); 4441 inc_syscr(current); 4442 return ret; 4443 } 4444 4445 static int btrfs_ioctl_encoded_write(struct file *file, void __user *argp, bool compat) 4446 { 4447 struct btrfs_ioctl_encoded_io_args args; 4448 struct iovec iovstack[UIO_FASTIOV]; 4449 struct iovec *iov = iovstack; 4450 struct iov_iter iter; 4451 loff_t pos; 4452 struct kiocb kiocb; 4453 ssize_t ret; 4454 4455 if (!capable(CAP_SYS_ADMIN)) { 4456 ret = -EPERM; 4457 goto out_acct; 4458 } 4459 4460 if (!(file->f_mode & FMODE_WRITE)) { 4461 ret = -EBADF; 4462 goto out_acct; 4463 } 4464 4465 if (compat) { 4466 #if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT) 4467 struct btrfs_ioctl_encoded_io_args_32 args32; 4468 4469 if (copy_from_user(&args32, argp, sizeof(args32))) { 4470 ret = -EFAULT; 4471 goto out_acct; 4472 } 4473 args.iov = compat_ptr(args32.iov); 4474 args.iovcnt = args32.iovcnt; 4475 args.offset = args32.offset; 4476 args.flags = args32.flags; 4477 args.len = args32.len; 4478 args.unencoded_len = args32.unencoded_len; 4479 args.unencoded_offset = args32.unencoded_offset; 4480 args.compression = args32.compression; 4481 args.encryption = args32.encryption; 4482 memcpy(args.reserved, args32.reserved, sizeof(args.reserved)); 4483 #else 4484 return -ENOTTY; 4485 #endif 4486 } else { 4487 if (copy_from_user(&args, argp, sizeof(args))) { 4488 ret = -EFAULT; 4489 goto out_acct; 4490 } 4491 } 4492 4493 ret = -EINVAL; 4494 if (args.flags != 0) 4495 goto out_acct; 4496 if (memchr_inv(args.reserved, 0, sizeof(args.reserved))) 4497 goto out_acct; 4498 if (args.compression == BTRFS_ENCODED_IO_COMPRESSION_NONE && 4499 args.encryption == BTRFS_ENCODED_IO_ENCRYPTION_NONE) 4500 goto out_acct; 4501 if (args.compression >= BTRFS_ENCODED_IO_COMPRESSION_TYPES || 4502 args.encryption >= BTRFS_ENCODED_IO_ENCRYPTION_TYPES) 4503 goto out_acct; 4504 if (args.unencoded_offset > args.unencoded_len) 4505 goto out_acct; 4506 if (args.len > args.unencoded_len - args.unencoded_offset) 4507 goto out_acct; 4508 4509 ret = import_iovec(ITER_SOURCE, args.iov, args.iovcnt, ARRAY_SIZE(iovstack), 4510 &iov, &iter); 4511 if (ret < 0) 4512 goto out_acct; 4513 4514 if (iov_iter_count(&iter) == 0) { 4515 ret = 0; 4516 goto out_iov; 4517 } 4518 pos = args.offset; 4519 ret = rw_verify_area(WRITE, file, &pos, args.len); 4520 if (ret < 0) 4521 goto out_iov; 4522 4523 init_sync_kiocb(&kiocb, file); 4524 ret = kiocb_set_rw_flags(&kiocb, 0, WRITE); 4525 if (ret) 4526 goto out_iov; 4527 kiocb.ki_pos = pos; 4528 4529 file_start_write(file); 4530 4531 ret = btrfs_do_write_iter(&kiocb, &iter, &args); 4532 if (ret > 0) 4533 fsnotify_modify(file); 4534 4535 file_end_write(file); 4536 out_iov: 4537 kfree(iov); 4538 out_acct: 4539 if (ret > 0) 4540 add_wchar(current, ret); 4541 inc_syscw(current); 4542 return ret; 4543 } 4544 4545 struct btrfs_uring_encoded_data { 4546 struct btrfs_ioctl_encoded_io_args args; 4547 struct iovec iovstack[UIO_FASTIOV]; 4548 struct iovec *iov; 4549 struct iov_iter iter; 4550 }; 4551 4552 /* 4553 * Context that's attached to an encoded read io_uring command, in cmd->pdu. It 4554 * contains the fields in btrfs_uring_read_extent that are necessary to finish 4555 * off and cleanup the I/O in btrfs_uring_read_finished. 4556 */ 4557 struct btrfs_uring_priv { 4558 struct io_uring_cmd *cmd; 4559 struct page **pages; 4560 unsigned long nr_pages; 4561 struct kiocb iocb; 4562 struct iovec *iov; 4563 struct iov_iter iter; 4564 struct extent_state *cached_state; 4565 u64 count; 4566 u64 start; 4567 u64 lockend; 4568 int err; 4569 bool compressed; 4570 }; 4571 4572 struct io_btrfs_cmd { 4573 struct btrfs_uring_encoded_data *data; 4574 struct btrfs_uring_priv *priv; 4575 }; 4576 4577 static void btrfs_uring_read_finished(struct io_tw_req tw_req, io_tw_token_t tw) 4578 { 4579 struct io_uring_cmd *cmd = io_uring_cmd_from_tw(tw_req); 4580 struct io_btrfs_cmd *bc = io_uring_cmd_to_pdu(cmd, struct io_btrfs_cmd); 4581 struct btrfs_uring_priv *priv = bc->priv; 4582 struct btrfs_inode *inode = BTRFS_I(file_inode(priv->iocb.ki_filp)); 4583 struct extent_io_tree *io_tree = &inode->io_tree; 4584 pgoff_t index; 4585 u64 cur; 4586 size_t page_offset; 4587 ssize_t ret; 4588 4589 /* The inode lock has already been acquired in btrfs_uring_read_extent. */ 4590 btrfs_lockdep_inode_acquire(inode, i_rwsem); 4591 4592 if (priv->err) { 4593 ret = priv->err; 4594 goto out; 4595 } 4596 4597 if (priv->compressed) { 4598 index = 0; 4599 page_offset = 0; 4600 } else { 4601 index = (priv->iocb.ki_pos - priv->start) >> PAGE_SHIFT; 4602 page_offset = offset_in_page(priv->iocb.ki_pos - priv->start); 4603 } 4604 cur = 0; 4605 while (cur < priv->count) { 4606 size_t bytes = min_t(size_t, priv->count - cur, PAGE_SIZE - page_offset); 4607 4608 if (copy_page_to_iter(priv->pages[index], page_offset, bytes, 4609 &priv->iter) != bytes) { 4610 ret = -EFAULT; 4611 goto out; 4612 } 4613 4614 index++; 4615 cur += bytes; 4616 page_offset = 0; 4617 } 4618 ret = priv->count; 4619 4620 out: 4621 btrfs_unlock_extent(io_tree, priv->start, priv->lockend, &priv->cached_state); 4622 btrfs_inode_unlock(inode, BTRFS_ILOCK_SHARED); 4623 4624 io_uring_cmd_done(cmd, ret, IO_URING_CMD_TASK_WORK_ISSUE_FLAGS); 4625 add_rchar(current, ret); 4626 4627 for (index = 0; index < priv->nr_pages; index++) 4628 __free_page(priv->pages[index]); 4629 4630 kfree(priv->pages); 4631 kfree(priv->iov); 4632 kfree(priv); 4633 kfree(bc->data); 4634 } 4635 4636 void btrfs_uring_read_extent_endio(void *ctx, int err) 4637 { 4638 struct btrfs_uring_priv *priv = ctx; 4639 struct io_btrfs_cmd *bc = io_uring_cmd_to_pdu(priv->cmd, struct io_btrfs_cmd); 4640 4641 priv->err = err; 4642 bc->priv = priv; 4643 4644 io_uring_cmd_complete_in_task(priv->cmd, btrfs_uring_read_finished); 4645 } 4646 4647 static int btrfs_uring_read_extent(struct kiocb *iocb, struct iov_iter *iter, 4648 u64 start, u64 lockend, 4649 struct extent_state *cached_state, 4650 u64 disk_bytenr, u64 disk_io_size, 4651 size_t count, bool compressed, 4652 struct iovec *iov, struct io_uring_cmd *cmd) 4653 { 4654 struct btrfs_inode *inode = BTRFS_I(file_inode(iocb->ki_filp)); 4655 struct extent_io_tree *io_tree = &inode->io_tree; 4656 struct page **pages = NULL; 4657 struct btrfs_uring_priv *priv = NULL; 4658 unsigned long nr_pages; 4659 int ret; 4660 4661 nr_pages = DIV_ROUND_UP(disk_io_size, PAGE_SIZE); 4662 pages = kzalloc_objs(struct page *, nr_pages, GFP_NOFS); 4663 if (!pages) 4664 return -ENOMEM; 4665 ret = btrfs_alloc_page_array(nr_pages, pages, GFP_NOFS); 4666 if (ret) { 4667 ret = -ENOMEM; 4668 goto out_fail; 4669 } 4670 4671 priv = kmalloc_obj(*priv, GFP_NOFS); 4672 if (!priv) { 4673 ret = -ENOMEM; 4674 goto out_fail; 4675 } 4676 4677 priv->iocb = *iocb; 4678 priv->iov = iov; 4679 priv->iter = *iter; 4680 priv->count = count; 4681 priv->cmd = cmd; 4682 priv->cached_state = cached_state; 4683 priv->compressed = compressed; 4684 priv->nr_pages = nr_pages; 4685 priv->pages = pages; 4686 priv->start = start; 4687 priv->lockend = lockend; 4688 priv->err = 0; 4689 4690 ret = btrfs_encoded_read_regular_fill_pages(inode, disk_bytenr, 4691 disk_io_size, pages, priv); 4692 if (ret && ret != -EIOCBQUEUED) 4693 goto out_fail; 4694 4695 /* 4696 * If we return -EIOCBQUEUED, we're deferring the cleanup to 4697 * btrfs_uring_read_finished(), which will handle unlocking the extent 4698 * and inode and freeing the allocations. 4699 */ 4700 4701 /* 4702 * We're returning to userspace with the inode lock held, and that's 4703 * okay - it'll get unlocked in a worker thread. Call 4704 * btrfs_lockdep_inode_release() to avoid confusing lockdep. 4705 */ 4706 btrfs_lockdep_inode_release(inode, i_rwsem); 4707 4708 return -EIOCBQUEUED; 4709 4710 out_fail: 4711 btrfs_unlock_extent(io_tree, start, lockend, &cached_state); 4712 btrfs_inode_unlock(inode, BTRFS_ILOCK_SHARED); 4713 kfree(priv); 4714 for (int i = 0; i < nr_pages; i++) { 4715 if (pages[i]) 4716 __free_page(pages[i]); 4717 } 4718 kfree(pages); 4719 return ret; 4720 } 4721 4722 static int btrfs_uring_encoded_read(struct io_uring_cmd *cmd, unsigned int issue_flags) 4723 { 4724 struct file *file = cmd->file; 4725 struct btrfs_inode *inode = BTRFS_I(file->f_inode); 4726 struct extent_io_tree *io_tree = &inode->io_tree; 4727 struct btrfs_fs_info *fs_info = inode->root->fs_info; 4728 size_t copy_end_kernel = offsetofend(struct btrfs_ioctl_encoded_io_args, flags); 4729 size_t copy_end; 4730 int ret; 4731 u64 disk_bytenr, disk_io_size; 4732 loff_t pos; 4733 struct kiocb kiocb; 4734 struct extent_state *cached_state = NULL; 4735 u64 start, lockend; 4736 void __user *sqe_addr; 4737 struct io_btrfs_cmd *bc = io_uring_cmd_to_pdu(cmd, struct io_btrfs_cmd); 4738 struct btrfs_uring_encoded_data *data = NULL; 4739 4740 if (cmd->flags & IORING_URING_CMD_REISSUE) 4741 data = bc->data; 4742 4743 if (!capable(CAP_SYS_ADMIN)) { 4744 ret = -EPERM; 4745 goto out_acct; 4746 } 4747 sqe_addr = u64_to_user_ptr(READ_ONCE(cmd->sqe->addr)); 4748 4749 if (issue_flags & IO_URING_F_COMPAT) { 4750 #if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT) 4751 copy_end = offsetofend(struct btrfs_ioctl_encoded_io_args_32, flags); 4752 #else 4753 ret = -ENOTTY; 4754 goto out_acct; 4755 #endif 4756 } else { 4757 copy_end = copy_end_kernel; 4758 } 4759 4760 if (!data) { 4761 data = kzalloc_obj(*data, GFP_NOFS); 4762 if (!data) { 4763 ret = -ENOMEM; 4764 goto out_acct; 4765 } 4766 4767 bc->data = data; 4768 4769 if (issue_flags & IO_URING_F_COMPAT) { 4770 #if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT) 4771 struct btrfs_ioctl_encoded_io_args_32 args32; 4772 4773 if (copy_from_user(&args32, sqe_addr, copy_end)) { 4774 ret = -EFAULT; 4775 goto out_acct; 4776 } 4777 4778 data->args.iov = compat_ptr(args32.iov); 4779 data->args.iovcnt = args32.iovcnt; 4780 data->args.offset = args32.offset; 4781 data->args.flags = args32.flags; 4782 #endif 4783 } else { 4784 if (copy_from_user(&data->args, sqe_addr, copy_end)) { 4785 ret = -EFAULT; 4786 goto out_acct; 4787 } 4788 } 4789 4790 if (data->args.flags != 0) { 4791 ret = -EINVAL; 4792 goto out_acct; 4793 } 4794 4795 data->iov = data->iovstack; 4796 ret = import_iovec(ITER_DEST, data->args.iov, data->args.iovcnt, 4797 ARRAY_SIZE(data->iovstack), &data->iov, 4798 &data->iter); 4799 if (ret < 0) 4800 goto out_acct; 4801 4802 if (iov_iter_count(&data->iter) == 0) { 4803 ret = 0; 4804 goto out_free; 4805 } 4806 } 4807 4808 pos = data->args.offset; 4809 ret = rw_verify_area(READ, file, &pos, data->args.len); 4810 if (ret < 0) 4811 goto out_free; 4812 4813 init_sync_kiocb(&kiocb, file); 4814 kiocb.ki_pos = pos; 4815 4816 if (issue_flags & IO_URING_F_NONBLOCK) 4817 kiocb.ki_flags |= IOCB_NOWAIT; 4818 4819 start = ALIGN_DOWN(pos, fs_info->sectorsize); 4820 lockend = start + BTRFS_MAX_UNCOMPRESSED - 1; 4821 4822 ret = btrfs_encoded_read(&kiocb, &data->iter, &data->args, &cached_state, 4823 &disk_bytenr, &disk_io_size); 4824 if (ret == -EAGAIN) 4825 goto out_acct; 4826 if (ret < 0 && ret != -EIOCBQUEUED) 4827 goto out_free; 4828 4829 file_accessed(file); 4830 4831 if (copy_to_user(sqe_addr + copy_end, 4832 (const char *)&data->args + copy_end_kernel, 4833 sizeof(data->args) - copy_end_kernel)) { 4834 if (ret == -EIOCBQUEUED) { 4835 btrfs_unlock_extent(io_tree, start, lockend, &cached_state); 4836 btrfs_inode_unlock(inode, BTRFS_ILOCK_SHARED); 4837 } 4838 ret = -EFAULT; 4839 goto out_free; 4840 } 4841 4842 if (ret == -EIOCBQUEUED) { 4843 u64 count = min_t(u64, iov_iter_count(&data->iter), disk_io_size); 4844 4845 /* Match ioctl by not returning past EOF if uncompressed. */ 4846 if (!data->args.compression) 4847 count = min_t(u64, count, data->args.len); 4848 4849 ret = btrfs_uring_read_extent(&kiocb, &data->iter, start, lockend, 4850 cached_state, disk_bytenr, disk_io_size, 4851 count, data->args.compression, 4852 data->iov, cmd); 4853 4854 goto out_acct; 4855 } 4856 4857 out_free: 4858 kfree(data->iov); 4859 4860 out_acct: 4861 if (ret > 0) 4862 add_rchar(current, ret); 4863 inc_syscr(current); 4864 4865 if (ret != -EIOCBQUEUED && ret != -EAGAIN) 4866 kfree(data); 4867 4868 return ret; 4869 } 4870 4871 static int btrfs_uring_encoded_write(struct io_uring_cmd *cmd, unsigned int issue_flags) 4872 { 4873 struct file *file = cmd->file; 4874 loff_t pos; 4875 struct kiocb kiocb; 4876 ssize_t ret; 4877 void __user *sqe_addr; 4878 struct io_btrfs_cmd *bc = io_uring_cmd_to_pdu(cmd, struct io_btrfs_cmd); 4879 struct btrfs_uring_encoded_data *data = NULL; 4880 4881 if (cmd->flags & IORING_URING_CMD_REISSUE) 4882 data = bc->data; 4883 4884 if (!capable(CAP_SYS_ADMIN)) { 4885 ret = -EPERM; 4886 goto out_acct; 4887 } 4888 sqe_addr = u64_to_user_ptr(READ_ONCE(cmd->sqe->addr)); 4889 4890 if (!(file->f_mode & FMODE_WRITE)) { 4891 ret = -EBADF; 4892 goto out_acct; 4893 } 4894 4895 if (!data) { 4896 data = kzalloc_obj(*data, GFP_NOFS); 4897 if (!data) { 4898 ret = -ENOMEM; 4899 goto out_acct; 4900 } 4901 4902 bc->data = data; 4903 4904 if (issue_flags & IO_URING_F_COMPAT) { 4905 #if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT) 4906 struct btrfs_ioctl_encoded_io_args_32 args32; 4907 4908 if (copy_from_user(&args32, sqe_addr, sizeof(args32))) { 4909 ret = -EFAULT; 4910 goto out_acct; 4911 } 4912 data->args.iov = compat_ptr(args32.iov); 4913 data->args.iovcnt = args32.iovcnt; 4914 data->args.offset = args32.offset; 4915 data->args.flags = args32.flags; 4916 data->args.len = args32.len; 4917 data->args.unencoded_len = args32.unencoded_len; 4918 data->args.unencoded_offset = args32.unencoded_offset; 4919 data->args.compression = args32.compression; 4920 data->args.encryption = args32.encryption; 4921 memcpy(data->args.reserved, args32.reserved, 4922 sizeof(data->args.reserved)); 4923 #else 4924 ret = -ENOTTY; 4925 goto out_acct; 4926 #endif 4927 } else { 4928 if (copy_from_user(&data->args, sqe_addr, sizeof(data->args))) { 4929 ret = -EFAULT; 4930 goto out_acct; 4931 } 4932 } 4933 4934 ret = -EINVAL; 4935 if (data->args.flags != 0) 4936 goto out_acct; 4937 if (memchr_inv(data->args.reserved, 0, sizeof(data->args.reserved))) 4938 goto out_acct; 4939 if (data->args.compression == BTRFS_ENCODED_IO_COMPRESSION_NONE && 4940 data->args.encryption == BTRFS_ENCODED_IO_ENCRYPTION_NONE) 4941 goto out_acct; 4942 if (data->args.compression >= BTRFS_ENCODED_IO_COMPRESSION_TYPES || 4943 data->args.encryption >= BTRFS_ENCODED_IO_ENCRYPTION_TYPES) 4944 goto out_acct; 4945 if (data->args.unencoded_offset > data->args.unencoded_len) 4946 goto out_acct; 4947 if (data->args.len > data->args.unencoded_len - data->args.unencoded_offset) 4948 goto out_acct; 4949 4950 data->iov = data->iovstack; 4951 ret = import_iovec(ITER_SOURCE, data->args.iov, data->args.iovcnt, 4952 ARRAY_SIZE(data->iovstack), &data->iov, 4953 &data->iter); 4954 if (ret < 0) 4955 goto out_acct; 4956 4957 if (iov_iter_count(&data->iter) == 0) { 4958 ret = 0; 4959 goto out_iov; 4960 } 4961 } 4962 4963 if (issue_flags & IO_URING_F_NONBLOCK) { 4964 ret = -EAGAIN; 4965 goto out_acct; 4966 } 4967 4968 pos = data->args.offset; 4969 ret = rw_verify_area(WRITE, file, &pos, data->args.len); 4970 if (ret < 0) 4971 goto out_iov; 4972 4973 init_sync_kiocb(&kiocb, file); 4974 ret = kiocb_set_rw_flags(&kiocb, 0, WRITE); 4975 if (ret) 4976 goto out_iov; 4977 kiocb.ki_pos = pos; 4978 4979 file_start_write(file); 4980 4981 ret = btrfs_do_write_iter(&kiocb, &data->iter, &data->args); 4982 if (ret > 0) 4983 fsnotify_modify(file); 4984 4985 file_end_write(file); 4986 out_iov: 4987 kfree(data->iov); 4988 out_acct: 4989 if (ret > 0) 4990 add_wchar(current, ret); 4991 inc_syscw(current); 4992 4993 if (ret != -EAGAIN) 4994 kfree(data); 4995 return ret; 4996 } 4997 4998 int btrfs_uring_cmd(struct io_uring_cmd *cmd, unsigned int issue_flags) 4999 { 5000 if (btrfs_is_shutdown(inode_to_fs_info(file_inode(cmd->file)))) 5001 return -EIO; 5002 5003 switch (cmd->cmd_op) { 5004 case BTRFS_IOC_ENCODED_READ: 5005 #if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT) 5006 case BTRFS_IOC_ENCODED_READ_32: 5007 #endif 5008 return btrfs_uring_encoded_read(cmd, issue_flags); 5009 5010 case BTRFS_IOC_ENCODED_WRITE: 5011 #if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT) 5012 case BTRFS_IOC_ENCODED_WRITE_32: 5013 #endif 5014 return btrfs_uring_encoded_write(cmd, issue_flags); 5015 } 5016 5017 return -EINVAL; 5018 } 5019 5020 static int btrfs_ioctl_subvol_sync(struct btrfs_fs_info *fs_info, void __user *argp) 5021 { 5022 struct btrfs_root *root; 5023 struct btrfs_ioctl_subvol_wait args = { 0 }; 5024 signed long sched_ret; 5025 int refs; 5026 u64 root_flags; 5027 bool wait_for_deletion = false; 5028 bool found = false; 5029 5030 if (copy_from_user(&args, argp, sizeof(args))) 5031 return -EFAULT; 5032 5033 switch (args.mode) { 5034 case BTRFS_SUBVOL_SYNC_WAIT_FOR_QUEUED: 5035 /* 5036 * Wait for the first one deleted that waits until all previous 5037 * are cleaned. 5038 */ 5039 spin_lock(&fs_info->trans_lock); 5040 if (!list_empty(&fs_info->dead_roots)) { 5041 root = list_last_entry(&fs_info->dead_roots, 5042 struct btrfs_root, root_list); 5043 args.subvolid = btrfs_root_id(root); 5044 found = true; 5045 } 5046 spin_unlock(&fs_info->trans_lock); 5047 if (!found) 5048 return -ENOENT; 5049 5050 fallthrough; 5051 case BTRFS_SUBVOL_SYNC_WAIT_FOR_ONE: 5052 if ((0 < args.subvolid && args.subvolid < BTRFS_FIRST_FREE_OBJECTID) || 5053 BTRFS_LAST_FREE_OBJECTID < args.subvolid) 5054 return -EINVAL; 5055 break; 5056 case BTRFS_SUBVOL_SYNC_COUNT: 5057 spin_lock(&fs_info->trans_lock); 5058 args.count = list_count_nodes(&fs_info->dead_roots); 5059 spin_unlock(&fs_info->trans_lock); 5060 if (copy_to_user(argp, &args, sizeof(args))) 5061 return -EFAULT; 5062 return 0; 5063 case BTRFS_SUBVOL_SYNC_PEEK_FIRST: 5064 spin_lock(&fs_info->trans_lock); 5065 /* Last in the list was deleted first. */ 5066 if (!list_empty(&fs_info->dead_roots)) { 5067 root = list_last_entry(&fs_info->dead_roots, 5068 struct btrfs_root, root_list); 5069 args.subvolid = btrfs_root_id(root); 5070 } else { 5071 args.subvolid = 0; 5072 } 5073 spin_unlock(&fs_info->trans_lock); 5074 if (copy_to_user(argp, &args, sizeof(args))) 5075 return -EFAULT; 5076 return 0; 5077 case BTRFS_SUBVOL_SYNC_PEEK_LAST: 5078 spin_lock(&fs_info->trans_lock); 5079 /* First in the list was deleted last. */ 5080 if (!list_empty(&fs_info->dead_roots)) { 5081 root = list_first_entry(&fs_info->dead_roots, 5082 struct btrfs_root, root_list); 5083 args.subvolid = btrfs_root_id(root); 5084 } else { 5085 args.subvolid = 0; 5086 } 5087 spin_unlock(&fs_info->trans_lock); 5088 if (copy_to_user(argp, &args, sizeof(args))) 5089 return -EFAULT; 5090 return 0; 5091 default: 5092 return -EINVAL; 5093 } 5094 5095 /* 32bit limitation: fs_roots_radix key is not wide enough. */ 5096 if (sizeof(unsigned long) != sizeof(u64) && args.subvolid > U32_MAX) 5097 return -EOVERFLOW; 5098 5099 while (1) { 5100 /* Wait for the specific one. */ 5101 if (down_read_interruptible(&fs_info->subvol_sem) == -EINTR) 5102 return -EINTR; 5103 refs = -1; 5104 spin_lock(&fs_info->fs_roots_radix_lock); 5105 root = radix_tree_lookup(&fs_info->fs_roots_radix, 5106 (unsigned long)args.subvolid); 5107 if (root) { 5108 spin_lock(&root->root_item_lock); 5109 refs = btrfs_root_refs(&root->root_item); 5110 root_flags = btrfs_root_flags(&root->root_item); 5111 spin_unlock(&root->root_item_lock); 5112 } 5113 spin_unlock(&fs_info->fs_roots_radix_lock); 5114 up_read(&fs_info->subvol_sem); 5115 5116 /* Subvolume does not exist. */ 5117 if (!root) 5118 return -ENOENT; 5119 5120 /* Subvolume not deleted at all. */ 5121 if (refs > 0) 5122 return -EEXIST; 5123 /* We've waited and now the subvolume is gone. */ 5124 if (wait_for_deletion && refs == -1) { 5125 /* Return the one we waited for as the last one. */ 5126 if (copy_to_user(argp, &args, sizeof(args))) 5127 return -EFAULT; 5128 return 0; 5129 } 5130 5131 /* Subvolume not found on the first try (deleted or never existed). */ 5132 if (refs == -1) 5133 return -ENOENT; 5134 5135 wait_for_deletion = true; 5136 ASSERT(root_flags & BTRFS_ROOT_SUBVOL_DEAD, "root_flags=0x%llx", 5137 root_flags); 5138 sched_ret = schedule_timeout_interruptible(HZ); 5139 /* Early wake up or error. */ 5140 if (sched_ret != 0) 5141 return -EINTR; 5142 } 5143 5144 return 0; 5145 } 5146 5147 static int btrfs_ioctl_shutdown(struct btrfs_fs_info *fs_info, unsigned long arg) 5148 { 5149 int ret = 0; 5150 u32 flags; 5151 5152 if (!capable(CAP_SYS_ADMIN)) 5153 return -EPERM; 5154 5155 if (get_user(flags, (u32 __user *)arg)) 5156 return -EFAULT; 5157 5158 if (flags >= BTRFS_SHUTDOWN_FLAGS_LAST) 5159 return -EINVAL; 5160 5161 if (btrfs_is_shutdown(fs_info)) 5162 return 0; 5163 5164 switch (flags) { 5165 case BTRFS_SHUTDOWN_FLAGS_LOGFLUSH: 5166 case BTRFS_SHUTDOWN_FLAGS_DEFAULT: 5167 ret = freeze_super(fs_info->sb, FREEZE_HOLDER_KERNEL, NULL); 5168 if (ret) 5169 return ret; 5170 btrfs_force_shutdown(fs_info); 5171 ret = thaw_super(fs_info->sb, FREEZE_HOLDER_KERNEL, NULL); 5172 if (ret) 5173 return ret; 5174 break; 5175 case BTRFS_SHUTDOWN_FLAGS_NOLOGFLUSH: 5176 btrfs_force_shutdown(fs_info); 5177 break; 5178 default: 5179 ret = -EINVAL; 5180 break; 5181 } 5182 return ret; 5183 } 5184 5185 #define GET_CSUMS_BUF_MAX SZ_16M 5186 5187 static int copy_csums_to_user(struct btrfs_fs_info *fs_info, u64 disk_bytenr, 5188 u64 len, u8 __user *buf) 5189 { 5190 struct btrfs_root *csum_root; 5191 struct btrfs_ordered_sum *sums; 5192 LIST_HEAD(list); 5193 const u32 csum_size = fs_info->csum_size; 5194 int ret; 5195 5196 csum_root = btrfs_csum_root(fs_info, disk_bytenr); 5197 if (unlikely(!csum_root)) { 5198 btrfs_err(fs_info, "missing csum root for extent at bytenr %llu", disk_bytenr); 5199 return -EUCLEAN; 5200 } 5201 5202 ret = btrfs_lookup_csums_list(csum_root, disk_bytenr, 5203 disk_bytenr + len - 1, &list, false); 5204 if (ret < 0) 5205 return ret; 5206 5207 ret = 0; 5208 while (!list_empty(&list)) { 5209 u64 offset; 5210 size_t copy_size; 5211 5212 sums = list_first_entry(&list, struct btrfs_ordered_sum, list); 5213 list_del(&sums->list); 5214 5215 offset = ((sums->logical - disk_bytenr) >> fs_info->sectorsize_bits) * csum_size; 5216 copy_size = (sums->len >> fs_info->sectorsize_bits) * csum_size; 5217 5218 if (copy_to_user(buf + offset, sums->sums, copy_size)) { 5219 kfree(sums); 5220 ret = -EFAULT; 5221 goto out; 5222 } 5223 5224 kfree(sums); 5225 } 5226 5227 out: 5228 while (!list_empty(&list)) { 5229 sums = list_first_entry(&list, struct btrfs_ordered_sum, list); 5230 list_del(&sums->list); 5231 kfree(sums); 5232 } 5233 return ret; 5234 } 5235 5236 static int btrfs_ioctl_get_csums(struct file *file, void __user *argp) 5237 { 5238 struct inode *vfs_inode = file_inode(file); 5239 struct btrfs_inode *inode = BTRFS_I(vfs_inode); 5240 struct btrfs_fs_info *fs_info = inode->root->fs_info; 5241 struct btrfs_root *root = inode->root; 5242 struct btrfs_ioctl_get_csums_args args = { 0 }; 5243 BTRFS_PATH_AUTO_FREE(path); 5244 const u64 ino = btrfs_ino(inode); 5245 const u32 csum_size = fs_info->csum_size; 5246 u8 __user *ubuf; 5247 u64 buf_limit; 5248 u64 buf_used = 0; 5249 u64 cur_offset; 5250 u64 end_offset; 5251 u64 prev_extent_end; 5252 struct btrfs_key key; 5253 int ret; 5254 5255 if (!(file->f_mode & FMODE_READ)) 5256 return -EBADF; 5257 5258 if (!S_ISREG(vfs_inode->i_mode)) 5259 return -EINVAL; 5260 5261 if (copy_from_user(&args, argp, sizeof(args))) 5262 return -EFAULT; 5263 5264 if (!IS_ALIGNED(args.offset, fs_info->sectorsize) || 5265 !IS_ALIGNED(args.length, fs_info->sectorsize)) 5266 return -EINVAL; 5267 if (args.length == 0) 5268 return -EINVAL; 5269 if (args.offset + args.length < args.offset) 5270 return -EOVERFLOW; 5271 if (args.flags != 0) 5272 return -EINVAL; 5273 if (args.buf_size < sizeof(struct btrfs_ioctl_get_csums_entry)) 5274 return -EINVAL; 5275 5276 buf_limit = min_t(u64, args.buf_size, GET_CSUMS_BUF_MAX); 5277 ubuf = (u8 __user *)(argp + offsetof(struct btrfs_ioctl_get_csums_args, buf)); 5278 5279 if (clear_user(ubuf, buf_limit)) 5280 return -EFAULT; 5281 5282 cur_offset = args.offset; 5283 end_offset = args.offset + args.length; 5284 5285 path = btrfs_alloc_path(); 5286 if (!path) 5287 return -ENOMEM; 5288 5289 ret = btrfs_wait_ordered_range(inode, cur_offset, args.length); 5290 if (ret) 5291 return ret; 5292 5293 ret = down_read_interruptible(&vfs_inode->i_rwsem); 5294 if (ret) 5295 return ret; 5296 5297 ret = btrfs_wait_ordered_range(inode, cur_offset, args.length); 5298 if (ret) 5299 goto out_unlock; 5300 5301 /* NODATASUM early exit. */ 5302 if (inode->flags & BTRFS_INODE_NODATASUM) { 5303 struct btrfs_ioctl_get_csums_entry entry = { 5304 .offset = cur_offset, 5305 .length = end_offset - cur_offset, 5306 .type = BTRFS_GET_CSUMS_NODATASUM, 5307 }; 5308 5309 if (copy_to_user(ubuf, &entry, sizeof(entry))) { 5310 ret = -EFAULT; 5311 goto out_unlock; 5312 } 5313 5314 buf_used = sizeof(entry); 5315 cur_offset = end_offset; 5316 goto done; 5317 } 5318 5319 prev_extent_end = cur_offset; 5320 5321 while (cur_offset < end_offset) { 5322 struct btrfs_file_extent_item *ei; 5323 struct extent_buffer *leaf; 5324 struct btrfs_ioctl_get_csums_entry entry = { 0 }; 5325 u64 extent_end; 5326 u64 disk_bytenr = 0; 5327 u64 extent_offset = 0; 5328 u64 range_start, range_len; 5329 u64 entry_csum_size; 5330 u64 key_offset; 5331 int extent_type; 5332 u8 compression; 5333 u8 encryption; 5334 5335 /* Search for the extent at or before cur_offset. */ 5336 key.objectid = ino; 5337 key.type = BTRFS_EXTENT_DATA_KEY; 5338 key.offset = cur_offset; 5339 5340 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); 5341 if (ret < 0) 5342 goto out_unlock; 5343 5344 if (ret > 0 && path->slots[0] > 0) { 5345 btrfs_item_key_to_cpu(path->nodes[0], &key, 5346 path->slots[0] - 1); 5347 if (key.objectid == ino && key.type == BTRFS_EXTENT_DATA_KEY) { 5348 path->slots[0]--; 5349 if (btrfs_file_extent_end(path) <= cur_offset) 5350 path->slots[0]++; 5351 } 5352 } 5353 5354 if (path->slots[0] >= btrfs_header_nritems(path->nodes[0])) { 5355 ret = btrfs_next_leaf(root, path); 5356 if (ret < 0) 5357 goto out_unlock; 5358 if (ret > 0) { 5359 ret = 0; 5360 btrfs_release_path(path); 5361 break; 5362 } 5363 } 5364 5365 leaf = path->nodes[0]; 5366 5367 btrfs_item_key_to_cpu(leaf, &key, path->slots[0]); 5368 if (key.objectid != ino || key.type != BTRFS_EXTENT_DATA_KEY) { 5369 btrfs_release_path(path); 5370 break; 5371 } 5372 5373 extent_end = btrfs_file_extent_end(path); 5374 key_offset = key.offset; 5375 5376 /* Read extent fields before releasing the path. */ 5377 ei = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_file_extent_item); 5378 extent_type = btrfs_file_extent_type(leaf, ei); 5379 compression = btrfs_file_extent_compression(leaf, ei); 5380 encryption = btrfs_file_extent_encryption(leaf, ei); 5381 5382 if (extent_type != BTRFS_FILE_EXTENT_INLINE) { 5383 disk_bytenr = btrfs_file_extent_disk_bytenr(leaf, ei); 5384 if (disk_bytenr && compression == BTRFS_COMPRESS_NONE) 5385 extent_offset = btrfs_file_extent_offset(leaf, ei); 5386 } 5387 5388 btrfs_release_path(path); 5389 5390 /* Implicit hole (NO_HOLES feature). */ 5391 if (prev_extent_end < key_offset) { 5392 u64 hole_end = min(key_offset, end_offset); 5393 u64 hole_len = hole_end - prev_extent_end; 5394 5395 if (prev_extent_end >= cur_offset) { 5396 entry.offset = prev_extent_end; 5397 entry.length = hole_len; 5398 entry.type = BTRFS_GET_CSUMS_ZEROED; 5399 5400 if (buf_used + sizeof(entry) > buf_limit) 5401 goto done; 5402 if (copy_to_user(ubuf + buf_used, &entry, sizeof(entry))) { 5403 ret = -EFAULT; 5404 goto out_unlock; 5405 } 5406 buf_used += sizeof(entry); 5407 cur_offset = hole_end; 5408 } 5409 5410 if (key_offset >= end_offset) { 5411 cur_offset = end_offset; 5412 break; 5413 } 5414 } 5415 5416 /* Clamp to our query range. */ 5417 range_start = max(cur_offset, key_offset); 5418 range_len = min(extent_end, end_offset) - range_start; 5419 5420 entry.offset = range_start; 5421 entry.length = range_len; 5422 5423 if (extent_type == BTRFS_FILE_EXTENT_INLINE) { 5424 entry.type = BTRFS_GET_CSUMS_INLINE; 5425 if (compression != BTRFS_COMPRESS_NONE) 5426 entry.type |= BTRFS_GET_CSUMS_COMPRESSED; 5427 if (encryption != 0) 5428 entry.type |= BTRFS_GET_CSUMS_ENCRYPTED; 5429 entry_csum_size = 0; 5430 } else if (extent_type == BTRFS_FILE_EXTENT_PREALLOC) { 5431 entry.type = BTRFS_GET_CSUMS_ZEROED; 5432 entry_csum_size = 0; 5433 } else { 5434 /* BTRFS_FILE_EXTENT_REG */ 5435 if (disk_bytenr == 0) { 5436 /* Explicit hole. */ 5437 entry.type = BTRFS_GET_CSUMS_ZEROED; 5438 entry_csum_size = 0; 5439 } else if (encryption != 0 || compression != BTRFS_COMPRESS_NONE) { 5440 entry.type = 0; 5441 if (encryption != 0) 5442 entry.type |= BTRFS_GET_CSUMS_ENCRYPTED; 5443 if (compression != BTRFS_COMPRESS_NONE) 5444 entry.type |= BTRFS_GET_CSUMS_COMPRESSED; 5445 entry_csum_size = 0; 5446 } else { 5447 entry.type = BTRFS_GET_CSUMS_HAS_CSUMS; 5448 entry_csum_size = (range_len >> fs_info->sectorsize_bits) * csum_size; 5449 } 5450 } 5451 5452 /* Check if this entry (+ csum data) fits in the buffer. */ 5453 if (buf_used + sizeof(entry) + entry_csum_size > buf_limit) { 5454 if (buf_used == 0) { 5455 ret = -EOVERFLOW; 5456 goto out_unlock; 5457 } 5458 goto done; 5459 } 5460 5461 if (copy_to_user(ubuf + buf_used, &entry, sizeof(entry))) { 5462 ret = -EFAULT; 5463 goto out_unlock; 5464 } 5465 buf_used += sizeof(entry); 5466 5467 if (entry.type == BTRFS_GET_CSUMS_HAS_CSUMS) { 5468 ret = copy_csums_to_user(fs_info, 5469 disk_bytenr + extent_offset + (range_start - key_offset), 5470 range_len, ubuf + buf_used); 5471 if (ret) 5472 goto out_unlock; 5473 buf_used += entry_csum_size; 5474 } 5475 5476 cur_offset = range_start + range_len; 5477 prev_extent_end = extent_end; 5478 5479 if (fatal_signal_pending(current)) { 5480 if (buf_used == 0) { 5481 ret = -EINTR; 5482 goto out_unlock; 5483 } 5484 goto done; 5485 } 5486 5487 cond_resched(); 5488 } 5489 5490 /* Handle trailing implicit hole. */ 5491 if (cur_offset < end_offset) { 5492 struct btrfs_ioctl_get_csums_entry entry = { 5493 .offset = prev_extent_end, 5494 .length = end_offset - prev_extent_end, 5495 .type = BTRFS_GET_CSUMS_ZEROED, 5496 }; 5497 5498 if (buf_used + sizeof(entry) <= buf_limit) { 5499 if (copy_to_user(ubuf + buf_used, &entry, sizeof(entry))) { 5500 ret = -EFAULT; 5501 goto out_unlock; 5502 } 5503 buf_used += sizeof(entry); 5504 cur_offset = end_offset; 5505 } 5506 } 5507 5508 done: 5509 args.offset = cur_offset; 5510 args.length = (cur_offset < end_offset) ? end_offset - cur_offset : 0; 5511 args.buf_size = buf_used; 5512 5513 if (copy_to_user(argp, &args, sizeof(args))) 5514 ret = -EFAULT; 5515 5516 out_unlock: 5517 up_read(&vfs_inode->i_rwsem); 5518 return ret; 5519 } 5520 5521 long btrfs_ioctl(struct file *file, unsigned int 5522 cmd, unsigned long arg) 5523 { 5524 struct inode *inode = file_inode(file); 5525 struct btrfs_fs_info *fs_info = inode_to_fs_info(inode); 5526 struct btrfs_root *root = BTRFS_I(inode)->root; 5527 void __user *argp = (void __user *)arg; 5528 5529 switch (cmd) { 5530 case FS_IOC_GETVERSION: 5531 return btrfs_ioctl_getversion(inode, argp); 5532 case FS_IOC_GETFSLABEL: 5533 return btrfs_ioctl_get_fslabel(fs_info, argp); 5534 case FS_IOC_SETFSLABEL: 5535 return btrfs_ioctl_set_fslabel(file, argp); 5536 case FITRIM: 5537 return btrfs_ioctl_fitrim(fs_info, argp); 5538 case BTRFS_IOC_SNAP_CREATE: 5539 return btrfs_ioctl_snap_create(file, argp, false); 5540 case BTRFS_IOC_SNAP_CREATE_V2: 5541 return btrfs_ioctl_snap_create_v2(file, argp, false); 5542 case BTRFS_IOC_SUBVOL_CREATE: 5543 return btrfs_ioctl_snap_create(file, argp, true); 5544 case BTRFS_IOC_SUBVOL_CREATE_V2: 5545 return btrfs_ioctl_snap_create_v2(file, argp, true); 5546 case BTRFS_IOC_SNAP_DESTROY: 5547 return btrfs_ioctl_snap_destroy(file, argp, false); 5548 case BTRFS_IOC_SNAP_DESTROY_V2: 5549 return btrfs_ioctl_snap_destroy(file, argp, true); 5550 case BTRFS_IOC_SUBVOL_GETFLAGS: 5551 return btrfs_ioctl_subvol_getflags(BTRFS_I(inode), argp); 5552 case BTRFS_IOC_SUBVOL_SETFLAGS: 5553 return btrfs_ioctl_subvol_setflags(file, argp); 5554 case BTRFS_IOC_DEFAULT_SUBVOL: 5555 return btrfs_ioctl_default_subvol(file, argp); 5556 case BTRFS_IOC_DEFRAG: 5557 return btrfs_ioctl_defrag(file, NULL); 5558 case BTRFS_IOC_DEFRAG_RANGE: 5559 return btrfs_ioctl_defrag(file, argp); 5560 case BTRFS_IOC_RESIZE: 5561 return btrfs_ioctl_resize(file, argp); 5562 case BTRFS_IOC_ADD_DEV: 5563 return btrfs_ioctl_add_dev(fs_info, argp); 5564 case BTRFS_IOC_RM_DEV: 5565 return btrfs_ioctl_rm_dev(file, argp); 5566 case BTRFS_IOC_RM_DEV_V2: 5567 return btrfs_ioctl_rm_dev_v2(file, argp); 5568 case BTRFS_IOC_FS_INFO: 5569 return btrfs_ioctl_fs_info(fs_info, argp); 5570 case BTRFS_IOC_DEV_INFO: 5571 return btrfs_ioctl_dev_info(fs_info, argp); 5572 case BTRFS_IOC_TREE_SEARCH: 5573 return btrfs_ioctl_tree_search(root, argp); 5574 case BTRFS_IOC_TREE_SEARCH_V2: 5575 return btrfs_ioctl_tree_search_v2(root, argp); 5576 case BTRFS_IOC_INO_LOOKUP: 5577 return btrfs_ioctl_ino_lookup(root, argp); 5578 case BTRFS_IOC_INO_PATHS: 5579 return btrfs_ioctl_ino_to_path(root, argp); 5580 case BTRFS_IOC_LOGICAL_INO: 5581 return btrfs_ioctl_logical_to_ino(fs_info, argp, 1); 5582 case BTRFS_IOC_LOGICAL_INO_V2: 5583 return btrfs_ioctl_logical_to_ino(fs_info, argp, 2); 5584 case BTRFS_IOC_SPACE_INFO: 5585 return btrfs_ioctl_space_info(fs_info, argp); 5586 case BTRFS_IOC_SYNC: { 5587 int ret; 5588 5589 ret = btrfs_start_delalloc_roots(fs_info, LONG_MAX, false); 5590 if (ret) 5591 return ret; 5592 ret = btrfs_sync_fs(inode->i_sb, 1); 5593 /* 5594 * There may be work for the cleaner kthread to do (subvolume 5595 * deletion, delayed iputs, defrag inodes, etc), so wake it up. 5596 */ 5597 wake_up_process(fs_info->cleaner_kthread); 5598 return ret; 5599 } 5600 case BTRFS_IOC_START_SYNC: 5601 return btrfs_ioctl_start_sync(root, argp); 5602 case BTRFS_IOC_WAIT_SYNC: 5603 return btrfs_ioctl_wait_sync(fs_info, argp); 5604 case BTRFS_IOC_SCRUB: 5605 return btrfs_ioctl_scrub(file, argp); 5606 case BTRFS_IOC_SCRUB_CANCEL: 5607 return btrfs_ioctl_scrub_cancel(fs_info); 5608 case BTRFS_IOC_SCRUB_PROGRESS: 5609 return btrfs_ioctl_scrub_progress(fs_info, argp); 5610 case BTRFS_IOC_BALANCE_V2: 5611 return btrfs_ioctl_balance(file, argp); 5612 case BTRFS_IOC_BALANCE_CTL: 5613 return btrfs_ioctl_balance_ctl(fs_info, arg); 5614 case BTRFS_IOC_BALANCE_PROGRESS: 5615 return btrfs_ioctl_balance_progress(fs_info, argp); 5616 case BTRFS_IOC_SET_RECEIVED_SUBVOL: 5617 return btrfs_ioctl_set_received_subvol(file, argp); 5618 #ifdef CONFIG_64BIT 5619 case BTRFS_IOC_SET_RECEIVED_SUBVOL_32: 5620 return btrfs_ioctl_set_received_subvol_32(file, argp); 5621 #endif 5622 case BTRFS_IOC_SEND: 5623 return _btrfs_ioctl_send(root, argp, false); 5624 #if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT) 5625 case BTRFS_IOC_SEND_32: 5626 return _btrfs_ioctl_send(root, argp, true); 5627 #endif 5628 case BTRFS_IOC_GET_DEV_STATS: 5629 return btrfs_ioctl_get_dev_stats(fs_info, argp); 5630 case BTRFS_IOC_QUOTA_CTL: 5631 return btrfs_ioctl_quota_ctl(file, argp); 5632 case BTRFS_IOC_QGROUP_ASSIGN: 5633 return btrfs_ioctl_qgroup_assign(file, argp); 5634 case BTRFS_IOC_QGROUP_CREATE: 5635 return btrfs_ioctl_qgroup_create(file, argp); 5636 case BTRFS_IOC_QGROUP_LIMIT: 5637 return btrfs_ioctl_qgroup_limit(file, argp); 5638 case BTRFS_IOC_QUOTA_RESCAN: 5639 return btrfs_ioctl_quota_rescan(file, argp); 5640 case BTRFS_IOC_QUOTA_RESCAN_STATUS: 5641 return btrfs_ioctl_quota_rescan_status(fs_info, argp); 5642 case BTRFS_IOC_QUOTA_RESCAN_WAIT: 5643 return btrfs_ioctl_quota_rescan_wait(fs_info); 5644 case BTRFS_IOC_DEV_REPLACE: 5645 return btrfs_ioctl_dev_replace(fs_info, argp); 5646 case BTRFS_IOC_GET_SUPPORTED_FEATURES: 5647 return btrfs_ioctl_get_supported_features(argp); 5648 case BTRFS_IOC_GET_FEATURES: 5649 return btrfs_ioctl_get_features(fs_info, argp); 5650 case BTRFS_IOC_SET_FEATURES: 5651 return btrfs_ioctl_set_features(file, argp); 5652 case BTRFS_IOC_GET_SUBVOL_INFO: 5653 return btrfs_ioctl_get_subvol_info(inode, argp); 5654 #ifdef CONFIG_64BIT 5655 case BTRFS_IOC_GET_SUBVOL_INFO_32: 5656 return btrfs_ioctl_get_subvol_info_32(inode, argp); 5657 #endif 5658 case BTRFS_IOC_GET_SUBVOL_ROOTREF: 5659 return btrfs_ioctl_get_subvol_rootref(root, argp); 5660 case BTRFS_IOC_INO_LOOKUP_USER: 5661 return btrfs_ioctl_ino_lookup_user(file, argp); 5662 case FS_IOC_ENABLE_VERITY: 5663 return fsverity_ioctl_enable(file, (const void __user *)argp); 5664 case FS_IOC_MEASURE_VERITY: 5665 return fsverity_ioctl_measure(file, argp); 5666 case FS_IOC_READ_VERITY_METADATA: 5667 return fsverity_ioctl_read_metadata(file, argp); 5668 case BTRFS_IOC_ENCODED_READ: 5669 return btrfs_ioctl_encoded_read(file, argp, false); 5670 case BTRFS_IOC_ENCODED_WRITE: 5671 return btrfs_ioctl_encoded_write(file, argp, false); 5672 #if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT) 5673 case BTRFS_IOC_ENCODED_READ_32: 5674 return btrfs_ioctl_encoded_read(file, argp, true); 5675 case BTRFS_IOC_ENCODED_WRITE_32: 5676 return btrfs_ioctl_encoded_write(file, argp, true); 5677 #endif 5678 case BTRFS_IOC_SUBVOL_SYNC_WAIT: 5679 return btrfs_ioctl_subvol_sync(fs_info, argp); 5680 case BTRFS_IOC_SHUTDOWN: 5681 return btrfs_ioctl_shutdown(fs_info, arg); 5682 case BTRFS_IOC_GET_CSUMS: 5683 return btrfs_ioctl_get_csums(file, argp); 5684 } 5685 5686 return -ENOTTY; 5687 } 5688 5689 #ifdef CONFIG_COMPAT 5690 long btrfs_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg) 5691 { 5692 /* 5693 * These all access 32-bit values anyway so no further 5694 * handling is necessary. 5695 */ 5696 switch (cmd) { 5697 case FS_IOC32_GETVERSION: 5698 cmd = FS_IOC_GETVERSION; 5699 break; 5700 } 5701 5702 return btrfs_ioctl(file, cmd, (unsigned long) compat_ptr(arg)); 5703 } 5704 #endif 5705