1 // SPDX-License-Identifier: GPL-2.0+
3 * BTRFS filesystem implementation for U-Boot
5 * 2017 Marek Behun, CZ.NIC, marek.behun@nic.cz
8 #include <linux/kernel.h>
16 * Read the content of symlink inode @ino of @root, into @target.
17 * NOTE: @target will not be \0 termiated, caller should handle it properly.
19 * Return the number of read data.
20 * Return <0 for error.
22 int btrfs_readlink(struct btrfs_root *root, u64 ino, char *target)
24 struct btrfs_path path;
26 struct btrfs_file_extent_item *fi;
30 key.type = BTRFS_EXTENT_DATA_KEY;
32 btrfs_init_path(&path);
34 ret = btrfs_search_slot(NULL, root, &key, &path, 0, 0);
41 fi = btrfs_item_ptr(path.nodes[0], path.slots[0],
42 struct btrfs_file_extent_item);
43 if (btrfs_file_extent_type(path.nodes[0], fi) !=
44 BTRFS_FILE_EXTENT_INLINE) {
46 error("Extent for symlink %llu must be INLINE type!", ino);
49 if (btrfs_file_extent_compression(path.nodes[0], fi) !=
50 BTRFS_COMPRESS_NONE) {
52 error("Extent for symlink %llu must not be compressed!", ino);
55 if (btrfs_file_extent_ram_bytes(path.nodes[0], fi) >=
56 root->fs_info->sectorsize) {
58 error("Symlink %llu extent data too large (%llu)!\n",
59 ino, btrfs_file_extent_ram_bytes(path.nodes[0], fi));
62 read_extent_buffer(path.nodes[0], target,
63 btrfs_file_extent_inline_start(fi),
64 btrfs_file_extent_ram_bytes(path.nodes[0], fi));
65 ret = btrfs_file_extent_ram_bytes(path.nodes[0], fi);
67 btrfs_release_path(&path);
71 static int lookup_root_ref(struct btrfs_fs_info *fs_info,
72 u64 rootid, u64 *root_ret, u64 *dir_ret)
74 struct btrfs_root *root = fs_info->tree_root;
75 struct btrfs_root_ref *root_ref;
76 struct btrfs_path path;
80 btrfs_init_path(&path);
81 key.objectid = rootid;
82 key.type = BTRFS_ROOT_BACKREF_KEY;
85 ret = btrfs_search_slot(NULL, root, &key, &path, 0, 0);
88 /* Should not happen */
93 ret = btrfs_previous_item(root, &path, rootid, BTRFS_ROOT_BACKREF_KEY);
100 btrfs_item_key_to_cpu(path.nodes[0], &key, path.slots[0]);
101 root_ref = btrfs_item_ptr(path.nodes[0], path.slots[0],
102 struct btrfs_root_ref);
103 *root_ret = key.offset;
104 *dir_ret = btrfs_root_ref_dirid(path.nodes[0], root_ref);
106 btrfs_release_path(&path);
111 * To get the parent inode of @ino of @root.
113 * @root_ret and @ino_ret will be filled.
115 * NOTE: This function is not reliable. It can only get one parent inode.
116 * The get the proper parent inode, we need a full VFS inodes stack to
119 static int get_parent_inode(struct btrfs_root *root, u64 ino,
120 struct btrfs_root **root_ret, u64 *ino_ret)
122 struct btrfs_fs_info *fs_info = root->fs_info;
123 struct btrfs_path path;
124 struct btrfs_key key;
127 if (ino == BTRFS_FIRST_FREE_OBJECTID) {
128 u64 parent_root = -1;
130 /* It's top level already, no more parent */
131 if (root->root_key.objectid == BTRFS_FS_TREE_OBJECTID) {
132 *root_ret = fs_info->fs_root;
133 *ino_ret = BTRFS_FIRST_FREE_OBJECTID;
137 ret = lookup_root_ref(fs_info, root->root_key.objectid,
138 &parent_root, ino_ret);
142 key.objectid = parent_root;
143 key.type = BTRFS_ROOT_ITEM_KEY;
144 key.offset = (u64)-1;
145 *root_ret = btrfs_read_fs_root(fs_info, &key);
146 if (IS_ERR(*root_ret))
147 return PTR_ERR(*root_ret);
152 btrfs_init_path(&path);
154 key.type = BTRFS_INODE_REF_KEY;
155 key.offset = (u64)-1;
157 ret = btrfs_search_slot(NULL, root, &key, &path, 0, 0);
160 /* Should not happen */
165 ret = btrfs_previous_item(root, &path, ino, BTRFS_INODE_REF_KEY);
172 btrfs_item_key_to_cpu(path.nodes[0], &key, path.slots[0]);
174 *ino_ret = key.offset;
176 btrfs_release_path(&path);
180 static inline int next_length(const char *path)
183 while (*path != '\0' && *path != '/') {
186 if (res > BTRFS_NAME_LEN)
192 static inline const char *skip_current_directories(const char *cur)
197 else if (cur[0] == '.' && cur[1] == '/')
207 * Resolve one filename of @ino of @root.
209 * key_ret: The child key (either INODE_ITEM or ROOT_ITEM type)
210 * type_ret: BTRFS_FT_* of the child inode.
212 * Return 0 with above members filled.
213 * Return <0 for error.
215 static int resolve_one_filename(struct btrfs_root *root, u64 ino,
216 const char *name, int namelen,
217 struct btrfs_key *key_ret, u8 *type_ret)
219 struct btrfs_dir_item *dir_item;
220 struct btrfs_path path;
223 btrfs_init_path(&path);
225 dir_item = btrfs_lookup_dir_item(NULL, root, &path, ino, name,
227 if (IS_ERR(dir_item)) {
228 ret = PTR_ERR(dir_item);
232 btrfs_dir_item_key_to_cpu(path.nodes[0], dir_item, key_ret);
233 *type_ret = btrfs_dir_type(path.nodes[0], dir_item);
235 btrfs_release_path(&path);
240 * Resolve a full path @filename. The start point is @ino of @root.
242 * The result will be filled into @root_ret, @ino_ret and @type_ret.
244 int btrfs_lookup_path(struct btrfs_root *root, u64 ino, const char *filename,
245 struct btrfs_root **root_ret, u64 *ino_ret,
246 u8 *type_ret, int symlink_limit)
248 struct btrfs_fs_info *fs_info = root->fs_info;
249 struct btrfs_root *next_root;
250 struct btrfs_key key;
251 const char *cur = filename;
254 u8 type = BTRFS_FT_UNKNOWN;
258 /* If the path is absolute path, also search from fs root */
260 root = fs_info->fs_root;
261 ino = btrfs_root_dirid(&root->root_item);
265 while (*cur != '\0') {
266 cur = skip_current_directories(cur);
268 len = next_length(cur);
269 if (len > BTRFS_NAME_LEN) {
270 error("%s: Name too long at \"%.*s\"", __func__,
271 BTRFS_NAME_LEN, cur);
272 return -ENAMETOOLONG;
275 if (len == 1 && cur[0] == '.')
278 if (len == 2 && cur[0] == '.' && cur[1] == '.') {
279 /* Go one level up */
280 ret = get_parent_inode(root, ino, &next_root, &next_ino);
291 ret = resolve_one_filename(root, ino, cur, len, &key, &type);
295 if (key.type == BTRFS_ROOT_ITEM_KEY) {
296 /* Child inode is a subvolume */
298 next_root = btrfs_read_fs_root(fs_info, &key);
299 if (IS_ERR(next_root))
300 return PTR_ERR(next_root);
302 ino = btrfs_root_dirid(&root->root_item);
303 } else if (type == BTRFS_FT_SYMLINK && symlink_limit >= 0) {
304 /* Child inode is a symlink */
308 if (symlink_limit == 0) {
309 error("%s: Too much symlinks!", __func__);
312 target = malloc(fs_info->sectorsize);
315 ret = btrfs_readlink(root, key.objectid, target);
322 ret = btrfs_lookup_path(root, ino, target, &next_root,
323 &next_ino, &next_type,
331 /* Child inode is an inode */
338 /* We haven't found anything, but still get no error? */
339 if (type == BTRFS_FT_UNKNOWN && !ret)
352 * Read out inline extent.
354 * Since inline extent should only exist for offset 0, no need for extra
356 * Truncating should be handled by the caller.
358 * Return the number of bytes read.
359 * Return <0 for error.
361 int btrfs_read_extent_inline(struct btrfs_path *path,
362 struct btrfs_file_extent_item *fi, char *dest)
364 struct extent_buffer *leaf = path->nodes[0];
365 int slot = path->slots[0];
372 csize = btrfs_file_extent_inline_item_len(leaf, btrfs_item_nr(slot));
373 if (btrfs_file_extent_compression(leaf, fi) == BTRFS_COMPRESS_NONE) {
374 /* Uncompressed, just read it out */
375 read_extent_buffer(leaf, dest,
376 btrfs_file_extent_inline_start(fi),
381 /* Compressed extent, prepare the compressed and data buffer */
382 dsize = btrfs_file_extent_ram_bytes(leaf, fi);
383 cbuf = malloc(csize);
384 dbuf = malloc(dsize);
385 if (!cbuf || !dbuf) {
389 read_extent_buffer(leaf, cbuf, btrfs_file_extent_inline_start(fi),
391 ret = btrfs_decompress(btrfs_file_extent_compression(leaf, fi),
392 cbuf, csize, dbuf, dsize);
393 if (ret < 0 || ret != dsize) {
397 memcpy(dest, dbuf, dsize);
406 * Read out regular extent.
408 * Truncating should be handled by the caller.
410 * @offset and @len should not cross the extent boundary.
411 * Return the number of bytes read.
412 * Return <0 for error.
414 int btrfs_read_extent_reg(struct btrfs_path *path,
415 struct btrfs_file_extent_item *fi, u64 offset,
418 struct extent_buffer *leaf = path->nodes[0];
419 struct btrfs_fs_info *fs_info = leaf->fs_info;
420 struct btrfs_key key;
421 u64 extent_num_bytes;
428 bool finished = false;
431 int slot = path->slots[0];
434 btrfs_item_key_to_cpu(leaf, &key, slot);
435 extent_num_bytes = btrfs_file_extent_num_bytes(leaf, fi);
436 ASSERT(IS_ALIGNED(offset, fs_info->sectorsize) &&
437 IS_ALIGNED(len, fs_info->sectorsize));
438 ASSERT(offset >= key.offset &&
439 offset + len <= key.offset + extent_num_bytes);
441 /* Preallocated or hole , fill @dest with zero */
442 if (btrfs_file_extent_type(leaf, fi) == BTRFS_FILE_EXTENT_PREALLOC ||
443 btrfs_file_extent_disk_bytenr(leaf, fi) == 0) {
444 memset(dest, 0, len);
448 if (btrfs_file_extent_compression(leaf, fi) == BTRFS_COMPRESS_NONE) {
451 logical = btrfs_file_extent_disk_bytenr(leaf, fi) +
452 btrfs_file_extent_offset(leaf, fi) +
456 num_copies = btrfs_num_copies(fs_info, logical, len);
457 for (i = 1; i <= num_copies; i++) {
458 ret = read_extent_data(fs_info, dest, logical, &read, i);
459 if (ret < 0 || read != len)
469 csize = btrfs_file_extent_disk_num_bytes(leaf, fi);
470 dsize = btrfs_file_extent_ram_bytes(leaf, fi);
471 disk_bytenr = btrfs_file_extent_disk_bytenr(leaf, fi);
472 num_copies = btrfs_num_copies(fs_info, disk_bytenr, csize);
474 cbuf = malloc_cache_aligned(csize);
475 dbuf = malloc_cache_aligned(dsize);
476 if (!cbuf || !dbuf) {
480 /* For compressed extent, we must read the whole on-disk extent */
481 for (i = 1; i <= num_copies; i++) {
483 ret = read_extent_data(fs_info, cbuf, disk_bytenr,
485 if (ret < 0 || read != csize)
495 ret = btrfs_decompress(btrfs_file_extent_compression(leaf, fi), cbuf,
501 /* Then copy the needed part */
502 memcpy(dest, dbuf + btrfs_file_extent_offset(leaf, fi), len);
511 * Get the first file extent that covers bytenr @file_offset.
513 * @file_offset must be aligned to sectorsize.
515 * return 0 for found, and path points to the file extent.
516 * return >0 for not found, and fill @next_offset.
517 * @next_offset can be 0 if there is no next file extent.
518 * return <0 for error.
520 static int lookup_data_extent(struct btrfs_root *root, struct btrfs_path *path,
521 u64 ino, u64 file_offset, u64 *next_offset)
523 struct btrfs_key key;
524 struct btrfs_file_extent_item *fi;
528 ASSERT(IS_ALIGNED(file_offset, root->fs_info->sectorsize));
530 key.type = BTRFS_EXTENT_DATA_KEY;
531 key.offset = file_offset;
533 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
534 /* Error or we're already at the file extent */
538 /* Check previous file extent */
539 ret = btrfs_previous_item(root, path, ino,
540 BTRFS_EXTENT_DATA_KEY);
546 /* Now the key.offset must be smaller than @file_offset */
547 btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
548 if (key.objectid != ino ||
549 key.type != BTRFS_EXTENT_DATA_KEY)
552 fi = btrfs_item_ptr(path->nodes[0], path->slots[0],
553 struct btrfs_file_extent_item);
554 extent_type = btrfs_file_extent_type(path->nodes[0], fi);
555 if (extent_type == BTRFS_FILE_EXTENT_INLINE) {
556 if (file_offset == 0)
558 /* Inline extent should be the only extent, no next extent. */
563 /* This file extent covers @file_offset */
564 if (key.offset <= file_offset && key.offset +
565 btrfs_file_extent_num_bytes(path->nodes[0], fi) > file_offset)
568 ret = btrfs_next_item(root, path);
576 btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
577 fi = btrfs_item_ptr(path->nodes[0], path->slots[0],
578 struct btrfs_file_extent_item);
579 /* Next next data extent */
580 if (key.objectid != ino ||
581 key.type != BTRFS_EXTENT_DATA_KEY) {
585 /* Current file extent already beyond @file_offset */
586 if (key.offset > file_offset) {
587 *next_offset = key.offset;
590 /* This file extent covers @file_offset */
591 if (key.offset <= file_offset && key.offset +
592 btrfs_file_extent_num_bytes(path->nodes[0], fi) > file_offset)
594 /* This file extent ends before @file_offset, check next */
595 ret = btrfs_next_item(root, path);
602 btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
603 if (key.type != BTRFS_EXTENT_DATA_KEY || key.objectid != ino) {
607 *next_offset = key.offset;
611 static int read_and_truncate_page(struct btrfs_path *path,
612 struct btrfs_file_extent_item *fi,
613 int start, int len, char *dest)
615 struct extent_buffer *leaf = path->nodes[0];
616 struct btrfs_fs_info *fs_info = leaf->fs_info;
617 u64 aligned_start = round_down(start, fs_info->sectorsize);
620 int page_off = start - aligned_start;
621 int page_len = fs_info->sectorsize - page_off;
624 ASSERT(start + len <= aligned_start + fs_info->sectorsize);
625 buf = malloc_cache_aligned(fs_info->sectorsize);
629 extent_type = btrfs_file_extent_type(leaf, fi);
630 if (extent_type == BTRFS_FILE_EXTENT_INLINE) {
631 ret = btrfs_read_extent_inline(path, fi, buf);
632 memcpy(dest, buf + page_off, min(page_len, ret));
637 ret = btrfs_read_extent_reg(path, fi,
638 round_down(start, fs_info->sectorsize),
639 fs_info->sectorsize, buf);
644 memcpy(dest, buf + page_off, page_len);
649 int btrfs_file_read(struct btrfs_root *root, u64 ino, u64 file_offset, u64 len,
652 struct btrfs_fs_info *fs_info = root->fs_info;
653 struct btrfs_file_extent_item *fi;
654 struct btrfs_path path;
655 struct btrfs_key key;
656 u64 aligned_start = round_down(file_offset, fs_info->sectorsize);
657 u64 aligned_end = round_down(file_offset + len, fs_info->sectorsize);
659 u64 cur = aligned_start;
662 btrfs_init_path(&path);
664 /* Set the whole dest all zero, so we won't need to bother holes */
665 memset(dest, 0, len);
667 /* Read out the leading unaligned part */
668 if (aligned_start != file_offset) {
669 ret = lookup_data_extent(root, &path, ino, aligned_start,
674 /* Read the unaligned part out*/
675 fi = btrfs_item_ptr(path.nodes[0], path.slots[0],
676 struct btrfs_file_extent_item);
677 ret = read_and_truncate_page(&path, fi, file_offset,
678 round_up(file_offset, fs_info->sectorsize) -
682 cur += fs_info->sectorsize;
684 /* The whole file is a hole */
686 memset(dest, 0, len);
693 /* Read the aligned part */
694 while (cur < aligned_end) {
695 u64 extent_num_bytes;
698 btrfs_release_path(&path);
699 ret = lookup_data_extent(root, &path, ino, cur, &next_offset);
703 /* No next, direct exit */
709 fi = btrfs_item_ptr(path.nodes[0], path.slots[0],
710 struct btrfs_file_extent_item);
711 btrfs_item_key_to_cpu(path.nodes[0], &key, path.slots[0]);
712 type = btrfs_file_extent_type(path.nodes[0], fi);
713 if (type == BTRFS_FILE_EXTENT_INLINE) {
714 ret = btrfs_read_extent_inline(&path, fi, dest);
717 /* Skip holes, as we have zeroed the dest */
718 if (type == BTRFS_FILE_EXTENT_PREALLOC ||
719 btrfs_file_extent_disk_bytenr(path.nodes[0], fi) == 0) {
720 cur = key.offset + btrfs_file_extent_num_bytes(
725 /* Read the remaining part of the extent */
726 extent_num_bytes = btrfs_file_extent_num_bytes(path.nodes[0],
728 ret = btrfs_read_extent_reg(&path, fi, cur,
729 min(extent_num_bytes, aligned_end - cur),
730 dest + cur - file_offset);
733 cur += min(extent_num_bytes, aligned_end - cur);
736 /* Read the tailing unaligned part*/
737 if (file_offset + len != aligned_end) {
738 btrfs_release_path(&path);
739 ret = lookup_data_extent(root, &path, ino, aligned_end,
741 /* <0 is error, >0 means no extent */
744 fi = btrfs_item_ptr(path.nodes[0], path.slots[0],
745 struct btrfs_file_extent_item);
746 ret = read_and_truncate_page(&path, fi, aligned_end,
747 file_offset + len - aligned_end,
748 dest + aligned_end - file_offset);
751 btrfs_release_path(&path);