blob: 019d532a1a4bc2f41288f02cfea4259f78d206f0 [file] [log] [blame]
Tom Rini10e47792018-05-06 17:58:06 -04001// SPDX-License-Identifier: GPL-2.0+
Marek Behún29387542017-09-03 17:00:28 +02002/*
3 * BTRFS filesystem implementation for U-Boot
4 *
5 * 2017 Marek Behun, CZ.NIC, marek.behun@nic.cz
Marek Behún29387542017-09-03 17:00:28 +02006 */
7
Qu Wenruo83b04df2020-06-24 18:03:09 +02008#include <linux/kernel.h>
Marek Behún29387542017-09-03 17:00:28 +02009#include <malloc.h>
Qu Wenruo83b04df2020-06-24 18:03:09 +020010#include <memalign.h>
Qu Wenruo49fe0f02020-06-24 18:03:03 +020011#include "btrfs.h"
12#include "disk-io.h"
Qu Wenruo83b04df2020-06-24 18:03:09 +020013#include "volumes.h"
Marek Behún29387542017-09-03 17:00:28 +020014
Qu Wenruo49fe0f02020-06-24 18:03:03 +020015/*
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.
18 *
19 * Return the number of read data.
20 * Return <0 for error.
21 */
22int btrfs_readlink(struct btrfs_root *root, u64 ino, char *target)
Marek Behún29387542017-09-03 17:00:28 +020023{
Qu Wenruo49fe0f02020-06-24 18:03:03 +020024 struct btrfs_path path;
Marek Behún29387542017-09-03 17:00:28 +020025 struct btrfs_key key;
Qu Wenruo49fe0f02020-06-24 18:03:03 +020026 struct btrfs_file_extent_item *fi;
27 int ret;
Marek Behún29387542017-09-03 17:00:28 +020028
Qu Wenruo49fe0f02020-06-24 18:03:03 +020029 key.objectid = ino;
Marek Behún29387542017-09-03 17:00:28 +020030 key.type = BTRFS_EXTENT_DATA_KEY;
31 key.offset = 0;
Qu Wenruo49fe0f02020-06-24 18:03:03 +020032 btrfs_init_path(&path);
Marek Behún29387542017-09-03 17:00:28 +020033
Qu Wenruo49fe0f02020-06-24 18:03:03 +020034 ret = btrfs_search_slot(NULL, root, &key, &path, 0, 0);
35 if (ret < 0)
36 return ret;
37 if (ret > 0) {
38 ret = -ENOENT;
Marek Behún29387542017-09-03 17:00:28 +020039 goto out;
40 }
Qu Wenruo49fe0f02020-06-24 18:03:03 +020041 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) {
45 ret = -EUCLEAN;
46 error("Extent for symlink %llu must be INLINE type!", ino);
Marek Behún29387542017-09-03 17:00:28 +020047 goto out;
Qu Wenruo49fe0f02020-06-24 18:03:03 +020048 }
49 if (btrfs_file_extent_compression(path.nodes[0], fi) !=
50 BTRFS_COMPRESS_NONE) {
51 ret = -EUCLEAN;
52 error("Extent for symlink %llu must not be compressed!", ino);
Marek Behún29387542017-09-03 17:00:28 +020053 goto out;
Qu Wenruo49fe0f02020-06-24 18:03:03 +020054 }
55 if (btrfs_file_extent_ram_bytes(path.nodes[0], fi) >=
56 root->fs_info->sectorsize) {
57 ret = -EUCLEAN;
58 error("Symlink %llu extent data too large (%llu)!\n",
59 ino, btrfs_file_extent_ram_bytes(path.nodes[0], fi));
Marek Behún29387542017-09-03 17:00:28 +020060 goto out;
61 }
Qu Wenruo49fe0f02020-06-24 18:03:03 +020062 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);
66out:
67 btrfs_release_path(&path);
68 return ret;
69}
Marek Behún29387542017-09-03 17:00:28 +020070
Qu Wenruoe565bd42020-06-24 18:03:05 +020071static int lookup_root_ref(struct btrfs_fs_info *fs_info,
72 u64 rootid, u64 *root_ret, u64 *dir_ret)
73{
74 struct btrfs_root *root = fs_info->tree_root;
75 struct btrfs_root_ref *root_ref;
76 struct btrfs_path path;
77 struct btrfs_key key;
78 int ret;
79
80 btrfs_init_path(&path);
81 key.objectid = rootid;
82 key.type = BTRFS_ROOT_BACKREF_KEY;
83 key.offset = (u64)-1;
84
85 ret = btrfs_search_slot(NULL, root, &key, &path, 0, 0);
86 if (ret < 0)
87 return ret;
88 /* Should not happen */
89 if (ret == 0) {
90 ret = -EUCLEAN;
91 goto out;
92 }
93 ret = btrfs_previous_item(root, &path, rootid, BTRFS_ROOT_BACKREF_KEY);
94 if (ret < 0)
95 goto out;
96 if (ret > 0) {
97 ret = -ENOENT;
98 goto out;
99 }
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);
105out:
106 btrfs_release_path(&path);
107 return ret;
108}
109
110/*
111 * To get the parent inode of @ino of @root.
112 *
113 * @root_ret and @ino_ret will be filled.
114 *
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
117 * resolve properly.
118 */
119static int get_parent_inode(struct btrfs_root *root, u64 ino,
120 struct btrfs_root **root_ret, u64 *ino_ret)
121{
122 struct btrfs_fs_info *fs_info = root->fs_info;
123 struct btrfs_path path;
124 struct btrfs_key key;
125 int ret;
126
127 if (ino == BTRFS_FIRST_FREE_OBJECTID) {
128 u64 parent_root = -1;
129
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;
134 return 0;
135 }
136
137 ret = lookup_root_ref(fs_info, root->root_key.objectid,
138 &parent_root, ino_ret);
139 if (ret < 0)
140 return ret;
141
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);
148
149 return 0;
150 }
151
152 btrfs_init_path(&path);
153 key.objectid = ino;
154 key.type = BTRFS_INODE_REF_KEY;
155 key.offset = (u64)-1;
156
157 ret = btrfs_search_slot(NULL, root, &key, &path, 0, 0);
158 if (ret < 0)
159 return ret;
160 /* Should not happen */
161 if (ret == 0) {
162 ret = -EUCLEAN;
163 goto out;
164 }
165 ret = btrfs_previous_item(root, &path, ino, BTRFS_INODE_REF_KEY);
166 if (ret < 0)
167 goto out;
168 if (ret > 0) {
169 ret = -ENOENT;
170 goto out;
171 }
172 btrfs_item_key_to_cpu(path.nodes[0], &key, path.slots[0]);
173 *root_ret = root;
174 *ino_ret = key.offset;
175out:
176 btrfs_release_path(&path);
177 return ret;
178}
179
Marek Behún29387542017-09-03 17:00:28 +0200180static inline int next_length(const char *path)
181{
182 int res = 0;
Qu Wenruob086e252020-06-24 18:03:04 +0200183 while (*path != '\0' && *path != '/') {
184 ++res;
185 ++path;
186 if (res > BTRFS_NAME_LEN)
187 break;
188 }
Marek Behún29387542017-09-03 17:00:28 +0200189 return res;
190}
191
192static inline const char *skip_current_directories(const char *cur)
193{
194 while (1) {
195 if (cur[0] == '/')
196 ++cur;
197 else if (cur[0] == '.' && cur[1] == '/')
198 cur += 2;
199 else
200 break;
201 }
202
203 return cur;
204}
205
Qu Wenruoe565bd42020-06-24 18:03:05 +0200206/*
207 * Resolve one filename of @ino of @root.
208 *
209 * key_ret: The child key (either INODE_ITEM or ROOT_ITEM type)
210 * type_ret: BTRFS_FT_* of the child inode.
211 *
212 * Return 0 with above members filled.
213 * Return <0 for error.
214 */
215static 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)
218{
219 struct btrfs_dir_item *dir_item;
220 struct btrfs_path path;
221 int ret = 0;
222
223 btrfs_init_path(&path);
224
225 dir_item = btrfs_lookup_dir_item(NULL, root, &path, ino, name,
226 namelen, 0);
227 if (IS_ERR(dir_item)) {
228 ret = PTR_ERR(dir_item);
229 goto out;
230 }
231
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);
234out:
235 btrfs_release_path(&path);
236 return ret;
237}
238
239/*
240 * Resolve a full path @filename. The start point is @ino of @root.
241 *
242 * The result will be filled into @root_ret, @ino_ret and @type_ret.
243 */
244int 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)
247{
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;
252 u64 next_ino;
253 u8 next_type;
Qu Wenruo11acf0f2020-10-31 09:07:49 +0800254 u8 type = BTRFS_FT_UNKNOWN;
Qu Wenruoe565bd42020-06-24 18:03:05 +0200255 int len;
256 int ret = 0;
257
258 /* If the path is absolute path, also search from fs root */
259 if (*cur == '/') {
260 root = fs_info->fs_root;
261 ino = btrfs_root_dirid(&root->root_item);
262 type = BTRFS_FT_DIR;
263 }
264
265 while (*cur != '\0') {
266 cur = skip_current_directories(cur);
267
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;
273 }
274
275 if (len == 1 && cur[0] == '.')
276 break;
277
278 if (len == 2 && cur[0] == '.' && cur[1] == '.') {
279 /* Go one level up */
280 ret = get_parent_inode(root, ino, &next_root, &next_ino);
281 if (ret < 0)
282 return ret;
283 root = next_root;
284 ino = next_ino;
285 goto next;
286 }
287
288 if (!*cur)
289 break;
290
291 ret = resolve_one_filename(root, ino, cur, len, &key, &type);
292 if (ret < 0)
293 return ret;
294
295 if (key.type == BTRFS_ROOT_ITEM_KEY) {
296 /* Child inode is a subvolume */
297
298 next_root = btrfs_read_fs_root(fs_info, &key);
299 if (IS_ERR(next_root))
300 return PTR_ERR(next_root);
301 root = 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 */
305
306 char *target;
307
308 if (symlink_limit == 0) {
309 error("%s: Too much symlinks!", __func__);
310 return -EMLINK;
311 }
312 target = malloc(fs_info->sectorsize);
313 if (!target)
314 return -ENOMEM;
315 ret = btrfs_readlink(root, key.objectid, target);
316 if (ret < 0) {
317 free(target);
318 return ret;
319 }
320 target[ret] = '\0';
321
322 ret = btrfs_lookup_path(root, ino, target, &next_root,
323 &next_ino, &next_type,
324 symlink_limit);
325 if (ret < 0)
326 return ret;
327 root = next_root;
328 ino = next_ino;
329 type = next_type;
330 } else {
331 /* Child inode is an inode */
332 ino = key.objectid;
333 }
334next:
335 cur += len;
336 }
337
Qu Wenruo11acf0f2020-10-31 09:07:49 +0800338 /* We haven't found anything, but still get no error? */
339 if (type == BTRFS_FT_UNKNOWN && !ret)
340 ret = -EUCLEAN;
341
Qu Wenruoe565bd42020-06-24 18:03:05 +0200342 if (!ret) {
343 *root_ret = root;
344 *ino_ret = ino;
345 *type_ret = type;
346 }
347
348 return ret;
349}
350
Qu Wenruo83b04df2020-06-24 18:03:09 +0200351/*
352 * Read out inline extent.
353 *
354 * Since inline extent should only exist for offset 0, no need for extra
355 * parameters.
356 * Truncating should be handled by the caller.
357 *
358 * Return the number of bytes read.
359 * Return <0 for error.
360 */
361int btrfs_read_extent_inline(struct btrfs_path *path,
362 struct btrfs_file_extent_item *fi, char *dest)
363{
364 struct extent_buffer *leaf = path->nodes[0];
365 int slot = path->slots[0];
366 char *cbuf = NULL;
367 char *dbuf = NULL;
368 u32 csize;
369 u32 dsize;
370 int ret;
371
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),
377 csize);
378 return csize;
379 }
380
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) {
386 ret = -ENOMEM;
387 goto out;
388 }
389 read_extent_buffer(leaf, cbuf, btrfs_file_extent_inline_start(fi),
390 csize);
391 ret = btrfs_decompress(btrfs_file_extent_compression(leaf, fi),
392 cbuf, csize, dbuf, dsize);
393 if (ret < 0 || ret != dsize) {
394 ret = -EIO;
395 goto out;
396 }
397 memcpy(dest, dbuf, dsize);
398 ret = dsize;
399out:
400 free(cbuf);
401 free(dbuf);
402 return ret;
403}
404
405/*
406 * Read out regular extent.
407 *
408 * Truncating should be handled by the caller.
409 *
410 * @offset and @len should not cross the extent boundary.
411 * Return the number of bytes read.
412 * Return <0 for error.
413 */
414int btrfs_read_extent_reg(struct btrfs_path *path,
415 struct btrfs_file_extent_item *fi, u64 offset,
416 int len, char *dest)
417{
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;
422 u64 disk_bytenr;
423 u64 read;
424 char *cbuf = NULL;
425 char *dbuf = NULL;
426 u32 csize;
427 u32 dsize;
428 bool finished = false;
429 int num_copies;
430 int i;
431 int slot = path->slots[0];
432 int ret;
433
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);
440
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);
445 return len;
446 }
447
448 if (btrfs_file_extent_compression(leaf, fi) == BTRFS_COMPRESS_NONE) {
449 u64 logical;
450
451 logical = btrfs_file_extent_disk_bytenr(leaf, fi) +
452 btrfs_file_extent_offset(leaf, fi) +
453 offset - key.offset;
454 read = len;
455
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)
460 continue;
461 finished = true;
462 break;
463 }
464 if (!finished)
465 return -EIO;
466 return len;
467 }
468
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);
473
474 cbuf = malloc_cache_aligned(csize);
475 dbuf = malloc_cache_aligned(dsize);
476 if (!cbuf || !dbuf) {
477 ret = -ENOMEM;
478 goto out;
479 }
480 /* For compressed extent, we must read the whole on-disk extent */
481 for (i = 1; i <= num_copies; i++) {
482 read = csize;
483 ret = read_extent_data(fs_info, cbuf, disk_bytenr,
484 &read, i);
485 if (ret < 0 || read != csize)
486 continue;
487 finished = true;
488 break;
489 }
490 if (!finished) {
491 ret = -EIO;
492 goto out;
493 }
494
495 ret = btrfs_decompress(btrfs_file_extent_compression(leaf, fi), cbuf,
496 csize, dbuf, dsize);
497 if (ret != dsize) {
498 ret = -EIO;
499 goto out;
500 }
501 /* Then copy the needed part */
502 memcpy(dest, dbuf + btrfs_file_extent_offset(leaf, fi), len);
503 ret = len;
504out:
505 free(cbuf);
506 free(dbuf);
507 return ret;
508}
Qu Wenruo4650ed12020-06-24 18:03:10 +0200509
510/*
511 * Get the first file extent that covers bytenr @file_offset.
512 *
513 * @file_offset must be aligned to sectorsize.
514 *
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.
519 */
520static int lookup_data_extent(struct btrfs_root *root, struct btrfs_path *path,
521 u64 ino, u64 file_offset, u64 *next_offset)
522{
523 struct btrfs_key key;
524 struct btrfs_file_extent_item *fi;
525 u8 extent_type;
526 int ret = 0;
527
528 ASSERT(IS_ALIGNED(file_offset, root->fs_info->sectorsize));
529 key.objectid = ino;
530 key.type = BTRFS_EXTENT_DATA_KEY;
531 key.offset = file_offset;
532
533 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
534 /* Error or we're already at the file extent */
535 if (ret <= 0)
536 return ret;
537 if (ret > 0) {
538 /* Check previous file extent */
539 ret = btrfs_previous_item(root, path, ino,
540 BTRFS_EXTENT_DATA_KEY);
541 if (ret < 0)
542 return ret;
543 if (ret > 0)
544 goto check_next;
545 }
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)
550 goto check_next;
551
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)
557 return 0;
558 /* Inline extent should be the only extent, no next extent. */
559 *next_offset = 0;
560 return 1;
561 }
562
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)
566 return 0;
567check_next:
568 ret = btrfs_next_item(root, path);
569 if (ret < 0)
570 return ret;
571 if (ret > 0) {
572 *next_offset = 0;
573 return 1;
574 }
575
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) {
582 *next_offset = 0;
583 return 1;
584 }
585 /* Current file extent already beyond @file_offset */
586 if (key.offset > file_offset) {
587 *next_offset = key.offset;
588 return 1;
589 }
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)
593 return 0;
594 /* This file extent ends before @file_offset, check next */
595 ret = btrfs_next_item(root, path);
596 if (ret < 0)
597 return ret;
598 if (ret > 0) {
599 *next_offset = 0;
600 return 1;
601 }
602 btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
603 if (key.type != BTRFS_EXTENT_DATA_KEY || key.objectid != ino) {
604 *next_offset = 0;
605 return 1;
606 }
607 *next_offset = key.offset;
608 return 1;
609}
Qu Wenruocf5a3a02020-06-24 18:03:11 +0200610
611static int read_and_truncate_page(struct btrfs_path *path,
612 struct btrfs_file_extent_item *fi,
613 int start, int len, char *dest)
614{
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);
618 u8 extent_type;
619 char *buf;
620 int page_off = start - aligned_start;
621 int page_len = fs_info->sectorsize - page_off;
622 int ret;
623
624 ASSERT(start + len <= aligned_start + fs_info->sectorsize);
625 buf = malloc_cache_aligned(fs_info->sectorsize);
626 if (!buf)
627 return -ENOMEM;
628
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));
633 free(buf);
634 return len;
635 }
636
637 ret = btrfs_read_extent_reg(path, fi,
638 round_down(start, fs_info->sectorsize),
639 fs_info->sectorsize, buf);
640 if (ret < 0) {
641 free(buf);
642 return ret;
643 }
644 memcpy(dest, buf + page_off, page_len);
645 free(buf);
646 return len;
647}
648
649int btrfs_file_read(struct btrfs_root *root, u64 ino, u64 file_offset, u64 len,
650 char *dest)
651{
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);
658 u64 next_offset;
659 u64 cur = aligned_start;
660 int ret = 0;
661
662 btrfs_init_path(&path);
663
664 /* Set the whole dest all zero, so we won't need to bother holes */
665 memset(dest, 0, len);
666
667 /* Read out the leading unaligned part */
668 if (aligned_start != file_offset) {
669 ret = lookup_data_extent(root, &path, ino, aligned_start,
670 &next_offset);
671 if (ret < 0)
672 goto out;
673 if (ret == 0) {
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) -
679 file_offset, dest);
680 if (ret < 0)
681 goto out;
682 cur += fs_info->sectorsize;
683 } else {
684 /* The whole file is a hole */
685 if (!next_offset) {
686 memset(dest, 0, len);
687 return len;
688 }
689 cur = next_offset;
690 }
691 }
692
693 /* Read the aligned part */
694 while (cur < aligned_end) {
695 u64 extent_num_bytes;
696 u8 type;
697
698 btrfs_release_path(&path);
699 ret = lookup_data_extent(root, &path, ino, cur, &next_offset);
700 if (ret < 0)
701 goto out;
702 if (ret > 0) {
703 /* No next, direct exit */
704 if (!next_offset) {
705 ret = 0;
706 goto out;
707 }
708 }
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);
715 goto out;
716 }
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(
721 path.nodes[0], fi);
722 continue;
723 }
724
725 /* Read the remaining part of the extent */
726 extent_num_bytes = btrfs_file_extent_num_bytes(path.nodes[0],
727 fi);
728 ret = btrfs_read_extent_reg(&path, fi, cur,
729 min(extent_num_bytes, aligned_end - cur),
730 dest + cur - file_offset);
731 if (ret < 0)
732 goto out;
733 cur += min(extent_num_bytes, aligned_end - cur);
734 }
735
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,
740 &next_offset);
741 /* <0 is error, >0 means no extent */
742 if (ret)
743 goto out;
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);
749 }
750out:
751 btrfs_release_path(&path);
752 if (ret < 0)
753 return ret;
754 return len;
755}