1 // SPDX-License-Identifier: GPL-2.0
2 /*
3  * Copyright (C) 2007 Oracle.  All rights reserved.
4  */
5 
6 #include <linux/bio.h>
7 #include <linux/slab.h>
8 #include <linux/pagemap.h>
9 #include <linux/highmem.h>
10 #include <linux/sched/mm.h>
11 #include <crypto/hash.h>
12 #include "misc.h"
13 #include "ctree.h"
14 #include "disk-io.h"
15 #include "transaction.h"
16 #include "volumes.h"
17 #include "print-tree.h"
18 #include "compression.h"
19 
20 #define __MAX_CSUM_ITEMS(r, size) ((unsigned long)(((BTRFS_LEAF_DATA_SIZE(r) - \
21 				   sizeof(struct btrfs_item) * 2) / \
22 				  size) - 1))
23 
24 #define MAX_CSUM_ITEMS(r, size) (min_t(u32, __MAX_CSUM_ITEMS(r, size), \
25 				       PAGE_SIZE))
26 
27 /**
28  * Set inode's size according to filesystem options
29  *
30  * @inode:      inode we want to update the disk_i_size for
31  * @new_i_size: i_size we want to set to, 0 if we use i_size
32  *
33  * With NO_HOLES set this simply sets the disk_is_size to whatever i_size_read()
34  * returns as it is perfectly fine with a file that has holes without hole file
35  * extent items.
36  *
37  * However without NO_HOLES we need to only return the area that is contiguous
38  * from the 0 offset of the file.  Otherwise we could end up adjust i_size up
39  * to an extent that has a gap in between.
40  *
41  * Finally new_i_size should only be set in the case of truncate where we're not
42  * ready to use i_size_read() as the limiter yet.
43  */
btrfs_inode_safe_disk_i_size_write(struct btrfs_inode * inode,u64 new_i_size)44 void btrfs_inode_safe_disk_i_size_write(struct btrfs_inode *inode, u64 new_i_size)
45 {
46 	struct btrfs_fs_info *fs_info = inode->root->fs_info;
47 	u64 start, end, i_size;
48 	int ret;
49 
50 	i_size = new_i_size ?: i_size_read(&inode->vfs_inode);
51 	if (btrfs_fs_incompat(fs_info, NO_HOLES)) {
52 		inode->disk_i_size = i_size;
53 		return;
54 	}
55 
56 	spin_lock(&inode->lock);
57 	ret = find_contiguous_extent_bit(&inode->file_extent_tree, 0, &start,
58 					 &end, EXTENT_DIRTY);
59 	if (!ret && start == 0)
60 		i_size = min(i_size, end + 1);
61 	else
62 		i_size = 0;
63 	inode->disk_i_size = i_size;
64 	spin_unlock(&inode->lock);
65 }
66 
67 /**
68  * Mark range within a file as having a new extent inserted
69  *
70  * @inode: inode being modified
71  * @start: start file offset of the file extent we've inserted
72  * @len:   logical length of the file extent item
73  *
74  * Call when we are inserting a new file extent where there was none before.
75  * Does not need to call this in the case where we're replacing an existing file
76  * extent, however if not sure it's fine to call this multiple times.
77  *
78  * The start and len must match the file extent item, so thus must be sectorsize
79  * aligned.
80  */
btrfs_inode_set_file_extent_range(struct btrfs_inode * inode,u64 start,u64 len)81 int btrfs_inode_set_file_extent_range(struct btrfs_inode *inode, u64 start,
82 				      u64 len)
83 {
84 	if (len == 0)
85 		return 0;
86 
87 	ASSERT(IS_ALIGNED(start + len, inode->root->fs_info->sectorsize));
88 
89 	if (btrfs_fs_incompat(inode->root->fs_info, NO_HOLES))
90 		return 0;
91 	return set_extent_bits(&inode->file_extent_tree, start, start + len - 1,
92 			       EXTENT_DIRTY);
93 }
94 
95 /**
96  * Marks an inode range as not having a backing extent
97  *
98  * @inode: inode being modified
99  * @start: start file offset of the file extent we've inserted
100  * @len:   logical length of the file extent item
101  *
102  * Called when we drop a file extent, for example when we truncate.  Doesn't
103  * need to be called for cases where we're replacing a file extent, like when
104  * we've COWed a file extent.
105  *
106  * The start and len must match the file extent item, so thus must be sectorsize
107  * aligned.
108  */
btrfs_inode_clear_file_extent_range(struct btrfs_inode * inode,u64 start,u64 len)109 int btrfs_inode_clear_file_extent_range(struct btrfs_inode *inode, u64 start,
110 					u64 len)
111 {
112 	if (len == 0)
113 		return 0;
114 
115 	ASSERT(IS_ALIGNED(start + len, inode->root->fs_info->sectorsize) ||
116 	       len == (u64)-1);
117 
118 	if (btrfs_fs_incompat(inode->root->fs_info, NO_HOLES))
119 		return 0;
120 	return clear_extent_bit(&inode->file_extent_tree, start,
121 				start + len - 1, EXTENT_DIRTY, 0, 0, NULL);
122 }
123 
max_ordered_sum_bytes(struct btrfs_fs_info * fs_info,u16 csum_size)124 static inline u32 max_ordered_sum_bytes(struct btrfs_fs_info *fs_info,
125 					u16 csum_size)
126 {
127 	u32 ncsums = (PAGE_SIZE - sizeof(struct btrfs_ordered_sum)) / csum_size;
128 
129 	return ncsums * fs_info->sectorsize;
130 }
131 
btrfs_insert_file_extent(struct btrfs_trans_handle * trans,struct btrfs_root * root,u64 objectid,u64 pos,u64 disk_offset,u64 disk_num_bytes,u64 num_bytes,u64 offset,u64 ram_bytes,u8 compression,u8 encryption,u16 other_encoding)132 int btrfs_insert_file_extent(struct btrfs_trans_handle *trans,
133 			     struct btrfs_root *root,
134 			     u64 objectid, u64 pos,
135 			     u64 disk_offset, u64 disk_num_bytes,
136 			     u64 num_bytes, u64 offset, u64 ram_bytes,
137 			     u8 compression, u8 encryption, u16 other_encoding)
138 {
139 	int ret = 0;
140 	struct btrfs_file_extent_item *item;
141 	struct btrfs_key file_key;
142 	struct btrfs_path *path;
143 	struct extent_buffer *leaf;
144 
145 	path = btrfs_alloc_path();
146 	if (!path)
147 		return -ENOMEM;
148 	file_key.objectid = objectid;
149 	file_key.offset = pos;
150 	file_key.type = BTRFS_EXTENT_DATA_KEY;
151 
152 	ret = btrfs_insert_empty_item(trans, root, path, &file_key,
153 				      sizeof(*item));
154 	if (ret < 0)
155 		goto out;
156 	BUG_ON(ret); /* Can't happen */
157 	leaf = path->nodes[0];
158 	item = btrfs_item_ptr(leaf, path->slots[0],
159 			      struct btrfs_file_extent_item);
160 	btrfs_set_file_extent_disk_bytenr(leaf, item, disk_offset);
161 	btrfs_set_file_extent_disk_num_bytes(leaf, item, disk_num_bytes);
162 	btrfs_set_file_extent_offset(leaf, item, offset);
163 	btrfs_set_file_extent_num_bytes(leaf, item, num_bytes);
164 	btrfs_set_file_extent_ram_bytes(leaf, item, ram_bytes);
165 	btrfs_set_file_extent_generation(leaf, item, trans->transid);
166 	btrfs_set_file_extent_type(leaf, item, BTRFS_FILE_EXTENT_REG);
167 	btrfs_set_file_extent_compression(leaf, item, compression);
168 	btrfs_set_file_extent_encryption(leaf, item, encryption);
169 	btrfs_set_file_extent_other_encoding(leaf, item, other_encoding);
170 
171 	btrfs_mark_buffer_dirty(leaf);
172 out:
173 	btrfs_free_path(path);
174 	return ret;
175 }
176 
177 static struct btrfs_csum_item *
btrfs_lookup_csum(struct btrfs_trans_handle * trans,struct btrfs_root * root,struct btrfs_path * path,u64 bytenr,int cow)178 btrfs_lookup_csum(struct btrfs_trans_handle *trans,
179 		  struct btrfs_root *root,
180 		  struct btrfs_path *path,
181 		  u64 bytenr, int cow)
182 {
183 	struct btrfs_fs_info *fs_info = root->fs_info;
184 	int ret;
185 	struct btrfs_key file_key;
186 	struct btrfs_key found_key;
187 	struct btrfs_csum_item *item;
188 	struct extent_buffer *leaf;
189 	u64 csum_offset = 0;
190 	const u32 csum_size = fs_info->csum_size;
191 	int csums_in_item;
192 
193 	file_key.objectid = BTRFS_EXTENT_CSUM_OBJECTID;
194 	file_key.offset = bytenr;
195 	file_key.type = BTRFS_EXTENT_CSUM_KEY;
196 	ret = btrfs_search_slot(trans, root, &file_key, path, 0, cow);
197 	if (ret < 0)
198 		goto fail;
199 	leaf = path->nodes[0];
200 	if (ret > 0) {
201 		ret = 1;
202 		if (path->slots[0] == 0)
203 			goto fail;
204 		path->slots[0]--;
205 		btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
206 		if (found_key.type != BTRFS_EXTENT_CSUM_KEY)
207 			goto fail;
208 
209 		csum_offset = (bytenr - found_key.offset) >>
210 				fs_info->sectorsize_bits;
211 		csums_in_item = btrfs_item_size(leaf, path->slots[0]);
212 		csums_in_item /= csum_size;
213 
214 		if (csum_offset == csums_in_item) {
215 			ret = -EFBIG;
216 			goto fail;
217 		} else if (csum_offset > csums_in_item) {
218 			goto fail;
219 		}
220 	}
221 	item = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_csum_item);
222 	item = (struct btrfs_csum_item *)((unsigned char *)item +
223 					  csum_offset * csum_size);
224 	return item;
225 fail:
226 	if (ret > 0)
227 		ret = -ENOENT;
228 	return ERR_PTR(ret);
229 }
230 
btrfs_lookup_file_extent(struct btrfs_trans_handle * trans,struct btrfs_root * root,struct btrfs_path * path,u64 objectid,u64 offset,int mod)231 int btrfs_lookup_file_extent(struct btrfs_trans_handle *trans,
232 			     struct btrfs_root *root,
233 			     struct btrfs_path *path, u64 objectid,
234 			     u64 offset, int mod)
235 {
236 	struct btrfs_key file_key;
237 	int ins_len = mod < 0 ? -1 : 0;
238 	int cow = mod != 0;
239 
240 	file_key.objectid = objectid;
241 	file_key.offset = offset;
242 	file_key.type = BTRFS_EXTENT_DATA_KEY;
243 
244 	return btrfs_search_slot(trans, root, &file_key, path, ins_len, cow);
245 }
246 
247 /*
248  * Find checksums for logical bytenr range [disk_bytenr, disk_bytenr + len) and
249  * estore the result to @dst.
250  *
251  * Return >0 for the number of sectors we found.
252  * Return 0 for the range [disk_bytenr, disk_bytenr + sectorsize) has no csum
253  * for it. Caller may want to try next sector until one range is hit.
254  * Return <0 for fatal error.
255  */
search_csum_tree(struct btrfs_fs_info * fs_info,struct btrfs_path * path,u64 disk_bytenr,u64 len,u8 * dst)256 static int search_csum_tree(struct btrfs_fs_info *fs_info,
257 			    struct btrfs_path *path, u64 disk_bytenr,
258 			    u64 len, u8 *dst)
259 {
260 	struct btrfs_root *csum_root;
261 	struct btrfs_csum_item *item = NULL;
262 	struct btrfs_key key;
263 	const u32 sectorsize = fs_info->sectorsize;
264 	const u32 csum_size = fs_info->csum_size;
265 	u32 itemsize;
266 	int ret;
267 	u64 csum_start;
268 	u64 csum_len;
269 
270 	ASSERT(IS_ALIGNED(disk_bytenr, sectorsize) &&
271 	       IS_ALIGNED(len, sectorsize));
272 
273 	/* Check if the current csum item covers disk_bytenr */
274 	if (path->nodes[0]) {
275 		item = btrfs_item_ptr(path->nodes[0], path->slots[0],
276 				      struct btrfs_csum_item);
277 		btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
278 		itemsize = btrfs_item_size(path->nodes[0], path->slots[0]);
279 
280 		csum_start = key.offset;
281 		csum_len = (itemsize / csum_size) * sectorsize;
282 
283 		if (in_range(disk_bytenr, csum_start, csum_len))
284 			goto found;
285 	}
286 
287 	/* Current item doesn't contain the desired range, search again */
288 	btrfs_release_path(path);
289 	csum_root = btrfs_csum_root(fs_info, disk_bytenr);
290 	item = btrfs_lookup_csum(NULL, csum_root, path, disk_bytenr, 0);
291 	if (IS_ERR(item)) {
292 		ret = PTR_ERR(item);
293 		goto out;
294 	}
295 	btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
296 	itemsize = btrfs_item_size(path->nodes[0], path->slots[0]);
297 
298 	csum_start = key.offset;
299 	csum_len = (itemsize / csum_size) * sectorsize;
300 	ASSERT(in_range(disk_bytenr, csum_start, csum_len));
301 
302 found:
303 	ret = (min(csum_start + csum_len, disk_bytenr + len) -
304 		   disk_bytenr) >> fs_info->sectorsize_bits;
305 	read_extent_buffer(path->nodes[0], dst, (unsigned long)item,
306 			ret * csum_size);
307 out:
308 	if (ret == -ENOENT || ret == -EFBIG)
309 		ret = 0;
310 	return ret;
311 }
312 
313 /*
314  * Locate the file_offset of @cur_disk_bytenr of a @bio.
315  *
316  * Bio of btrfs represents read range of
317  * [bi_sector << 9, bi_sector << 9 + bi_size).
318  * Knowing this, we can iterate through each bvec to locate the page belong to
319  * @cur_disk_bytenr and get the file offset.
320  *
321  * @inode is used to determine if the bvec page really belongs to @inode.
322  *
323  * Return 0 if we can't find the file offset
324  * Return >0 if we find the file offset and restore it to @file_offset_ret
325  */
search_file_offset_in_bio(struct bio * bio,struct inode * inode,u64 disk_bytenr,u64 * file_offset_ret)326 static int search_file_offset_in_bio(struct bio *bio, struct inode *inode,
327 				     u64 disk_bytenr, u64 *file_offset_ret)
328 {
329 	struct bvec_iter iter;
330 	struct bio_vec bvec;
331 	u64 cur = bio->bi_iter.bi_sector << SECTOR_SHIFT;
332 	int ret = 0;
333 
334 	bio_for_each_segment(bvec, bio, iter) {
335 		struct page *page = bvec.bv_page;
336 
337 		if (cur > disk_bytenr)
338 			break;
339 		if (cur + bvec.bv_len <= disk_bytenr) {
340 			cur += bvec.bv_len;
341 			continue;
342 		}
343 		ASSERT(in_range(disk_bytenr, cur, bvec.bv_len));
344 		if (page->mapping && page->mapping->host &&
345 		    page->mapping->host == inode) {
346 			ret = 1;
347 			*file_offset_ret = page_offset(page) + bvec.bv_offset +
348 					   disk_bytenr - cur;
349 			break;
350 		}
351 	}
352 	return ret;
353 }
354 
355 /**
356  * Lookup the checksum for the read bio in csum tree.
357  *
358  * @inode: inode that the bio is for.
359  * @bio: bio to look up.
360  * @dst: Buffer of size nblocks * btrfs_super_csum_size() used to return
361  *       checksum (nblocks = bio->bi_iter.bi_size / fs_info->sectorsize). If
362  *       NULL, the checksum buffer is allocated and returned in
363  *       btrfs_bio(bio)->csum instead.
364  *
365  * Return: BLK_STS_RESOURCE if allocating memory fails, BLK_STS_OK otherwise.
366  */
btrfs_lookup_bio_sums(struct inode * inode,struct bio * bio,u8 * dst)367 blk_status_t btrfs_lookup_bio_sums(struct inode *inode, struct bio *bio, u8 *dst)
368 {
369 	struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
370 	struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
371 	struct btrfs_bio *bbio = NULL;
372 	struct btrfs_path *path;
373 	const u32 sectorsize = fs_info->sectorsize;
374 	const u32 csum_size = fs_info->csum_size;
375 	u32 orig_len = bio->bi_iter.bi_size;
376 	u64 orig_disk_bytenr = bio->bi_iter.bi_sector << SECTOR_SHIFT;
377 	u64 cur_disk_bytenr;
378 	u8 *csum;
379 	const unsigned int nblocks = orig_len >> fs_info->sectorsize_bits;
380 	int count = 0;
381 	blk_status_t ret = BLK_STS_OK;
382 
383 	if ((BTRFS_I(inode)->flags & BTRFS_INODE_NODATASUM) ||
384 	    test_bit(BTRFS_FS_STATE_NO_CSUMS, &fs_info->fs_state))
385 		return BLK_STS_OK;
386 
387 	/*
388 	 * This function is only called for read bio.
389 	 *
390 	 * This means two things:
391 	 * - All our csums should only be in csum tree
392 	 *   No ordered extents csums, as ordered extents are only for write
393 	 *   path.
394 	 * - No need to bother any other info from bvec
395 	 *   Since we're looking up csums, the only important info is the
396 	 *   disk_bytenr and the length, which can be extracted from bi_iter
397 	 *   directly.
398 	 */
399 	ASSERT(bio_op(bio) == REQ_OP_READ);
400 	path = btrfs_alloc_path();
401 	if (!path)
402 		return BLK_STS_RESOURCE;
403 
404 	if (!dst) {
405 		bbio = btrfs_bio(bio);
406 
407 		if (nblocks * csum_size > BTRFS_BIO_INLINE_CSUM_SIZE) {
408 			bbio->csum = kmalloc_array(nblocks, csum_size, GFP_NOFS);
409 			if (!bbio->csum) {
410 				btrfs_free_path(path);
411 				return BLK_STS_RESOURCE;
412 			}
413 		} else {
414 			bbio->csum = bbio->csum_inline;
415 		}
416 		csum = bbio->csum;
417 	} else {
418 		csum = dst;
419 	}
420 
421 	/*
422 	 * If requested number of sectors is larger than one leaf can contain,
423 	 * kick the readahead for csum tree.
424 	 */
425 	if (nblocks > fs_info->csums_per_leaf)
426 		path->reada = READA_FORWARD;
427 
428 	/*
429 	 * the free space stuff is only read when it hasn't been
430 	 * updated in the current transaction.  So, we can safely
431 	 * read from the commit root and sidestep a nasty deadlock
432 	 * between reading the free space cache and updating the csum tree.
433 	 */
434 	if (btrfs_is_free_space_inode(BTRFS_I(inode))) {
435 		path->search_commit_root = 1;
436 		path->skip_locking = 1;
437 	}
438 
439 	for (cur_disk_bytenr = orig_disk_bytenr;
440 	     cur_disk_bytenr < orig_disk_bytenr + orig_len;
441 	     cur_disk_bytenr += (count * sectorsize)) {
442 		u64 search_len = orig_disk_bytenr + orig_len - cur_disk_bytenr;
443 		unsigned int sector_offset;
444 		u8 *csum_dst;
445 
446 		/*
447 		 * Although both cur_disk_bytenr and orig_disk_bytenr is u64,
448 		 * we're calculating the offset to the bio start.
449 		 *
450 		 * Bio size is limited to UINT_MAX, thus unsigned int is large
451 		 * enough to contain the raw result, not to mention the right
452 		 * shifted result.
453 		 */
454 		ASSERT(cur_disk_bytenr - orig_disk_bytenr < UINT_MAX);
455 		sector_offset = (cur_disk_bytenr - orig_disk_bytenr) >>
456 				fs_info->sectorsize_bits;
457 		csum_dst = csum + sector_offset * csum_size;
458 
459 		count = search_csum_tree(fs_info, path, cur_disk_bytenr,
460 					 search_len, csum_dst);
461 		if (count < 0) {
462 			ret = errno_to_blk_status(count);
463 			if (bbio)
464 				btrfs_bio_free_csum(bbio);
465 			break;
466 		}
467 
468 		/*
469 		 * We didn't find a csum for this range.  We need to make sure
470 		 * we complain loudly about this, because we are not NODATASUM.
471 		 *
472 		 * However for the DATA_RELOC inode we could potentially be
473 		 * relocating data extents for a NODATASUM inode, so the inode
474 		 * itself won't be marked with NODATASUM, but the extent we're
475 		 * copying is in fact NODATASUM.  If we don't find a csum we
476 		 * assume this is the case.
477 		 */
478 		if (count == 0) {
479 			memset(csum_dst, 0, csum_size);
480 			count = 1;
481 
482 			if (BTRFS_I(inode)->root->root_key.objectid ==
483 			    BTRFS_DATA_RELOC_TREE_OBJECTID) {
484 				u64 file_offset;
485 				int ret;
486 
487 				ret = search_file_offset_in_bio(bio, inode,
488 						cur_disk_bytenr, &file_offset);
489 				if (ret)
490 					set_extent_bits(io_tree, file_offset,
491 						file_offset + sectorsize - 1,
492 						EXTENT_NODATASUM);
493 			} else {
494 				btrfs_warn_rl(fs_info,
495 			"csum hole found for disk bytenr range [%llu, %llu)",
496 				cur_disk_bytenr, cur_disk_bytenr + sectorsize);
497 			}
498 		}
499 	}
500 
501 	btrfs_free_path(path);
502 	return ret;
503 }
504 
btrfs_lookup_csums_range(struct btrfs_root * root,u64 start,u64 end,struct list_head * list,int search_commit)505 int btrfs_lookup_csums_range(struct btrfs_root *root, u64 start, u64 end,
506 			     struct list_head *list, int search_commit)
507 {
508 	struct btrfs_fs_info *fs_info = root->fs_info;
509 	struct btrfs_key key;
510 	struct btrfs_path *path;
511 	struct extent_buffer *leaf;
512 	struct btrfs_ordered_sum *sums;
513 	struct btrfs_csum_item *item;
514 	LIST_HEAD(tmplist);
515 	unsigned long offset;
516 	int ret;
517 	size_t size;
518 	u64 csum_end;
519 	const u32 csum_size = fs_info->csum_size;
520 
521 	ASSERT(IS_ALIGNED(start, fs_info->sectorsize) &&
522 	       IS_ALIGNED(end + 1, fs_info->sectorsize));
523 
524 	path = btrfs_alloc_path();
525 	if (!path)
526 		return -ENOMEM;
527 
528 	if (search_commit) {
529 		path->skip_locking = 1;
530 		path->reada = READA_FORWARD;
531 		path->search_commit_root = 1;
532 	}
533 
534 	key.objectid = BTRFS_EXTENT_CSUM_OBJECTID;
535 	key.offset = start;
536 	key.type = BTRFS_EXTENT_CSUM_KEY;
537 
538 	ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
539 	if (ret < 0)
540 		goto fail;
541 	if (ret > 0 && path->slots[0] > 0) {
542 		leaf = path->nodes[0];
543 		btrfs_item_key_to_cpu(leaf, &key, path->slots[0] - 1);
544 		if (key.objectid == BTRFS_EXTENT_CSUM_OBJECTID &&
545 		    key.type == BTRFS_EXTENT_CSUM_KEY) {
546 			offset = (start - key.offset) >> fs_info->sectorsize_bits;
547 			if (offset * csum_size <
548 			    btrfs_item_size(leaf, path->slots[0] - 1))
549 				path->slots[0]--;
550 		}
551 	}
552 
553 	while (start <= end) {
554 		leaf = path->nodes[0];
555 		if (path->slots[0] >= btrfs_header_nritems(leaf)) {
556 			ret = btrfs_next_leaf(root, path);
557 			if (ret < 0)
558 				goto fail;
559 			if (ret > 0)
560 				break;
561 			leaf = path->nodes[0];
562 		}
563 
564 		btrfs_item_key_to_cpu(leaf, &key, path->slots[0]);
565 		if (key.objectid != BTRFS_EXTENT_CSUM_OBJECTID ||
566 		    key.type != BTRFS_EXTENT_CSUM_KEY ||
567 		    key.offset > end)
568 			break;
569 
570 		if (key.offset > start)
571 			start = key.offset;
572 
573 		size = btrfs_item_size(leaf, path->slots[0]);
574 		csum_end = key.offset + (size / csum_size) * fs_info->sectorsize;
575 		if (csum_end <= start) {
576 			path->slots[0]++;
577 			continue;
578 		}
579 
580 		csum_end = min(csum_end, end + 1);
581 		item = btrfs_item_ptr(path->nodes[0], path->slots[0],
582 				      struct btrfs_csum_item);
583 		while (start < csum_end) {
584 			size = min_t(size_t, csum_end - start,
585 				     max_ordered_sum_bytes(fs_info, csum_size));
586 			sums = kzalloc(btrfs_ordered_sum_size(fs_info, size),
587 				       GFP_NOFS);
588 			if (!sums) {
589 				ret = -ENOMEM;
590 				goto fail;
591 			}
592 
593 			sums->bytenr = start;
594 			sums->len = (int)size;
595 
596 			offset = (start - key.offset) >> fs_info->sectorsize_bits;
597 			offset *= csum_size;
598 			size >>= fs_info->sectorsize_bits;
599 
600 			read_extent_buffer(path->nodes[0],
601 					   sums->sums,
602 					   ((unsigned long)item) + offset,
603 					   csum_size * size);
604 
605 			start += fs_info->sectorsize * size;
606 			list_add_tail(&sums->list, &tmplist);
607 		}
608 		path->slots[0]++;
609 	}
610 	ret = 0;
611 fail:
612 	while (ret < 0 && !list_empty(&tmplist)) {
613 		sums = list_entry(tmplist.next, struct btrfs_ordered_sum, list);
614 		list_del(&sums->list);
615 		kfree(sums);
616 	}
617 	list_splice_tail(&tmplist, list);
618 
619 	btrfs_free_path(path);
620 	return ret;
621 }
622 
623 /**
624  * Calculate checksums of the data contained inside a bio
625  *
626  * @inode:	 Owner of the data inside the bio
627  * @bio:	 Contains the data to be checksummed
628  * @offset:      If (u64)-1, @bio may contain discontiguous bio vecs, so the
629  *               file offsets are determined from the page offsets in the bio.
630  *               Otherwise, this is the starting file offset of the bio vecs in
631  *               @bio, which must be contiguous.
632  * @one_ordered: If true, @bio only refers to one ordered extent.
633  */
btrfs_csum_one_bio(struct btrfs_inode * inode,struct bio * bio,u64 offset,bool one_ordered)634 blk_status_t btrfs_csum_one_bio(struct btrfs_inode *inode, struct bio *bio,
635 				u64 offset, bool one_ordered)
636 {
637 	struct btrfs_fs_info *fs_info = inode->root->fs_info;
638 	SHASH_DESC_ON_STACK(shash, fs_info->csum_shash);
639 	struct btrfs_ordered_sum *sums;
640 	struct btrfs_ordered_extent *ordered = NULL;
641 	const bool use_page_offsets = (offset == (u64)-1);
642 	char *data;
643 	struct bvec_iter iter;
644 	struct bio_vec bvec;
645 	int index;
646 	unsigned int blockcount;
647 	unsigned long total_bytes = 0;
648 	unsigned long this_sum_bytes = 0;
649 	int i;
650 	unsigned nofs_flag;
651 
652 	nofs_flag = memalloc_nofs_save();
653 	sums = kvzalloc(btrfs_ordered_sum_size(fs_info, bio->bi_iter.bi_size),
654 		       GFP_KERNEL);
655 	memalloc_nofs_restore(nofs_flag);
656 
657 	if (!sums)
658 		return BLK_STS_RESOURCE;
659 
660 	sums->len = bio->bi_iter.bi_size;
661 	INIT_LIST_HEAD(&sums->list);
662 
663 	sums->bytenr = bio->bi_iter.bi_sector << 9;
664 	index = 0;
665 
666 	shash->tfm = fs_info->csum_shash;
667 
668 	bio_for_each_segment(bvec, bio, iter) {
669 		if (use_page_offsets)
670 			offset = page_offset(bvec.bv_page) + bvec.bv_offset;
671 
672 		if (!ordered) {
673 			ordered = btrfs_lookup_ordered_extent(inode, offset);
674 			/*
675 			 * The bio range is not covered by any ordered extent,
676 			 * must be a code logic error.
677 			 */
678 			if (unlikely(!ordered)) {
679 				WARN(1, KERN_WARNING
680 			"no ordered extent for root %llu ino %llu offset %llu\n",
681 				     inode->root->root_key.objectid,
682 				     btrfs_ino(inode), offset);
683 				kvfree(sums);
684 				return BLK_STS_IOERR;
685 			}
686 		}
687 
688 		blockcount = BTRFS_BYTES_TO_BLKS(fs_info,
689 						 bvec.bv_len + fs_info->sectorsize
690 						 - 1);
691 
692 		for (i = 0; i < blockcount; i++) {
693 			if (!one_ordered &&
694 			    !in_range(offset, ordered->file_offset,
695 				      ordered->num_bytes)) {
696 				unsigned long bytes_left;
697 
698 				sums->len = this_sum_bytes;
699 				this_sum_bytes = 0;
700 				btrfs_add_ordered_sum(ordered, sums);
701 				btrfs_put_ordered_extent(ordered);
702 
703 				bytes_left = bio->bi_iter.bi_size - total_bytes;
704 
705 				nofs_flag = memalloc_nofs_save();
706 				sums = kvzalloc(btrfs_ordered_sum_size(fs_info,
707 						      bytes_left), GFP_KERNEL);
708 				memalloc_nofs_restore(nofs_flag);
709 				BUG_ON(!sums); /* -ENOMEM */
710 				sums->len = bytes_left;
711 				ordered = btrfs_lookup_ordered_extent(inode,
712 								offset);
713 				ASSERT(ordered); /* Logic error */
714 				sums->bytenr = (bio->bi_iter.bi_sector << 9)
715 					+ total_bytes;
716 				index = 0;
717 			}
718 
719 			data = bvec_kmap_local(&bvec);
720 			crypto_shash_digest(shash,
721 					    data + (i * fs_info->sectorsize),
722 					    fs_info->sectorsize,
723 					    sums->sums + index);
724 			kunmap_local(data);
725 			index += fs_info->csum_size;
726 			offset += fs_info->sectorsize;
727 			this_sum_bytes += fs_info->sectorsize;
728 			total_bytes += fs_info->sectorsize;
729 		}
730 
731 	}
732 	this_sum_bytes = 0;
733 	btrfs_add_ordered_sum(ordered, sums);
734 	btrfs_put_ordered_extent(ordered);
735 	return 0;
736 }
737 
738 /*
739  * helper function for csum removal, this expects the
740  * key to describe the csum pointed to by the path, and it expects
741  * the csum to overlap the range [bytenr, len]
742  *
743  * The csum should not be entirely contained in the range and the
744  * range should not be entirely contained in the csum.
745  *
746  * This calls btrfs_truncate_item with the correct args based on the
747  * overlap, and fixes up the key as required.
748  */
truncate_one_csum(struct btrfs_fs_info * fs_info,struct btrfs_path * path,struct btrfs_key * key,u64 bytenr,u64 len)749 static noinline void truncate_one_csum(struct btrfs_fs_info *fs_info,
750 				       struct btrfs_path *path,
751 				       struct btrfs_key *key,
752 				       u64 bytenr, u64 len)
753 {
754 	struct extent_buffer *leaf;
755 	const u32 csum_size = fs_info->csum_size;
756 	u64 csum_end;
757 	u64 end_byte = bytenr + len;
758 	u32 blocksize_bits = fs_info->sectorsize_bits;
759 
760 	leaf = path->nodes[0];
761 	csum_end = btrfs_item_size(leaf, path->slots[0]) / csum_size;
762 	csum_end <<= blocksize_bits;
763 	csum_end += key->offset;
764 
765 	if (key->offset < bytenr && csum_end <= end_byte) {
766 		/*
767 		 *         [ bytenr - len ]
768 		 *         [   ]
769 		 *   [csum     ]
770 		 *   A simple truncate off the end of the item
771 		 */
772 		u32 new_size = (bytenr - key->offset) >> blocksize_bits;
773 		new_size *= csum_size;
774 		btrfs_truncate_item(path, new_size, 1);
775 	} else if (key->offset >= bytenr && csum_end > end_byte &&
776 		   end_byte > key->offset) {
777 		/*
778 		 *         [ bytenr - len ]
779 		 *                 [ ]
780 		 *                 [csum     ]
781 		 * we need to truncate from the beginning of the csum
782 		 */
783 		u32 new_size = (csum_end - end_byte) >> blocksize_bits;
784 		new_size *= csum_size;
785 
786 		btrfs_truncate_item(path, new_size, 0);
787 
788 		key->offset = end_byte;
789 		btrfs_set_item_key_safe(fs_info, path, key);
790 	} else {
791 		BUG();
792 	}
793 }
794 
795 /*
796  * deletes the csum items from the csum tree for a given
797  * range of bytes.
798  */
btrfs_del_csums(struct btrfs_trans_handle * trans,struct btrfs_root * root,u64 bytenr,u64 len)799 int btrfs_del_csums(struct btrfs_trans_handle *trans,
800 		    struct btrfs_root *root, u64 bytenr, u64 len)
801 {
802 	struct btrfs_fs_info *fs_info = trans->fs_info;
803 	struct btrfs_path *path;
804 	struct btrfs_key key;
805 	u64 end_byte = bytenr + len;
806 	u64 csum_end;
807 	struct extent_buffer *leaf;
808 	int ret = 0;
809 	const u32 csum_size = fs_info->csum_size;
810 	u32 blocksize_bits = fs_info->sectorsize_bits;
811 
812 	ASSERT(root->root_key.objectid == BTRFS_CSUM_TREE_OBJECTID ||
813 	       root->root_key.objectid == BTRFS_TREE_LOG_OBJECTID);
814 
815 	path = btrfs_alloc_path();
816 	if (!path)
817 		return -ENOMEM;
818 
819 	while (1) {
820 		key.objectid = BTRFS_EXTENT_CSUM_OBJECTID;
821 		key.offset = end_byte - 1;
822 		key.type = BTRFS_EXTENT_CSUM_KEY;
823 
824 		ret = btrfs_search_slot(trans, root, &key, path, -1, 1);
825 		if (ret > 0) {
826 			ret = 0;
827 			if (path->slots[0] == 0)
828 				break;
829 			path->slots[0]--;
830 		} else if (ret < 0) {
831 			break;
832 		}
833 
834 		leaf = path->nodes[0];
835 		btrfs_item_key_to_cpu(leaf, &key, path->slots[0]);
836 
837 		if (key.objectid != BTRFS_EXTENT_CSUM_OBJECTID ||
838 		    key.type != BTRFS_EXTENT_CSUM_KEY) {
839 			break;
840 		}
841 
842 		if (key.offset >= end_byte)
843 			break;
844 
845 		csum_end = btrfs_item_size(leaf, path->slots[0]) / csum_size;
846 		csum_end <<= blocksize_bits;
847 		csum_end += key.offset;
848 
849 		/* this csum ends before we start, we're done */
850 		if (csum_end <= bytenr)
851 			break;
852 
853 		/* delete the entire item, it is inside our range */
854 		if (key.offset >= bytenr && csum_end <= end_byte) {
855 			int del_nr = 1;
856 
857 			/*
858 			 * Check how many csum items preceding this one in this
859 			 * leaf correspond to our range and then delete them all
860 			 * at once.
861 			 */
862 			if (key.offset > bytenr && path->slots[0] > 0) {
863 				int slot = path->slots[0] - 1;
864 
865 				while (slot >= 0) {
866 					struct btrfs_key pk;
867 
868 					btrfs_item_key_to_cpu(leaf, &pk, slot);
869 					if (pk.offset < bytenr ||
870 					    pk.type != BTRFS_EXTENT_CSUM_KEY ||
871 					    pk.objectid !=
872 					    BTRFS_EXTENT_CSUM_OBJECTID)
873 						break;
874 					path->slots[0] = slot;
875 					del_nr++;
876 					key.offset = pk.offset;
877 					slot--;
878 				}
879 			}
880 			ret = btrfs_del_items(trans, root, path,
881 					      path->slots[0], del_nr);
882 			if (ret)
883 				break;
884 			if (key.offset == bytenr)
885 				break;
886 		} else if (key.offset < bytenr && csum_end > end_byte) {
887 			unsigned long offset;
888 			unsigned long shift_len;
889 			unsigned long item_offset;
890 			/*
891 			 *        [ bytenr - len ]
892 			 *     [csum                ]
893 			 *
894 			 * Our bytes are in the middle of the csum,
895 			 * we need to split this item and insert a new one.
896 			 *
897 			 * But we can't drop the path because the
898 			 * csum could change, get removed, extended etc.
899 			 *
900 			 * The trick here is the max size of a csum item leaves
901 			 * enough room in the tree block for a single
902 			 * item header.  So, we split the item in place,
903 			 * adding a new header pointing to the existing
904 			 * bytes.  Then we loop around again and we have
905 			 * a nicely formed csum item that we can neatly
906 			 * truncate.
907 			 */
908 			offset = (bytenr - key.offset) >> blocksize_bits;
909 			offset *= csum_size;
910 
911 			shift_len = (len >> blocksize_bits) * csum_size;
912 
913 			item_offset = btrfs_item_ptr_offset(leaf,
914 							    path->slots[0]);
915 
916 			memzero_extent_buffer(leaf, item_offset + offset,
917 					     shift_len);
918 			key.offset = bytenr;
919 
920 			/*
921 			 * btrfs_split_item returns -EAGAIN when the
922 			 * item changed size or key
923 			 */
924 			ret = btrfs_split_item(trans, root, path, &key, offset);
925 			if (ret && ret != -EAGAIN) {
926 				btrfs_abort_transaction(trans, ret);
927 				break;
928 			}
929 			ret = 0;
930 
931 			key.offset = end_byte - 1;
932 		} else {
933 			truncate_one_csum(fs_info, path, &key, bytenr, len);
934 			if (key.offset < bytenr)
935 				break;
936 		}
937 		btrfs_release_path(path);
938 	}
939 	btrfs_free_path(path);
940 	return ret;
941 }
942 
find_next_csum_offset(struct btrfs_root * root,struct btrfs_path * path,u64 * next_offset)943 static int find_next_csum_offset(struct btrfs_root *root,
944 				 struct btrfs_path *path,
945 				 u64 *next_offset)
946 {
947 	const u32 nritems = btrfs_header_nritems(path->nodes[0]);
948 	struct btrfs_key found_key;
949 	int slot = path->slots[0] + 1;
950 	int ret;
951 
952 	if (nritems == 0 || slot >= nritems) {
953 		ret = btrfs_next_leaf(root, path);
954 		if (ret < 0) {
955 			return ret;
956 		} else if (ret > 0) {
957 			*next_offset = (u64)-1;
958 			return 0;
959 		}
960 		slot = path->slots[0];
961 	}
962 
963 	btrfs_item_key_to_cpu(path->nodes[0], &found_key, slot);
964 
965 	if (found_key.objectid != BTRFS_EXTENT_CSUM_OBJECTID ||
966 	    found_key.type != BTRFS_EXTENT_CSUM_KEY)
967 		*next_offset = (u64)-1;
968 	else
969 		*next_offset = found_key.offset;
970 
971 	return 0;
972 }
973 
btrfs_csum_file_blocks(struct btrfs_trans_handle * trans,struct btrfs_root * root,struct btrfs_ordered_sum * sums)974 int btrfs_csum_file_blocks(struct btrfs_trans_handle *trans,
975 			   struct btrfs_root *root,
976 			   struct btrfs_ordered_sum *sums)
977 {
978 	struct btrfs_fs_info *fs_info = root->fs_info;
979 	struct btrfs_key file_key;
980 	struct btrfs_key found_key;
981 	struct btrfs_path *path;
982 	struct btrfs_csum_item *item;
983 	struct btrfs_csum_item *item_end;
984 	struct extent_buffer *leaf = NULL;
985 	u64 next_offset;
986 	u64 total_bytes = 0;
987 	u64 csum_offset;
988 	u64 bytenr;
989 	u32 ins_size;
990 	int index = 0;
991 	int found_next;
992 	int ret;
993 	const u32 csum_size = fs_info->csum_size;
994 
995 	path = btrfs_alloc_path();
996 	if (!path)
997 		return -ENOMEM;
998 again:
999 	next_offset = (u64)-1;
1000 	found_next = 0;
1001 	bytenr = sums->bytenr + total_bytes;
1002 	file_key.objectid = BTRFS_EXTENT_CSUM_OBJECTID;
1003 	file_key.offset = bytenr;
1004 	file_key.type = BTRFS_EXTENT_CSUM_KEY;
1005 
1006 	item = btrfs_lookup_csum(trans, root, path, bytenr, 1);
1007 	if (!IS_ERR(item)) {
1008 		ret = 0;
1009 		leaf = path->nodes[0];
1010 		item_end = btrfs_item_ptr(leaf, path->slots[0],
1011 					  struct btrfs_csum_item);
1012 		item_end = (struct btrfs_csum_item *)((char *)item_end +
1013 			   btrfs_item_size(leaf, path->slots[0]));
1014 		goto found;
1015 	}
1016 	ret = PTR_ERR(item);
1017 	if (ret != -EFBIG && ret != -ENOENT)
1018 		goto out;
1019 
1020 	if (ret == -EFBIG) {
1021 		u32 item_size;
1022 		/* we found one, but it isn't big enough yet */
1023 		leaf = path->nodes[0];
1024 		item_size = btrfs_item_size(leaf, path->slots[0]);
1025 		if ((item_size / csum_size) >=
1026 		    MAX_CSUM_ITEMS(fs_info, csum_size)) {
1027 			/* already at max size, make a new one */
1028 			goto insert;
1029 		}
1030 	} else {
1031 		/* We didn't find a csum item, insert one. */
1032 		ret = find_next_csum_offset(root, path, &next_offset);
1033 		if (ret < 0)
1034 			goto out;
1035 		found_next = 1;
1036 		goto insert;
1037 	}
1038 
1039 	/*
1040 	 * At this point, we know the tree has a checksum item that ends at an
1041 	 * offset matching the start of the checksum range we want to insert.
1042 	 * We try to extend that item as much as possible and then add as many
1043 	 * checksums to it as they fit.
1044 	 *
1045 	 * First check if the leaf has enough free space for at least one
1046 	 * checksum. If it has go directly to the item extension code, otherwise
1047 	 * release the path and do a search for insertion before the extension.
1048 	 */
1049 	if (btrfs_leaf_free_space(leaf) >= csum_size) {
1050 		btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
1051 		csum_offset = (bytenr - found_key.offset) >>
1052 			fs_info->sectorsize_bits;
1053 		goto extend_csum;
1054 	}
1055 
1056 	btrfs_release_path(path);
1057 	path->search_for_extension = 1;
1058 	ret = btrfs_search_slot(trans, root, &file_key, path,
1059 				csum_size, 1);
1060 	path->search_for_extension = 0;
1061 	if (ret < 0)
1062 		goto out;
1063 
1064 	if (ret > 0) {
1065 		if (path->slots[0] == 0)
1066 			goto insert;
1067 		path->slots[0]--;
1068 	}
1069 
1070 	leaf = path->nodes[0];
1071 	btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
1072 	csum_offset = (bytenr - found_key.offset) >> fs_info->sectorsize_bits;
1073 
1074 	if (found_key.type != BTRFS_EXTENT_CSUM_KEY ||
1075 	    found_key.objectid != BTRFS_EXTENT_CSUM_OBJECTID ||
1076 	    csum_offset >= MAX_CSUM_ITEMS(fs_info, csum_size)) {
1077 		goto insert;
1078 	}
1079 
1080 extend_csum:
1081 	if (csum_offset == btrfs_item_size(leaf, path->slots[0]) /
1082 	    csum_size) {
1083 		int extend_nr;
1084 		u64 tmp;
1085 		u32 diff;
1086 
1087 		tmp = sums->len - total_bytes;
1088 		tmp >>= fs_info->sectorsize_bits;
1089 		WARN_ON(tmp < 1);
1090 		extend_nr = max_t(int, 1, tmp);
1091 
1092 		/*
1093 		 * A log tree can already have checksum items with a subset of
1094 		 * the checksums we are trying to log. This can happen after
1095 		 * doing a sequence of partial writes into prealloc extents and
1096 		 * fsyncs in between, with a full fsync logging a larger subrange
1097 		 * of an extent for which a previous fast fsync logged a smaller
1098 		 * subrange. And this happens in particular due to merging file
1099 		 * extent items when we complete an ordered extent for a range
1100 		 * covered by a prealloc extent - this is done at
1101 		 * btrfs_mark_extent_written().
1102 		 *
1103 		 * So if we try to extend the previous checksum item, which has
1104 		 * a range that ends at the start of the range we want to insert,
1105 		 * make sure we don't extend beyond the start offset of the next
1106 		 * checksum item. If we are at the last item in the leaf, then
1107 		 * forget the optimization of extending and add a new checksum
1108 		 * item - it is not worth the complexity of releasing the path,
1109 		 * getting the first key for the next leaf, repeat the btree
1110 		 * search, etc, because log trees are temporary anyway and it
1111 		 * would only save a few bytes of leaf space.
1112 		 */
1113 		if (root->root_key.objectid == BTRFS_TREE_LOG_OBJECTID) {
1114 			if (path->slots[0] + 1 >=
1115 			    btrfs_header_nritems(path->nodes[0])) {
1116 				ret = find_next_csum_offset(root, path, &next_offset);
1117 				if (ret < 0)
1118 					goto out;
1119 				found_next = 1;
1120 				goto insert;
1121 			}
1122 
1123 			ret = find_next_csum_offset(root, path, &next_offset);
1124 			if (ret < 0)
1125 				goto out;
1126 
1127 			tmp = (next_offset - bytenr) >> fs_info->sectorsize_bits;
1128 			if (tmp <= INT_MAX)
1129 				extend_nr = min_t(int, extend_nr, tmp);
1130 		}
1131 
1132 		diff = (csum_offset + extend_nr) * csum_size;
1133 		diff = min(diff,
1134 			   MAX_CSUM_ITEMS(fs_info, csum_size) * csum_size);
1135 
1136 		diff = diff - btrfs_item_size(leaf, path->slots[0]);
1137 		diff = min_t(u32, btrfs_leaf_free_space(leaf), diff);
1138 		diff /= csum_size;
1139 		diff *= csum_size;
1140 
1141 		btrfs_extend_item(path, diff);
1142 		ret = 0;
1143 		goto csum;
1144 	}
1145 
1146 insert:
1147 	btrfs_release_path(path);
1148 	csum_offset = 0;
1149 	if (found_next) {
1150 		u64 tmp;
1151 
1152 		tmp = sums->len - total_bytes;
1153 		tmp >>= fs_info->sectorsize_bits;
1154 		tmp = min(tmp, (next_offset - file_key.offset) >>
1155 					 fs_info->sectorsize_bits);
1156 
1157 		tmp = max_t(u64, 1, tmp);
1158 		tmp = min_t(u64, tmp, MAX_CSUM_ITEMS(fs_info, csum_size));
1159 		ins_size = csum_size * tmp;
1160 	} else {
1161 		ins_size = csum_size;
1162 	}
1163 	ret = btrfs_insert_empty_item(trans, root, path, &file_key,
1164 				      ins_size);
1165 	if (ret < 0)
1166 		goto out;
1167 	if (WARN_ON(ret != 0))
1168 		goto out;
1169 	leaf = path->nodes[0];
1170 csum:
1171 	item = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_csum_item);
1172 	item_end = (struct btrfs_csum_item *)((unsigned char *)item +
1173 				      btrfs_item_size(leaf, path->slots[0]));
1174 	item = (struct btrfs_csum_item *)((unsigned char *)item +
1175 					  csum_offset * csum_size);
1176 found:
1177 	ins_size = (u32)(sums->len - total_bytes) >> fs_info->sectorsize_bits;
1178 	ins_size *= csum_size;
1179 	ins_size = min_t(u32, (unsigned long)item_end - (unsigned long)item,
1180 			      ins_size);
1181 	write_extent_buffer(leaf, sums->sums + index, (unsigned long)item,
1182 			    ins_size);
1183 
1184 	index += ins_size;
1185 	ins_size /= csum_size;
1186 	total_bytes += ins_size * fs_info->sectorsize;
1187 
1188 	btrfs_mark_buffer_dirty(path->nodes[0]);
1189 	if (total_bytes < sums->len) {
1190 		btrfs_release_path(path);
1191 		cond_resched();
1192 		goto again;
1193 	}
1194 out:
1195 	btrfs_free_path(path);
1196 	return ret;
1197 }
1198 
btrfs_extent_item_to_extent_map(struct btrfs_inode * inode,const struct btrfs_path * path,struct btrfs_file_extent_item * fi,const bool new_inline,struct extent_map * em)1199 void btrfs_extent_item_to_extent_map(struct btrfs_inode *inode,
1200 				     const struct btrfs_path *path,
1201 				     struct btrfs_file_extent_item *fi,
1202 				     const bool new_inline,
1203 				     struct extent_map *em)
1204 {
1205 	struct btrfs_fs_info *fs_info = inode->root->fs_info;
1206 	struct btrfs_root *root = inode->root;
1207 	struct extent_buffer *leaf = path->nodes[0];
1208 	const int slot = path->slots[0];
1209 	struct btrfs_key key;
1210 	u64 extent_start, extent_end;
1211 	u64 bytenr;
1212 	u8 type = btrfs_file_extent_type(leaf, fi);
1213 	int compress_type = btrfs_file_extent_compression(leaf, fi);
1214 
1215 	btrfs_item_key_to_cpu(leaf, &key, slot);
1216 	extent_start = key.offset;
1217 	extent_end = btrfs_file_extent_end(path);
1218 	em->ram_bytes = btrfs_file_extent_ram_bytes(leaf, fi);
1219 	em->generation = btrfs_file_extent_generation(leaf, fi);
1220 	if (type == BTRFS_FILE_EXTENT_REG ||
1221 	    type == BTRFS_FILE_EXTENT_PREALLOC) {
1222 		em->start = extent_start;
1223 		em->len = extent_end - extent_start;
1224 		em->orig_start = extent_start -
1225 			btrfs_file_extent_offset(leaf, fi);
1226 		em->orig_block_len = btrfs_file_extent_disk_num_bytes(leaf, fi);
1227 		bytenr = btrfs_file_extent_disk_bytenr(leaf, fi);
1228 		if (bytenr == 0) {
1229 			em->block_start = EXTENT_MAP_HOLE;
1230 			return;
1231 		}
1232 		if (compress_type != BTRFS_COMPRESS_NONE) {
1233 			set_bit(EXTENT_FLAG_COMPRESSED, &em->flags);
1234 			em->compress_type = compress_type;
1235 			em->block_start = bytenr;
1236 			em->block_len = em->orig_block_len;
1237 		} else {
1238 			bytenr += btrfs_file_extent_offset(leaf, fi);
1239 			em->block_start = bytenr;
1240 			em->block_len = em->len;
1241 			if (type == BTRFS_FILE_EXTENT_PREALLOC)
1242 				set_bit(EXTENT_FLAG_PREALLOC, &em->flags);
1243 		}
1244 	} else if (type == BTRFS_FILE_EXTENT_INLINE) {
1245 		em->block_start = EXTENT_MAP_INLINE;
1246 		em->start = extent_start;
1247 		em->len = extent_end - extent_start;
1248 		/*
1249 		 * Initialize orig_start and block_len with the same values
1250 		 * as in inode.c:btrfs_get_extent().
1251 		 */
1252 		em->orig_start = EXTENT_MAP_HOLE;
1253 		em->block_len = (u64)-1;
1254 		if (!new_inline && compress_type != BTRFS_COMPRESS_NONE) {
1255 			set_bit(EXTENT_FLAG_COMPRESSED, &em->flags);
1256 			em->compress_type = compress_type;
1257 		}
1258 	} else {
1259 		btrfs_err(fs_info,
1260 			  "unknown file extent item type %d, inode %llu, offset %llu, "
1261 			  "root %llu", type, btrfs_ino(inode), extent_start,
1262 			  root->root_key.objectid);
1263 	}
1264 }
1265 
1266 /*
1267  * Returns the end offset (non inclusive) of the file extent item the given path
1268  * points to. If it points to an inline extent, the returned offset is rounded
1269  * up to the sector size.
1270  */
btrfs_file_extent_end(const struct btrfs_path * path)1271 u64 btrfs_file_extent_end(const struct btrfs_path *path)
1272 {
1273 	const struct extent_buffer *leaf = path->nodes[0];
1274 	const int slot = path->slots[0];
1275 	struct btrfs_file_extent_item *fi;
1276 	struct btrfs_key key;
1277 	u64 end;
1278 
1279 	btrfs_item_key_to_cpu(leaf, &key, slot);
1280 	ASSERT(key.type == BTRFS_EXTENT_DATA_KEY);
1281 	fi = btrfs_item_ptr(leaf, slot, struct btrfs_file_extent_item);
1282 
1283 	if (btrfs_file_extent_type(leaf, fi) == BTRFS_FILE_EXTENT_INLINE) {
1284 		end = btrfs_file_extent_ram_bytes(leaf, fi);
1285 		end = ALIGN(key.offset + end, leaf->fs_info->sectorsize);
1286 	} else {
1287 		end = key.offset + btrfs_file_extent_num_bytes(leaf, fi);
1288 	}
1289 
1290 	return end;
1291 }
1292