xref: /openbmc/linux/fs/btrfs/file-item.c (revision 9fb29c73)
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 "ctree.h"
11 #include "disk-io.h"
12 #include "transaction.h"
13 #include "volumes.h"
14 #include "print-tree.h"
15 #include "compression.h"
16 
17 #define __MAX_CSUM_ITEMS(r, size) ((unsigned long)(((BTRFS_LEAF_DATA_SIZE(r) - \
18 				   sizeof(struct btrfs_item) * 2) / \
19 				  size) - 1))
20 
21 #define MAX_CSUM_ITEMS(r, size) (min_t(u32, __MAX_CSUM_ITEMS(r, size), \
22 				       PAGE_SIZE))
23 
24 #define MAX_ORDERED_SUM_BYTES(fs_info) ((PAGE_SIZE - \
25 				   sizeof(struct btrfs_ordered_sum)) / \
26 				   sizeof(u32) * (fs_info)->sectorsize)
27 
28 int btrfs_insert_file_extent(struct btrfs_trans_handle *trans,
29 			     struct btrfs_root *root,
30 			     u64 objectid, u64 pos,
31 			     u64 disk_offset, u64 disk_num_bytes,
32 			     u64 num_bytes, u64 offset, u64 ram_bytes,
33 			     u8 compression, u8 encryption, u16 other_encoding)
34 {
35 	int ret = 0;
36 	struct btrfs_file_extent_item *item;
37 	struct btrfs_key file_key;
38 	struct btrfs_path *path;
39 	struct extent_buffer *leaf;
40 
41 	path = btrfs_alloc_path();
42 	if (!path)
43 		return -ENOMEM;
44 	file_key.objectid = objectid;
45 	file_key.offset = pos;
46 	file_key.type = BTRFS_EXTENT_DATA_KEY;
47 
48 	path->leave_spinning = 1;
49 	ret = btrfs_insert_empty_item(trans, root, path, &file_key,
50 				      sizeof(*item));
51 	if (ret < 0)
52 		goto out;
53 	BUG_ON(ret); /* Can't happen */
54 	leaf = path->nodes[0];
55 	item = btrfs_item_ptr(leaf, path->slots[0],
56 			      struct btrfs_file_extent_item);
57 	btrfs_set_file_extent_disk_bytenr(leaf, item, disk_offset);
58 	btrfs_set_file_extent_disk_num_bytes(leaf, item, disk_num_bytes);
59 	btrfs_set_file_extent_offset(leaf, item, offset);
60 	btrfs_set_file_extent_num_bytes(leaf, item, num_bytes);
61 	btrfs_set_file_extent_ram_bytes(leaf, item, ram_bytes);
62 	btrfs_set_file_extent_generation(leaf, item, trans->transid);
63 	btrfs_set_file_extent_type(leaf, item, BTRFS_FILE_EXTENT_REG);
64 	btrfs_set_file_extent_compression(leaf, item, compression);
65 	btrfs_set_file_extent_encryption(leaf, item, encryption);
66 	btrfs_set_file_extent_other_encoding(leaf, item, other_encoding);
67 
68 	btrfs_mark_buffer_dirty(leaf);
69 out:
70 	btrfs_free_path(path);
71 	return ret;
72 }
73 
74 static struct btrfs_csum_item *
75 btrfs_lookup_csum(struct btrfs_trans_handle *trans,
76 		  struct btrfs_root *root,
77 		  struct btrfs_path *path,
78 		  u64 bytenr, int cow)
79 {
80 	struct btrfs_fs_info *fs_info = root->fs_info;
81 	int ret;
82 	struct btrfs_key file_key;
83 	struct btrfs_key found_key;
84 	struct btrfs_csum_item *item;
85 	struct extent_buffer *leaf;
86 	u64 csum_offset = 0;
87 	u16 csum_size = btrfs_super_csum_size(fs_info->super_copy);
88 	int csums_in_item;
89 
90 	file_key.objectid = BTRFS_EXTENT_CSUM_OBJECTID;
91 	file_key.offset = bytenr;
92 	file_key.type = BTRFS_EXTENT_CSUM_KEY;
93 	ret = btrfs_search_slot(trans, root, &file_key, path, 0, cow);
94 	if (ret < 0)
95 		goto fail;
96 	leaf = path->nodes[0];
97 	if (ret > 0) {
98 		ret = 1;
99 		if (path->slots[0] == 0)
100 			goto fail;
101 		path->slots[0]--;
102 		btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
103 		if (found_key.type != BTRFS_EXTENT_CSUM_KEY)
104 			goto fail;
105 
106 		csum_offset = (bytenr - found_key.offset) >>
107 				fs_info->sb->s_blocksize_bits;
108 		csums_in_item = btrfs_item_size_nr(leaf, path->slots[0]);
109 		csums_in_item /= csum_size;
110 
111 		if (csum_offset == csums_in_item) {
112 			ret = -EFBIG;
113 			goto fail;
114 		} else if (csum_offset > csums_in_item) {
115 			goto fail;
116 		}
117 	}
118 	item = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_csum_item);
119 	item = (struct btrfs_csum_item *)((unsigned char *)item +
120 					  csum_offset * csum_size);
121 	return item;
122 fail:
123 	if (ret > 0)
124 		ret = -ENOENT;
125 	return ERR_PTR(ret);
126 }
127 
128 int btrfs_lookup_file_extent(struct btrfs_trans_handle *trans,
129 			     struct btrfs_root *root,
130 			     struct btrfs_path *path, u64 objectid,
131 			     u64 offset, int mod)
132 {
133 	int ret;
134 	struct btrfs_key file_key;
135 	int ins_len = mod < 0 ? -1 : 0;
136 	int cow = mod != 0;
137 
138 	file_key.objectid = objectid;
139 	file_key.offset = offset;
140 	file_key.type = BTRFS_EXTENT_DATA_KEY;
141 	ret = btrfs_search_slot(trans, root, &file_key, path, ins_len, cow);
142 	return ret;
143 }
144 
145 static blk_status_t __btrfs_lookup_bio_sums(struct inode *inode, struct bio *bio,
146 				   u64 logical_offset, u32 *dst, int dio)
147 {
148 	struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
149 	struct bio_vec bvec;
150 	struct bvec_iter iter;
151 	struct btrfs_io_bio *btrfs_bio = btrfs_io_bio(bio);
152 	struct btrfs_csum_item *item = NULL;
153 	struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
154 	struct btrfs_path *path;
155 	u8 *csum;
156 	u64 offset = 0;
157 	u64 item_start_offset = 0;
158 	u64 item_last_offset = 0;
159 	u64 disk_bytenr;
160 	u64 page_bytes_left;
161 	u32 diff;
162 	int nblocks;
163 	int count = 0;
164 	u16 csum_size = btrfs_super_csum_size(fs_info->super_copy);
165 
166 	path = btrfs_alloc_path();
167 	if (!path)
168 		return BLK_STS_RESOURCE;
169 
170 	nblocks = bio->bi_iter.bi_size >> inode->i_sb->s_blocksize_bits;
171 	if (!dst) {
172 		if (nblocks * csum_size > BTRFS_BIO_INLINE_CSUM_SIZE) {
173 			btrfs_bio->csum = kmalloc_array(nblocks, csum_size,
174 							GFP_NOFS);
175 			if (!btrfs_bio->csum) {
176 				btrfs_free_path(path);
177 				return BLK_STS_RESOURCE;
178 			}
179 		} else {
180 			btrfs_bio->csum = btrfs_bio->csum_inline;
181 		}
182 		csum = btrfs_bio->csum;
183 	} else {
184 		csum = (u8 *)dst;
185 	}
186 
187 	if (bio->bi_iter.bi_size > PAGE_SIZE * 8)
188 		path->reada = READA_FORWARD;
189 
190 	/*
191 	 * the free space stuff is only read when it hasn't been
192 	 * updated in the current transaction.  So, we can safely
193 	 * read from the commit root and sidestep a nasty deadlock
194 	 * between reading the free space cache and updating the csum tree.
195 	 */
196 	if (btrfs_is_free_space_inode(BTRFS_I(inode))) {
197 		path->search_commit_root = 1;
198 		path->skip_locking = 1;
199 	}
200 
201 	disk_bytenr = (u64)bio->bi_iter.bi_sector << 9;
202 	if (dio)
203 		offset = logical_offset;
204 
205 	bio_for_each_segment(bvec, bio, iter) {
206 		page_bytes_left = bvec.bv_len;
207 		if (count)
208 			goto next;
209 
210 		if (!dio)
211 			offset = page_offset(bvec.bv_page) + bvec.bv_offset;
212 		count = btrfs_find_ordered_sum(inode, offset, disk_bytenr,
213 					       (u32 *)csum, nblocks);
214 		if (count)
215 			goto found;
216 
217 		if (!item || disk_bytenr < item_start_offset ||
218 		    disk_bytenr >= item_last_offset) {
219 			struct btrfs_key found_key;
220 			u32 item_size;
221 
222 			if (item)
223 				btrfs_release_path(path);
224 			item = btrfs_lookup_csum(NULL, fs_info->csum_root,
225 						 path, disk_bytenr, 0);
226 			if (IS_ERR(item)) {
227 				count = 1;
228 				memset(csum, 0, csum_size);
229 				if (BTRFS_I(inode)->root->root_key.objectid ==
230 				    BTRFS_DATA_RELOC_TREE_OBJECTID) {
231 					set_extent_bits(io_tree, offset,
232 						offset + fs_info->sectorsize - 1,
233 						EXTENT_NODATASUM);
234 				} else {
235 					btrfs_info_rl(fs_info,
236 						   "no csum found for inode %llu start %llu",
237 					       btrfs_ino(BTRFS_I(inode)), offset);
238 				}
239 				item = NULL;
240 				btrfs_release_path(path);
241 				goto found;
242 			}
243 			btrfs_item_key_to_cpu(path->nodes[0], &found_key,
244 					      path->slots[0]);
245 
246 			item_start_offset = found_key.offset;
247 			item_size = btrfs_item_size_nr(path->nodes[0],
248 						       path->slots[0]);
249 			item_last_offset = item_start_offset +
250 				(item_size / csum_size) *
251 				fs_info->sectorsize;
252 			item = btrfs_item_ptr(path->nodes[0], path->slots[0],
253 					      struct btrfs_csum_item);
254 		}
255 		/*
256 		 * this byte range must be able to fit inside
257 		 * a single leaf so it will also fit inside a u32
258 		 */
259 		diff = disk_bytenr - item_start_offset;
260 		diff = diff / fs_info->sectorsize;
261 		diff = diff * csum_size;
262 		count = min_t(int, nblocks, (item_last_offset - disk_bytenr) >>
263 					    inode->i_sb->s_blocksize_bits);
264 		read_extent_buffer(path->nodes[0], csum,
265 				   ((unsigned long)item) + diff,
266 				   csum_size * count);
267 found:
268 		csum += count * csum_size;
269 		nblocks -= count;
270 next:
271 		while (count--) {
272 			disk_bytenr += fs_info->sectorsize;
273 			offset += fs_info->sectorsize;
274 			page_bytes_left -= fs_info->sectorsize;
275 			if (!page_bytes_left)
276 				break; /* move to next bio */
277 		}
278 	}
279 
280 	WARN_ON_ONCE(count);
281 	btrfs_free_path(path);
282 	return 0;
283 }
284 
285 blk_status_t btrfs_lookup_bio_sums(struct inode *inode, struct bio *bio, u32 *dst)
286 {
287 	return __btrfs_lookup_bio_sums(inode, bio, 0, dst, 0);
288 }
289 
290 blk_status_t btrfs_lookup_bio_sums_dio(struct inode *inode, struct bio *bio, u64 offset)
291 {
292 	return __btrfs_lookup_bio_sums(inode, bio, offset, NULL, 1);
293 }
294 
295 int btrfs_lookup_csums_range(struct btrfs_root *root, u64 start, u64 end,
296 			     struct list_head *list, int search_commit)
297 {
298 	struct btrfs_fs_info *fs_info = root->fs_info;
299 	struct btrfs_key key;
300 	struct btrfs_path *path;
301 	struct extent_buffer *leaf;
302 	struct btrfs_ordered_sum *sums;
303 	struct btrfs_csum_item *item;
304 	LIST_HEAD(tmplist);
305 	unsigned long offset;
306 	int ret;
307 	size_t size;
308 	u64 csum_end;
309 	u16 csum_size = btrfs_super_csum_size(fs_info->super_copy);
310 
311 	ASSERT(IS_ALIGNED(start, fs_info->sectorsize) &&
312 	       IS_ALIGNED(end + 1, fs_info->sectorsize));
313 
314 	path = btrfs_alloc_path();
315 	if (!path)
316 		return -ENOMEM;
317 
318 	if (search_commit) {
319 		path->skip_locking = 1;
320 		path->reada = READA_FORWARD;
321 		path->search_commit_root = 1;
322 	}
323 
324 	key.objectid = BTRFS_EXTENT_CSUM_OBJECTID;
325 	key.offset = start;
326 	key.type = BTRFS_EXTENT_CSUM_KEY;
327 
328 	ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
329 	if (ret < 0)
330 		goto fail;
331 	if (ret > 0 && path->slots[0] > 0) {
332 		leaf = path->nodes[0];
333 		btrfs_item_key_to_cpu(leaf, &key, path->slots[0] - 1);
334 		if (key.objectid == BTRFS_EXTENT_CSUM_OBJECTID &&
335 		    key.type == BTRFS_EXTENT_CSUM_KEY) {
336 			offset = (start - key.offset) >>
337 				 fs_info->sb->s_blocksize_bits;
338 			if (offset * csum_size <
339 			    btrfs_item_size_nr(leaf, path->slots[0] - 1))
340 				path->slots[0]--;
341 		}
342 	}
343 
344 	while (start <= end) {
345 		leaf = path->nodes[0];
346 		if (path->slots[0] >= btrfs_header_nritems(leaf)) {
347 			ret = btrfs_next_leaf(root, path);
348 			if (ret < 0)
349 				goto fail;
350 			if (ret > 0)
351 				break;
352 			leaf = path->nodes[0];
353 		}
354 
355 		btrfs_item_key_to_cpu(leaf, &key, path->slots[0]);
356 		if (key.objectid != BTRFS_EXTENT_CSUM_OBJECTID ||
357 		    key.type != BTRFS_EXTENT_CSUM_KEY ||
358 		    key.offset > end)
359 			break;
360 
361 		if (key.offset > start)
362 			start = key.offset;
363 
364 		size = btrfs_item_size_nr(leaf, path->slots[0]);
365 		csum_end = key.offset + (size / csum_size) * fs_info->sectorsize;
366 		if (csum_end <= start) {
367 			path->slots[0]++;
368 			continue;
369 		}
370 
371 		csum_end = min(csum_end, end + 1);
372 		item = btrfs_item_ptr(path->nodes[0], path->slots[0],
373 				      struct btrfs_csum_item);
374 		while (start < csum_end) {
375 			size = min_t(size_t, csum_end - start,
376 				     MAX_ORDERED_SUM_BYTES(fs_info));
377 			sums = kzalloc(btrfs_ordered_sum_size(fs_info, size),
378 				       GFP_NOFS);
379 			if (!sums) {
380 				ret = -ENOMEM;
381 				goto fail;
382 			}
383 
384 			sums->bytenr = start;
385 			sums->len = (int)size;
386 
387 			offset = (start - key.offset) >>
388 				fs_info->sb->s_blocksize_bits;
389 			offset *= csum_size;
390 			size >>= fs_info->sb->s_blocksize_bits;
391 
392 			read_extent_buffer(path->nodes[0],
393 					   sums->sums,
394 					   ((unsigned long)item) + offset,
395 					   csum_size * size);
396 
397 			start += fs_info->sectorsize * size;
398 			list_add_tail(&sums->list, &tmplist);
399 		}
400 		path->slots[0]++;
401 	}
402 	ret = 0;
403 fail:
404 	while (ret < 0 && !list_empty(&tmplist)) {
405 		sums = list_entry(tmplist.next, struct btrfs_ordered_sum, list);
406 		list_del(&sums->list);
407 		kfree(sums);
408 	}
409 	list_splice_tail(&tmplist, list);
410 
411 	btrfs_free_path(path);
412 	return ret;
413 }
414 
415 blk_status_t btrfs_csum_one_bio(struct inode *inode, struct bio *bio,
416 		       u64 file_start, int contig)
417 {
418 	struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
419 	struct btrfs_ordered_sum *sums;
420 	struct btrfs_ordered_extent *ordered = NULL;
421 	char *data;
422 	struct bvec_iter iter;
423 	struct bio_vec bvec;
424 	int index;
425 	int nr_sectors;
426 	unsigned long total_bytes = 0;
427 	unsigned long this_sum_bytes = 0;
428 	int i;
429 	u64 offset;
430 
431 	sums = kzalloc(btrfs_ordered_sum_size(fs_info, bio->bi_iter.bi_size),
432 		       GFP_NOFS);
433 	if (!sums)
434 		return BLK_STS_RESOURCE;
435 
436 	sums->len = bio->bi_iter.bi_size;
437 	INIT_LIST_HEAD(&sums->list);
438 
439 	if (contig)
440 		offset = file_start;
441 	else
442 		offset = 0; /* shut up gcc */
443 
444 	sums->bytenr = (u64)bio->bi_iter.bi_sector << 9;
445 	index = 0;
446 
447 	bio_for_each_segment(bvec, bio, iter) {
448 		if (!contig)
449 			offset = page_offset(bvec.bv_page) + bvec.bv_offset;
450 
451 		if (!ordered) {
452 			ordered = btrfs_lookup_ordered_extent(inode, offset);
453 			BUG_ON(!ordered); /* Logic error */
454 		}
455 
456 		data = kmap_atomic(bvec.bv_page);
457 
458 		nr_sectors = BTRFS_BYTES_TO_BLKS(fs_info,
459 						 bvec.bv_len + fs_info->sectorsize
460 						 - 1);
461 
462 		for (i = 0; i < nr_sectors; i++) {
463 			if (offset >= ordered->file_offset + ordered->len ||
464 				offset < ordered->file_offset) {
465 				unsigned long bytes_left;
466 
467 				kunmap_atomic(data);
468 				sums->len = this_sum_bytes;
469 				this_sum_bytes = 0;
470 				btrfs_add_ordered_sum(inode, ordered, sums);
471 				btrfs_put_ordered_extent(ordered);
472 
473 				bytes_left = bio->bi_iter.bi_size - total_bytes;
474 
475 				sums = kzalloc(btrfs_ordered_sum_size(fs_info, bytes_left),
476 					       GFP_NOFS);
477 				BUG_ON(!sums); /* -ENOMEM */
478 				sums->len = bytes_left;
479 				ordered = btrfs_lookup_ordered_extent(inode,
480 								offset);
481 				ASSERT(ordered); /* Logic error */
482 				sums->bytenr = ((u64)bio->bi_iter.bi_sector << 9)
483 					+ total_bytes;
484 				index = 0;
485 
486 				data = kmap_atomic(bvec.bv_page);
487 			}
488 
489 			sums->sums[index] = ~(u32)0;
490 			sums->sums[index]
491 				= btrfs_csum_data(data + bvec.bv_offset
492 						+ (i * fs_info->sectorsize),
493 						sums->sums[index],
494 						fs_info->sectorsize);
495 			btrfs_csum_final(sums->sums[index],
496 					(char *)(sums->sums + index));
497 			index++;
498 			offset += fs_info->sectorsize;
499 			this_sum_bytes += fs_info->sectorsize;
500 			total_bytes += fs_info->sectorsize;
501 		}
502 
503 		kunmap_atomic(data);
504 	}
505 	this_sum_bytes = 0;
506 	btrfs_add_ordered_sum(inode, ordered, sums);
507 	btrfs_put_ordered_extent(ordered);
508 	return 0;
509 }
510 
511 /*
512  * helper function for csum removal, this expects the
513  * key to describe the csum pointed to by the path, and it expects
514  * the csum to overlap the range [bytenr, len]
515  *
516  * The csum should not be entirely contained in the range and the
517  * range should not be entirely contained in the csum.
518  *
519  * This calls btrfs_truncate_item with the correct args based on the
520  * overlap, and fixes up the key as required.
521  */
522 static noinline void truncate_one_csum(struct btrfs_fs_info *fs_info,
523 				       struct btrfs_path *path,
524 				       struct btrfs_key *key,
525 				       u64 bytenr, u64 len)
526 {
527 	struct extent_buffer *leaf;
528 	u16 csum_size = btrfs_super_csum_size(fs_info->super_copy);
529 	u64 csum_end;
530 	u64 end_byte = bytenr + len;
531 	u32 blocksize_bits = fs_info->sb->s_blocksize_bits;
532 
533 	leaf = path->nodes[0];
534 	csum_end = btrfs_item_size_nr(leaf, path->slots[0]) / csum_size;
535 	csum_end <<= fs_info->sb->s_blocksize_bits;
536 	csum_end += key->offset;
537 
538 	if (key->offset < bytenr && csum_end <= end_byte) {
539 		/*
540 		 *         [ bytenr - len ]
541 		 *         [   ]
542 		 *   [csum     ]
543 		 *   A simple truncate off the end of the item
544 		 */
545 		u32 new_size = (bytenr - key->offset) >> blocksize_bits;
546 		new_size *= csum_size;
547 		btrfs_truncate_item(fs_info, path, new_size, 1);
548 	} else if (key->offset >= bytenr && csum_end > end_byte &&
549 		   end_byte > key->offset) {
550 		/*
551 		 *         [ bytenr - len ]
552 		 *                 [ ]
553 		 *                 [csum     ]
554 		 * we need to truncate from the beginning of the csum
555 		 */
556 		u32 new_size = (csum_end - end_byte) >> blocksize_bits;
557 		new_size *= csum_size;
558 
559 		btrfs_truncate_item(fs_info, path, new_size, 0);
560 
561 		key->offset = end_byte;
562 		btrfs_set_item_key_safe(fs_info, path, key);
563 	} else {
564 		BUG();
565 	}
566 }
567 
568 /*
569  * deletes the csum items from the csum tree for a given
570  * range of bytes.
571  */
572 int btrfs_del_csums(struct btrfs_trans_handle *trans,
573 		    struct btrfs_fs_info *fs_info, u64 bytenr, u64 len)
574 {
575 	struct btrfs_root *root = fs_info->csum_root;
576 	struct btrfs_path *path;
577 	struct btrfs_key key;
578 	u64 end_byte = bytenr + len;
579 	u64 csum_end;
580 	struct extent_buffer *leaf;
581 	int ret;
582 	u16 csum_size = btrfs_super_csum_size(fs_info->super_copy);
583 	int blocksize_bits = fs_info->sb->s_blocksize_bits;
584 
585 	path = btrfs_alloc_path();
586 	if (!path)
587 		return -ENOMEM;
588 
589 	while (1) {
590 		key.objectid = BTRFS_EXTENT_CSUM_OBJECTID;
591 		key.offset = end_byte - 1;
592 		key.type = BTRFS_EXTENT_CSUM_KEY;
593 
594 		path->leave_spinning = 1;
595 		ret = btrfs_search_slot(trans, root, &key, path, -1, 1);
596 		if (ret > 0) {
597 			if (path->slots[0] == 0)
598 				break;
599 			path->slots[0]--;
600 		} else if (ret < 0) {
601 			break;
602 		}
603 
604 		leaf = path->nodes[0];
605 		btrfs_item_key_to_cpu(leaf, &key, path->slots[0]);
606 
607 		if (key.objectid != BTRFS_EXTENT_CSUM_OBJECTID ||
608 		    key.type != BTRFS_EXTENT_CSUM_KEY) {
609 			break;
610 		}
611 
612 		if (key.offset >= end_byte)
613 			break;
614 
615 		csum_end = btrfs_item_size_nr(leaf, path->slots[0]) / csum_size;
616 		csum_end <<= blocksize_bits;
617 		csum_end += key.offset;
618 
619 		/* this csum ends before we start, we're done */
620 		if (csum_end <= bytenr)
621 			break;
622 
623 		/* delete the entire item, it is inside our range */
624 		if (key.offset >= bytenr && csum_end <= end_byte) {
625 			int del_nr = 1;
626 
627 			/*
628 			 * Check how many csum items preceding this one in this
629 			 * leaf correspond to our range and then delete them all
630 			 * at once.
631 			 */
632 			if (key.offset > bytenr && path->slots[0] > 0) {
633 				int slot = path->slots[0] - 1;
634 
635 				while (slot >= 0) {
636 					struct btrfs_key pk;
637 
638 					btrfs_item_key_to_cpu(leaf, &pk, slot);
639 					if (pk.offset < bytenr ||
640 					    pk.type != BTRFS_EXTENT_CSUM_KEY ||
641 					    pk.objectid !=
642 					    BTRFS_EXTENT_CSUM_OBJECTID)
643 						break;
644 					path->slots[0] = slot;
645 					del_nr++;
646 					key.offset = pk.offset;
647 					slot--;
648 				}
649 			}
650 			ret = btrfs_del_items(trans, root, path,
651 					      path->slots[0], del_nr);
652 			if (ret)
653 				goto out;
654 			if (key.offset == bytenr)
655 				break;
656 		} else if (key.offset < bytenr && csum_end > end_byte) {
657 			unsigned long offset;
658 			unsigned long shift_len;
659 			unsigned long item_offset;
660 			/*
661 			 *        [ bytenr - len ]
662 			 *     [csum                ]
663 			 *
664 			 * Our bytes are in the middle of the csum,
665 			 * we need to split this item and insert a new one.
666 			 *
667 			 * But we can't drop the path because the
668 			 * csum could change, get removed, extended etc.
669 			 *
670 			 * The trick here is the max size of a csum item leaves
671 			 * enough room in the tree block for a single
672 			 * item header.  So, we split the item in place,
673 			 * adding a new header pointing to the existing
674 			 * bytes.  Then we loop around again and we have
675 			 * a nicely formed csum item that we can neatly
676 			 * truncate.
677 			 */
678 			offset = (bytenr - key.offset) >> blocksize_bits;
679 			offset *= csum_size;
680 
681 			shift_len = (len >> blocksize_bits) * csum_size;
682 
683 			item_offset = btrfs_item_ptr_offset(leaf,
684 							    path->slots[0]);
685 
686 			memzero_extent_buffer(leaf, item_offset + offset,
687 					     shift_len);
688 			key.offset = bytenr;
689 
690 			/*
691 			 * btrfs_split_item returns -EAGAIN when the
692 			 * item changed size or key
693 			 */
694 			ret = btrfs_split_item(trans, root, path, &key, offset);
695 			if (ret && ret != -EAGAIN) {
696 				btrfs_abort_transaction(trans, ret);
697 				goto out;
698 			}
699 
700 			key.offset = end_byte - 1;
701 		} else {
702 			truncate_one_csum(fs_info, path, &key, bytenr, len);
703 			if (key.offset < bytenr)
704 				break;
705 		}
706 		btrfs_release_path(path);
707 	}
708 	ret = 0;
709 out:
710 	btrfs_free_path(path);
711 	return ret;
712 }
713 
714 int btrfs_csum_file_blocks(struct btrfs_trans_handle *trans,
715 			   struct btrfs_root *root,
716 			   struct btrfs_ordered_sum *sums)
717 {
718 	struct btrfs_fs_info *fs_info = root->fs_info;
719 	struct btrfs_key file_key;
720 	struct btrfs_key found_key;
721 	struct btrfs_path *path;
722 	struct btrfs_csum_item *item;
723 	struct btrfs_csum_item *item_end;
724 	struct extent_buffer *leaf = NULL;
725 	u64 next_offset;
726 	u64 total_bytes = 0;
727 	u64 csum_offset;
728 	u64 bytenr;
729 	u32 nritems;
730 	u32 ins_size;
731 	int index = 0;
732 	int found_next;
733 	int ret;
734 	u16 csum_size = btrfs_super_csum_size(fs_info->super_copy);
735 
736 	path = btrfs_alloc_path();
737 	if (!path)
738 		return -ENOMEM;
739 again:
740 	next_offset = (u64)-1;
741 	found_next = 0;
742 	bytenr = sums->bytenr + total_bytes;
743 	file_key.objectid = BTRFS_EXTENT_CSUM_OBJECTID;
744 	file_key.offset = bytenr;
745 	file_key.type = BTRFS_EXTENT_CSUM_KEY;
746 
747 	item = btrfs_lookup_csum(trans, root, path, bytenr, 1);
748 	if (!IS_ERR(item)) {
749 		ret = 0;
750 		leaf = path->nodes[0];
751 		item_end = btrfs_item_ptr(leaf, path->slots[0],
752 					  struct btrfs_csum_item);
753 		item_end = (struct btrfs_csum_item *)((char *)item_end +
754 			   btrfs_item_size_nr(leaf, path->slots[0]));
755 		goto found;
756 	}
757 	ret = PTR_ERR(item);
758 	if (ret != -EFBIG && ret != -ENOENT)
759 		goto fail_unlock;
760 
761 	if (ret == -EFBIG) {
762 		u32 item_size;
763 		/* we found one, but it isn't big enough yet */
764 		leaf = path->nodes[0];
765 		item_size = btrfs_item_size_nr(leaf, path->slots[0]);
766 		if ((item_size / csum_size) >=
767 		    MAX_CSUM_ITEMS(fs_info, csum_size)) {
768 			/* already at max size, make a new one */
769 			goto insert;
770 		}
771 	} else {
772 		int slot = path->slots[0] + 1;
773 		/* we didn't find a csum item, insert one */
774 		nritems = btrfs_header_nritems(path->nodes[0]);
775 		if (!nritems || (path->slots[0] >= nritems - 1)) {
776 			ret = btrfs_next_leaf(root, path);
777 			if (ret == 1)
778 				found_next = 1;
779 			if (ret != 0)
780 				goto insert;
781 			slot = path->slots[0];
782 		}
783 		btrfs_item_key_to_cpu(path->nodes[0], &found_key, slot);
784 		if (found_key.objectid != BTRFS_EXTENT_CSUM_OBJECTID ||
785 		    found_key.type != BTRFS_EXTENT_CSUM_KEY) {
786 			found_next = 1;
787 			goto insert;
788 		}
789 		next_offset = found_key.offset;
790 		found_next = 1;
791 		goto insert;
792 	}
793 
794 	/*
795 	 * at this point, we know the tree has an item, but it isn't big
796 	 * enough yet to put our csum in.  Grow it
797 	 */
798 	btrfs_release_path(path);
799 	ret = btrfs_search_slot(trans, root, &file_key, path,
800 				csum_size, 1);
801 	if (ret < 0)
802 		goto fail_unlock;
803 
804 	if (ret > 0) {
805 		if (path->slots[0] == 0)
806 			goto insert;
807 		path->slots[0]--;
808 	}
809 
810 	leaf = path->nodes[0];
811 	btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
812 	csum_offset = (bytenr - found_key.offset) >>
813 			fs_info->sb->s_blocksize_bits;
814 
815 	if (found_key.type != BTRFS_EXTENT_CSUM_KEY ||
816 	    found_key.objectid != BTRFS_EXTENT_CSUM_OBJECTID ||
817 	    csum_offset >= MAX_CSUM_ITEMS(fs_info, csum_size)) {
818 		goto insert;
819 	}
820 
821 	if (csum_offset == btrfs_item_size_nr(leaf, path->slots[0]) /
822 	    csum_size) {
823 		int extend_nr;
824 		u64 tmp;
825 		u32 diff;
826 		u32 free_space;
827 
828 		if (btrfs_leaf_free_space(fs_info, leaf) <
829 				 sizeof(struct btrfs_item) + csum_size * 2)
830 			goto insert;
831 
832 		free_space = btrfs_leaf_free_space(fs_info, leaf) -
833 					 sizeof(struct btrfs_item) - csum_size;
834 		tmp = sums->len - total_bytes;
835 		tmp >>= fs_info->sb->s_blocksize_bits;
836 		WARN_ON(tmp < 1);
837 
838 		extend_nr = max_t(int, 1, (int)tmp);
839 		diff = (csum_offset + extend_nr) * csum_size;
840 		diff = min(diff,
841 			   MAX_CSUM_ITEMS(fs_info, csum_size) * csum_size);
842 
843 		diff = diff - btrfs_item_size_nr(leaf, path->slots[0]);
844 		diff = min(free_space, diff);
845 		diff /= csum_size;
846 		diff *= csum_size;
847 
848 		btrfs_extend_item(fs_info, path, diff);
849 		ret = 0;
850 		goto csum;
851 	}
852 
853 insert:
854 	btrfs_release_path(path);
855 	csum_offset = 0;
856 	if (found_next) {
857 		u64 tmp;
858 
859 		tmp = sums->len - total_bytes;
860 		tmp >>= fs_info->sb->s_blocksize_bits;
861 		tmp = min(tmp, (next_offset - file_key.offset) >>
862 					 fs_info->sb->s_blocksize_bits);
863 
864 		tmp = max_t(u64, 1, tmp);
865 		tmp = min_t(u64, tmp, MAX_CSUM_ITEMS(fs_info, csum_size));
866 		ins_size = csum_size * tmp;
867 	} else {
868 		ins_size = csum_size;
869 	}
870 	path->leave_spinning = 1;
871 	ret = btrfs_insert_empty_item(trans, root, path, &file_key,
872 				      ins_size);
873 	path->leave_spinning = 0;
874 	if (ret < 0)
875 		goto fail_unlock;
876 	if (WARN_ON(ret != 0))
877 		goto fail_unlock;
878 	leaf = path->nodes[0];
879 csum:
880 	item = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_csum_item);
881 	item_end = (struct btrfs_csum_item *)((unsigned char *)item +
882 				      btrfs_item_size_nr(leaf, path->slots[0]));
883 	item = (struct btrfs_csum_item *)((unsigned char *)item +
884 					  csum_offset * csum_size);
885 found:
886 	ins_size = (u32)(sums->len - total_bytes) >>
887 		   fs_info->sb->s_blocksize_bits;
888 	ins_size *= csum_size;
889 	ins_size = min_t(u32, (unsigned long)item_end - (unsigned long)item,
890 			      ins_size);
891 	write_extent_buffer(leaf, sums->sums + index, (unsigned long)item,
892 			    ins_size);
893 
894 	ins_size /= csum_size;
895 	total_bytes += ins_size * fs_info->sectorsize;
896 	index += ins_size;
897 
898 	btrfs_mark_buffer_dirty(path->nodes[0]);
899 	if (total_bytes < sums->len) {
900 		btrfs_release_path(path);
901 		cond_resched();
902 		goto again;
903 	}
904 out:
905 	btrfs_free_path(path);
906 	return ret;
907 
908 fail_unlock:
909 	goto out;
910 }
911 
912 void btrfs_extent_item_to_extent_map(struct btrfs_inode *inode,
913 				     const struct btrfs_path *path,
914 				     struct btrfs_file_extent_item *fi,
915 				     const bool new_inline,
916 				     struct extent_map *em)
917 {
918 	struct btrfs_fs_info *fs_info = inode->root->fs_info;
919 	struct btrfs_root *root = inode->root;
920 	struct extent_buffer *leaf = path->nodes[0];
921 	const int slot = path->slots[0];
922 	struct btrfs_key key;
923 	u64 extent_start, extent_end;
924 	u64 bytenr;
925 	u8 type = btrfs_file_extent_type(leaf, fi);
926 	int compress_type = btrfs_file_extent_compression(leaf, fi);
927 
928 	em->bdev = fs_info->fs_devices->latest_bdev;
929 	btrfs_item_key_to_cpu(leaf, &key, slot);
930 	extent_start = key.offset;
931 
932 	if (type == BTRFS_FILE_EXTENT_REG ||
933 	    type == BTRFS_FILE_EXTENT_PREALLOC) {
934 		extent_end = extent_start +
935 			btrfs_file_extent_num_bytes(leaf, fi);
936 	} else if (type == BTRFS_FILE_EXTENT_INLINE) {
937 		size_t size;
938 		size = btrfs_file_extent_ram_bytes(leaf, fi);
939 		extent_end = ALIGN(extent_start + size,
940 				   fs_info->sectorsize);
941 	}
942 
943 	em->ram_bytes = btrfs_file_extent_ram_bytes(leaf, fi);
944 	if (type == BTRFS_FILE_EXTENT_REG ||
945 	    type == BTRFS_FILE_EXTENT_PREALLOC) {
946 		em->start = extent_start;
947 		em->len = extent_end - extent_start;
948 		em->orig_start = extent_start -
949 			btrfs_file_extent_offset(leaf, fi);
950 		em->orig_block_len = btrfs_file_extent_disk_num_bytes(leaf, fi);
951 		bytenr = btrfs_file_extent_disk_bytenr(leaf, fi);
952 		if (bytenr == 0) {
953 			em->block_start = EXTENT_MAP_HOLE;
954 			return;
955 		}
956 		if (compress_type != BTRFS_COMPRESS_NONE) {
957 			set_bit(EXTENT_FLAG_COMPRESSED, &em->flags);
958 			em->compress_type = compress_type;
959 			em->block_start = bytenr;
960 			em->block_len = em->orig_block_len;
961 		} else {
962 			bytenr += btrfs_file_extent_offset(leaf, fi);
963 			em->block_start = bytenr;
964 			em->block_len = em->len;
965 			if (type == BTRFS_FILE_EXTENT_PREALLOC)
966 				set_bit(EXTENT_FLAG_PREALLOC, &em->flags);
967 		}
968 	} else if (type == BTRFS_FILE_EXTENT_INLINE) {
969 		em->block_start = EXTENT_MAP_INLINE;
970 		em->start = extent_start;
971 		em->len = extent_end - extent_start;
972 		/*
973 		 * Initialize orig_start and block_len with the same values
974 		 * as in inode.c:btrfs_get_extent().
975 		 */
976 		em->orig_start = EXTENT_MAP_HOLE;
977 		em->block_len = (u64)-1;
978 		if (!new_inline && compress_type != BTRFS_COMPRESS_NONE) {
979 			set_bit(EXTENT_FLAG_COMPRESSED, &em->flags);
980 			em->compress_type = compress_type;
981 		}
982 	} else {
983 		btrfs_err(fs_info,
984 			  "unknown file extent item type %d, inode %llu, offset %llu, "
985 			  "root %llu", type, btrfs_ino(inode), extent_start,
986 			  root->root_key.objectid);
987 	}
988 }
989