file-item.c 22.7 KB
Newer Older
C
Chris Mason 已提交
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18
/*
 * Copyright (C) 2007 Oracle.  All rights reserved.
 *
 * This program is free software; you can redistribute it and/or
 * modify it under the terms of the GNU General Public
 * License v2 as published by the Free Software Foundation.
 *
 * This program is distributed in the hope that it will be useful,
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
 * General Public License for more details.
 *
 * You should have received a copy of the GNU General Public
 * License along with this program; if not, write to the
 * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
 * Boston, MA 021110-1307, USA.
 */

19
#include <linux/bio.h>
20
#include <linux/slab.h>
21 22
#include <linux/pagemap.h>
#include <linux/highmem.h>
C
Chris Mason 已提交
23
#include "ctree.h"
C
Chris Mason 已提交
24
#include "disk-io.h"
25
#include "transaction.h"
C
Chris Mason 已提交
26
#include "print-tree.h"
C
Chris Mason 已提交
27

C
Chris Mason 已提交
28
#define MAX_CSUM_ITEMS(r, size) ((((BTRFS_LEAF_DATA_SIZE(r) - \
29 30
				   sizeof(struct btrfs_item) * 2) / \
				  size) - 1))
Y
Yan Zheng 已提交
31 32 33 34 35 36

#define MAX_ORDERED_SUM_BYTES(r) ((PAGE_SIZE - \
				   sizeof(struct btrfs_ordered_sum)) / \
				   sizeof(struct btrfs_sector_sum) * \
				   (r)->sectorsize - (r)->sectorsize)

C
Chris Mason 已提交
37
int btrfs_insert_file_extent(struct btrfs_trans_handle *trans,
S
Sage Weil 已提交
38 39 40
			     struct btrfs_root *root,
			     u64 objectid, u64 pos,
			     u64 disk_offset, u64 disk_num_bytes,
C
Chris Mason 已提交
41 42
			     u64 num_bytes, u64 offset, u64 ram_bytes,
			     u8 compression, u8 encryption, u16 other_encoding)
43
{
C
Chris Mason 已提交
44 45 46
	int ret = 0;
	struct btrfs_file_extent_item *item;
	struct btrfs_key file_key;
47
	struct btrfs_path *path;
48
	struct extent_buffer *leaf;
C
Chris Mason 已提交
49

50
	path = btrfs_alloc_path();
T
Tsutomu Itoh 已提交
51 52
	if (!path)
		return -ENOMEM;
C
Chris Mason 已提交
53
	file_key.objectid = objectid;
C
Chris Mason 已提交
54
	file_key.offset = pos;
C
Chris Mason 已提交
55 56
	btrfs_set_key_type(&file_key, BTRFS_EXTENT_DATA_KEY);

57
	path->leave_spinning = 1;
58
	ret = btrfs_insert_empty_item(trans, root, path, &file_key,
C
Chris Mason 已提交
59
				      sizeof(*item));
60 61
	if (ret < 0)
		goto out;
62
	BUG_ON(ret); /* Can't happen */
63 64
	leaf = path->nodes[0];
	item = btrfs_item_ptr(leaf, path->slots[0],
C
Chris Mason 已提交
65
			      struct btrfs_file_extent_item);
S
Sage Weil 已提交
66
	btrfs_set_file_extent_disk_bytenr(leaf, item, disk_offset);
67
	btrfs_set_file_extent_disk_num_bytes(leaf, item, disk_num_bytes);
S
Sage Weil 已提交
68
	btrfs_set_file_extent_offset(leaf, item, offset);
69
	btrfs_set_file_extent_num_bytes(leaf, item, num_bytes);
C
Chris Mason 已提交
70
	btrfs_set_file_extent_ram_bytes(leaf, item, ram_bytes);
71 72
	btrfs_set_file_extent_generation(leaf, item, trans->transid);
	btrfs_set_file_extent_type(leaf, item, BTRFS_FILE_EXTENT_REG);
C
Chris Mason 已提交
73 74 75 76
	btrfs_set_file_extent_compression(leaf, item, compression);
	btrfs_set_file_extent_encryption(leaf, item, encryption);
	btrfs_set_file_extent_other_encoding(leaf, item, other_encoding);

77
	btrfs_mark_buffer_dirty(leaf);
78
out:
79
	btrfs_free_path(path);
80
	return ret;
81
}
C
Chris Mason 已提交
82

C
Chris Mason 已提交
83 84 85
struct btrfs_csum_item *btrfs_lookup_csum(struct btrfs_trans_handle *trans,
					  struct btrfs_root *root,
					  struct btrfs_path *path,
86
					  u64 bytenr, int cow)
87 88 89 90 91
{
	int ret;
	struct btrfs_key file_key;
	struct btrfs_key found_key;
	struct btrfs_csum_item *item;
92
	struct extent_buffer *leaf;
93
	u64 csum_offset = 0;
94
	u16 csum_size = btrfs_super_csum_size(root->fs_info->super_copy);
95
	int csums_in_item;
96

97 98 99
	file_key.objectid = BTRFS_EXTENT_CSUM_OBJECTID;
	file_key.offset = bytenr;
	btrfs_set_key_type(&file_key, BTRFS_EXTENT_CSUM_KEY);
C
Chris Mason 已提交
100
	ret = btrfs_search_slot(trans, root, &file_key, path, 0, cow);
101 102
	if (ret < 0)
		goto fail;
103
	leaf = path->nodes[0];
104 105
	if (ret > 0) {
		ret = 1;
106
		if (path->slots[0] == 0)
107 108
			goto fail;
		path->slots[0]--;
109
		btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
110
		if (btrfs_key_type(&found_key) != BTRFS_EXTENT_CSUM_KEY)
111
			goto fail;
112 113

		csum_offset = (bytenr - found_key.offset) >>
114
				root->fs_info->sb->s_blocksize_bits;
115
		csums_in_item = btrfs_item_size_nr(leaf, path->slots[0]);
116
		csums_in_item /= csum_size;
117 118 119

		if (csum_offset >= csums_in_item) {
			ret = -EFBIG;
120 121 122 123
			goto fail;
		}
	}
	item = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_csum_item);
124
	item = (struct btrfs_csum_item *)((unsigned char *)item +
125
					  csum_offset * csum_size);
126 127 128
	return item;
fail:
	if (ret > 0)
C
Chris Mason 已提交
129
		ret = -ENOENT;
130 131 132 133
	return ERR_PTR(ret);
}


C
Chris Mason 已提交
134 135 136
int btrfs_lookup_file_extent(struct btrfs_trans_handle *trans,
			     struct btrfs_root *root,
			     struct btrfs_path *path, u64 objectid,
C
Chris Mason 已提交
137
			     u64 offset, int mod)
C
Chris Mason 已提交
138 139 140 141 142 143 144
{
	int ret;
	struct btrfs_key file_key;
	int ins_len = mod < 0 ? -1 : 0;
	int cow = mod != 0;

	file_key.objectid = objectid;
145
	file_key.offset = offset;
C
Chris Mason 已提交
146 147 148 149
	btrfs_set_key_type(&file_key, BTRFS_EXTENT_DATA_KEY);
	ret = btrfs_search_slot(trans, root, &file_key, path, ins_len, cow);
	return ret;
}
C
Chris Mason 已提交
150

151

152 153 154
static int __btrfs_lookup_bio_sums(struct btrfs_root *root,
				   struct inode *inode, struct bio *bio,
				   u64 logical_offset, u32 *dst, int dio)
155 156 157 158
{
	u32 sum;
	struct bio_vec *bvec = bio->bi_io_vec;
	int bio_index = 0;
159
	u64 offset = 0;
160 161
	u64 item_start_offset = 0;
	u64 item_last_offset = 0;
162
	u64 disk_bytenr;
163
	u32 diff;
164
	u16 csum_size = btrfs_super_csum_size(root->fs_info->super_copy);
165 166 167 168 169 170
	int ret;
	struct btrfs_path *path;
	struct btrfs_csum_item *item = NULL;
	struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;

	path = btrfs_alloc_path();
171 172
	if (!path)
		return -ENOMEM;
173 174
	if (bio->bi_size > PAGE_CACHE_SIZE * 8)
		path->reada = 2;
175 176 177

	WARN_ON(bio->bi_vcnt <= 0);

178 179 180 181 182 183
	/*
	 * the free space stuff is only read when it hasn't been
	 * updated in the current transaction.  So, we can safely
	 * read from the commit root and sidestep a nasty deadlock
	 * between reading the free space cache and updating the csum tree.
	 */
184
	if (btrfs_is_free_space_inode(root, inode)) {
185
		path->search_commit_root = 1;
186 187
		path->skip_locking = 1;
	}
188

189
	disk_bytenr = (u64)bio->bi_sector << 9;
190 191
	if (dio)
		offset = logical_offset;
C
Chris Mason 已提交
192
	while (bio_index < bio->bi_vcnt) {
193 194
		if (!dio)
			offset = page_offset(bvec->bv_page) + bvec->bv_offset;
195
		ret = btrfs_find_ordered_sum(inode, offset, disk_bytenr, &sum);
196 197 198
		if (ret == 0)
			goto found;

199 200
		if (!item || disk_bytenr < item_start_offset ||
		    disk_bytenr >= item_last_offset) {
201 202 203 204
			struct btrfs_key found_key;
			u32 item_size;

			if (item)
205
				btrfs_release_path(path);
206 207
			item = btrfs_lookup_csum(NULL, root->fs_info->csum_root,
						 path, disk_bytenr, 0);
208 209 210 211 212
			if (IS_ERR(item)) {
				ret = PTR_ERR(item);
				if (ret == -ENOENT || ret == -EFBIG)
					ret = 0;
				sum = 0;
213 214 215 216 217 218
				if (BTRFS_I(inode)->root->root_key.objectid ==
				    BTRFS_DATA_RELOC_TREE_OBJECTID) {
					set_extent_bits(io_tree, offset,
						offset + bvec->bv_len - 1,
						EXTENT_NODATASUM, GFP_NOFS);
				} else {
C
Chris Mason 已提交
219
					printk(KERN_INFO "btrfs no csum found "
L
Li Zefan 已提交
220 221 222
					       "for inode %llu start %llu\n",
					       (unsigned long long)
					       btrfs_ino(inode),
223 224
					       (unsigned long long)offset);
				}
225
				item = NULL;
226
				btrfs_release_path(path);
227 228 229 230 231 232 233 234 235
				goto found;
			}
			btrfs_item_key_to_cpu(path->nodes[0], &found_key,
					      path->slots[0]);

			item_start_offset = found_key.offset;
			item_size = btrfs_item_size_nr(path->nodes[0],
						       path->slots[0]);
			item_last_offset = item_start_offset +
236
				(item_size / csum_size) *
237 238 239 240 241 242 243 244
				root->sectorsize;
			item = btrfs_item_ptr(path->nodes[0], path->slots[0],
					      struct btrfs_csum_item);
		}
		/*
		 * this byte range must be able to fit inside
		 * a single leaf so it will also fit inside a u32
		 */
245
		diff = disk_bytenr - item_start_offset;
246
		diff = diff / root->sectorsize;
247
		diff = diff * csum_size;
248 249

		read_extent_buffer(path->nodes[0], &sum,
250
				   ((unsigned long)item) + diff,
251
				   csum_size);
252
found:
253 254 255 256 257
		if (dst)
			*dst++ = sum;
		else
			set_state_private(io_tree, offset, sum);
		disk_bytenr += bvec->bv_len;
258
		offset += bvec->bv_len;
259 260 261 262 263 264 265
		bio_index++;
		bvec++;
	}
	btrfs_free_path(path);
	return 0;
}

266 267 268 269 270 271 272 273 274 275 276 277
int btrfs_lookup_bio_sums(struct btrfs_root *root, struct inode *inode,
			  struct bio *bio, u32 *dst)
{
	return __btrfs_lookup_bio_sums(root, inode, bio, 0, dst, 0);
}

int btrfs_lookup_bio_sums_dio(struct btrfs_root *root, struct inode *inode,
			      struct bio *bio, u64 offset, u32 *dst)
{
	return __btrfs_lookup_bio_sums(root, inode, bio, offset, dst, 1);
}

278
int btrfs_lookup_csums_range(struct btrfs_root *root, u64 start, u64 end,
A
Arne Jansen 已提交
279
			     struct list_head *list, int search_commit)
280 281 282 283 284 285 286
{
	struct btrfs_key key;
	struct btrfs_path *path;
	struct extent_buffer *leaf;
	struct btrfs_ordered_sum *sums;
	struct btrfs_sector_sum *sector_sum;
	struct btrfs_csum_item *item;
287
	LIST_HEAD(tmplist);
288 289 290 291
	unsigned long offset;
	int ret;
	size_t size;
	u64 csum_end;
292
	u16 csum_size = btrfs_super_csum_size(root->fs_info->super_copy);
293 294

	path = btrfs_alloc_path();
295 296
	if (!path)
		return -ENOMEM;
297

A
Arne Jansen 已提交
298 299 300 301 302 303
	if (search_commit) {
		path->skip_locking = 1;
		path->reada = 2;
		path->search_commit_root = 1;
	}

304 305 306 307
	key.objectid = BTRFS_EXTENT_CSUM_OBJECTID;
	key.offset = start;
	key.type = BTRFS_EXTENT_CSUM_KEY;

Y
Yan Zheng 已提交
308
	ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326
	if (ret < 0)
		goto fail;
	if (ret > 0 && path->slots[0] > 0) {
		leaf = path->nodes[0];
		btrfs_item_key_to_cpu(leaf, &key, path->slots[0] - 1);
		if (key.objectid == BTRFS_EXTENT_CSUM_OBJECTID &&
		    key.type == BTRFS_EXTENT_CSUM_KEY) {
			offset = (start - key.offset) >>
				 root->fs_info->sb->s_blocksize_bits;
			if (offset * csum_size <
			    btrfs_item_size_nr(leaf, path->slots[0] - 1))
				path->slots[0]--;
		}
	}

	while (start <= end) {
		leaf = path->nodes[0];
		if (path->slots[0] >= btrfs_header_nritems(leaf)) {
Y
Yan Zheng 已提交
327
			ret = btrfs_next_leaf(root, path);
328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348
			if (ret < 0)
				goto fail;
			if (ret > 0)
				break;
			leaf = path->nodes[0];
		}

		btrfs_item_key_to_cpu(leaf, &key, path->slots[0]);
		if (key.objectid != BTRFS_EXTENT_CSUM_OBJECTID ||
		    key.type != BTRFS_EXTENT_CSUM_KEY)
			break;

		btrfs_item_key_to_cpu(leaf, &key, path->slots[0]);
		if (key.offset > end)
			break;

		if (key.offset > start)
			start = key.offset;

		size = btrfs_item_size_nr(leaf, path->slots[0]);
		csum_end = key.offset + (size / csum_size) * root->sectorsize;
349 350 351 352
		if (csum_end <= start) {
			path->slots[0]++;
			continue;
		}
353

Y
Yan Zheng 已提交
354
		csum_end = min(csum_end, end + 1);
355 356
		item = btrfs_item_ptr(path->nodes[0], path->slots[0],
				      struct btrfs_csum_item);
Y
Yan Zheng 已提交
357 358 359 360 361
		while (start < csum_end) {
			size = min_t(size_t, csum_end - start,
					MAX_ORDERED_SUM_BYTES(root));
			sums = kzalloc(btrfs_ordered_sum_size(root, size),
					GFP_NOFS);
362 363 364 365
			if (!sums) {
				ret = -ENOMEM;
				goto fail;
			}
366

Y
Yan Zheng 已提交
367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386
			sector_sum = sums->sums;
			sums->bytenr = start;
			sums->len = size;

			offset = (start - key.offset) >>
				root->fs_info->sb->s_blocksize_bits;
			offset *= csum_size;

			while (size > 0) {
				read_extent_buffer(path->nodes[0],
						&sector_sum->sum,
						((unsigned long)item) +
						offset, csum_size);
				sector_sum->bytenr = start;

				size -= root->sectorsize;
				start += root->sectorsize;
				offset += csum_size;
				sector_sum++;
			}
387
			list_add_tail(&sums->list, &tmplist);
Y
Yan Zheng 已提交
388
		}
389 390 391 392
		path->slots[0]++;
	}
	ret = 0;
fail:
393 394 395 396 397 398 399
	while (ret < 0 && !list_empty(&tmplist)) {
		sums = list_entry(&tmplist, struct btrfs_ordered_sum, list);
		list_del(&sums->list);
		kfree(sums);
	}
	list_splice_tail(&tmplist, list);

400 401 402 403
	btrfs_free_path(path);
	return ret;
}

404
int btrfs_csum_one_bio(struct btrfs_root *root, struct inode *inode,
405
		       struct bio *bio, u64 file_start, int contig)
406
{
407 408
	struct btrfs_ordered_sum *sums;
	struct btrfs_sector_sum *sector_sum;
409
	struct btrfs_ordered_extent *ordered;
410 411 412
	char *data;
	struct bio_vec *bvec = bio->bi_io_vec;
	int bio_index = 0;
413 414 415
	unsigned long total_bytes = 0;
	unsigned long this_sum_bytes = 0;
	u64 offset;
416
	u64 disk_bytenr;
417

418 419
	WARN_ON(bio->bi_vcnt <= 0);
	sums = kzalloc(btrfs_ordered_sum_size(root, bio->bi_size), GFP_NOFS);
420 421
	if (!sums)
		return -ENOMEM;
422

423
	sector_sum = sums->sums;
424
	disk_bytenr = (u64)bio->bi_sector << 9;
425 426
	sums->len = bio->bi_size;
	INIT_LIST_HEAD(&sums->list);
427 428 429 430 431 432 433

	if (contig)
		offset = file_start;
	else
		offset = page_offset(bvec->bv_page) + bvec->bv_offset;

	ordered = btrfs_lookup_ordered_extent(inode, offset);
434
	BUG_ON(!ordered); /* Logic error */
435
	sums->bytenr = ordered->start;
436

C
Chris Mason 已提交
437
	while (bio_index < bio->bi_vcnt) {
438 439 440 441 442
		if (!contig)
			offset = page_offset(bvec->bv_page) + bvec->bv_offset;

		if (!contig && (offset >= ordered->file_offset + ordered->len ||
		    offset < ordered->file_offset)) {
443 444 445 446 447 448 449 450 451 452
			unsigned long bytes_left;
			sums->len = this_sum_bytes;
			this_sum_bytes = 0;
			btrfs_add_ordered_sum(inode, ordered, sums);
			btrfs_put_ordered_extent(ordered);

			bytes_left = bio->bi_size - total_bytes;

			sums = kzalloc(btrfs_ordered_sum_size(root, bytes_left),
				       GFP_NOFS);
453
			BUG_ON(!sums); /* -ENOMEM */
454
			sector_sum = sums->sums;
455
			sums->len = bytes_left;
456
			ordered = btrfs_lookup_ordered_extent(inode, offset);
457
			BUG_ON(!ordered); /* Logic error */
458
			sums->bytenr = ordered->start;
459 460
		}

461
		data = kmap_atomic(bvec->bv_page, KM_USER0);
462 463 464 465 466
		sector_sum->sum = ~(u32)0;
		sector_sum->sum = btrfs_csum_data(root,
						  data + bvec->bv_offset,
						  sector_sum->sum,
						  bvec->bv_len);
467
		kunmap_atomic(data, KM_USER0);
468 469
		btrfs_csum_final(sector_sum->sum,
				 (char *)&sector_sum->sum);
470
		sector_sum->bytenr = disk_bytenr;
471

472
		sector_sum++;
473
		bio_index++;
474 475
		total_bytes += bvec->bv_len;
		this_sum_bytes += bvec->bv_len;
476 477
		disk_bytenr += bvec->bv_len;
		offset += bvec->bv_len;
478 479
		bvec++;
	}
480
	this_sum_bytes = 0;
481 482
	btrfs_add_ordered_sum(inode, ordered, sums);
	btrfs_put_ordered_extent(ordered);
483 484 485
	return 0;
}

486 487 488 489 490 491 492 493 494 495 496
/*
 * helper function for csum removal, this expects the
 * key to describe the csum pointed to by the path, and it expects
 * the csum to overlap the range [bytenr, len]
 *
 * The csum should not be entirely contained in the range and the
 * range should not be entirely contained in the csum.
 *
 * This calls btrfs_truncate_item with the correct args based on the
 * overlap, and fixes up the key as required.
 */
497 498 499 500 501
static noinline void truncate_one_csum(struct btrfs_trans_handle *trans,
				       struct btrfs_root *root,
				       struct btrfs_path *path,
				       struct btrfs_key *key,
				       u64 bytenr, u64 len)
502 503
{
	struct extent_buffer *leaf;
504
	u16 csum_size = btrfs_super_csum_size(root->fs_info->super_copy);
505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522
	u64 csum_end;
	u64 end_byte = bytenr + len;
	u32 blocksize_bits = root->fs_info->sb->s_blocksize_bits;

	leaf = path->nodes[0];
	csum_end = btrfs_item_size_nr(leaf, path->slots[0]) / csum_size;
	csum_end <<= root->fs_info->sb->s_blocksize_bits;
	csum_end += key->offset;

	if (key->offset < bytenr && csum_end <= end_byte) {
		/*
		 *         [ bytenr - len ]
		 *         [   ]
		 *   [csum     ]
		 *   A simple truncate off the end of the item
		 */
		u32 new_size = (bytenr - key->offset) >> blocksize_bits;
		new_size *= csum_size;
523
		btrfs_truncate_item(trans, root, path, new_size, 1);
524 525 526 527 528 529 530 531 532 533 534
	} else if (key->offset >= bytenr && csum_end > end_byte &&
		   end_byte > key->offset) {
		/*
		 *         [ bytenr - len ]
		 *                 [ ]
		 *                 [csum     ]
		 * we need to truncate from the beginning of the csum
		 */
		u32 new_size = (csum_end - end_byte) >> blocksize_bits;
		new_size *= csum_size;

535
		btrfs_truncate_item(trans, root, path, new_size, 0);
536 537

		key->offset = end_byte;
538
		btrfs_set_item_key_safe(trans, root, path, key);
539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556
	} else {
		BUG();
	}
}

/*
 * deletes the csum items from the csum tree for a given
 * range of bytes.
 */
int btrfs_del_csums(struct btrfs_trans_handle *trans,
		    struct btrfs_root *root, u64 bytenr, u64 len)
{
	struct btrfs_path *path;
	struct btrfs_key key;
	u64 end_byte = bytenr + len;
	u64 csum_end;
	struct extent_buffer *leaf;
	int ret;
557
	u16 csum_size = btrfs_super_csum_size(root->fs_info->super_copy);
558 559 560 561 562
	int blocksize_bits = root->fs_info->sb->s_blocksize_bits;

	root = root->fs_info->csum_root;

	path = btrfs_alloc_path();
563 564
	if (!path)
		return -ENOMEM;
565

C
Chris Mason 已提交
566
	while (1) {
567 568 569 570
		key.objectid = BTRFS_EXTENT_CSUM_OBJECTID;
		key.offset = end_byte - 1;
		key.type = BTRFS_EXTENT_CSUM_KEY;

571
		path->leave_spinning = 1;
572 573 574
		ret = btrfs_search_slot(trans, root, &key, path, -1, 1);
		if (ret > 0) {
			if (path->slots[0] == 0)
575
				break;
576
			path->slots[0]--;
577
		} else if (ret < 0) {
578
			break;
579
		}
580

581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602
		leaf = path->nodes[0];
		btrfs_item_key_to_cpu(leaf, &key, path->slots[0]);

		if (key.objectid != BTRFS_EXTENT_CSUM_OBJECTID ||
		    key.type != BTRFS_EXTENT_CSUM_KEY) {
			break;
		}

		if (key.offset >= end_byte)
			break;

		csum_end = btrfs_item_size_nr(leaf, path->slots[0]) / csum_size;
		csum_end <<= blocksize_bits;
		csum_end += key.offset;

		/* this csum ends before we start, we're done */
		if (csum_end <= bytenr)
			break;

		/* delete the entire item, it is inside our range */
		if (key.offset >= bytenr && csum_end <= end_byte) {
			ret = btrfs_del_item(trans, root, path);
603 604
			if (ret)
				goto out;
605 606
			if (key.offset == bytenr)
				break;
607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645
		} else if (key.offset < bytenr && csum_end > end_byte) {
			unsigned long offset;
			unsigned long shift_len;
			unsigned long item_offset;
			/*
			 *        [ bytenr - len ]
			 *     [csum                ]
			 *
			 * Our bytes are in the middle of the csum,
			 * we need to split this item and insert a new one.
			 *
			 * But we can't drop the path because the
			 * csum could change, get removed, extended etc.
			 *
			 * The trick here is the max size of a csum item leaves
			 * enough room in the tree block for a single
			 * item header.  So, we split the item in place,
			 * adding a new header pointing to the existing
			 * bytes.  Then we loop around again and we have
			 * a nicely formed csum item that we can neatly
			 * truncate.
			 */
			offset = (bytenr - key.offset) >> blocksize_bits;
			offset *= csum_size;

			shift_len = (len >> blocksize_bits) * csum_size;

			item_offset = btrfs_item_ptr_offset(leaf,
							    path->slots[0]);

			memset_extent_buffer(leaf, 0, item_offset + offset,
					     shift_len);
			key.offset = bytenr;

			/*
			 * btrfs_split_item returns -EAGAIN when the
			 * item changed size or key
			 */
			ret = btrfs_split_item(trans, root, path, &key, offset);
646 647 648 649
			if (ret && ret != -EAGAIN) {
				btrfs_abort_transaction(trans, root, ret);
				goto out;
			}
650 651 652

			key.offset = end_byte - 1;
		} else {
653
			truncate_one_csum(trans, root, path, &key, bytenr, len);
654 655
			if (key.offset < bytenr)
				break;
656
		}
657
		btrfs_release_path(path);
658
	}
659
	ret = 0;
660 661
out:
	btrfs_free_path(path);
662
	return ret;
663 664
}

665
int btrfs_csum_file_blocks(struct btrfs_trans_handle *trans,
666
			   struct btrfs_root *root,
667
			   struct btrfs_ordered_sum *sums)
C
Chris Mason 已提交
668
{
669
	u64 bytenr;
C
Chris Mason 已提交
670 671
	int ret;
	struct btrfs_key file_key;
672
	struct btrfs_key found_key;
673
	u64 next_offset;
674
	u64 total_bytes = 0;
675
	int found_next;
676
	struct btrfs_path *path;
C
Chris Mason 已提交
677
	struct btrfs_csum_item *item;
678
	struct btrfs_csum_item *item_end;
679
	struct extent_buffer *leaf = NULL;
680
	u64 csum_offset;
681
	struct btrfs_sector_sum *sector_sum;
682 683
	u32 nritems;
	u32 ins_size;
684
	u16 csum_size = btrfs_super_csum_size(root->fs_info->super_copy);
685

686
	path = btrfs_alloc_path();
687 688 689
	if (!path)
		return -ENOMEM;

690
	sector_sum = sums->sums;
691 692 693
again:
	next_offset = (u64)-1;
	found_next = 0;
694 695 696 697
	file_key.objectid = BTRFS_EXTENT_CSUM_OBJECTID;
	file_key.offset = sector_sum->bytenr;
	bytenr = sector_sum->bytenr;
	btrfs_set_key_type(&file_key, BTRFS_EXTENT_CSUM_KEY);
698

699
	item = btrfs_lookup_csum(trans, root, path, sector_sum->bytenr, 1);
700 701
	if (!IS_ERR(item)) {
		leaf = path->nodes[0];
702
		ret = 0;
703
		goto found;
704
	}
705
	ret = PTR_ERR(item);
706 707 708
	if (ret != -EFBIG && ret != -ENOENT)
		goto fail_unlock;

709 710 711
	if (ret == -EFBIG) {
		u32 item_size;
		/* we found one, but it isn't big enough yet */
712 713
		leaf = path->nodes[0];
		item_size = btrfs_item_size_nr(leaf, path->slots[0]);
714 715
		if ((item_size / csum_size) >=
		    MAX_CSUM_ITEMS(root, csum_size)) {
716 717 718 719
			/* already at max size, make a new one */
			goto insert;
		}
	} else {
720
		int slot = path->slots[0] + 1;
721
		/* we didn't find a csum item, insert one */
722 723 724
		nritems = btrfs_header_nritems(path->nodes[0]);
		if (path->slots[0] >= nritems - 1) {
			ret = btrfs_next_leaf(root, path);
Y
Yan 已提交
725
			if (ret == 1)
726
				found_next = 1;
Y
Yan 已提交
727
			if (ret != 0)
728
				goto insert;
Y
Yan 已提交
729
			slot = 0;
730 731
		}
		btrfs_item_key_to_cpu(path->nodes[0], &found_key, slot);
732 733
		if (found_key.objectid != BTRFS_EXTENT_CSUM_OBJECTID ||
		    found_key.type != BTRFS_EXTENT_CSUM_KEY) {
734 735 736 737 738
			found_next = 1;
			goto insert;
		}
		next_offset = found_key.offset;
		found_next = 1;
739 740 741 742 743 744 745
		goto insert;
	}

	/*
	 * at this point, we know the tree has an item, but it isn't big
	 * enough yet to put our csum in.  Grow it
	 */
746
	btrfs_release_path(path);
747
	ret = btrfs_search_slot(trans, root, &file_key, path,
748
				csum_size, 1);
749
	if (ret < 0)
750
		goto fail_unlock;
751 752 753 754 755

	if (ret > 0) {
		if (path->slots[0] == 0)
			goto insert;
		path->slots[0]--;
756
	}
757

758 759
	leaf = path->nodes[0];
	btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
760
	csum_offset = (bytenr - found_key.offset) >>
761
			root->fs_info->sb->s_blocksize_bits;
762

763 764
	if (btrfs_key_type(&found_key) != BTRFS_EXTENT_CSUM_KEY ||
	    found_key.objectid != BTRFS_EXTENT_CSUM_OBJECTID ||
765
	    csum_offset >= MAX_CSUM_ITEMS(root, csum_size)) {
766 767
		goto insert;
	}
768

769
	if (csum_offset >= btrfs_item_size_nr(leaf, path->slots[0]) /
770 771
	    csum_size) {
		u32 diff = (csum_offset + 1) * csum_size;
772 773 774 775 776 777 778 779

		/*
		 * is the item big enough already?  we dropped our lock
		 * before and need to recheck
		 */
		if (diff < btrfs_item_size_nr(leaf, path->slots[0]))
			goto csum;

780
		diff = diff - btrfs_item_size_nr(leaf, path->slots[0]);
C
Chris Mason 已提交
781
		if (diff != csum_size)
C
Chris Mason 已提交
782
			goto insert;
783

784
		btrfs_extend_item(trans, root, path, diff);
785 786 787 788
		goto csum;
	}

insert:
789
	btrfs_release_path(path);
790
	csum_offset = 0;
791
	if (found_next) {
792 793 794 795
		u64 tmp = total_bytes + root->sectorsize;
		u64 next_sector = sector_sum->bytenr;
		struct btrfs_sector_sum *next = sector_sum + 1;

C
Chris Mason 已提交
796
		while (tmp < sums->len) {
797 798 799 800 801 802 803
			if (next_sector + root->sectorsize != next->bytenr)
				break;
			tmp += root->sectorsize;
			next_sector = next->bytenr;
			next++;
		}
		tmp = min(tmp, next_offset - file_key.offset);
804 805
		tmp >>= root->fs_info->sb->s_blocksize_bits;
		tmp = max((u64)1, tmp);
806 807
		tmp = min(tmp, (u64)MAX_CSUM_ITEMS(root, csum_size));
		ins_size = csum_size * tmp;
808
	} else {
809
		ins_size = csum_size;
810
	}
811
	path->leave_spinning = 1;
812
	ret = btrfs_insert_empty_item(trans, root, path, &file_key,
813
				      ins_size);
814
	path->leave_spinning = 0;
815
	if (ret < 0)
816
		goto fail_unlock;
817 818
	if (ret != 0) {
		WARN_ON(1);
819
		goto fail_unlock;
820
	}
821
csum:
822 823
	leaf = path->nodes[0];
	item = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_csum_item);
C
Chris Mason 已提交
824
	ret = 0;
825
	item = (struct btrfs_csum_item *)((unsigned char *)item +
826
					  csum_offset * csum_size);
C
Chris Mason 已提交
827
found:
828 829 830
	item_end = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_csum_item);
	item_end = (struct btrfs_csum_item *)((unsigned char *)item_end +
				      btrfs_item_size_nr(leaf, path->slots[0]));
831
next_sector:
832

833
	write_extent_buffer(leaf, &sector_sum->sum, (unsigned long)item, csum_size);
834

835 836 837
	total_bytes += root->sectorsize;
	sector_sum++;
	if (total_bytes < sums->len) {
838
		item = (struct btrfs_csum_item *)((char *)item +
839
						  csum_size);
840 841 842
		if (item < item_end && bytenr + PAGE_CACHE_SIZE ==
		    sector_sum->bytenr) {
			bytenr = sector_sum->bytenr;
843
			goto next_sector;
844
		}
845
	}
846

847
	btrfs_mark_buffer_dirty(path->nodes[0]);
848
	if (total_bytes < sums->len) {
849
		btrfs_release_path(path);
850
		cond_resched();
851 852
		goto again;
	}
853
out:
854
	btrfs_free_path(path);
C
Chris Mason 已提交
855
	return ret;
856 857 858

fail_unlock:
	goto out;
C
Chris Mason 已提交
859
}