file-item.c 22.8 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

28
#define __MAX_CSUM_ITEMS(r, size) ((unsigned long)(((BTRFS_LEAF_DATA_SIZE(r) - \
29 30
				   sizeof(struct btrfs_item) * 2) / \
				  size) - 1))
Y
Yan Zheng 已提交
31

32 33
#define MAX_CSUM_ITEMS(r, size) (min_t(u32, __MAX_CSUM_ITEMS(r, size), \
				       PAGE_CACHE_SIZE))
34

Y
Yan Zheng 已提交
35 36 37 38 39
#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
40
int btrfs_insert_file_extent(struct btrfs_trans_handle *trans,
S
Sage Weil 已提交
41 42 43
			     struct btrfs_root *root,
			     u64 objectid, u64 pos,
			     u64 disk_offset, u64 disk_num_bytes,
C
Chris Mason 已提交
44 45
			     u64 num_bytes, u64 offset, u64 ram_bytes,
			     u8 compression, u8 encryption, u16 other_encoding)
46
{
C
Chris Mason 已提交
47 48 49
	int ret = 0;
	struct btrfs_file_extent_item *item;
	struct btrfs_key file_key;
50
	struct btrfs_path *path;
51
	struct extent_buffer *leaf;
C
Chris Mason 已提交
52

53
	path = btrfs_alloc_path();
T
Tsutomu Itoh 已提交
54 55
	if (!path)
		return -ENOMEM;
C
Chris Mason 已提交
56
	file_key.objectid = objectid;
C
Chris Mason 已提交
57
	file_key.offset = pos;
C
Chris Mason 已提交
58 59
	btrfs_set_key_type(&file_key, BTRFS_EXTENT_DATA_KEY);

60
	path->leave_spinning = 1;
61
	ret = btrfs_insert_empty_item(trans, root, path, &file_key,
C
Chris Mason 已提交
62
				      sizeof(*item));
63 64
	if (ret < 0)
		goto out;
65
	BUG_ON(ret); /* Can't happen */
66 67
	leaf = path->nodes[0];
	item = btrfs_item_ptr(leaf, path->slots[0],
C
Chris Mason 已提交
68
			      struct btrfs_file_extent_item);
S
Sage Weil 已提交
69
	btrfs_set_file_extent_disk_bytenr(leaf, item, disk_offset);
70
	btrfs_set_file_extent_disk_num_bytes(leaf, item, disk_num_bytes);
S
Sage Weil 已提交
71
	btrfs_set_file_extent_offset(leaf, item, offset);
72
	btrfs_set_file_extent_num_bytes(leaf, item, num_bytes);
C
Chris Mason 已提交
73
	btrfs_set_file_extent_ram_bytes(leaf, item, ram_bytes);
74 75
	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
76 77 78 79
	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);

80
	btrfs_mark_buffer_dirty(leaf);
81
out:
82
	btrfs_free_path(path);
83
	return ret;
84
}
C
Chris Mason 已提交
85

C
Chris Mason 已提交
86 87 88
struct btrfs_csum_item *btrfs_lookup_csum(struct btrfs_trans_handle *trans,
					  struct btrfs_root *root,
					  struct btrfs_path *path,
89
					  u64 bytenr, int cow)
90 91 92 93 94
{
	int ret;
	struct btrfs_key file_key;
	struct btrfs_key found_key;
	struct btrfs_csum_item *item;
95
	struct extent_buffer *leaf;
96
	u64 csum_offset = 0;
97
	u16 csum_size = btrfs_super_csum_size(root->fs_info->super_copy);
98
	int csums_in_item;
99

100 101 102
	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
103
	ret = btrfs_search_slot(trans, root, &file_key, path, 0, cow);
104 105
	if (ret < 0)
		goto fail;
106
	leaf = path->nodes[0];
107 108
	if (ret > 0) {
		ret = 1;
109
		if (path->slots[0] == 0)
110 111
			goto fail;
		path->slots[0]--;
112
		btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
113
		if (btrfs_key_type(&found_key) != BTRFS_EXTENT_CSUM_KEY)
114
			goto fail;
115 116

		csum_offset = (bytenr - found_key.offset) >>
117
				root->fs_info->sb->s_blocksize_bits;
118
		csums_in_item = btrfs_item_size_nr(leaf, path->slots[0]);
119
		csums_in_item /= csum_size;
120 121 122

		if (csum_offset >= csums_in_item) {
			ret = -EFBIG;
123 124 125 126
			goto fail;
		}
	}
	item = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_csum_item);
127
	item = (struct btrfs_csum_item *)((unsigned char *)item +
128
					  csum_offset * csum_size);
129 130 131
	return item;
fail:
	if (ret > 0)
C
Chris Mason 已提交
132
		ret = -ENOENT;
133 134 135 136
	return ERR_PTR(ret);
}


C
Chris Mason 已提交
137 138 139
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
140
			     u64 offset, int mod)
C
Chris Mason 已提交
141 142 143 144 145 146 147
{
	int ret;
	struct btrfs_key file_key;
	int ins_len = mod < 0 ? -1 : 0;
	int cow = mod != 0;

	file_key.objectid = objectid;
148
	file_key.offset = offset;
C
Chris Mason 已提交
149 150 151 152
	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
153

154

155 156 157
static int __btrfs_lookup_bio_sums(struct btrfs_root *root,
				   struct inode *inode, struct bio *bio,
				   u64 logical_offset, u32 *dst, int dio)
158 159 160 161
{
	u32 sum;
	struct bio_vec *bvec = bio->bi_io_vec;
	int bio_index = 0;
162
	u64 offset = 0;
163 164
	u64 item_start_offset = 0;
	u64 item_last_offset = 0;
165
	u64 disk_bytenr;
166
	u32 diff;
167
	u16 csum_size = btrfs_super_csum_size(root->fs_info->super_copy);
168 169 170 171 172 173
	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();
174 175
	if (!path)
		return -ENOMEM;
176 177
	if (bio->bi_size > PAGE_CACHE_SIZE * 8)
		path->reada = 2;
178 179 180

	WARN_ON(bio->bi_vcnt <= 0);

181 182 183 184 185 186
	/*
	 * 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.
	 */
187
	if (btrfs_is_free_space_inode(inode)) {
188
		path->search_commit_root = 1;
189 190
		path->skip_locking = 1;
	}
191

192
	disk_bytenr = (u64)bio->bi_sector << 9;
193 194
	if (dio)
		offset = logical_offset;
C
Chris Mason 已提交
195
	while (bio_index < bio->bi_vcnt) {
196 197
		if (!dio)
			offset = page_offset(bvec->bv_page) + bvec->bv_offset;
198
		ret = btrfs_find_ordered_sum(inode, offset, disk_bytenr, &sum);
199 200 201
		if (ret == 0)
			goto found;

202 203
		if (!item || disk_bytenr < item_start_offset ||
		    disk_bytenr >= item_last_offset) {
204 205 206 207
			struct btrfs_key found_key;
			u32 item_size;

			if (item)
208
				btrfs_release_path(path);
209 210
			item = btrfs_lookup_csum(NULL, root->fs_info->csum_root,
						 path, disk_bytenr, 0);
211 212 213 214 215
			if (IS_ERR(item)) {
				ret = PTR_ERR(item);
				if (ret == -ENOENT || ret == -EFBIG)
					ret = 0;
				sum = 0;
216 217 218 219 220 221
				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
222
					printk(KERN_INFO "btrfs no csum found "
L
Li Zefan 已提交
223 224 225
					       "for inode %llu start %llu\n",
					       (unsigned long long)
					       btrfs_ino(inode),
226 227
					       (unsigned long long)offset);
				}
228
				item = NULL;
229
				btrfs_release_path(path);
230 231 232 233 234 235 236 237 238
				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 +
239
				(item_size / csum_size) *
240 241 242 243 244 245 246 247
				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
		 */
248
		diff = disk_bytenr - item_start_offset;
249
		diff = diff / root->sectorsize;
250
		diff = diff * csum_size;
251 252

		read_extent_buffer(path->nodes[0], &sum,
253
				   ((unsigned long)item) + diff,
254
				   csum_size);
255
found:
256 257 258 259 260
		if (dst)
			*dst++ = sum;
		else
			set_state_private(io_tree, offset, sum);
		disk_bytenr += bvec->bv_len;
261
		offset += bvec->bv_len;
262 263 264 265 266 267 268
		bio_index++;
		bvec++;
	}
	btrfs_free_path(path);
	return 0;
}

269 270 271 272 273 274 275
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,
276
			      struct bio *bio, u64 offset)
277
{
278
	return __btrfs_lookup_bio_sums(root, inode, bio, offset, NULL, 1);
279 280
}

281
int btrfs_lookup_csums_range(struct btrfs_root *root, u64 start, u64 end,
A
Arne Jansen 已提交
282
			     struct list_head *list, int search_commit)
283 284 285 286 287 288 289
{
	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;
290
	LIST_HEAD(tmplist);
291 292 293 294
	unsigned long offset;
	int ret;
	size_t size;
	u64 csum_end;
295
	u16 csum_size = btrfs_super_csum_size(root->fs_info->super_copy);
296 297

	path = btrfs_alloc_path();
298 299
	if (!path)
		return -ENOMEM;
300

A
Arne Jansen 已提交
301 302 303 304 305 306
	if (search_commit) {
		path->skip_locking = 1;
		path->reada = 2;
		path->search_commit_root = 1;
	}

307 308 309 310
	key.objectid = BTRFS_EXTENT_CSUM_OBJECTID;
	key.offset = start;
	key.type = BTRFS_EXTENT_CSUM_KEY;

Y
Yan Zheng 已提交
311
	ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329
	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
330
			ret = btrfs_next_leaf(root, path);
331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351
			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;
352 353 354 355
		if (csum_end <= start) {
			path->slots[0]++;
			continue;
		}
356

Y
Yan Zheng 已提交
357
		csum_end = min(csum_end, end + 1);
358 359
		item = btrfs_item_ptr(path->nodes[0], path->slots[0],
				      struct btrfs_csum_item);
Y
Yan Zheng 已提交
360 361 362 363 364
		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);
365 366 367 368
			if (!sums) {
				ret = -ENOMEM;
				goto fail;
			}
369

Y
Yan Zheng 已提交
370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389
			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++;
			}
390
			list_add_tail(&sums->list, &tmplist);
Y
Yan Zheng 已提交
391
		}
392 393 394 395
		path->slots[0]++;
	}
	ret = 0;
fail:
396 397 398 399 400 401 402
	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);

403 404 405 406
	btrfs_free_path(path);
	return ret;
}

407
int btrfs_csum_one_bio(struct btrfs_root *root, struct inode *inode,
408
		       struct bio *bio, u64 file_start, int contig)
409
{
410 411
	struct btrfs_ordered_sum *sums;
	struct btrfs_sector_sum *sector_sum;
412
	struct btrfs_ordered_extent *ordered;
413 414 415
	char *data;
	struct bio_vec *bvec = bio->bi_io_vec;
	int bio_index = 0;
416 417 418
	unsigned long total_bytes = 0;
	unsigned long this_sum_bytes = 0;
	u64 offset;
419
	u64 disk_bytenr;
420

421 422
	WARN_ON(bio->bi_vcnt <= 0);
	sums = kzalloc(btrfs_ordered_sum_size(root, bio->bi_size), GFP_NOFS);
423 424
	if (!sums)
		return -ENOMEM;
425

426
	sector_sum = sums->sums;
427
	disk_bytenr = (u64)bio->bi_sector << 9;
428 429
	sums->len = bio->bi_size;
	INIT_LIST_HEAD(&sums->list);
430 431 432 433 434 435 436

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

	ordered = btrfs_lookup_ordered_extent(inode, offset);
437
	BUG_ON(!ordered); /* Logic error */
438
	sums->bytenr = ordered->start;
439

C
Chris Mason 已提交
440
	while (bio_index < bio->bi_vcnt) {
441 442 443 444 445
		if (!contig)
			offset = page_offset(bvec->bv_page) + bvec->bv_offset;

		if (!contig && (offset >= ordered->file_offset + ordered->len ||
		    offset < ordered->file_offset)) {
446 447 448 449 450 451 452 453 454 455
			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);
456
			BUG_ON(!sums); /* -ENOMEM */
457
			sector_sum = sums->sums;
458
			sums->len = bytes_left;
459
			ordered = btrfs_lookup_ordered_extent(inode, offset);
460
			BUG_ON(!ordered); /* Logic error */
461
			sums->bytenr = ordered->start;
462 463
		}

464
		data = kmap_atomic(bvec->bv_page);
465 466 467 468 469
		sector_sum->sum = ~(u32)0;
		sector_sum->sum = btrfs_csum_data(root,
						  data + bvec->bv_offset,
						  sector_sum->sum,
						  bvec->bv_len);
470
		kunmap_atomic(data);
471 472
		btrfs_csum_final(sector_sum->sum,
				 (char *)&sector_sum->sum);
473
		sector_sum->bytenr = disk_bytenr;
474

475
		sector_sum++;
476
		bio_index++;
477 478
		total_bytes += bvec->bv_len;
		this_sum_bytes += bvec->bv_len;
479 480
		disk_bytenr += bvec->bv_len;
		offset += bvec->bv_len;
481 482
		bvec++;
	}
483
	this_sum_bytes = 0;
484 485
	btrfs_add_ordered_sum(inode, ordered, sums);
	btrfs_put_ordered_extent(ordered);
486 487 488
	return 0;
}

489 490 491 492 493 494 495 496 497 498 499
/*
 * 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.
 */
500 501 502 503 504
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)
505 506
{
	struct extent_buffer *leaf;
507
	u16 csum_size = btrfs_super_csum_size(root->fs_info->super_copy);
508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525
	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;
526
		btrfs_truncate_item(trans, root, path, new_size, 1);
527 528 529 530 531 532 533 534 535 536 537
	} 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;

538
		btrfs_truncate_item(trans, root, path, new_size, 0);
539 540

		key->offset = end_byte;
541
		btrfs_set_item_key_safe(trans, root, path, key);
542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559
	} 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;
560
	u16 csum_size = btrfs_super_csum_size(root->fs_info->super_copy);
561 562 563 564 565
	int blocksize_bits = root->fs_info->sb->s_blocksize_bits;

	root = root->fs_info->csum_root;

	path = btrfs_alloc_path();
566 567
	if (!path)
		return -ENOMEM;
568

C
Chris Mason 已提交
569
	while (1) {
570 571 572 573
		key.objectid = BTRFS_EXTENT_CSUM_OBJECTID;
		key.offset = end_byte - 1;
		key.type = BTRFS_EXTENT_CSUM_KEY;

574
		path->leave_spinning = 1;
575 576 577
		ret = btrfs_search_slot(trans, root, &key, path, -1, 1);
		if (ret > 0) {
			if (path->slots[0] == 0)
578
				break;
579
			path->slots[0]--;
580
		} else if (ret < 0) {
581
			break;
582
		}
583

584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605
		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);
606 607
			if (ret)
				goto out;
608 609
			if (key.offset == bytenr)
				break;
610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648
		} 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);
649 650 651 652
			if (ret && ret != -EAGAIN) {
				btrfs_abort_transaction(trans, root, ret);
				goto out;
			}
653 654 655

			key.offset = end_byte - 1;
		} else {
656
			truncate_one_csum(trans, root, path, &key, bytenr, len);
657 658
			if (key.offset < bytenr)
				break;
659
		}
660
		btrfs_release_path(path);
661
	}
662
	ret = 0;
663 664
out:
	btrfs_free_path(path);
665
	return ret;
666 667
}

668
int btrfs_csum_file_blocks(struct btrfs_trans_handle *trans,
669
			   struct btrfs_root *root,
670
			   struct btrfs_ordered_sum *sums)
C
Chris Mason 已提交
671
{
672
	u64 bytenr;
C
Chris Mason 已提交
673 674
	int ret;
	struct btrfs_key file_key;
675
	struct btrfs_key found_key;
676
	u64 next_offset;
677
	u64 total_bytes = 0;
678
	int found_next;
679
	struct btrfs_path *path;
C
Chris Mason 已提交
680
	struct btrfs_csum_item *item;
681
	struct btrfs_csum_item *item_end;
682
	struct extent_buffer *leaf = NULL;
683
	u64 csum_offset;
684
	struct btrfs_sector_sum *sector_sum;
685 686
	u32 nritems;
	u32 ins_size;
687
	u16 csum_size = btrfs_super_csum_size(root->fs_info->super_copy);
688

689
	path = btrfs_alloc_path();
690 691 692
	if (!path)
		return -ENOMEM;

693
	sector_sum = sums->sums;
694
	trans->adding_csums = 1;
695 696 697
again:
	next_offset = (u64)-1;
	found_next = 0;
698 699 700 701
	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);
702

703
	item = btrfs_lookup_csum(trans, root, path, sector_sum->bytenr, 1);
704 705
	if (!IS_ERR(item)) {
		leaf = path->nodes[0];
706
		ret = 0;
707
		goto found;
708
	}
709
	ret = PTR_ERR(item);
710 711 712
	if (ret != -EFBIG && ret != -ENOENT)
		goto fail_unlock;

713 714 715
	if (ret == -EFBIG) {
		u32 item_size;
		/* we found one, but it isn't big enough yet */
716 717
		leaf = path->nodes[0];
		item_size = btrfs_item_size_nr(leaf, path->slots[0]);
718 719
		if ((item_size / csum_size) >=
		    MAX_CSUM_ITEMS(root, csum_size)) {
720 721 722 723
			/* already at max size, make a new one */
			goto insert;
		}
	} else {
724
		int slot = path->slots[0] + 1;
725
		/* we didn't find a csum item, insert one */
726 727 728
		nritems = btrfs_header_nritems(path->nodes[0]);
		if (path->slots[0] >= nritems - 1) {
			ret = btrfs_next_leaf(root, path);
Y
Yan 已提交
729
			if (ret == 1)
730
				found_next = 1;
Y
Yan 已提交
731
			if (ret != 0)
732
				goto insert;
Y
Yan 已提交
733
			slot = 0;
734 735
		}
		btrfs_item_key_to_cpu(path->nodes[0], &found_key, slot);
736 737
		if (found_key.objectid != BTRFS_EXTENT_CSUM_OBJECTID ||
		    found_key.type != BTRFS_EXTENT_CSUM_KEY) {
738 739 740 741 742
			found_next = 1;
			goto insert;
		}
		next_offset = found_key.offset;
		found_next = 1;
743 744 745 746 747 748 749
		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
	 */
750
	btrfs_release_path(path);
751
	ret = btrfs_search_slot(trans, root, &file_key, path,
752
				csum_size, 1);
753
	if (ret < 0)
754
		goto fail_unlock;
755 756 757 758 759

	if (ret > 0) {
		if (path->slots[0] == 0)
			goto insert;
		path->slots[0]--;
760
	}
761

762 763
	leaf = path->nodes[0];
	btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
764
	csum_offset = (bytenr - found_key.offset) >>
765
			root->fs_info->sb->s_blocksize_bits;
766

767 768
	if (btrfs_key_type(&found_key) != BTRFS_EXTENT_CSUM_KEY ||
	    found_key.objectid != BTRFS_EXTENT_CSUM_OBJECTID ||
769
	    csum_offset >= MAX_CSUM_ITEMS(root, csum_size)) {
770 771
		goto insert;
	}
772

773
	if (csum_offset >= btrfs_item_size_nr(leaf, path->slots[0]) /
774 775
	    csum_size) {
		u32 diff = (csum_offset + 1) * csum_size;
776 777 778 779 780 781 782 783

		/*
		 * 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;

784
		diff = diff - btrfs_item_size_nr(leaf, path->slots[0]);
C
Chris Mason 已提交
785
		if (diff != csum_size)
C
Chris Mason 已提交
786
			goto insert;
787

788
		btrfs_extend_item(trans, root, path, diff);
789 790 791 792
		goto csum;
	}

insert:
793
	btrfs_release_path(path);
794
	csum_offset = 0;
795
	if (found_next) {
796 797 798 799
		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
800
		while (tmp < sums->len) {
801 802 803 804 805 806 807
			if (next_sector + root->sectorsize != next->bytenr)
				break;
			tmp += root->sectorsize;
			next_sector = next->bytenr;
			next++;
		}
		tmp = min(tmp, next_offset - file_key.offset);
808 809
		tmp >>= root->fs_info->sb->s_blocksize_bits;
		tmp = max((u64)1, tmp);
810 811
		tmp = min(tmp, (u64)MAX_CSUM_ITEMS(root, csum_size));
		ins_size = csum_size * tmp;
812
	} else {
813
		ins_size = csum_size;
814
	}
815
	path->leave_spinning = 1;
816
	ret = btrfs_insert_empty_item(trans, root, path, &file_key,
817
				      ins_size);
818
	path->leave_spinning = 0;
819
	if (ret < 0)
820
		goto fail_unlock;
821 822
	if (ret != 0) {
		WARN_ON(1);
823
		goto fail_unlock;
824
	}
825
csum:
826 827
	leaf = path->nodes[0];
	item = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_csum_item);
C
Chris Mason 已提交
828
	ret = 0;
829
	item = (struct btrfs_csum_item *)((unsigned char *)item +
830
					  csum_offset * csum_size);
C
Chris Mason 已提交
831
found:
832 833 834
	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]));
835
next_sector:
836

837
	write_extent_buffer(leaf, &sector_sum->sum, (unsigned long)item, csum_size);
838

839 840 841
	total_bytes += root->sectorsize;
	sector_sum++;
	if (total_bytes < sums->len) {
842
		item = (struct btrfs_csum_item *)((char *)item +
843
						  csum_size);
844 845 846
		if (item < item_end && bytenr + PAGE_CACHE_SIZE ==
		    sector_sum->bytenr) {
			bytenr = sector_sum->bytenr;
847
			goto next_sector;
848
		}
849
	}
850

851
	btrfs_mark_buffer_dirty(path->nodes[0]);
852
	if (total_bytes < sums->len) {
853
		btrfs_release_path(path);
854
		cond_resched();
855 856
		goto again;
	}
857
out:
858
	trans->adding_csums = 0;
859
	btrfs_free_path(path);
C
Chris Mason 已提交
860
	return ret;
861 862 863

fail_unlock:
	goto out;
C
Chris Mason 已提交
864
}