file-item.c 22.6 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;
C
Chris Mason 已提交
62
	BUG_ON(ret);
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 95
	u16 csum_size =
		btrfs_super_csum_size(&root->fs_info->super_copy);
96
	int csums_in_item;
97

98 99 100
	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
101
	ret = btrfs_search_slot(trans, root, &file_key, path, 0, cow);
102 103
	if (ret < 0)
		goto fail;
104
	leaf = path->nodes[0];
105 106
	if (ret > 0) {
		ret = 1;
107
		if (path->slots[0] == 0)
108 109
			goto fail;
		path->slots[0]--;
110
		btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
111
		if (btrfs_key_type(&found_key) != BTRFS_EXTENT_CSUM_KEY)
112
			goto fail;
113 114

		csum_offset = (bytenr - found_key.offset) >>
115
				root->fs_info->sb->s_blocksize_bits;
116
		csums_in_item = btrfs_item_size_nr(leaf, path->slots[0]);
117
		csums_in_item /= csum_size;
118 119 120

		if (csum_offset >= csums_in_item) {
			ret = -EFBIG;
121 122 123 124
			goto fail;
		}
	}
	item = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_csum_item);
125
	item = (struct btrfs_csum_item *)((unsigned char *)item +
126
					  csum_offset * csum_size);
127 128 129
	return item;
fail:
	if (ret > 0)
C
Chris Mason 已提交
130
		ret = -ENOENT;
131 132 133 134
	return ERR_PTR(ret);
}


C
Chris Mason 已提交
135 136 137
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
138
			     u64 offset, int mod)
C
Chris Mason 已提交
139 140 141 142 143 144 145
{
	int ret;
	struct btrfs_key file_key;
	int ins_len = mod < 0 ? -1 : 0;
	int cow = mod != 0;

	file_key.objectid = objectid;
146
	file_key.offset = offset;
C
Chris Mason 已提交
147 148 149 150
	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
151

152

153 154 155
static int __btrfs_lookup_bio_sums(struct btrfs_root *root,
				   struct inode *inode, struct bio *bio,
				   u64 logical_offset, u32 *dst, int dio)
156 157 158 159
{
	u32 sum;
	struct bio_vec *bvec = bio->bi_io_vec;
	int bio_index = 0;
160
	u64 offset = 0;
161 162
	u64 item_start_offset = 0;
	u64 item_last_offset = 0;
163
	u64 disk_bytenr;
164
	u32 diff;
165 166
	u16 csum_size =
		btrfs_super_csum_size(&root->fs_info->super_copy);
167 168 169 170 171 172
	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();
173 174
	if (!path)
		return -ENOMEM;
175 176
	if (bio->bi_size > PAGE_CACHE_SIZE * 8)
		path->reada = 2;
177 178 179

	WARN_ON(bio->bi_vcnt <= 0);

180
	disk_bytenr = (u64)bio->bi_sector << 9;
181 182
	if (dio)
		offset = logical_offset;
C
Chris Mason 已提交
183
	while (bio_index < bio->bi_vcnt) {
184 185
		if (!dio)
			offset = page_offset(bvec->bv_page) + bvec->bv_offset;
186
		ret = btrfs_find_ordered_sum(inode, offset, disk_bytenr, &sum);
187 188 189
		if (ret == 0)
			goto found;

190 191
		if (!item || disk_bytenr < item_start_offset ||
		    disk_bytenr >= item_last_offset) {
192 193 194 195 196
			struct btrfs_key found_key;
			u32 item_size;

			if (item)
				btrfs_release_path(root, path);
197 198
			item = btrfs_lookup_csum(NULL, root->fs_info->csum_root,
						 path, disk_bytenr, 0);
199 200 201 202 203
			if (IS_ERR(item)) {
				ret = PTR_ERR(item);
				if (ret == -ENOENT || ret == -EFBIG)
					ret = 0;
				sum = 0;
204 205 206 207 208 209
				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
210 211 212
					printk(KERN_INFO "btrfs no csum found "
					       "for inode %lu start %llu\n",
					       inode->i_ino,
213 214
					       (unsigned long long)offset);
				}
215
				item = NULL;
216
				btrfs_release_path(root, path);
217 218 219 220 221 222 223 224 225
				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 +
226
				(item_size / csum_size) *
227 228 229 230 231 232 233 234
				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
		 */
235
		diff = disk_bytenr - item_start_offset;
236
		diff = diff / root->sectorsize;
237
		diff = diff * csum_size;
238 239

		read_extent_buffer(path->nodes[0], &sum,
240
				   ((unsigned long)item) + diff,
241
				   csum_size);
242
found:
243 244 245 246 247
		if (dst)
			*dst++ = sum;
		else
			set_state_private(io_tree, offset, sum);
		disk_bytenr += bvec->bv_len;
248
		offset += bvec->bv_len;
249 250 251 252 253 254 255
		bio_index++;
		bvec++;
	}
	btrfs_free_path(path);
	return 0;
}

256 257 258 259 260 261 262 263 264 265 266 267
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);
}

268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289
int btrfs_lookup_csums_range(struct btrfs_root *root, u64 start, u64 end,
			     struct list_head *list)
{
	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;
	unsigned long offset;
	int ret;
	size_t size;
	u64 csum_end;
	u16 csum_size = btrfs_super_csum_size(&root->fs_info->super_copy);

	path = btrfs_alloc_path();
	BUG_ON(!path);

	key.objectid = BTRFS_EXTENT_CSUM_OBJECTID;
	key.offset = start;
	key.type = BTRFS_EXTENT_CSUM_KEY;

Y
Yan Zheng 已提交
290
	ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308
	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
309
			ret = btrfs_next_leaf(root, path);
310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330
			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;
331 332 333 334
		if (csum_end <= start) {
			path->slots[0]++;
			continue;
		}
335

Y
Yan Zheng 已提交
336
		csum_end = min(csum_end, end + 1);
337 338
		item = btrfs_item_ptr(path->nodes[0], path->slots[0],
				      struct btrfs_csum_item);
Y
Yan Zheng 已提交
339 340 341 342 343 344
		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);
			BUG_ON(!sums);
345

Y
Yan Zheng 已提交
346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367
			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++;
			}
			list_add_tail(&sums->list, list);
		}
368 369 370 371 372 373 374 375
		path->slots[0]++;
	}
	ret = 0;
fail:
	btrfs_free_path(path);
	return ret;
}

376
int btrfs_csum_one_bio(struct btrfs_root *root, struct inode *inode,
377
		       struct bio *bio, u64 file_start, int contig)
378
{
379 380
	struct btrfs_ordered_sum *sums;
	struct btrfs_sector_sum *sector_sum;
381
	struct btrfs_ordered_extent *ordered;
382 383 384
	char *data;
	struct bio_vec *bvec = bio->bi_io_vec;
	int bio_index = 0;
385 386 387
	unsigned long total_bytes = 0;
	unsigned long this_sum_bytes = 0;
	u64 offset;
388
	u64 disk_bytenr;
389

390 391
	WARN_ON(bio->bi_vcnt <= 0);
	sums = kzalloc(btrfs_ordered_sum_size(root, bio->bi_size), GFP_NOFS);
392 393
	if (!sums)
		return -ENOMEM;
394

395
	sector_sum = sums->sums;
396
	disk_bytenr = (u64)bio->bi_sector << 9;
397 398
	sums->len = bio->bi_size;
	INIT_LIST_HEAD(&sums->list);
399 400 401 402 403 404 405

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

	ordered = btrfs_lookup_ordered_extent(inode, offset);
406
	BUG_ON(!ordered);
407
	sums->bytenr = ordered->start;
408

C
Chris Mason 已提交
409
	while (bio_index < bio->bi_vcnt) {
410 411 412 413 414
		if (!contig)
			offset = page_offset(bvec->bv_page) + bvec->bv_offset;

		if (!contig && (offset >= ordered->file_offset + ordered->len ||
		    offset < ordered->file_offset)) {
415 416 417 418 419 420 421 422 423 424 425
			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);
			BUG_ON(!sums);
426
			sector_sum = sums->sums;
427
			sums->len = bytes_left;
428
			ordered = btrfs_lookup_ordered_extent(inode, offset);
429
			BUG_ON(!ordered);
430
			sums->bytenr = ordered->start;
431 432
		}

433
		data = kmap_atomic(bvec->bv_page, KM_USER0);
434 435 436 437 438
		sector_sum->sum = ~(u32)0;
		sector_sum->sum = btrfs_csum_data(root,
						  data + bvec->bv_offset,
						  sector_sum->sum,
						  bvec->bv_len);
439
		kunmap_atomic(data, KM_USER0);
440 441
		btrfs_csum_final(sector_sum->sum,
				 (char *)&sector_sum->sum);
442
		sector_sum->bytenr = disk_bytenr;
443

444
		sector_sum++;
445
		bio_index++;
446 447
		total_bytes += bvec->bv_len;
		this_sum_bytes += bvec->bv_len;
448 449
		disk_bytenr += bvec->bv_len;
		offset += bvec->bv_len;
450 451
		bvec++;
	}
452
	this_sum_bytes = 0;
453 454
	btrfs_add_ordered_sum(inode, ordered, sums);
	btrfs_put_ordered_extent(ordered);
455 456 457
	return 0;
}

458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541
/*
 * 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.
 */
static noinline int truncate_one_csum(struct btrfs_trans_handle *trans,
				      struct btrfs_root *root,
				      struct btrfs_path *path,
				      struct btrfs_key *key,
				      u64 bytenr, u64 len)
{
	struct extent_buffer *leaf;
	u16 csum_size =
		btrfs_super_csum_size(&root->fs_info->super_copy);
	u64 csum_end;
	u64 end_byte = bytenr + len;
	u32 blocksize_bits = root->fs_info->sb->s_blocksize_bits;
	int ret;

	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;
		ret = btrfs_truncate_item(trans, root, path, new_size, 1);
		BUG_ON(ret);
	} 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;

		ret = btrfs_truncate_item(trans, root, path, new_size, 0);
		BUG_ON(ret);

		key->offset = end_byte;
		ret = btrfs_set_item_key_safe(trans, root, path, key);
		BUG_ON(ret);
	} else {
		BUG();
	}
	return 0;
}

/*
 * 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;
	u16 csum_size =
		btrfs_super_csum_size(&root->fs_info->super_copy);
	int blocksize_bits = root->fs_info->sb->s_blocksize_bits;

	root = root->fs_info->csum_root;

	path = btrfs_alloc_path();
542 543
	if (!path)
		return -ENOMEM;
544

C
Chris Mason 已提交
545
	while (1) {
546 547 548 549
		key.objectid = BTRFS_EXTENT_CSUM_OBJECTID;
		key.offset = end_byte - 1;
		key.type = BTRFS_EXTENT_CSUM_KEY;

550
		path->leave_spinning = 1;
551 552 553 554 555
		ret = btrfs_search_slot(trans, root, &key, path, -1, 1);
		if (ret > 0) {
			if (path->slots[0] == 0)
				goto out;
			path->slots[0]--;
556 557
		} else if (ret < 0) {
			goto out;
558
		}
559

560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582
		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);
			BUG_ON(ret);
583 584
			if (key.offset == bytenr)
				break;
585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630
		} 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);
			BUG_ON(ret && ret != -EAGAIN);

			key.offset = end_byte - 1;
		} else {
			ret = truncate_one_csum(trans, root, path,
						&key, bytenr, len);
			BUG_ON(ret);
631 632
			if (key.offset < bytenr)
				break;
633 634 635 636 637 638 639 640
		}
		btrfs_release_path(root, path);
	}
out:
	btrfs_free_path(path);
	return 0;
}

641
int btrfs_csum_file_blocks(struct btrfs_trans_handle *trans,
642
			   struct btrfs_root *root,
643
			   struct btrfs_ordered_sum *sums)
C
Chris Mason 已提交
644
{
645
	u64 bytenr;
C
Chris Mason 已提交
646 647
	int ret;
	struct btrfs_key file_key;
648
	struct btrfs_key found_key;
649
	u64 next_offset;
650
	u64 total_bytes = 0;
651
	int found_next;
652
	struct btrfs_path *path;
C
Chris Mason 已提交
653
	struct btrfs_csum_item *item;
654
	struct btrfs_csum_item *item_end;
655
	struct extent_buffer *leaf = NULL;
656
	u64 csum_offset;
657
	struct btrfs_sector_sum *sector_sum;
658 659
	u32 nritems;
	u32 ins_size;
660 661 662 663
	char *eb_map;
	char *eb_token;
	unsigned long map_len;
	unsigned long map_start;
664 665
	u16 csum_size =
		btrfs_super_csum_size(&root->fs_info->super_copy);
666

667 668
	path = btrfs_alloc_path();
	BUG_ON(!path);
669
	sector_sum = sums->sums;
670 671 672
again:
	next_offset = (u64)-1;
	found_next = 0;
673 674 675 676
	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);
677

678
	item = btrfs_lookup_csum(trans, root, path, sector_sum->bytenr, 1);
679 680
	if (!IS_ERR(item)) {
		leaf = path->nodes[0];
681
		ret = 0;
682
		goto found;
683
	}
684
	ret = PTR_ERR(item);
685 686 687
	if (ret != -EFBIG && ret != -ENOENT)
		goto fail_unlock;

688 689 690
	if (ret == -EFBIG) {
		u32 item_size;
		/* we found one, but it isn't big enough yet */
691 692
		leaf = path->nodes[0];
		item_size = btrfs_item_size_nr(leaf, path->slots[0]);
693 694
		if ((item_size / csum_size) >=
		    MAX_CSUM_ITEMS(root, csum_size)) {
695 696 697 698
			/* already at max size, make a new one */
			goto insert;
		}
	} else {
699
		int slot = path->slots[0] + 1;
700
		/* we didn't find a csum item, insert one */
701 702 703
		nritems = btrfs_header_nritems(path->nodes[0]);
		if (path->slots[0] >= nritems - 1) {
			ret = btrfs_next_leaf(root, path);
Y
Yan 已提交
704
			if (ret == 1)
705
				found_next = 1;
Y
Yan 已提交
706
			if (ret != 0)
707
				goto insert;
Y
Yan 已提交
708
			slot = 0;
709 710
		}
		btrfs_item_key_to_cpu(path->nodes[0], &found_key, slot);
711 712
		if (found_key.objectid != BTRFS_EXTENT_CSUM_OBJECTID ||
		    found_key.type != BTRFS_EXTENT_CSUM_KEY) {
713 714 715 716 717
			found_next = 1;
			goto insert;
		}
		next_offset = found_key.offset;
		found_next = 1;
718 719 720 721 722 723 724 725
		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
	 */
	btrfs_release_path(root, path);
726
	ret = btrfs_search_slot(trans, root, &file_key, path,
727
				csum_size, 1);
728
	if (ret < 0)
729
		goto fail_unlock;
730 731 732 733 734

	if (ret > 0) {
		if (path->slots[0] == 0)
			goto insert;
		path->slots[0]--;
735
	}
736

737 738
	leaf = path->nodes[0];
	btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
739
	csum_offset = (bytenr - found_key.offset) >>
740
			root->fs_info->sb->s_blocksize_bits;
741

742 743
	if (btrfs_key_type(&found_key) != BTRFS_EXTENT_CSUM_KEY ||
	    found_key.objectid != BTRFS_EXTENT_CSUM_OBJECTID ||
744
	    csum_offset >= MAX_CSUM_ITEMS(root, csum_size)) {
745 746
		goto insert;
	}
747

748
	if (csum_offset >= btrfs_item_size_nr(leaf, path->slots[0]) /
749 750
	    csum_size) {
		u32 diff = (csum_offset + 1) * csum_size;
751 752 753 754 755 756 757 758

		/*
		 * 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;

759
		diff = diff - btrfs_item_size_nr(leaf, path->slots[0]);
C
Chris Mason 已提交
760
		if (diff != csum_size)
C
Chris Mason 已提交
761
			goto insert;
762

763
		ret = btrfs_extend_item(trans, root, path, diff);
764 765 766 767 768
		BUG_ON(ret);
		goto csum;
	}

insert:
769
	btrfs_release_path(root, path);
770
	csum_offset = 0;
771
	if (found_next) {
772 773 774 775
		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
776
		while (tmp < sums->len) {
777 778 779 780 781 782 783
			if (next_sector + root->sectorsize != next->bytenr)
				break;
			tmp += root->sectorsize;
			next_sector = next->bytenr;
			next++;
		}
		tmp = min(tmp, next_offset - file_key.offset);
784 785
		tmp >>= root->fs_info->sb->s_blocksize_bits;
		tmp = max((u64)1, tmp);
786 787
		tmp = min(tmp, (u64)MAX_CSUM_ITEMS(root, csum_size));
		ins_size = csum_size * tmp;
788
	} else {
789
		ins_size = csum_size;
790
	}
791
	path->leave_spinning = 1;
792
	ret = btrfs_insert_empty_item(trans, root, path, &file_key,
793
				      ins_size);
794
	path->leave_spinning = 0;
795
	if (ret < 0)
796
		goto fail_unlock;
797 798
	if (ret != 0) {
		WARN_ON(1);
799
		goto fail_unlock;
800
	}
801
csum:
802 803
	leaf = path->nodes[0];
	item = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_csum_item);
C
Chris Mason 已提交
804
	ret = 0;
805
	item = (struct btrfs_csum_item *)((unsigned char *)item +
806
					  csum_offset * csum_size);
C
Chris Mason 已提交
807
found:
808 809 810
	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]));
811
	eb_token = NULL;
812
next_sector:
813

814
	if (!eb_token ||
815
	   (unsigned long)item + csum_size >= map_start + map_len) {
816 817 818
		int err;

		if (eb_token)
819
			unmap_extent_buffer(leaf, eb_token, KM_USER1);
820 821
		eb_token = NULL;
		err = map_private_extent_buffer(leaf, (unsigned long)item,
822
						csum_size,
823
						&eb_token, &eb_map,
824
						&map_start, &map_len, KM_USER1);
825 826 827 828 829
		if (err)
			eb_token = NULL;
	}
	if (eb_token) {
		memcpy(eb_token + ((unsigned long)item & (PAGE_CACHE_SIZE - 1)),
830
		       &sector_sum->sum, csum_size);
831
	} else {
832
		write_extent_buffer(leaf, &sector_sum->sum,
833
				    (unsigned long)item, csum_size);
834
	}
835

836 837 838
	total_bytes += root->sectorsize;
	sector_sum++;
	if (total_bytes < sums->len) {
839
		item = (struct btrfs_csum_item *)((char *)item +
840
						  csum_size);
841 842 843
		if (item < item_end && bytenr + PAGE_CACHE_SIZE ==
		    sector_sum->bytenr) {
			bytenr = sector_sum->bytenr;
844
			goto next_sector;
845
		}
846
	}
847
	if (eb_token) {
848
		unmap_extent_buffer(leaf, eb_token, KM_USER1);
849 850
		eb_token = NULL;
	}
851
	btrfs_mark_buffer_dirty(path->nodes[0]);
852
	if (total_bytes < sums->len) {
853
		btrfs_release_path(root, path);
854
		cond_resched();
855 856
		goto again;
	}
857
out:
858
	btrfs_free_path(path);
C
Chris Mason 已提交
859
	return ret;
860 861 862

fail_unlock:
	goto out;
C
Chris Mason 已提交
863
}