super.c 80.2 KB
Newer Older
1
#include <linux/module.h>
C
Chris Mason 已提交
2
#include <linux/buffer_head.h>
3 4 5 6 7 8 9 10
#include <linux/fs.h>
#include <linux/pagemap.h>
#include <linux/highmem.h>
#include <linux/time.h>
#include <linux/init.h>
#include <linux/string.h>
#include <linux/smp_lock.h>
#include <linux/backing-dev.h>
C
Chris Mason 已提交
11
#include <linux/mpage.h>
C
Chris Mason 已提交
12 13
#include <linux/swap.h>
#include <linux/writeback.h>
C
Chris Mason 已提交
14
#include <linux/statfs.h>
15
#include "ctree.h"
C
Chris Mason 已提交
16
#include "disk-io.h"
17
#include "transaction.h"
C
Chris Mason 已提交
18
#include "btrfs_inode.h"
C
Chris Mason 已提交
19
#include "ioctl.h"
C
Chris Mason 已提交
20
#include "print-tree.h"
21

C
Chris Mason 已提交
22 23 24 25 26
struct btrfs_iget_args {
	u64 ino;
	struct btrfs_root *root;
};

27
#define BTRFS_SUPER_MAGIC 0x9123682E
C
Chris Mason 已提交
28 29

static struct inode_operations btrfs_dir_inode_operations;
C
Chris Mason 已提交
30
static struct inode_operations btrfs_symlink_inode_operations;
31
static struct inode_operations btrfs_dir_ro_inode_operations;
C
Chris Mason 已提交
32 33
static struct super_operations btrfs_super_ops;
static struct file_operations btrfs_dir_file_operations;
C
Chris Mason 已提交
34 35
static struct inode_operations btrfs_file_inode_operations;
static struct address_space_operations btrfs_aops;
C
Chris Mason 已提交
36
static struct address_space_operations btrfs_symlink_aops;
C
Chris Mason 已提交
37
static struct file_operations btrfs_file_operations;
C
Chris Mason 已提交
38

C
Chris Mason 已提交
39 40 41 42 43 44 45
static int drop_extents(struct btrfs_trans_handle *trans,
			  struct btrfs_root *root,
			  struct inode *inode,
			  u64 start, u64 end, u64 *hint_block);
static int btrfs_get_block(struct inode *inode, sector_t iblock,
			   struct buffer_head *result, int create);

C
Chris Mason 已提交
46 47 48 49 50 51 52 53 54 55 56 57

#define S_SHIFT 12
static unsigned char btrfs_type_by_mode[S_IFMT >> S_SHIFT] = {
	[S_IFREG >> S_SHIFT]	= BTRFS_FT_REG_FILE,
	[S_IFDIR >> S_SHIFT]	= BTRFS_FT_DIR,
	[S_IFCHR >> S_SHIFT]	= BTRFS_FT_CHRDEV,
	[S_IFBLK >> S_SHIFT]	= BTRFS_FT_BLKDEV,
	[S_IFIFO >> S_SHIFT]	= BTRFS_FT_FIFO,
	[S_IFSOCK >> S_SHIFT]	= BTRFS_FT_SOCK,
	[S_IFLNK >> S_SHIFT]	= BTRFS_FT_SYMLINK,
};

C
Chris Mason 已提交
58
static void btrfs_read_locked_inode(struct inode *inode)
59
{
60
	struct btrfs_path *path;
C
Chris Mason 已提交
61
	struct btrfs_inode_item *inode_item;
62 63
	struct btrfs_root *root = BTRFS_I(inode)->root;
	struct btrfs_key location;
64 65
	struct btrfs_block_group_cache *alloc_group;
	u64 alloc_group_block;
C
Chris Mason 已提交
66
	int ret;
C
Chris Mason 已提交
67

68 69 70
	path = btrfs_alloc_path();
	BUG_ON(!path);
	btrfs_init_path(path);
C
Chris Mason 已提交
71 72
	mutex_lock(&root->fs_info->fs_mutex);

73 74
	memcpy(&location, &BTRFS_I(inode)->location, sizeof(location));
	ret = btrfs_lookup_inode(NULL, root, path, &location, 0);
C
Chris Mason 已提交
75
	if (ret) {
C
Chris Mason 已提交
76
		btrfs_free_path(path);
77
		goto make_bad;
C
Chris Mason 已提交
78
	}
79 80
	inode_item = btrfs_item_ptr(btrfs_buffer_leaf(path->nodes[0]),
				  path->slots[0],
C
Chris Mason 已提交
81
				  struct btrfs_inode_item);
82

C
Chris Mason 已提交
83 84 85 86 87 88 89 90 91 92 93 94 95
	inode->i_mode = btrfs_inode_mode(inode_item);
	inode->i_nlink = btrfs_inode_nlink(inode_item);
	inode->i_uid = btrfs_inode_uid(inode_item);
	inode->i_gid = btrfs_inode_gid(inode_item);
	inode->i_size = btrfs_inode_size(inode_item);
	inode->i_atime.tv_sec = btrfs_timespec_sec(&inode_item->atime);
	inode->i_atime.tv_nsec = btrfs_timespec_nsec(&inode_item->atime);
	inode->i_mtime.tv_sec = btrfs_timespec_sec(&inode_item->mtime);
	inode->i_mtime.tv_nsec = btrfs_timespec_nsec(&inode_item->mtime);
	inode->i_ctime.tv_sec = btrfs_timespec_sec(&inode_item->ctime);
	inode->i_ctime.tv_nsec = btrfs_timespec_nsec(&inode_item->ctime);
	inode->i_blocks = btrfs_inode_nblocks(inode_item);
	inode->i_generation = btrfs_inode_generation(inode_item);
96 97 98 99 100 101
	alloc_group_block = btrfs_inode_block_group(inode_item);
	ret = radix_tree_gang_lookup(&root->fs_info->block_group_radix,
				     (void **)&alloc_group,
				     alloc_group_block, 1);
	BUG_ON(!ret);
	BTRFS_I(inode)->block_group = alloc_group;
102 103 104 105

	btrfs_free_path(path);
	inode_item = NULL;

C
Chris Mason 已提交
106
	mutex_unlock(&root->fs_info->fs_mutex);
C
Chris Mason 已提交
107

C
Chris Mason 已提交
108 109 110 111 112 113 114 115
	switch (inode->i_mode & S_IFMT) {
#if 0
	default:
		init_special_inode(inode, inode->i_mode,
				   btrfs_inode_rdev(inode_item));
		break;
#endif
	case S_IFREG:
C
Chris Mason 已提交
116 117 118
		inode->i_mapping->a_ops = &btrfs_aops;
		inode->i_fop = &btrfs_file_operations;
		inode->i_op = &btrfs_file_inode_operations;
C
Chris Mason 已提交
119 120 121
		break;
	case S_IFDIR:
		inode->i_fop = &btrfs_dir_file_operations;
122 123 124 125
		if (root == root->fs_info->tree_root)
			inode->i_op = &btrfs_dir_ro_inode_operations;
		else
			inode->i_op = &btrfs_dir_inode_operations;
C
Chris Mason 已提交
126 127
		break;
	case S_IFLNK:
C
Chris Mason 已提交
128 129
		inode->i_op = &btrfs_symlink_inode_operations;
		inode->i_mapping->a_ops = &btrfs_symlink_aops;
C
Chris Mason 已提交
130
		break;
131
	}
C
Chris Mason 已提交
132
	return;
133 134 135 136 137 138

make_bad:
	btrfs_release_path(root, path);
	btrfs_free_path(path);
	mutex_unlock(&root->fs_info->fs_mutex);
	make_bad_inode(inode);
139 140
}

141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156
static void fill_inode_item(struct btrfs_inode_item *item,
			    struct inode *inode)
{
	btrfs_set_inode_uid(item, inode->i_uid);
	btrfs_set_inode_gid(item, inode->i_gid);
	btrfs_set_inode_size(item, inode->i_size);
	btrfs_set_inode_mode(item, inode->i_mode);
	btrfs_set_inode_nlink(item, inode->i_nlink);
	btrfs_set_timespec_sec(&item->atime, inode->i_atime.tv_sec);
	btrfs_set_timespec_nsec(&item->atime, inode->i_atime.tv_nsec);
	btrfs_set_timespec_sec(&item->mtime, inode->i_mtime.tv_sec);
	btrfs_set_timespec_nsec(&item->mtime, inode->i_mtime.tv_nsec);
	btrfs_set_timespec_sec(&item->ctime, inode->i_ctime.tv_sec);
	btrfs_set_timespec_nsec(&item->ctime, inode->i_ctime.tv_nsec);
	btrfs_set_inode_nblocks(item, inode->i_blocks);
	btrfs_set_inode_generation(item, inode->i_generation);
157 158
	btrfs_set_inode_block_group(item,
				    BTRFS_I(inode)->block_group->key.objectid);
159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193
}

static int btrfs_update_inode(struct btrfs_trans_handle *trans,
			      struct btrfs_root *root,
			      struct inode *inode)
{
	struct btrfs_inode_item *inode_item;
	struct btrfs_path *path;
	int ret;

	path = btrfs_alloc_path();
	BUG_ON(!path);
	btrfs_init_path(path);
	ret = btrfs_lookup_inode(trans, root, path,
				 &BTRFS_I(inode)->location, 1);
	if (ret) {
		if (ret > 0)
			ret = -ENOENT;
		goto failed;
	}

	inode_item = btrfs_item_ptr(btrfs_buffer_leaf(path->nodes[0]),
				  path->slots[0],
				  struct btrfs_inode_item);

	fill_inode_item(inode_item, inode);
	btrfs_mark_buffer_dirty(path->nodes[0]);
	ret = 0;
failed:
	btrfs_release_path(root, path);
	btrfs_free_path(path);
	return ret;
}


C
Chris Mason 已提交
194 195 196 197
static int btrfs_unlink_trans(struct btrfs_trans_handle *trans,
			      struct btrfs_root *root,
			      struct inode *dir,
			      struct dentry *dentry)
C
Chris Mason 已提交
198
{
199
	struct btrfs_path *path;
C
Chris Mason 已提交
200 201
	const char *name = dentry->d_name.name;
	int name_len = dentry->d_name.len;
202
	int ret = 0;
C
Chris Mason 已提交
203 204 205
	u64 objectid;
	struct btrfs_dir_item *di;

206 207 208
	path = btrfs_alloc_path();
	BUG_ON(!path);
	btrfs_init_path(path);
209
	di = btrfs_lookup_dir_item(trans, root, path, dir->i_ino,
C
Chris Mason 已提交
210
				    name, name_len, -1);
211 212
	if (IS_ERR(di)) {
		ret = PTR_ERR(di);
C
Chris Mason 已提交
213
		goto err;
214 215
	}
	if (!di) {
C
Chris Mason 已提交
216 217 218
		ret = -ENOENT;
		goto err;
	}
219
	objectid = btrfs_disk_key_objectid(&di->location);
220 221 222
	ret = btrfs_delete_one_dir_name(trans, root, path, di);
	BUG_ON(ret);
	btrfs_release_path(root, path);
C
Chris Mason 已提交
223

224 225 226 227 228 229 230 231 232 233 234
	di = btrfs_lookup_dir_index_item(trans, root, path, dir->i_ino,
					 objectid, name, name_len, -1);
	if (IS_ERR(di)) {
		ret = PTR_ERR(di);
		goto err;
	}
	if (!di) {
		ret = -ENOENT;
		goto err;
	}
	ret = btrfs_delete_one_dir_name(trans, root, path, di);
235 236
	BUG_ON(ret);

C
Chris Mason 已提交
237 238
	dentry->d_inode->i_ctime = dir->i_ctime;
err:
239
	btrfs_free_path(path);
240
	if (!ret) {
241
		dir->i_size -= name_len * 2;
242 243 244
		btrfs_update_inode(trans, root, dir);
		drop_nlink(dentry->d_inode);
		btrfs_update_inode(trans, root, dentry->d_inode);
C
Chris Mason 已提交
245
		dir->i_sb->s_dirt = 1;
246
	}
C
Chris Mason 已提交
247 248 249
	return ret;
}

C
Chris Mason 已提交
250 251 252 253 254 255
static int btrfs_unlink(struct inode *dir, struct dentry *dentry)
{
	struct btrfs_root *root;
	struct btrfs_trans_handle *trans;
	int ret;

256
	root = BTRFS_I(dir)->root;
C
Chris Mason 已提交
257 258
	mutex_lock(&root->fs_info->fs_mutex);
	trans = btrfs_start_transaction(root, 1);
259
	btrfs_set_trans_block_group(trans, dir);
C
Chris Mason 已提交
260 261 262
	ret = btrfs_unlink_trans(trans, root, dir, dentry);
	btrfs_end_transaction(trans, root);
	mutex_unlock(&root->fs_info->fs_mutex);
C
Chris Mason 已提交
263
	btrfs_btree_balance_dirty(root);
C
Chris Mason 已提交
264 265 266 267 268 269 270 271
	return ret;
}

static int btrfs_rmdir(struct inode *dir, struct dentry *dentry)
{
	struct inode *inode = dentry->d_inode;
	int err;
	int ret;
272
	struct btrfs_root *root = BTRFS_I(dir)->root;
273
	struct btrfs_path *path;
C
Chris Mason 已提交
274 275
	struct btrfs_key key;
	struct btrfs_trans_handle *trans;
276 277
	struct btrfs_key found_key;
	int found_type;
C
Chris Mason 已提交
278
	struct btrfs_leaf *leaf;
279
	char *goodnames = "..";
C
Chris Mason 已提交
280

281 282 283
	path = btrfs_alloc_path();
	BUG_ON(!path);
	btrfs_init_path(path);
C
Chris Mason 已提交
284 285
	mutex_lock(&root->fs_info->fs_mutex);
	trans = btrfs_start_transaction(root, 1);
286
	btrfs_set_trans_block_group(trans, dir);
C
Chris Mason 已提交
287 288
	key.objectid = inode->i_ino;
	key.offset = (u64)-1;
289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318
	key.flags = (u32)-1;
	while(1) {
		ret = btrfs_search_slot(trans, root, &key, path, -1, 1);
		if (ret < 0) {
			err = ret;
			goto out;
		}
		BUG_ON(ret == 0);
		if (path->slots[0] == 0) {
			err = -ENOENT;
			goto out;
		}
		path->slots[0]--;
		leaf = btrfs_buffer_leaf(path->nodes[0]);
		btrfs_disk_key_to_cpu(&found_key,
				      &leaf->items[path->slots[0]].key);
		found_type = btrfs_key_type(&found_key);
		if (found_key.objectid != inode->i_ino) {
			err = -ENOENT;
			goto out;
		}
		if ((found_type != BTRFS_DIR_ITEM_KEY &&
		     found_type != BTRFS_DIR_INDEX_KEY) ||
	            (!btrfs_match_dir_item_name(root, path, goodnames, 2) &&
	            !btrfs_match_dir_item_name(root, path, goodnames, 1))) {
			err = -ENOTEMPTY;
			goto out;
		}
		ret = btrfs_del_item(trans, root, path);
		BUG_ON(ret);
C
Chris Mason 已提交
319

320 321 322
		if (found_type == BTRFS_DIR_ITEM_KEY && found_key.offset == 1)
			break;
		btrfs_release_path(root, path);
C
Chris Mason 已提交
323
	}
324
	ret = 0;
325
	btrfs_release_path(root, path);
C
Chris Mason 已提交
326 327 328 329 330 331 332

	/* now the directory is empty */
	err = btrfs_unlink_trans(trans, root, dir, dentry);
	if (!err) {
		inode->i_size = 0;
	}
out:
C
Chris Mason 已提交
333 334
	btrfs_release_path(root, path);
	btrfs_free_path(path);
C
Chris Mason 已提交
335 336
	mutex_unlock(&root->fs_info->fs_mutex);
	ret = btrfs_end_transaction(trans, root);
C
Chris Mason 已提交
337
	btrfs_btree_balance_dirty(root);
C
Chris Mason 已提交
338 339 340 341 342
	if (ret && !err)
		err = ret;
	return err;
}

C
Chris Mason 已提交
343 344 345 346
static int btrfs_free_inode(struct btrfs_trans_handle *trans,
			    struct btrfs_root *root,
			    struct inode *inode)
{
347
	struct btrfs_path *path;
C
Chris Mason 已提交
348
	int ret;
349

C
Chris Mason 已提交
350
	clear_inode(inode);
351 352 353 354

	path = btrfs_alloc_path();
	BUG_ON(!path);
	btrfs_init_path(path);
355 356
	ret = btrfs_lookup_inode(trans, root, path,
				 &BTRFS_I(inode)->location, -1);
C
Chris Mason 已提交
357
	BUG_ON(ret);
358
	ret = btrfs_del_item(trans, root, path);
C
Chris Mason 已提交
359
	BUG_ON(ret);
360
	btrfs_free_path(path);
C
Chris Mason 已提交
361 362 363
	return ret;
}

364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392
static void reada_truncate(struct btrfs_root *root, struct btrfs_path *path,
			   u64 objectid)
{
	struct btrfs_node *node;
	int i;
	int nritems;
	u64 item_objectid;
	u64 blocknr;
	int slot;
	int ret;

	if (!path->nodes[1])
		return;
	node = btrfs_buffer_node(path->nodes[1]);
	slot = path->slots[1];
	if (slot == 0)
		return;
	nritems = btrfs_header_nritems(&node->header);
	for (i = slot - 1; i >= 0; i--) {
		item_objectid = btrfs_disk_key_objectid(&node->ptrs[i].key);
		if (item_objectid != objectid)
			break;
		blocknr = btrfs_node_blockptr(node, i);
		ret = readahead_tree_block(root, blocknr);
		if (ret)
			break;
	}
}

C
Chris Mason 已提交
393 394 395 396 397
static int btrfs_truncate_in_trans(struct btrfs_trans_handle *trans,
				   struct btrfs_root *root,
				   struct inode *inode)
{
	int ret;
398
	struct btrfs_path *path;
C
Chris Mason 已提交
399 400
	struct btrfs_key key;
	struct btrfs_disk_key *found_key;
C
Chris Mason 已提交
401
	u32 found_type;
C
Chris Mason 已提交
402
	struct btrfs_leaf *leaf;
C
Chris Mason 已提交
403
	struct btrfs_file_extent_item *fi;
C
Chris Mason 已提交
404 405
	u64 extent_start = 0;
	u64 extent_num_blocks = 0;
C
Chris Mason 已提交
406
	u64 item_end = 0;
C
Chris Mason 已提交
407
	int found_extent;
C
Chris Mason 已提交
408
	int del_item;
C
Chris Mason 已提交
409

410 411
	path = btrfs_alloc_path();
	BUG_ON(!path);
C
Chris Mason 已提交
412 413 414
	/* FIXME, add redo link to tree so we don't leak on crash */
	key.objectid = inode->i_ino;
	key.offset = (u64)-1;
C
Chris Mason 已提交
415
	key.flags = (u32)-1;
C
Chris Mason 已提交
416
	while(1) {
417
		btrfs_init_path(path);
C
Chris Mason 已提交
418
		fi = NULL;
419
		ret = btrfs_search_slot(trans, root, &key, path, -1, 1);
C
Chris Mason 已提交
420 421 422 423
		if (ret < 0) {
			goto error;
		}
		if (ret > 0) {
424 425
			BUG_ON(path->slots[0] == 0);
			path->slots[0]--;
C
Chris Mason 已提交
426
		}
427
		reada_truncate(root, path, inode->i_ino);
428 429
		leaf = btrfs_buffer_leaf(path->nodes[0]);
		found_key = &leaf->items[path->slots[0]].key;
C
Chris Mason 已提交
430
		found_type = btrfs_disk_key_type(found_key);
C
Chris Mason 已提交
431 432
		if (btrfs_disk_key_objectid(found_key) != inode->i_ino)
			break;
C
Chris Mason 已提交
433 434 435 436
		if (found_type != BTRFS_CSUM_ITEM_KEY &&
		    found_type != BTRFS_DIR_ITEM_KEY &&
		    found_type != BTRFS_DIR_INDEX_KEY &&
		    found_type != BTRFS_EXTENT_DATA_KEY)
C
Chris Mason 已提交
437
			break;
C
Chris Mason 已提交
438 439
		item_end = btrfs_disk_key_offset(found_key);
		if (found_type == BTRFS_EXTENT_DATA_KEY) {
440 441
			fi = btrfs_item_ptr(btrfs_buffer_leaf(path->nodes[0]),
					    path->slots[0],
C
Chris Mason 已提交
442
					    struct btrfs_file_extent_item);
443 444
			if (btrfs_file_extent_type(fi) !=
			    BTRFS_FILE_EXTENT_INLINE) {
C
Chris Mason 已提交
445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483
				item_end += btrfs_file_extent_num_blocks(fi) <<
						inode->i_blkbits;
			}
		}
		if (found_type == BTRFS_CSUM_ITEM_KEY) {
			ret = btrfs_csum_truncate(trans, root, path,
						  inode->i_size);
			BUG_ON(ret);
		}
		if (item_end < inode->i_size) {
			if (found_type) {
				btrfs_set_key_type(&key, found_type - 1);
				continue;
			}
			break;
		}
		if (btrfs_disk_key_offset(found_key) >= inode->i_size)
			del_item = 1;
		else
			del_item = 0;
		found_extent = 0;

		if (found_type == BTRFS_EXTENT_DATA_KEY &&
			   btrfs_file_extent_type(fi) !=
			   BTRFS_FILE_EXTENT_INLINE) {
			u64 num_dec;
			if (!del_item) {
				u64 orig_num_blocks =
					btrfs_file_extent_num_blocks(fi);
				extent_num_blocks = inode->i_size -
					btrfs_disk_key_offset(found_key) +
					root->blocksize - 1;
				extent_num_blocks >>= inode->i_blkbits;
				btrfs_set_file_extent_num_blocks(fi,
							 extent_num_blocks);
				inode->i_blocks -= (orig_num_blocks -
					extent_num_blocks) << 3;
				mark_buffer_dirty(path->nodes[0]);
			} else {
484 485 486 487 488
				extent_start =
					btrfs_file_extent_disk_blocknr(fi);
				extent_num_blocks =
					btrfs_file_extent_disk_num_blocks(fi);
				/* FIXME blocksize != 4096 */
C
Chris Mason 已提交
489 490 491 492 493
				num_dec = btrfs_file_extent_num_blocks(fi) << 3;
				if (extent_start != 0) {
					found_extent = 1;
					inode->i_blocks -= num_dec;
				}
494
			}
C
Chris Mason 已提交
495
		}
C
Chris Mason 已提交
496 497 498 499 500 501
		if (del_item) {
			ret = btrfs_del_item(trans, root, path);
			BUG_ON(ret);
		} else {
			break;
		}
502
		btrfs_release_path(root, path);
C
Chris Mason 已提交
503 504 505 506 507
		if (found_extent) {
			ret = btrfs_free_extent(trans, root, extent_start,
						extent_num_blocks, 0);
			BUG_ON(ret);
		}
C
Chris Mason 已提交
508 509 510
	}
	ret = 0;
error:
511 512
	btrfs_release_path(root, path);
	btrfs_free_path(path);
C
Chris Mason 已提交
513
	inode->i_sb->s_dirt = 1;
C
Chris Mason 已提交
514 515 516
	return ret;
}

C
Chris Mason 已提交
517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578
static int btrfs_truncate_page(struct address_space *mapping, loff_t from)
{
	struct inode *inode = mapping->host;
	unsigned blocksize = 1 << inode->i_blkbits;
	pgoff_t index = from >> PAGE_CACHE_SHIFT;
	unsigned offset = from & (PAGE_CACHE_SIZE-1);
	struct page *page;
	char *kaddr;
	int ret = 0;
	struct btrfs_root *root = BTRFS_I(inode)->root;
	u64 alloc_hint;
	struct btrfs_key ins;
	struct btrfs_trans_handle *trans;

	if ((offset & (blocksize - 1)) == 0)
		goto out;

	ret = -ENOMEM;
	page = grab_cache_page(mapping, index);
	if (!page)
		goto out;

	if (!PageUptodate(page)) {
		ret = mpage_readpage(page, btrfs_get_block);
		lock_page(page);
		if (!PageUptodate(page)) {
			ret = -EIO;
			goto out;
		}
	}
	mutex_lock(&root->fs_info->fs_mutex);
	trans = btrfs_start_transaction(root, 1);
	btrfs_set_trans_block_group(trans, inode);

	ret = drop_extents(trans, root, inode, page->index << PAGE_CACHE_SHIFT,
			   (page->index + 1) << PAGE_CACHE_SHIFT, &alloc_hint);
	BUG_ON(ret);
	ret = btrfs_alloc_extent(trans, root, inode->i_ino, 1,
				 alloc_hint, (u64)-1, &ins, 1);
	BUG_ON(ret);
	ret = btrfs_insert_file_extent(trans, root, inode->i_ino,
				       page->index << PAGE_CACHE_SHIFT,
				       ins.objectid, 1, 1);
	BUG_ON(ret);
	SetPageChecked(page);
	kaddr = kmap(page);
	memset(kaddr + offset, 0, PAGE_CACHE_SIZE - offset);
	flush_dcache_page(page);
	btrfs_csum_file_block(trans, root, inode->i_ino,
			      page->index << PAGE_CACHE_SHIFT,
			      kaddr, PAGE_CACHE_SIZE);
	kunmap(page);
	btrfs_end_transaction(trans, root);
	mutex_unlock(&root->fs_info->fs_mutex);

	set_page_dirty(page);
	unlock_page(page);
	page_cache_release(page);
out:
	return ret;
}

C
Chris Mason 已提交
579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595
static int btrfs_setattr(struct dentry *dentry, struct iattr *attr)
{
	struct inode *inode = dentry->d_inode;
	int err;

	err = inode_change_ok(inode, attr);
	if (err)
		return err;

	if (S_ISREG(inode->i_mode) &&
	    attr->ia_valid & ATTR_SIZE && attr->ia_size > inode->i_size) {
		struct btrfs_trans_handle *trans;
		struct btrfs_root *root = BTRFS_I(inode)->root;
		u64 mask = root->blocksize - 1;
		u64 pos = (inode->i_size + mask) & ~mask;
		u64 hole_size;

C
Chris Mason 已提交
596
		if (attr->ia_size <= pos)
C
Chris Mason 已提交
597
			goto out;
C
Chris Mason 已提交
598 599 600

		btrfs_truncate_page(inode->i_mapping, inode->i_size);

C
Chris Mason 已提交
601 602 603 604 605 606 607 608 609 610 611 612
		hole_size = (attr->ia_size - pos + mask) & ~mask;
		hole_size >>= inode->i_blkbits;

		mutex_lock(&root->fs_info->fs_mutex);
		trans = btrfs_start_transaction(root, 1);
		btrfs_set_trans_block_group(trans, inode);
		err = btrfs_insert_file_extent(trans, root, inode->i_ino,
					       pos, 0, 0, hole_size);
		BUG_ON(err);
		btrfs_end_transaction(trans, root);
		mutex_unlock(&root->fs_info->fs_mutex);
	}
C
Chris Mason 已提交
613
out:
C
Chris Mason 已提交
614 615 616 617
	err = inode_setattr(inode, attr);

	return err;
}
C
Chris Mason 已提交
618 619 620
static void btrfs_delete_inode(struct inode *inode)
{
	struct btrfs_trans_handle *trans;
621
	struct btrfs_root *root = BTRFS_I(inode)->root;
C
Chris Mason 已提交
622 623
	int ret;

C
Chris Mason 已提交
624 625 626 627 628 629 630
	truncate_inode_pages(&inode->i_data, 0);
	if (is_bad_inode(inode)) {
		goto no_delete;
	}
	inode->i_size = 0;
	mutex_lock(&root->fs_info->fs_mutex);
	trans = btrfs_start_transaction(root, 1);
631
	btrfs_set_trans_block_group(trans, inode);
C
Chris Mason 已提交
632 633
	ret = btrfs_truncate_in_trans(trans, root, inode);
	BUG_ON(ret);
C
Chris Mason 已提交
634 635 636
	btrfs_free_inode(trans, root, inode);
	btrfs_end_transaction(trans, root);
	mutex_unlock(&root->fs_info->fs_mutex);
C
Chris Mason 已提交
637
	btrfs_btree_balance_dirty(root);
C
Chris Mason 已提交
638 639 640 641 642
	return;
no_delete:
	clear_inode(inode);
}

C
Chris Mason 已提交
643
static int btrfs_inode_by_name(struct inode *dir, struct dentry *dentry,
644
			       struct btrfs_key *location)
C
Chris Mason 已提交
645 646 647 648
{
	const char *name = dentry->d_name.name;
	int namelen = dentry->d_name.len;
	struct btrfs_dir_item *di;
649
	struct btrfs_path *path;
650
	struct btrfs_root *root = BTRFS_I(dir)->root;
C
Chris Mason 已提交
651 652
	int ret;

653 654 655
	path = btrfs_alloc_path();
	BUG_ON(!path);
	btrfs_init_path(path);
656
	di = btrfs_lookup_dir_item(NULL, root, path, dir->i_ino, name,
C
Chris Mason 已提交
657
				    namelen, 0);
658
	if (!di || IS_ERR(di)) {
659
		location->objectid = 0;
C
Chris Mason 已提交
660
		ret = 0;
C
Chris Mason 已提交
661 662
		goto out;
	}
663
	btrfs_disk_key_to_cpu(location, &di->location);
C
Chris Mason 已提交
664
out:
665 666
	btrfs_release_path(root, path);
	btrfs_free_path(path);
C
Chris Mason 已提交
667 668 669
	return ret;
}

C
Chris Mason 已提交
670
static int fixup_tree_root_location(struct btrfs_root *root,
671 672 673 674 675 676 677 678 679 680 681 682 683 684 685
			     struct btrfs_key *location,
			     struct btrfs_root **sub_root)
{
	struct btrfs_path *path;
	struct btrfs_root_item *ri;

	if (btrfs_key_type(location) != BTRFS_ROOT_ITEM_KEY)
		return 0;
	if (location->objectid == BTRFS_ROOT_TREE_OBJECTID)
		return 0;

	path = btrfs_alloc_path();
	BUG_ON(!path);
	mutex_lock(&root->fs_info->fs_mutex);

686 687 688 689 690
	*sub_root = btrfs_read_fs_root(root->fs_info, location);
	if (IS_ERR(*sub_root))
		return PTR_ERR(*sub_root);

	ri = &(*sub_root)->root_item;
691 692 693 694
	location->objectid = btrfs_root_dirid(ri);
	location->flags = 0;
	btrfs_set_key_type(location, BTRFS_INODE_ITEM_KEY);
	location->offset = 0;
695

696 697
	btrfs_free_path(path);
	mutex_unlock(&root->fs_info->fs_mutex);
698
	return 0;
699 700
}

C
Chris Mason 已提交
701
static int btrfs_init_locked_inode(struct inode *inode, void *p)
C
Chris Mason 已提交
702 703 704 705 706 707 708
{
	struct btrfs_iget_args *args = p;
	inode->i_ino = args->ino;
	BTRFS_I(inode)->root = args->root;
	return 0;
}

C
Chris Mason 已提交
709
static int btrfs_find_actor(struct inode *inode, void *opaque)
C
Chris Mason 已提交
710 711 712 713 714 715
{
	struct btrfs_iget_args *args = opaque;
	return (args->ino == inode->i_ino &&
		args->root == BTRFS_I(inode)->root);
}

C
Chris Mason 已提交
716 717
static struct inode *btrfs_iget_locked(struct super_block *s, u64 objectid,
				       struct btrfs_root *root)
C
Chris Mason 已提交
718 719 720 721 722 723 724 725 726 727 728
{
	struct inode *inode;
	struct btrfs_iget_args args;
	args.ino = objectid;
	args.root = root;

	inode = iget5_locked(s, objectid, btrfs_find_actor,
			     btrfs_init_locked_inode,
			     (void *)&args);
	return inode;
}
729

C
Chris Mason 已提交
730 731 732 733
static struct dentry *btrfs_lookup(struct inode *dir, struct dentry *dentry,
				   struct nameidata *nd)
{
	struct inode * inode;
734 735 736 737
	struct btrfs_inode *bi = BTRFS_I(dir);
	struct btrfs_root *root = bi->root;
	struct btrfs_root *sub_root = root;
	struct btrfs_key location;
C
Chris Mason 已提交
738 739 740 741
	int ret;

	if (dentry->d_name.len > BTRFS_NAME_LEN)
		return ERR_PTR(-ENAMETOOLONG);
C
Chris Mason 已提交
742
	mutex_lock(&root->fs_info->fs_mutex);
743
	ret = btrfs_inode_by_name(dir, dentry, &location);
C
Chris Mason 已提交
744
	mutex_unlock(&root->fs_info->fs_mutex);
C
Chris Mason 已提交
745 746 747
	if (ret < 0)
		return ERR_PTR(ret);
	inode = NULL;
748 749 750 751 752 753
	if (location.objectid) {
		ret = fixup_tree_root_location(root, &location, &sub_root);
		if (ret < 0)
			return ERR_PTR(ret);
		if (ret > 0)
			return ERR_PTR(-ENOENT);
C
Chris Mason 已提交
754 755
		inode = btrfs_iget_locked(dir->i_sb, location.objectid,
					  sub_root);
C
Chris Mason 已提交
756 757
		if (!inode)
			return ERR_PTR(-EACCES);
758
		if (inode->i_state & I_NEW) {
759
			if (sub_root != root) {
C
Chris Mason 已提交
760
printk("adding new root for inode %lu root %p (found %p)\n", inode->i_ino, sub_root, BTRFS_I(inode)->root);
761 762 763
				igrab(inode);
				sub_root->inode = inode;
			}
764 765 766 767 768 769
			BTRFS_I(inode)->root = sub_root;
			memcpy(&BTRFS_I(inode)->location, &location,
			       sizeof(location));
			btrfs_read_locked_inode(inode);
			unlock_new_inode(inode);
		}
C
Chris Mason 已提交
770 771 772 773
	}
	return d_splice_alias(inode, dentry);
}

774 775
static void reada_leaves(struct btrfs_root *root, struct btrfs_path *path,
			 u64 objectid)
C
Chris Mason 已提交
776 777 778
{
	struct btrfs_node *node;
	int i;
779
	u32 nritems;
C
Chris Mason 已提交
780 781 782
	u64 item_objectid;
	u64 blocknr;
	int slot;
783
	int ret;
C
Chris Mason 已提交
784 785 786 787 788 789

	if (!path->nodes[1])
		return;
	node = btrfs_buffer_node(path->nodes[1]);
	slot = path->slots[1];
	nritems = btrfs_header_nritems(&node->header);
790
	for (i = slot + 1; i < nritems; i++) {
C
Chris Mason 已提交
791 792 793 794
		item_objectid = btrfs_disk_key_objectid(&node->ptrs[i].key);
		if (item_objectid != objectid)
			break;
		blocknr = btrfs_node_blockptr(node, i);
795 796 797
		ret = readahead_tree_block(root, blocknr);
		if (ret)
			break;
C
Chris Mason 已提交
798 799
	}
}
C
Chris Mason 已提交
800 801 802
static unsigned char btrfs_filetype_table[] = {
	DT_UNKNOWN, DT_REG, DT_DIR, DT_CHR, DT_BLK, DT_FIFO, DT_SOCK, DT_LNK
};
C
Chris Mason 已提交
803

C
Chris Mason 已提交
804 805 806
static int btrfs_readdir(struct file *filp, void *dirent, filldir_t filldir)
{
	struct inode *inode = filp->f_path.dentry->d_inode;
807
	struct btrfs_root *root = BTRFS_I(inode)->root;
C
Chris Mason 已提交
808 809 810
	struct btrfs_item *item;
	struct btrfs_dir_item *di;
	struct btrfs_key key;
811
	struct btrfs_path *path;
C
Chris Mason 已提交
812 813 814 815 816
	int ret;
	u32 nritems;
	struct btrfs_leaf *leaf;
	int slot;
	int advance;
C
Chris Mason 已提交
817
	unsigned char d_type;
818
	int over = 0;
819 820 821 822
	u32 di_cur;
	u32 di_total;
	u32 di_len;
	int key_type = BTRFS_DIR_INDEX_KEY;
823 824 825 826

	/* FIXME, use a real flag for deciding about the key type */
	if (root->fs_info->tree_root == root)
		key_type = BTRFS_DIR_ITEM_KEY;
C
Chris Mason 已提交
827
	mutex_lock(&root->fs_info->fs_mutex);
C
Chris Mason 已提交
828 829
	key.objectid = inode->i_ino;
	key.flags = 0;
830
	btrfs_set_key_type(&key, key_type);
C
Chris Mason 已提交
831
	key.offset = filp->f_pos;
832 833 834
	path = btrfs_alloc_path();
	btrfs_init_path(path);
	ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
C
Chris Mason 已提交
835
	if (ret < 0)
C
Chris Mason 已提交
836
		goto err;
837
	advance = 0;
838
	reada_leaves(root, path, inode->i_ino);
C
Chris Mason 已提交
839
	while(1) {
840
		leaf = btrfs_buffer_leaf(path->nodes[0]);
C
Chris Mason 已提交
841
		nritems = btrfs_header_nritems(&leaf->header);
842
		slot = path->slots[0];
C
Chris Mason 已提交
843 844
		if (advance || slot >= nritems) {
			if (slot >= nritems -1) {
845
				reada_leaves(root, path, inode->i_ino);
846
				ret = btrfs_next_leaf(root, path);
C
Chris Mason 已提交
847 848
				if (ret)
					break;
849
				leaf = btrfs_buffer_leaf(path->nodes[0]);
C
Chris Mason 已提交
850
				nritems = btrfs_header_nritems(&leaf->header);
851
				slot = path->slots[0];
C
Chris Mason 已提交
852 853
			} else {
				slot++;
854
				path->slots[0]++;
C
Chris Mason 已提交
855 856 857 858 859 860
			}
		}
		advance = 1;
		item = leaf->items + slot;
		if (btrfs_disk_key_objectid(&item->key) != key.objectid)
			break;
861
		if (btrfs_disk_key_type(&item->key) != key_type)
862
			break;
863 864
		if (btrfs_disk_key_offset(&item->key) < filp->f_pos)
			continue;
865
		filp->f_pos = btrfs_disk_key_offset(&item->key);
C
Chris Mason 已提交
866
		advance = 1;
C
Chris Mason 已提交
867
		di = btrfs_item_ptr(leaf, slot, struct btrfs_dir_item);
868 869 870
		di_cur = 0;
		di_total = btrfs_item_size(leaf->items + slot);
		while(di_cur < di_total) {
C
Chris Mason 已提交
871
			d_type = btrfs_filetype_table[btrfs_dir_type(di)];
872 873 874 875 876 877 878 879 880 881 882
			over = filldir(dirent, (const char *)(di + 1),
				       btrfs_dir_name_len(di),
				       btrfs_disk_key_offset(&item->key),
				       btrfs_disk_key_objectid(&di->location),
				       d_type);
			if (over)
				goto nopos;
			di_len = btrfs_dir_name_len(di) + sizeof(*di);
			di_cur += di_len;
			di = (struct btrfs_dir_item *)((char *)di + di_len);
		}
C
Chris Mason 已提交
883
	}
884 885
	filp->f_pos++;
nopos:
C
Chris Mason 已提交
886 887
	ret = 0;
err:
888 889
	btrfs_release_path(root, path);
	btrfs_free_path(path);
C
Chris Mason 已提交
890
	mutex_unlock(&root->fs_info->fs_mutex);
C
Chris Mason 已提交
891 892 893 894 895 896 897 898 899 900 901 902 903 904
	return ret;
}

static void btrfs_put_super (struct super_block * sb)
{
	struct btrfs_root *root = btrfs_sb(sb);
	int ret;

	ret = close_ctree(root);
	if (ret) {
		printk("close ctree returns %d\n", ret);
	}
	sb->s_fs_info = NULL;
}
905 906 907 908

static int btrfs_fill_super(struct super_block * sb, void * data, int silent)
{
	struct inode * inode;
C
Chris Mason 已提交
909 910
	struct dentry * root_dentry;
	struct btrfs_super_block *disk_super;
911
	struct btrfs_root *tree_root;
912
	struct btrfs_inode *bi;
913 914 915

	sb->s_maxbytes = MAX_LFS_FILESIZE;
	sb->s_magic = BTRFS_SUPER_MAGIC;
C
Chris Mason 已提交
916
	sb->s_op = &btrfs_super_ops;
917
	sb->s_time_gran = 1;
C
Chris Mason 已提交
918

919
	tree_root = open_ctree(sb);
920

921
	if (!tree_root) {
C
Chris Mason 已提交
922 923 924
		printk("btrfs: open_ctree failed\n");
		return -EIO;
	}
925 926
	sb->s_fs_info = tree_root;
	disk_super = tree_root->fs_info->disk_super;
C
Chris Mason 已提交
927 928 929 930
	printk("read in super total blocks %Lu root %Lu\n",
	       btrfs_super_total_blocks(disk_super),
	       btrfs_super_root_dir(disk_super));

C
Chris Mason 已提交
931 932
	inode = btrfs_iget_locked(sb, btrfs_super_root_dir(disk_super),
				  tree_root);
933 934 935 936
	bi = BTRFS_I(inode);
	bi->location.objectid = inode->i_ino;
	bi->location.offset = 0;
	bi->location.flags = 0;
937
	bi->root = tree_root;
938 939
	btrfs_set_key_type(&bi->location, BTRFS_INODE_ITEM_KEY);

940 941
	if (!inode)
		return -ENOMEM;
C
Chris Mason 已提交
942 943 944 945
	if (inode->i_state & I_NEW) {
		btrfs_read_locked_inode(inode);
		unlock_new_inode(inode);
	}
946

C
Chris Mason 已提交
947 948
	root_dentry = d_alloc_root(inode);
	if (!root_dentry) {
949 950 951
		iput(inode);
		return -ENOMEM;
	}
C
Chris Mason 已提交
952 953
	sb->s_root = root_dentry;

954 955 956
	return 0;
}

C
Chris Mason 已提交
957 958
static int btrfs_write_inode(struct inode *inode, int wait)
{
959
	struct btrfs_root *root = BTRFS_I(inode)->root;
C
Chris Mason 已提交
960
	struct btrfs_trans_handle *trans;
C
Chris Mason 已提交
961 962 963 964 965
	int ret = 0;

	if (wait) {
		mutex_lock(&root->fs_info->fs_mutex);
		trans = btrfs_start_transaction(root, 1);
966
		btrfs_set_trans_block_group(trans, inode);
C
Chris Mason 已提交
967 968 969 970 971 972 973 974 975 976
		ret = btrfs_commit_transaction(trans, root);
		mutex_unlock(&root->fs_info->fs_mutex);
	}
	return ret;
}

static void btrfs_dirty_inode(struct inode *inode)
{
	struct btrfs_root *root = BTRFS_I(inode)->root;
	struct btrfs_trans_handle *trans;
C
Chris Mason 已提交
977 978 979

	mutex_lock(&root->fs_info->fs_mutex);
	trans = btrfs_start_transaction(root, 1);
980
	btrfs_set_trans_block_group(trans, inode);
C
Chris Mason 已提交
981 982
	btrfs_update_inode(trans, root, inode);
	btrfs_end_transaction(trans, root);
C
Chris Mason 已提交
983
	mutex_unlock(&root->fs_info->fs_mutex);
C
Chris Mason 已提交
984
	btrfs_btree_balance_dirty(root);
C
Chris Mason 已提交
985 986
}

987
static struct inode *btrfs_new_inode(struct btrfs_trans_handle *trans,
C
Chris Mason 已提交
988
				     struct btrfs_root *root,
989 990 991
				     u64 objectid,
				     struct btrfs_block_group_cache *group,
				     int mode)
992 993 994
{
	struct inode *inode;
	struct btrfs_inode_item inode_item;
C
Chris Mason 已提交
995
	struct btrfs_key *location;
996
	int ret;
997
	int owner;
998

C
Chris Mason 已提交
999
	inode = new_inode(root->fs_info->sb);
1000 1001 1002
	if (!inode)
		return ERR_PTR(-ENOMEM);

C
Chris Mason 已提交
1003
	BTRFS_I(inode)->root = root;
1004 1005 1006 1007 1008
	if (mode & S_IFDIR)
		owner = 0;
	else
		owner = 1;
	group = btrfs_find_block_group(root, group, 0, 0, owner);
1009
	BTRFS_I(inode)->block_group = group;
1010 1011 1012 1013 1014 1015

	inode->i_uid = current->fsuid;
	inode->i_gid = current->fsgid;
	inode->i_mode = mode;
	inode->i_ino = objectid;
	inode->i_blocks = 0;
C
Chris Mason 已提交
1016
	inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
1017
	fill_inode_item(&inode_item, inode);
C
Chris Mason 已提交
1018 1019 1020 1021 1022
	location = &BTRFS_I(inode)->location;
	location->objectid = objectid;
	location->flags = 0;
	location->offset = 0;
	btrfs_set_key_type(location, BTRFS_INODE_ITEM_KEY);
1023 1024 1025 1026 1027 1028 1029 1030

	ret = btrfs_insert_inode(trans, root, objectid, &inode_item);
	BUG_ON(ret);

	insert_inode_hash(inode);
	return inode;
}

C
Chris Mason 已提交
1031 1032 1033 1034 1035
static inline u8 btrfs_inode_type(struct inode *inode)
{
	return btrfs_type_by_mode[(inode->i_mode & S_IFMT) >> S_SHIFT];
}

1036 1037 1038 1039
static int btrfs_add_link(struct btrfs_trans_handle *trans,
			    struct dentry *dentry, struct inode *inode)
{
	int ret;
1040 1041 1042 1043 1044 1045 1046 1047
	struct btrfs_key key;
	struct btrfs_root *root = BTRFS_I(dentry->d_parent->d_inode)->root;
	key.objectid = inode->i_ino;
	key.flags = 0;
	btrfs_set_key_type(&key, BTRFS_INODE_ITEM_KEY);
	key.offset = 0;

	ret = btrfs_insert_dir_item(trans, root,
1048 1049
				    dentry->d_name.name, dentry->d_name.len,
				    dentry->d_parent->d_inode->i_ino,
C
Chris Mason 已提交
1050
				    &key, btrfs_inode_type(inode));
C
Chris Mason 已提交
1051
	if (ret == 0) {
1052
		dentry->d_parent->d_inode->i_size += dentry->d_name.len * 2;
1053
		ret = btrfs_update_inode(trans, root,
C
Chris Mason 已提交
1054 1055
					 dentry->d_parent->d_inode);
	}
1056 1057 1058 1059 1060 1061 1062 1063 1064 1065 1066
	return ret;
}

static int btrfs_add_nondir(struct btrfs_trans_handle *trans,
			    struct dentry *dentry, struct inode *inode)
{
	int err = btrfs_add_link(trans, dentry, inode);
	if (!err) {
		d_instantiate(dentry, inode);
		return 0;
	}
C
Chris Mason 已提交
1067 1068
	if (err > 0)
		err = -EEXIST;
1069 1070 1071 1072 1073 1074 1075
	return err;
}

static int btrfs_create(struct inode *dir, struct dentry *dentry,
			int mode, struct nameidata *nd)
{
	struct btrfs_trans_handle *trans;
1076
	struct btrfs_root *root = BTRFS_I(dir)->root;
1077 1078
	struct inode *inode;
	int err;
C
Chris Mason 已提交
1079
	int drop_inode = 0;
C
Chris Mason 已提交
1080
	u64 objectid;
1081

C
Chris Mason 已提交
1082
	mutex_lock(&root->fs_info->fs_mutex);
1083
	trans = btrfs_start_transaction(root, 1);
1084
	btrfs_set_trans_block_group(trans, dir);
C
Chris Mason 已提交
1085 1086 1087 1088 1089 1090 1091

	err = btrfs_find_free_objectid(trans, root, dir->i_ino, &objectid);
	if (err) {
		err = -ENOSPC;
		goto out_unlock;
	}

1092 1093
	inode = btrfs_new_inode(trans, root, objectid,
				BTRFS_I(dir)->block_group, mode);
1094 1095
	err = PTR_ERR(inode);
	if (IS_ERR(inode))
C
Chris Mason 已提交
1096
		goto out_unlock;
1097 1098

	btrfs_set_trans_block_group(trans, inode);
1099
	err = btrfs_add_nondir(trans, dentry, inode);
C
Chris Mason 已提交
1100 1101
	if (err)
		drop_inode = 1;
C
Chris Mason 已提交
1102 1103 1104 1105 1106
	else {
		inode->i_mapping->a_ops = &btrfs_aops;
		inode->i_fop = &btrfs_file_operations;
		inode->i_op = &btrfs_file_inode_operations;
	}
1107
	dir->i_sb->s_dirt = 1;
1108 1109
	btrfs_update_inode_block_group(trans, inode);
	btrfs_update_inode_block_group(trans, dir);
C
Chris Mason 已提交
1110
out_unlock:
C
Chris Mason 已提交
1111
	btrfs_end_transaction(trans, root);
C
Chris Mason 已提交
1112
	mutex_unlock(&root->fs_info->fs_mutex);
C
Chris Mason 已提交
1113

C
Chris Mason 已提交
1114 1115 1116 1117
	if (drop_inode) {
		inode_dec_link_count(inode);
		iput(inode);
	}
C
Chris Mason 已提交
1118
	btrfs_btree_balance_dirty(root);
1119 1120 1121
	return err;
}

C
Chris Mason 已提交
1122 1123 1124 1125 1126 1127 1128 1129 1130 1131 1132 1133 1134 1135 1136 1137 1138 1139 1140 1141 1142 1143 1144 1145 1146 1147 1148 1149 1150 1151 1152 1153 1154 1155 1156
static int btrfs_link(struct dentry *old_dentry, struct inode *dir,
		      struct dentry *dentry)
{
	struct btrfs_trans_handle *trans;
	struct btrfs_root *root = BTRFS_I(dir)->root;
	struct inode *inode = old_dentry->d_inode;
	int err;
	int drop_inode = 0;

	if (inode->i_nlink == 0)
		return -ENOENT;

	inc_nlink(inode);
	mutex_lock(&root->fs_info->fs_mutex);
	trans = btrfs_start_transaction(root, 1);
	btrfs_set_trans_block_group(trans, dir);
	atomic_inc(&inode->i_count);
	err = btrfs_add_nondir(trans, dentry, inode);
	if (err)
		drop_inode = 1;
	dir->i_sb->s_dirt = 1;
	btrfs_update_inode_block_group(trans, dir);
	btrfs_update_inode(trans, root, inode);

	btrfs_end_transaction(trans, root);
	mutex_unlock(&root->fs_info->fs_mutex);

	if (drop_inode) {
		inode_dec_link_count(inode);
		iput(inode);
	}
	btrfs_btree_balance_dirty(root);
	return err;
}

C
Chris Mason 已提交
1157
static int btrfs_make_empty_dir(struct btrfs_trans_handle *trans,
C
Chris Mason 已提交
1158 1159
				struct btrfs_root *root,
				u64 objectid, u64 dirid)
C
Chris Mason 已提交
1160 1161 1162
{
	int ret;
	char buf[2];
1163 1164
	struct btrfs_key key;

C
Chris Mason 已提交
1165 1166 1167
	buf[0] = '.';
	buf[1] = '.';

C
Chris Mason 已提交
1168
	key.objectid = objectid;
1169 1170 1171 1172
	key.offset = 0;
	key.flags = 0;
	btrfs_set_key_type(&key, BTRFS_INODE_ITEM_KEY);

C
Chris Mason 已提交
1173
	ret = btrfs_insert_dir_item(trans, root, buf, 1, objectid,
C
Chris Mason 已提交
1174
				    &key, BTRFS_FT_DIR);
C
Chris Mason 已提交
1175 1176
	if (ret)
		goto error;
C
Chris Mason 已提交
1177 1178
	key.objectid = dirid;
	ret = btrfs_insert_dir_item(trans, root, buf, 2, objectid,
C
Chris Mason 已提交
1179
				    &key, BTRFS_FT_DIR);
C
Chris Mason 已提交
1180 1181
	if (ret)
		goto error;
C
Chris Mason 已提交
1182 1183 1184 1185 1186 1187 1188 1189
error:
	return ret;
}

static int btrfs_mkdir(struct inode *dir, struct dentry *dentry, int mode)
{
	struct inode *inode;
	struct btrfs_trans_handle *trans;
1190
	struct btrfs_root *root = BTRFS_I(dir)->root;
C
Chris Mason 已提交
1191 1192
	int err = 0;
	int drop_on_err = 0;
C
Chris Mason 已提交
1193
	u64 objectid;
C
Chris Mason 已提交
1194 1195 1196

	mutex_lock(&root->fs_info->fs_mutex);
	trans = btrfs_start_transaction(root, 1);
1197
	btrfs_set_trans_block_group(trans, dir);
C
Chris Mason 已提交
1198 1199 1200 1201
	if (IS_ERR(trans)) {
		err = PTR_ERR(trans);
		goto out_unlock;
	}
C
Chris Mason 已提交
1202 1203 1204 1205 1206 1207 1208

	err = btrfs_find_free_objectid(trans, root, dir->i_ino, &objectid);
	if (err) {
		err = -ENOSPC;
		goto out_unlock;
	}

1209 1210
	inode = btrfs_new_inode(trans, root, objectid,
				BTRFS_I(dir)->block_group, S_IFDIR | mode);
C
Chris Mason 已提交
1211 1212 1213 1214 1215 1216 1217
	if (IS_ERR(inode)) {
		err = PTR_ERR(inode);
		goto out_fail;
	}
	drop_on_err = 1;
	inode->i_op = &btrfs_dir_inode_operations;
	inode->i_fop = &btrfs_dir_file_operations;
1218
	btrfs_set_trans_block_group(trans, inode);
C
Chris Mason 已提交
1219

C
Chris Mason 已提交
1220 1221 1222 1223 1224 1225
	err = btrfs_make_empty_dir(trans, root, inode->i_ino, dir->i_ino);
	if (err)
		goto out_fail;

	inode->i_size = 6;
	err = btrfs_update_inode(trans, root, inode);
C
Chris Mason 已提交
1226 1227 1228 1229 1230 1231 1232
	if (err)
		goto out_fail;
	err = btrfs_add_link(trans, dentry, inode);
	if (err)
		goto out_fail;
	d_instantiate(dentry, inode);
	drop_on_err = 0;
C
Chris Mason 已提交
1233
	dir->i_sb->s_dirt = 1;
1234 1235
	btrfs_update_inode_block_group(trans, inode);
	btrfs_update_inode_block_group(trans, dir);
C
Chris Mason 已提交
1236 1237 1238 1239 1240 1241 1242

out_fail:
	btrfs_end_transaction(trans, root);
out_unlock:
	mutex_unlock(&root->fs_info->fs_mutex);
	if (drop_on_err)
		iput(inode);
C
Chris Mason 已提交
1243
	btrfs_btree_balance_dirty(root);
C
Chris Mason 已提交
1244 1245 1246
	return err;
}

C
Chris Mason 已提交
1247 1248 1249 1250 1251 1252 1253 1254 1255 1256 1257 1258 1259 1260 1261 1262 1263 1264 1265 1266
static int btrfs_sync_file(struct file *file,
			   struct dentry *dentry, int datasync)
{
	struct inode *inode = dentry->d_inode;
	struct btrfs_root *root = BTRFS_I(inode)->root;
	int ret;
	struct btrfs_trans_handle *trans;

	mutex_lock(&root->fs_info->fs_mutex);
	trans = btrfs_start_transaction(root, 1);
	if (!trans) {
		ret = -ENOMEM;
		goto out;
	}
	ret = btrfs_commit_transaction(trans, root);
	mutex_unlock(&root->fs_info->fs_mutex);
out:
	return ret > 0 ? EIO : ret;
}

1267 1268 1269 1270 1271
static int btrfs_sync_fs(struct super_block *sb, int wait)
{
	struct btrfs_trans_handle *trans;
	struct btrfs_root *root;
	int ret;
1272
	root = btrfs_sb(sb);
1273

1274
	sb->s_dirt = 0;
C
Chris Mason 已提交
1275
	if (!wait) {
C
Chris Mason 已提交
1276
		filemap_flush(root->fs_info->btree_inode->i_mapping);
C
Chris Mason 已提交
1277 1278 1279
		return 0;
	}
	mutex_lock(&root->fs_info->fs_mutex);
1280 1281 1282 1283 1284
	trans = btrfs_start_transaction(root, 1);
	ret = btrfs_commit_transaction(trans, root);
	sb->s_dirt = 0;
	BUG_ON(ret);
printk("btrfs sync_fs\n");
C
Chris Mason 已提交
1285
	mutex_unlock(&root->fs_info->fs_mutex);
1286 1287 1288
	return 0;
}

C
Chris Mason 已提交
1289 1290 1291 1292
#define BTRFS_GET_BLOCK_NO_CREATE 0
#define BTRFS_GET_BLOCK_CREATE 1
#define BTRFS_GET_BLOCK_NO_DIRECT 2

C
Chris Mason 已提交
1293
static int btrfs_get_block_lock(struct inode *inode, sector_t iblock,
C
Chris Mason 已提交
1294 1295 1296 1297 1298 1299 1300 1301
			   struct buffer_head *result, int create)
{
	int ret;
	int err = 0;
	u64 blocknr;
	u64 extent_start = 0;
	u64 extent_end = 0;
	u64 objectid = inode->i_ino;
1302
	u32 found_type;
C
Chris Mason 已提交
1303
	u64 alloc_hint = 0;
1304
	struct btrfs_path *path;
1305
	struct btrfs_root *root = BTRFS_I(inode)->root;
C
Chris Mason 已提交
1306 1307 1308
	struct btrfs_file_extent_item *item;
	struct btrfs_leaf *leaf;
	struct btrfs_disk_key *found_key;
C
Chris Mason 已提交
1309
	struct btrfs_trans_handle *trans = NULL;
C
Chris Mason 已提交
1310

1311 1312 1313
	path = btrfs_alloc_path();
	BUG_ON(!path);
	btrfs_init_path(path);
C
Chris Mason 已提交
1314
	if (create & BTRFS_GET_BLOCK_CREATE) {
1315
		WARN_ON(1);
C
Chris Mason 已提交
1316 1317 1318 1319 1320 1321 1322 1323 1324 1325 1326
		/* this almost but not quite works */
		trans = btrfs_start_transaction(root, 1);
		if (!trans) {
			err = -ENOMEM;
			goto out;
		}
		ret = drop_extents(trans, root, inode,
				   iblock << inode->i_blkbits,
				   (iblock + 1) << inode->i_blkbits,
				   &alloc_hint);
		BUG_ON(ret);
1327
	}
C
Chris Mason 已提交
1328

1329
	ret = btrfs_lookup_file_extent(NULL, root, path,
C
Chris Mason 已提交
1330
				       inode->i_ino,
1331
				       iblock << inode->i_blkbits, 0);
C
Chris Mason 已提交
1332 1333 1334 1335 1336 1337
	if (ret < 0) {
		err = ret;
		goto out;
	}

	if (ret != 0) {
1338 1339
		if (path->slots[0] == 0) {
			btrfs_release_path(root, path);
C
Chris Mason 已提交
1340
			goto not_found;
C
Chris Mason 已提交
1341
		}
1342
		path->slots[0]--;
C
Chris Mason 已提交
1343 1344
	}

1345
	item = btrfs_item_ptr(btrfs_buffer_leaf(path->nodes[0]), path->slots[0],
C
Chris Mason 已提交
1346
			      struct btrfs_file_extent_item);
1347
	leaf = btrfs_buffer_leaf(path->nodes[0]);
C
Chris Mason 已提交
1348 1349 1350 1351
	blocknr = btrfs_file_extent_disk_blocknr(item);
	blocknr += btrfs_file_extent_offset(item);

	/* are we inside the extent that was found? */
1352
	found_key = &leaf->items[path->slots[0]].key;
1353
	found_type = btrfs_disk_key_type(found_key);
C
Chris Mason 已提交
1354
	if (btrfs_disk_key_objectid(found_key) != objectid ||
1355
	    found_type != BTRFS_EXTENT_DATA_KEY) {
C
Chris Mason 已提交
1356 1357
		extent_end = 0;
		extent_start = 0;
C
Chris Mason 已提交
1358
		goto not_found;
C
Chris Mason 已提交
1359
	}
1360 1361 1362 1363 1364
	found_type = btrfs_file_extent_type(item);
	extent_start = btrfs_disk_key_offset(&leaf->items[path->slots[0]].key);
	if (found_type == BTRFS_FILE_EXTENT_REG) {
		extent_start = extent_start >> inode->i_blkbits;
		extent_end = extent_start + btrfs_file_extent_num_blocks(item);
C
Chris Mason 已提交
1365
		err = 0;
C
Chris Mason 已提交
1366
		if (btrfs_file_extent_disk_blocknr(item) == 0)
C
Chris Mason 已提交
1367
			goto out;
1368 1369 1370 1371 1372 1373 1374 1375 1376
		if (iblock >= extent_start && iblock < extent_end) {
			btrfs_map_bh_to_logical(root, result, blocknr +
						iblock - extent_start);
			goto out;
		}
	} else if (found_type == BTRFS_FILE_EXTENT_INLINE) {
		char *ptr;
		char *map;
		u32 size;
C
Chris Mason 已提交
1377 1378 1379 1380 1381

		if (create & BTRFS_GET_BLOCK_NO_DIRECT) {
			err = -EINVAL;
			goto out;
		}
1382 1383 1384 1385 1386
		size = btrfs_file_extent_inline_len(leaf->items +
						    path->slots[0]);
		extent_end = (extent_start + size) >> inode->i_blkbits;
		extent_start >>= inode->i_blkbits;
		if (iblock < extent_start || iblock > extent_end) {
C
Chris Mason 已提交
1387
			goto not_found;
1388 1389 1390 1391 1392 1393 1394 1395 1396 1397
		}
		ptr = btrfs_file_extent_inline_start(item);
		map = kmap(result->b_page);
		memcpy(map, ptr, size);
		memset(map + size, 0, PAGE_CACHE_SIZE - size);
		flush_dcache_page(result->b_page);
		kunmap(result->b_page);
		set_buffer_uptodate(result);
		SetPageChecked(result->b_page);
		btrfs_map_bh_to_logical(root, result, 0);
C
Chris Mason 已提交
1398
	}
C
Chris Mason 已提交
1399
not_found:
C
Chris Mason 已提交
1400
	if (create & BTRFS_GET_BLOCK_CREATE) {
C
Chris Mason 已提交
1401 1402 1403 1404 1405 1406 1407 1408 1409 1410 1411 1412 1413
		struct btrfs_key ins;
		ret = btrfs_alloc_extent(trans, root, inode->i_ino,
					 1, alloc_hint, (u64)-1,
					 &ins, 1);
		BUG_ON(ret);
		ret = btrfs_insert_file_extent(trans, root, inode->i_ino,
					       iblock << inode->i_blkbits,
					       ins.objectid, ins.offset,
					       ins.offset);
		BUG_ON(ret);
		SetPageChecked(result->b_page);
		btrfs_map_bh_to_logical(root, result, ins.objectid);
	}
C
Chris Mason 已提交
1414
out:
C
Chris Mason 已提交
1415 1416
	if (trans)
		err = btrfs_end_transaction(trans, root);
1417
	btrfs_free_path(path);
C
Chris Mason 已提交
1418 1419 1420 1421 1422 1423 1424
	return err;
}

static int btrfs_get_block(struct inode *inode, sector_t iblock,
			   struct buffer_head *result, int create)
{
	int err;
1425
	struct btrfs_root *root = BTRFS_I(inode)->root;
C
Chris Mason 已提交
1426
	mutex_lock(&root->fs_info->fs_mutex);
1427
	err = btrfs_get_block_lock(inode, iblock, result, create);
C
Chris Mason 已提交
1428 1429 1430 1431
	mutex_unlock(&root->fs_info->fs_mutex);
	return err;
}

C
Chris Mason 已提交
1432 1433 1434 1435 1436 1437 1438 1439 1440 1441 1442 1443 1444 1445 1446
static int btrfs_get_block_bmap(struct inode *inode, sector_t iblock,
			   struct buffer_head *result, int create)
{
	struct btrfs_root *root = BTRFS_I(inode)->root;
	mutex_lock(&root->fs_info->fs_mutex);
	btrfs_get_block_lock(inode, iblock, result, BTRFS_GET_BLOCK_NO_DIRECT);
	mutex_unlock(&root->fs_info->fs_mutex);
	return 0;
}

static sector_t btrfs_bmap(struct address_space *as, sector_t block)
{
	return generic_block_bmap(as, block, btrfs_get_block_bmap);
}

C
Chris Mason 已提交
1447 1448 1449
static int btrfs_prepare_write(struct file *file, struct page *page,
			       unsigned from, unsigned to)
{
C
Chris Mason 已提交
1450
	return block_prepare_write(page, from, to, btrfs_get_block);
C
Chris Mason 已提交
1451 1452
}

C
Chris Mason 已提交
1453 1454 1455 1456 1457
static void btrfs_write_super(struct super_block *sb)
{
	btrfs_sync_fs(sb, 1);
}

C
Chris Mason 已提交
1458 1459 1460 1461 1462
static int btrfs_readpage(struct file *file, struct page *page)
{
	return mpage_readpage(page, btrfs_get_block);
}

1463 1464 1465 1466 1467 1468 1469 1470 1471 1472 1473 1474 1475 1476 1477 1478 1479 1480 1481 1482 1483 1484 1485 1486 1487 1488 1489 1490 1491 1492 1493 1494 1495 1496 1497 1498 1499 1500 1501 1502 1503 1504 1505 1506 1507 1508 1509 1510 1511 1512 1513 1514 1515 1516 1517 1518 1519 1520 1521 1522 1523 1524 1525 1526
/*
 * While block_write_full_page is writing back the dirty buffers under
 * the page lock, whoever dirtied the buffers may decide to clean them
 * again at any time.  We handle that by only looking at the buffer
 * state inside lock_buffer().
 *
 * If block_write_full_page() is called for regular writeback
 * (wbc->sync_mode == WB_SYNC_NONE) then it will redirty a page which has a
 * locked buffer.   This only can happen if someone has written the buffer
 * directly, with submit_bh().  At the address_space level PageWriteback
 * prevents this contention from occurring.
 */
static int __btrfs_write_full_page(struct inode *inode, struct page *page,
				   struct writeback_control *wbc)
{
	int err;
	sector_t block;
	sector_t last_block;
	struct buffer_head *bh, *head;
	const unsigned blocksize = 1 << inode->i_blkbits;
	int nr_underway = 0;

	BUG_ON(!PageLocked(page));

	last_block = (i_size_read(inode) - 1) >> inode->i_blkbits;

	if (!page_has_buffers(page)) {
		create_empty_buffers(page, blocksize,
					(1 << BH_Dirty)|(1 << BH_Uptodate));
	}

	/*
	 * Be very careful.  We have no exclusion from __set_page_dirty_buffers
	 * here, and the (potentially unmapped) buffers may become dirty at
	 * any time.  If a buffer becomes dirty here after we've inspected it
	 * then we just miss that fact, and the page stays dirty.
	 *
	 * Buffers outside i_size may be dirtied by __set_page_dirty_buffers;
	 * handle that here by just cleaning them.
	 */

	block = (sector_t)page->index << (PAGE_CACHE_SHIFT - inode->i_blkbits);
	head = page_buffers(page);
	bh = head;

	/*
	 * Get all the dirty buffers mapped to disk addresses and
	 * handle any aliases from the underlying blockdev's mapping.
	 */
	do {
		if (block > last_block) {
			/*
			 * mapped buffers outside i_size will occur, because
			 * this page can be outside i_size when there is a
			 * truncate in progress.
			 */
			/*
			 * The buffer was zeroed by block_write_full_page()
			 */
			clear_buffer_dirty(bh);
			set_buffer_uptodate(bh);
		} else if (!buffer_mapped(bh) && buffer_dirty(bh)) {
			WARN_ON(bh->b_size != blocksize);
			err = btrfs_get_block(inode, block, bh, 0);
C
Chris Mason 已提交
1527 1528
			if (err) {
printk("writepage going to recovery err %d\n", err);
1529
				goto recover;
C
Chris Mason 已提交
1530
			}
1531 1532 1533 1534 1535 1536 1537 1538 1539 1540 1541 1542 1543 1544 1545 1546 1547 1548 1549 1550 1551 1552 1553 1554 1555 1556 1557 1558 1559 1560 1561 1562 1563 1564 1565 1566 1567 1568 1569 1570 1571 1572 1573 1574 1575 1576 1577 1578 1579 1580 1581 1582 1583 1584 1585 1586 1587 1588 1589 1590 1591 1592 1593 1594 1595 1596 1597 1598 1599 1600 1601 1602 1603 1604 1605 1606 1607 1608 1609 1610 1611 1612 1613 1614 1615 1616 1617 1618 1619 1620 1621 1622 1623 1624 1625 1626 1627 1628 1629 1630 1631 1632 1633 1634 1635 1636 1637 1638 1639 1640 1641
			if (buffer_new(bh)) {
				/* blockdev mappings never come here */
				clear_buffer_new(bh);
			}
		}
		bh = bh->b_this_page;
		block++;
	} while (bh != head);

	do {
		if (!buffer_mapped(bh))
			continue;
		/*
		 * If it's a fully non-blocking write attempt and we cannot
		 * lock the buffer then redirty the page.  Note that this can
		 * potentially cause a busy-wait loop from pdflush and kswapd
		 * activity, but those code paths have their own higher-level
		 * throttling.
		 */
		if (wbc->sync_mode != WB_SYNC_NONE || !wbc->nonblocking) {
			lock_buffer(bh);
		} else if (test_set_buffer_locked(bh)) {
			redirty_page_for_writepage(wbc, page);
			continue;
		}
		if (test_clear_buffer_dirty(bh) && bh->b_blocknr != 0) {
			mark_buffer_async_write(bh);
		} else {
			unlock_buffer(bh);
		}
	} while ((bh = bh->b_this_page) != head);

	/*
	 * The page and its buffers are protected by PageWriteback(), so we can
	 * drop the bh refcounts early.
	 */
	BUG_ON(PageWriteback(page));
	set_page_writeback(page);

	do {
		struct buffer_head *next = bh->b_this_page;
		if (buffer_async_write(bh)) {
			submit_bh(WRITE, bh);
			nr_underway++;
		}
		bh = next;
	} while (bh != head);
	unlock_page(page);

	err = 0;
done:
	if (nr_underway == 0) {
		/*
		 * The page was marked dirty, but the buffers were
		 * clean.  Someone wrote them back by hand with
		 * ll_rw_block/submit_bh.  A rare case.
		 */
		int uptodate = 1;
		do {
			if (!buffer_uptodate(bh)) {
				uptodate = 0;
				break;
			}
			bh = bh->b_this_page;
		} while (bh != head);
		if (uptodate)
			SetPageUptodate(page);
		end_page_writeback(page);
	}
	return err;

recover:
	/*
	 * ENOSPC, or some other error.  We may already have added some
	 * blocks to the file, so we need to write these out to avoid
	 * exposing stale data.
	 * The page is currently locked and not marked for writeback
	 */
	bh = head;
	/* Recovery: lock and submit the mapped buffers */
	do {
		if (buffer_mapped(bh) && buffer_dirty(bh)) {
			lock_buffer(bh);
			mark_buffer_async_write(bh);
		} else {
			/*
			 * The buffer may have been set dirty during
			 * attachment to a dirty page.
			 */
			clear_buffer_dirty(bh);
		}
	} while ((bh = bh->b_this_page) != head);
	SetPageError(page);
	BUG_ON(PageWriteback(page));
	set_page_writeback(page);
	do {
		struct buffer_head *next = bh->b_this_page;
		if (buffer_async_write(bh)) {
			clear_buffer_dirty(bh);
			submit_bh(WRITE, bh);
			nr_underway++;
		}
		bh = next;
	} while (bh != head);
	unlock_page(page);
	goto done;
}

/*
 * The generic ->writepage function for buffer-backed address_spaces
 */
C
Chris Mason 已提交
1642 1643
static int btrfs_writepage(struct page *page, struct writeback_control *wbc)
{
1644 1645 1646 1647 1648 1649 1650 1651 1652 1653 1654 1655 1656 1657 1658 1659 1660 1661 1662 1663 1664 1665 1666 1667 1668 1669 1670 1671 1672 1673 1674 1675 1676 1677 1678
	struct inode * const inode = page->mapping->host;
	loff_t i_size = i_size_read(inode);
	const pgoff_t end_index = i_size >> PAGE_CACHE_SHIFT;
	unsigned offset;
	void *kaddr;

	/* Is the page fully inside i_size? */
	if (page->index < end_index)
		return __btrfs_write_full_page(inode, page, wbc);

	/* Is the page fully outside i_size? (truncate in progress) */
	offset = i_size & (PAGE_CACHE_SIZE-1);
	if (page->index >= end_index+1 || !offset) {
		/*
		 * The page may have dirty, unmapped buffers.  For example,
		 * they may have been added in ext3_writepage().  Make them
		 * freeable here, so the page does not leak.
		 */
		block_invalidatepage(page, 0);
		unlock_page(page);
		return 0; /* don't care */
	}

	/*
	 * The page straddles i_size.  It must be zeroed out on each and every
	 * writepage invokation because it may be mmapped.  "A file is mapped
	 * in multiples of the page size.  For a file that is not a multiple of
	 * the  page size, the remaining memory is zeroed when mapped, and
	 * writes to that region are not written out to the file."
	 */
	kaddr = kmap_atomic(page, KM_USER0);
	memset(kaddr + offset, 0, PAGE_CACHE_SIZE - offset);
	flush_dcache_page(page);
	kunmap_atomic(kaddr, KM_USER0);
	return __btrfs_write_full_page(inode, page, wbc);
C
Chris Mason 已提交
1679
}
C
Chris Mason 已提交
1680

C
Chris Mason 已提交
1681 1682
static void btrfs_truncate(struct inode *inode)
{
1683
	struct btrfs_root *root = BTRFS_I(inode)->root;
C
Chris Mason 已提交
1684 1685 1686 1687 1688 1689 1690 1691
	int ret;
	struct btrfs_trans_handle *trans;

	if (!S_ISREG(inode->i_mode))
		return;
	if (IS_APPEND(inode) || IS_IMMUTABLE(inode))
		return;

C
Chris Mason 已提交
1692
	btrfs_truncate_page(inode->i_mapping, inode->i_size);
C
Chris Mason 已提交
1693 1694 1695

	mutex_lock(&root->fs_info->fs_mutex);
	trans = btrfs_start_transaction(root, 1);
1696
	btrfs_set_trans_block_group(trans, inode);
C
Chris Mason 已提交
1697 1698

	/* FIXME, add redo link to tree so we don't leak on crash */
C
Chris Mason 已提交
1699 1700
	ret = btrfs_truncate_in_trans(trans, root, inode);
	BUG_ON(ret);
C
Chris Mason 已提交
1701
	btrfs_update_inode(trans, root, inode);
C
Chris Mason 已提交
1702 1703 1704
	ret = btrfs_end_transaction(trans, root);
	BUG_ON(ret);
	mutex_unlock(&root->fs_info->fs_mutex);
C
Chris Mason 已提交
1705
	btrfs_btree_balance_dirty(root);
C
Chris Mason 已提交
1706 1707
}

1708 1709 1710 1711 1712 1713 1714 1715 1716
static int btrfs_commit_write(struct file *file, struct page *page,
			      unsigned from, unsigned to)
{
	struct inode *inode = page->mapping->host;
	struct buffer_head *bh;
	loff_t pos = ((loff_t)page->index << PAGE_CACHE_SHIFT) + to;

	SetPageUptodate(page);
	bh = page_buffers(page);
C
Chris Mason 已提交
1717
	set_buffer_uptodate(bh);
1718 1719 1720 1721 1722 1723 1724 1725 1726 1727
	if (buffer_mapped(bh) && bh->b_blocknr != 0) {
		set_page_dirty(page);
	}
	if (pos > inode->i_size) {
		i_size_write(inode, pos);
		mark_inode_dirty(inode);
	}
	return 0;
}

C
Chris Mason 已提交
1728 1729 1730 1731 1732 1733 1734 1735 1736 1737 1738 1739 1740 1741 1742 1743 1744 1745 1746 1747 1748 1749 1750 1751 1752 1753 1754 1755 1756 1757 1758 1759 1760 1761 1762 1763 1764 1765 1766 1767 1768 1769 1770 1771 1772 1773 1774 1775 1776 1777 1778 1779 1780 1781
static int btrfs_copy_from_user(loff_t pos, int num_pages, int write_bytes,
				struct page **prepared_pages,
				const char __user * buf)
{
	long page_fault = 0;
	int i;
	int offset = pos & (PAGE_CACHE_SIZE - 1);

	for (i = 0; i < num_pages && write_bytes > 0; i++, offset = 0) {
		size_t count = min_t(size_t,
				     PAGE_CACHE_SIZE - offset, write_bytes);
		struct page *page = prepared_pages[i];
		fault_in_pages_readable(buf, count);

		/* Copy data from userspace to the current page */
		kmap(page);
		page_fault = __copy_from_user(page_address(page) + offset,
					      buf, count);
		/* Flush processor's dcache for this page */
		flush_dcache_page(page);
		kunmap(page);
		buf += count;
		write_bytes -= count;

		if (page_fault)
			break;
	}
	return page_fault ? -EFAULT : 0;
}

static void btrfs_drop_pages(struct page **pages, size_t num_pages)
{
	size_t i;
	for (i = 0; i < num_pages; i++) {
		if (!pages[i])
			break;
		unlock_page(pages[i]);
		mark_page_accessed(pages[i]);
		page_cache_release(pages[i]);
	}
}
static int dirty_and_release_pages(struct btrfs_trans_handle *trans,
				   struct btrfs_root *root,
				   struct file *file,
				   struct page **pages,
				   size_t num_pages,
				   loff_t pos,
				   size_t write_bytes)
{
	int i;
	int offset;
	int err = 0;
	int ret;
	int this_write;
C
Chris Mason 已提交
1782
	struct inode *inode = file->f_path.dentry->d_inode;
1783 1784
	struct buffer_head *bh;
	struct btrfs_file_extent_item *ei;
C
Chris Mason 已提交
1785 1786 1787 1788

	for (i = 0; i < num_pages; i++) {
		offset = pos & (PAGE_CACHE_SIZE -1);
		this_write = min(PAGE_CACHE_SIZE - offset, write_bytes);
C
Chris Mason 已提交
1789 1790 1791 1792
		/* FIXME, one block at a time */

		mutex_lock(&root->fs_info->fs_mutex);
		trans = btrfs_start_transaction(root, 1);
1793
		btrfs_set_trans_block_group(trans, inode);
1794 1795 1796 1797 1798 1799 1800 1801 1802 1803 1804 1805 1806 1807 1808 1809 1810 1811 1812 1813 1814 1815 1816 1817 1818 1819

		bh = page_buffers(pages[i]);
		if (buffer_mapped(bh) && bh->b_blocknr == 0) {
			struct btrfs_key key;
			struct btrfs_path *path;
			char *ptr;
			u32 datasize;

			path = btrfs_alloc_path();
			BUG_ON(!path);
			key.objectid = inode->i_ino;
			key.offset = pages[i]->index << PAGE_CACHE_SHIFT;
			key.flags = 0;
			btrfs_set_key_type(&key, BTRFS_EXTENT_DATA_KEY);
			BUG_ON(write_bytes >= PAGE_CACHE_SIZE);
			datasize = offset +
				btrfs_file_extent_calc_inline_size(write_bytes);
			ret = btrfs_insert_empty_item(trans, root, path, &key,
						      datasize);
			BUG_ON(ret);
			ei = btrfs_item_ptr(btrfs_buffer_leaf(path->nodes[0]),
			       path->slots[0], struct btrfs_file_extent_item);
			btrfs_set_file_extent_generation(ei, trans->transid);
			btrfs_set_file_extent_type(ei,
						   BTRFS_FILE_EXTENT_INLINE);
			ptr = btrfs_file_extent_inline_start(ei);
1820 1821
			btrfs_memcpy(root, path->nodes[0]->b_data,
				     ptr, bh->b_data, offset + write_bytes);
1822 1823
			mark_buffer_dirty(path->nodes[0]);
			btrfs_free_path(path);
C
Chris Mason 已提交
1824
		} else if (buffer_mapped(bh)) {
1825
			btrfs_csum_file_block(trans, root, inode->i_ino,
C
Chris Mason 已提交
1826 1827
				      pages[i]->index << PAGE_CACHE_SHIFT,
				      kmap(pages[i]), PAGE_CACHE_SIZE);
1828 1829
			kunmap(pages[i]);
		}
C
Chris Mason 已提交
1830
		SetPageChecked(pages[i]);
1831
		// btrfs_update_inode_block_group(trans, inode);
C
Chris Mason 已提交
1832 1833 1834 1835
		ret = btrfs_end_transaction(trans, root);
		BUG_ON(ret);
		mutex_unlock(&root->fs_info->fs_mutex);

1836
		ret = btrfs_commit_write(file, pages[i], offset,
C
Chris Mason 已提交
1837 1838 1839 1840 1841 1842 1843 1844 1845 1846 1847 1848 1849
					 offset + this_write);
		pos += this_write;
		if (ret) {
			err = ret;
			goto failed;
		}
		WARN_ON(this_write > write_bytes);
		write_bytes -= this_write;
	}
failed:
	return err;
}

C
Chris Mason 已提交
1850 1851 1852
static int drop_extents(struct btrfs_trans_handle *trans,
			  struct btrfs_root *root,
			  struct inode *inode,
1853
			  u64 start, u64 end, u64 *hint_block)
C
Chris Mason 已提交
1854 1855 1856 1857 1858 1859
{
	int ret;
	struct btrfs_key key;
	struct btrfs_leaf *leaf;
	int slot;
	struct btrfs_file_extent_item *extent;
1860
	u64 extent_end = 0;
C
Chris Mason 已提交
1861 1862 1863 1864 1865
	int keep;
	struct btrfs_file_extent_item old;
	struct btrfs_path *path;
	u64 search_start = start;
	int bookend;
1866 1867 1868 1869
	int found_type;
	int found_extent;
	int found_inline;

C
Chris Mason 已提交
1870 1871 1872
	path = btrfs_alloc_path();
	if (!path)
		return -ENOMEM;
1873 1874 1875 1876 1877
	while(1) {
		btrfs_release_path(root, path);
		ret = btrfs_lookup_file_extent(trans, root, path, inode->i_ino,
					       search_start, -1);
		if (ret < 0)
1878
			goto out;
1879 1880
		if (ret > 0) {
			if (path->slots[0] == 0) {
1881
				ret = 0;
1882 1883 1884
				goto out;
			}
			path->slots[0]--;
1885
		}
C
Chris Mason 已提交
1886 1887
		keep = 0;
		bookend = 0;
1888 1889 1890
		found_extent = 0;
		found_inline = 0;
		extent = NULL;
C
Chris Mason 已提交
1891 1892 1893 1894 1895 1896 1897
		leaf = btrfs_buffer_leaf(path->nodes[0]);
		slot = path->slots[0];
		btrfs_disk_key_to_cpu(&key, &leaf->items[slot].key);
		if (key.offset >= end || key.objectid != inode->i_ino) {
			ret = 0;
			goto out;
		}
1898 1899 1900 1901 1902 1903 1904 1905 1906 1907 1908 1909 1910 1911 1912 1913 1914 1915 1916 1917
		if (btrfs_key_type(&key) != BTRFS_EXTENT_DATA_KEY) {
			ret = 0;
			goto out;
		}
		extent = btrfs_item_ptr(leaf, slot,
					struct btrfs_file_extent_item);
		found_type = btrfs_file_extent_type(extent);
		if (found_type == BTRFS_FILE_EXTENT_REG) {
			extent_end = key.offset +
				(btrfs_file_extent_num_blocks(extent) <<
				 inode->i_blkbits);
			found_extent = 1;
		} else if (found_type == BTRFS_FILE_EXTENT_INLINE) {
			found_inline = 1;
			extent_end = key.offset +
			     btrfs_file_extent_inline_len(leaf->items + slot);
		}

		if (!found_extent && !found_inline) {
			ret = 0;
1918
			goto out;
1919 1920 1921 1922
		}

		if (search_start >= extent_end) {
			ret = 0;
1923
			goto out;
1924 1925
		}

C
Chris Mason 已提交
1926 1927 1928 1929 1930
		if (found_inline) {
			u64 mask = root->blocksize - 1;
			search_start = (extent_end + mask) & ~mask;
		} else
			search_start = extent_end;
C
Chris Mason 已提交
1931 1932

		if (end < extent_end && end >= key.offset) {
1933
			if (found_extent) {
C
Chris Mason 已提交
1934 1935 1936 1937
				u64 disk_blocknr =
					btrfs_file_extent_disk_blocknr(extent);
				u64 disk_num_blocks =
				      btrfs_file_extent_disk_num_blocks(extent);
1938
				memcpy(&old, extent, sizeof(old));
C
Chris Mason 已提交
1939 1940 1941 1942 1943
				if (disk_blocknr != 0) {
					ret = btrfs_inc_extent_ref(trans, root,
					         disk_blocknr, disk_num_blocks);
					BUG_ON(ret);
				}
1944 1945
			}
			WARN_ON(found_inline);
C
Chris Mason 已提交
1946 1947 1948 1949 1950
			bookend = 1;
		}

		if (start > key.offset) {
			u64 new_num;
1951
			u64 old_num;
C
Chris Mason 已提交
1952 1953 1954
			/* truncate existing extent */
			keep = 1;
			WARN_ON(start & (root->blocksize - 1));
1955 1956 1957 1958
			if (found_extent) {
				new_num = (start - key.offset) >>
					inode->i_blkbits;
				old_num = btrfs_file_extent_num_blocks(extent);
1959 1960
				*hint_block =
					btrfs_file_extent_disk_blocknr(extent);
C
Chris Mason 已提交
1961 1962 1963 1964
				if (btrfs_file_extent_disk_blocknr(extent)) {
					inode->i_blocks -=
						(old_num - new_num) << 3;
				}
1965 1966 1967 1968 1969 1970
				btrfs_set_file_extent_num_blocks(extent,
								 new_num);
				mark_buffer_dirty(path->nodes[0]);
			} else {
				WARN_ON(1);
			}
C
Chris Mason 已提交
1971 1972
		}
		if (!keep) {
1973 1974 1975 1976 1977 1978 1979 1980 1981 1982
			u64 disk_blocknr = 0;
			u64 disk_num_blocks = 0;
			u64 extent_num_blocks = 0;
			if (found_extent) {
				disk_blocknr =
				      btrfs_file_extent_disk_blocknr(extent);
				disk_num_blocks =
				      btrfs_file_extent_disk_num_blocks(extent);
				extent_num_blocks =
				      btrfs_file_extent_num_blocks(extent);
1983 1984
				*hint_block =
					btrfs_file_extent_disk_blocknr(extent);
1985
			}
C
Chris Mason 已提交
1986 1987 1988
			ret = btrfs_del_item(trans, root, path);
			BUG_ON(ret);
			btrfs_release_path(root, path);
1989
			extent = NULL;
C
Chris Mason 已提交
1990
			if (found_extent && disk_blocknr != 0) {
1991
				inode->i_blocks -= extent_num_blocks << 3;
1992 1993 1994 1995
				ret = btrfs_free_extent(trans, root,
							disk_blocknr,
							disk_num_blocks, 0);
			}
C
Chris Mason 已提交
1996 1997 1998 1999 2000 2001 2002

			BUG_ON(ret);
			if (!bookend && search_start >= end) {
				ret = 0;
				goto out;
			}
			if (!bookend)
2003
				continue;
C
Chris Mason 已提交
2004
		}
2005
		if (bookend && found_extent) {
C
Chris Mason 已提交
2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 2020 2021 2022 2023 2024 2025 2026 2027 2028 2029
			/* create bookend */
			struct btrfs_key ins;
			ins.objectid = inode->i_ino;
			ins.offset = end;
			ins.flags = 0;
			btrfs_set_key_type(&ins, BTRFS_EXTENT_DATA_KEY);

			btrfs_release_path(root, path);
			ret = btrfs_insert_empty_item(trans, root, path, &ins,
						      sizeof(*extent));
			BUG_ON(ret);
			extent = btrfs_item_ptr(
				    btrfs_buffer_leaf(path->nodes[0]),
				    path->slots[0],
				    struct btrfs_file_extent_item);
			btrfs_set_file_extent_disk_blocknr(extent,
				    btrfs_file_extent_disk_blocknr(&old));
			btrfs_set_file_extent_disk_num_blocks(extent,
				    btrfs_file_extent_disk_num_blocks(&old));

			btrfs_set_file_extent_offset(extent,
				    btrfs_file_extent_offset(&old) +
				    ((end - key.offset) >> inode->i_blkbits));
			WARN_ON(btrfs_file_extent_num_blocks(&old) <
C
Chris Mason 已提交
2030
				(extent_end - end) >> inode->i_blkbits);
C
Chris Mason 已提交
2031
			btrfs_set_file_extent_num_blocks(extent,
C
Chris Mason 已提交
2032
				    (extent_end - end) >> inode->i_blkbits);
C
Chris Mason 已提交
2033

2034 2035
			btrfs_set_file_extent_type(extent,
						   BTRFS_FILE_EXTENT_REG);
C
Chris Mason 已提交
2036 2037 2038
			btrfs_set_file_extent_generation(extent,
				    btrfs_file_extent_generation(&old));
			btrfs_mark_buffer_dirty(path->nodes[0]);
C
Chris Mason 已提交
2039 2040 2041 2042
			if (btrfs_file_extent_disk_blocknr(&old) != 0) {
				inode->i_blocks +=
				      btrfs_file_extent_num_blocks(extent) << 3;
			}
C
Chris Mason 已提交
2043
			ret = 0;
2044
			goto out;
C
Chris Mason 已提交
2045 2046 2047 2048 2049 2050 2051 2052
		}
	}
out:
	btrfs_free_path(path);
	return ret;
}

static int prepare_pages(struct btrfs_root *root,
C
Chris Mason 已提交
2053 2054 2055 2056
			 struct file *file,
			 struct page **pages,
			 size_t num_pages,
			 loff_t pos,
2057 2058
			 unsigned long first_index,
			 unsigned long last_index,
2059 2060
			 size_t write_bytes,
			 u64 alloc_extent_start)
C
Chris Mason 已提交
2061 2062 2063 2064 2065 2066 2067
{
	int i;
	unsigned long index = pos >> PAGE_CACHE_SHIFT;
	struct inode *inode = file->f_path.dentry->d_inode;
	int offset;
	int err = 0;
	int this_write;
2068 2069
	struct buffer_head *bh;
	struct buffer_head *head;
C
Chris Mason 已提交
2070 2071 2072 2073 2074 2075 2076 2077 2078 2079
	loff_t isize = i_size_read(inode);

	memset(pages, 0, num_pages * sizeof(struct page *));

	for (i = 0; i < num_pages; i++) {
		pages[i] = grab_cache_page(inode->i_mapping, index + i);
		if (!pages[i]) {
			err = -ENOMEM;
			goto failed_release;
		}
C
Chris Mason 已提交
2080 2081
		cancel_dirty_page(pages[i], PAGE_CACHE_SIZE);
		wait_on_page_writeback(pages[i]);
C
Chris Mason 已提交
2082 2083
		offset = pos & (PAGE_CACHE_SIZE -1);
		this_write = min(PAGE_CACHE_SIZE - offset, write_bytes);
C
Chris Mason 已提交
2084 2085 2086 2087 2088
		if (!page_has_buffers(pages[i])) {
			create_empty_buffers(pages[i],
					     root->fs_info->sb->s_blocksize,
					     (1 << BH_Uptodate));
		}
2089 2090 2091 2092 2093 2094 2095 2096 2097
		head = page_buffers(pages[i]);
		bh = head;
		do {
			err = btrfs_map_bh_to_logical(root, bh,
						      alloc_extent_start);
			BUG_ON(err);
			if (err)
				goto failed_truncate;
			bh = bh->b_this_page;
2098 2099
			if (alloc_extent_start)
				alloc_extent_start++;
2100
		} while (bh != head);
C
Chris Mason 已提交
2101 2102 2103 2104 2105 2106 2107 2108 2109 2110 2111 2112 2113 2114 2115 2116 2117 2118 2119 2120 2121 2122 2123 2124 2125
		pos += this_write;
		WARN_ON(this_write > write_bytes);
		write_bytes -= this_write;
	}
	return 0;

failed_release:
	btrfs_drop_pages(pages, num_pages);
	return err;

failed_truncate:
	btrfs_drop_pages(pages, num_pages);
	if (pos > isize)
		vmtruncate(inode, isize);
	return err;
}

static ssize_t btrfs_file_write(struct file *file, const char __user *buf,
				size_t count, loff_t *ppos)
{
	loff_t pos;
	size_t num_written = 0;
	int err = 0;
	int ret = 0;
	struct inode *inode = file->f_path.dentry->d_inode;
2126
	struct btrfs_root *root = BTRFS_I(inode)->root;
C
Chris Mason 已提交
2127
	struct page *pages[8];
C
Chris Mason 已提交
2128
	struct page *pinned[2];
2129 2130
	unsigned long first_index;
	unsigned long last_index;
2131 2132 2133
	u64 start_pos;
	u64 num_blocks;
	u64 alloc_extent_start;
2134
	u64 hint_block;
2135
	struct btrfs_trans_handle *trans;
C
Chris Mason 已提交
2136
	struct btrfs_key ins;
C
Chris Mason 已提交
2137 2138
	pinned[0] = NULL;
	pinned[1] = NULL;
C
Chris Mason 已提交
2139 2140 2141 2142 2143 2144 2145 2146 2147 2148 2149 2150 2151 2152
	if (file->f_flags & O_DIRECT)
		return -EINVAL;
	pos = *ppos;
	vfs_check_frozen(inode->i_sb, SB_FREEZE_WRITE);
	current->backing_dev_info = inode->i_mapping->backing_dev_info;
	err = generic_write_checks(file, &pos, &count, S_ISBLK(inode->i_mode));
	if (err)
		goto out;
	if (count == 0)
		goto out;
	err = remove_suid(file->f_path.dentry);
	if (err)
		goto out;
	file_update_time(file);
2153 2154 2155 2156 2157

	start_pos = pos & ~((u64)PAGE_CACHE_SIZE - 1);
	num_blocks = (count + pos - start_pos + root->blocksize - 1) >>
			inode->i_blkbits;

C
Chris Mason 已提交
2158
	mutex_lock(&inode->i_mutex);
2159 2160
	first_index = pos >> PAGE_CACHE_SHIFT;
	last_index = (pos + count) >> PAGE_CACHE_SHIFT;
2161

C
Chris Mason 已提交
2162
	if ((pos & (PAGE_CACHE_SIZE - 1))) {
2163 2164 2165 2166
		pinned[0] = grab_cache_page(inode->i_mapping, first_index);
		if (!PageUptodate(pinned[0])) {
			ret = mpage_readpage(pinned[0], btrfs_get_block);
			BUG_ON(ret);
C
Chris Mason 已提交
2167
			wait_on_page_locked(pinned[0]);
2168 2169 2170 2171
		} else {
			unlock_page(pinned[0]);
		}
	}
C
Chris Mason 已提交
2172
	if ((pos + count) & (PAGE_CACHE_SIZE - 1)) {
2173 2174 2175 2176
		pinned[1] = grab_cache_page(inode->i_mapping, last_index);
		if (!PageUptodate(pinned[1])) {
			ret = mpage_readpage(pinned[1], btrfs_get_block);
			BUG_ON(ret);
C
Chris Mason 已提交
2177
			wait_on_page_locked(pinned[1]);
2178 2179 2180 2181 2182
		} else {
			unlock_page(pinned[1]);
		}
	}

2183 2184 2185 2186
	mutex_lock(&root->fs_info->fs_mutex);
	trans = btrfs_start_transaction(root, 1);
	if (!trans) {
		err = -ENOMEM;
C
Chris Mason 已提交
2187
		mutex_unlock(&root->fs_info->fs_mutex);
2188 2189
		goto out_unlock;
	}
2190
	btrfs_set_trans_block_group(trans, inode);
2191 2192
	/* FIXME blocksize != 4096 */
	inode->i_blocks += num_blocks << 3;
2193
	hint_block = 0;
C
Chris Mason 已提交
2194
	if (start_pos < inode->i_size) {
2195
		/* FIXME blocksize != pagesize */
C
Chris Mason 已提交
2196 2197 2198
		ret = drop_extents(trans, root, inode,
				   start_pos,
				   (pos + count + root->blocksize -1) &
2199
				   ~((u64)root->blocksize - 1), &hint_block);
2200
		BUG_ON(ret);
C
Chris Mason 已提交
2201
	}
C
Chris Mason 已提交
2202 2203 2204 2205 2206 2207 2208 2209 2210 2211 2212 2213 2214 2215 2216
	if (inode->i_size < start_pos) {
		u64 last_pos_in_file;
		u64 hole_size;
		u64 mask = root->blocksize - 1;
		last_pos_in_file = (inode->i_size + mask) & ~mask;
		hole_size = (start_pos - last_pos_in_file + mask) & ~mask;
		hole_size >>= inode->i_blkbits;
		if (last_pos_in_file < start_pos) {
			ret = btrfs_insert_file_extent(trans, root,
						       inode->i_ino,
						       last_pos_in_file,
						       0, 0, hole_size);
		}
		BUG_ON(ret);
	}
2217 2218
	if (inode->i_size >= PAGE_CACHE_SIZE || pos + count < inode->i_size ||
	    pos + count - start_pos > BTRFS_MAX_INLINE_DATA_SIZE(root)) {
2219
		ret = btrfs_alloc_extent(trans, root, inode->i_ino,
2220 2221
					 num_blocks, hint_block, (u64)-1,
					 &ins, 1);
2222 2223
		BUG_ON(ret);
		ret = btrfs_insert_file_extent(trans, root, inode->i_ino,
C
Chris Mason 已提交
2224 2225
				       start_pos, ins.objectid, ins.offset,
				       ins.offset);
2226 2227 2228 2229 2230
		BUG_ON(ret);
	} else {
		ins.offset = 0;
		ins.objectid = 0;
	}
2231
	BUG_ON(ret);
C
Chris Mason 已提交
2232
	alloc_extent_start = ins.objectid;
2233
	// btrfs_update_inode_block_group(trans, inode);
C
Chris Mason 已提交
2234
	ret = btrfs_end_transaction(trans, root);
2235 2236
	mutex_unlock(&root->fs_info->fs_mutex);

C
Chris Mason 已提交
2237 2238 2239 2240 2241
	while(count > 0) {
		size_t offset = pos & (PAGE_CACHE_SIZE - 1);
		size_t write_bytes = min(count, PAGE_CACHE_SIZE - offset);
		size_t num_pages = (write_bytes + PAGE_CACHE_SIZE - 1) >>
					PAGE_CACHE_SHIFT;
C
Chris Mason 已提交
2242 2243 2244

		memset(pages, 0, sizeof(pages));
		ret = prepare_pages(root, file, pages, num_pages,
2245 2246
				    pos, first_index, last_index,
				    write_bytes, alloc_extent_start);
C
Chris Mason 已提交
2247
		BUG_ON(ret);
C
Chris Mason 已提交
2248

2249
		/* FIXME blocks != pagesize */
2250 2251
		if (alloc_extent_start)
			alloc_extent_start += num_pages;
C
Chris Mason 已提交
2252 2253 2254 2255
		ret = btrfs_copy_from_user(pos, num_pages,
					   write_bytes, pages, buf);
		BUG_ON(ret);

C
Chris Mason 已提交
2256
		ret = dirty_and_release_pages(NULL, root, file, pages,
2257
					      num_pages, pos, write_bytes);
C
Chris Mason 已提交
2258 2259 2260 2261 2262 2263 2264 2265 2266
		BUG_ON(ret);
		btrfs_drop_pages(pages, num_pages);

		buf += write_bytes;
		count -= write_bytes;
		pos += write_bytes;
		num_written += write_bytes;

		balance_dirty_pages_ratelimited(inode->i_mapping);
C
Chris Mason 已提交
2267
		btrfs_btree_balance_dirty(root);
C
Chris Mason 已提交
2268 2269
		cond_resched();
	}
2270
out_unlock:
C
Chris Mason 已提交
2271 2272
	mutex_unlock(&inode->i_mutex);
out:
2273 2274 2275 2276
	if (pinned[0])
		page_cache_release(pinned[0]);
	if (pinned[1])
		page_cache_release(pinned[1]);
C
Chris Mason 已提交
2277 2278
	*ppos = pos;
	current->backing_dev_info = NULL;
2279
	mark_inode_dirty(inode);
C
Chris Mason 已提交
2280 2281 2282
	return num_written ? num_written : err;
}

C
Chris Mason 已提交
2283 2284 2285 2286 2287
static int btrfs_read_actor(read_descriptor_t *desc, struct page *page,
			unsigned long offset, unsigned long size)
{
	char *kaddr;
	unsigned long left, count = desc->count;
2288
	struct inode *inode = page->mapping->host;
C
Chris Mason 已提交
2289 2290 2291 2292 2293 2294

	if (size > count)
		size = count;

	if (!PageChecked(page)) {
		/* FIXME, do it per block */
2295
		struct btrfs_root *root = BTRFS_I(inode)->root;
C
Chris Mason 已提交
2296 2297 2298 2299 2300 2301 2302 2303 2304 2305
		int ret;
		struct buffer_head *bh;

		if (page_has_buffers(page)) {
			bh = page_buffers(page);
			if (!buffer_mapped(bh)) {
				SetPageChecked(page);
				goto checked;
			}
		}
2306

C
Chris Mason 已提交
2307
		ret = btrfs_csum_verify_file_block(root,
2308 2309 2310
				  page->mapping->host->i_ino,
				  page->index << PAGE_CACHE_SHIFT,
				  kmap(page), PAGE_CACHE_SIZE);
C
Chris Mason 已提交
2311
		if (ret) {
C
Chris Mason 已提交
2312
			if (ret != -ENOENT) {
C
Chris Mason 已提交
2313
				printk("failed to verify ino %lu page %lu ret %d\n",
C
Chris Mason 已提交
2314
				       page->mapping->host->i_ino,
C
Chris Mason 已提交
2315 2316
				       page->index, ret);
				memset(page_address(page), 1, PAGE_CACHE_SIZE);
C
Chris Mason 已提交
2317 2318
				flush_dcache_page(page);
			}
C
Chris Mason 已提交
2319 2320 2321 2322
		}
		SetPageChecked(page);
		kunmap(page);
	}
C
Chris Mason 已提交
2323
checked:
C
Chris Mason 已提交
2324 2325 2326 2327 2328 2329 2330 2331 2332 2333 2334 2335 2336 2337 2338 2339 2340 2341 2342 2343 2344 2345 2346 2347 2348 2349 2350 2351 2352 2353 2354 2355 2356 2357 2358 2359 2360 2361 2362 2363 2364 2365 2366 2367 2368 2369 2370 2371 2372 2373 2374 2375 2376 2377 2378 2379 2380 2381 2382 2383 2384 2385 2386 2387 2388 2389 2390 2391 2392 2393 2394 2395 2396 2397 2398 2399 2400 2401 2402 2403 2404 2405 2406 2407 2408 2409 2410
	/*
	 * Faults on the destination of a read are common, so do it before
	 * taking the kmap.
	 */
	if (!fault_in_pages_writeable(desc->arg.buf, size)) {
		kaddr = kmap_atomic(page, KM_USER0);
		left = __copy_to_user_inatomic(desc->arg.buf,
						kaddr + offset, size);
		kunmap_atomic(kaddr, KM_USER0);
		if (left == 0)
			goto success;
	}

	/* Do it the slow way */
	kaddr = kmap(page);
	left = __copy_to_user(desc->arg.buf, kaddr + offset, size);
	kunmap(page);

	if (left) {
		size -= left;
		desc->error = -EFAULT;
	}
success:
	desc->count = count - size;
	desc->written += size;
	desc->arg.buf += size;
	return size;
}

/**
 * btrfs_file_aio_read - filesystem read routine
 * @iocb:	kernel I/O control block
 * @iov:	io vector request
 * @nr_segs:	number of segments in the iovec
 * @pos:	current file position
 */
static ssize_t btrfs_file_aio_read(struct kiocb *iocb, const struct iovec *iov,
				   unsigned long nr_segs, loff_t pos)
{
	struct file *filp = iocb->ki_filp;
	ssize_t retval;
	unsigned long seg;
	size_t count;
	loff_t *ppos = &iocb->ki_pos;

	count = 0;
	for (seg = 0; seg < nr_segs; seg++) {
		const struct iovec *iv = &iov[seg];

		/*
		 * If any segment has a negative length, or the cumulative
		 * length ever wraps negative then return -EINVAL.
		 */
		count += iv->iov_len;
		if (unlikely((ssize_t)(count|iv->iov_len) < 0))
			return -EINVAL;
		if (access_ok(VERIFY_WRITE, iv->iov_base, iv->iov_len))
			continue;
		if (seg == 0)
			return -EFAULT;
		nr_segs = seg;
		count -= iv->iov_len;	/* This segment is no good */
		break;
	}
	retval = 0;
	if (count) {
		for (seg = 0; seg < nr_segs; seg++) {
			read_descriptor_t desc;

			desc.written = 0;
			desc.arg.buf = iov[seg].iov_base;
			desc.count = iov[seg].iov_len;
			if (desc.count == 0)
				continue;
			desc.error = 0;
			do_generic_file_read(filp, ppos, &desc,
					     btrfs_read_actor);
			retval += desc.written;
			if (desc.error) {
				retval = retval ?: desc.error;
				break;
			}
		}
	}
	return retval;
}

C
Chris Mason 已提交
2411 2412 2413 2414 2415 2416 2417 2418 2419 2420
static int create_subvol(struct btrfs_root *root, char *name, int namelen)
{
	struct btrfs_trans_handle *trans;
	struct btrfs_key key;
	struct btrfs_root_item root_item;
	struct btrfs_inode_item *inode_item;
	struct buffer_head *subvol;
	struct btrfs_leaf *leaf;
	struct btrfs_root *new_root;
	struct inode *inode;
2421
	struct inode *dir;
C
Chris Mason 已提交
2422 2423 2424 2425 2426 2427 2428 2429
	int ret;
	u64 objectid;
	u64 new_dirid = BTRFS_FIRST_FREE_OBJECTID;

	mutex_lock(&root->fs_info->fs_mutex);
	trans = btrfs_start_transaction(root, 1);
	BUG_ON(!trans);

2430
	subvol = btrfs_alloc_free_block(trans, root, 0);
C
Chris Mason 已提交
2431 2432
	if (subvol == NULL)
		return -ENOSPC;
C
Chris Mason 已提交
2433 2434 2435
	leaf = btrfs_buffer_leaf(subvol);
	btrfs_set_header_nritems(&leaf->header, 0);
	btrfs_set_header_level(&leaf->header, 0);
2436
	btrfs_set_header_blocknr(&leaf->header, bh_blocknr(subvol));
C
Chris Mason 已提交
2437
	btrfs_set_header_generation(&leaf->header, trans->transid);
2438
	btrfs_set_header_owner(&leaf->header, root->root_key.objectid);
C
Chris Mason 已提交
2439 2440
	memcpy(leaf->header.fsid, root->fs_info->disk_super->fsid,
	       sizeof(leaf->header.fsid));
2441
	mark_buffer_dirty(subvol);
C
Chris Mason 已提交
2442 2443 2444 2445 2446 2447 2448 2449 2450

	inode_item = &root_item.inode;
	memset(inode_item, 0, sizeof(*inode_item));
	btrfs_set_inode_generation(inode_item, 1);
	btrfs_set_inode_size(inode_item, 3);
	btrfs_set_inode_nlink(inode_item, 1);
	btrfs_set_inode_nblocks(inode_item, 1);
	btrfs_set_inode_mode(inode_item, S_IFDIR | 0755);

2451
	btrfs_set_root_blocknr(&root_item, bh_blocknr(subvol));
C
Chris Mason 已提交
2452
	btrfs_set_root_refs(&root_item, 1);
C
Chris Mason 已提交
2453 2454
	brelse(subvol);
	subvol = NULL;
C
Chris Mason 已提交
2455 2456 2457 2458 2459 2460 2461 2462 2463 2464 2465 2466 2467 2468 2469 2470 2471 2472 2473

	ret = btrfs_find_free_objectid(trans, root->fs_info->tree_root,
				       0, &objectid);
	BUG_ON(ret);

	btrfs_set_root_dirid(&root_item, new_dirid);

	key.objectid = objectid;
	key.offset = 1;
	key.flags = 0;
	btrfs_set_key_type(&key, BTRFS_ROOT_ITEM_KEY);
	ret = btrfs_insert_root(trans, root->fs_info->tree_root, &key,
				&root_item);
	BUG_ON(ret);

	/*
	 * insert the directory item
	 */
	key.offset = (u64)-1;
2474
	dir = root->fs_info->sb->s_root->d_inode;
C
Chris Mason 已提交
2475
	ret = btrfs_insert_dir_item(trans, root->fs_info->tree_root,
C
Chris Mason 已提交
2476 2477
				    name, namelen, dir->i_ino, &key,
				    BTRFS_FT_DIR);
C
Chris Mason 已提交
2478 2479 2480 2481 2482 2483 2484 2485 2486 2487 2488
	BUG_ON(ret);

	ret = btrfs_commit_transaction(trans, root);
	BUG_ON(ret);

	new_root = btrfs_read_fs_root(root->fs_info, &key);
	BUG_ON(!new_root);

	trans = btrfs_start_transaction(new_root, 1);
	BUG_ON(!trans);

2489 2490
	inode = btrfs_new_inode(trans, new_root, new_dirid,
				BTRFS_I(dir)->block_group, S_IFDIR | 0700);
C
Chris Mason 已提交
2491 2492 2493 2494 2495 2496 2497 2498 2499 2500 2501 2502 2503 2504 2505 2506 2507
	inode->i_op = &btrfs_dir_inode_operations;
	inode->i_fop = &btrfs_dir_file_operations;

	ret = btrfs_make_empty_dir(trans, new_root, new_dirid, new_dirid);
	BUG_ON(ret);

	inode->i_nlink = 1;
	inode->i_size = 6;
	ret = btrfs_update_inode(trans, new_root, inode);
	BUG_ON(ret);

	ret = btrfs_commit_transaction(trans, new_root);
	BUG_ON(ret);

	iput(inode);

	mutex_unlock(&root->fs_info->fs_mutex);
C
Chris Mason 已提交
2508
	btrfs_btree_balance_dirty(root);
C
Chris Mason 已提交
2509 2510 2511
	return 0;
}

C
Chris Mason 已提交
2512 2513 2514 2515 2516 2517 2518 2519
static int create_snapshot(struct btrfs_root *root, char *name, int namelen)
{
	struct btrfs_trans_handle *trans;
	struct btrfs_key key;
	struct btrfs_root_item new_root_item;
	int ret;
	u64 objectid;

C
Chris Mason 已提交
2520 2521 2522
	if (!root->ref_cows)
		return -EINVAL;

C
Chris Mason 已提交
2523 2524 2525 2526 2527 2528 2529
	mutex_lock(&root->fs_info->fs_mutex);
	trans = btrfs_start_transaction(root, 1);
	BUG_ON(!trans);

	ret = btrfs_update_inode(trans, root, root->inode);
	BUG_ON(ret);

C
Chris Mason 已提交
2530 2531
	ret = btrfs_find_free_objectid(trans, root->fs_info->tree_root,
				       0, &objectid);
C
Chris Mason 已提交
2532 2533 2534 2535 2536 2537 2538 2539 2540
	BUG_ON(ret);

	memcpy(&new_root_item, &root->root_item,
	       sizeof(new_root_item));

	key.objectid = objectid;
	key.offset = 1;
	key.flags = 0;
	btrfs_set_key_type(&key, BTRFS_ROOT_ITEM_KEY);
2541
	btrfs_set_root_blocknr(&new_root_item, bh_blocknr(root->node));
C
Chris Mason 已提交
2542 2543 2544 2545 2546 2547 2548 2549 2550 2551 2552 2553

	ret = btrfs_insert_root(trans, root->fs_info->tree_root, &key,
				&new_root_item);
	BUG_ON(ret);

	/*
	 * insert the directory item
	 */
	key.offset = (u64)-1;
	ret = btrfs_insert_dir_item(trans, root->fs_info->tree_root,
				    name, namelen,
				    root->fs_info->sb->s_root->d_inode->i_ino,
C
Chris Mason 已提交
2554
				    &key, BTRFS_FT_DIR);
C
Chris Mason 已提交
2555 2556 2557 2558 2559 2560 2561 2562 2563

	BUG_ON(ret);

	ret = btrfs_inc_root_ref(trans, root);
	BUG_ON(ret);

	ret = btrfs_commit_transaction(trans, root);
	BUG_ON(ret);
	mutex_unlock(&root->fs_info->fs_mutex);
C
Chris Mason 已提交
2564
	btrfs_btree_balance_dirty(root);
C
Chris Mason 已提交
2565 2566 2567
	return 0;
}

2568 2569 2570 2571 2572 2573 2574 2575 2576 2577 2578 2579
static int add_disk(struct btrfs_root *root, char *name, int namelen)
{
	struct block_device *bdev;
	struct btrfs_path *path;
	struct super_block *sb = root->fs_info->sb;
	struct btrfs_root *dev_root = root->fs_info->dev_root;
	struct btrfs_trans_handle *trans;
	struct btrfs_device_item *dev_item;
	struct btrfs_key key;
	u16 item_size;
	u64 num_blocks;
	u64 new_blocks;
2580
	u64 device_id;
2581
	int ret;
2582

2583 2584 2585 2586 2587 2588 2589 2590 2591 2592 2593 2594 2595 2596 2597 2598 2599 2600 2601 2602 2603 2604 2605 2606 2607 2608 2609 2610 2611 2612 2613 2614 2615 2616
printk("adding disk %s\n", name);
	path = btrfs_alloc_path();
	if (!path)
		return -ENOMEM;
	num_blocks = btrfs_super_total_blocks(root->fs_info->disk_super);
	bdev = open_bdev_excl(name, O_RDWR, sb);
	if (IS_ERR(bdev)) {
		ret = PTR_ERR(bdev);
printk("open bdev excl failed ret %d\n", ret);
		goto out_nolock;
	}
	set_blocksize(bdev, sb->s_blocksize);
	new_blocks = bdev->bd_inode->i_size >> sb->s_blocksize_bits;
	key.objectid = num_blocks;
	key.offset = new_blocks;
	key.flags = 0;
	btrfs_set_key_type(&key, BTRFS_DEV_ITEM_KEY);

	mutex_lock(&dev_root->fs_info->fs_mutex);
	trans = btrfs_start_transaction(dev_root, 1);
	item_size = sizeof(*dev_item) + namelen;
printk("insert empty on %Lu %Lu %u size %d\n", num_blocks, new_blocks, key.flags, item_size);
	ret = btrfs_insert_empty_item(trans, dev_root, path, &key, item_size);
	if (ret) {
printk("insert failed %d\n", ret);
		close_bdev_excl(bdev);
		if (ret > 0)
			ret = -EEXIST;
		goto out;
	}
	dev_item = btrfs_item_ptr(btrfs_buffer_leaf(path->nodes[0]),
				  path->slots[0], struct btrfs_device_item);
	btrfs_set_device_pathlen(dev_item, namelen);
	memcpy(dev_item + 1, name, namelen);
2617 2618 2619 2620

	device_id = btrfs_super_last_device_id(root->fs_info->disk_super) + 1;
	btrfs_set_super_last_device_id(root->fs_info->disk_super, device_id);
	btrfs_set_device_id(dev_item, device_id);
2621 2622
	mark_buffer_dirty(path->nodes[0]);

2623 2624
	ret = btrfs_insert_dev_radix(root, bdev, device_id, num_blocks,
				     new_blocks);
2625 2626 2627 2628 2629 2630 2631 2632 2633 2634 2635 2636 2637 2638 2639

	if (!ret) {
		btrfs_set_super_total_blocks(root->fs_info->disk_super,
					     num_blocks + new_blocks);
		i_size_write(root->fs_info->btree_inode,
			     (num_blocks + new_blocks) <<
			     root->fs_info->btree_inode->i_blkbits);
	}

out:
	ret = btrfs_commit_transaction(trans, dev_root);
	BUG_ON(ret);
	mutex_unlock(&root->fs_info->fs_mutex);
out_nolock:
	btrfs_free_path(path);
C
Chris Mason 已提交
2640
	btrfs_btree_balance_dirty(root);
2641 2642 2643 2644

	return ret;
}

C
Chris Mason 已提交
2645 2646 2647 2648 2649
static int btrfs_ioctl(struct inode *inode, struct file *filp, unsigned int
		       cmd, unsigned long arg)
{
	struct btrfs_root *root = BTRFS_I(inode)->root;
	struct btrfs_ioctl_vol_args vol_args;
2650
	int ret = 0;
2651
	struct btrfs_dir_item *di;
C
Chris Mason 已提交
2652
	int namelen;
C
Chris Mason 已提交
2653 2654
	struct btrfs_path *path;
	u64 root_dirid;
C
Chris Mason 已提交
2655 2656 2657 2658 2659 2660 2661 2662 2663 2664

	switch (cmd) {
	case BTRFS_IOC_SNAP_CREATE:
		if (copy_from_user(&vol_args,
				   (struct btrfs_ioctl_vol_args __user *)arg,
				   sizeof(vol_args)))
			return -EFAULT;
		namelen = strlen(vol_args.name);
		if (namelen > BTRFS_VOL_NAME_MAX)
			return -EINVAL;
C
Chris Mason 已提交
2665 2666 2667
		path = btrfs_alloc_path();
		if (!path)
			return -ENOMEM;
2668
		root_dirid = root->fs_info->sb->s_root->d_inode->i_ino,
C
Chris Mason 已提交
2669
		mutex_lock(&root->fs_info->fs_mutex);
2670
		di = btrfs_lookup_dir_item(NULL, root->fs_info->tree_root,
C
Chris Mason 已提交
2671 2672 2673
				    path, root_dirid,
				    vol_args.name, namelen, 0);
		mutex_unlock(&root->fs_info->fs_mutex);
2674
		btrfs_free_path(path);
2675
		if (di && !IS_ERR(di))
C
Chris Mason 已提交
2676 2677 2678 2679 2680 2681
			return -EEXIST;

		if (root == root->fs_info->tree_root)
			ret = create_subvol(root, vol_args.name, namelen);
		else
			ret = create_snapshot(root, vol_args.name, namelen);
C
Chris Mason 已提交
2682 2683
		WARN_ON(ret);
		break;
2684 2685 2686 2687 2688 2689 2690 2691 2692 2693 2694
	case BTRFS_IOC_ADD_DISK:
		if (copy_from_user(&vol_args,
				   (struct btrfs_ioctl_vol_args __user *)arg,
				   sizeof(vol_args)))
			return -EFAULT;
		namelen = strlen(vol_args.name);
		if (namelen > BTRFS_VOL_NAME_MAX)
			return -EINVAL;
		vol_args.name[namelen] = '\0';
		ret = add_disk(root, vol_args.name, namelen);
		break;
C
Chris Mason 已提交
2695 2696 2697
	default:
		return -ENOTTY;
	}
2698
	return ret;
C
Chris Mason 已提交
2699 2700
}

C
Chris Mason 已提交
2701 2702 2703 2704 2705 2706 2707 2708 2709 2710 2711 2712 2713 2714 2715 2716 2717 2718 2719 2720 2721 2722 2723 2724 2725 2726 2727 2728 2729 2730 2731 2732
static struct kmem_cache *btrfs_inode_cachep;
struct kmem_cache *btrfs_trans_handle_cachep;
struct kmem_cache *btrfs_transaction_cachep;
struct kmem_cache *btrfs_bit_radix_cachep;
struct kmem_cache *btrfs_path_cachep;

/*
 * Called inside transaction, so use GFP_NOFS
 */
static struct inode *btrfs_alloc_inode(struct super_block *sb)
{
	struct btrfs_inode *ei;

	ei = kmem_cache_alloc(btrfs_inode_cachep, GFP_NOFS);
	if (!ei)
		return NULL;
	return &ei->vfs_inode;
}

static void btrfs_destroy_inode(struct inode *inode)
{
	WARN_ON(!list_empty(&inode->i_dentry));
	WARN_ON(inode->i_data.nrpages);

	kmem_cache_free(btrfs_inode_cachep, BTRFS_I(inode));
}

static void init_once(void * foo, struct kmem_cache * cachep,
		      unsigned long flags)
{
	struct btrfs_inode *ei = (struct btrfs_inode *) foo;

C
Chris Mason 已提交
2733
	if ((flags & (SLAB_CTOR_CONSTRUCTOR)) ==
C
Chris Mason 已提交
2734 2735 2736 2737 2738 2739 2740 2741 2742 2743 2744 2745 2746 2747 2748 2749 2750 2751 2752 2753 2754 2755 2756 2757 2758 2759 2760 2761 2762 2763 2764 2765 2766 2767 2768 2769 2770 2771 2772 2773 2774 2775 2776 2777 2778 2779 2780 2781
	    SLAB_CTOR_CONSTRUCTOR) {
		inode_init_once(&ei->vfs_inode);
	}
}

static int init_inodecache(void)
{
	btrfs_inode_cachep = kmem_cache_create("btrfs_inode_cache",
					     sizeof(struct btrfs_inode),
					     0, (SLAB_RECLAIM_ACCOUNT|
						SLAB_MEM_SPREAD),
					     init_once, NULL);
	btrfs_trans_handle_cachep = kmem_cache_create("btrfs_trans_handle_cache",
					     sizeof(struct btrfs_trans_handle),
					     0, (SLAB_RECLAIM_ACCOUNT|
						SLAB_MEM_SPREAD),
					     NULL, NULL);
	btrfs_transaction_cachep = kmem_cache_create("btrfs_transaction_cache",
					     sizeof(struct btrfs_transaction),
					     0, (SLAB_RECLAIM_ACCOUNT|
						SLAB_MEM_SPREAD),
					     NULL, NULL);
	btrfs_path_cachep = kmem_cache_create("btrfs_path_cache",
					     sizeof(struct btrfs_transaction),
					     0, (SLAB_RECLAIM_ACCOUNT|
						SLAB_MEM_SPREAD),
					     NULL, NULL);
	btrfs_bit_radix_cachep = kmem_cache_create("btrfs_radix",
					     256,
					     0, (SLAB_RECLAIM_ACCOUNT|
						SLAB_MEM_SPREAD |
						SLAB_DESTROY_BY_RCU),
					     NULL, NULL);
	if (btrfs_inode_cachep == NULL || btrfs_trans_handle_cachep == NULL ||
	    btrfs_transaction_cachep == NULL || btrfs_bit_radix_cachep == NULL)
		return -ENOMEM;
	return 0;
}

static void destroy_inodecache(void)
{
	kmem_cache_destroy(btrfs_inode_cachep);
	kmem_cache_destroy(btrfs_trans_handle_cachep);
	kmem_cache_destroy(btrfs_transaction_cachep);
	kmem_cache_destroy(btrfs_bit_radix_cachep);
	kmem_cache_destroy(btrfs_path_cachep);
}

2782 2783 2784 2785 2786 2787 2788
static int btrfs_get_sb(struct file_system_type *fs_type,
	int flags, const char *dev_name, void *data, struct vfsmount *mnt)
{
	return get_sb_bdev(fs_type, flags, dev_name, data,
			   btrfs_fill_super, mnt);
}

2789 2790 2791 2792 2793 2794 2795 2796 2797
static int btrfs_getattr(struct vfsmount *mnt,
			 struct dentry *dentry, struct kstat *stat)
{
	struct inode *inode = dentry->d_inode;
	generic_fillattr(inode, stat);
	stat->blksize = 256 * 1024;
	return 0;
}

C
Chris Mason 已提交
2798 2799 2800 2801 2802 2803 2804 2805 2806 2807 2808 2809 2810
static int btrfs_statfs(struct dentry *dentry, struct kstatfs *buf)
{
	struct btrfs_root *root = btrfs_sb(dentry->d_sb);
	struct btrfs_super_block *disk_super = root->fs_info->disk_super;

	buf->f_namelen = BTRFS_NAME_LEN;
	buf->f_blocks = btrfs_super_total_blocks(disk_super);
	buf->f_bfree = buf->f_blocks - btrfs_super_blocks_used(disk_super);
	buf->f_bavail = buf->f_bfree;
	buf->f_bsize = dentry->d_sb->s_blocksize;
	buf->f_type = BTRFS_SUPER_MAGIC;
	return 0;
}
C
Chris Mason 已提交
2811

C
Chris Mason 已提交
2812 2813 2814 2815 2816 2817 2818 2819 2820 2821 2822 2823 2824 2825 2826 2827 2828 2829 2830 2831 2832 2833 2834 2835 2836 2837 2838 2839 2840 2841 2842 2843 2844 2845 2846 2847 2848 2849 2850 2851 2852 2853 2854 2855 2856 2857 2858 2859 2860 2861 2862 2863 2864 2865 2866 2867 2868 2869 2870 2871 2872 2873 2874 2875 2876 2877 2878 2879 2880 2881
static int btrfs_rename(struct inode * old_dir, struct dentry *old_dentry,
			   struct inode * new_dir,struct dentry *new_dentry)
{
	struct btrfs_trans_handle *trans;
	struct btrfs_root *root = BTRFS_I(old_dir)->root;
	struct inode *new_inode = new_dentry->d_inode;
	struct inode *old_inode = old_dentry->d_inode;
	struct timespec ctime = CURRENT_TIME;
	struct btrfs_path *path;
	struct btrfs_dir_item *di;
	int ret;

	if (S_ISDIR(old_inode->i_mode) && new_inode &&
	    new_inode->i_size > BTRFS_EMPTY_DIR_SIZE) {
		return -ENOTEMPTY;
	}
	mutex_lock(&root->fs_info->fs_mutex);
	trans = btrfs_start_transaction(root, 1);
	btrfs_set_trans_block_group(trans, new_dir);
	path = btrfs_alloc_path();
	if (!path) {
		ret = -ENOMEM;
		goto out_fail;
	}

	old_dentry->d_inode->i_nlink++;
	old_dir->i_ctime = old_dir->i_mtime = ctime;
	new_dir->i_ctime = new_dir->i_mtime = ctime;
	old_inode->i_ctime = ctime;
	if (S_ISDIR(old_inode->i_mode) && old_dir != new_dir) {
		struct btrfs_key *location = &BTRFS_I(new_dir)->location;
		u64 old_parent_oid;
		di = btrfs_lookup_dir_item(trans, root, path, old_inode->i_ino,
					   "..", 2, -1);
		if (IS_ERR(di)) {
			ret = PTR_ERR(di);
			goto out_fail;
		}
		if (!di) {
			ret = -ENOENT;
			goto out_fail;
		}
		old_parent_oid = btrfs_disk_key_objectid(&di->location);
		ret = btrfs_del_item(trans, root, path);
		if (ret) {
			ret = -EIO;
			goto out_fail;
		}
		btrfs_release_path(root, path);

		di = btrfs_lookup_dir_index_item(trans, root, path,
						 old_inode->i_ino,
						 old_parent_oid,
						 "..", 2, -1);
		if (IS_ERR(di)) {
			ret = PTR_ERR(di);
			goto out_fail;
		}
		if (!di) {
			ret = -ENOENT;
			goto out_fail;
		}
		ret = btrfs_del_item(trans, root, path);
		if (ret) {
			ret = -EIO;
			goto out_fail;
		}
		btrfs_release_path(root, path);

		ret = btrfs_insert_dir_item(trans, root, "..", 2,
C
Chris Mason 已提交
2882 2883
					    old_inode->i_ino, location,
					    BTRFS_FT_DIR);
C
Chris Mason 已提交
2884 2885 2886 2887 2888 2889 2890 2891 2892 2893 2894
		if (ret)
			goto out_fail;
	}


	ret = btrfs_unlink_trans(trans, root, old_dir, old_dentry);
	if (ret)
		goto out_fail;

	if (new_inode) {
		new_inode->i_ctime = CURRENT_TIME;
C
Chris Mason 已提交
2895 2896 2897
		ret = btrfs_unlink_trans(trans, root, new_dir, new_dentry);
		if (ret)
			goto out_fail;
C
Chris Mason 已提交
2898 2899 2900 2901 2902 2903
		if (S_ISDIR(new_inode->i_mode))
			clear_nlink(new_inode);
		else
			drop_nlink(new_inode);
		btrfs_update_inode(trans, root, new_inode);
	}
C
Chris Mason 已提交
2904 2905 2906 2907
	ret = btrfs_add_link(trans, new_dentry, old_inode);
	if (ret)
		goto out_fail;

C
Chris Mason 已提交
2908 2909 2910 2911 2912 2913 2914
out_fail:
	btrfs_free_path(path);
	btrfs_end_transaction(trans, root);
	mutex_unlock(&root->fs_info->fs_mutex);
	return ret;
}

C
Chris Mason 已提交
2915 2916 2917 2918 2919 2920 2921 2922 2923 2924 2925 2926 2927 2928 2929 2930 2931 2932 2933 2934 2935 2936 2937 2938 2939 2940 2941 2942 2943 2944 2945 2946 2947 2948 2949 2950 2951 2952 2953 2954 2955 2956 2957 2958 2959 2960 2961 2962 2963 2964 2965 2966 2967 2968 2969 2970 2971 2972 2973 2974 2975 2976 2977 2978 2979 2980 2981 2982 2983 2984 2985 2986 2987 2988 2989 2990 2991 2992 2993 2994 2995 2996 2997 2998 2999 3000 3001 3002
static int btrfs_symlink(struct inode *dir, struct dentry *dentry,
			 const char *symname)
{
	struct btrfs_trans_handle *trans;
	struct btrfs_root *root = BTRFS_I(dir)->root;
	struct btrfs_path *path;
	struct btrfs_key key;
	struct inode *inode;
	int err;
	int drop_inode = 0;
	u64 objectid;
	int name_len;
	int datasize;
	char *ptr;
	struct btrfs_file_extent_item *ei;

	name_len = strlen(symname) + 1;
	if (name_len > BTRFS_MAX_INLINE_DATA_SIZE(root))
		return -ENAMETOOLONG;
	mutex_lock(&root->fs_info->fs_mutex);
	trans = btrfs_start_transaction(root, 1);
	btrfs_set_trans_block_group(trans, dir);

	err = btrfs_find_free_objectid(trans, root, dir->i_ino, &objectid);
	if (err) {
		err = -ENOSPC;
		goto out_unlock;
	}

	inode = btrfs_new_inode(trans, root, objectid,
				BTRFS_I(dir)->block_group, S_IFLNK|S_IRWXUGO);
	err = PTR_ERR(inode);
	if (IS_ERR(inode))
		goto out_unlock;

	btrfs_set_trans_block_group(trans, inode);
	err = btrfs_add_nondir(trans, dentry, inode);
	if (err)
		drop_inode = 1;
	else {
		inode->i_mapping->a_ops = &btrfs_aops;
		inode->i_fop = &btrfs_file_operations;
		inode->i_op = &btrfs_file_inode_operations;
	}
	dir->i_sb->s_dirt = 1;
	btrfs_update_inode_block_group(trans, inode);
	btrfs_update_inode_block_group(trans, dir);
	if (drop_inode)
		goto out_unlock;

	path = btrfs_alloc_path();
	BUG_ON(!path);
	key.objectid = inode->i_ino;
	key.offset = 0;
	key.flags = 0;
	btrfs_set_key_type(&key, BTRFS_EXTENT_DATA_KEY);
	datasize = btrfs_file_extent_calc_inline_size(name_len);
	err = btrfs_insert_empty_item(trans, root, path, &key,
				      datasize);
	BUG_ON(err);
	ei = btrfs_item_ptr(btrfs_buffer_leaf(path->nodes[0]),
	       path->slots[0], struct btrfs_file_extent_item);
	btrfs_set_file_extent_generation(ei, trans->transid);
	btrfs_set_file_extent_type(ei,
				   BTRFS_FILE_EXTENT_INLINE);
	ptr = btrfs_file_extent_inline_start(ei);
	btrfs_memcpy(root, path->nodes[0]->b_data,
		     ptr, symname, name_len);
	mark_buffer_dirty(path->nodes[0]);
	btrfs_free_path(path);
	inode->i_op = &btrfs_symlink_inode_operations;
	inode->i_mapping->a_ops = &btrfs_symlink_aops;
	inode->i_size = name_len - 1;
	btrfs_update_inode(trans, root, inode);
	err = 0;

out_unlock:
	btrfs_end_transaction(trans, root);
	mutex_unlock(&root->fs_info->fs_mutex);

	if (drop_inode) {
		inode_dec_link_count(inode);
		iput(inode);
	}
	btrfs_btree_balance_dirty(root);
	return err;
}

3003 3004 3005 3006 3007 3008 3009 3010
static struct file_system_type btrfs_fs_type = {
	.owner		= THIS_MODULE,
	.name		= "btrfs",
	.get_sb		= btrfs_get_sb,
	.kill_sb	= kill_block_super,
	.fs_flags	= FS_REQUIRES_DEV,
};

C
Chris Mason 已提交
3011
static struct super_operations btrfs_super_ops = {
C
Chris Mason 已提交
3012
	.delete_inode	= btrfs_delete_inode,
C
Chris Mason 已提交
3013 3014
	.put_super	= btrfs_put_super,
	.read_inode	= btrfs_read_locked_inode,
3015 3016
	.write_super	= btrfs_write_super,
	.sync_fs	= btrfs_sync_fs,
C
Chris Mason 已提交
3017
	.write_inode	= btrfs_write_inode,
C
Chris Mason 已提交
3018
	.dirty_inode	= btrfs_dirty_inode,
C
Chris Mason 已提交
3019 3020
	.alloc_inode	= btrfs_alloc_inode,
	.destroy_inode	= btrfs_destroy_inode,
C
Chris Mason 已提交
3021
	.statfs		= btrfs_statfs,
C
Chris Mason 已提交
3022 3023 3024 3025
};

static struct inode_operations btrfs_dir_inode_operations = {
	.lookup		= btrfs_lookup,
3026
	.create		= btrfs_create,
C
Chris Mason 已提交
3027
	.unlink		= btrfs_unlink,
C
Chris Mason 已提交
3028
	.link		= btrfs_link,
C
Chris Mason 已提交
3029
	.mkdir		= btrfs_mkdir,
C
Chris Mason 已提交
3030
	.rmdir		= btrfs_rmdir,
C
Chris Mason 已提交
3031
	.rename		= btrfs_rename,
C
Chris Mason 已提交
3032
	.symlink	= btrfs_symlink,
C
Chris Mason 已提交
3033
	.setattr	= btrfs_setattr,
C
Chris Mason 已提交
3034 3035
};

3036 3037 3038 3039
static struct inode_operations btrfs_dir_ro_inode_operations = {
	.lookup		= btrfs_lookup,
};

C
Chris Mason 已提交
3040 3041 3042 3043
static struct file_operations btrfs_dir_file_operations = {
	.llseek		= generic_file_llseek,
	.read		= generic_read_dir,
	.readdir	= btrfs_readdir,
C
Chris Mason 已提交
3044
	.ioctl		= btrfs_ioctl,
C
Chris Mason 已提交
3045 3046
};

C
Chris Mason 已提交
3047 3048 3049 3050 3051
static struct address_space_operations btrfs_aops = {
	.readpage	= btrfs_readpage,
	.writepage	= btrfs_writepage,
	.sync_page	= block_sync_page,
	.prepare_write	= btrfs_prepare_write,
C
Chris Mason 已提交
3052
	.commit_write	= btrfs_commit_write,
C
Chris Mason 已提交
3053
	.bmap		= btrfs_bmap,
C
Chris Mason 已提交
3054 3055
};

C
Chris Mason 已提交
3056 3057 3058 3059 3060
static struct address_space_operations btrfs_symlink_aops = {
	.readpage	= btrfs_readpage,
	.writepage	= btrfs_writepage,
};

C
Chris Mason 已提交
3061
static struct inode_operations btrfs_file_inode_operations = {
C
Chris Mason 已提交
3062
	.truncate	= btrfs_truncate,
3063
	.getattr	= btrfs_getattr,
C
Chris Mason 已提交
3064
	.setattr	= btrfs_setattr,
C
Chris Mason 已提交
3065 3066 3067 3068 3069
};

static struct file_operations btrfs_file_operations = {
	.llseek		= generic_file_llseek,
	.read		= do_sync_read,
3070 3071
	.aio_read       = btrfs_file_aio_read,
	.write		= btrfs_file_write,
C
Chris Mason 已提交
3072 3073
	.mmap		= generic_file_mmap,
	.open		= generic_file_open,
C
Chris Mason 已提交
3074
	.ioctl		= btrfs_ioctl,
C
Chris Mason 已提交
3075
	.fsync		= btrfs_sync_file,
C
Chris Mason 已提交
3076
};
C
Chris Mason 已提交
3077

C
Chris Mason 已提交
3078 3079 3080 3081 3082 3083
static struct inode_operations btrfs_symlink_inode_operations = {
	.readlink	= generic_readlink,
	.follow_link	= page_follow_link_light,
	.put_link	= page_put_link,
};

3084 3085
static int __init init_btrfs_fs(void)
{
C
Chris Mason 已提交
3086
	int err;
3087
	printk("btrfs loaded!\n");
C
Chris Mason 已提交
3088 3089 3090
	err = init_inodecache();
	if (err)
		return err;
3091
	return register_filesystem(&btrfs_fs_type);
3092 3093
	destroy_inodecache();
	return err;
3094 3095 3096 3097
}

static void __exit exit_btrfs_fs(void)
{
C
Chris Mason 已提交
3098
	destroy_inodecache();
3099 3100 3101 3102 3103 3104 3105 3106
	unregister_filesystem(&btrfs_fs_type);
	printk("btrfs unloaded\n");
}

module_init(init_btrfs_fs)
module_exit(exit_btrfs_fs)

MODULE_LICENSE("GPL");