inode.c 203.5 KB
Newer Older
C
Chris Mason 已提交
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18
/*
 * Copyright (C) 2007 Oracle.  All rights reserved.
 *
 * This program is free software; you can redistribute it and/or
 * modify it under the terms of the GNU General Public
 * License v2 as published by the Free Software Foundation.
 *
 * This program is distributed in the hope that it will be useful,
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
 * General Public License for more details.
 *
 * You should have received a copy of the GNU General Public
 * License along with this program; if not, write to the
 * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
 * Boston, MA 021110-1307, USA.
 */

19
#include <linux/kernel.h>
20
#include <linux/bio.h>
C
Chris Mason 已提交
21
#include <linux/buffer_head.h>
S
Sage Weil 已提交
22
#include <linux/file.h>
C
Chris Mason 已提交
23 24 25 26 27 28 29 30 31 32 33 34
#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/backing-dev.h>
#include <linux/mpage.h>
#include <linux/swap.h>
#include <linux/writeback.h>
#include <linux/statfs.h>
#include <linux/compat.h>
C
Chris Mason 已提交
35
#include <linux/bit_spinlock.h>
J
Josef Bacik 已提交
36
#include <linux/xattr.h>
J
Josef Bacik 已提交
37
#include <linux/posix_acl.h>
Y
Yan Zheng 已提交
38
#include <linux/falloc.h>
39
#include <linux/slab.h>
40
#include <linux/ratelimit.h>
41
#include <linux/mount.h>
C
Chris Mason 已提交
42
#include "compat.h"
C
Chris Mason 已提交
43 44 45 46 47 48
#include "ctree.h"
#include "disk-io.h"
#include "transaction.h"
#include "btrfs_inode.h"
#include "ioctl.h"
#include "print-tree.h"
49
#include "ordered-data.h"
50
#include "xattr.h"
51
#include "tree-log.h"
52
#include "volumes.h"
C
Chris Mason 已提交
53
#include "compression.h"
54
#include "locking.h"
55
#include "free-space-cache.h"
56
#include "inode-map.h"
C
Chris Mason 已提交
57 58 59 60 61 62

struct btrfs_iget_args {
	u64 ino;
	struct btrfs_root *root;
};

63 64 65 66 67
static const struct inode_operations btrfs_dir_inode_operations;
static const struct inode_operations btrfs_symlink_inode_operations;
static const struct inode_operations btrfs_dir_ro_inode_operations;
static const struct inode_operations btrfs_special_inode_operations;
static const struct inode_operations btrfs_file_inode_operations;
68 69
static const struct address_space_operations btrfs_aops;
static const struct address_space_operations btrfs_symlink_aops;
70
static const struct file_operations btrfs_dir_file_operations;
71
static struct extent_io_ops btrfs_extent_io_ops;
C
Chris Mason 已提交
72 73 74 75 76

static struct kmem_cache *btrfs_inode_cachep;
struct kmem_cache *btrfs_trans_handle_cachep;
struct kmem_cache *btrfs_transaction_cachep;
struct kmem_cache *btrfs_path_cachep;
77
struct kmem_cache *btrfs_free_space_cachep;
C
Chris Mason 已提交
78 79 80 81 82 83 84 85 86 87 88 89

#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,
};

90 91
static int btrfs_setsize(struct inode *inode, loff_t newsize);
static int btrfs_truncate(struct inode *inode);
92
static int btrfs_finish_ordered_io(struct btrfs_ordered_extent *ordered_extent);
93 94 95 96
static noinline int cow_file_range(struct inode *inode,
				   struct page *locked_page,
				   u64 start, u64 end, int *page_started,
				   unsigned long *nr_written, int unlock);
97 98
static noinline int btrfs_update_inode_fallback(struct btrfs_trans_handle *trans,
				struct btrfs_root *root, struct inode *inode);
99

100
static int btrfs_init_inode_security(struct btrfs_trans_handle *trans,
101 102
				     struct inode *inode,  struct inode *dir,
				     const struct qstr *qstr)
J
Jim Owens 已提交
103 104 105
{
	int err;

106
	err = btrfs_init_acl(trans, inode, dir);
J
Jim Owens 已提交
107
	if (!err)
108
		err = btrfs_xattr_security_init(trans, inode, dir, qstr);
J
Jim Owens 已提交
109 110 111
	return err;
}

C
Chris Mason 已提交
112 113 114 115 116
/*
 * this does all the hard work for inserting an inline extent into
 * the btree.  The caller should have done a btrfs_drop_extents so that
 * no overlapping inline items exist in the btree
 */
C
Chris Mason 已提交
117
static noinline int insert_inline_extent(struct btrfs_trans_handle *trans,
C
Chris Mason 已提交
118 119
				struct btrfs_root *root, struct inode *inode,
				u64 start, size_t size, size_t compressed_size,
120
				int compress_type,
C
Chris Mason 已提交
121 122 123 124 125 126 127 128 129 130 131 132 133 134 135
				struct page **compressed_pages)
{
	struct btrfs_key key;
	struct btrfs_path *path;
	struct extent_buffer *leaf;
	struct page *page = NULL;
	char *kaddr;
	unsigned long ptr;
	struct btrfs_file_extent_item *ei;
	int err = 0;
	int ret;
	size_t cur_size = size;
	size_t datasize;
	unsigned long offset;

136
	if (compressed_size && compressed_pages)
C
Chris Mason 已提交
137 138
		cur_size = compressed_size;

C
Chris Mason 已提交
139 140
	path = btrfs_alloc_path();
	if (!path)
C
Chris Mason 已提交
141 142
		return -ENOMEM;

143
	path->leave_spinning = 1;
C
Chris Mason 已提交
144

L
Li Zefan 已提交
145
	key.objectid = btrfs_ino(inode);
C
Chris Mason 已提交
146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166
	key.offset = start;
	btrfs_set_key_type(&key, BTRFS_EXTENT_DATA_KEY);
	datasize = btrfs_file_extent_calc_inline_size(cur_size);

	inode_add_bytes(inode, size);
	ret = btrfs_insert_empty_item(trans, root, path, &key,
				      datasize);
	if (ret) {
		err = ret;
		goto fail;
	}
	leaf = path->nodes[0];
	ei = btrfs_item_ptr(leaf, path->slots[0],
			    struct btrfs_file_extent_item);
	btrfs_set_file_extent_generation(leaf, ei, trans->transid);
	btrfs_set_file_extent_type(leaf, ei, BTRFS_FILE_EXTENT_INLINE);
	btrfs_set_file_extent_encryption(leaf, ei, 0);
	btrfs_set_file_extent_other_encoding(leaf, ei, 0);
	btrfs_set_file_extent_ram_bytes(leaf, ei, size);
	ptr = btrfs_file_extent_inline_start(ei);

167
	if (compress_type != BTRFS_COMPRESS_NONE) {
C
Chris Mason 已提交
168 169
		struct page *cpage;
		int i = 0;
C
Chris Mason 已提交
170
		while (compressed_size > 0) {
C
Chris Mason 已提交
171
			cpage = compressed_pages[i];
172
			cur_size = min_t(unsigned long, compressed_size,
C
Chris Mason 已提交
173 174
				       PAGE_CACHE_SIZE);

175
			kaddr = kmap_atomic(cpage);
C
Chris Mason 已提交
176
			write_extent_buffer(leaf, kaddr, ptr, cur_size);
177
			kunmap_atomic(kaddr);
C
Chris Mason 已提交
178 179 180 181 182 183

			i++;
			ptr += cur_size;
			compressed_size -= cur_size;
		}
		btrfs_set_file_extent_compression(leaf, ei,
184
						  compress_type);
C
Chris Mason 已提交
185 186 187 188
	} else {
		page = find_get_page(inode->i_mapping,
				     start >> PAGE_CACHE_SHIFT);
		btrfs_set_file_extent_compression(leaf, ei, 0);
189
		kaddr = kmap_atomic(page);
C
Chris Mason 已提交
190 191
		offset = start & (PAGE_CACHE_SIZE - 1);
		write_extent_buffer(leaf, kaddr + offset, ptr, size);
192
		kunmap_atomic(kaddr);
C
Chris Mason 已提交
193 194 195 196 197
		page_cache_release(page);
	}
	btrfs_mark_buffer_dirty(leaf);
	btrfs_free_path(path);

198 199 200 201 202 203 204 205 206
	/*
	 * we're an inline extent, so nobody can
	 * extend the file past i_size without locking
	 * a page we already have locked.
	 *
	 * We must do any isize and inode updates
	 * before we unlock the pages.  Otherwise we
	 * could end up racing with unlink.
	 */
C
Chris Mason 已提交
207
	BTRFS_I(inode)->disk_i_size = inode->i_size;
208
	ret = btrfs_update_inode(trans, root, inode);
209

210
	return ret;
C
Chris Mason 已提交
211 212 213 214 215 216 217 218 219 220 221
fail:
	btrfs_free_path(path);
	return err;
}


/*
 * conditionally insert an inline extent into the file.  This
 * does the checks required to make sure the data is small enough
 * to fit as an inline extent.
 */
222
static noinline int cow_file_range_inline(struct btrfs_trans_handle *trans,
C
Chris Mason 已提交
223 224
				 struct btrfs_root *root,
				 struct inode *inode, u64 start, u64 end,
225
				 size_t compressed_size, int compress_type,
C
Chris Mason 已提交
226 227 228 229 230 231 232 233 234 235 236 237 238 239 240
				 struct page **compressed_pages)
{
	u64 isize = i_size_read(inode);
	u64 actual_end = min(end + 1, isize);
	u64 inline_len = actual_end - start;
	u64 aligned_end = (end + root->sectorsize - 1) &
			~((u64)root->sectorsize - 1);
	u64 hint_byte;
	u64 data_len = inline_len;
	int ret;

	if (compressed_size)
		data_len = compressed_size;

	if (start > 0 ||
C
Chris Mason 已提交
241
	    actual_end >= PAGE_CACHE_SIZE ||
C
Chris Mason 已提交
242 243 244 245 246 247 248 249
	    data_len >= BTRFS_MAX_INLINE_DATA_SIZE(root) ||
	    (!compressed_size &&
	    (actual_end & (root->sectorsize - 1)) == 0) ||
	    end + 1 < isize ||
	    data_len > root->fs_info->max_inline) {
		return 1;
	}

Y
Yan, Zheng 已提交
250
	ret = btrfs_drop_extents(trans, inode, start, aligned_end,
C
Chris Mason 已提交
251
				 &hint_byte, 1);
252 253
	if (ret)
		return ret;
C
Chris Mason 已提交
254 255 256 257 258

	if (isize > actual_end)
		inline_len = min_t(u64, isize, actual_end);
	ret = insert_inline_extent(trans, root, inode, start,
				   inline_len, compressed_size,
259
				   compress_type, compressed_pages);
260
	if (ret && ret != -ENOSPC) {
261 262
		btrfs_abort_transaction(trans, root, ret);
		return ret;
263 264
	} else if (ret == -ENOSPC) {
		return 1;
265
	}
266

267
	btrfs_delalloc_release_metadata(inode, end + 1 - start);
C
Chris Mason 已提交
268
	btrfs_drop_extent_cache(inode, start, aligned_end - 1, 0);
C
Chris Mason 已提交
269 270 271
	return 0;
}

272 273 274 275 276 277
struct async_extent {
	u64 start;
	u64 ram_size;
	u64 compressed_size;
	struct page **pages;
	unsigned long nr_pages;
278
	int compress_type;
279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295
	struct list_head list;
};

struct async_cow {
	struct inode *inode;
	struct btrfs_root *root;
	struct page *locked_page;
	u64 start;
	u64 end;
	struct list_head extents;
	struct btrfs_work work;
};

static noinline int add_async_extent(struct async_cow *cow,
				     u64 start, u64 ram_size,
				     u64 compressed_size,
				     struct page **pages,
296 297
				     unsigned long nr_pages,
				     int compress_type)
298 299 300 301
{
	struct async_extent *async_extent;

	async_extent = kmalloc(sizeof(*async_extent), GFP_NOFS);
302
	BUG_ON(!async_extent); /* -ENOMEM */
303 304 305 306 307
	async_extent->start = start;
	async_extent->ram_size = ram_size;
	async_extent->compressed_size = compressed_size;
	async_extent->pages = pages;
	async_extent->nr_pages = nr_pages;
308
	async_extent->compress_type = compress_type;
309 310 311 312
	list_add_tail(&async_extent->list, &cow->extents);
	return 0;
}

C
Chris Mason 已提交
313
/*
314 315 316
 * we create compressed extents in two phases.  The first
 * phase compresses a range of pages that have already been
 * locked (both pages and state bits are locked).
C
Chris Mason 已提交
317
 *
318 319 320 321 322
 * This is done inside an ordered work queue, and the compression
 * is spread across many cpus.  The actual IO submission is step
 * two, and the ordered work queue takes care of making sure that
 * happens in the same order things were put onto the queue by
 * writepages and friends.
C
Chris Mason 已提交
323
 *
324 325 326 327
 * If this code finds it can't get good compression, it puts an
 * entry onto the work queue to write the uncompressed bytes.  This
 * makes sure that both compressed inodes and uncompressed inodes
 * are written in the same order that pdflush sent them down.
C
Chris Mason 已提交
328
 */
329 330 331 332 333
static noinline int compress_file_range(struct inode *inode,
					struct page *locked_page,
					u64 start, u64 end,
					struct async_cow *async_cow,
					int *num_added)
334 335 336
{
	struct btrfs_root *root = BTRFS_I(inode)->root;
	struct btrfs_trans_handle *trans;
337 338
	u64 num_bytes;
	u64 blocksize = root->sectorsize;
C
Chris Mason 已提交
339
	u64 actual_end;
340
	u64 isize = i_size_read(inode);
341
	int ret = 0;
C
Chris Mason 已提交
342 343 344 345 346 347
	struct page **pages = NULL;
	unsigned long nr_pages;
	unsigned long nr_pages_ret = 0;
	unsigned long total_compressed = 0;
	unsigned long total_in = 0;
	unsigned long max_compressed = 128 * 1024;
348
	unsigned long max_uncompressed = 128 * 1024;
C
Chris Mason 已提交
349 350
	int i;
	int will_compress;
351
	int compress_type = root->fs_info->compress_type;
352

353 354 355
	/* if this is a small write inside eof, kick off a defrag */
	if ((end - start + 1) < 16 * 1024 &&
	    (start > 0 || end + 1 < BTRFS_I(inode)->disk_i_size))
C
Chris Mason 已提交
356 357
		btrfs_add_inode_defrag(NULL, inode);

358
	actual_end = min_t(u64, isize, end + 1);
C
Chris Mason 已提交
359 360 361 362
again:
	will_compress = 0;
	nr_pages = (end >> PAGE_CACHE_SHIFT) - (start >> PAGE_CACHE_SHIFT) + 1;
	nr_pages = min(nr_pages, (128 * 1024UL) / PAGE_CACHE_SIZE);
363

364 365 366 367 368 369 370 371 372 373 374 375 376
	/*
	 * we don't want to send crud past the end of i_size through
	 * compression, that's just a waste of CPU time.  So, if the
	 * end of the file is before the start of our current
	 * requested range of bytes, we bail out to the uncompressed
	 * cleanup code that can deal with all of this.
	 *
	 * It isn't really the fastest way to fix things, but this is a
	 * very uncommon corner.
	 */
	if (actual_end <= start)
		goto cleanup_and_bail_uncompressed;

C
Chris Mason 已提交
377 378 379 380
	total_compressed = actual_end - start;

	/* we want to make sure that amount of ram required to uncompress
	 * an extent is reasonable, so we limit the total size in ram
381 382 383 384 385 386 387
	 * of a compressed extent to 128k.  This is a crucial number
	 * because it also controls how easily we can spread reads across
	 * cpus for decompression.
	 *
	 * We also want to make sure the amount of IO required to do
	 * a random read is reasonably small, so we limit the size of
	 * a compressed extent to 128k.
C
Chris Mason 已提交
388 389
	 */
	total_compressed = min(total_compressed, max_uncompressed);
390
	num_bytes = (end - start + blocksize) & ~(blocksize - 1);
391
	num_bytes = max(blocksize,  num_bytes);
C
Chris Mason 已提交
392 393
	total_in = 0;
	ret = 0;
394

395 396 397 398
	/*
	 * we do compression for mount -o compress and when the
	 * inode has not been flagged as nocompress.  This flag can
	 * change at any time if we discover bad compression ratios.
C
Chris Mason 已提交
399
	 */
400
	if (!(BTRFS_I(inode)->flags & BTRFS_INODE_NOCOMPRESS) &&
C
Chris Mason 已提交
401
	    (btrfs_test_opt(root, COMPRESS) ||
402 403
	     (BTRFS_I(inode)->force_compress) ||
	     (BTRFS_I(inode)->flags & BTRFS_INODE_COMPRESS))) {
C
Chris Mason 已提交
404
		WARN_ON(pages);
405
		pages = kzalloc(sizeof(struct page *) * nr_pages, GFP_NOFS);
406 407 408 409
		if (!pages) {
			/* just bail out to the uncompressed code */
			goto cont;
		}
C
Chris Mason 已提交
410

411 412 413 414 415 416 417 418 419 420
		if (BTRFS_I(inode)->force_compress)
			compress_type = BTRFS_I(inode)->force_compress;

		ret = btrfs_compress_pages(compress_type,
					   inode->i_mapping, start,
					   total_compressed, pages,
					   nr_pages, &nr_pages_ret,
					   &total_in,
					   &total_compressed,
					   max_compressed);
C
Chris Mason 已提交
421 422 423 424 425 426 427 428 429 430 431

		if (!ret) {
			unsigned long offset = total_compressed &
				(PAGE_CACHE_SIZE - 1);
			struct page *page = pages[nr_pages_ret - 1];
			char *kaddr;

			/* zero the tail end of the last page, we might be
			 * sending it down to disk
			 */
			if (offset) {
432
				kaddr = kmap_atomic(page);
C
Chris Mason 已提交
433 434
				memset(kaddr + offset, 0,
				       PAGE_CACHE_SIZE - offset);
435
				kunmap_atomic(kaddr);
C
Chris Mason 已提交
436 437 438 439
			}
			will_compress = 1;
		}
	}
440
cont:
C
Chris Mason 已提交
441
	if (start == 0) {
442
		trans = btrfs_join_transaction(root);
443 444 445 446 447
		if (IS_ERR(trans)) {
			ret = PTR_ERR(trans);
			trans = NULL;
			goto cleanup_and_out;
		}
448
		trans->block_rsv = &root->fs_info->delalloc_block_rsv;
449

C
Chris Mason 已提交
450
		/* lets try to make an inline extent */
451
		if (ret || total_in < (actual_end - start)) {
C
Chris Mason 已提交
452
			/* we didn't compress the entire range, try
453
			 * to make an uncompressed inline extent.
C
Chris Mason 已提交
454 455
			 */
			ret = cow_file_range_inline(trans, root, inode,
456
						    start, end, 0, 0, NULL);
C
Chris Mason 已提交
457
		} else {
458
			/* try making a compressed inline extent */
C
Chris Mason 已提交
459 460
			ret = cow_file_range_inline(trans, root, inode,
						    start, end,
461 462
						    total_compressed,
						    compress_type, pages);
C
Chris Mason 已提交
463
		}
464
		if (ret <= 0) {
465
			/*
466 467 468
			 * inline extent creation worked or returned error,
			 * we don't need to create any more async work items.
			 * Unlock and free up our temp pages.
469
			 */
C
Chris Mason 已提交
470
			extent_clear_unlock_delalloc(inode,
471 472 473
			     &BTRFS_I(inode)->io_tree,
			     start, end, NULL,
			     EXTENT_CLEAR_UNLOCK_PAGE | EXTENT_CLEAR_DIRTY |
474
			     EXTENT_CLEAR_DELALLOC |
475
			     EXTENT_SET_WRITEBACK | EXTENT_END_WRITEBACK);
476 477

			btrfs_end_transaction(trans, root);
C
Chris Mason 已提交
478 479
			goto free_pages_out;
		}
480
		btrfs_end_transaction(trans, root);
C
Chris Mason 已提交
481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509
	}

	if (will_compress) {
		/*
		 * we aren't doing an inline extent round the compressed size
		 * up to a block size boundary so the allocator does sane
		 * things
		 */
		total_compressed = (total_compressed + blocksize - 1) &
			~(blocksize - 1);

		/*
		 * one last check to make sure the compression is really a
		 * win, compare the page count read with the blocks on disk
		 */
		total_in = (total_in + PAGE_CACHE_SIZE - 1) &
			~(PAGE_CACHE_SIZE - 1);
		if (total_compressed >= total_in) {
			will_compress = 0;
		} else {
			num_bytes = total_in;
		}
	}
	if (!will_compress && pages) {
		/*
		 * the compression code ran but failed to make things smaller,
		 * free any pages it allocated and our page pointer array
		 */
		for (i = 0; i < nr_pages_ret; i++) {
C
Chris Mason 已提交
510
			WARN_ON(pages[i]->mapping);
C
Chris Mason 已提交
511 512 513 514 515 516 517 518
			page_cache_release(pages[i]);
		}
		kfree(pages);
		pages = NULL;
		total_compressed = 0;
		nr_pages_ret = 0;

		/* flag the file so we don't compress in the future */
C
Chris Mason 已提交
519 520
		if (!btrfs_test_opt(root, FORCE_COMPRESS) &&
		    !(BTRFS_I(inode)->force_compress)) {
C
Chris Mason 已提交
521
			BTRFS_I(inode)->flags |= BTRFS_INODE_NOCOMPRESS;
C
Chris Mason 已提交
522
		}
C
Chris Mason 已提交
523
	}
524 525
	if (will_compress) {
		*num_added += 1;
C
Chris Mason 已提交
526

527 528 529 530 531
		/* the async work queues will take care of doing actual
		 * allocation on disk for these compressed pages,
		 * and will submit them to the elevator.
		 */
		add_async_extent(async_cow, start, num_bytes,
532 533
				 total_compressed, pages, nr_pages_ret,
				 compress_type);
534

535
		if (start + num_bytes < end) {
536 537 538 539 540 541
			start += num_bytes;
			pages = NULL;
			cond_resched();
			goto again;
		}
	} else {
542
cleanup_and_bail_uncompressed:
543 544 545 546 547 548 549 550 551 552 553 554
		/*
		 * No compression, but we still need to write the pages in
		 * the file we've been given so far.  redirty the locked
		 * page if it corresponds to our extent and set things up
		 * for the async work queue to run cow_file_range to do
		 * the normal delalloc dance
		 */
		if (page_offset(locked_page) >= start &&
		    page_offset(locked_page) <= end) {
			__set_page_dirty_nobuffers(locked_page);
			/* unlocked later on in the async handlers */
		}
555 556
		add_async_extent(async_cow, start, end - start + 1,
				 0, NULL, 0, BTRFS_COMPRESS_NONE);
557 558
		*num_added += 1;
	}
559

560
out:
561
	return ret;
562 563 564 565 566 567

free_pages_out:
	for (i = 0; i < nr_pages_ret; i++) {
		WARN_ON(pages[i]->mapping);
		page_cache_release(pages[i]);
	}
C
Chris Mason 已提交
568
	kfree(pages);
569 570

	goto out;
571 572 573 574 575 576 577 578 579 580 581 582 583 584

cleanup_and_out:
	extent_clear_unlock_delalloc(inode, &BTRFS_I(inode)->io_tree,
				     start, end, NULL,
				     EXTENT_CLEAR_UNLOCK_PAGE |
				     EXTENT_CLEAR_DIRTY |
				     EXTENT_CLEAR_DELALLOC |
				     EXTENT_SET_WRITEBACK |
				     EXTENT_END_WRITEBACK);
	if (!trans || IS_ERR(trans))
		btrfs_error(root->fs_info, ret, "Failed to join transaction");
	else
		btrfs_abort_transaction(trans, root, ret);
	goto free_pages_out;
585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603
}

/*
 * phase two of compressed writeback.  This is the ordered portion
 * of the code, which only gets called in the order the work was
 * queued.  We walk all the async extents created by compress_file_range
 * and send them down to the disk.
 */
static noinline int submit_compressed_extents(struct inode *inode,
					      struct async_cow *async_cow)
{
	struct async_extent *async_extent;
	u64 alloc_hint = 0;
	struct btrfs_trans_handle *trans;
	struct btrfs_key ins;
	struct extent_map *em;
	struct btrfs_root *root = BTRFS_I(inode)->root;
	struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
	struct extent_io_tree *io_tree;
604
	int ret = 0;
605 606 607 608 609

	if (list_empty(&async_cow->extents))
		return 0;


C
Chris Mason 已提交
610
	while (!list_empty(&async_cow->extents)) {
611 612 613
		async_extent = list_entry(async_cow->extents.next,
					  struct async_extent, list);
		list_del(&async_extent->list);
C
Chris Mason 已提交
614

615 616
		io_tree = &BTRFS_I(inode)->io_tree;

617
retry:
618 619 620 621 622 623
		/* did the compression code fall back to uncompressed IO? */
		if (!async_extent->pages) {
			int page_started = 0;
			unsigned long nr_written = 0;

			lock_extent(io_tree, async_extent->start,
624
					 async_extent->start +
625
					 async_extent->ram_size - 1);
626 627

			/* allocate blocks */
628 629 630 631 632
			ret = cow_file_range(inode, async_cow->locked_page,
					     async_extent->start,
					     async_extent->start +
					     async_extent->ram_size - 1,
					     &page_started, &nr_written, 0);
633

634 635
			/* JDM XXX */

636 637 638 639 640 641
			/*
			 * if page_started, cow_file_range inserted an
			 * inline extent and took care of all the unlocking
			 * and IO for us.  Otherwise, we need to submit
			 * all those pages down to the drive.
			 */
642
			if (!page_started && !ret)
643 644
				extent_write_locked_range(io_tree,
						  inode, async_extent->start,
C
Chris Mason 已提交
645
						  async_extent->start +
646 647 648 649 650 651 652 653 654
						  async_extent->ram_size - 1,
						  btrfs_get_extent,
						  WB_SYNC_ALL);
			kfree(async_extent);
			cond_resched();
			continue;
		}

		lock_extent(io_tree, async_extent->start,
655
			    async_extent->start + async_extent->ram_size - 1);
656

657
		trans = btrfs_join_transaction(root);
658 659 660 661 662
		if (IS_ERR(trans)) {
			ret = PTR_ERR(trans);
		} else {
			trans->block_rsv = &root->fs_info->delalloc_block_rsv;
			ret = btrfs_reserve_extent(trans, root,
663 664
					   async_extent->compressed_size,
					   async_extent->compressed_size,
665
					   0, alloc_hint, &ins, 1);
666 667 668 669
			if (ret)
				btrfs_abort_transaction(trans, root, ret);
			btrfs_end_transaction(trans, root);
		}
670

671 672 673 674 675 676 677 678 679 680 681
		if (ret) {
			int i;
			for (i = 0; i < async_extent->nr_pages; i++) {
				WARN_ON(async_extent->pages[i]->mapping);
				page_cache_release(async_extent->pages[i]);
			}
			kfree(async_extent->pages);
			async_extent->nr_pages = 0;
			async_extent->pages = NULL;
			unlock_extent(io_tree, async_extent->start,
				      async_extent->start +
682
				      async_extent->ram_size - 1);
683 684 685
			if (ret == -ENOSPC)
				goto retry;
			goto out_free; /* JDM: Requeue? */
686 687
		}

688 689 690 691 692 693 694 695
		/*
		 * here we're doing allocation and writeback of the
		 * compressed pages
		 */
		btrfs_drop_extent_cache(inode, async_extent->start,
					async_extent->start +
					async_extent->ram_size - 1, 0);

696
		em = alloc_extent_map();
697
		BUG_ON(!em); /* -ENOMEM */
698 699
		em->start = async_extent->start;
		em->len = async_extent->ram_size;
700
		em->orig_start = em->start;
C
Chris Mason 已提交
701

702 703 704
		em->block_start = ins.objectid;
		em->block_len = ins.offset;
		em->bdev = root->fs_info->fs_devices->latest_bdev;
705
		em->compress_type = async_extent->compress_type;
706 707 708
		set_bit(EXTENT_FLAG_PINNED, &em->flags);
		set_bit(EXTENT_FLAG_COMPRESSED, &em->flags);

C
Chris Mason 已提交
709
		while (1) {
710
			write_lock(&em_tree->lock);
711
			ret = add_extent_mapping(em_tree, em);
712
			write_unlock(&em_tree->lock);
713 714 715 716 717 718 719 720 721
			if (ret != -EEXIST) {
				free_extent_map(em);
				break;
			}
			btrfs_drop_extent_cache(inode, async_extent->start,
						async_extent->start +
						async_extent->ram_size - 1, 0);
		}

722 723 724 725 726 727 728
		ret = btrfs_add_ordered_extent_compress(inode,
						async_extent->start,
						ins.objectid,
						async_extent->ram_size,
						ins.offset,
						BTRFS_ORDERED_COMPRESSED,
						async_extent->compress_type);
729
		BUG_ON(ret); /* -ENOMEM */
730 731 732 733 734

		/*
		 * clear dirty, set writeback and unlock the pages.
		 */
		extent_clear_unlock_delalloc(inode,
735 736 737 738 739 740
				&BTRFS_I(inode)->io_tree,
				async_extent->start,
				async_extent->start +
				async_extent->ram_size - 1,
				NULL, EXTENT_CLEAR_UNLOCK_PAGE |
				EXTENT_CLEAR_UNLOCK |
741
				EXTENT_CLEAR_DELALLOC |
742
				EXTENT_CLEAR_DIRTY | EXTENT_SET_WRITEBACK);
743 744

		ret = btrfs_submit_compressed_write(inode,
C
Chris Mason 已提交
745 746 747 748 749
				    async_extent->start,
				    async_extent->ram_size,
				    ins.objectid,
				    ins.offset, async_extent->pages,
				    async_extent->nr_pages);
750

751
		BUG_ON(ret); /* -ENOMEM */
752 753 754 755
		alloc_hint = ins.objectid + ins.offset;
		kfree(async_extent);
		cond_resched();
	}
756 757 758 759 760 761
	ret = 0;
out:
	return ret;
out_free:
	kfree(async_extent);
	goto out;
762 763
}

764 765 766 767 768 769 770 771 772 773 774 775 776 777 778 779 780 781 782 783 784 785 786 787 788 789 790 791 792 793 794 795
static u64 get_extent_allocation_hint(struct inode *inode, u64 start,
				      u64 num_bytes)
{
	struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
	struct extent_map *em;
	u64 alloc_hint = 0;

	read_lock(&em_tree->lock);
	em = search_extent_mapping(em_tree, start, num_bytes);
	if (em) {
		/*
		 * if block start isn't an actual block number then find the
		 * first block in this inode and use that as a hint.  If that
		 * block is also bogus then just don't worry about it.
		 */
		if (em->block_start >= EXTENT_MAP_LAST_BYTE) {
			free_extent_map(em);
			em = search_extent_mapping(em_tree, 0, 0);
			if (em && em->block_start < EXTENT_MAP_LAST_BYTE)
				alloc_hint = em->block_start;
			if (em)
				free_extent_map(em);
		} else {
			alloc_hint = em->block_start;
			free_extent_map(em);
		}
	}
	read_unlock(&em_tree->lock);

	return alloc_hint;
}

796 797 798 799 800 801 802 803 804 805 806 807 808 809 810 811 812 813 814 815 816 817 818 819 820 821 822 823 824 825 826 827
/*
 * when extent_io.c finds a delayed allocation range in the file,
 * the call backs end up in this code.  The basic idea is to
 * allocate extents on disk for the range, and create ordered data structs
 * in ram to track those extents.
 *
 * locked_page is the page that writepage had locked already.  We use
 * it to make sure we don't do extra locks or unlocks.
 *
 * *page_started is set to one if we unlock locked_page and do everything
 * required to start IO on it.  It may be clean and already done with
 * IO when we return.
 */
static noinline int cow_file_range(struct inode *inode,
				   struct page *locked_page,
				   u64 start, u64 end, int *page_started,
				   unsigned long *nr_written,
				   int unlock)
{
	struct btrfs_root *root = BTRFS_I(inode)->root;
	struct btrfs_trans_handle *trans;
	u64 alloc_hint = 0;
	u64 num_bytes;
	unsigned long ram_size;
	u64 disk_num_bytes;
	u64 cur_alloc_size;
	u64 blocksize = root->sectorsize;
	struct btrfs_key ins;
	struct extent_map *em;
	struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
	int ret = 0;

828
	BUG_ON(btrfs_is_free_space_inode(root, inode));
829
	trans = btrfs_join_transaction(root);
830 831 832
	if (IS_ERR(trans)) {
		extent_clear_unlock_delalloc(inode,
			     &BTRFS_I(inode)->io_tree,
833
			     start, end, locked_page,
834 835 836 837 838 839 840 841
			     EXTENT_CLEAR_UNLOCK_PAGE |
			     EXTENT_CLEAR_UNLOCK |
			     EXTENT_CLEAR_DELALLOC |
			     EXTENT_CLEAR_DIRTY |
			     EXTENT_SET_WRITEBACK |
			     EXTENT_END_WRITEBACK);
		return PTR_ERR(trans);
	}
842
	trans->block_rsv = &root->fs_info->delalloc_block_rsv;
843 844 845 846 847 848

	num_bytes = (end - start + blocksize) & ~(blocksize - 1);
	num_bytes = max(blocksize,  num_bytes);
	disk_num_bytes = num_bytes;
	ret = 0;

C
Chris Mason 已提交
849
	/* if this is a small write inside eof, kick off defrag */
850 851
	if (num_bytes < 64 * 1024 &&
	    (start > 0 || end + 1 < BTRFS_I(inode)->disk_i_size))
C
Chris Mason 已提交
852 853
		btrfs_add_inode_defrag(trans, inode);

854 855 856
	if (start == 0) {
		/* lets try to make an inline extent */
		ret = cow_file_range_inline(trans, root, inode,
857
					    start, end, 0, 0, NULL);
858 859
		if (ret == 0) {
			extent_clear_unlock_delalloc(inode,
860 861 862 863 864 865 866 867
				     &BTRFS_I(inode)->io_tree,
				     start, end, NULL,
				     EXTENT_CLEAR_UNLOCK_PAGE |
				     EXTENT_CLEAR_UNLOCK |
				     EXTENT_CLEAR_DELALLOC |
				     EXTENT_CLEAR_DIRTY |
				     EXTENT_SET_WRITEBACK |
				     EXTENT_END_WRITEBACK);
868

869 870 871 872
			*nr_written = *nr_written +
			     (end - start + PAGE_CACHE_SIZE) / PAGE_CACHE_SIZE;
			*page_started = 1;
			goto out;
873 874 875
		} else if (ret < 0) {
			btrfs_abort_transaction(trans, root, ret);
			goto out_unlock;
876 877 878 879
		}
	}

	BUG_ON(disk_num_bytes >
880
	       btrfs_super_total_bytes(root->fs_info->super_copy));
881

882
	alloc_hint = get_extent_allocation_hint(inode, start, num_bytes);
883 884
	btrfs_drop_extent_cache(inode, start, start + num_bytes - 1, 0);

C
Chris Mason 已提交
885
	while (disk_num_bytes > 0) {
886 887
		unsigned long op;

888
		cur_alloc_size = disk_num_bytes;
889
		ret = btrfs_reserve_extent(trans, root, cur_alloc_size,
890
					   root->sectorsize, 0, alloc_hint,
891
					   &ins, 1);
892 893 894 895
		if (ret < 0) {
			btrfs_abort_transaction(trans, root, ret);
			goto out_unlock;
		}
C
Chris Mason 已提交
896

897
		em = alloc_extent_map();
898
		BUG_ON(!em); /* -ENOMEM */
899
		em->start = start;
900
		em->orig_start = em->start;
901 902
		ram_size = ins.offset;
		em->len = ins.offset;
C
Chris Mason 已提交
903

904
		em->block_start = ins.objectid;
C
Chris Mason 已提交
905
		em->block_len = ins.offset;
906
		em->bdev = root->fs_info->fs_devices->latest_bdev;
907
		set_bit(EXTENT_FLAG_PINNED, &em->flags);
C
Chris Mason 已提交
908

C
Chris Mason 已提交
909
		while (1) {
910
			write_lock(&em_tree->lock);
911
			ret = add_extent_mapping(em_tree, em);
912
			write_unlock(&em_tree->lock);
913 914 915 916 917
			if (ret != -EEXIST) {
				free_extent_map(em);
				break;
			}
			btrfs_drop_extent_cache(inode, start,
C
Chris Mason 已提交
918
						start + ram_size - 1, 0);
919 920
		}

921
		cur_alloc_size = ins.offset;
922
		ret = btrfs_add_ordered_extent(inode, start, ins.objectid,
923
					       ram_size, cur_alloc_size, 0);
924
		BUG_ON(ret); /* -ENOMEM */
C
Chris Mason 已提交
925

926 927 928 929
		if (root->root_key.objectid ==
		    BTRFS_DATA_RELOC_TREE_OBJECTID) {
			ret = btrfs_reloc_clone_csums(inode, start,
						      cur_alloc_size);
930 931 932 933
			if (ret) {
				btrfs_abort_transaction(trans, root, ret);
				goto out_unlock;
			}
934 935
		}

C
Chris Mason 已提交
936
		if (disk_num_bytes < cur_alloc_size)
937
			break;
C
Chris Mason 已提交
938

C
Chris Mason 已提交
939 940 941
		/* we're not doing compressed IO, don't unlock the first
		 * page (which the caller expects to stay locked), don't
		 * clear any dirty bits and don't set any writeback bits
942 943 944
		 *
		 * Do set the Private2 bit so we know this page was properly
		 * setup for writepage
C
Chris Mason 已提交
945
		 */
946 947 948 949
		op = unlock ? EXTENT_CLEAR_UNLOCK_PAGE : 0;
		op |= EXTENT_CLEAR_UNLOCK | EXTENT_CLEAR_DELALLOC |
			EXTENT_SET_PRIVATE2;

C
Chris Mason 已提交
950 951
		extent_clear_unlock_delalloc(inode, &BTRFS_I(inode)->io_tree,
					     start, start + ram_size - 1,
952
					     locked_page, op);
C
Chris Mason 已提交
953
		disk_num_bytes -= cur_alloc_size;
954 955 956
		num_bytes -= cur_alloc_size;
		alloc_hint = ins.objectid + ins.offset;
		start += cur_alloc_size;
957
	}
958
	ret = 0;
959
out:
960
	btrfs_end_transaction(trans, root);
C
Chris Mason 已提交
961

962
	return ret;
963 964 965
out_unlock:
	extent_clear_unlock_delalloc(inode,
		     &BTRFS_I(inode)->io_tree,
966
		     start, end, locked_page,
967 968 969 970 971 972 973 974
		     EXTENT_CLEAR_UNLOCK_PAGE |
		     EXTENT_CLEAR_UNLOCK |
		     EXTENT_CLEAR_DELALLOC |
		     EXTENT_CLEAR_DIRTY |
		     EXTENT_SET_WRITEBACK |
		     EXTENT_END_WRITEBACK);

	goto out;
975
}
C
Chris Mason 已提交
976

977 978 979 980 981 982 983 984 985 986 987 988
/*
 * work queue call back to started compression on a file and pages
 */
static noinline void async_cow_start(struct btrfs_work *work)
{
	struct async_cow *async_cow;
	int num_added = 0;
	async_cow = container_of(work, struct async_cow, work);

	compress_file_range(async_cow->inode, async_cow->locked_page,
			    async_cow->start, async_cow->end, async_cow,
			    &num_added);
989
	if (num_added == 0) {
990
		btrfs_add_delayed_iput(async_cow->inode);
991
		async_cow->inode = NULL;
992
	}
993 994 995 996 997 998 999 1000 1001 1002 1003 1004 1005 1006 1007 1008 1009 1010 1011 1012 1013 1014 1015 1016
}

/*
 * work queue call back to submit previously compressed pages
 */
static noinline void async_cow_submit(struct btrfs_work *work)
{
	struct async_cow *async_cow;
	struct btrfs_root *root;
	unsigned long nr_pages;

	async_cow = container_of(work, struct async_cow, work);

	root = async_cow->root;
	nr_pages = (async_cow->end - async_cow->start + PAGE_CACHE_SIZE) >>
		PAGE_CACHE_SHIFT;

	atomic_sub(nr_pages, &root->fs_info->async_delalloc_pages);

	if (atomic_read(&root->fs_info->async_delalloc_pages) <
	    5 * 1042 * 1024 &&
	    waitqueue_active(&root->fs_info->async_submit_wait))
		wake_up(&root->fs_info->async_submit_wait);

C
Chris Mason 已提交
1017
	if (async_cow->inode)
1018 1019
		submit_compressed_extents(async_cow->inode, async_cow);
}
C
Chris Mason 已提交
1020

1021 1022 1023 1024
static noinline void async_cow_free(struct btrfs_work *work)
{
	struct async_cow *async_cow;
	async_cow = container_of(work, struct async_cow, work);
1025
	if (async_cow->inode)
1026
		btrfs_add_delayed_iput(async_cow->inode);
1027 1028 1029 1030 1031 1032 1033 1034 1035 1036 1037 1038 1039
	kfree(async_cow);
}

static int cow_file_range_async(struct inode *inode, struct page *locked_page,
				u64 start, u64 end, int *page_started,
				unsigned long *nr_written)
{
	struct async_cow *async_cow;
	struct btrfs_root *root = BTRFS_I(inode)->root;
	unsigned long nr_pages;
	u64 cur_end;
	int limit = 10 * 1024 * 1042;

1040 1041
	clear_extent_bit(&BTRFS_I(inode)->io_tree, start, end, EXTENT_LOCKED,
			 1, 0, NULL, GFP_NOFS);
C
Chris Mason 已提交
1042
	while (start < end) {
1043
		async_cow = kmalloc(sizeof(*async_cow), GFP_NOFS);
1044
		BUG_ON(!async_cow); /* -ENOMEM */
1045
		async_cow->inode = igrab(inode);
1046 1047 1048 1049
		async_cow->root = root;
		async_cow->locked_page = locked_page;
		async_cow->start = start;

1050
		if (BTRFS_I(inode)->flags & BTRFS_INODE_NOCOMPRESS)
1051 1052 1053 1054 1055 1056 1057 1058 1059 1060 1061 1062 1063 1064 1065 1066 1067 1068 1069 1070 1071 1072 1073 1074 1075
			cur_end = end;
		else
			cur_end = min(end, start + 512 * 1024 - 1);

		async_cow->end = cur_end;
		INIT_LIST_HEAD(&async_cow->extents);

		async_cow->work.func = async_cow_start;
		async_cow->work.ordered_func = async_cow_submit;
		async_cow->work.ordered_free = async_cow_free;
		async_cow->work.flags = 0;

		nr_pages = (cur_end - start + PAGE_CACHE_SIZE) >>
			PAGE_CACHE_SHIFT;
		atomic_add(nr_pages, &root->fs_info->async_delalloc_pages);

		btrfs_queue_worker(&root->fs_info->delalloc_workers,
				   &async_cow->work);

		if (atomic_read(&root->fs_info->async_delalloc_pages) > limit) {
			wait_event(root->fs_info->async_submit_wait,
			   (atomic_read(&root->fs_info->async_delalloc_pages) <
			    limit));
		}

C
Chris Mason 已提交
1076
		while (atomic_read(&root->fs_info->async_submit_draining) &&
1077 1078 1079 1080 1081 1082 1083 1084 1085 1086 1087
		      atomic_read(&root->fs_info->async_delalloc_pages)) {
			wait_event(root->fs_info->async_submit_wait,
			  (atomic_read(&root->fs_info->async_delalloc_pages) ==
			   0));
		}

		*nr_written += nr_pages;
		start = cur_end + 1;
	}
	*page_started = 1;
	return 0;
1088 1089
}

C
Chris Mason 已提交
1090
static noinline int csum_exist_in_range(struct btrfs_root *root,
1091 1092 1093 1094 1095 1096
					u64 bytenr, u64 num_bytes)
{
	int ret;
	struct btrfs_ordered_sum *sums;
	LIST_HEAD(list);

Y
Yan Zheng 已提交
1097
	ret = btrfs_lookup_csums_range(root->fs_info->csum_root, bytenr,
A
Arne Jansen 已提交
1098
				       bytenr + num_bytes - 1, &list, 0);
1099 1100 1101 1102 1103 1104 1105 1106 1107 1108 1109
	if (ret == 0 && list_empty(&list))
		return 0;

	while (!list_empty(&list)) {
		sums = list_entry(list.next, struct btrfs_ordered_sum, list);
		list_del(&sums->list);
		kfree(sums);
	}
	return 1;
}

C
Chris Mason 已提交
1110 1111 1112 1113 1114 1115 1116
/*
 * when nowcow writeback call back.  This checks for snapshots or COW copies
 * of the extents that exist in the file, and COWs the file as required.
 *
 * If no cow copies or snapshots exist, we write directly to the existing
 * blocks on disk
 */
1117 1118
static noinline int run_delalloc_nocow(struct inode *inode,
				       struct page *locked_page,
1119 1120
			      u64 start, u64 end, int *page_started, int force,
			      unsigned long *nr_written)
1121 1122
{
	struct btrfs_root *root = BTRFS_I(inode)->root;
1123
	struct btrfs_trans_handle *trans;
1124 1125
	struct extent_buffer *leaf;
	struct btrfs_path *path;
Y
Yan Zheng 已提交
1126
	struct btrfs_file_extent_item *fi;
1127
	struct btrfs_key found_key;
Y
Yan Zheng 已提交
1128 1129 1130
	u64 cow_start;
	u64 cur_offset;
	u64 extent_end;
1131
	u64 extent_offset;
Y
Yan Zheng 已提交
1132 1133 1134
	u64 disk_bytenr;
	u64 num_bytes;
	int extent_type;
1135
	int ret, err;
Y
Yan Zheng 已提交
1136
	int type;
Y
Yan Zheng 已提交
1137 1138
	int nocow;
	int check_prev = 1;
1139
	bool nolock;
L
Li Zefan 已提交
1140
	u64 ino = btrfs_ino(inode);
1141 1142

	path = btrfs_alloc_path();
1143 1144 1145 1146 1147 1148 1149 1150 1151 1152
	if (!path) {
		extent_clear_unlock_delalloc(inode,
			     &BTRFS_I(inode)->io_tree,
			     start, end, locked_page,
			     EXTENT_CLEAR_UNLOCK_PAGE |
			     EXTENT_CLEAR_UNLOCK |
			     EXTENT_CLEAR_DELALLOC |
			     EXTENT_CLEAR_DIRTY |
			     EXTENT_SET_WRITEBACK |
			     EXTENT_END_WRITEBACK);
1153
		return -ENOMEM;
1154
	}
1155

1156
	nolock = btrfs_is_free_space_inode(root, inode);
1157 1158

	if (nolock)
1159
		trans = btrfs_join_transaction_nolock(root);
1160
	else
1161
		trans = btrfs_join_transaction(root);
C
Chris Mason 已提交
1162

1163
	if (IS_ERR(trans)) {
1164 1165 1166 1167 1168 1169 1170 1171 1172
		extent_clear_unlock_delalloc(inode,
			     &BTRFS_I(inode)->io_tree,
			     start, end, locked_page,
			     EXTENT_CLEAR_UNLOCK_PAGE |
			     EXTENT_CLEAR_UNLOCK |
			     EXTENT_CLEAR_DELALLOC |
			     EXTENT_CLEAR_DIRTY |
			     EXTENT_SET_WRITEBACK |
			     EXTENT_END_WRITEBACK);
1173 1174 1175 1176
		btrfs_free_path(path);
		return PTR_ERR(trans);
	}

1177
	trans->block_rsv = &root->fs_info->delalloc_block_rsv;
1178

Y
Yan Zheng 已提交
1179 1180 1181
	cow_start = (u64)-1;
	cur_offset = start;
	while (1) {
L
Li Zefan 已提交
1182
		ret = btrfs_lookup_file_extent(trans, root, path, ino,
Y
Yan Zheng 已提交
1183
					       cur_offset, 0);
1184 1185 1186 1187
		if (ret < 0) {
			btrfs_abort_transaction(trans, root, ret);
			goto error;
		}
Y
Yan Zheng 已提交
1188 1189 1190 1191
		if (ret > 0 && path->slots[0] > 0 && check_prev) {
			leaf = path->nodes[0];
			btrfs_item_key_to_cpu(leaf, &found_key,
					      path->slots[0] - 1);
L
Li Zefan 已提交
1192
			if (found_key.objectid == ino &&
Y
Yan Zheng 已提交
1193 1194 1195 1196 1197 1198 1199 1200
			    found_key.type == BTRFS_EXTENT_DATA_KEY)
				path->slots[0]--;
		}
		check_prev = 0;
next_slot:
		leaf = path->nodes[0];
		if (path->slots[0] >= btrfs_header_nritems(leaf)) {
			ret = btrfs_next_leaf(root, path);
1201 1202 1203 1204
			if (ret < 0) {
				btrfs_abort_transaction(trans, root, ret);
				goto error;
			}
Y
Yan Zheng 已提交
1205 1206 1207 1208
			if (ret > 0)
				break;
			leaf = path->nodes[0];
		}
1209

Y
Yan Zheng 已提交
1210 1211
		nocow = 0;
		disk_bytenr = 0;
1212
		num_bytes = 0;
Y
Yan Zheng 已提交
1213 1214
		btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);

L
Li Zefan 已提交
1215
		if (found_key.objectid > ino ||
Y
Yan Zheng 已提交
1216 1217 1218 1219 1220 1221
		    found_key.type > BTRFS_EXTENT_DATA_KEY ||
		    found_key.offset > end)
			break;

		if (found_key.offset > cur_offset) {
			extent_end = found_key.offset;
1222
			extent_type = 0;
Y
Yan Zheng 已提交
1223 1224 1225 1226 1227 1228 1229
			goto out_check;
		}

		fi = btrfs_item_ptr(leaf, path->slots[0],
				    struct btrfs_file_extent_item);
		extent_type = btrfs_file_extent_type(leaf, fi);

Y
Yan Zheng 已提交
1230 1231
		if (extent_type == BTRFS_FILE_EXTENT_REG ||
		    extent_type == BTRFS_FILE_EXTENT_PREALLOC) {
Y
Yan Zheng 已提交
1232
			disk_bytenr = btrfs_file_extent_disk_bytenr(leaf, fi);
1233
			extent_offset = btrfs_file_extent_offset(leaf, fi);
Y
Yan Zheng 已提交
1234 1235 1236 1237 1238 1239
			extent_end = found_key.offset +
				btrfs_file_extent_num_bytes(leaf, fi);
			if (extent_end <= start) {
				path->slots[0]++;
				goto next_slot;
			}
1240 1241
			if (disk_bytenr == 0)
				goto out_check;
Y
Yan Zheng 已提交
1242 1243 1244 1245
			if (btrfs_file_extent_compression(leaf, fi) ||
			    btrfs_file_extent_encryption(leaf, fi) ||
			    btrfs_file_extent_other_encoding(leaf, fi))
				goto out_check;
Y
Yan Zheng 已提交
1246 1247
			if (extent_type == BTRFS_FILE_EXTENT_REG && !force)
				goto out_check;
1248
			if (btrfs_extent_readonly(root, disk_bytenr))
Y
Yan Zheng 已提交
1249
				goto out_check;
L
Li Zefan 已提交
1250
			if (btrfs_cross_ref_exist(trans, root, ino,
1251 1252
						  found_key.offset -
						  extent_offset, disk_bytenr))
1253
				goto out_check;
1254
			disk_bytenr += extent_offset;
1255 1256 1257 1258 1259 1260 1261 1262 1263
			disk_bytenr += cur_offset - found_key.offset;
			num_bytes = min(end + 1, extent_end) - cur_offset;
			/*
			 * force cow if csum exists in the range.
			 * this ensure that csum for a given extent are
			 * either valid or do not exist.
			 */
			if (csum_exist_in_range(root, disk_bytenr, num_bytes))
				goto out_check;
Y
Yan Zheng 已提交
1264 1265 1266 1267 1268 1269 1270 1271 1272 1273 1274 1275 1276 1277 1278 1279 1280 1281 1282 1283 1284
			nocow = 1;
		} else if (extent_type == BTRFS_FILE_EXTENT_INLINE) {
			extent_end = found_key.offset +
				btrfs_file_extent_inline_len(leaf, fi);
			extent_end = ALIGN(extent_end, root->sectorsize);
		} else {
			BUG_ON(1);
		}
out_check:
		if (extent_end <= start) {
			path->slots[0]++;
			goto next_slot;
		}
		if (!nocow) {
			if (cow_start == (u64)-1)
				cow_start = cur_offset;
			cur_offset = extent_end;
			if (cur_offset > end)
				break;
			path->slots[0]++;
			goto next_slot;
1285 1286
		}

1287
		btrfs_release_path(path);
Y
Yan Zheng 已提交
1288 1289
		if (cow_start != (u64)-1) {
			ret = cow_file_range(inode, locked_page, cow_start,
1290 1291
					found_key.offset - 1, page_started,
					nr_written, 1);
1292 1293 1294 1295
			if (ret) {
				btrfs_abort_transaction(trans, root, ret);
				goto error;
			}
Y
Yan Zheng 已提交
1296
			cow_start = (u64)-1;
1297
		}
Y
Yan Zheng 已提交
1298

Y
Yan Zheng 已提交
1299 1300 1301 1302
		if (extent_type == BTRFS_FILE_EXTENT_PREALLOC) {
			struct extent_map *em;
			struct extent_map_tree *em_tree;
			em_tree = &BTRFS_I(inode)->extent_tree;
1303
			em = alloc_extent_map();
1304
			BUG_ON(!em); /* -ENOMEM */
Y
Yan Zheng 已提交
1305
			em->start = cur_offset;
1306
			em->orig_start = em->start;
Y
Yan Zheng 已提交
1307 1308 1309 1310 1311 1312
			em->len = num_bytes;
			em->block_len = num_bytes;
			em->block_start = disk_bytenr;
			em->bdev = root->fs_info->fs_devices->latest_bdev;
			set_bit(EXTENT_FLAG_PINNED, &em->flags);
			while (1) {
1313
				write_lock(&em_tree->lock);
Y
Yan Zheng 已提交
1314
				ret = add_extent_mapping(em_tree, em);
1315
				write_unlock(&em_tree->lock);
Y
Yan Zheng 已提交
1316 1317 1318 1319 1320 1321 1322 1323 1324 1325 1326
				if (ret != -EEXIST) {
					free_extent_map(em);
					break;
				}
				btrfs_drop_extent_cache(inode, em->start,
						em->start + em->len - 1, 0);
			}
			type = BTRFS_ORDERED_PREALLOC;
		} else {
			type = BTRFS_ORDERED_NOCOW;
		}
Y
Yan Zheng 已提交
1327 1328

		ret = btrfs_add_ordered_extent(inode, cur_offset, disk_bytenr,
Y
Yan Zheng 已提交
1329
					       num_bytes, num_bytes, type);
1330
		BUG_ON(ret); /* -ENOMEM */
1331

1332 1333 1334 1335
		if (root->root_key.objectid ==
		    BTRFS_DATA_RELOC_TREE_OBJECTID) {
			ret = btrfs_reloc_clone_csums(inode, cur_offset,
						      num_bytes);
1336 1337 1338 1339
			if (ret) {
				btrfs_abort_transaction(trans, root, ret);
				goto error;
			}
1340 1341
		}

Y
Yan Zheng 已提交
1342
		extent_clear_unlock_delalloc(inode, &BTRFS_I(inode)->io_tree,
1343 1344 1345 1346
				cur_offset, cur_offset + num_bytes - 1,
				locked_page, EXTENT_CLEAR_UNLOCK_PAGE |
				EXTENT_CLEAR_UNLOCK | EXTENT_CLEAR_DELALLOC |
				EXTENT_SET_PRIVATE2);
Y
Yan Zheng 已提交
1347 1348 1349
		cur_offset = extent_end;
		if (cur_offset > end)
			break;
1350
	}
1351
	btrfs_release_path(path);
Y
Yan Zheng 已提交
1352

1353
	if (cur_offset <= end && cow_start == (u64)-1) {
Y
Yan Zheng 已提交
1354
		cow_start = cur_offset;
1355 1356 1357
		cur_offset = end;
	}

Y
Yan Zheng 已提交
1358 1359
	if (cow_start != (u64)-1) {
		ret = cow_file_range(inode, locked_page, cow_start, end,
1360
				     page_started, nr_written, 1);
1361 1362 1363 1364
		if (ret) {
			btrfs_abort_transaction(trans, root, ret);
			goto error;
		}
Y
Yan Zheng 已提交
1365 1366
	}

1367
error:
J
Josef Bacik 已提交
1368
	if (nolock) {
1369
		err = btrfs_end_transaction_nolock(trans, root);
J
Josef Bacik 已提交
1370
	} else {
1371
		err = btrfs_end_transaction(trans, root);
J
Josef Bacik 已提交
1372
	}
1373 1374 1375
	if (!ret)
		ret = err;

1376 1377 1378 1379 1380 1381 1382 1383 1384 1385 1386
	if (ret && cur_offset < end)
		extent_clear_unlock_delalloc(inode,
			     &BTRFS_I(inode)->io_tree,
			     cur_offset, end, locked_page,
			     EXTENT_CLEAR_UNLOCK_PAGE |
			     EXTENT_CLEAR_UNLOCK |
			     EXTENT_CLEAR_DELALLOC |
			     EXTENT_CLEAR_DIRTY |
			     EXTENT_SET_WRITEBACK |
			     EXTENT_END_WRITEBACK);

1387
	btrfs_free_path(path);
1388
	return ret;
1389 1390
}

C
Chris Mason 已提交
1391 1392 1393
/*
 * extent_io.c call back to do delayed allocation processing
 */
C
Chris Mason 已提交
1394
static int run_delalloc_range(struct inode *inode, struct page *locked_page,
1395 1396
			      u64 start, u64 end, int *page_started,
			      unsigned long *nr_written)
1397 1398
{
	int ret;
1399
	struct btrfs_root *root = BTRFS_I(inode)->root;
1400

1401
	if (BTRFS_I(inode)->flags & BTRFS_INODE_NODATACOW) {
C
Chris Mason 已提交
1402
		ret = run_delalloc_nocow(inode, locked_page, start, end,
C
Chris Mason 已提交
1403
					 page_started, 1, nr_written);
1404
	} else if (BTRFS_I(inode)->flags & BTRFS_INODE_PREALLOC) {
Y
Yan Zheng 已提交
1405
		ret = run_delalloc_nocow(inode, locked_page, start, end,
C
Chris Mason 已提交
1406
					 page_started, 0, nr_written);
1407 1408 1409
	} else if (!btrfs_test_opt(root, COMPRESS) &&
		   !(BTRFS_I(inode)->force_compress) &&
		   !(BTRFS_I(inode)->flags & BTRFS_INODE_COMPRESS)) {
1410 1411
		ret = cow_file_range(inode, locked_page, start, end,
				      page_started, nr_written, 1);
1412 1413 1414
	} else {
		set_bit(BTRFS_INODE_HAS_ASYNC_EXTENT,
			&BTRFS_I(inode)->runtime_flags);
1415
		ret = cow_file_range_async(inode, locked_page, start, end,
C
Chris Mason 已提交
1416
					   page_started, nr_written);
1417
	}
1418 1419 1420
	return ret;
}

1421 1422
static void btrfs_split_extent_hook(struct inode *inode,
				    struct extent_state *orig, u64 split)
J
Josef Bacik 已提交
1423
{
1424
	/* not delalloc, ignore it */
J
Josef Bacik 已提交
1425
	if (!(orig->state & EXTENT_DELALLOC))
1426
		return;
J
Josef Bacik 已提交
1427

1428 1429 1430
	spin_lock(&BTRFS_I(inode)->lock);
	BTRFS_I(inode)->outstanding_extents++;
	spin_unlock(&BTRFS_I(inode)->lock);
J
Josef Bacik 已提交
1431 1432 1433 1434 1435 1436 1437 1438
}

/*
 * extent_io.c merge_extent_hook, used to track merged delayed allocation
 * extents so we can keep track of new extents that are just merged onto old
 * extents, such as when we are doing sequential writes, so we can properly
 * account for the metadata space we'll need.
 */
1439 1440 1441
static void btrfs_merge_extent_hook(struct inode *inode,
				    struct extent_state *new,
				    struct extent_state *other)
J
Josef Bacik 已提交
1442 1443 1444
{
	/* not delalloc, ignore it */
	if (!(other->state & EXTENT_DELALLOC))
1445
		return;
J
Josef Bacik 已提交
1446

1447 1448 1449
	spin_lock(&BTRFS_I(inode)->lock);
	BTRFS_I(inode)->outstanding_extents--;
	spin_unlock(&BTRFS_I(inode)->lock);
J
Josef Bacik 已提交
1450 1451
}

C
Chris Mason 已提交
1452 1453 1454 1455 1456
/*
 * extent_io.c set_bit_hook, used to track delayed allocation
 * bytes in this file, and to maintain the list of inodes that
 * have pending delalloc work to be done.
 */
1457 1458
static void btrfs_set_bit_hook(struct inode *inode,
			       struct extent_state *state, int *bits)
1459
{
J
Josef Bacik 已提交
1460

1461 1462
	/*
	 * set_bit and clear bit hooks normally require _irqsave/restore
1463
	 * but in this case, we are only testing for the DELALLOC
1464 1465
	 * bit, which is only set or cleared with irqs on
	 */
1466
	if (!(state->state & EXTENT_DELALLOC) && (*bits & EXTENT_DELALLOC)) {
1467
		struct btrfs_root *root = BTRFS_I(inode)->root;
1468
		u64 len = state->end + 1 - state->start;
1469
		bool do_list = !btrfs_is_free_space_inode(root, inode);
J
Josef Bacik 已提交
1470

1471
		if (*bits & EXTENT_FIRST_DELALLOC) {
1472
			*bits &= ~EXTENT_FIRST_DELALLOC;
1473 1474 1475 1476 1477
		} else {
			spin_lock(&BTRFS_I(inode)->lock);
			BTRFS_I(inode)->outstanding_extents++;
			spin_unlock(&BTRFS_I(inode)->lock);
		}
1478

1479
		spin_lock(&root->fs_info->delalloc_lock);
1480 1481
		BTRFS_I(inode)->delalloc_bytes += len;
		root->fs_info->delalloc_bytes += len;
J
Josef Bacik 已提交
1482
		if (do_list && list_empty(&BTRFS_I(inode)->delalloc_inodes)) {
1483 1484 1485
			list_add_tail(&BTRFS_I(inode)->delalloc_inodes,
				      &root->fs_info->delalloc_inodes);
		}
1486
		spin_unlock(&root->fs_info->delalloc_lock);
1487 1488 1489
	}
}

C
Chris Mason 已提交
1490 1491 1492
/*
 * extent_io.c clear_bit_hook, see set_bit_hook for why
 */
1493 1494
static void btrfs_clear_bit_hook(struct inode *inode,
				 struct extent_state *state, int *bits)
1495
{
1496 1497
	/*
	 * set_bit and clear bit hooks normally require _irqsave/restore
1498
	 * but in this case, we are only testing for the DELALLOC
1499 1500
	 * bit, which is only set or cleared with irqs on
	 */
1501
	if ((state->state & EXTENT_DELALLOC) && (*bits & EXTENT_DELALLOC)) {
1502
		struct btrfs_root *root = BTRFS_I(inode)->root;
1503
		u64 len = state->end + 1 - state->start;
1504
		bool do_list = !btrfs_is_free_space_inode(root, inode);
1505

1506
		if (*bits & EXTENT_FIRST_DELALLOC) {
1507
			*bits &= ~EXTENT_FIRST_DELALLOC;
1508 1509 1510 1511 1512
		} else if (!(*bits & EXTENT_DO_ACCOUNTING)) {
			spin_lock(&BTRFS_I(inode)->lock);
			BTRFS_I(inode)->outstanding_extents--;
			spin_unlock(&BTRFS_I(inode)->lock);
		}
1513 1514 1515 1516

		if (*bits & EXTENT_DO_ACCOUNTING)
			btrfs_delalloc_release_metadata(inode, len);

J
Josef Bacik 已提交
1517 1518
		if (root->root_key.objectid != BTRFS_DATA_RELOC_TREE_OBJECTID
		    && do_list)
1519
			btrfs_free_reserved_data_space(inode, len);
J
Josef Bacik 已提交
1520

1521
		spin_lock(&root->fs_info->delalloc_lock);
1522 1523 1524
		root->fs_info->delalloc_bytes -= len;
		BTRFS_I(inode)->delalloc_bytes -= len;

J
Josef Bacik 已提交
1525
		if (do_list && BTRFS_I(inode)->delalloc_bytes == 0 &&
1526 1527 1528
		    !list_empty(&BTRFS_I(inode)->delalloc_inodes)) {
			list_del_init(&BTRFS_I(inode)->delalloc_inodes);
		}
1529
		spin_unlock(&root->fs_info->delalloc_lock);
1530 1531 1532
	}
}

C
Chris Mason 已提交
1533 1534 1535 1536
/*
 * extent_io.c merge_bio_hook, this must check the chunk tree to make sure
 * we don't create bios that span stripes or chunks
 */
1537
int btrfs_merge_bio_hook(struct page *page, unsigned long offset,
C
Chris Mason 已提交
1538 1539
			 size_t size, struct bio *bio,
			 unsigned long bio_flags)
1540 1541 1542
{
	struct btrfs_root *root = BTRFS_I(page->mapping->host)->root;
	struct btrfs_mapping_tree *map_tree;
1543
	u64 logical = (u64)bio->bi_sector << 9;
1544 1545 1546 1547
	u64 length = 0;
	u64 map_length;
	int ret;

1548 1549 1550
	if (bio_flags & EXTENT_BIO_COMPRESSED)
		return 0;

1551
	length = bio->bi_size;
1552 1553
	map_tree = &root->fs_info->mapping_tree;
	map_length = length;
1554
	ret = btrfs_map_block(map_tree, READ, logical,
1555
			      &map_length, NULL, 0);
1556 1557
	/* Will always return 0 or 1 with map_multi == NULL */
	BUG_ON(ret < 0);
C
Chris Mason 已提交
1558
	if (map_length < length + size)
1559
		return 1;
1560
	return 0;
1561 1562
}

C
Chris Mason 已提交
1563 1564 1565 1566 1567 1568 1569 1570
/*
 * in order to insert checksums into the metadata in large chunks,
 * we wait until bio submission time.   All the pages in the bio are
 * checksummed and sums are attached onto the ordered extent record.
 *
 * At IO completion time the cums attached on the ordered extent record
 * are inserted into the btree
 */
C
Chris Mason 已提交
1571 1572
static int __btrfs_submit_bio_start(struct inode *inode, int rw,
				    struct bio *bio, int mirror_num,
1573 1574
				    unsigned long bio_flags,
				    u64 bio_offset)
1575 1576 1577
{
	struct btrfs_root *root = BTRFS_I(inode)->root;
	int ret = 0;
1578

1579
	ret = btrfs_csum_one_bio(root, inode, bio, 0, 0);
1580
	BUG_ON(ret); /* -ENOMEM */
C
Chris Mason 已提交
1581 1582
	return 0;
}
1583

C
Chris Mason 已提交
1584 1585 1586 1587 1588 1589 1590 1591
/*
 * in order to insert checksums into the metadata in large chunks,
 * we wait until bio submission time.   All the pages in the bio are
 * checksummed and sums are attached onto the ordered extent record.
 *
 * At IO completion time the cums attached on the ordered extent record
 * are inserted into the btree
 */
1592
static int __btrfs_submit_bio_done(struct inode *inode, int rw, struct bio *bio,
1593 1594
			  int mirror_num, unsigned long bio_flags,
			  u64 bio_offset)
C
Chris Mason 已提交
1595 1596
{
	struct btrfs_root *root = BTRFS_I(inode)->root;
1597
	return btrfs_map_bio(root, rw, bio, mirror_num, 1);
1598 1599
}

C
Chris Mason 已提交
1600
/*
1601 1602
 * extent_io.c submission hook. This does the right thing for csum calculation
 * on write, or reading the csums from the tree before a read
C
Chris Mason 已提交
1603
 */
1604
static int btrfs_submit_bio_hook(struct inode *inode, int rw, struct bio *bio,
1605 1606
			  int mirror_num, unsigned long bio_flags,
			  u64 bio_offset)
1607 1608 1609
{
	struct btrfs_root *root = BTRFS_I(inode)->root;
	int ret = 0;
1610
	int skip_sum;
1611
	int metadata = 0;
1612

1613
	skip_sum = BTRFS_I(inode)->flags & BTRFS_INODE_NODATASUM;
1614

1615
	if (btrfs_is_free_space_inode(root, inode))
1616 1617
		metadata = 2;

1618
	if (!(rw & REQ_WRITE)) {
1619 1620 1621 1622
		ret = btrfs_bio_wq_end_io(root->fs_info, bio, metadata);
		if (ret)
			return ret;

1623
		if (bio_flags & EXTENT_BIO_COMPRESSED) {
C
Chris Mason 已提交
1624 1625
			return btrfs_submit_compressed_read(inode, bio,
						    mirror_num, bio_flags);
1626 1627 1628 1629 1630
		} else if (!skip_sum) {
			ret = btrfs_lookup_bio_sums(root, inode, bio, NULL);
			if (ret)
				return ret;
		}
1631
		goto mapit;
1632
	} else if (!skip_sum) {
1633 1634 1635
		/* csum items have already been cloned */
		if (root->root_key.objectid == BTRFS_DATA_RELOC_TREE_OBJECTID)
			goto mapit;
1636 1637
		/* we're doing a write, do the async checksumming */
		return btrfs_wq_submit_bio(BTRFS_I(inode)->root->fs_info,
1638
				   inode, rw, bio, mirror_num,
1639 1640
				   bio_flags, bio_offset,
				   __btrfs_submit_bio_start,
C
Chris Mason 已提交
1641
				   __btrfs_submit_bio_done);
1642 1643
	}

1644
mapit:
1645
	return btrfs_map_bio(root, rw, bio, mirror_num, 0);
1646
}
C
Chris Mason 已提交
1647

C
Chris Mason 已提交
1648 1649 1650 1651
/*
 * given a list of ordered sums record them in the inode.  This happens
 * at IO completion time based on sums calculated at bio submission time.
 */
1652
static noinline int add_pending_csums(struct btrfs_trans_handle *trans,
1653 1654 1655 1656 1657
			     struct inode *inode, u64 file_offset,
			     struct list_head *list)
{
	struct btrfs_ordered_sum *sum;

Q
Qinghuang Feng 已提交
1658
	list_for_each_entry(sum, list, list) {
1659 1660
		btrfs_csum_file_blocks(trans,
		       BTRFS_I(inode)->root->fs_info->csum_root, sum);
1661 1662 1663 1664
	}
	return 0;
}

1665 1666
int btrfs_set_extent_delalloc(struct inode *inode, u64 start, u64 end,
			      struct extent_state **cached_state)
1667
{
C
Chris Mason 已提交
1668
	if ((end & (PAGE_CACHE_SIZE - 1)) == 0)
1669
		WARN_ON(1);
1670
	return set_extent_delalloc(&BTRFS_I(inode)->io_tree, start, end,
1671
				   cached_state, GFP_NOFS);
1672 1673
}

C
Chris Mason 已提交
1674
/* see btrfs_writepage_start_hook for details on why this is required */
1675 1676 1677 1678 1679
struct btrfs_writepage_fixup {
	struct page *page;
	struct btrfs_work work;
};

1680
static void btrfs_writepage_fixup_worker(struct btrfs_work *work)
1681 1682 1683
{
	struct btrfs_writepage_fixup *fixup;
	struct btrfs_ordered_extent *ordered;
1684
	struct extent_state *cached_state = NULL;
1685 1686 1687 1688
	struct page *page;
	struct inode *inode;
	u64 page_start;
	u64 page_end;
1689
	int ret;
1690 1691 1692

	fixup = container_of(work, struct btrfs_writepage_fixup, work);
	page = fixup->page;
C
Chris Mason 已提交
1693
again:
1694 1695 1696 1697 1698 1699 1700 1701 1702 1703
	lock_page(page);
	if (!page->mapping || !PageDirty(page) || !PageChecked(page)) {
		ClearPageChecked(page);
		goto out_page;
	}

	inode = page->mapping->host;
	page_start = page_offset(page);
	page_end = page_offset(page) + PAGE_CACHE_SIZE - 1;

1704
	lock_extent_bits(&BTRFS_I(inode)->io_tree, page_start, page_end, 0,
1705
			 &cached_state);
C
Chris Mason 已提交
1706 1707

	/* already ordered? We're done */
1708
	if (PagePrivate2(page))
1709
		goto out;
C
Chris Mason 已提交
1710 1711 1712

	ordered = btrfs_lookup_ordered_extent(inode, page_start);
	if (ordered) {
1713 1714
		unlock_extent_cached(&BTRFS_I(inode)->io_tree, page_start,
				     page_end, &cached_state, GFP_NOFS);
C
Chris Mason 已提交
1715 1716
		unlock_page(page);
		btrfs_start_ordered_extent(inode, ordered, 1);
1717
		btrfs_put_ordered_extent(ordered);
C
Chris Mason 已提交
1718 1719
		goto again;
	}
1720

1721 1722 1723 1724 1725 1726 1727 1728
	ret = btrfs_delalloc_reserve_space(inode, PAGE_CACHE_SIZE);
	if (ret) {
		mapping_set_error(page->mapping, ret);
		end_extent_writepage(page, ret, page_start, page_end);
		ClearPageChecked(page);
		goto out;
	 }

1729
	btrfs_set_extent_delalloc(inode, page_start, page_end, &cached_state);
1730
	ClearPageChecked(page);
1731
	set_page_dirty(page);
1732
out:
1733 1734
	unlock_extent_cached(&BTRFS_I(inode)->io_tree, page_start, page_end,
			     &cached_state, GFP_NOFS);
1735 1736 1737
out_page:
	unlock_page(page);
	page_cache_release(page);
1738
	kfree(fixup);
1739 1740 1741 1742 1743 1744 1745 1746
}

/*
 * There are a few paths in the higher layers of the kernel that directly
 * set the page dirty bit without asking the filesystem if it is a
 * good idea.  This causes problems because we want to make sure COW
 * properly happens and the data=ordered rules are followed.
 *
C
Chris Mason 已提交
1747
 * In our case any range that doesn't have the ORDERED bit set
1748 1749 1750 1751
 * hasn't been properly setup for IO.  We kick off an async process
 * to fix it up.  The async helper will wait for ordered extents, set
 * the delalloc bit and make it safe to write the page.
 */
1752
static int btrfs_writepage_start_hook(struct page *page, u64 start, u64 end)
1753 1754 1755 1756 1757
{
	struct inode *inode = page->mapping->host;
	struct btrfs_writepage_fixup *fixup;
	struct btrfs_root *root = BTRFS_I(inode)->root;

1758 1759
	/* this page is properly in the ordered list */
	if (TestClearPagePrivate2(page))
1760 1761 1762 1763 1764 1765 1766 1767
		return 0;

	if (PageChecked(page))
		return -EAGAIN;

	fixup = kzalloc(sizeof(*fixup), GFP_NOFS);
	if (!fixup)
		return -EAGAIN;
1768

1769 1770 1771 1772 1773
	SetPageChecked(page);
	page_cache_get(page);
	fixup->work.func = btrfs_writepage_fixup_worker;
	fixup->page = page;
	btrfs_queue_worker(&root->fs_info->fixup_workers, &fixup->work);
1774
	return -EBUSY;
1775 1776
}

Y
Yan Zheng 已提交
1777 1778 1779 1780 1781 1782 1783 1784 1785 1786 1787 1788 1789 1790 1791 1792
static int insert_reserved_file_extent(struct btrfs_trans_handle *trans,
				       struct inode *inode, u64 file_pos,
				       u64 disk_bytenr, u64 disk_num_bytes,
				       u64 num_bytes, u64 ram_bytes,
				       u8 compression, u8 encryption,
				       u16 other_encoding, int extent_type)
{
	struct btrfs_root *root = BTRFS_I(inode)->root;
	struct btrfs_file_extent_item *fi;
	struct btrfs_path *path;
	struct extent_buffer *leaf;
	struct btrfs_key ins;
	u64 hint;
	int ret;

	path = btrfs_alloc_path();
1793 1794
	if (!path)
		return -ENOMEM;
Y
Yan Zheng 已提交
1795

1796
	path->leave_spinning = 1;
C
Chris Mason 已提交
1797 1798 1799 1800 1801 1802 1803 1804 1805 1806

	/*
	 * we may be replacing one extent in the tree with another.
	 * The new extent is pinned in the extent map, and we don't want
	 * to drop it from the cache until it is completely in the btree.
	 *
	 * So, tell btrfs_drop_extents to leave this extent in the cache.
	 * the caller is expected to unpin it and allow it to be merged
	 * with the others.
	 */
Y
Yan, Zheng 已提交
1807 1808
	ret = btrfs_drop_extents(trans, inode, file_pos, file_pos + num_bytes,
				 &hint, 0);
1809 1810
	if (ret)
		goto out;
Y
Yan Zheng 已提交
1811

L
Li Zefan 已提交
1812
	ins.objectid = btrfs_ino(inode);
Y
Yan Zheng 已提交
1813 1814 1815
	ins.offset = file_pos;
	ins.type = BTRFS_EXTENT_DATA_KEY;
	ret = btrfs_insert_empty_item(trans, root, path, &ins, sizeof(*fi));
1816 1817
	if (ret)
		goto out;
Y
Yan Zheng 已提交
1818 1819 1820 1821 1822 1823 1824 1825 1826 1827 1828 1829 1830
	leaf = path->nodes[0];
	fi = btrfs_item_ptr(leaf, path->slots[0],
			    struct btrfs_file_extent_item);
	btrfs_set_file_extent_generation(leaf, fi, trans->transid);
	btrfs_set_file_extent_type(leaf, fi, extent_type);
	btrfs_set_file_extent_disk_bytenr(leaf, fi, disk_bytenr);
	btrfs_set_file_extent_disk_num_bytes(leaf, fi, disk_num_bytes);
	btrfs_set_file_extent_offset(leaf, fi, 0);
	btrfs_set_file_extent_num_bytes(leaf, fi, num_bytes);
	btrfs_set_file_extent_ram_bytes(leaf, fi, ram_bytes);
	btrfs_set_file_extent_compression(leaf, fi, compression);
	btrfs_set_file_extent_encryption(leaf, fi, encryption);
	btrfs_set_file_extent_other_encoding(leaf, fi, other_encoding);
1831 1832 1833 1834

	btrfs_unlock_up_safe(path, 1);
	btrfs_set_lock_blocking(leaf);

Y
Yan Zheng 已提交
1835 1836 1837 1838 1839 1840 1841
	btrfs_mark_buffer_dirty(leaf);

	inode_add_bytes(inode, num_bytes);

	ins.objectid = disk_bytenr;
	ins.offset = disk_num_bytes;
	ins.type = BTRFS_EXTENT_ITEM_KEY;
1842 1843
	ret = btrfs_alloc_reserved_file_extent(trans, root,
					root->root_key.objectid,
L
Li Zefan 已提交
1844
					btrfs_ino(inode), file_pos, &ins);
1845
out:
Y
Yan Zheng 已提交
1846
	btrfs_free_path(path);
1847

1848
	return ret;
Y
Yan Zheng 已提交
1849 1850
}

1851 1852 1853 1854 1855 1856
/*
 * helper function for btrfs_finish_ordered_io, this
 * just reads in some of the csum leaves to prime them into ram
 * before we start the transaction.  It limits the amount of btree
 * reads required while inside the transaction.
 */
C
Chris Mason 已提交
1857 1858 1859 1860
/* as ordered data IO finishes, this gets called so we can finish
 * an ordered extent if the range of bytes in the file it covers are
 * fully written.
 */
1861
static int btrfs_finish_ordered_io(struct btrfs_ordered_extent *ordered_extent)
1862
{
1863
	struct inode *inode = ordered_extent->inode;
1864
	struct btrfs_root *root = BTRFS_I(inode)->root;
1865
	struct btrfs_trans_handle *trans = NULL;
1866
	struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
1867
	struct extent_state *cached_state = NULL;
1868
	int compress_type = 0;
1869
	int ret;
1870
	bool nolock;
1871

1872
	nolock = btrfs_is_free_space_inode(root, inode);
J
Josef Bacik 已提交
1873

1874 1875 1876 1877 1878
	if (test_bit(BTRFS_ORDERED_IOERR, &ordered_extent->flags)) {
		ret = -EIO;
		goto out;
	}

1879
	if (test_bit(BTRFS_ORDERED_NOCOW, &ordered_extent->flags)) {
1880
		BUG_ON(!list_empty(&ordered_extent->list)); /* Logic error */
1881 1882
		ret = btrfs_ordered_update_i_size(inode, 0, ordered_extent);
		if (!ret) {
J
Josef Bacik 已提交
1883
			if (nolock)
1884
				trans = btrfs_join_transaction_nolock(root);
J
Josef Bacik 已提交
1885
			else
1886
				trans = btrfs_join_transaction(root);
1887 1888
			if (IS_ERR(trans))
				return PTR_ERR(trans);
1889
			trans->block_rsv = &root->fs_info->delalloc_block_rsv;
1890
			ret = btrfs_update_inode_fallback(trans, root, inode);
1891 1892
			if (ret) /* -ENOMEM or corruption */
				btrfs_abort_transaction(trans, root, ret);
1893 1894 1895
		}
		goto out;
	}
1896

1897 1898
	lock_extent_bits(io_tree, ordered_extent->file_offset,
			 ordered_extent->file_offset + ordered_extent->len - 1,
1899
			 0, &cached_state);
1900

J
Josef Bacik 已提交
1901
	if (nolock)
1902
		trans = btrfs_join_transaction_nolock(root);
J
Josef Bacik 已提交
1903
	else
1904
		trans = btrfs_join_transaction(root);
1905 1906 1907 1908 1909
	if (IS_ERR(trans)) {
		ret = PTR_ERR(trans);
		trans = NULL;
		goto out_unlock;
	}
1910
	trans->block_rsv = &root->fs_info->delalloc_block_rsv;
1911

C
Chris Mason 已提交
1912
	if (test_bit(BTRFS_ORDERED_COMPRESSED, &ordered_extent->flags))
1913
		compress_type = ordered_extent->compress_type;
Y
Yan Zheng 已提交
1914
	if (test_bit(BTRFS_ORDERED_PREALLOC, &ordered_extent->flags)) {
1915
		BUG_ON(compress_type);
Y
Yan, Zheng 已提交
1916
		ret = btrfs_mark_extent_written(trans, inode,
Y
Yan Zheng 已提交
1917 1918 1919 1920
						ordered_extent->file_offset,
						ordered_extent->file_offset +
						ordered_extent->len);
	} else {
1921
		BUG_ON(root == root->fs_info->tree_root);
Y
Yan Zheng 已提交
1922 1923 1924 1925 1926 1927
		ret = insert_reserved_file_extent(trans, inode,
						ordered_extent->file_offset,
						ordered_extent->start,
						ordered_extent->disk_len,
						ordered_extent->len,
						ordered_extent->len,
1928
						compress_type, 0, 0,
Y
Yan Zheng 已提交
1929
						BTRFS_FILE_EXTENT_REG);
C
Chris Mason 已提交
1930 1931 1932
		unpin_extent_cache(&BTRFS_I(inode)->extent_tree,
				   ordered_extent->file_offset,
				   ordered_extent->len);
Y
Yan Zheng 已提交
1933
	}
1934

1935 1936
	if (ret < 0) {
		btrfs_abort_transaction(trans, root, ret);
1937
		goto out_unlock;
1938
	}
1939

1940 1941 1942
	add_pending_csums(trans, inode, ordered_extent->file_offset,
			  &ordered_extent->list);

1943
	ret = btrfs_ordered_update_i_size(inode, 0, ordered_extent);
1944
	if (!ret || !test_bit(BTRFS_ORDERED_PREALLOC, &ordered_extent->flags)) {
1945
		ret = btrfs_update_inode_fallback(trans, root, inode);
1946 1947
		if (ret) { /* -ENOMEM or corruption */
			btrfs_abort_transaction(trans, root, ret);
1948
			goto out_unlock;
1949
		}
1950 1951
	}
	ret = 0;
1952 1953 1954 1955
out_unlock:
	unlock_extent_cached(io_tree, ordered_extent->file_offset,
			     ordered_extent->file_offset +
			     ordered_extent->len - 1, &cached_state, GFP_NOFS);
1956
out:
1957
	if (root != root->fs_info->tree_root)
J
Josef Bacik 已提交
1958
		btrfs_delalloc_release_metadata(inode, ordered_extent->len);
1959 1960
	if (trans) {
		if (nolock)
J
Josef Bacik 已提交
1961
			btrfs_end_transaction_nolock(trans, root);
1962
		else
J
Josef Bacik 已提交
1963 1964 1965
			btrfs_end_transaction(trans, root);
	}

1966 1967 1968 1969 1970 1971 1972 1973 1974 1975 1976
	if (ret)
		clear_extent_uptodate(io_tree, ordered_extent->file_offset,
				      ordered_extent->file_offset +
				      ordered_extent->len - 1, NULL, GFP_NOFS);

	/*
	 * This needs to be dont to make sure anybody waiting knows we are done
	 * upating everything for this ordered extent.
	 */
	btrfs_remove_ordered_extent(inode, ordered_extent);

1977 1978 1979 1980 1981
	/* once for us */
	btrfs_put_ordered_extent(ordered_extent);
	/* once for the tree */
	btrfs_put_ordered_extent(ordered_extent);

1982 1983 1984 1985 1986 1987 1988 1989
	return ret;
}

static void finish_ordered_fn(struct btrfs_work *work)
{
	struct btrfs_ordered_extent *ordered_extent;
	ordered_extent = container_of(work, struct btrfs_ordered_extent, work);
	btrfs_finish_ordered_io(ordered_extent);
1990 1991
}

1992
static int btrfs_writepage_end_io_hook(struct page *page, u64 start, u64 end,
1993 1994
				struct extent_state *state, int uptodate)
{
1995 1996 1997 1998 1999
	struct inode *inode = page->mapping->host;
	struct btrfs_root *root = BTRFS_I(inode)->root;
	struct btrfs_ordered_extent *ordered_extent = NULL;
	struct btrfs_workers *workers;

2000 2001
	trace_btrfs_writepage_end_io_hook(page, start, end, uptodate);

2002
	ClearPagePrivate2(page);
2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016
	if (!btrfs_dec_test_ordered_pending(inode, &ordered_extent, start,
					    end - start + 1, uptodate))
		return 0;

	ordered_extent->work.func = finish_ordered_fn;
	ordered_extent->work.flags = 0;

	if (btrfs_is_free_space_inode(root, inode))
		workers = &root->fs_info->endio_freespace_worker;
	else
		workers = &root->fs_info->endio_write_workers;
	btrfs_queue_worker(workers, &ordered_extent->work);

	return 0;
2017 2018
}

C
Chris Mason 已提交
2019 2020
/*
 * when reads are done, we need to check csums to verify the data is correct
2021 2022
 * if there's a match, we allow the bio to finish.  If not, the code in
 * extent_io.c will try to find good copies for us.
C
Chris Mason 已提交
2023
 */
2024
static int btrfs_readpage_end_io_hook(struct page *page, u64 start, u64 end,
2025
			       struct extent_state *state, int mirror)
2026
{
2027
	size_t offset = start - ((u64)page->index << PAGE_CACHE_SHIFT);
2028
	struct inode *inode = page->mapping->host;
2029
	struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
2030
	char *kaddr;
2031
	u64 private = ~(u32)0;
2032
	int ret;
2033 2034
	struct btrfs_root *root = BTRFS_I(inode)->root;
	u32 csum = ~(u32)0;
2035

2036 2037 2038 2039
	if (PageChecked(page)) {
		ClearPageChecked(page);
		goto good;
	}
2040 2041

	if (BTRFS_I(inode)->flags & BTRFS_INODE_NODATASUM)
2042
		goto good;
2043 2044

	if (root->root_key.objectid == BTRFS_DATA_RELOC_TREE_OBJECTID &&
2045
	    test_range_bit(io_tree, start, end, EXTENT_NODATASUM, 1, NULL)) {
2046 2047
		clear_extent_bits(io_tree, start, end, EXTENT_NODATASUM,
				  GFP_NOFS);
2048
		return 0;
2049
	}
2050

Y
Yan 已提交
2051
	if (state && state->start == start) {
2052 2053 2054 2055 2056
		private = state->private;
		ret = 0;
	} else {
		ret = get_state_private(io_tree, start, &private);
	}
2057
	kaddr = kmap_atomic(page);
C
Chris Mason 已提交
2058
	if (ret)
2059
		goto zeroit;
C
Chris Mason 已提交
2060

2061 2062
	csum = btrfs_csum_data(root, kaddr + offset, csum,  end - start + 1);
	btrfs_csum_final(csum, (char *)&csum);
C
Chris Mason 已提交
2063
	if (csum != private)
2064
		goto zeroit;
C
Chris Mason 已提交
2065

2066
	kunmap_atomic(kaddr);
2067
good:
2068 2069 2070
	return 0;

zeroit:
2071
	printk_ratelimited(KERN_INFO "btrfs csum failed ino %llu off %llu csum %u "
L
Li Zefan 已提交
2072 2073
		       "private %llu\n",
		       (unsigned long long)btrfs_ino(page->mapping->host),
C
Chris Mason 已提交
2074 2075
		       (unsigned long long)start, csum,
		       (unsigned long long)private);
2076 2077
	memset(kaddr + offset, 1, end - start + 1);
	flush_dcache_page(page);
2078
	kunmap_atomic(kaddr);
2079 2080
	if (private == 0)
		return 0;
2081
	return -EIO;
2082
}
2083

Y
Yan, Zheng 已提交
2084 2085 2086 2087 2088
struct delayed_iput {
	struct list_head list;
	struct inode *inode;
};

2089 2090
/* JDM: If this is fs-wide, why can't we add a pointer to
 * btrfs_inode instead and avoid the allocation? */
Y
Yan, Zheng 已提交
2091 2092 2093 2094 2095 2096 2097 2098 2099 2100 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 2126 2127 2128 2129 2130 2131 2132 2133
void btrfs_add_delayed_iput(struct inode *inode)
{
	struct btrfs_fs_info *fs_info = BTRFS_I(inode)->root->fs_info;
	struct delayed_iput *delayed;

	if (atomic_add_unless(&inode->i_count, -1, 1))
		return;

	delayed = kmalloc(sizeof(*delayed), GFP_NOFS | __GFP_NOFAIL);
	delayed->inode = inode;

	spin_lock(&fs_info->delayed_iput_lock);
	list_add_tail(&delayed->list, &fs_info->delayed_iputs);
	spin_unlock(&fs_info->delayed_iput_lock);
}

void btrfs_run_delayed_iputs(struct btrfs_root *root)
{
	LIST_HEAD(list);
	struct btrfs_fs_info *fs_info = root->fs_info;
	struct delayed_iput *delayed;
	int empty;

	spin_lock(&fs_info->delayed_iput_lock);
	empty = list_empty(&fs_info->delayed_iputs);
	spin_unlock(&fs_info->delayed_iput_lock);
	if (empty)
		return;

	down_read(&root->fs_info->cleanup_work_sem);
	spin_lock(&fs_info->delayed_iput_lock);
	list_splice_init(&fs_info->delayed_iputs, &list);
	spin_unlock(&fs_info->delayed_iput_lock);

	while (!list_empty(&list)) {
		delayed = list_entry(list.next, struct delayed_iput, list);
		list_del(&delayed->list);
		iput(delayed->inode);
		kfree(delayed);
	}
	up_read(&root->fs_info->cleanup_work_sem);
}

2134 2135 2136 2137 2138 2139
enum btrfs_orphan_cleanup_state {
	ORPHAN_CLEANUP_STARTED	= 1,
	ORPHAN_CLEANUP_DONE	= 2,
};

/*
2140
 * This is called in transaction commit time. If there are no orphan
2141 2142 2143 2144 2145 2146
 * files in the subvolume, it removes orphan item and frees block_rsv
 * structure.
 */
void btrfs_orphan_commit_root(struct btrfs_trans_handle *trans,
			      struct btrfs_root *root)
{
2147
	struct btrfs_block_rsv *block_rsv;
2148 2149
	int ret;

2150
	if (atomic_read(&root->orphan_inodes) ||
2151 2152 2153
	    root->orphan_cleanup_state != ORPHAN_CLEANUP_DONE)
		return;

2154
	spin_lock(&root->orphan_lock);
2155
	if (atomic_read(&root->orphan_inodes)) {
2156 2157 2158 2159 2160 2161 2162 2163 2164 2165 2166 2167 2168
		spin_unlock(&root->orphan_lock);
		return;
	}

	if (root->orphan_cleanup_state != ORPHAN_CLEANUP_DONE) {
		spin_unlock(&root->orphan_lock);
		return;
	}

	block_rsv = root->orphan_block_rsv;
	root->orphan_block_rsv = NULL;
	spin_unlock(&root->orphan_lock);

2169 2170 2171 2172 2173 2174 2175 2176
	if (root->orphan_item_inserted &&
	    btrfs_root_refs(&root->root_item) > 0) {
		ret = btrfs_del_orphan_item(trans, root->fs_info->tree_root,
					    root->root_key.objectid);
		BUG_ON(ret);
		root->orphan_item_inserted = 0;
	}

2177 2178 2179
	if (block_rsv) {
		WARN_ON(block_rsv->size > 0);
		btrfs_free_block_rsv(root, block_rsv);
2180 2181 2182
	}
}

2183 2184 2185
/*
 * This creates an orphan entry for the given inode in case something goes
 * wrong in the middle of an unlink/truncate.
2186 2187 2188
 *
 * NOTE: caller of this function should reserve 5 units of metadata for
 *	 this function.
2189 2190 2191 2192
 */
int btrfs_orphan_add(struct btrfs_trans_handle *trans, struct inode *inode)
{
	struct btrfs_root *root = BTRFS_I(inode)->root;
2193 2194 2195 2196
	struct btrfs_block_rsv *block_rsv = NULL;
	int reserve = 0;
	int insert = 0;
	int ret;
2197

2198 2199
	if (!root->orphan_block_rsv) {
		block_rsv = btrfs_alloc_block_rsv(root);
2200 2201
		if (!block_rsv)
			return -ENOMEM;
2202
	}
2203

2204 2205 2206 2207 2208 2209
	spin_lock(&root->orphan_lock);
	if (!root->orphan_block_rsv) {
		root->orphan_block_rsv = block_rsv;
	} else if (block_rsv) {
		btrfs_free_block_rsv(root, block_rsv);
		block_rsv = NULL;
2210 2211
	}

2212 2213
	if (!test_and_set_bit(BTRFS_INODE_HAS_ORPHAN_ITEM,
			      &BTRFS_I(inode)->runtime_flags)) {
2214 2215 2216 2217 2218 2219 2220 2221 2222 2223 2224 2225
#if 0
		/*
		 * For proper ENOSPC handling, we should do orphan
		 * cleanup when mounting. But this introduces backward
		 * compatibility issue.
		 */
		if (!xchg(&root->orphan_item_inserted, 1))
			insert = 2;
		else
			insert = 1;
#endif
		insert = 1;
2226
		atomic_dec(&root->orphan_inodes);
2227 2228
	}

2229 2230
	if (!test_and_set_bit(BTRFS_INODE_ORPHAN_META_RESERVED,
			      &BTRFS_I(inode)->runtime_flags))
2231 2232
		reserve = 1;
	spin_unlock(&root->orphan_lock);
2233

2234 2235 2236
	/* grab metadata reservation from transaction handle */
	if (reserve) {
		ret = btrfs_orphan_reserve_metadata(trans, inode);
2237
		BUG_ON(ret); /* -ENOSPC in reservation; Logic error? JDM */
2238
	}
2239

2240 2241
	/* insert an orphan item to track this unlinked/truncated file */
	if (insert >= 1) {
L
Li Zefan 已提交
2242
		ret = btrfs_insert_orphan_item(trans, root, btrfs_ino(inode));
2243
		if (ret && ret != -EEXIST) {
2244 2245
			clear_bit(BTRFS_INODE_HAS_ORPHAN_ITEM,
				  &BTRFS_I(inode)->runtime_flags);
2246 2247 2248 2249
			btrfs_abort_transaction(trans, root, ret);
			return ret;
		}
		ret = 0;
2250 2251 2252 2253 2254 2255
	}

	/* insert an orphan item to track subvolume contains orphan files */
	if (insert >= 2) {
		ret = btrfs_insert_orphan_item(trans, root->fs_info->tree_root,
					       root->root_key.objectid);
2256 2257 2258 2259
		if (ret && ret != -EEXIST) {
			btrfs_abort_transaction(trans, root, ret);
			return ret;
		}
2260 2261
	}
	return 0;
2262 2263 2264 2265 2266 2267 2268 2269 2270
}

/*
 * We have done the truncate/delete so we can go ahead and remove the orphan
 * item for this particular inode.
 */
int btrfs_orphan_del(struct btrfs_trans_handle *trans, struct inode *inode)
{
	struct btrfs_root *root = BTRFS_I(inode)->root;
2271 2272
	int delete_item = 0;
	int release_rsv = 0;
2273 2274
	int ret = 0;

2275
	spin_lock(&root->orphan_lock);
2276 2277
	if (test_and_clear_bit(BTRFS_INODE_HAS_ORPHAN_ITEM,
			       &BTRFS_I(inode)->runtime_flags))
2278
		delete_item = 1;
2279

2280 2281
	if (test_and_clear_bit(BTRFS_INODE_ORPHAN_META_RESERVED,
			       &BTRFS_I(inode)->runtime_flags))
2282 2283
		release_rsv = 1;
	spin_unlock(&root->orphan_lock);
2284

2285
	if (trans && delete_item) {
L
Li Zefan 已提交
2286
		ret = btrfs_del_orphan_item(trans, root, btrfs_ino(inode));
2287
		BUG_ON(ret); /* -ENOMEM or corruption (JDM: Recheck) */
2288
	}
2289

2290
	if (release_rsv) {
2291
		btrfs_orphan_release_metadata(inode);
2292 2293
		atomic_dec(&root->orphan_inodes);
	}
2294

2295
	return 0;
2296 2297 2298 2299 2300 2301
}

/*
 * this cleans up any orphans that may be left on the list from the last use
 * of this root.
 */
2302
int btrfs_orphan_cleanup(struct btrfs_root *root)
2303 2304 2305 2306 2307 2308
{
	struct btrfs_path *path;
	struct extent_buffer *leaf;
	struct btrfs_key key, found_key;
	struct btrfs_trans_handle *trans;
	struct inode *inode;
2309
	u64 last_objectid = 0;
2310 2311
	int ret = 0, nr_unlink = 0, nr_truncate = 0;

2312
	if (cmpxchg(&root->orphan_cleanup_state, 0, ORPHAN_CLEANUP_STARTED))
2313
		return 0;
2314 2315

	path = btrfs_alloc_path();
2316 2317 2318 2319
	if (!path) {
		ret = -ENOMEM;
		goto out;
	}
2320 2321 2322 2323 2324 2325 2326 2327
	path->reada = -1;

	key.objectid = BTRFS_ORPHAN_OBJECTID;
	btrfs_set_key_type(&key, BTRFS_ORPHAN_ITEM_KEY);
	key.offset = (u64)-1;

	while (1) {
		ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
2328 2329
		if (ret < 0)
			goto out;
2330 2331 2332

		/*
		 * if ret == 0 means we found what we were searching for, which
L
Lucas De Marchi 已提交
2333
		 * is weird, but possible, so only screw with path if we didn't
2334 2335 2336
		 * find the key and see if we have stuff that matches
		 */
		if (ret > 0) {
2337
			ret = 0;
2338 2339 2340 2341 2342 2343 2344 2345 2346 2347 2348 2349 2350 2351 2352 2353
			if (path->slots[0] == 0)
				break;
			path->slots[0]--;
		}

		/* pull out the item */
		leaf = path->nodes[0];
		btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);

		/* make sure the item matches what we want */
		if (found_key.objectid != BTRFS_ORPHAN_OBJECTID)
			break;
		if (btrfs_key_type(&found_key) != BTRFS_ORPHAN_ITEM_KEY)
			break;

		/* release the path since we're done with it */
2354
		btrfs_release_path(path);
2355 2356 2357 2358 2359 2360

		/*
		 * this is where we are basically btrfs_lookup, without the
		 * crossing root thing.  we store the inode number in the
		 * offset of the orphan item.
		 */
2361 2362 2363 2364 2365 2366 2367 2368 2369 2370

		if (found_key.offset == last_objectid) {
			printk(KERN_ERR "btrfs: Error removing orphan entry, "
			       "stopping orphan cleanup\n");
			ret = -EINVAL;
			goto out;
		}

		last_objectid = found_key.offset;

2371 2372 2373
		found_key.objectid = found_key.offset;
		found_key.type = BTRFS_INODE_ITEM_KEY;
		found_key.offset = 0;
2374
		inode = btrfs_iget(root->fs_info->sb, &found_key, root, NULL);
J
Josef Bacik 已提交
2375 2376
		ret = PTR_RET(inode);
		if (ret && ret != -ESTALE)
2377
			goto out;
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
		if (ret == -ESTALE && root == root->fs_info->tree_root) {
			struct btrfs_root *dead_root;
			struct btrfs_fs_info *fs_info = root->fs_info;
			int is_dead_root = 0;

			/*
			 * this is an orphan in the tree root. Currently these
			 * could come from 2 sources:
			 *  a) a snapshot deletion in progress
			 *  b) a free space cache inode
			 * We need to distinguish those two, as the snapshot
			 * orphan must not get deleted.
			 * find_dead_roots already ran before us, so if this
			 * is a snapshot deletion, we should find the root
			 * in the dead_roots list
			 */
			spin_lock(&fs_info->trans_lock);
			list_for_each_entry(dead_root, &fs_info->dead_roots,
					    root_list) {
				if (dead_root->root_key.objectid ==
				    found_key.objectid) {
					is_dead_root = 1;
					break;
				}
			}
			spin_unlock(&fs_info->trans_lock);
			if (is_dead_root) {
				/* prevent this orphan from being found again */
				key.offset = found_key.objectid - 1;
				continue;
			}
		}
2411
		/*
J
Josef Bacik 已提交
2412 2413
		 * Inode is already gone but the orphan item is still there,
		 * kill the orphan item.
2414
		 */
J
Josef Bacik 已提交
2415 2416
		if (ret == -ESTALE) {
			trans = btrfs_start_transaction(root, 1);
2417 2418 2419 2420
			if (IS_ERR(trans)) {
				ret = PTR_ERR(trans);
				goto out;
			}
2421 2422
			printk(KERN_ERR "auto deleting %Lu\n",
			       found_key.objectid);
J
Josef Bacik 已提交
2423 2424
			ret = btrfs_del_orphan_item(trans, root,
						    found_key.objectid);
2425
			BUG_ON(ret); /* -ENOMEM or corruption (JDM: Recheck) */
2426
			btrfs_end_transaction(trans, root);
2427 2428 2429
			continue;
		}

J
Josef Bacik 已提交
2430 2431 2432 2433
		/*
		 * add this inode to the orphan list so btrfs_orphan_del does
		 * the proper thing when we hit it
		 */
2434 2435
		set_bit(BTRFS_INODE_HAS_ORPHAN_ITEM,
			&BTRFS_I(inode)->runtime_flags);
J
Josef Bacik 已提交
2436

2437 2438
		/* if we have links, this was a truncate, lets do that */
		if (inode->i_nlink) {
2439 2440 2441 2442 2443
			if (!S_ISREG(inode->i_mode)) {
				WARN_ON(1);
				iput(inode);
				continue;
			}
2444
			nr_truncate++;
2445
			ret = btrfs_truncate(inode);
2446 2447 2448 2449 2450 2451
		} else {
			nr_unlink++;
		}

		/* this will do delete_inode and everything for us */
		iput(inode);
2452 2453
		if (ret)
			goto out;
2454
	}
2455 2456 2457
	/* release the path since we're done with it */
	btrfs_release_path(path);

2458 2459 2460 2461 2462 2463 2464
	root->orphan_cleanup_state = ORPHAN_CLEANUP_DONE;

	if (root->orphan_block_rsv)
		btrfs_block_rsv_release(root, root->orphan_block_rsv,
					(u64)-1);

	if (root->orphan_block_rsv || root->orphan_item_inserted) {
2465
		trans = btrfs_join_transaction(root);
2466 2467
		if (!IS_ERR(trans))
			btrfs_end_transaction(trans, root);
2468
	}
2469 2470 2471 2472 2473

	if (nr_unlink)
		printk(KERN_INFO "btrfs: unlinked %d orphans\n", nr_unlink);
	if (nr_truncate)
		printk(KERN_INFO "btrfs: truncated %d orphans\n", nr_truncate);
2474 2475 2476 2477 2478 2479

out:
	if (ret)
		printk(KERN_CRIT "btrfs: could not do orphan cleanup %d\n", ret);
	btrfs_free_path(path);
	return ret;
2480 2481
}

2482 2483 2484 2485 2486 2487 2488 2489 2490 2491 2492 2493 2494 2495 2496 2497 2498 2499 2500 2501 2502 2503 2504 2505 2506 2507 2508 2509 2510 2511 2512 2513 2514 2515 2516 2517 2518 2519 2520 2521 2522 2523 2524 2525 2526 2527 2528 2529 2530 2531 2532
/*
 * very simple check to peek ahead in the leaf looking for xattrs.  If we
 * don't find any xattrs, we know there can't be any acls.
 *
 * slot is the slot the inode is in, objectid is the objectid of the inode
 */
static noinline int acls_after_inode_item(struct extent_buffer *leaf,
					  int slot, u64 objectid)
{
	u32 nritems = btrfs_header_nritems(leaf);
	struct btrfs_key found_key;
	int scanned = 0;

	slot++;
	while (slot < nritems) {
		btrfs_item_key_to_cpu(leaf, &found_key, slot);

		/* we found a different objectid, there must not be acls */
		if (found_key.objectid != objectid)
			return 0;

		/* we found an xattr, assume we've got an acl */
		if (found_key.type == BTRFS_XATTR_ITEM_KEY)
			return 1;

		/*
		 * we found a key greater than an xattr key, there can't
		 * be any acls later on
		 */
		if (found_key.type > BTRFS_XATTR_ITEM_KEY)
			return 0;

		slot++;
		scanned++;

		/*
		 * it goes inode, inode backrefs, xattrs, extents,
		 * so if there are a ton of hard links to an inode there can
		 * be a lot of backrefs.  Don't waste time searching too hard,
		 * this is just an optimization
		 */
		if (scanned >= 8)
			break;
	}
	/* we hit the end of the leaf before we found an xattr or
	 * something larger than an xattr.  We have to assume the inode
	 * has acls
	 */
	return 1;
}

C
Chris Mason 已提交
2533 2534 2535
/*
 * read an inode from the btree into the in-memory inode
 */
2536
static void btrfs_read_locked_inode(struct inode *inode)
C
Chris Mason 已提交
2537 2538
{
	struct btrfs_path *path;
2539
	struct extent_buffer *leaf;
C
Chris Mason 已提交
2540
	struct btrfs_inode_item *inode_item;
2541
	struct btrfs_timespec *tspec;
C
Chris Mason 已提交
2542 2543
	struct btrfs_root *root = BTRFS_I(inode)->root;
	struct btrfs_key location;
2544
	int maybe_acls;
J
Josef Bacik 已提交
2545
	u32 rdev;
C
Chris Mason 已提交
2546
	int ret;
2547 2548 2549 2550 2551
	bool filled = false;

	ret = btrfs_fill_inode(inode, &rdev);
	if (!ret)
		filled = true;
C
Chris Mason 已提交
2552 2553

	path = btrfs_alloc_path();
2554 2555 2556
	if (!path)
		goto make_bad;

2557
	path->leave_spinning = 1;
C
Chris Mason 已提交
2558
	memcpy(&location, &BTRFS_I(inode)->location, sizeof(location));
C
Chris Mason 已提交
2559

C
Chris Mason 已提交
2560
	ret = btrfs_lookup_inode(NULL, root, path, &location, 0);
2561
	if (ret)
C
Chris Mason 已提交
2562 2563
		goto make_bad;

2564
	leaf = path->nodes[0];
2565 2566 2567 2568

	if (filled)
		goto cache_acl;

2569 2570 2571
	inode_item = btrfs_item_ptr(leaf, path->slots[0],
				    struct btrfs_inode_item);
	inode->i_mode = btrfs_inode_mode(leaf, inode_item);
M
Miklos Szeredi 已提交
2572
	set_nlink(inode, btrfs_inode_nlink(leaf, inode_item));
2573 2574
	inode->i_uid = btrfs_inode_uid(leaf, inode_item);
	inode->i_gid = btrfs_inode_gid(leaf, inode_item);
2575
	btrfs_i_size_write(inode, btrfs_inode_size(leaf, inode_item));
2576 2577 2578 2579 2580 2581 2582 2583 2584 2585 2586 2587 2588

	tspec = btrfs_inode_atime(inode_item);
	inode->i_atime.tv_sec = btrfs_timespec_sec(leaf, tspec);
	inode->i_atime.tv_nsec = btrfs_timespec_nsec(leaf, tspec);

	tspec = btrfs_inode_mtime(inode_item);
	inode->i_mtime.tv_sec = btrfs_timespec_sec(leaf, tspec);
	inode->i_mtime.tv_nsec = btrfs_timespec_nsec(leaf, tspec);

	tspec = btrfs_inode_ctime(inode_item);
	inode->i_ctime.tv_sec = btrfs_timespec_sec(leaf, tspec);
	inode->i_ctime.tv_nsec = btrfs_timespec_nsec(leaf, tspec);

2589
	inode_set_bytes(inode, btrfs_inode_nbytes(leaf, inode_item));
2590
	BTRFS_I(inode)->generation = btrfs_inode_generation(leaf, inode_item);
2591
	inode->i_version = btrfs_inode_sequence(leaf, inode_item);
2592
	inode->i_generation = BTRFS_I(inode)->generation;
J
Josef Bacik 已提交
2593
	inode->i_rdev = 0;
2594 2595
	rdev = btrfs_inode_rdev(leaf, inode_item);

2596
	BTRFS_I(inode)->index_cnt = (u64)-1;
2597
	BTRFS_I(inode)->flags = btrfs_inode_flags(leaf, inode_item);
2598
cache_acl:
2599 2600 2601 2602
	/*
	 * try to precache a NULL acl entry for files that don't have
	 * any xattrs or acls
	 */
L
Li Zefan 已提交
2603 2604
	maybe_acls = acls_after_inode_item(leaf, path->slots[0],
					   btrfs_ino(inode));
2605 2606
	if (!maybe_acls)
		cache_no_acl(inode);
2607

C
Chris Mason 已提交
2608 2609 2610 2611 2612
	btrfs_free_path(path);

	switch (inode->i_mode & S_IFMT) {
	case S_IFREG:
		inode->i_mapping->a_ops = &btrfs_aops;
C
Chris Mason 已提交
2613
		inode->i_mapping->backing_dev_info = &root->fs_info->bdi;
2614
		BTRFS_I(inode)->io_tree.ops = &btrfs_extent_io_ops;
C
Chris Mason 已提交
2615 2616 2617 2618 2619 2620 2621 2622 2623 2624 2625 2626 2627
		inode->i_fop = &btrfs_file_operations;
		inode->i_op = &btrfs_file_inode_operations;
		break;
	case S_IFDIR:
		inode->i_fop = &btrfs_dir_file_operations;
		if (root == root->fs_info->tree_root)
			inode->i_op = &btrfs_dir_ro_inode_operations;
		else
			inode->i_op = &btrfs_dir_inode_operations;
		break;
	case S_IFLNK:
		inode->i_op = &btrfs_symlink_inode_operations;
		inode->i_mapping->a_ops = &btrfs_symlink_aops;
C
Chris Mason 已提交
2628
		inode->i_mapping->backing_dev_info = &root->fs_info->bdi;
C
Chris Mason 已提交
2629
		break;
J
Josef Bacik 已提交
2630
	default:
J
Jim Owens 已提交
2631
		inode->i_op = &btrfs_special_inode_operations;
J
Josef Bacik 已提交
2632 2633
		init_special_inode(inode, inode->i_mode, rdev);
		break;
C
Chris Mason 已提交
2634
	}
2635 2636

	btrfs_update_iflags(inode);
C
Chris Mason 已提交
2637 2638 2639 2640 2641 2642 2643
	return;

make_bad:
	btrfs_free_path(path);
	make_bad_inode(inode);
}

C
Chris Mason 已提交
2644 2645 2646
/*
 * given a leaf and an inode, copy the inode fields into the leaf
 */
2647 2648
static void fill_inode_item(struct btrfs_trans_handle *trans,
			    struct extent_buffer *leaf,
2649
			    struct btrfs_inode_item *item,
C
Chris Mason 已提交
2650 2651
			    struct inode *inode)
{
2652 2653
	btrfs_set_inode_uid(leaf, item, inode->i_uid);
	btrfs_set_inode_gid(leaf, item, inode->i_gid);
2654
	btrfs_set_inode_size(leaf, item, BTRFS_I(inode)->disk_i_size);
2655 2656 2657 2658 2659 2660 2661 2662 2663 2664 2665 2666 2667 2668 2669 2670 2671 2672
	btrfs_set_inode_mode(leaf, item, inode->i_mode);
	btrfs_set_inode_nlink(leaf, item, inode->i_nlink);

	btrfs_set_timespec_sec(leaf, btrfs_inode_atime(item),
			       inode->i_atime.tv_sec);
	btrfs_set_timespec_nsec(leaf, btrfs_inode_atime(item),
				inode->i_atime.tv_nsec);

	btrfs_set_timespec_sec(leaf, btrfs_inode_mtime(item),
			       inode->i_mtime.tv_sec);
	btrfs_set_timespec_nsec(leaf, btrfs_inode_mtime(item),
				inode->i_mtime.tv_nsec);

	btrfs_set_timespec_sec(leaf, btrfs_inode_ctime(item),
			       inode->i_ctime.tv_sec);
	btrfs_set_timespec_nsec(leaf, btrfs_inode_ctime(item),
				inode->i_ctime.tv_nsec);

2673
	btrfs_set_inode_nbytes(leaf, item, inode_get_bytes(inode));
2674
	btrfs_set_inode_generation(leaf, item, BTRFS_I(inode)->generation);
2675
	btrfs_set_inode_sequence(leaf, item, inode->i_version);
2676
	btrfs_set_inode_transid(leaf, item, trans->transid);
2677
	btrfs_set_inode_rdev(leaf, item, inode->i_rdev);
Y
Yan 已提交
2678
	btrfs_set_inode_flags(leaf, item, BTRFS_I(inode)->flags);
2679
	btrfs_set_inode_block_group(leaf, item, 0);
C
Chris Mason 已提交
2680 2681
}

C
Chris Mason 已提交
2682 2683 2684
/*
 * copy everything in the in-memory inode into the btree.
 */
2685
static noinline int btrfs_update_inode_item(struct btrfs_trans_handle *trans,
C
Chris Mason 已提交
2686
				struct btrfs_root *root, struct inode *inode)
C
Chris Mason 已提交
2687 2688 2689
{
	struct btrfs_inode_item *inode_item;
	struct btrfs_path *path;
2690
	struct extent_buffer *leaf;
C
Chris Mason 已提交
2691 2692 2693
	int ret;

	path = btrfs_alloc_path();
2694 2695 2696
	if (!path)
		return -ENOMEM;

2697
	path->leave_spinning = 1;
2698 2699
	ret = btrfs_lookup_inode(trans, root, path, &BTRFS_I(inode)->location,
				 1);
C
Chris Mason 已提交
2700 2701 2702 2703 2704 2705
	if (ret) {
		if (ret > 0)
			ret = -ENOENT;
		goto failed;
	}

2706
	btrfs_unlock_up_safe(path, 1);
2707 2708
	leaf = path->nodes[0];
	inode_item = btrfs_item_ptr(leaf, path->slots[0],
2709
				    struct btrfs_inode_item);
C
Chris Mason 已提交
2710

2711
	fill_inode_item(trans, leaf, inode_item, inode);
2712
	btrfs_mark_buffer_dirty(leaf);
2713
	btrfs_set_inode_last_trans(trans, inode);
C
Chris Mason 已提交
2714 2715 2716 2717 2718 2719
	ret = 0;
failed:
	btrfs_free_path(path);
	return ret;
}

2720 2721 2722 2723 2724 2725 2726 2727 2728 2729 2730 2731 2732 2733 2734 2735 2736 2737 2738 2739 2740 2741 2742 2743 2744 2745 2746 2747 2748 2749 2750 2751 2752 2753 2754 2755 2756
/*
 * copy everything in the in-memory inode into the btree.
 */
noinline int btrfs_update_inode(struct btrfs_trans_handle *trans,
				struct btrfs_root *root, struct inode *inode)
{
	int ret;

	/*
	 * If the inode is a free space inode, we can deadlock during commit
	 * if we put it into the delayed code.
	 *
	 * The data relocation inode should also be directly updated
	 * without delay
	 */
	if (!btrfs_is_free_space_inode(root, inode)
	    && root->root_key.objectid != BTRFS_DATA_RELOC_TREE_OBJECTID) {
		ret = btrfs_delayed_update_inode(trans, root, inode);
		if (!ret)
			btrfs_set_inode_last_trans(trans, inode);
		return ret;
	}

	return btrfs_update_inode_item(trans, root, inode);
}

static noinline int btrfs_update_inode_fallback(struct btrfs_trans_handle *trans,
				struct btrfs_root *root, struct inode *inode)
{
	int ret;

	ret = btrfs_update_inode(trans, root, inode);
	if (ret == -ENOSPC)
		return btrfs_update_inode_item(trans, root, inode);
	return ret;
}

C
Chris Mason 已提交
2757 2758 2759 2760 2761
/*
 * unlink helper that gets used here in inode.c and in the tree logging
 * recovery code.  It remove a link in a directory with a given name, and
 * also drops the back refs in the inode to the directory
 */
2762 2763 2764 2765
static int __btrfs_unlink_inode(struct btrfs_trans_handle *trans,
				struct btrfs_root *root,
				struct inode *dir, struct inode *inode,
				const char *name, int name_len)
C
Chris Mason 已提交
2766 2767 2768
{
	struct btrfs_path *path;
	int ret = 0;
2769
	struct extent_buffer *leaf;
C
Chris Mason 已提交
2770
	struct btrfs_dir_item *di;
2771
	struct btrfs_key key;
2772
	u64 index;
L
Li Zefan 已提交
2773 2774
	u64 ino = btrfs_ino(inode);
	u64 dir_ino = btrfs_ino(dir);
C
Chris Mason 已提交
2775 2776

	path = btrfs_alloc_path();
2777 2778
	if (!path) {
		ret = -ENOMEM;
2779
		goto out;
2780 2781
	}

2782
	path->leave_spinning = 1;
L
Li Zefan 已提交
2783
	di = btrfs_lookup_dir_item(trans, root, path, dir_ino,
C
Chris Mason 已提交
2784 2785 2786 2787 2788 2789 2790 2791 2792
				    name, name_len, -1);
	if (IS_ERR(di)) {
		ret = PTR_ERR(di);
		goto err;
	}
	if (!di) {
		ret = -ENOENT;
		goto err;
	}
2793 2794
	leaf = path->nodes[0];
	btrfs_dir_item_key_to_cpu(leaf, di, &key);
C
Chris Mason 已提交
2795
	ret = btrfs_delete_one_dir_name(trans, root, path, di);
2796 2797
	if (ret)
		goto err;
2798
	btrfs_release_path(path);
C
Chris Mason 已提交
2799

L
Li Zefan 已提交
2800 2801
	ret = btrfs_del_inode_ref(trans, root, name, name_len, ino,
				  dir_ino, &index);
2802
	if (ret) {
C
Chris Mason 已提交
2803
		printk(KERN_INFO "btrfs failed to delete reference to %.*s, "
L
Li Zefan 已提交
2804 2805
		       "inode %llu parent %llu\n", name_len, name,
		       (unsigned long long)ino, (unsigned long long)dir_ino);
2806
		btrfs_abort_transaction(trans, root, ret);
2807 2808 2809
		goto err;
	}

2810
	ret = btrfs_delete_delayed_dir_index(trans, root, dir, index);
2811 2812
	if (ret) {
		btrfs_abort_transaction(trans, root, ret);
C
Chris Mason 已提交
2813
		goto err;
2814
	}
C
Chris Mason 已提交
2815

2816
	ret = btrfs_del_inode_ref_in_log(trans, root, name, name_len,
L
Li Zefan 已提交
2817
					 inode, dir_ino);
2818 2819 2820 2821
	if (ret != 0 && ret != -ENOENT) {
		btrfs_abort_transaction(trans, root, ret);
		goto err;
	}
2822 2823 2824

	ret = btrfs_del_dir_entries_in_log(trans, root, name, name_len,
					   dir, index);
2825 2826
	if (ret == -ENOENT)
		ret = 0;
C
Chris Mason 已提交
2827 2828
err:
	btrfs_free_path(path);
2829 2830 2831 2832
	if (ret)
		goto out;

	btrfs_i_size_write(dir, dir->i_size - name_len * 2);
2833 2834
	inode_inc_iversion(inode);
	inode_inc_iversion(dir);
2835 2836 2837
	inode->i_ctime = dir->i_mtime = dir->i_ctime = CURRENT_TIME;
	btrfs_update_inode(trans, root, dir);
out:
C
Chris Mason 已提交
2838 2839 2840
	return ret;
}

2841 2842 2843 2844 2845 2846 2847 2848 2849 2850 2851 2852 2853 2854 2855
int btrfs_unlink_inode(struct btrfs_trans_handle *trans,
		       struct btrfs_root *root,
		       struct inode *dir, struct inode *inode,
		       const char *name, int name_len)
{
	int ret;
	ret = __btrfs_unlink_inode(trans, root, dir, inode, name, name_len);
	if (!ret) {
		btrfs_drop_nlink(inode);
		ret = btrfs_update_inode(trans, root, inode);
	}
	return ret;
}
		

2856 2857 2858
/* helper to check if there is any shared block in the path */
static int check_path_shared(struct btrfs_root *root,
			     struct btrfs_path *path)
C
Chris Mason 已提交
2859
{
2860 2861
	struct extent_buffer *eb;
	int level;
2862
	u64 refs = 1;
2863

2864
	for (level = 0; level < BTRFS_MAX_LEVEL; level++) {
2865 2866
		int ret;

2867 2868 2869 2870 2871 2872 2873 2874 2875
		if (!path->nodes[level])
			break;
		eb = path->nodes[level];
		if (!btrfs_block_can_be_shared(root, eb))
			continue;
		ret = btrfs_lookup_extent_info(NULL, root, eb->start, eb->len,
					       &refs, NULL);
		if (refs > 1)
			return 1;
2876
	}
2877
	return 0;
C
Chris Mason 已提交
2878 2879
}

2880 2881 2882 2883 2884 2885 2886 2887 2888
/*
 * helper to start transaction for unlink and rmdir.
 *
 * unlink and rmdir are special in btrfs, they do not always free space.
 * so in enospc case, we should make sure they will free space before
 * allowing them to use the global metadata reservation.
 */
static struct btrfs_trans_handle *__unlink_start_trans(struct inode *dir,
						       struct dentry *dentry)
2889
{
C
Chris Mason 已提交
2890
	struct btrfs_trans_handle *trans;
2891
	struct btrfs_root *root = BTRFS_I(dir)->root;
2892
	struct btrfs_path *path;
2893
	struct btrfs_inode_ref *ref;
2894
	struct btrfs_dir_item *di;
2895
	struct inode *inode = dentry->d_inode;
2896
	u64 index;
2897 2898
	int check_link = 1;
	int err = -ENOSPC;
2899
	int ret;
L
Li Zefan 已提交
2900 2901
	u64 ino = btrfs_ino(inode);
	u64 dir_ino = btrfs_ino(dir);
2902

2903 2904 2905 2906 2907 2908 2909 2910 2911 2912
	/*
	 * 1 for the possible orphan item
	 * 1 for the dir item
	 * 1 for the dir index
	 * 1 for the inode ref
	 * 1 for the inode ref in the tree log
	 * 2 for the dir entries in the log
	 * 1 for the inode
	 */
	trans = btrfs_start_transaction(root, 8);
2913 2914
	if (!IS_ERR(trans) || PTR_ERR(trans) != -ENOSPC)
		return trans;
2915

L
Li Zefan 已提交
2916
	if (ino == BTRFS_EMPTY_SUBVOL_DIR_OBJECTID)
2917
		return ERR_PTR(-ENOSPC);
2918

2919 2920 2921
	/* check if there is someone else holds reference */
	if (S_ISDIR(inode->i_mode) && atomic_read(&inode->i_count) > 1)
		return ERR_PTR(-ENOSPC);
2922

2923 2924
	if (atomic_read(&inode->i_count) > 2)
		return ERR_PTR(-ENOSPC);
2925

2926 2927 2928 2929 2930 2931 2932
	if (xchg(&root->fs_info->enospc_unlink, 1))
		return ERR_PTR(-ENOSPC);

	path = btrfs_alloc_path();
	if (!path) {
		root->fs_info->enospc_unlink = 0;
		return ERR_PTR(-ENOMEM);
2933 2934
	}

2935 2936
	/* 1 for the orphan item */
	trans = btrfs_start_transaction(root, 1);
2937
	if (IS_ERR(trans)) {
2938 2939 2940 2941
		btrfs_free_path(path);
		root->fs_info->enospc_unlink = 0;
		return trans;
	}
2942

2943 2944
	path->skip_locking = 1;
	path->search_commit_root = 1;
2945

2946 2947 2948 2949 2950 2951 2952 2953 2954 2955 2956
	ret = btrfs_lookup_inode(trans, root, path,
				&BTRFS_I(dir)->location, 0);
	if (ret < 0) {
		err = ret;
		goto out;
	}
	if (ret == 0) {
		if (check_path_shared(root, path))
			goto out;
	} else {
		check_link = 0;
2957
	}
2958
	btrfs_release_path(path);
2959 2960 2961 2962 2963 2964 2965 2966 2967 2968 2969 2970 2971

	ret = btrfs_lookup_inode(trans, root, path,
				&BTRFS_I(inode)->location, 0);
	if (ret < 0) {
		err = ret;
		goto out;
	}
	if (ret == 0) {
		if (check_path_shared(root, path))
			goto out;
	} else {
		check_link = 0;
	}
2972
	btrfs_release_path(path);
2973 2974 2975

	if (ret == 0 && S_ISREG(inode->i_mode)) {
		ret = btrfs_lookup_file_extent(trans, root, path,
L
Li Zefan 已提交
2976
					       ino, (u64)-1, 0);
2977 2978 2979 2980
		if (ret < 0) {
			err = ret;
			goto out;
		}
2981
		BUG_ON(ret == 0); /* Corruption */
2982 2983
		if (check_path_shared(root, path))
			goto out;
2984
		btrfs_release_path(path);
2985 2986 2987 2988 2989 2990 2991
	}

	if (!check_link) {
		err = 0;
		goto out;
	}

L
Li Zefan 已提交
2992
	di = btrfs_lookup_dir_item(trans, root, path, dir_ino,
2993 2994 2995 2996 2997 2998 2999 3000 3001 3002 3003 3004
				dentry->d_name.name, dentry->d_name.len, 0);
	if (IS_ERR(di)) {
		err = PTR_ERR(di);
		goto out;
	}
	if (di) {
		if (check_path_shared(root, path))
			goto out;
	} else {
		err = 0;
		goto out;
	}
3005
	btrfs_release_path(path);
3006 3007 3008

	ref = btrfs_lookup_inode_ref(trans, root, path,
				dentry->d_name.name, dentry->d_name.len,
L
Li Zefan 已提交
3009
				ino, dir_ino, 0);
3010 3011 3012 3013
	if (IS_ERR(ref)) {
		err = PTR_ERR(ref);
		goto out;
	}
3014
	BUG_ON(!ref); /* Logic error */
3015 3016 3017
	if (check_path_shared(root, path))
		goto out;
	index = btrfs_inode_ref_index(path->nodes[0], ref);
3018
	btrfs_release_path(path);
3019

3020 3021 3022 3023 3024 3025 3026 3027
	/*
	 * This is a commit root search, if we can lookup inode item and other
	 * relative items in the commit root, it means the transaction of
	 * dir/file creation has been committed, and the dir index item that we
	 * delay to insert has also been inserted into the commit root. So
	 * we needn't worry about the delayed insertion of the dir index item
	 * here.
	 */
L
Li Zefan 已提交
3028
	di = btrfs_lookup_dir_index_item(trans, root, path, dir_ino, index,
3029 3030 3031 3032 3033 3034 3035 3036 3037 3038 3039 3040
				dentry->d_name.name, dentry->d_name.len, 0);
	if (IS_ERR(di)) {
		err = PTR_ERR(di);
		goto out;
	}
	BUG_ON(ret == -ENOENT);
	if (check_path_shared(root, path))
		goto out;

	err = 0;
out:
	btrfs_free_path(path);
3041 3042 3043 3044
	/* Migrate the orphan reservation over */
	if (!err)
		err = btrfs_block_rsv_migrate(trans->block_rsv,
				&root->fs_info->global_block_rsv,
3045
				trans->bytes_reserved);
3046

3047 3048 3049 3050 3051 3052 3053 3054 3055 3056 3057 3058 3059 3060
	if (err) {
		btrfs_end_transaction(trans, root);
		root->fs_info->enospc_unlink = 0;
		return ERR_PTR(err);
	}

	trans->block_rsv = &root->fs_info->global_block_rsv;
	return trans;
}

static void __unlink_end_trans(struct btrfs_trans_handle *trans,
			       struct btrfs_root *root)
{
	if (trans->block_rsv == &root->fs_info->global_block_rsv) {
3061 3062 3063
		btrfs_block_rsv_release(root, trans->block_rsv,
					trans->bytes_reserved);
		trans->block_rsv = &root->fs_info->trans_block_rsv;
3064 3065 3066
		BUG_ON(!root->fs_info->enospc_unlink);
		root->fs_info->enospc_unlink = 0;
	}
3067
	btrfs_end_transaction(trans, root);
3068 3069 3070 3071 3072 3073 3074 3075 3076 3077 3078 3079 3080
}

static int btrfs_unlink(struct inode *dir, struct dentry *dentry)
{
	struct btrfs_root *root = BTRFS_I(dir)->root;
	struct btrfs_trans_handle *trans;
	struct inode *inode = dentry->d_inode;
	int ret;
	unsigned long nr = 0;

	trans = __unlink_start_trans(dir, dentry);
	if (IS_ERR(trans))
		return PTR_ERR(trans);
3081

3082 3083
	btrfs_record_unlink_dir(trans, dir, dentry->d_inode, 0);

3084 3085
	ret = btrfs_unlink_inode(trans, root, dir, dentry->d_inode,
				 dentry->d_name.name, dentry->d_name.len);
3086 3087
	if (ret)
		goto out;
3088

3089
	if (inode->i_nlink == 0) {
3090
		ret = btrfs_orphan_add(trans, inode);
3091 3092
		if (ret)
			goto out;
3093
	}
3094

3095
out:
3096
	nr = trans->blocks_used;
3097
	__unlink_end_trans(trans, root);
3098
	btrfs_btree_balance_dirty(root, nr);
C
Chris Mason 已提交
3099 3100 3101
	return ret;
}

3102 3103 3104 3105 3106 3107 3108 3109 3110 3111 3112
int btrfs_unlink_subvol(struct btrfs_trans_handle *trans,
			struct btrfs_root *root,
			struct inode *dir, u64 objectid,
			const char *name, int name_len)
{
	struct btrfs_path *path;
	struct extent_buffer *leaf;
	struct btrfs_dir_item *di;
	struct btrfs_key key;
	u64 index;
	int ret;
L
Li Zefan 已提交
3113
	u64 dir_ino = btrfs_ino(dir);
3114 3115 3116 3117 3118

	path = btrfs_alloc_path();
	if (!path)
		return -ENOMEM;

L
Li Zefan 已提交
3119
	di = btrfs_lookup_dir_item(trans, root, path, dir_ino,
3120
				   name, name_len, -1);
3121 3122 3123 3124 3125 3126 3127
	if (IS_ERR_OR_NULL(di)) {
		if (!di)
			ret = -ENOENT;
		else
			ret = PTR_ERR(di);
		goto out;
	}
3128 3129 3130 3131 3132

	leaf = path->nodes[0];
	btrfs_dir_item_key_to_cpu(leaf, di, &key);
	WARN_ON(key.type != BTRFS_ROOT_ITEM_KEY || key.objectid != objectid);
	ret = btrfs_delete_one_dir_name(trans, root, path, di);
3133 3134 3135 3136
	if (ret) {
		btrfs_abort_transaction(trans, root, ret);
		goto out;
	}
3137
	btrfs_release_path(path);
3138 3139 3140

	ret = btrfs_del_root_ref(trans, root->fs_info->tree_root,
				 objectid, root->root_key.objectid,
L
Li Zefan 已提交
3141
				 dir_ino, &index, name, name_len);
3142
	if (ret < 0) {
3143 3144 3145 3146
		if (ret != -ENOENT) {
			btrfs_abort_transaction(trans, root, ret);
			goto out;
		}
L
Li Zefan 已提交
3147
		di = btrfs_search_dir_index_item(root, path, dir_ino,
3148
						 name, name_len);
3149 3150 3151 3152 3153 3154 3155 3156
		if (IS_ERR_OR_NULL(di)) {
			if (!di)
				ret = -ENOENT;
			else
				ret = PTR_ERR(di);
			btrfs_abort_transaction(trans, root, ret);
			goto out;
		}
3157 3158 3159

		leaf = path->nodes[0];
		btrfs_item_key_to_cpu(leaf, &key, path->slots[0]);
3160
		btrfs_release_path(path);
3161 3162
		index = key.offset;
	}
3163
	btrfs_release_path(path);
3164

3165
	ret = btrfs_delete_delayed_dir_index(trans, root, dir, index);
3166 3167 3168 3169
	if (ret) {
		btrfs_abort_transaction(trans, root, ret);
		goto out;
	}
3170 3171

	btrfs_i_size_write(dir, dir->i_size - name_len * 2);
3172
	inode_inc_iversion(dir);
3173 3174
	dir->i_mtime = dir->i_ctime = CURRENT_TIME;
	ret = btrfs_update_inode(trans, root, dir);
3175 3176 3177
	if (ret)
		btrfs_abort_transaction(trans, root, ret);
out:
3178
	btrfs_free_path(path);
3179
	return ret;
3180 3181
}

C
Chris Mason 已提交
3182 3183 3184
static int btrfs_rmdir(struct inode *dir, struct dentry *dentry)
{
	struct inode *inode = dentry->d_inode;
3185
	int err = 0;
C
Chris Mason 已提交
3186 3187
	struct btrfs_root *root = BTRFS_I(dir)->root;
	struct btrfs_trans_handle *trans;
3188
	unsigned long nr = 0;
C
Chris Mason 已提交
3189

3190
	if (inode->i_size > BTRFS_EMPTY_DIR_SIZE ||
L
Li Zefan 已提交
3191
	    btrfs_ino(inode) == BTRFS_FIRST_FREE_OBJECTID)
Y
Yan 已提交
3192 3193
		return -ENOTEMPTY;

3194 3195
	trans = __unlink_start_trans(dir, dentry);
	if (IS_ERR(trans))
3196 3197
		return PTR_ERR(trans);

L
Li Zefan 已提交
3198
	if (unlikely(btrfs_ino(inode) == BTRFS_EMPTY_SUBVOL_DIR_OBJECTID)) {
3199 3200 3201 3202 3203 3204 3205
		err = btrfs_unlink_subvol(trans, root, dir,
					  BTRFS_I(inode)->location.objectid,
					  dentry->d_name.name,
					  dentry->d_name.len);
		goto out;
	}

3206 3207
	err = btrfs_orphan_add(trans, inode);
	if (err)
3208
		goto out;
3209

C
Chris Mason 已提交
3210
	/* now the directory is empty */
3211 3212
	err = btrfs_unlink_inode(trans, root, dir, dentry->d_inode,
				 dentry->d_name.name, dentry->d_name.len);
C
Chris Mason 已提交
3213
	if (!err)
3214
		btrfs_i_size_write(inode, 0);
3215
out:
3216
	nr = trans->blocks_used;
3217
	__unlink_end_trans(trans, root);
3218
	btrfs_btree_balance_dirty(root, nr);
3219

C
Chris Mason 已提交
3220 3221 3222 3223 3224 3225
	return err;
}

/*
 * this can truncate away extent items, csum items and directory items.
 * It starts at a high offset and removes keys until it can't find
C
Chris Mason 已提交
3226
 * any higher than new_size
C
Chris Mason 已提交
3227 3228 3229
 *
 * csum items that cross the new i_size are truncated to the new size
 * as well.
3230 3231 3232
 *
 * min_type is the minimum key type to truncate down to.  If set to 0, this
 * will kill all the items on this inode, including the INODE_ITEM_KEY.
C
Chris Mason 已提交
3233
 */
3234 3235 3236 3237
int btrfs_truncate_inode_items(struct btrfs_trans_handle *trans,
			       struct btrfs_root *root,
			       struct inode *inode,
			       u64 new_size, u32 min_type)
C
Chris Mason 已提交
3238 3239
{
	struct btrfs_path *path;
3240
	struct extent_buffer *leaf;
C
Chris Mason 已提交
3241
	struct btrfs_file_extent_item *fi;
3242 3243
	struct btrfs_key key;
	struct btrfs_key found_key;
C
Chris Mason 已提交
3244
	u64 extent_start = 0;
3245
	u64 extent_num_bytes = 0;
3246
	u64 extent_offset = 0;
C
Chris Mason 已提交
3247
	u64 item_end = 0;
3248 3249
	u64 mask = root->sectorsize - 1;
	u32 found_type = (u8)-1;
C
Chris Mason 已提交
3250 3251
	int found_extent;
	int del_item;
3252 3253
	int pending_del_nr = 0;
	int pending_del_slot = 0;
3254
	int extent_type = -1;
3255 3256
	int ret;
	int err = 0;
L
Li Zefan 已提交
3257
	u64 ino = btrfs_ino(inode);
3258 3259

	BUG_ON(new_size > 0 && min_type != BTRFS_EXTENT_DATA_KEY);
C
Chris Mason 已提交
3260

3261 3262 3263 3264 3265
	path = btrfs_alloc_path();
	if (!path)
		return -ENOMEM;
	path->reada = -1;

3266
	if (root->ref_cows || root == root->fs_info->tree_root)
3267
		btrfs_drop_extent_cache(inode, new_size & (~mask), (u64)-1, 0);
3268

3269 3270 3271 3272 3273 3274 3275 3276 3277
	/*
	 * This function is also used to drop the items in the log tree before
	 * we relog the inode, so if root != BTRFS_I(inode)->root, it means
	 * it is used to drop the loged items. So we shouldn't kill the delayed
	 * items.
	 */
	if (min_type == 0 && root == BTRFS_I(inode)->root)
		btrfs_kill_delayed_inode_items(inode);

L
Li Zefan 已提交
3278
	key.objectid = ino;
C
Chris Mason 已提交
3279
	key.offset = (u64)-1;
3280 3281
	key.type = (u8)-1;

3282
search_again:
3283
	path->leave_spinning = 1;
3284
	ret = btrfs_search_slot(trans, root, &key, path, -1, 1);
3285 3286 3287 3288
	if (ret < 0) {
		err = ret;
		goto out;
	}
C
Chris Mason 已提交
3289

3290
	if (ret > 0) {
3291 3292 3293
		/* there are no items in the tree for us to truncate, we're
		 * done
		 */
3294 3295
		if (path->slots[0] == 0)
			goto out;
3296 3297 3298
		path->slots[0]--;
	}

C
Chris Mason 已提交
3299
	while (1) {
C
Chris Mason 已提交
3300
		fi = NULL;
3301 3302 3303
		leaf = path->nodes[0];
		btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
		found_type = btrfs_key_type(&found_key);
C
Chris Mason 已提交
3304

L
Li Zefan 已提交
3305
		if (found_key.objectid != ino)
C
Chris Mason 已提交
3306
			break;
3307

3308
		if (found_type < min_type)
C
Chris Mason 已提交
3309 3310
			break;

3311
		item_end = found_key.offset;
C
Chris Mason 已提交
3312
		if (found_type == BTRFS_EXTENT_DATA_KEY) {
3313
			fi = btrfs_item_ptr(leaf, path->slots[0],
C
Chris Mason 已提交
3314
					    struct btrfs_file_extent_item);
3315 3316
			extent_type = btrfs_file_extent_type(leaf, fi);
			if (extent_type != BTRFS_FILE_EXTENT_INLINE) {
3317
				item_end +=
3318
				    btrfs_file_extent_num_bytes(leaf, fi);
3319 3320
			} else if (extent_type == BTRFS_FILE_EXTENT_INLINE) {
				item_end += btrfs_file_extent_inline_len(leaf,
C
Chris Mason 已提交
3321
									 fi);
C
Chris Mason 已提交
3322
			}
3323
			item_end--;
C
Chris Mason 已提交
3324
		}
3325 3326 3327 3328
		if (found_type > min_type) {
			del_item = 1;
		} else {
			if (item_end < new_size)
3329
				break;
3330 3331 3332 3333
			if (found_key.offset >= new_size)
				del_item = 1;
			else
				del_item = 0;
C
Chris Mason 已提交
3334 3335 3336
		}
		found_extent = 0;
		/* FIXME, shrink the extent if the ref count is only 1 */
3337 3338 3339 3340
		if (found_type != BTRFS_EXTENT_DATA_KEY)
			goto delete;

		if (extent_type != BTRFS_FILE_EXTENT_INLINE) {
C
Chris Mason 已提交
3341
			u64 num_dec;
3342
			extent_start = btrfs_file_extent_disk_bytenr(leaf, fi);
3343
			if (!del_item) {
3344 3345
				u64 orig_num_bytes =
					btrfs_file_extent_num_bytes(leaf, fi);
3346
				extent_num_bytes = new_size -
3347
					found_key.offset + root->sectorsize - 1;
3348 3349
				extent_num_bytes = extent_num_bytes &
					~((u64)root->sectorsize - 1);
3350 3351 3352
				btrfs_set_file_extent_num_bytes(leaf, fi,
							 extent_num_bytes);
				num_dec = (orig_num_bytes -
C
Chris Mason 已提交
3353
					   extent_num_bytes);
3354
				if (root->ref_cows && extent_start != 0)
3355
					inode_sub_bytes(inode, num_dec);
3356
				btrfs_mark_buffer_dirty(leaf);
C
Chris Mason 已提交
3357
			} else {
3358 3359 3360
				extent_num_bytes =
					btrfs_file_extent_disk_num_bytes(leaf,
									 fi);
3361 3362 3363
				extent_offset = found_key.offset -
					btrfs_file_extent_offset(leaf, fi);

C
Chris Mason 已提交
3364
				/* FIXME blocksize != 4096 */
C
Chris Mason 已提交
3365
				num_dec = btrfs_file_extent_num_bytes(leaf, fi);
C
Chris Mason 已提交
3366 3367
				if (extent_start != 0) {
					found_extent = 1;
3368
					if (root->ref_cows)
3369
						inode_sub_bytes(inode, num_dec);
3370
				}
C
Chris Mason 已提交
3371
			}
C
Chris Mason 已提交
3372
		} else if (extent_type == BTRFS_FILE_EXTENT_INLINE) {
C
Chris Mason 已提交
3373 3374 3375 3376 3377 3378 3379 3380
			/*
			 * we can't truncate inline items that have had
			 * special encodings
			 */
			if (!del_item &&
			    btrfs_file_extent_compression(leaf, fi) == 0 &&
			    btrfs_file_extent_encryption(leaf, fi) == 0 &&
			    btrfs_file_extent_other_encoding(leaf, fi) == 0) {
3381 3382 3383
				u32 size = new_size - found_key.offset;

				if (root->ref_cows) {
3384 3385
					inode_sub_bytes(inode, item_end + 1 -
							new_size);
3386 3387 3388
				}
				size =
				    btrfs_file_extent_calc_inline_size(size);
3389 3390
				btrfs_truncate_item(trans, root, path,
						    size, 1);
3391
			} else if (root->ref_cows) {
3392 3393
				inode_sub_bytes(inode, item_end + 1 -
						found_key.offset);
C
Chris Mason 已提交
3394
			}
C
Chris Mason 已提交
3395
		}
3396
delete:
C
Chris Mason 已提交
3397
		if (del_item) {
3398 3399 3400 3401 3402 3403 3404 3405 3406 3407
			if (!pending_del_nr) {
				/* no pending yet, add ourselves */
				pending_del_slot = path->slots[0];
				pending_del_nr = 1;
			} else if (pending_del_nr &&
				   path->slots[0] + 1 == pending_del_slot) {
				/* hop on the pending chunk */
				pending_del_nr++;
				pending_del_slot = path->slots[0];
			} else {
C
Chris Mason 已提交
3408
				BUG();
3409
			}
C
Chris Mason 已提交
3410 3411 3412
		} else {
			break;
		}
3413 3414
		if (found_extent && (root->ref_cows ||
				     root == root->fs_info->tree_root)) {
3415
			btrfs_set_path_blocking(path);
C
Chris Mason 已提交
3416
			ret = btrfs_free_extent(trans, root, extent_start,
3417 3418
						extent_num_bytes, 0,
						btrfs_header_owner(leaf),
A
Arne Jansen 已提交
3419
						ino, extent_offset, 0);
C
Chris Mason 已提交
3420 3421
			BUG_ON(ret);
		}
3422

3423 3424 3425 3426 3427
		if (found_type == BTRFS_INODE_ITEM_KEY)
			break;

		if (path->slots[0] == 0 ||
		    path->slots[0] != pending_del_slot) {
3428 3429 3430
			if (root->ref_cows &&
			    BTRFS_I(inode)->location.objectid !=
						BTRFS_FREE_INO_OBJECTID) {
3431 3432 3433 3434 3435 3436 3437
				err = -EAGAIN;
				goto out;
			}
			if (pending_del_nr) {
				ret = btrfs_del_items(trans, root, path,
						pending_del_slot,
						pending_del_nr);
3438 3439 3440 3441 3442
				if (ret) {
					btrfs_abort_transaction(trans,
								root, ret);
					goto error;
				}
3443 3444
				pending_del_nr = 0;
			}
3445
			btrfs_release_path(path);
3446
			goto search_again;
3447 3448
		} else {
			path->slots[0]--;
3449
		}
C
Chris Mason 已提交
3450
	}
3451
out:
3452 3453 3454
	if (pending_del_nr) {
		ret = btrfs_del_items(trans, root, path, pending_del_slot,
				      pending_del_nr);
3455 3456
		if (ret)
			btrfs_abort_transaction(trans, root, ret);
3457
	}
3458
error:
C
Chris Mason 已提交
3459
	btrfs_free_path(path);
3460
	return err;
C
Chris Mason 已提交
3461 3462 3463 3464 3465 3466 3467 3468 3469
}

/*
 * taken from block_truncate_page, but does cow as it zeros out
 * any bytes left in the last page in the file.
 */
static int btrfs_truncate_page(struct address_space *mapping, loff_t from)
{
	struct inode *inode = mapping->host;
3470
	struct btrfs_root *root = BTRFS_I(inode)->root;
3471 3472
	struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
	struct btrfs_ordered_extent *ordered;
3473
	struct extent_state *cached_state = NULL;
3474
	char *kaddr;
3475
	u32 blocksize = root->sectorsize;
C
Chris Mason 已提交
3476 3477 3478
	pgoff_t index = from >> PAGE_CACHE_SHIFT;
	unsigned offset = from & (PAGE_CACHE_SIZE-1);
	struct page *page;
3479
	gfp_t mask = btrfs_alloc_write_mask(mapping);
C
Chris Mason 已提交
3480
	int ret = 0;
3481
	u64 page_start;
3482
	u64 page_end;
C
Chris Mason 已提交
3483 3484 3485

	if ((offset & (blocksize - 1)) == 0)
		goto out;
3486
	ret = btrfs_delalloc_reserve_space(inode, PAGE_CACHE_SIZE);
3487 3488
	if (ret)
		goto out;
C
Chris Mason 已提交
3489 3490

	ret = -ENOMEM;
3491
again:
3492
	page = find_or_create_page(mapping, index, mask);
3493
	if (!page) {
3494
		btrfs_delalloc_release_space(inode, PAGE_CACHE_SIZE);
C
Chris Mason 已提交
3495
		goto out;
3496
	}
3497 3498 3499 3500

	page_start = page_offset(page);
	page_end = page_start + PAGE_CACHE_SIZE - 1;

C
Chris Mason 已提交
3501
	if (!PageUptodate(page)) {
C
Chris Mason 已提交
3502
		ret = btrfs_readpage(NULL, page);
C
Chris Mason 已提交
3503
		lock_page(page);
3504 3505 3506 3507 3508
		if (page->mapping != mapping) {
			unlock_page(page);
			page_cache_release(page);
			goto again;
		}
C
Chris Mason 已提交
3509 3510
		if (!PageUptodate(page)) {
			ret = -EIO;
3511
			goto out_unlock;
C
Chris Mason 已提交
3512 3513
		}
	}
3514
	wait_on_page_writeback(page);
3515

3516
	lock_extent_bits(io_tree, page_start, page_end, 0, &cached_state);
3517 3518 3519 3520
	set_page_extent_mapped(page);

	ordered = btrfs_lookup_ordered_extent(inode, page_start);
	if (ordered) {
3521 3522
		unlock_extent_cached(io_tree, page_start, page_end,
				     &cached_state, GFP_NOFS);
3523 3524
		unlock_page(page);
		page_cache_release(page);
3525
		btrfs_start_ordered_extent(inode, ordered, 1);
3526 3527 3528 3529
		btrfs_put_ordered_extent(ordered);
		goto again;
	}

3530
	clear_extent_bit(&BTRFS_I(inode)->io_tree, page_start, page_end,
3531
			  EXTENT_DIRTY | EXTENT_DELALLOC | EXTENT_DO_ACCOUNTING,
3532
			  0, 0, &cached_state, GFP_NOFS);
3533

3534 3535
	ret = btrfs_set_extent_delalloc(inode, page_start, page_end,
					&cached_state);
J
Josef Bacik 已提交
3536
	if (ret) {
3537 3538
		unlock_extent_cached(io_tree, page_start, page_end,
				     &cached_state, GFP_NOFS);
J
Josef Bacik 已提交
3539 3540 3541
		goto out_unlock;
	}

3542 3543 3544 3545 3546 3547 3548
	ret = 0;
	if (offset != PAGE_CACHE_SIZE) {
		kaddr = kmap(page);
		memset(kaddr + offset, 0, PAGE_CACHE_SIZE - offset);
		flush_dcache_page(page);
		kunmap(page);
	}
3549
	ClearPageChecked(page);
3550
	set_page_dirty(page);
3551 3552
	unlock_extent_cached(io_tree, page_start, page_end, &cached_state,
			     GFP_NOFS);
C
Chris Mason 已提交
3553

3554
out_unlock:
3555
	if (ret)
3556
		btrfs_delalloc_release_space(inode, PAGE_CACHE_SIZE);
C
Chris Mason 已提交
3557 3558 3559 3560 3561 3562
	unlock_page(page);
	page_cache_release(page);
out:
	return ret;
}

3563 3564 3565 3566 3567 3568
/*
 * This function puts in dummy file extents for the area we're creating a hole
 * for.  So if we are truncating this file to a larger size we need to insert
 * these file extents so that btrfs_get_extent will return a EXTENT_MAP_HOLE for
 * the range between oldsize and size
 */
3569
int btrfs_cont_expand(struct inode *inode, loff_t oldsize, loff_t size)
C
Chris Mason 已提交
3570
{
Y
Yan Zheng 已提交
3571 3572 3573
	struct btrfs_trans_handle *trans;
	struct btrfs_root *root = BTRFS_I(inode)->root;
	struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
3574
	struct extent_map *em = NULL;
3575
	struct extent_state *cached_state = NULL;
Y
Yan Zheng 已提交
3576
	u64 mask = root->sectorsize - 1;
3577
	u64 hole_start = (oldsize + mask) & ~mask;
Y
Yan Zheng 已提交
3578 3579 3580 3581
	u64 block_end = (size + mask) & ~mask;
	u64 last_byte;
	u64 cur_offset;
	u64 hole_size;
J
Josef Bacik 已提交
3582
	int err = 0;
C
Chris Mason 已提交
3583

Y
Yan Zheng 已提交
3584 3585 3586 3587 3588 3589 3590
	if (size <= hole_start)
		return 0;

	while (1) {
		struct btrfs_ordered_extent *ordered;
		btrfs_wait_ordered_range(inode, hole_start,
					 block_end - hole_start);
3591
		lock_extent_bits(io_tree, hole_start, block_end - 1, 0,
3592
				 &cached_state);
Y
Yan Zheng 已提交
3593 3594 3595
		ordered = btrfs_lookup_ordered_extent(inode, hole_start);
		if (!ordered)
			break;
3596 3597
		unlock_extent_cached(io_tree, hole_start, block_end - 1,
				     &cached_state, GFP_NOFS);
Y
Yan Zheng 已提交
3598 3599
		btrfs_put_ordered_extent(ordered);
	}
C
Chris Mason 已提交
3600

Y
Yan Zheng 已提交
3601 3602 3603 3604
	cur_offset = hole_start;
	while (1) {
		em = btrfs_get_extent(inode, NULL, 0, cur_offset,
				block_end - cur_offset, 0);
3605 3606 3607 3608
		if (IS_ERR(em)) {
			err = PTR_ERR(em);
			break;
		}
Y
Yan Zheng 已提交
3609 3610
		last_byte = min(extent_map_end(em), block_end);
		last_byte = (last_byte + mask) & ~mask;
3611
		if (!test_bit(EXTENT_FLAG_PREALLOC, &em->flags)) {
3612
			u64 hint_byte = 0;
Y
Yan Zheng 已提交
3613
			hole_size = last_byte - cur_offset;
J
Josef Bacik 已提交
3614

3615
			trans = btrfs_start_transaction(root, 3);
3616 3617
			if (IS_ERR(trans)) {
				err = PTR_ERR(trans);
J
Josef Bacik 已提交
3618
				break;
3619
			}
3620 3621 3622 3623

			err = btrfs_drop_extents(trans, inode, cur_offset,
						 cur_offset + hole_size,
						 &hint_byte, 1);
3624
			if (err) {
3625
				btrfs_abort_transaction(trans, root, err);
3626
				btrfs_end_transaction(trans, root);
3627
				break;
3628
			}
3629

Y
Yan Zheng 已提交
3630
			err = btrfs_insert_file_extent(trans, root,
L
Li Zefan 已提交
3631
					btrfs_ino(inode), cur_offset, 0,
Y
Yan Zheng 已提交
3632 3633
					0, hole_size, 0, hole_size,
					0, 0, 0);
3634
			if (err) {
3635
				btrfs_abort_transaction(trans, root, err);
3636
				btrfs_end_transaction(trans, root);
3637
				break;
3638
			}
3639

Y
Yan Zheng 已提交
3640 3641
			btrfs_drop_extent_cache(inode, hole_start,
					last_byte - 1, 0);
3642

3643
			btrfs_update_inode(trans, root, inode);
3644
			btrfs_end_transaction(trans, root);
Y
Yan Zheng 已提交
3645 3646
		}
		free_extent_map(em);
3647
		em = NULL;
Y
Yan Zheng 已提交
3648
		cur_offset = last_byte;
3649
		if (cur_offset >= block_end)
Y
Yan Zheng 已提交
3650 3651
			break;
	}
3652

3653
	free_extent_map(em);
3654 3655
	unlock_extent_cached(io_tree, hole_start, block_end - 1, &cached_state,
			     GFP_NOFS);
Y
Yan Zheng 已提交
3656 3657
	return err;
}
C
Chris Mason 已提交
3658

3659
static int btrfs_setsize(struct inode *inode, loff_t newsize)
3660
{
3661 3662
	struct btrfs_root *root = BTRFS_I(inode)->root;
	struct btrfs_trans_handle *trans;
3663
	loff_t oldsize = i_size_read(inode);
3664 3665
	int ret;

3666
	if (newsize == oldsize)
3667 3668
		return 0;

3669 3670 3671
	if (newsize > oldsize) {
		truncate_pagecache(inode, oldsize, newsize);
		ret = btrfs_cont_expand(inode, oldsize, newsize);
3672
		if (ret)
3673 3674
			return ret;

3675 3676 3677 3678 3679 3680 3681
		trans = btrfs_start_transaction(root, 1);
		if (IS_ERR(trans))
			return PTR_ERR(trans);

		i_size_write(inode, newsize);
		btrfs_ordered_update_i_size(inode, i_size_read(inode), NULL);
		ret = btrfs_update_inode(trans, root, inode);
3682
		btrfs_end_transaction(trans, root);
3683
	} else {
3684

3685 3686 3687 3688 3689 3690
		/*
		 * We're truncating a file that used to have good data down to
		 * zero. Make sure it gets into the ordered flush list so that
		 * any new writes get down to disk quickly.
		 */
		if (newsize == 0)
3691 3692
			set_bit(BTRFS_INODE_ORDERED_DATA_CLOSE,
				&BTRFS_I(inode)->runtime_flags);
3693

3694 3695 3696
		/* we don't support swapfiles, so vmtruncate shouldn't fail */
		truncate_setsize(inode, newsize);
		ret = btrfs_truncate(inode);
3697 3698
	}

3699
	return ret;
3700 3701
}

Y
Yan Zheng 已提交
3702 3703 3704
static int btrfs_setattr(struct dentry *dentry, struct iattr *attr)
{
	struct inode *inode = dentry->d_inode;
L
Li Zefan 已提交
3705
	struct btrfs_root *root = BTRFS_I(inode)->root;
Y
Yan Zheng 已提交
3706
	int err;
C
Chris Mason 已提交
3707

L
Li Zefan 已提交
3708 3709 3710
	if (btrfs_root_readonly(root))
		return -EROFS;

Y
Yan Zheng 已提交
3711 3712 3713
	err = inode_change_ok(inode, attr);
	if (err)
		return err;
C
Chris Mason 已提交
3714

3715
	if (S_ISREG(inode->i_mode) && (attr->ia_valid & ATTR_SIZE)) {
3716
		err = btrfs_setsize(inode, attr->ia_size);
3717 3718
		if (err)
			return err;
C
Chris Mason 已提交
3719
	}
Y
Yan Zheng 已提交
3720

C
Christoph Hellwig 已提交
3721 3722
	if (attr->ia_valid) {
		setattr_copy(inode, attr);
3723
		inode_inc_iversion(inode);
3724
		err = btrfs_dirty_inode(inode);
C
Christoph Hellwig 已提交
3725

3726
		if (!err && attr->ia_valid & ATTR_MODE)
C
Christoph Hellwig 已提交
3727 3728
			err = btrfs_acl_chmod(inode);
	}
J
Josef Bacik 已提交
3729

C
Chris Mason 已提交
3730 3731
	return err;
}
3732

A
Al Viro 已提交
3733
void btrfs_evict_inode(struct inode *inode)
C
Chris Mason 已提交
3734 3735 3736
{
	struct btrfs_trans_handle *trans;
	struct btrfs_root *root = BTRFS_I(inode)->root;
3737
	struct btrfs_block_rsv *rsv, *global_rsv;
3738
	u64 min_size = btrfs_calc_trunc_metadata_size(root, 1);
3739
	unsigned long nr;
C
Chris Mason 已提交
3740 3741
	int ret;

3742 3743
	trace_btrfs_inode_evict(inode);

C
Chris Mason 已提交
3744
	truncate_inode_pages(&inode->i_data, 0);
3745
	if (inode->i_nlink && (btrfs_root_refs(&root->root_item) != 0 ||
3746
			       btrfs_is_free_space_inode(root, inode)))
A
Al Viro 已提交
3747 3748
		goto no_delete;

C
Chris Mason 已提交
3749
	if (is_bad_inode(inode)) {
3750
		btrfs_orphan_del(NULL, inode);
C
Chris Mason 已提交
3751 3752
		goto no_delete;
	}
A
Al Viro 已提交
3753
	/* do we really want it for ->i_nlink > 0 and zero btrfs_root_refs? */
C
Chris Mason 已提交
3754
	btrfs_wait_ordered_range(inode, 0, (u64)-1);
3755

3756
	if (root->fs_info->log_root_recovering) {
3757
		BUG_ON(test_bit(BTRFS_INODE_HAS_ORPHAN_ITEM,
3758
				 &BTRFS_I(inode)->runtime_flags));
3759 3760 3761
		goto no_delete;
	}

3762 3763 3764 3765 3766
	if (inode->i_nlink > 0) {
		BUG_ON(btrfs_root_refs(&root->root_item) != 0);
		goto no_delete;
	}

3767 3768 3769 3770 3771
	rsv = btrfs_alloc_block_rsv(root);
	if (!rsv) {
		btrfs_orphan_del(NULL, inode);
		goto no_delete;
	}
J
Josef Bacik 已提交
3772
	rsv->size = min_size;
3773
	global_rsv = &root->fs_info->global_block_rsv;
3774

3775
	btrfs_i_size_write(inode, 0);
3776

3777 3778 3779 3780 3781 3782 3783 3784 3785 3786 3787
	/*
	 * This is a bit simpler than btrfs_truncate since
	 *
	 * 1) We've already reserved our space for our orphan item in the
	 *    unlink.
	 * 2) We're going to delete the inode item, so we don't need to update
	 *    it at all.
	 *
	 * So we just need to reserve some slack space in case we add bytes when
	 * doing the truncate.
	 */
3788
	while (1) {
3789
		ret = btrfs_block_rsv_refill_noflush(root, rsv, min_size);
3790 3791 3792 3793 3794 3795 3796 3797

		/*
		 * Try and steal from the global reserve since we will
		 * likely not use this space anyway, we want to try as
		 * hard as possible to get this to work.
		 */
		if (ret)
			ret = btrfs_block_rsv_migrate(global_rsv, rsv, min_size);
3798 3799

		if (ret) {
3800
			printk(KERN_WARNING "Could not get space for a "
3801
			       "delete, will truncate on mount %d\n", ret);
3802 3803 3804
			btrfs_orphan_del(NULL, inode);
			btrfs_free_block_rsv(root, rsv);
			goto no_delete;
3805
		}
3806

3807 3808 3809 3810 3811
		trans = btrfs_start_transaction(root, 0);
		if (IS_ERR(trans)) {
			btrfs_orphan_del(NULL, inode);
			btrfs_free_block_rsv(root, rsv);
			goto no_delete;
3812
		}
3813

3814 3815
		trans->block_rsv = rsv;

3816
		ret = btrfs_truncate_inode_items(trans, root, inode, 0, 0);
3817 3818
		if (ret != -EAGAIN)
			break;
3819

3820 3821 3822 3823 3824
		nr = trans->blocks_used;
		btrfs_end_transaction(trans, root);
		trans = NULL;
		btrfs_btree_balance_dirty(root, nr);
	}
3825

3826 3827
	btrfs_free_block_rsv(root, rsv);

3828
	if (ret == 0) {
3829
		trans->block_rsv = root->orphan_block_rsv;
3830 3831 3832
		ret = btrfs_orphan_del(trans, inode);
		BUG_ON(ret);
	}
3833

3834
	trans->block_rsv = &root->fs_info->trans_block_rsv;
3835 3836
	if (!(root == root->fs_info->tree_root ||
	      root->root_key.objectid == BTRFS_TREE_RELOC_OBJECTID))
L
Li Zefan 已提交
3837
		btrfs_return_ino(root, btrfs_ino(inode));
3838

3839
	nr = trans->blocks_used;
3840
	btrfs_end_transaction(trans, root);
3841
	btrfs_btree_balance_dirty(root, nr);
C
Chris Mason 已提交
3842
no_delete:
3843
	clear_inode(inode);
3844
	return;
C
Chris Mason 已提交
3845 3846 3847 3848 3849 3850 3851 3852 3853 3854 3855 3856 3857 3858
}

/*
 * this returns the key found in the dir entry in the location pointer.
 * If no dir entries were found, location->objectid is 0.
 */
static int btrfs_inode_by_name(struct inode *dir, struct dentry *dentry,
			       struct btrfs_key *location)
{
	const char *name = dentry->d_name.name;
	int namelen = dentry->d_name.len;
	struct btrfs_dir_item *di;
	struct btrfs_path *path;
	struct btrfs_root *root = BTRFS_I(dir)->root;
Y
Yan 已提交
3859
	int ret = 0;
C
Chris Mason 已提交
3860 3861

	path = btrfs_alloc_path();
3862 3863
	if (!path)
		return -ENOMEM;
3864

L
Li Zefan 已提交
3865
	di = btrfs_lookup_dir_item(NULL, root, path, btrfs_ino(dir), name,
C
Chris Mason 已提交
3866
				    namelen, 0);
Y
Yan 已提交
3867 3868
	if (IS_ERR(di))
		ret = PTR_ERR(di);
C
Chris Mason 已提交
3869

3870
	if (IS_ERR_OR_NULL(di))
3871
		goto out_err;
C
Chris Mason 已提交
3872

3873
	btrfs_dir_item_key_to_cpu(path->nodes[0], di, location);
C
Chris Mason 已提交
3874 3875 3876
out:
	btrfs_free_path(path);
	return ret;
3877 3878 3879
out_err:
	location->objectid = 0;
	goto out;
C
Chris Mason 已提交
3880 3881 3882 3883 3884 3885 3886 3887
}

/*
 * when we hit a tree root in a directory, the btrfs part of the inode
 * needs to be changed to reflect the root directory of the tree root.  This
 * is kind of like crossing a mount point.
 */
static int fixup_tree_root_location(struct btrfs_root *root,
3888 3889 3890 3891
				    struct inode *dir,
				    struct dentry *dentry,
				    struct btrfs_key *location,
				    struct btrfs_root **sub_root)
C
Chris Mason 已提交
3892
{
3893 3894 3895 3896 3897 3898
	struct btrfs_path *path;
	struct btrfs_root *new_root;
	struct btrfs_root_ref *ref;
	struct extent_buffer *leaf;
	int ret;
	int err = 0;
C
Chris Mason 已提交
3899

3900 3901 3902 3903 3904
	path = btrfs_alloc_path();
	if (!path) {
		err = -ENOMEM;
		goto out;
	}
C
Chris Mason 已提交
3905

3906 3907 3908 3909 3910 3911 3912 3913 3914
	err = -ENOENT;
	ret = btrfs_find_root_ref(root->fs_info->tree_root, path,
				  BTRFS_I(dir)->root->root_key.objectid,
				  location->objectid);
	if (ret) {
		if (ret < 0)
			err = ret;
		goto out;
	}
C
Chris Mason 已提交
3915

3916 3917
	leaf = path->nodes[0];
	ref = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_root_ref);
L
Li Zefan 已提交
3918
	if (btrfs_root_ref_dirid(leaf, ref) != btrfs_ino(dir) ||
3919 3920
	    btrfs_root_ref_name_len(leaf, ref) != dentry->d_name.len)
		goto out;
C
Chris Mason 已提交
3921

3922 3923 3924 3925 3926 3927
	ret = memcmp_extent_buffer(leaf, dentry->d_name.name,
				   (unsigned long)(ref + 1),
				   dentry->d_name.len);
	if (ret)
		goto out;

3928
	btrfs_release_path(path);
3929 3930 3931 3932 3933 3934 3935 3936 3937 3938 3939 3940 3941 3942 3943 3944 3945 3946 3947 3948

	new_root = btrfs_read_fs_root_no_name(root->fs_info, location);
	if (IS_ERR(new_root)) {
		err = PTR_ERR(new_root);
		goto out;
	}

	if (btrfs_root_refs(&new_root->root_item) == 0) {
		err = -ENOENT;
		goto out;
	}

	*sub_root = new_root;
	location->objectid = btrfs_root_dirid(&new_root->root_item);
	location->type = BTRFS_INODE_ITEM_KEY;
	location->offset = 0;
	err = 0;
out:
	btrfs_free_path(path);
	return err;
C
Chris Mason 已提交
3949 3950
}

3951 3952 3953 3954
static void inode_tree_add(struct inode *inode)
{
	struct btrfs_root *root = BTRFS_I(inode)->root;
	struct btrfs_inode *entry;
3955 3956
	struct rb_node **p;
	struct rb_node *parent;
L
Li Zefan 已提交
3957
	u64 ino = btrfs_ino(inode);
3958 3959 3960
again:
	p = &root->inode_tree.rb_node;
	parent = NULL;
3961

A
Al Viro 已提交
3962
	if (inode_unhashed(inode))
3963 3964
		return;

3965 3966 3967 3968 3969
	spin_lock(&root->inode_lock);
	while (*p) {
		parent = *p;
		entry = rb_entry(parent, struct btrfs_inode, rb_node);

L
Li Zefan 已提交
3970
		if (ino < btrfs_ino(&entry->vfs_inode))
3971
			p = &parent->rb_left;
L
Li Zefan 已提交
3972
		else if (ino > btrfs_ino(&entry->vfs_inode))
3973
			p = &parent->rb_right;
3974 3975
		else {
			WARN_ON(!(entry->vfs_inode.i_state &
A
Al Viro 已提交
3976
				  (I_WILL_FREE | I_FREEING)));
3977 3978 3979 3980
			rb_erase(parent, &root->inode_tree);
			RB_CLEAR_NODE(parent);
			spin_unlock(&root->inode_lock);
			goto again;
3981 3982 3983 3984 3985 3986 3987 3988 3989 3990
		}
	}
	rb_link_node(&BTRFS_I(inode)->rb_node, parent, p);
	rb_insert_color(&BTRFS_I(inode)->rb_node, &root->inode_tree);
	spin_unlock(&root->inode_lock);
}

static void inode_tree_del(struct inode *inode)
{
	struct btrfs_root *root = BTRFS_I(inode)->root;
3991
	int empty = 0;
3992

3993
	spin_lock(&root->inode_lock);
3994 3995 3996
	if (!RB_EMPTY_NODE(&BTRFS_I(inode)->rb_node)) {
		rb_erase(&BTRFS_I(inode)->rb_node, &root->inode_tree);
		RB_CLEAR_NODE(&BTRFS_I(inode)->rb_node);
3997
		empty = RB_EMPTY_ROOT(&root->inode_tree);
3998
	}
3999
	spin_unlock(&root->inode_lock);
4000

4001 4002 4003 4004 4005 4006 4007 4008
	/*
	 * Free space cache has inodes in the tree root, but the tree root has a
	 * root_refs of 0, so this could end up dropping the tree root as a
	 * snapshot, so we need the extra !root->fs_info->tree_root check to
	 * make sure we don't drop it.
	 */
	if (empty && btrfs_root_refs(&root->root_item) == 0 &&
	    root != root->fs_info->tree_root) {
4009 4010 4011 4012 4013 4014 4015 4016 4017
		synchronize_srcu(&root->fs_info->subvol_srcu);
		spin_lock(&root->inode_lock);
		empty = RB_EMPTY_ROOT(&root->inode_tree);
		spin_unlock(&root->inode_lock);
		if (empty)
			btrfs_add_dead_root(root);
	}
}

4018
void btrfs_invalidate_inodes(struct btrfs_root *root)
4019 4020 4021 4022 4023 4024 4025 4026 4027 4028 4029 4030 4031 4032 4033 4034 4035
{
	struct rb_node *node;
	struct rb_node *prev;
	struct btrfs_inode *entry;
	struct inode *inode;
	u64 objectid = 0;

	WARN_ON(btrfs_root_refs(&root->root_item) != 0);

	spin_lock(&root->inode_lock);
again:
	node = root->inode_tree.rb_node;
	prev = NULL;
	while (node) {
		prev = node;
		entry = rb_entry(node, struct btrfs_inode, rb_node);

L
Li Zefan 已提交
4036
		if (objectid < btrfs_ino(&entry->vfs_inode))
4037
			node = node->rb_left;
L
Li Zefan 已提交
4038
		else if (objectid > btrfs_ino(&entry->vfs_inode))
4039 4040 4041 4042 4043 4044 4045
			node = node->rb_right;
		else
			break;
	}
	if (!node) {
		while (prev) {
			entry = rb_entry(prev, struct btrfs_inode, rb_node);
L
Li Zefan 已提交
4046
			if (objectid <= btrfs_ino(&entry->vfs_inode)) {
4047 4048 4049 4050 4051 4052 4053 4054
				node = prev;
				break;
			}
			prev = rb_next(prev);
		}
	}
	while (node) {
		entry = rb_entry(node, struct btrfs_inode, rb_node);
L
Li Zefan 已提交
4055
		objectid = btrfs_ino(&entry->vfs_inode) + 1;
4056 4057 4058 4059 4060 4061
		inode = igrab(&entry->vfs_inode);
		if (inode) {
			spin_unlock(&root->inode_lock);
			if (atomic_read(&inode->i_count) > 1)
				d_prune_aliases(inode);
			/*
4062
			 * btrfs_drop_inode will have it removed from
4063 4064 4065 4066 4067 4068 4069 4070 4071 4072 4073 4074 4075 4076 4077
			 * the inode cache when its usage count
			 * hits zero.
			 */
			iput(inode);
			cond_resched();
			spin_lock(&root->inode_lock);
			goto again;
		}

		if (cond_resched_lock(&root->inode_lock))
			goto again;

		node = rb_next(node);
	}
	spin_unlock(&root->inode_lock);
4078 4079
}

4080 4081 4082 4083 4084
static int btrfs_init_locked_inode(struct inode *inode, void *p)
{
	struct btrfs_iget_args *args = p;
	inode->i_ino = args->ino;
	BTRFS_I(inode)->root = args->root;
J
Josef Bacik 已提交
4085
	btrfs_set_inode_space_info(args->root, inode);
C
Chris Mason 已提交
4086 4087 4088 4089 4090 4091
	return 0;
}

static int btrfs_find_actor(struct inode *inode, void *opaque)
{
	struct btrfs_iget_args *args = opaque;
L
Li Zefan 已提交
4092
	return args->ino == btrfs_ino(inode) &&
C
Chris Mason 已提交
4093
		args->root == BTRFS_I(inode)->root;
C
Chris Mason 已提交
4094 4095
}

4096 4097 4098
static struct inode *btrfs_iget_locked(struct super_block *s,
				       u64 objectid,
				       struct btrfs_root *root)
C
Chris Mason 已提交
4099 4100 4101 4102 4103 4104 4105 4106 4107 4108 4109 4110
{
	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;
}

B
Balaji Rao 已提交
4111 4112 4113 4114
/* Get an inode object given its location and corresponding root.
 * Returns in *is_new if the inode was read from disk
 */
struct inode *btrfs_iget(struct super_block *s, struct btrfs_key *location,
4115
			 struct btrfs_root *root, int *new)
B
Balaji Rao 已提交
4116 4117 4118 4119 4120
{
	struct inode *inode;

	inode = btrfs_iget_locked(s, location->objectid, root);
	if (!inode)
4121
		return ERR_PTR(-ENOMEM);
B
Balaji Rao 已提交
4122 4123 4124 4125 4126

	if (inode->i_state & I_NEW) {
		BTRFS_I(inode)->root = root;
		memcpy(&BTRFS_I(inode)->location, location, sizeof(*location));
		btrfs_read_locked_inode(inode);
4127 4128 4129 4130 4131 4132
		if (!is_bad_inode(inode)) {
			inode_tree_add(inode);
			unlock_new_inode(inode);
			if (new)
				*new = 1;
		} else {
4133 4134 4135
			unlock_new_inode(inode);
			iput(inode);
			inode = ERR_PTR(-ESTALE);
4136 4137 4138
		}
	}

B
Balaji Rao 已提交
4139 4140 4141
	return inode;
}

4142 4143 4144 4145 4146 4147 4148 4149 4150 4151 4152
static struct inode *new_simple_dir(struct super_block *s,
				    struct btrfs_key *key,
				    struct btrfs_root *root)
{
	struct inode *inode = new_inode(s);

	if (!inode)
		return ERR_PTR(-ENOMEM);

	BTRFS_I(inode)->root = root;
	memcpy(&BTRFS_I(inode)->location, key, sizeof(*key));
4153
	set_bit(BTRFS_INODE_DUMMY, &BTRFS_I(inode)->runtime_flags);
4154 4155

	inode->i_ino = BTRFS_EMPTY_SUBVOL_DIR_OBJECTID;
L
Li Zefan 已提交
4156
	inode->i_op = &btrfs_dir_ro_inode_operations;
4157 4158 4159 4160 4161 4162 4163
	inode->i_fop = &simple_dir_operations;
	inode->i_mode = S_IFDIR | S_IRUGO | S_IWUSR | S_IXUGO;
	inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;

	return inode;
}

4164
struct inode *btrfs_lookup_dentry(struct inode *dir, struct dentry *dentry)
C
Chris Mason 已提交
4165
{
C
Chris Mason 已提交
4166
	struct inode *inode;
4167
	struct btrfs_root *root = BTRFS_I(dir)->root;
C
Chris Mason 已提交
4168 4169
	struct btrfs_root *sub_root = root;
	struct btrfs_key location;
4170
	int index;
4171
	int ret = 0;
C
Chris Mason 已提交
4172 4173 4174

	if (dentry->d_name.len > BTRFS_NAME_LEN)
		return ERR_PTR(-ENAMETOOLONG);
4175

4176 4177 4178 4179
	if (unlikely(d_need_lookup(dentry))) {
		memcpy(&location, dentry->d_fsdata, sizeof(struct btrfs_key));
		kfree(dentry->d_fsdata);
		dentry->d_fsdata = NULL;
4180 4181
		/* This thing is hashed, drop it for now */
		d_drop(dentry);
4182 4183 4184
	} else {
		ret = btrfs_inode_by_name(dir, dentry, &location);
	}
4185

C
Chris Mason 已提交
4186 4187
	if (ret < 0)
		return ERR_PTR(ret);
4188

4189 4190 4191 4192
	if (location.objectid == 0)
		return NULL;

	if (location.type == BTRFS_INODE_ITEM_KEY) {
4193
		inode = btrfs_iget(dir->i_sb, &location, root, NULL);
4194 4195 4196 4197 4198
		return inode;
	}

	BUG_ON(location.type != BTRFS_ROOT_ITEM_KEY);

4199
	index = srcu_read_lock(&root->fs_info->subvol_srcu);
4200 4201 4202 4203 4204 4205 4206 4207
	ret = fixup_tree_root_location(root, dir, dentry,
				       &location, &sub_root);
	if (ret < 0) {
		if (ret != -ENOENT)
			inode = ERR_PTR(ret);
		else
			inode = new_simple_dir(dir->i_sb, &location, sub_root);
	} else {
4208
		inode = btrfs_iget(dir->i_sb, &location, sub_root, NULL);
C
Chris Mason 已提交
4209
	}
4210 4211
	srcu_read_unlock(&root->fs_info->subvol_srcu, index);

4212
	if (!IS_ERR(inode) && root != sub_root) {
4213 4214
		down_read(&root->fs_info->cleanup_work_sem);
		if (!(inode->i_sb->s_flags & MS_RDONLY))
4215
			ret = btrfs_orphan_cleanup(sub_root);
4216
		up_read(&root->fs_info->cleanup_work_sem);
4217 4218
		if (ret)
			inode = ERR_PTR(ret);
4219 4220
	}

4221 4222 4223
	return inode;
}

N
Nick Piggin 已提交
4224
static int btrfs_dentry_delete(const struct dentry *dentry)
4225 4226
{
	struct btrfs_root *root;
L
Li Zefan 已提交
4227
	struct inode *inode = dentry->d_inode;
4228

L
Li Zefan 已提交
4229 4230
	if (!inode && !IS_ROOT(dentry))
		inode = dentry->d_parent->d_inode;
4231

L
Li Zefan 已提交
4232 4233
	if (inode) {
		root = BTRFS_I(inode)->root;
4234 4235
		if (btrfs_root_refs(&root->root_item) == 0)
			return 1;
L
Li Zefan 已提交
4236 4237 4238

		if (btrfs_ino(inode) == BTRFS_EMPTY_SUBVOL_DIR_OBJECTID)
			return 1;
4239
	}
4240 4241 4242
	return 0;
}

4243 4244 4245 4246 4247 4248
static void btrfs_dentry_release(struct dentry *dentry)
{
	if (dentry->d_fsdata)
		kfree(dentry->d_fsdata);
}

4249
static struct dentry *btrfs_lookup(struct inode *dir, struct dentry *dentry,
A
Al Viro 已提交
4250
				   unsigned int flags)
4251
{
4252 4253 4254 4255 4256 4257 4258 4259 4260
	struct dentry *ret;

	ret = d_splice_alias(btrfs_lookup_dentry(dir, dentry), dentry);
	if (unlikely(d_need_lookup(dentry))) {
		spin_lock(&dentry->d_lock);
		dentry->d_flags &= ~DCACHE_NEED_LOOKUP;
		spin_unlock(&dentry->d_lock);
	}
	return ret;
C
Chris Mason 已提交
4261 4262
}

4263
unsigned char btrfs_filetype_table[] = {
C
Chris Mason 已提交
4264 4265 4266
	DT_UNKNOWN, DT_REG, DT_DIR, DT_CHR, DT_BLK, DT_FIFO, DT_SOCK, DT_LNK
};

4267 4268
static int btrfs_real_readdir(struct file *filp, void *dirent,
			      filldir_t filldir)
C
Chris Mason 已提交
4269
{
4270
	struct inode *inode = filp->f_dentry->d_inode;
C
Chris Mason 已提交
4271 4272 4273 4274
	struct btrfs_root *root = BTRFS_I(inode)->root;
	struct btrfs_item *item;
	struct btrfs_dir_item *di;
	struct btrfs_key key;
4275
	struct btrfs_key found_key;
C
Chris Mason 已提交
4276
	struct btrfs_path *path;
4277 4278
	struct list_head ins_list;
	struct list_head del_list;
C
Chris Mason 已提交
4279
	int ret;
4280
	struct extent_buffer *leaf;
C
Chris Mason 已提交
4281 4282 4283 4284 4285 4286 4287
	int slot;
	unsigned char d_type;
	int over = 0;
	u32 di_cur;
	u32 di_total;
	u32 di_len;
	int key_type = BTRFS_DIR_INDEX_KEY;
4288 4289 4290
	char tmp_name[32];
	char *name_ptr;
	int name_len;
4291
	int is_curr = 0;	/* filp->f_pos points to the current index? */
C
Chris Mason 已提交
4292 4293 4294 4295

	/* FIXME, use a real flag for deciding about the key type */
	if (root->fs_info->tree_root == root)
		key_type = BTRFS_DIR_ITEM_KEY;
4296

4297 4298
	/* special case for "." */
	if (filp->f_pos == 0) {
4299 4300
		over = filldir(dirent, ".", 1,
			       filp->f_pos, btrfs_ino(inode), DT_DIR);
4301 4302 4303 4304 4305 4306
		if (over)
			return 0;
		filp->f_pos = 1;
	}
	/* special case for .., just use the back ref */
	if (filp->f_pos == 1) {
4307
		u64 pino = parent_ino(filp->f_path.dentry);
4308
		over = filldir(dirent, "..", 2,
4309
			       filp->f_pos, pino, DT_DIR);
4310
		if (over)
4311
			return 0;
4312 4313
		filp->f_pos = 2;
	}
4314
	path = btrfs_alloc_path();
4315 4316
	if (!path)
		return -ENOMEM;
C
Chris Mason 已提交
4317

J
Josef Bacik 已提交
4318
	path->reada = 1;
4319

4320 4321 4322 4323 4324 4325
	if (key_type == BTRFS_DIR_INDEX_KEY) {
		INIT_LIST_HEAD(&ins_list);
		INIT_LIST_HEAD(&del_list);
		btrfs_get_delayed_items(inode, &ins_list, &del_list);
	}

C
Chris Mason 已提交
4326 4327
	btrfs_set_key_type(&key, key_type);
	key.offset = filp->f_pos;
L
Li Zefan 已提交
4328
	key.objectid = btrfs_ino(inode);
4329

C
Chris Mason 已提交
4330 4331 4332
	ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
	if (ret < 0)
		goto err;
4333 4334

	while (1) {
4335
		leaf = path->nodes[0];
C
Chris Mason 已提交
4336
		slot = path->slots[0];
4337 4338 4339 4340 4341 4342 4343
		if (slot >= btrfs_header_nritems(leaf)) {
			ret = btrfs_next_leaf(root, path);
			if (ret < 0)
				goto err;
			else if (ret > 0)
				break;
			continue;
C
Chris Mason 已提交
4344
		}
4345

4346 4347 4348 4349
		item = btrfs_item_nr(leaf, slot);
		btrfs_item_key_to_cpu(leaf, &found_key, slot);

		if (found_key.objectid != key.objectid)
C
Chris Mason 已提交
4350
			break;
4351
		if (btrfs_key_type(&found_key) != key_type)
C
Chris Mason 已提交
4352
			break;
4353
		if (found_key.offset < filp->f_pos)
4354
			goto next;
4355 4356 4357 4358
		if (key_type == BTRFS_DIR_INDEX_KEY &&
		    btrfs_should_delete_dir_index(&del_list,
						  found_key.offset))
			goto next;
4359 4360

		filp->f_pos = found_key.offset;
4361
		is_curr = 1;
4362

C
Chris Mason 已提交
4363 4364
		di = btrfs_item_ptr(leaf, slot, struct btrfs_dir_item);
		di_cur = 0;
4365
		di_total = btrfs_item_size(leaf, item);
4366 4367

		while (di_cur < di_total) {
4368 4369
			struct btrfs_key location;

4370 4371 4372
			if (verify_dir_item(root, leaf, di))
				break;

4373
			name_len = btrfs_dir_name_len(leaf, di);
4374
			if (name_len <= sizeof(tmp_name)) {
4375 4376 4377
				name_ptr = tmp_name;
			} else {
				name_ptr = kmalloc(name_len, GFP_NOFS);
4378 4379 4380 4381
				if (!name_ptr) {
					ret = -ENOMEM;
					goto err;
				}
4382 4383 4384 4385 4386 4387
			}
			read_extent_buffer(leaf, name_ptr,
					   (unsigned long)(di + 1), name_len);

			d_type = btrfs_filetype_table[btrfs_dir_type(leaf, di)];
			btrfs_dir_item_key_to_cpu(leaf, di, &location);
4388

4389

4390
			/* is this a reference to our own snapshot? If so
A
Arne Jansen 已提交
4391 4392 4393 4394 4395 4396 4397
			 * skip it.
			 *
			 * In contrast to old kernels, we insert the snapshot's
			 * dir item and dir index after it has been created, so
			 * we won't find a reference to our own snapshot. We
			 * still keep the following code for backward
			 * compatibility.
4398 4399 4400 4401 4402 4403
			 */
			if (location.type == BTRFS_ROOT_ITEM_KEY &&
			    location.objectid == root->root_key.objectid) {
				over = 0;
				goto skip;
			}
4404
			over = filldir(dirent, name_ptr, name_len,
4405
				       found_key.offset, location.objectid,
C
Chris Mason 已提交
4406
				       d_type);
4407

4408
skip:
4409 4410 4411
			if (name_ptr != tmp_name)
				kfree(name_ptr);

C
Chris Mason 已提交
4412 4413
			if (over)
				goto nopos;
J
Josef Bacik 已提交
4414
			di_len = btrfs_dir_name_len(leaf, di) +
4415
				 btrfs_dir_data_len(leaf, di) + sizeof(*di);
C
Chris Mason 已提交
4416 4417 4418
			di_cur += di_len;
			di = (struct btrfs_dir_item *)((char *)di + di_len);
		}
4419 4420
next:
		path->slots[0]++;
C
Chris Mason 已提交
4421
	}
4422

4423 4424 4425 4426 4427 4428 4429 4430 4431
	if (key_type == BTRFS_DIR_INDEX_KEY) {
		if (is_curr)
			filp->f_pos++;
		ret = btrfs_readdir_delayed_dir_index(filp, dirent, filldir,
						      &ins_list);
		if (ret)
			goto nopos;
	}

4432
	/* Reached end of directory/root. Bump pos past the last item. */
4433
	if (key_type == BTRFS_DIR_INDEX_KEY)
4434 4435 4436 4437 4438
		/*
		 * 32-bit glibc will use getdents64, but then strtol -
		 * so the last number we can serve is this.
		 */
		filp->f_pos = 0x7fffffff;
4439 4440
	else
		filp->f_pos++;
C
Chris Mason 已提交
4441 4442 4443
nopos:
	ret = 0;
err:
4444 4445
	if (key_type == BTRFS_DIR_INDEX_KEY)
		btrfs_put_delayed_items(&ins_list, &del_list);
C
Chris Mason 已提交
4446 4447 4448 4449
	btrfs_free_path(path);
	return ret;
}

4450
int btrfs_write_inode(struct inode *inode, struct writeback_control *wbc)
C
Chris Mason 已提交
4451 4452 4453 4454
{
	struct btrfs_root *root = BTRFS_I(inode)->root;
	struct btrfs_trans_handle *trans;
	int ret = 0;
4455
	bool nolock = false;
C
Chris Mason 已提交
4456

4457
	if (test_bit(BTRFS_INODE_DUMMY, &BTRFS_I(inode)->runtime_flags))
4458 4459
		return 0;

4460
	if (btrfs_fs_closing(root->fs_info) && btrfs_is_free_space_inode(root, inode))
4461
		nolock = true;
4462

4463
	if (wbc->sync_mode == WB_SYNC_ALL) {
4464
		if (nolock)
4465
			trans = btrfs_join_transaction_nolock(root);
4466
		else
4467
			trans = btrfs_join_transaction(root);
4468 4469
		if (IS_ERR(trans))
			return PTR_ERR(trans);
4470 4471 4472 4473
		if (nolock)
			ret = btrfs_end_transaction_nolock(trans, root);
		else
			ret = btrfs_commit_transaction(trans, root);
C
Chris Mason 已提交
4474 4475 4476 4477 4478
	}
	return ret;
}

/*
4479
 * This is somewhat expensive, updating the tree every time the
C
Chris Mason 已提交
4480 4481 4482 4483
 * inode changes.  But, it is most likely to find the inode in cache.
 * FIXME, needs more benchmarking...there are no reasons other than performance
 * to keep or drop this code.
 */
4484
int btrfs_dirty_inode(struct inode *inode)
C
Chris Mason 已提交
4485 4486 4487
{
	struct btrfs_root *root = BTRFS_I(inode)->root;
	struct btrfs_trans_handle *trans;
4488 4489
	int ret;

4490
	if (test_bit(BTRFS_INODE_DUMMY, &BTRFS_I(inode)->runtime_flags))
4491
		return 0;
C
Chris Mason 已提交
4492

4493
	trans = btrfs_join_transaction(root);
4494 4495
	if (IS_ERR(trans))
		return PTR_ERR(trans);
4496 4497

	ret = btrfs_update_inode(trans, root, inode);
4498 4499 4500 4501
	if (ret && ret == -ENOSPC) {
		/* whoops, lets try again with the full transaction */
		btrfs_end_transaction(trans, root);
		trans = btrfs_start_transaction(root, 1);
4502 4503
		if (IS_ERR(trans))
			return PTR_ERR(trans);
4504

4505 4506
		ret = btrfs_update_inode(trans, root, inode);
	}
C
Chris Mason 已提交
4507
	btrfs_end_transaction(trans, root);
4508 4509
	if (BTRFS_I(inode)->delayed_node)
		btrfs_balance_delayed_items(root);
4510 4511 4512 4513 4514 4515 4516 4517

	return ret;
}

/*
 * This is a copy of file_update_time.  We need this so we can return error on
 * ENOSPC for updating the inode in the case of file write and mmap writes.
 */
4518 4519
static int btrfs_update_time(struct inode *inode, struct timespec *now,
			     int flags)
4520
{
4521
	if (flags & S_VERSION)
4522
		inode_inc_iversion(inode);
4523 4524 4525 4526 4527 4528 4529
	if (flags & S_CTIME)
		inode->i_ctime = *now;
	if (flags & S_MTIME)
		inode->i_mtime = *now;
	if (flags & S_ATIME)
		inode->i_atime = *now;
	return btrfs_dirty_inode(inode);
C
Chris Mason 已提交
4530 4531
}

C
Chris Mason 已提交
4532 4533 4534 4535 4536
/*
 * find the highest existing sequence number in a directory
 * and then set the in-memory index_cnt variable to reflect
 * free sequence numbers
 */
4537 4538 4539 4540 4541 4542 4543 4544
static int btrfs_set_inode_index_count(struct inode *inode)
{
	struct btrfs_root *root = BTRFS_I(inode)->root;
	struct btrfs_key key, found_key;
	struct btrfs_path *path;
	struct extent_buffer *leaf;
	int ret;

L
Li Zefan 已提交
4545
	key.objectid = btrfs_ino(inode);
4546 4547 4548 4549 4550 4551 4552 4553 4554 4555 4556 4557 4558 4559 4560 4561 4562 4563 4564 4565 4566 4567 4568 4569 4570 4571 4572 4573 4574 4575 4576
	btrfs_set_key_type(&key, BTRFS_DIR_INDEX_KEY);
	key.offset = (u64)-1;

	path = btrfs_alloc_path();
	if (!path)
		return -ENOMEM;

	ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
	if (ret < 0)
		goto out;
	/* FIXME: we should be able to handle this */
	if (ret == 0)
		goto out;
	ret = 0;

	/*
	 * MAGIC NUMBER EXPLANATION:
	 * since we search a directory based on f_pos we have to start at 2
	 * since '.' and '..' have f_pos of 0 and 1 respectively, so everybody
	 * else has to start at 2
	 */
	if (path->slots[0] == 0) {
		BTRFS_I(inode)->index_cnt = 2;
		goto out;
	}

	path->slots[0]--;

	leaf = path->nodes[0];
	btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);

L
Li Zefan 已提交
4577
	if (found_key.objectid != btrfs_ino(inode) ||
4578 4579 4580 4581 4582 4583 4584 4585 4586 4587 4588
	    btrfs_key_type(&found_key) != BTRFS_DIR_INDEX_KEY) {
		BTRFS_I(inode)->index_cnt = 2;
		goto out;
	}

	BTRFS_I(inode)->index_cnt = found_key.offset + 1;
out:
	btrfs_free_path(path);
	return ret;
}

C
Chris Mason 已提交
4589 4590 4591 4592
/*
 * helper to find a free sequence number in a given directory.  This current
 * code is very simple, later versions will do smarter things in the btree
 */
4593
int btrfs_set_inode_index(struct inode *dir, u64 *index)
4594 4595 4596 4597
{
	int ret = 0;

	if (BTRFS_I(dir)->index_cnt == (u64)-1) {
4598 4599 4600 4601 4602 4603
		ret = btrfs_inode_delayed_dir_index_count(dir);
		if (ret) {
			ret = btrfs_set_inode_index_count(dir);
			if (ret)
				return ret;
		}
4604 4605
	}

4606
	*index = BTRFS_I(dir)->index_cnt;
4607 4608 4609 4610 4611
	BTRFS_I(dir)->index_cnt++;

	return ret;
}

C
Chris Mason 已提交
4612 4613
static struct inode *btrfs_new_inode(struct btrfs_trans_handle *trans,
				     struct btrfs_root *root,
4614
				     struct inode *dir,
4615
				     const char *name, int name_len,
A
Al Viro 已提交
4616 4617
				     u64 ref_objectid, u64 objectid,
				     umode_t mode, u64 *index)
C
Chris Mason 已提交
4618 4619
{
	struct inode *inode;
4620
	struct btrfs_inode_item *inode_item;
C
Chris Mason 已提交
4621
	struct btrfs_key *location;
4622
	struct btrfs_path *path;
4623 4624 4625 4626
	struct btrfs_inode_ref *ref;
	struct btrfs_key key[2];
	u32 sizes[2];
	unsigned long ptr;
C
Chris Mason 已提交
4627 4628 4629
	int ret;
	int owner;

4630
	path = btrfs_alloc_path();
4631 4632
	if (!path)
		return ERR_PTR(-ENOMEM);
4633

C
Chris Mason 已提交
4634
	inode = new_inode(root->fs_info->sb);
4635 4636
	if (!inode) {
		btrfs_free_path(path);
C
Chris Mason 已提交
4637
		return ERR_PTR(-ENOMEM);
4638
	}
C
Chris Mason 已提交
4639

4640 4641 4642 4643 4644 4645
	/*
	 * we have to initialize this early, so we can reclaim the inode
	 * number if we fail afterwards in this function.
	 */
	inode->i_ino = objectid;

4646
	if (dir) {
4647 4648
		trace_btrfs_inode_request(dir);

4649
		ret = btrfs_set_inode_index(dir, index);
4650
		if (ret) {
4651
			btrfs_free_path(path);
4652
			iput(inode);
4653
			return ERR_PTR(ret);
4654
		}
4655 4656 4657 4658 4659 4660 4661
	}
	/*
	 * index_cnt is ignored for everything but a dir,
	 * btrfs_get_inode_index_count has an explanation for the magic
	 * number
	 */
	BTRFS_I(inode)->index_cnt = 2;
C
Chris Mason 已提交
4662
	BTRFS_I(inode)->root = root;
4663
	BTRFS_I(inode)->generation = trans->transid;
4664
	inode->i_generation = BTRFS_I(inode)->generation;
J
Josef Bacik 已提交
4665
	btrfs_set_inode_space_info(root, inode);
4666

4667
	if (S_ISDIR(mode))
C
Chris Mason 已提交
4668 4669 4670
		owner = 0;
	else
		owner = 1;
4671 4672 4673 4674 4675 4676 4677 4678 4679 4680 4681 4682

	key[0].objectid = objectid;
	btrfs_set_key_type(&key[0], BTRFS_INODE_ITEM_KEY);
	key[0].offset = 0;

	key[1].objectid = objectid;
	btrfs_set_key_type(&key[1], BTRFS_INODE_REF_KEY);
	key[1].offset = ref_objectid;

	sizes[0] = sizeof(struct btrfs_inode_item);
	sizes[1] = name_len + sizeof(*ref);

4683
	path->leave_spinning = 1;
4684 4685
	ret = btrfs_insert_empty_items(trans, root, path, key, sizes, 2);
	if (ret != 0)
4686 4687
		goto fail;

4688
	inode_init_owner(inode, dir, mode);
4689
	inode_set_bytes(inode, 0);
C
Chris Mason 已提交
4690
	inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
4691 4692
	inode_item = btrfs_item_ptr(path->nodes[0], path->slots[0],
				  struct btrfs_inode_item);
4693
	fill_inode_item(trans, path->nodes[0], inode_item, inode);
4694 4695 4696 4697

	ref = btrfs_item_ptr(path->nodes[0], path->slots[0] + 1,
			     struct btrfs_inode_ref);
	btrfs_set_inode_ref_name_len(path->nodes[0], ref, name_len);
4698
	btrfs_set_inode_ref_index(path->nodes[0], ref, *index);
4699 4700 4701
	ptr = (unsigned long)(ref + 1);
	write_extent_buffer(path->nodes[0], name, ptr, name_len);

4702 4703 4704
	btrfs_mark_buffer_dirty(path->nodes[0]);
	btrfs_free_path(path);

C
Chris Mason 已提交
4705 4706 4707 4708 4709
	location = &BTRFS_I(inode)->location;
	location->objectid = objectid;
	location->offset = 0;
	btrfs_set_key_type(location, BTRFS_INODE_ITEM_KEY);

4710 4711
	btrfs_inherit_iflags(inode, dir);

4712
	if (S_ISREG(mode)) {
4713 4714
		if (btrfs_test_opt(root, NODATASUM))
			BTRFS_I(inode)->flags |= BTRFS_INODE_NODATASUM;
4715 4716
		if (btrfs_test_opt(root, NODATACOW) ||
		    (BTRFS_I(dir)->flags & BTRFS_INODE_NODATACOW))
4717 4718 4719
			BTRFS_I(inode)->flags |= BTRFS_INODE_NODATACOW;
	}

C
Chris Mason 已提交
4720
	insert_inode_hash(inode);
4721
	inode_tree_add(inode);
4722 4723

	trace_btrfs_inode_new(inode);
4724
	btrfs_set_inode_last_trans(trans, inode);
4725

C
Chris Mason 已提交
4726
	return inode;
4727
fail:
4728 4729
	if (dir)
		BTRFS_I(dir)->index_cnt--;
4730
	btrfs_free_path(path);
4731
	iput(inode);
4732
	return ERR_PTR(ret);
C
Chris Mason 已提交
4733 4734 4735 4736 4737 4738 4739
}

static inline u8 btrfs_inode_type(struct inode *inode)
{
	return btrfs_type_by_mode[(inode->i_mode & S_IFMT) >> S_SHIFT];
}

C
Chris Mason 已提交
4740 4741 4742 4743 4744 4745
/*
 * utility function to add 'inode' into 'parent_inode' with
 * a give name and a given sequence number.
 * if 'add_backref' is true, also insert a backref from the
 * inode to the parent directory.
 */
4746 4747 4748
int btrfs_add_link(struct btrfs_trans_handle *trans,
		   struct inode *parent_inode, struct inode *inode,
		   const char *name, int name_len, int add_backref, u64 index)
C
Chris Mason 已提交
4749
{
4750
	int ret = 0;
C
Chris Mason 已提交
4751
	struct btrfs_key key;
4752
	struct btrfs_root *root = BTRFS_I(parent_inode)->root;
L
Li Zefan 已提交
4753 4754
	u64 ino = btrfs_ino(inode);
	u64 parent_ino = btrfs_ino(parent_inode);
4755

L
Li Zefan 已提交
4756
	if (unlikely(ino == BTRFS_FIRST_FREE_OBJECTID)) {
4757 4758
		memcpy(&key, &BTRFS_I(inode)->root->root_key, sizeof(key));
	} else {
L
Li Zefan 已提交
4759
		key.objectid = ino;
4760 4761 4762 4763
		btrfs_set_key_type(&key, BTRFS_INODE_ITEM_KEY);
		key.offset = 0;
	}

L
Li Zefan 已提交
4764
	if (unlikely(ino == BTRFS_FIRST_FREE_OBJECTID)) {
4765 4766
		ret = btrfs_add_root_ref(trans, root->fs_info->tree_root,
					 key.objectid, root->root_key.objectid,
L
Li Zefan 已提交
4767
					 parent_ino, index, name, name_len);
4768
	} else if (add_backref) {
L
Li Zefan 已提交
4769 4770
		ret = btrfs_insert_inode_ref(trans, root, name, name_len, ino,
					     parent_ino, index);
4771
	}
C
Chris Mason 已提交
4772

4773 4774 4775
	/* Nothing to clean up yet */
	if (ret)
		return ret;
4776

4777 4778 4779 4780 4781 4782 4783 4784
	ret = btrfs_insert_dir_item(trans, root, name, name_len,
				    parent_inode, &key,
				    btrfs_inode_type(inode), index);
	if (ret == -EEXIST)
		goto fail_dir_item;
	else if (ret) {
		btrfs_abort_transaction(trans, root, ret);
		return ret;
C
Chris Mason 已提交
4785
	}
4786 4787 4788

	btrfs_i_size_write(parent_inode, parent_inode->i_size +
			   name_len * 2);
4789
	inode_inc_iversion(parent_inode);
4790 4791 4792 4793
	parent_inode->i_mtime = parent_inode->i_ctime = CURRENT_TIME;
	ret = btrfs_update_inode(trans, root, parent_inode);
	if (ret)
		btrfs_abort_transaction(trans, root, ret);
C
Chris Mason 已提交
4794
	return ret;
4795 4796 4797 4798 4799 4800 4801 4802 4803 4804 4805 4806 4807 4808 4809 4810 4811

fail_dir_item:
	if (unlikely(ino == BTRFS_FIRST_FREE_OBJECTID)) {
		u64 local_index;
		int err;
		err = btrfs_del_root_ref(trans, root->fs_info->tree_root,
				 key.objectid, root->root_key.objectid,
				 parent_ino, &local_index, name, name_len);

	} else if (add_backref) {
		u64 local_index;
		int err;

		err = btrfs_del_inode_ref(trans, root, name, name_len,
					  ino, parent_ino, &local_index);
	}
	return ret;
C
Chris Mason 已提交
4812 4813 4814
}

static int btrfs_add_nondir(struct btrfs_trans_handle *trans,
4815 4816
			    struct inode *dir, struct dentry *dentry,
			    struct inode *inode, int backref, u64 index)
C
Chris Mason 已提交
4817
{
4818 4819 4820
	int err = btrfs_add_link(trans, dir, inode,
				 dentry->d_name.name, dentry->d_name.len,
				 backref, index);
C
Chris Mason 已提交
4821 4822 4823 4824 4825
	if (err > 0)
		err = -EEXIST;
	return err;
}

J
Josef Bacik 已提交
4826
static int btrfs_mknod(struct inode *dir, struct dentry *dentry,
A
Al Viro 已提交
4827
			umode_t mode, dev_t rdev)
J
Josef Bacik 已提交
4828 4829 4830
{
	struct btrfs_trans_handle *trans;
	struct btrfs_root *root = BTRFS_I(dir)->root;
4831
	struct inode *inode = NULL;
J
Josef Bacik 已提交
4832 4833 4834
	int err;
	int drop_inode = 0;
	u64 objectid;
4835
	unsigned long nr = 0;
4836
	u64 index = 0;
J
Josef Bacik 已提交
4837 4838 4839 4840

	if (!new_valid_dev(rdev))
		return -EINVAL;

J
Josef Bacik 已提交
4841 4842 4843 4844 4845
	/*
	 * 2 for inode item and ref
	 * 2 for dir items
	 * 1 for xattr if selinux is on
	 */
4846 4847 4848
	trans = btrfs_start_transaction(root, 5);
	if (IS_ERR(trans))
		return PTR_ERR(trans);
4849

4850 4851 4852 4853
	err = btrfs_find_free_ino(root, &objectid);
	if (err)
		goto out_unlock;

4854
	inode = btrfs_new_inode(trans, root, dir, dentry->d_name.name,
L
Li Zefan 已提交
4855
				dentry->d_name.len, btrfs_ino(dir), objectid,
4856
				mode, &index);
4857 4858
	if (IS_ERR(inode)) {
		err = PTR_ERR(inode);
J
Josef Bacik 已提交
4859
		goto out_unlock;
4860
	}
J
Josef Bacik 已提交
4861

4862
	err = btrfs_init_inode_security(trans, inode, dir, &dentry->d_name);
J
Josef Bacik 已提交
4863 4864 4865 4866 4867
	if (err) {
		drop_inode = 1;
		goto out_unlock;
	}

4868 4869 4870 4871 4872 4873 4874 4875
	/*
	* If the active LSM wants to access the inode during
	* d_instantiate it needs these. Smack checks to see
	* if the filesystem supports xattrs by looking at the
	* ops vector.
	*/

	inode->i_op = &btrfs_special_inode_operations;
4876
	err = btrfs_add_nondir(trans, dir, dentry, inode, 0, index);
J
Josef Bacik 已提交
4877 4878 4879 4880
	if (err)
		drop_inode = 1;
	else {
		init_special_inode(inode, inode->i_mode, rdev);
4881
		btrfs_update_inode(trans, root, inode);
4882
		d_instantiate(dentry, inode);
J
Josef Bacik 已提交
4883 4884
	}
out_unlock:
4885
	nr = trans->blocks_used;
4886
	btrfs_end_transaction(trans, root);
4887
	btrfs_btree_balance_dirty(root, nr);
J
Josef Bacik 已提交
4888 4889 4890 4891 4892 4893 4894
	if (drop_inode) {
		inode_dec_link_count(inode);
		iput(inode);
	}
	return err;
}

C
Chris Mason 已提交
4895
static int btrfs_create(struct inode *dir, struct dentry *dentry,
A
Al Viro 已提交
4896
			umode_t mode, bool excl)
C
Chris Mason 已提交
4897 4898 4899
{
	struct btrfs_trans_handle *trans;
	struct btrfs_root *root = BTRFS_I(dir)->root;
4900
	struct inode *inode = NULL;
C
Chris Mason 已提交
4901
	int drop_inode = 0;
4902
	int err;
4903
	unsigned long nr = 0;
C
Chris Mason 已提交
4904
	u64 objectid;
4905
	u64 index = 0;
C
Chris Mason 已提交
4906

J
Josef Bacik 已提交
4907 4908 4909 4910 4911
	/*
	 * 2 for inode item and ref
	 * 2 for dir items
	 * 1 for xattr if selinux is on
	 */
4912 4913 4914
	trans = btrfs_start_transaction(root, 5);
	if (IS_ERR(trans))
		return PTR_ERR(trans);
J
Josef Bacik 已提交
4915

4916 4917 4918 4919
	err = btrfs_find_free_ino(root, &objectid);
	if (err)
		goto out_unlock;

4920
	inode = btrfs_new_inode(trans, root, dir, dentry->d_name.name,
L
Li Zefan 已提交
4921
				dentry->d_name.len, btrfs_ino(dir), objectid,
4922
				mode, &index);
4923 4924
	if (IS_ERR(inode)) {
		err = PTR_ERR(inode);
C
Chris Mason 已提交
4925
		goto out_unlock;
4926
	}
C
Chris Mason 已提交
4927

4928
	err = btrfs_init_inode_security(trans, inode, dir, &dentry->d_name);
J
Josef Bacik 已提交
4929 4930 4931 4932 4933
	if (err) {
		drop_inode = 1;
		goto out_unlock;
	}

4934 4935 4936 4937 4938 4939 4940 4941 4942
	/*
	* If the active LSM wants to access the inode during
	* d_instantiate it needs these. Smack checks to see
	* if the filesystem supports xattrs by looking at the
	* ops vector.
	*/
	inode->i_fop = &btrfs_file_operations;
	inode->i_op = &btrfs_file_inode_operations;

4943
	err = btrfs_add_nondir(trans, dir, dentry, inode, 0, index);
C
Chris Mason 已提交
4944 4945 4946 4947
	if (err)
		drop_inode = 1;
	else {
		inode->i_mapping->a_ops = &btrfs_aops;
C
Chris Mason 已提交
4948
		inode->i_mapping->backing_dev_info = &root->fs_info->bdi;
4949
		BTRFS_I(inode)->io_tree.ops = &btrfs_extent_io_ops;
4950
		d_instantiate(dentry, inode);
C
Chris Mason 已提交
4951 4952
	}
out_unlock:
4953
	nr = trans->blocks_used;
4954
	btrfs_end_transaction(trans, root);
C
Chris Mason 已提交
4955 4956 4957 4958
	if (drop_inode) {
		inode_dec_link_count(inode);
		iput(inode);
	}
4959
	btrfs_btree_balance_dirty(root, nr);
C
Chris Mason 已提交
4960 4961 4962 4963 4964 4965 4966 4967 4968
	return err;
}

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;
4969
	u64 index;
4970
	unsigned long nr = 0;
C
Chris Mason 已提交
4971 4972 4973
	int err;
	int drop_inode = 0;

4974 4975
	/* do not allow sys_link's with other subvols of the same device */
	if (root->objectid != BTRFS_I(inode)->root->objectid)
4976
		return -EXDEV;
4977

4978 4979
	if (inode->i_nlink == ~0U)
		return -EMLINK;
4980

4981
	err = btrfs_set_inode_index(dir, &index);
4982 4983 4984
	if (err)
		goto fail;

4985
	/*
M
Miao Xie 已提交
4986
	 * 2 items for inode and inode ref
4987
	 * 2 items for dir items
M
Miao Xie 已提交
4988
	 * 1 item for parent inode
4989
	 */
M
Miao Xie 已提交
4990
	trans = btrfs_start_transaction(root, 5);
4991 4992 4993 4994
	if (IS_ERR(trans)) {
		err = PTR_ERR(trans);
		goto fail;
	}
4995

4996
	btrfs_inc_nlink(inode);
4997
	inode_inc_iversion(inode);
4998
	inode->i_ctime = CURRENT_TIME;
A
Al Viro 已提交
4999
	ihold(inode);
5000

5001
	err = btrfs_add_nondir(trans, dir, dentry, inode, 1, index);
5002

5003
	if (err) {
5004
		drop_inode = 1;
5005
	} else {
5006
		struct dentry *parent = dentry->d_parent;
5007
		err = btrfs_update_inode(trans, root, inode);
5008 5009
		if (err)
			goto fail;
5010
		d_instantiate(dentry, inode);
5011
		btrfs_log_new_name(trans, inode, NULL, parent);
5012
	}
C
Chris Mason 已提交
5013

5014
	nr = trans->blocks_used;
5015
	btrfs_end_transaction(trans, root);
5016
fail:
C
Chris Mason 已提交
5017 5018 5019 5020
	if (drop_inode) {
		inode_dec_link_count(inode);
		iput(inode);
	}
5021
	btrfs_btree_balance_dirty(root, nr);
C
Chris Mason 已提交
5022 5023 5024
	return err;
}

5025
static int btrfs_mkdir(struct inode *dir, struct dentry *dentry, umode_t mode)
C
Chris Mason 已提交
5026
{
5027
	struct inode *inode = NULL;
C
Chris Mason 已提交
5028 5029 5030 5031
	struct btrfs_trans_handle *trans;
	struct btrfs_root *root = BTRFS_I(dir)->root;
	int err = 0;
	int drop_on_err = 0;
5032
	u64 objectid = 0;
5033
	u64 index = 0;
5034
	unsigned long nr = 1;
C
Chris Mason 已提交
5035

J
Josef Bacik 已提交
5036 5037 5038 5039 5040
	/*
	 * 2 items for inode and ref
	 * 2 items for dir items
	 * 1 for xattr if selinux is on
	 */
5041 5042 5043
	trans = btrfs_start_transaction(root, 5);
	if (IS_ERR(trans))
		return PTR_ERR(trans);
C
Chris Mason 已提交
5044

5045 5046 5047 5048
	err = btrfs_find_free_ino(root, &objectid);
	if (err)
		goto out_fail;

5049
	inode = btrfs_new_inode(trans, root, dir, dentry->d_name.name,
L
Li Zefan 已提交
5050
				dentry->d_name.len, btrfs_ino(dir), objectid,
5051
				S_IFDIR | mode, &index);
C
Chris Mason 已提交
5052 5053 5054 5055
	if (IS_ERR(inode)) {
		err = PTR_ERR(inode);
		goto out_fail;
	}
5056

C
Chris Mason 已提交
5057
	drop_on_err = 1;
J
Josef Bacik 已提交
5058

5059
	err = btrfs_init_inode_security(trans, inode, dir, &dentry->d_name);
J
Josef Bacik 已提交
5060 5061 5062
	if (err)
		goto out_fail;

C
Chris Mason 已提交
5063 5064 5065
	inode->i_op = &btrfs_dir_inode_operations;
	inode->i_fop = &btrfs_dir_file_operations;

5066
	btrfs_i_size_write(inode, 0);
C
Chris Mason 已提交
5067 5068 5069
	err = btrfs_update_inode(trans, root, inode);
	if (err)
		goto out_fail;
5070

5071 5072
	err = btrfs_add_link(trans, dir, inode, dentry->d_name.name,
			     dentry->d_name.len, 0, index);
C
Chris Mason 已提交
5073 5074
	if (err)
		goto out_fail;
5075

C
Chris Mason 已提交
5076 5077 5078 5079
	d_instantiate(dentry, inode);
	drop_on_err = 0;

out_fail:
5080
	nr = trans->blocks_used;
5081
	btrfs_end_transaction(trans, root);
C
Chris Mason 已提交
5082 5083
	if (drop_on_err)
		iput(inode);
5084
	btrfs_btree_balance_dirty(root, nr);
C
Chris Mason 已提交
5085 5086 5087
	return err;
}

C
Chris Mason 已提交
5088 5089 5090 5091
/* helper for btfs_get_extent.  Given an existing extent in the tree,
 * and an extent that you want to insert, deal with overlap and insert
 * the new extent into the tree.
 */
5092 5093
static int merge_extent_mapping(struct extent_map_tree *em_tree,
				struct extent_map *existing,
5094 5095
				struct extent_map *em,
				u64 map_start, u64 map_len)
5096 5097 5098
{
	u64 start_diff;

5099 5100 5101 5102
	BUG_ON(map_start < em->start || map_start >= extent_map_end(em));
	start_diff = map_start - em->start;
	em->start = map_start;
	em->len = map_len;
C
Chris Mason 已提交
5103 5104
	if (em->block_start < EXTENT_MAP_LAST_BYTE &&
	    !test_bit(EXTENT_FLAG_COMPRESSED, &em->flags)) {
5105
		em->block_start += start_diff;
C
Chris Mason 已提交
5106 5107
		em->block_len -= start_diff;
	}
5108
	return add_extent_mapping(em_tree, em);
5109 5110
}

C
Chris Mason 已提交
5111 5112 5113 5114 5115 5116 5117 5118 5119 5120 5121
static noinline int uncompress_inline(struct btrfs_path *path,
				      struct inode *inode, struct page *page,
				      size_t pg_offset, u64 extent_offset,
				      struct btrfs_file_extent_item *item)
{
	int ret;
	struct extent_buffer *leaf = path->nodes[0];
	char *tmp;
	size_t max_size;
	unsigned long inline_size;
	unsigned long ptr;
5122
	int compress_type;
C
Chris Mason 已提交
5123 5124

	WARN_ON(pg_offset != 0);
5125
	compress_type = btrfs_file_extent_compression(leaf, item);
C
Chris Mason 已提交
5126 5127 5128 5129
	max_size = btrfs_file_extent_ram_bytes(leaf, item);
	inline_size = btrfs_file_extent_inline_item_len(leaf,
					btrfs_item_nr(leaf, path->slots[0]));
	tmp = kmalloc(inline_size, GFP_NOFS);
5130 5131
	if (!tmp)
		return -ENOMEM;
C
Chris Mason 已提交
5132 5133 5134 5135
	ptr = btrfs_file_extent_inline_start(item);

	read_extent_buffer(leaf, tmp, ptr, inline_size);

5136
	max_size = min_t(unsigned long, PAGE_CACHE_SIZE, max_size);
5137 5138
	ret = btrfs_decompress(compress_type, tmp, page,
			       extent_offset, inline_size, max_size);
C
Chris Mason 已提交
5139
	if (ret) {
5140
		char *kaddr = kmap_atomic(page);
C
Chris Mason 已提交
5141 5142 5143 5144
		unsigned long copy_size = min_t(u64,
				  PAGE_CACHE_SIZE - pg_offset,
				  max_size - extent_offset);
		memset(kaddr + pg_offset, 0, copy_size);
5145
		kunmap_atomic(kaddr);
C
Chris Mason 已提交
5146 5147 5148 5149 5150
	}
	kfree(tmp);
	return 0;
}

C
Chris Mason 已提交
5151 5152
/*
 * a bit scary, this does extent mapping from logical file offset to the disk.
C
Chris Mason 已提交
5153 5154
 * the ugly parts come from merging extents from the disk with the in-ram
 * representation.  This gets more complex because of the data=ordered code,
C
Chris Mason 已提交
5155 5156 5157 5158
 * where the in-ram extents might be locked pending data=ordered completion.
 *
 * This also copies inline extents directly into the page.
 */
C
Chris Mason 已提交
5159

5160
struct extent_map *btrfs_get_extent(struct inode *inode, struct page *page,
5161
				    size_t pg_offset, u64 start, u64 len,
5162 5163 5164 5165
				    int create)
{
	int ret;
	int err = 0;
5166
	u64 bytenr;
5167 5168
	u64 extent_start = 0;
	u64 extent_end = 0;
L
Li Zefan 已提交
5169
	u64 objectid = btrfs_ino(inode);
5170
	u32 found_type;
5171
	struct btrfs_path *path = NULL;
5172 5173
	struct btrfs_root *root = BTRFS_I(inode)->root;
	struct btrfs_file_extent_item *item;
5174 5175
	struct extent_buffer *leaf;
	struct btrfs_key found_key;
5176 5177
	struct extent_map *em = NULL;
	struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
5178
	struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
5179
	struct btrfs_trans_handle *trans = NULL;
5180
	int compress_type;
5181 5182

again:
5183
	read_lock(&em_tree->lock);
5184
	em = lookup_extent_mapping(em_tree, start, len);
5185 5186
	if (em)
		em->bdev = root->fs_info->fs_devices->latest_bdev;
5187
	read_unlock(&em_tree->lock);
5188

5189
	if (em) {
5190 5191 5192
		if (em->start > start || em->start + em->len <= start)
			free_extent_map(em);
		else if (em->block_start == EXTENT_MAP_INLINE && page)
5193 5194 5195
			free_extent_map(em);
		else
			goto out;
5196
	}
5197
	em = alloc_extent_map();
5198
	if (!em) {
5199 5200
		err = -ENOMEM;
		goto out;
5201
	}
5202
	em->bdev = root->fs_info->fs_devices->latest_bdev;
5203
	em->start = EXTENT_MAP_HOLE;
5204
	em->orig_start = EXTENT_MAP_HOLE;
5205
	em->len = (u64)-1;
C
Chris Mason 已提交
5206
	em->block_len = (u64)-1;
5207 5208 5209

	if (!path) {
		path = btrfs_alloc_path();
J
Josef Bacik 已提交
5210 5211 5212 5213 5214 5215 5216 5217 5218
		if (!path) {
			err = -ENOMEM;
			goto out;
		}
		/*
		 * Chances are we'll be called again, so go ahead and do
		 * readahead
		 */
		path->reada = 1;
5219 5220
	}

5221 5222
	ret = btrfs_lookup_file_extent(trans, root, path,
				       objectid, start, trans != NULL);
5223 5224 5225 5226 5227 5228 5229 5230 5231 5232 5233
	if (ret < 0) {
		err = ret;
		goto out;
	}

	if (ret != 0) {
		if (path->slots[0] == 0)
			goto not_found;
		path->slots[0]--;
	}

5234 5235
	leaf = path->nodes[0];
	item = btrfs_item_ptr(leaf, path->slots[0],
5236 5237
			      struct btrfs_file_extent_item);
	/* are we inside the extent that was found? */
5238 5239 5240
	btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
	found_type = btrfs_key_type(&found_key);
	if (found_key.objectid != objectid ||
5241 5242 5243 5244
	    found_type != BTRFS_EXTENT_DATA_KEY) {
		goto not_found;
	}

5245 5246
	found_type = btrfs_file_extent_type(leaf, item);
	extent_start = found_key.offset;
5247
	compress_type = btrfs_file_extent_compression(leaf, item);
Y
Yan Zheng 已提交
5248 5249
	if (found_type == BTRFS_FILE_EXTENT_REG ||
	    found_type == BTRFS_FILE_EXTENT_PREALLOC) {
5250
		extent_end = extent_start +
5251
		       btrfs_file_extent_num_bytes(leaf, item);
Y
Yan Zheng 已提交
5252 5253 5254 5255 5256 5257 5258 5259 5260 5261 5262 5263 5264 5265
	} else if (found_type == BTRFS_FILE_EXTENT_INLINE) {
		size_t size;
		size = btrfs_file_extent_inline_len(leaf, item);
		extent_end = (extent_start + size + root->sectorsize - 1) &
			~((u64)root->sectorsize - 1);
	}

	if (start >= extent_end) {
		path->slots[0]++;
		if (path->slots[0] >= btrfs_header_nritems(leaf)) {
			ret = btrfs_next_leaf(root, path);
			if (ret < 0) {
				err = ret;
				goto out;
5266
			}
Y
Yan Zheng 已提交
5267 5268 5269
			if (ret > 0)
				goto not_found;
			leaf = path->nodes[0];
5270
		}
Y
Yan Zheng 已提交
5271 5272 5273 5274 5275 5276 5277 5278 5279 5280 5281
		btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
		if (found_key.objectid != objectid ||
		    found_key.type != BTRFS_EXTENT_DATA_KEY)
			goto not_found;
		if (start + len <= found_key.offset)
			goto not_found;
		em->start = start;
		em->len = found_key.offset - start;
		goto not_found_em;
	}

Y
Yan Zheng 已提交
5282 5283
	if (found_type == BTRFS_FILE_EXTENT_REG ||
	    found_type == BTRFS_FILE_EXTENT_PREALLOC) {
Y
Yan Zheng 已提交
5284 5285
		em->start = extent_start;
		em->len = extent_end - extent_start;
5286 5287
		em->orig_start = extent_start -
				 btrfs_file_extent_offset(leaf, item);
5288 5289
		bytenr = btrfs_file_extent_disk_bytenr(leaf, item);
		if (bytenr == 0) {
5290
			em->block_start = EXTENT_MAP_HOLE;
5291 5292
			goto insert;
		}
5293
		if (compress_type != BTRFS_COMPRESS_NONE) {
C
Chris Mason 已提交
5294
			set_bit(EXTENT_FLAG_COMPRESSED, &em->flags);
5295
			em->compress_type = compress_type;
C
Chris Mason 已提交
5296 5297 5298 5299 5300 5301 5302
			em->block_start = bytenr;
			em->block_len = btrfs_file_extent_disk_num_bytes(leaf,
									 item);
		} else {
			bytenr += btrfs_file_extent_offset(leaf, item);
			em->block_start = bytenr;
			em->block_len = em->len;
Y
Yan Zheng 已提交
5303 5304
			if (found_type == BTRFS_FILE_EXTENT_PREALLOC)
				set_bit(EXTENT_FLAG_PREALLOC, &em->flags);
C
Chris Mason 已提交
5305
		}
5306 5307
		goto insert;
	} else if (found_type == BTRFS_FILE_EXTENT_INLINE) {
5308
		unsigned long ptr;
5309
		char *map;
5310 5311 5312
		size_t size;
		size_t extent_offset;
		size_t copy_size;
5313

5314
		em->block_start = EXTENT_MAP_INLINE;
C
Chris Mason 已提交
5315
		if (!page || create) {
5316
			em->start = extent_start;
Y
Yan Zheng 已提交
5317
			em->len = extent_end - extent_start;
5318 5319
			goto out;
		}
5320

Y
Yan Zheng 已提交
5321 5322
		size = btrfs_file_extent_inline_len(leaf, item);
		extent_offset = page_offset(page) + pg_offset - extent_start;
5323
		copy_size = min_t(u64, PAGE_CACHE_SIZE - pg_offset,
5324 5325
				size - extent_offset);
		em->start = extent_start + extent_offset;
5326 5327
		em->len = (copy_size + root->sectorsize - 1) &
			~((u64)root->sectorsize - 1);
5328
		em->orig_start = EXTENT_MAP_INLINE;
5329
		if (compress_type) {
C
Chris Mason 已提交
5330
			set_bit(EXTENT_FLAG_COMPRESSED, &em->flags);
5331 5332
			em->compress_type = compress_type;
		}
5333
		ptr = btrfs_file_extent_inline_start(item) + extent_offset;
5334
		if (create == 0 && !PageUptodate(page)) {
5335 5336
			if (btrfs_file_extent_compression(leaf, item) !=
			    BTRFS_COMPRESS_NONE) {
C
Chris Mason 已提交
5337 5338 5339
				ret = uncompress_inline(path, inode, page,
							pg_offset,
							extent_offset, item);
5340
				BUG_ON(ret); /* -ENOMEM */
C
Chris Mason 已提交
5341 5342 5343 5344
			} else {
				map = kmap(page);
				read_extent_buffer(leaf, map + pg_offset, ptr,
						   copy_size);
5345 5346 5347 5348 5349
				if (pg_offset + copy_size < PAGE_CACHE_SIZE) {
					memset(map + pg_offset + copy_size, 0,
					       PAGE_CACHE_SIZE - pg_offset -
					       copy_size);
				}
C
Chris Mason 已提交
5350 5351
				kunmap(page);
			}
5352 5353
			flush_dcache_page(page);
		} else if (create && PageUptodate(page)) {
5354
			BUG();
5355 5356 5357 5358
			if (!trans) {
				kunmap(page);
				free_extent_map(em);
				em = NULL;
C
Chris Mason 已提交
5359

5360
				btrfs_release_path(path);
5361
				trans = btrfs_join_transaction(root);
C
Chris Mason 已提交
5362

5363 5364
				if (IS_ERR(trans))
					return ERR_CAST(trans);
5365 5366
				goto again;
			}
C
Chris Mason 已提交
5367
			map = kmap(page);
5368
			write_extent_buffer(leaf, map + pg_offset, ptr,
5369
					    copy_size);
C
Chris Mason 已提交
5370
			kunmap(page);
5371
			btrfs_mark_buffer_dirty(leaf);
5372
		}
5373
		set_extent_uptodate(io_tree, em->start,
5374
				    extent_map_end(em) - 1, NULL, GFP_NOFS);
5375 5376
		goto insert;
	} else {
C
Chris Mason 已提交
5377
		printk(KERN_ERR "btrfs unknown found_type %d\n", found_type);
5378 5379 5380 5381
		WARN_ON(1);
	}
not_found:
	em->start = start;
5382
	em->len = len;
5383
not_found_em:
5384
	em->block_start = EXTENT_MAP_HOLE;
Y
Yan Zheng 已提交
5385
	set_bit(EXTENT_FLAG_VACANCY, &em->flags);
5386
insert:
5387
	btrfs_release_path(path);
5388
	if (em->start > start || extent_map_end(em) <= start) {
C
Chris Mason 已提交
5389 5390 5391 5392 5393
		printk(KERN_ERR "Btrfs: bad extent! em: [%llu %llu] passed "
		       "[%llu %llu]\n", (unsigned long long)em->start,
		       (unsigned long long)em->len,
		       (unsigned long long)start,
		       (unsigned long long)len);
5394 5395 5396
		err = -EIO;
		goto out;
	}
5397 5398

	err = 0;
5399
	write_lock(&em_tree->lock);
5400
	ret = add_extent_mapping(em_tree, em);
5401 5402 5403 5404
	/* it is possible that someone inserted the extent into the tree
	 * while we had the lock dropped.  It is also possible that
	 * an overlapping map exists in the tree
	 */
5405
	if (ret == -EEXIST) {
5406
		struct extent_map *existing;
5407 5408 5409

		ret = 0;

5410
		existing = lookup_extent_mapping(em_tree, start, len);
5411 5412 5413 5414 5415
		if (existing && (existing->start > start ||
		    existing->start + existing->len <= start)) {
			free_extent_map(existing);
			existing = NULL;
		}
5416 5417 5418 5419 5420
		if (!existing) {
			existing = lookup_extent_mapping(em_tree, em->start,
							 em->len);
			if (existing) {
				err = merge_extent_mapping(em_tree, existing,
5421 5422
							   em, start,
							   root->sectorsize);
5423 5424 5425 5426 5427 5428 5429 5430 5431 5432 5433 5434 5435
				free_extent_map(existing);
				if (err) {
					free_extent_map(em);
					em = NULL;
				}
			} else {
				err = -EIO;
				free_extent_map(em);
				em = NULL;
			}
		} else {
			free_extent_map(em);
			em = existing;
5436
			err = 0;
5437 5438
		}
	}
5439
	write_unlock(&em_tree->lock);
5440
out:
5441 5442 5443

	trace_btrfs_get_extent(root, em);

5444 5445
	if (path)
		btrfs_free_path(path);
5446 5447
	if (trans) {
		ret = btrfs_end_transaction(trans, root);
C
Chris Mason 已提交
5448
		if (!err)
5449 5450 5451 5452 5453 5454
			err = ret;
	}
	if (err) {
		free_extent_map(em);
		return ERR_PTR(err);
	}
5455
	BUG_ON(!em); /* Error is always set */
5456 5457 5458
	return em;
}

5459 5460 5461 5462 5463 5464 5465 5466 5467 5468 5469 5470 5471 5472 5473 5474 5475 5476 5477 5478 5479 5480 5481 5482 5483 5484 5485 5486 5487 5488 5489 5490 5491 5492 5493 5494 5495 5496 5497 5498 5499 5500 5501 5502 5503 5504 5505 5506 5507 5508 5509 5510 5511 5512 5513 5514 5515 5516 5517 5518 5519 5520
struct extent_map *btrfs_get_extent_fiemap(struct inode *inode, struct page *page,
					   size_t pg_offset, u64 start, u64 len,
					   int create)
{
	struct extent_map *em;
	struct extent_map *hole_em = NULL;
	u64 range_start = start;
	u64 end;
	u64 found;
	u64 found_end;
	int err = 0;

	em = btrfs_get_extent(inode, page, pg_offset, start, len, create);
	if (IS_ERR(em))
		return em;
	if (em) {
		/*
		 * if our em maps to a hole, there might
		 * actually be delalloc bytes behind it
		 */
		if (em->block_start != EXTENT_MAP_HOLE)
			return em;
		else
			hole_em = em;
	}

	/* check to see if we've wrapped (len == -1 or similar) */
	end = start + len;
	if (end < start)
		end = (u64)-1;
	else
		end -= 1;

	em = NULL;

	/* ok, we didn't find anything, lets look for delalloc */
	found = count_range_bits(&BTRFS_I(inode)->io_tree, &range_start,
				 end, len, EXTENT_DELALLOC, 1);
	found_end = range_start + found;
	if (found_end < range_start)
		found_end = (u64)-1;

	/*
	 * we didn't find anything useful, return
	 * the original results from get_extent()
	 */
	if (range_start > end || found_end <= start) {
		em = hole_em;
		hole_em = NULL;
		goto out;
	}

	/* adjust the range_start to make sure it doesn't
	 * go backwards from the start they passed in
	 */
	range_start = max(start,range_start);
	found = found_end - range_start;

	if (found > 0) {
		u64 hole_start = start;
		u64 hole_len = len;

5521
		em = alloc_extent_map();
5522 5523 5524 5525 5526 5527 5528 5529 5530 5531 5532 5533 5534 5535 5536 5537 5538 5539 5540 5541 5542 5543 5544 5545 5546 5547 5548 5549 5550 5551 5552 5553 5554 5555 5556 5557 5558 5559 5560 5561 5562 5563 5564 5565 5566 5567 5568 5569 5570 5571 5572 5573 5574 5575 5576 5577 5578 5579 5580
		if (!em) {
			err = -ENOMEM;
			goto out;
		}
		/*
		 * when btrfs_get_extent can't find anything it
		 * returns one huge hole
		 *
		 * make sure what it found really fits our range, and
		 * adjust to make sure it is based on the start from
		 * the caller
		 */
		if (hole_em) {
			u64 calc_end = extent_map_end(hole_em);

			if (calc_end <= start || (hole_em->start > end)) {
				free_extent_map(hole_em);
				hole_em = NULL;
			} else {
				hole_start = max(hole_em->start, start);
				hole_len = calc_end - hole_start;
			}
		}
		em->bdev = NULL;
		if (hole_em && range_start > hole_start) {
			/* our hole starts before our delalloc, so we
			 * have to return just the parts of the hole
			 * that go until  the delalloc starts
			 */
			em->len = min(hole_len,
				      range_start - hole_start);
			em->start = hole_start;
			em->orig_start = hole_start;
			/*
			 * don't adjust block start at all,
			 * it is fixed at EXTENT_MAP_HOLE
			 */
			em->block_start = hole_em->block_start;
			em->block_len = hole_len;
		} else {
			em->start = range_start;
			em->len = found;
			em->orig_start = range_start;
			em->block_start = EXTENT_MAP_DELALLOC;
			em->block_len = found;
		}
	} else if (hole_em) {
		return hole_em;
	}
out:

	free_extent_map(hole_em);
	if (err) {
		free_extent_map(em);
		return ERR_PTR(err);
	}
	return em;
}

5581
static struct extent_map *btrfs_new_extent_direct(struct inode *inode,
5582
						  struct extent_map *em,
5583 5584 5585 5586 5587 5588 5589 5590
						  u64 start, u64 len)
{
	struct btrfs_root *root = BTRFS_I(inode)->root;
	struct btrfs_trans_handle *trans;
	struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
	struct btrfs_key ins;
	u64 alloc_hint;
	int ret;
5591
	bool insert = false;
5592

5593 5594 5595 5596 5597 5598 5599 5600 5601 5602 5603 5604 5605
	/*
	 * Ok if the extent map we looked up is a hole and is for the exact
	 * range we want, there is no reason to allocate a new one, however if
	 * it is not right then we need to free this one and drop the cache for
	 * our range.
	 */
	if (em->block_start != EXTENT_MAP_HOLE || em->start != start ||
	    em->len != len) {
		free_extent_map(em);
		em = NULL;
		insert = true;
		btrfs_drop_extent_cache(inode, start, start + len - 1, 0);
	}
5606

5607
	trans = btrfs_join_transaction(root);
5608 5609
	if (IS_ERR(trans))
		return ERR_CAST(trans);
5610

C
Chris Mason 已提交
5611 5612 5613
	if (start <= BTRFS_I(inode)->disk_i_size && len < 64 * 1024)
		btrfs_add_inode_defrag(trans, inode);

5614 5615 5616 5617
	trans->block_rsv = &root->fs_info->delalloc_block_rsv;

	alloc_hint = get_extent_allocation_hint(inode, start, len);
	ret = btrfs_reserve_extent(trans, root, len, root->sectorsize, 0,
5618
				   alloc_hint, &ins, 1);
5619 5620 5621 5622 5623 5624
	if (ret) {
		em = ERR_PTR(ret);
		goto out;
	}

	if (!em) {
5625
		em = alloc_extent_map();
5626 5627 5628 5629
		if (!em) {
			em = ERR_PTR(-ENOMEM);
			goto out;
		}
5630 5631 5632 5633 5634 5635 5636 5637 5638
	}

	em->start = start;
	em->orig_start = em->start;
	em->len = ins.offset;

	em->block_start = ins.objectid;
	em->block_len = ins.offset;
	em->bdev = root->fs_info->fs_devices->latest_bdev;
5639 5640 5641 5642 5643 5644

	/*
	 * We need to do this because if we're using the original em we searched
	 * for, we could have EXTENT_FLAG_VACANCY set, and we don't want that.
	 */
	em->flags = 0;
5645 5646
	set_bit(EXTENT_FLAG_PINNED, &em->flags);

5647
	while (insert) {
5648 5649 5650 5651 5652 5653 5654 5655 5656 5657 5658 5659 5660 5661 5662 5663 5664 5665 5666
		write_lock(&em_tree->lock);
		ret = add_extent_mapping(em_tree, em);
		write_unlock(&em_tree->lock);
		if (ret != -EEXIST)
			break;
		btrfs_drop_extent_cache(inode, start, start + em->len - 1, 0);
	}

	ret = btrfs_add_ordered_extent_dio(inode, start, ins.objectid,
					   ins.offset, ins.offset, 0);
	if (ret) {
		btrfs_free_reserved_extent(root, ins.objectid, ins.offset);
		em = ERR_PTR(ret);
	}
out:
	btrfs_end_transaction(trans, root);
	return em;
}

5667 5668 5669 5670 5671 5672 5673 5674 5675 5676 5677 5678 5679 5680 5681 5682 5683 5684 5685 5686 5687 5688 5689 5690
/*
 * returns 1 when the nocow is safe, < 1 on error, 0 if the
 * block must be cow'd
 */
static noinline int can_nocow_odirect(struct btrfs_trans_handle *trans,
				      struct inode *inode, u64 offset, u64 len)
{
	struct btrfs_path *path;
	int ret;
	struct extent_buffer *leaf;
	struct btrfs_root *root = BTRFS_I(inode)->root;
	struct btrfs_file_extent_item *fi;
	struct btrfs_key key;
	u64 disk_bytenr;
	u64 backref_offset;
	u64 extent_end;
	u64 num_bytes;
	int slot;
	int found_type;

	path = btrfs_alloc_path();
	if (!path)
		return -ENOMEM;

L
Li Zefan 已提交
5691
	ret = btrfs_lookup_file_extent(trans, root, path, btrfs_ino(inode),
5692 5693 5694 5695 5696 5697 5698 5699 5700 5701 5702 5703 5704 5705 5706 5707
				       offset, 0);
	if (ret < 0)
		goto out;

	slot = path->slots[0];
	if (ret == 1) {
		if (slot == 0) {
			/* can't find the item, must cow */
			ret = 0;
			goto out;
		}
		slot--;
	}
	ret = 0;
	leaf = path->nodes[0];
	btrfs_item_key_to_cpu(leaf, &key, slot);
L
Li Zefan 已提交
5708
	if (key.objectid != btrfs_ino(inode) ||
5709 5710 5711 5712 5713 5714 5715 5716 5717 5718 5719 5720 5721 5722 5723 5724 5725 5726 5727 5728 5729 5730 5731 5732 5733 5734 5735 5736 5737 5738 5739 5740 5741
	    key.type != BTRFS_EXTENT_DATA_KEY) {
		/* not our file or wrong item type, must cow */
		goto out;
	}

	if (key.offset > offset) {
		/* Wrong offset, must cow */
		goto out;
	}

	fi = btrfs_item_ptr(leaf, slot, struct btrfs_file_extent_item);
	found_type = btrfs_file_extent_type(leaf, fi);
	if (found_type != BTRFS_FILE_EXTENT_REG &&
	    found_type != BTRFS_FILE_EXTENT_PREALLOC) {
		/* not a regular extent, must cow */
		goto out;
	}
	disk_bytenr = btrfs_file_extent_disk_bytenr(leaf, fi);
	backref_offset = btrfs_file_extent_offset(leaf, fi);

	extent_end = key.offset + btrfs_file_extent_num_bytes(leaf, fi);
	if (extent_end < offset + len) {
		/* extent doesn't include our full range, must cow */
		goto out;
	}

	if (btrfs_extent_readonly(root, disk_bytenr))
		goto out;

	/*
	 * look for other files referencing this extent, if we
	 * find any we must cow
	 */
L
Li Zefan 已提交
5742
	if (btrfs_cross_ref_exist(trans, root, btrfs_ino(inode),
5743 5744 5745 5746 5747 5748 5749 5750 5751 5752 5753 5754 5755 5756 5757 5758 5759 5760 5761 5762 5763 5764 5765 5766
				  key.offset - backref_offset, disk_bytenr))
		goto out;

	/*
	 * adjust disk_bytenr and num_bytes to cover just the bytes
	 * in this extent we are about to write.  If there
	 * are any csums in that range we have to cow in order
	 * to keep the csums correct
	 */
	disk_bytenr += backref_offset;
	disk_bytenr += offset - key.offset;
	num_bytes = min(offset + len, extent_end) - offset;
	if (csum_exist_in_range(root, disk_bytenr, num_bytes))
				goto out;
	/*
	 * all of the above have passed, it is safe to overwrite this extent
	 * without cow
	 */
	ret = 1;
out:
	btrfs_free_path(path);
	return ret;
}

5767 5768 5769 5770 5771 5772 5773
static int btrfs_get_blocks_direct(struct inode *inode, sector_t iblock,
				   struct buffer_head *bh_result, int create)
{
	struct extent_map *em;
	struct btrfs_root *root = BTRFS_I(inode)->root;
	u64 start = iblock << inode->i_blkbits;
	u64 len = bh_result->b_size;
5774
	struct btrfs_trans_handle *trans;
5775 5776 5777 5778 5779 5780 5781 5782 5783 5784 5785 5786 5787 5788 5789 5790 5791 5792 5793 5794 5795 5796 5797 5798 5799 5800 5801 5802 5803 5804 5805

	em = btrfs_get_extent(inode, NULL, 0, start, len, 0);
	if (IS_ERR(em))
		return PTR_ERR(em);

	/*
	 * Ok for INLINE and COMPRESSED extents we need to fallback on buffered
	 * io.  INLINE is special, and we could probably kludge it in here, but
	 * it's still buffered so for safety lets just fall back to the generic
	 * buffered path.
	 *
	 * For COMPRESSED we _have_ to read the entire extent in so we can
	 * decompress it, so there will be buffering required no matter what we
	 * do, so go ahead and fallback to buffered.
	 *
	 * We return -ENOTBLK because thats what makes DIO go ahead and go back
	 * to buffered IO.  Don't blame me, this is the price we pay for using
	 * the generic code.
	 */
	if (test_bit(EXTENT_FLAG_COMPRESSED, &em->flags) ||
	    em->block_start == EXTENT_MAP_INLINE) {
		free_extent_map(em);
		return -ENOTBLK;
	}

	/* Just a good old fashioned hole, return */
	if (!create && (em->block_start == EXTENT_MAP_HOLE ||
			test_bit(EXTENT_FLAG_PREALLOC, &em->flags))) {
		free_extent_map(em);
		/* DIO will do one hole at a time, so just unlock a sector */
		unlock_extent(&BTRFS_I(inode)->io_tree, start,
5806
			      start + root->sectorsize - 1);
5807 5808 5809 5810 5811 5812 5813 5814 5815 5816 5817 5818
		return 0;
	}

	/*
	 * We don't allocate a new extent in the following cases
	 *
	 * 1) The inode is marked as NODATACOW.  In this case we'll just use the
	 * existing extent.
	 * 2) The extent is marked as PREALLOC.  We're good to go here and can
	 * just use the extent.
	 *
	 */
5819 5820
	if (!create) {
		len = em->len - (start - em->start);
5821
		goto map;
5822
	}
5823 5824 5825 5826 5827 5828

	if (test_bit(EXTENT_FLAG_PREALLOC, &em->flags) ||
	    ((BTRFS_I(inode)->flags & BTRFS_INODE_NODATACOW) &&
	     em->block_start != EXTENT_MAP_HOLE)) {
		int type;
		int ret;
5829
		u64 block_start;
5830 5831 5832 5833 5834

		if (test_bit(EXTENT_FLAG_PREALLOC, &em->flags))
			type = BTRFS_ORDERED_PREALLOC;
		else
			type = BTRFS_ORDERED_NOCOW;
5835
		len = min(len, em->len - (start - em->start));
5836
		block_start = em->block_start + (start - em->start);
5837 5838 5839 5840 5841 5842

		/*
		 * we're not going to log anything, but we do need
		 * to make sure the current transaction stays open
		 * while we look for nocow cross refs
		 */
5843
		trans = btrfs_join_transaction(root);
5844
		if (IS_ERR(trans))
5845 5846 5847 5848 5849 5850 5851 5852 5853 5854 5855
			goto must_cow;

		if (can_nocow_odirect(trans, inode, start, len) == 1) {
			ret = btrfs_add_ordered_extent_dio(inode, start,
					   block_start, len, len, type);
			btrfs_end_transaction(trans, root);
			if (ret) {
				free_extent_map(em);
				return ret;
			}
			goto unlock;
5856
		}
5857
		btrfs_end_transaction(trans, root);
5858
	}
5859 5860 5861 5862 5863 5864
must_cow:
	/*
	 * this will cow the extent, reset the len in case we changed
	 * it above
	 */
	len = bh_result->b_size;
5865
	em = btrfs_new_extent_direct(inode, em, start, len);
5866 5867 5868 5869
	if (IS_ERR(em))
		return PTR_ERR(em);
	len = min(len, em->len - (start - em->start));
unlock:
5870 5871 5872
	clear_extent_bit(&BTRFS_I(inode)->io_tree, start, start + len - 1,
			  EXTENT_LOCKED | EXTENT_DELALLOC | EXTENT_DIRTY, 1,
			  0, NULL, GFP_NOFS);
5873 5874 5875
map:
	bh_result->b_blocknr = (em->block_start + (start - em->start)) >>
		inode->i_blkbits;
5876
	bh_result->b_size = len;
5877 5878
	bh_result->b_bdev = em->bdev;
	set_buffer_mapped(bh_result);
5879 5880 5881 5882 5883 5884 5885 5886 5887 5888 5889
	if (create) {
		if (!test_bit(EXTENT_FLAG_PREALLOC, &em->flags))
			set_buffer_new(bh_result);

		/*
		 * Need to update the i_size under the extent lock so buffered
		 * readers will get the updated i_size when we unlock.
		 */
		if (start + len > i_size_read(inode))
			i_size_write(inode, start + len);
	}
5890 5891 5892 5893 5894 5895 5896 5897 5898 5899 5900 5901 5902

	free_extent_map(em);

	return 0;
}

struct btrfs_dio_private {
	struct inode *inode;
	u64 logical_offset;
	u64 disk_bytenr;
	u64 bytes;
	u32 *csums;
	void *private;
M
Miao Xie 已提交
5903 5904 5905 5906 5907 5908 5909 5910

	/* number of bios pending for this dio */
	atomic_t pending_bios;

	/* IO errors */
	int errors;

	struct bio *orig_bio;
5911 5912 5913 5914
};

static void btrfs_endio_direct_read(struct bio *bio, int err)
{
M
Miao Xie 已提交
5915
	struct btrfs_dio_private *dip = bio->bi_private;
5916 5917 5918 5919 5920 5921 5922 5923 5924 5925 5926 5927 5928 5929 5930 5931
	struct bio_vec *bvec_end = bio->bi_io_vec + bio->bi_vcnt - 1;
	struct bio_vec *bvec = bio->bi_io_vec;
	struct inode *inode = dip->inode;
	struct btrfs_root *root = BTRFS_I(inode)->root;
	u64 start;
	u32 *private = dip->csums;

	start = dip->logical_offset;
	do {
		if (!(BTRFS_I(inode)->flags & BTRFS_INODE_NODATASUM)) {
			struct page *page = bvec->bv_page;
			char *kaddr;
			u32 csum = ~(u32)0;
			unsigned long flags;

			local_irq_save(flags);
5932
			kaddr = kmap_atomic(page);
5933 5934 5935
			csum = btrfs_csum_data(root, kaddr + bvec->bv_offset,
					       csum, bvec->bv_len);
			btrfs_csum_final(csum, (char *)&csum);
5936
			kunmap_atomic(kaddr);
5937 5938 5939 5940
			local_irq_restore(flags);

			flush_dcache_page(bvec->bv_page);
			if (csum != *private) {
L
Li Zefan 已提交
5941
				printk(KERN_ERR "btrfs csum failed ino %llu off"
5942
				      " %llu csum %u private %u\n",
L
Li Zefan 已提交
5943 5944
				      (unsigned long long)btrfs_ino(inode),
				      (unsigned long long)start,
5945 5946 5947 5948 5949 5950 5951 5952 5953 5954 5955
				      csum, *private);
				err = -EIO;
			}
		}

		start += bvec->bv_len;
		private++;
		bvec++;
	} while (bvec <= bvec_end);

	unlock_extent(&BTRFS_I(inode)->io_tree, dip->logical_offset,
5956
		      dip->logical_offset + dip->bytes - 1);
5957 5958 5959 5960
	bio->bi_private = dip->private;

	kfree(dip->csums);
	kfree(dip);
5961 5962 5963 5964

	/* If we had a csum failure make sure to clear the uptodate flag */
	if (err)
		clear_bit(BIO_UPTODATE, &bio->bi_flags);
5965 5966 5967 5968 5969 5970 5971 5972 5973
	dio_end_io(bio, err);
}

static void btrfs_endio_direct_write(struct bio *bio, int err)
{
	struct btrfs_dio_private *dip = bio->bi_private;
	struct inode *inode = dip->inode;
	struct btrfs_root *root = BTRFS_I(inode)->root;
	struct btrfs_ordered_extent *ordered = NULL;
5974 5975
	u64 ordered_offset = dip->logical_offset;
	u64 ordered_bytes = dip->bytes;
5976 5977 5978 5979
	int ret;

	if (err)
		goto out_done;
5980 5981 5982
again:
	ret = btrfs_dec_test_first_ordered_pending(inode, &ordered,
						   &ordered_offset,
5983
						   ordered_bytes, !err);
5984
	if (!ret)
5985
		goto out_test;
5986

5987 5988 5989 5990
	ordered->work.func = finish_ordered_fn;
	ordered->work.flags = 0;
	btrfs_queue_worker(&root->fs_info->endio_write_workers,
			   &ordered->work);
5991 5992 5993 5994 5995 5996 5997 5998
out_test:
	/*
	 * our bio might span multiple ordered extents.  If we haven't
	 * completed the accounting for the whole dio, go back and try again
	 */
	if (ordered_offset < dip->logical_offset + dip->bytes) {
		ordered_bytes = dip->logical_offset + dip->bytes -
			ordered_offset;
5999
		ordered = NULL;
6000 6001
		goto again;
	}
6002 6003 6004 6005
out_done:
	bio->bi_private = dip->private;

	kfree(dip);
6006 6007 6008 6009

	/* If we had an error make sure to clear the uptodate flag */
	if (err)
		clear_bit(BIO_UPTODATE, &bio->bi_flags);
6010 6011 6012
	dio_end_io(bio, err);
}

6013 6014 6015 6016 6017 6018 6019
static int __btrfs_submit_bio_start_direct_io(struct inode *inode, int rw,
				    struct bio *bio, int mirror_num,
				    unsigned long bio_flags, u64 offset)
{
	int ret;
	struct btrfs_root *root = BTRFS_I(inode)->root;
	ret = btrfs_csum_one_bio(root, inode, bio, offset, 1);
6020
	BUG_ON(ret); /* -ENOMEM */
6021 6022 6023
	return 0;
}

M
Miao Xie 已提交
6024 6025 6026 6027 6028
static void btrfs_end_dio_bio(struct bio *bio, int err)
{
	struct btrfs_dio_private *dip = bio->bi_private;

	if (err) {
L
Li Zefan 已提交
6029
		printk(KERN_ERR "btrfs direct IO failed ino %llu rw %lu "
J
Jan Beulich 已提交
6030
		      "sector %#Lx len %u err no %d\n",
L
Li Zefan 已提交
6031
		      (unsigned long long)btrfs_ino(dip->inode), bio->bi_rw,
J
Jan Beulich 已提交
6032
		      (unsigned long long)bio->bi_sector, bio->bi_size, err);
M
Miao Xie 已提交
6033 6034 6035 6036 6037 6038 6039 6040 6041 6042 6043 6044 6045 6046 6047 6048 6049 6050 6051 6052 6053 6054 6055 6056 6057 6058 6059 6060 6061 6062 6063 6064
		dip->errors = 1;

		/*
		 * before atomic variable goto zero, we must make sure
		 * dip->errors is perceived to be set.
		 */
		smp_mb__before_atomic_dec();
	}

	/* if there are more bios still pending for this dio, just exit */
	if (!atomic_dec_and_test(&dip->pending_bios))
		goto out;

	if (dip->errors)
		bio_io_error(dip->orig_bio);
	else {
		set_bit(BIO_UPTODATE, &dip->orig_bio->bi_flags);
		bio_endio(dip->orig_bio, 0);
	}
out:
	bio_put(bio);
}

static struct bio *btrfs_dio_bio_alloc(struct block_device *bdev,
				       u64 first_sector, gfp_t gfp_flags)
{
	int nr_vecs = bio_get_nr_vecs(bdev);
	return btrfs_bio_alloc(bdev, first_sector, nr_vecs, gfp_flags);
}

static inline int __btrfs_submit_dio_bio(struct bio *bio, struct inode *inode,
					 int rw, u64 file_offset, int skip_sum,
6065
					 u32 *csums, int async_submit)
M
Miao Xie 已提交
6066 6067 6068 6069 6070 6071
{
	int write = rw & REQ_WRITE;
	struct btrfs_root *root = BTRFS_I(inode)->root;
	int ret;

	bio_get(bio);
6072 6073 6074 6075 6076 6077

	if (!write) {
		ret = btrfs_bio_wq_end_io(root->fs_info, bio, 0);
		if (ret)
			goto err;
	}
M
Miao Xie 已提交
6078

6079 6080 6081 6082
	if (skip_sum)
		goto map;

	if (write && async_submit) {
M
Miao Xie 已提交
6083 6084 6085 6086 6087 6088
		ret = btrfs_wq_submit_bio(root->fs_info,
				   inode, rw, bio, 0, 0,
				   file_offset,
				   __btrfs_submit_bio_start_direct_io,
				   __btrfs_submit_bio_done);
		goto err;
6089 6090 6091 6092 6093 6094 6095 6096
	} else if (write) {
		/*
		 * If we aren't doing async submit, calculate the csum of the
		 * bio now.
		 */
		ret = btrfs_csum_one_bio(root, inode, bio, file_offset, 1);
		if (ret)
			goto err;
6097 6098
	} else if (!skip_sum) {
		ret = btrfs_lookup_bio_sums_dio(root, inode, bio,
M
Miao Xie 已提交
6099
					  file_offset, csums);
6100 6101 6102
		if (ret)
			goto err;
	}
M
Miao Xie 已提交
6103

6104 6105
map:
	ret = btrfs_map_bio(root, rw, bio, 0, async_submit);
M
Miao Xie 已提交
6106 6107 6108 6109 6110 6111 6112 6113 6114 6115 6116 6117 6118 6119 6120 6121 6122 6123 6124 6125 6126
err:
	bio_put(bio);
	return ret;
}

static int btrfs_submit_direct_hook(int rw, struct btrfs_dio_private *dip,
				    int skip_sum)
{
	struct inode *inode = dip->inode;
	struct btrfs_root *root = BTRFS_I(inode)->root;
	struct btrfs_mapping_tree *map_tree = &root->fs_info->mapping_tree;
	struct bio *bio;
	struct bio *orig_bio = dip->orig_bio;
	struct bio_vec *bvec = orig_bio->bi_io_vec;
	u64 start_sector = orig_bio->bi_sector;
	u64 file_offset = dip->logical_offset;
	u64 submit_len = 0;
	u64 map_length;
	int nr_pages = 0;
	u32 *csums = dip->csums;
	int ret = 0;
6127
	int async_submit = 0;
6128
	int write = rw & REQ_WRITE;
M
Miao Xie 已提交
6129 6130 6131 6132 6133

	map_length = orig_bio->bi_size;
	ret = btrfs_map_block(map_tree, READ, start_sector << 9,
			      &map_length, NULL, 0);
	if (ret) {
6134
		bio_put(orig_bio);
M
Miao Xie 已提交
6135 6136 6137
		return -EIO;
	}

6138 6139 6140 6141 6142
	if (map_length >= orig_bio->bi_size) {
		bio = orig_bio;
		goto submit;
	}

6143
	async_submit = 1;
6144 6145 6146 6147 6148 6149 6150
	bio = btrfs_dio_bio_alloc(orig_bio->bi_bdev, start_sector, GFP_NOFS);
	if (!bio)
		return -ENOMEM;
	bio->bi_private = dip;
	bio->bi_end_io = btrfs_end_dio_bio;
	atomic_inc(&dip->pending_bios);

M
Miao Xie 已提交
6151 6152 6153 6154 6155 6156 6157 6158 6159 6160 6161 6162 6163
	while (bvec <= (orig_bio->bi_io_vec + orig_bio->bi_vcnt - 1)) {
		if (unlikely(map_length < submit_len + bvec->bv_len ||
		    bio_add_page(bio, bvec->bv_page, bvec->bv_len,
				 bvec->bv_offset) < bvec->bv_len)) {
			/*
			 * inc the count before we submit the bio so
			 * we know the end IO handler won't happen before
			 * we inc the count. Otherwise, the dip might get freed
			 * before we're done setting it up
			 */
			atomic_inc(&dip->pending_bios);
			ret = __btrfs_submit_dio_bio(bio, inode, rw,
						     file_offset, skip_sum,
6164
						     csums, async_submit);
M
Miao Xie 已提交
6165 6166 6167 6168 6169 6170
			if (ret) {
				bio_put(bio);
				atomic_dec(&dip->pending_bios);
				goto out_err;
			}

6171 6172
			/* Write's use the ordered csums */
			if (!write && !skip_sum)
M
Miao Xie 已提交
6173 6174 6175 6176 6177 6178 6179 6180 6181 6182 6183 6184 6185 6186 6187 6188 6189 6190 6191 6192 6193 6194 6195 6196 6197 6198 6199 6200
				csums = csums + nr_pages;
			start_sector += submit_len >> 9;
			file_offset += submit_len;

			submit_len = 0;
			nr_pages = 0;

			bio = btrfs_dio_bio_alloc(orig_bio->bi_bdev,
						  start_sector, GFP_NOFS);
			if (!bio)
				goto out_err;
			bio->bi_private = dip;
			bio->bi_end_io = btrfs_end_dio_bio;

			map_length = orig_bio->bi_size;
			ret = btrfs_map_block(map_tree, READ, start_sector << 9,
					      &map_length, NULL, 0);
			if (ret) {
				bio_put(bio);
				goto out_err;
			}
		} else {
			submit_len += bvec->bv_len;
			nr_pages ++;
			bvec++;
		}
	}

6201
submit:
M
Miao Xie 已提交
6202
	ret = __btrfs_submit_dio_bio(bio, inode, rw, file_offset, skip_sum,
6203
				     csums, async_submit);
M
Miao Xie 已提交
6204 6205 6206 6207 6208 6209 6210 6211 6212 6213 6214 6215 6216 6217 6218 6219 6220 6221
	if (!ret)
		return 0;

	bio_put(bio);
out_err:
	dip->errors = 1;
	/*
	 * before atomic variable goto zero, we must
	 * make sure dip->errors is perceived to be set.
	 */
	smp_mb__before_atomic_dec();
	if (atomic_dec_and_test(&dip->pending_bios))
		bio_io_error(dip->orig_bio);

	/* bio_end_io() will handle error, so we needn't return it */
	return 0;
}

6222 6223 6224 6225 6226 6227 6228
static void btrfs_submit_direct(int rw, struct bio *bio, struct inode *inode,
				loff_t file_offset)
{
	struct btrfs_root *root = BTRFS_I(inode)->root;
	struct btrfs_dio_private *dip;
	struct bio_vec *bvec = bio->bi_io_vec;
	int skip_sum;
6229
	int write = rw & REQ_WRITE;
6230 6231 6232 6233 6234 6235 6236 6237 6238 6239 6240
	int ret = 0;

	skip_sum = BTRFS_I(inode)->flags & BTRFS_INODE_NODATASUM;

	dip = kmalloc(sizeof(*dip), GFP_NOFS);
	if (!dip) {
		ret = -ENOMEM;
		goto free_ordered;
	}
	dip->csums = NULL;

6241 6242
	/* Write's use the ordered csum stuff, so we don't need dip->csums */
	if (!write && !skip_sum) {
6243 6244
		dip->csums = kmalloc(sizeof(u32) * bio->bi_vcnt, GFP_NOFS);
		if (!dip->csums) {
D
Daniel J Blueman 已提交
6245
			kfree(dip);
6246 6247 6248 6249 6250 6251 6252 6253 6254 6255 6256 6257 6258 6259 6260
			ret = -ENOMEM;
			goto free_ordered;
		}
	}

	dip->private = bio->bi_private;
	dip->inode = inode;
	dip->logical_offset = file_offset;

	dip->bytes = 0;
	do {
		dip->bytes += bvec->bv_len;
		bvec++;
	} while (bvec <= (bio->bi_io_vec + bio->bi_vcnt - 1));

6261
	dip->disk_bytenr = (u64)bio->bi_sector << 9;
6262
	bio->bi_private = dip;
M
Miao Xie 已提交
6263 6264 6265
	dip->errors = 0;
	dip->orig_bio = bio;
	atomic_set(&dip->pending_bios, 0);
6266 6267 6268 6269 6270 6271

	if (write)
		bio->bi_end_io = btrfs_endio_direct_write;
	else
		bio->bi_end_io = btrfs_endio_direct_read;

M
Miao Xie 已提交
6272 6273
	ret = btrfs_submit_direct_hook(rw, dip, skip_sum);
	if (!ret)
6274
		return;
6275 6276 6277 6278 6279 6280 6281
free_ordered:
	/*
	 * If this is a write, we need to clean up the reserved space and kill
	 * the ordered extent.
	 */
	if (write) {
		struct btrfs_ordered_extent *ordered;
6282
		ordered = btrfs_lookup_ordered_extent(inode, file_offset);
6283 6284 6285 6286 6287 6288 6289 6290 6291 6292
		if (!test_bit(BTRFS_ORDERED_PREALLOC, &ordered->flags) &&
		    !test_bit(BTRFS_ORDERED_NOCOW, &ordered->flags))
			btrfs_free_reserved_extent(root, ordered->start,
						   ordered->disk_len);
		btrfs_put_ordered_extent(ordered);
		btrfs_put_ordered_extent(ordered);
	}
	bio_endio(bio, ret);
}

C
Chris Mason 已提交
6293 6294 6295 6296 6297
static ssize_t check_direct_IO(struct btrfs_root *root, int rw, struct kiocb *iocb,
			const struct iovec *iov, loff_t offset,
			unsigned long nr_segs)
{
	int seg;
6298
	int i;
C
Chris Mason 已提交
6299 6300 6301 6302 6303 6304 6305 6306 6307 6308 6309 6310 6311 6312
	size_t size;
	unsigned long addr;
	unsigned blocksize_mask = root->sectorsize - 1;
	ssize_t retval = -EINVAL;
	loff_t end = offset;

	if (offset & blocksize_mask)
		goto out;

	/* Check the memory alignment.  Blocks cannot straddle pages */
	for (seg = 0; seg < nr_segs; seg++) {
		addr = (unsigned long)iov[seg].iov_base;
		size = iov[seg].iov_len;
		end += size;
6313
		if ((addr & blocksize_mask) || (size & blocksize_mask))
C
Chris Mason 已提交
6314
			goto out;
6315 6316 6317 6318 6319 6320 6321 6322 6323 6324 6325 6326 6327 6328

		/* If this is a write we don't need to check anymore */
		if (rw & WRITE)
			continue;

		/*
		 * Check to make sure we don't have duplicate iov_base's in this
		 * iovec, if so return EINVAL, otherwise we'll get csum errors
		 * when reading back.
		 */
		for (i = seg + 1; i < nr_segs; i++) {
			if (iov[seg].iov_base == iov[i].iov_base)
				goto out;
		}
C
Chris Mason 已提交
6329 6330 6331 6332 6333
	}
	retval = 0;
out:
	return retval;
}
6334 6335 6336 6337
static ssize_t btrfs_direct_IO(int rw, struct kiocb *iocb,
			const struct iovec *iov, loff_t offset,
			unsigned long nr_segs)
{
6338 6339 6340
	struct file *file = iocb->ki_filp;
	struct inode *inode = file->f_mapping->host;
	struct btrfs_ordered_extent *ordered;
6341
	struct extent_state *cached_state = NULL;
6342 6343
	u64 lockstart, lockend;
	ssize_t ret;
6344 6345
	int writing = rw & WRITE;
	int write_bits = 0;
6346
	size_t count = iov_length(iov, nr_segs);
6347

C
Chris Mason 已提交
6348 6349 6350 6351 6352
	if (check_direct_IO(BTRFS_I(inode)->root, rw, iocb, iov,
			    offset, nr_segs)) {
		return 0;
	}

6353
	lockstart = offset;
6354 6355 6356 6357 6358 6359 6360
	lockend = offset + count - 1;

	if (writing) {
		ret = btrfs_delalloc_reserve_space(inode, count);
		if (ret)
			goto out;
	}
6361

6362
	while (1) {
6363
		lock_extent_bits(&BTRFS_I(inode)->io_tree, lockstart, lockend,
6364
				 0, &cached_state);
6365 6366 6367 6368 6369 6370 6371
		/*
		 * We're concerned with the entire range that we're going to be
		 * doing DIO to, so we need to make sure theres no ordered
		 * extents in this range.
		 */
		ordered = btrfs_lookup_ordered_range(inode, lockstart,
						     lockend - lockstart + 1);
6372 6373 6374 6375 6376 6377 6378 6379 6380 6381 6382 6383

		/*
		 * We need to make sure there are no buffered pages in this
		 * range either, we could have raced between the invalidate in
		 * generic_file_direct_write and locking the extent.  The
		 * invalidate needs to happen so that reads after a write do not
		 * get stale data.
		 */
		if (!ordered && (!writing ||
		    !test_range_bit(&BTRFS_I(inode)->io_tree,
				    lockstart, lockend, EXTENT_UPTODATE, 0,
				    cached_state)))
6384
			break;
6385

6386 6387
		unlock_extent_cached(&BTRFS_I(inode)->io_tree, lockstart, lockend,
				     &cached_state, GFP_NOFS);
6388 6389 6390 6391 6392 6393 6394 6395 6396 6397 6398 6399 6400 6401 6402 6403 6404 6405 6406 6407 6408 6409 6410 6411 6412 6413

		if (ordered) {
			btrfs_start_ordered_extent(inode, ordered, 1);
			btrfs_put_ordered_extent(ordered);
		} else {
			/* Screw you mmap */
			ret = filemap_write_and_wait_range(file->f_mapping,
							   lockstart,
							   lockend);
			if (ret)
				goto out;

			/*
			 * If we found a page that couldn't be invalidated just
			 * fall back to buffered.
			 */
			ret = invalidate_inode_pages2_range(file->f_mapping,
					lockstart >> PAGE_CACHE_SHIFT,
					lockend >> PAGE_CACHE_SHIFT);
			if (ret) {
				if (ret == -EBUSY)
					ret = 0;
				goto out;
			}
		}

6414 6415 6416
		cond_resched();
	}

6417 6418 6419 6420 6421 6422 6423
	/*
	 * we don't use btrfs_set_extent_delalloc because we don't want
	 * the dirty or uptodate bits
	 */
	if (writing) {
		write_bits = EXTENT_DELALLOC | EXTENT_DO_ACCOUNTING;
		ret = set_extent_bit(&BTRFS_I(inode)->io_tree, lockstart, lockend,
J
Jeff Mahoney 已提交
6424
				     EXTENT_DELALLOC, NULL, &cached_state,
6425 6426 6427 6428 6429 6430 6431 6432 6433 6434 6435 6436
				     GFP_NOFS);
		if (ret) {
			clear_extent_bit(&BTRFS_I(inode)->io_tree, lockstart,
					 lockend, EXTENT_LOCKED | write_bits,
					 1, 0, &cached_state, GFP_NOFS);
			goto out;
		}
	}

	free_extent_state(cached_state);
	cached_state = NULL;

C
Chris Mason 已提交
6437 6438 6439 6440
	ret = __blockdev_direct_IO(rw, iocb, inode,
		   BTRFS_I(inode)->root->fs_info->fs_devices->latest_bdev,
		   iov, offset, nr_segs, btrfs_get_blocks_direct, NULL,
		   btrfs_submit_direct, 0);
6441 6442

	if (ret < 0 && ret != -EIOCBQUEUED) {
6443 6444 6445 6446
		clear_extent_bit(&BTRFS_I(inode)->io_tree, offset,
			      offset + iov_length(iov, nr_segs) - 1,
			      EXTENT_LOCKED | write_bits, 1, 0,
			      &cached_state, GFP_NOFS);
6447 6448 6449 6450 6451
	} else if (ret >= 0 && ret < iov_length(iov, nr_segs)) {
		/*
		 * We're falling back to buffered, unlock the section we didn't
		 * do IO on.
		 */
6452 6453 6454 6455
		clear_extent_bit(&BTRFS_I(inode)->io_tree, offset + ret,
			      offset + iov_length(iov, nr_segs) - 1,
			      EXTENT_LOCKED | write_bits, 1, 0,
			      &cached_state, GFP_NOFS);
6456
	}
6457 6458
out:
	free_extent_state(cached_state);
6459
	return ret;
6460 6461
}

Y
Yehuda Sadeh 已提交
6462 6463 6464
static int btrfs_fiemap(struct inode *inode, struct fiemap_extent_info *fieinfo,
		__u64 start, __u64 len)
{
6465
	return extent_fiemap(inode, fieinfo, start, len, btrfs_get_extent_fiemap);
Y
Yehuda Sadeh 已提交
6466 6467
}

6468
int btrfs_readpage(struct file *file, struct page *page)
C
Chris Mason 已提交
6469
{
6470 6471
	struct extent_io_tree *tree;
	tree = &BTRFS_I(page->mapping->host)->io_tree;
6472
	return extent_read_full_page(tree, page, btrfs_get_extent, 0);
C
Chris Mason 已提交
6473
}
6474

6475
static int btrfs_writepage(struct page *page, struct writeback_control *wbc)
C
Chris Mason 已提交
6476
{
6477
	struct extent_io_tree *tree;
6478 6479 6480 6481 6482 6483 6484


	if (current->flags & PF_MEMALLOC) {
		redirty_page_for_writepage(wbc, page);
		unlock_page(page);
		return 0;
	}
6485
	tree = &BTRFS_I(page->mapping->host)->io_tree;
6486
	return extent_write_full_page(tree, page, btrfs_get_extent, wbc);
C
Chris Mason 已提交
6487 6488
}

6489 6490
int btrfs_writepages(struct address_space *mapping,
		     struct writeback_control *wbc)
C
Chris Mason 已提交
6491
{
6492
	struct extent_io_tree *tree;
6493

6494
	tree = &BTRFS_I(mapping->host)->io_tree;
C
Chris Mason 已提交
6495 6496 6497
	return extent_writepages(tree, mapping, btrfs_get_extent, wbc);
}

C
Chris Mason 已提交
6498 6499 6500 6501
static int
btrfs_readpages(struct file *file, struct address_space *mapping,
		struct list_head *pages, unsigned nr_pages)
{
6502 6503
	struct extent_io_tree *tree;
	tree = &BTRFS_I(mapping->host)->io_tree;
C
Chris Mason 已提交
6504 6505 6506
	return extent_readpages(tree, mapping, pages, nr_pages,
				btrfs_get_extent);
}
6507
static int __btrfs_releasepage(struct page *page, gfp_t gfp_flags)
C
Chris Mason 已提交
6508
{
6509 6510
	struct extent_io_tree *tree;
	struct extent_map_tree *map;
6511
	int ret;
6512

6513 6514
	tree = &BTRFS_I(page->mapping->host)->io_tree;
	map = &BTRFS_I(page->mapping->host)->extent_tree;
6515
	ret = try_release_extent_mapping(map, tree, page, gfp_flags);
6516 6517 6518 6519
	if (ret == 1) {
		ClearPagePrivate(page);
		set_page_private(page, 0);
		page_cache_release(page);
C
Chris Mason 已提交
6520
	}
6521
	return ret;
C
Chris Mason 已提交
6522 6523
}

6524 6525
static int btrfs_releasepage(struct page *page, gfp_t gfp_flags)
{
6526 6527
	if (PageWriteback(page) || PageDirty(page))
		return 0;
6528
	return __btrfs_releasepage(page, gfp_flags & GFP_NOFS);
6529 6530
}

6531
static void btrfs_invalidatepage(struct page *page, unsigned long offset)
C
Chris Mason 已提交
6532
{
6533
	struct inode *inode = page->mapping->host;
6534
	struct extent_io_tree *tree;
6535
	struct btrfs_ordered_extent *ordered;
6536
	struct extent_state *cached_state = NULL;
6537 6538
	u64 page_start = page_offset(page);
	u64 page_end = page_start + PAGE_CACHE_SIZE - 1;
C
Chris Mason 已提交
6539

6540 6541 6542 6543 6544 6545 6546
	/*
	 * we have the page locked, so new writeback can't start,
	 * and the dirty bit won't be cleared while we are here.
	 *
	 * Wait for IO on this page so that we can safely clear
	 * the PagePrivate2 bit and do ordered accounting
	 */
6547
	wait_on_page_writeback(page);
6548

6549
	tree = &BTRFS_I(inode)->io_tree;
6550 6551 6552 6553
	if (offset) {
		btrfs_releasepage(page, GFP_NOFS);
		return;
	}
6554
	lock_extent_bits(tree, page_start, page_end, 0, &cached_state);
6555
	ordered = btrfs_lookup_ordered_extent(inode,
6556 6557
					   page_offset(page));
	if (ordered) {
6558 6559 6560 6561
		/*
		 * IO on this page will never be started, so we need
		 * to account for any ordered extents now
		 */
6562 6563
		clear_extent_bit(tree, page_start, page_end,
				 EXTENT_DIRTY | EXTENT_DELALLOC |
6564
				 EXTENT_LOCKED | EXTENT_DO_ACCOUNTING, 1, 0,
6565
				 &cached_state, GFP_NOFS);
6566 6567 6568 6569
		/*
		 * whoever cleared the private bit is responsible
		 * for the finish_ordered_io
		 */
6570 6571 6572 6573
		if (TestClearPagePrivate2(page) &&
		    btrfs_dec_test_ordered_pending(inode, &ordered, page_start,
						   PAGE_CACHE_SIZE, 1)) {
			btrfs_finish_ordered_io(ordered);
6574
		}
6575
		btrfs_put_ordered_extent(ordered);
6576
		cached_state = NULL;
6577
		lock_extent_bits(tree, page_start, page_end, 0, &cached_state);
6578 6579
	}
	clear_extent_bit(tree, page_start, page_end,
6580
		 EXTENT_LOCKED | EXTENT_DIRTY | EXTENT_DELALLOC |
6581
		 EXTENT_DO_ACCOUNTING, 1, 1, &cached_state, GFP_NOFS);
6582 6583
	__btrfs_releasepage(page, GFP_NOFS);

C
Chris Mason 已提交
6584
	ClearPageChecked(page);
6585 6586 6587 6588 6589
	if (PagePrivate(page)) {
		ClearPagePrivate(page);
		set_page_private(page, 0);
		page_cache_release(page);
	}
C
Chris Mason 已提交
6590 6591
}

C
Chris Mason 已提交
6592 6593 6594 6595 6596 6597 6598 6599 6600 6601 6602 6603 6604 6605 6606
/*
 * btrfs_page_mkwrite() is not allowed to change the file size as it gets
 * called from a page fault handler when a page is first dirtied. Hence we must
 * be careful to check for EOF conditions here. We set the page up correctly
 * for a written page which means we get ENOSPC checking when writing into
 * holes and correct delalloc and unwritten extent mapping on filesystems that
 * support these features.
 *
 * We are not allowed to take the i_mutex here so we have to play games to
 * protect against truncate races as the page could now be beyond EOF.  Because
 * vmtruncate() writes the inode size before removing pages, once we have the
 * page lock we can determine safely if the page is beyond EOF. If it is not
 * beyond EOF, then the page is guaranteed safe against truncation until we
 * unlock the page.
 */
6607
int btrfs_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf)
C
Chris Mason 已提交
6608
{
6609
	struct page *page = vmf->page;
6610
	struct inode *inode = fdentry(vma->vm_file)->d_inode;
6611
	struct btrfs_root *root = BTRFS_I(inode)->root;
6612 6613
	struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
	struct btrfs_ordered_extent *ordered;
6614
	struct extent_state *cached_state = NULL;
6615 6616
	char *kaddr;
	unsigned long zero_start;
C
Chris Mason 已提交
6617
	loff_t size;
6618
	int ret;
6619
	int reserved = 0;
6620
	u64 page_start;
6621
	u64 page_end;
C
Chris Mason 已提交
6622

6623
	ret  = btrfs_delalloc_reserve_space(inode, PAGE_CACHE_SIZE);
6624
	if (!ret) {
6625
		ret = file_update_time(vma->vm_file);
6626 6627
		reserved = 1;
	}
6628 6629 6630 6631 6632
	if (ret) {
		if (ret == -ENOMEM)
			ret = VM_FAULT_OOM;
		else /* -ENOSPC, -EIO, etc */
			ret = VM_FAULT_SIGBUS;
6633 6634 6635
		if (reserved)
			goto out;
		goto out_noreserve;
6636
	}
6637

6638
	ret = VM_FAULT_NOPAGE; /* make the VM retry the fault */
6639
again:
C
Chris Mason 已提交
6640 6641
	lock_page(page);
	size = i_size_read(inode);
6642 6643
	page_start = page_offset(page);
	page_end = page_start + PAGE_CACHE_SIZE - 1;
6644

C
Chris Mason 已提交
6645
	if ((page->mapping != inode->i_mapping) ||
6646
	    (page_start >= size)) {
C
Chris Mason 已提交
6647 6648 6649
		/* page got truncated out from underneath us */
		goto out_unlock;
	}
6650 6651
	wait_on_page_writeback(page);

6652
	lock_extent_bits(io_tree, page_start, page_end, 0, &cached_state);
6653 6654
	set_page_extent_mapped(page);

6655 6656 6657 6658
	/*
	 * we can't set the delalloc bits if there are pending ordered
	 * extents.  Drop our locks and wait for them to finish
	 */
6659 6660
	ordered = btrfs_lookup_ordered_extent(inode, page_start);
	if (ordered) {
6661 6662
		unlock_extent_cached(io_tree, page_start, page_end,
				     &cached_state, GFP_NOFS);
6663
		unlock_page(page);
6664
		btrfs_start_ordered_extent(inode, ordered, 1);
6665 6666 6667 6668
		btrfs_put_ordered_extent(ordered);
		goto again;
	}

J
Josef Bacik 已提交
6669 6670 6671 6672 6673 6674 6675
	/*
	 * XXX - page_mkwrite gets called every time the page is dirtied, even
	 * if it was already dirty, so for space accounting reasons we need to
	 * clear any delalloc bits for the range we are fixing to save.  There
	 * is probably a better way to do this, but for now keep consistent with
	 * prepare_pages in the normal write path.
	 */
6676
	clear_extent_bit(&BTRFS_I(inode)->io_tree, page_start, page_end,
6677
			  EXTENT_DIRTY | EXTENT_DELALLOC | EXTENT_DO_ACCOUNTING,
6678
			  0, 0, &cached_state, GFP_NOFS);
J
Josef Bacik 已提交
6679

6680 6681
	ret = btrfs_set_extent_delalloc(inode, page_start, page_end,
					&cached_state);
J
Josef Bacik 已提交
6682
	if (ret) {
6683 6684
		unlock_extent_cached(io_tree, page_start, page_end,
				     &cached_state, GFP_NOFS);
J
Josef Bacik 已提交
6685 6686 6687
		ret = VM_FAULT_SIGBUS;
		goto out_unlock;
	}
6688
	ret = 0;
C
Chris Mason 已提交
6689 6690

	/* page is wholly or partially inside EOF */
6691
	if (page_start + PAGE_CACHE_SIZE > size)
6692
		zero_start = size & ~PAGE_CACHE_MASK;
C
Chris Mason 已提交
6693
	else
6694
		zero_start = PAGE_CACHE_SIZE;
C
Chris Mason 已提交
6695

6696 6697 6698 6699 6700 6701
	if (zero_start != PAGE_CACHE_SIZE) {
		kaddr = kmap(page);
		memset(kaddr + zero_start, 0, PAGE_CACHE_SIZE - zero_start);
		flush_dcache_page(page);
		kunmap(page);
	}
6702
	ClearPageChecked(page);
6703
	set_page_dirty(page);
6704
	SetPageUptodate(page);
6705

6706 6707 6708
	BTRFS_I(inode)->last_trans = root->fs_info->generation;
	BTRFS_I(inode)->last_sub_trans = BTRFS_I(inode)->root->log_transid;

6709
	unlock_extent_cached(io_tree, page_start, page_end, &cached_state, GFP_NOFS);
C
Chris Mason 已提交
6710 6711

out_unlock:
6712 6713
	if (!ret)
		return VM_FAULT_LOCKED;
C
Chris Mason 已提交
6714
	unlock_page(page);
6715
out:
6716
	btrfs_delalloc_release_space(inode, PAGE_CACHE_SIZE);
6717
out_noreserve:
C
Chris Mason 已提交
6718 6719 6720
	return ret;
}

6721
static int btrfs_truncate(struct inode *inode)
C
Chris Mason 已提交
6722 6723
{
	struct btrfs_root *root = BTRFS_I(inode)->root;
6724
	struct btrfs_block_rsv *rsv;
C
Chris Mason 已提交
6725
	int ret;
6726
	int err = 0;
C
Chris Mason 已提交
6727
	struct btrfs_trans_handle *trans;
6728
	unsigned long nr;
6729
	u64 mask = root->sectorsize - 1;
6730
	u64 min_size = btrfs_calc_trunc_metadata_size(root, 1);
C
Chris Mason 已提交
6731

6732 6733
	ret = btrfs_truncate_page(inode->i_mapping, inode->i_size);
	if (ret)
6734
		return ret;
6735

C
Chris Mason 已提交
6736
	btrfs_wait_ordered_range(inode, inode->i_size & (~mask), (u64)-1);
6737
	btrfs_ordered_update_i_size(inode, inode->i_size, NULL);
C
Chris Mason 已提交
6738

6739 6740 6741 6742 6743 6744 6745 6746 6747 6748 6749 6750 6751 6752 6753 6754 6755 6756 6757 6758 6759 6760 6761 6762 6763 6764 6765 6766 6767 6768 6769 6770 6771 6772 6773 6774 6775 6776 6777
	/*
	 * Yes ladies and gentelment, this is indeed ugly.  The fact is we have
	 * 3 things going on here
	 *
	 * 1) We need to reserve space for our orphan item and the space to
	 * delete our orphan item.  Lord knows we don't want to have a dangling
	 * orphan item because we didn't reserve space to remove it.
	 *
	 * 2) We need to reserve space to update our inode.
	 *
	 * 3) We need to have something to cache all the space that is going to
	 * be free'd up by the truncate operation, but also have some slack
	 * space reserved in case it uses space during the truncate (thank you
	 * very much snapshotting).
	 *
	 * And we need these to all be seperate.  The fact is we can use alot of
	 * space doing the truncate, and we have no earthly idea how much space
	 * we will use, so we need the truncate reservation to be seperate so it
	 * doesn't end up using space reserved for updating the inode or
	 * removing the orphan item.  We also need to be able to stop the
	 * transaction and start a new one, which means we need to be able to
	 * update the inode several times, and we have no idea of knowing how
	 * many times that will be, so we can't just reserve 1 item for the
	 * entirety of the opration, so that has to be done seperately as well.
	 * Then there is the orphan item, which does indeed need to be held on
	 * to for the whole operation, and we need nobody to touch this reserved
	 * space except the orphan code.
	 *
	 * So that leaves us with
	 *
	 * 1) root->orphan_block_rsv - for the orphan deletion.
	 * 2) rsv - for the truncate reservation, which we will steal from the
	 * transaction reservation.
	 * 3) fs_info->trans_block_rsv - this will have 1 items worth left for
	 * updating the inode.
	 */
	rsv = btrfs_alloc_block_rsv(root);
	if (!rsv)
		return -ENOMEM;
J
Josef Bacik 已提交
6778
	rsv->size = min_size;
6779

6780
	/*
6781
	 * 1 for the truncate slack space
6782 6783 6784 6785
	 * 1 for the orphan item we're going to add
	 * 1 for the orphan item deletion
	 * 1 for updating the inode.
	 */
6786 6787 6788 6789 6790
	trans = btrfs_start_transaction(root, 4);
	if (IS_ERR(trans)) {
		err = PTR_ERR(trans);
		goto out;
	}
6791

6792 6793 6794
	/* Migrate the slack space for the truncate to our reserve */
	ret = btrfs_block_rsv_migrate(&root->fs_info->trans_block_rsv, rsv,
				      min_size);
6795
	BUG_ON(ret);
6796 6797 6798 6799

	ret = btrfs_orphan_add(trans, inode);
	if (ret) {
		btrfs_end_transaction(trans, root);
6800
		goto out;
6801 6802
	}

6803 6804 6805 6806 6807 6808 6809 6810 6811 6812 6813 6814 6815 6816 6817 6818 6819
	/*
	 * setattr is responsible for setting the ordered_data_close flag,
	 * but that is only tested during the last file release.  That
	 * could happen well after the next commit, leaving a great big
	 * window where new writes may get lost if someone chooses to write
	 * to this file after truncating to zero
	 *
	 * The inode doesn't have any dirty data here, and so if we commit
	 * this is a noop.  If someone immediately starts writing to the inode
	 * it is very likely we'll catch some of their writes in this
	 * transaction, and the commit will find this file on the ordered
	 * data list with good things to send down.
	 *
	 * This is a best effort solution, there is still a window where
	 * using truncate to replace the contents of the file will
	 * end up with a zero length file after a crash.
	 */
6820 6821
	if (inode->i_size == 0 && test_bit(BTRFS_INODE_ORDERED_DATA_CLOSE,
					   &BTRFS_I(inode)->runtime_flags))
6822 6823
		btrfs_add_ordered_operation(trans, root, inode);

6824
	while (1) {
6825
		ret = btrfs_block_rsv_refill(root, rsv, min_size);
6826 6827 6828 6829 6830 6831 6832 6833 6834 6835 6836 6837
		if (ret) {
			/*
			 * This can only happen with the original transaction we
			 * started above, every other time we shouldn't have a
			 * transaction started yet.
			 */
			if (ret == -EAGAIN)
				goto end_trans;
			err = ret;
			break;
		}

6838
		if (!trans) {
6839 6840
			/* Just need the 1 for updating the inode */
			trans = btrfs_start_transaction(root, 1);
6841
			if (IS_ERR(trans)) {
J
Josef Bacik 已提交
6842 6843 6844
				ret = err = PTR_ERR(trans);
				trans = NULL;
				break;
6845
			}
6846 6847
		}

6848 6849
		trans->block_rsv = rsv;

6850 6851 6852
		ret = btrfs_truncate_inode_items(trans, root, inode,
						 inode->i_size,
						 BTRFS_EXTENT_DATA_KEY);
6853 6854
		if (ret != -EAGAIN) {
			err = ret;
6855
			break;
6856
		}
C
Chris Mason 已提交
6857

6858
		trans->block_rsv = &root->fs_info->trans_block_rsv;
6859
		ret = btrfs_update_inode(trans, root, inode);
6860 6861 6862 6863
		if (ret) {
			err = ret;
			break;
		}
6864
end_trans:
6865 6866
		nr = trans->blocks_used;
		btrfs_end_transaction(trans, root);
6867
		trans = NULL;
6868 6869 6870 6871
		btrfs_btree_balance_dirty(root, nr);
	}

	if (ret == 0 && inode->i_nlink > 0) {
6872
		trans->block_rsv = root->orphan_block_rsv;
6873
		ret = btrfs_orphan_del(trans, inode);
6874 6875
		if (ret)
			err = ret;
6876 6877 6878 6879 6880 6881
	} else if (ret && inode->i_nlink > 0) {
		/*
		 * Failed to do the truncate, remove us from the in memory
		 * orphan list.
		 */
		ret = btrfs_orphan_del(NULL, inode);
6882 6883
	}

6884 6885 6886 6887 6888
	if (trans) {
		trans->block_rsv = &root->fs_info->trans_block_rsv;
		ret = btrfs_update_inode(trans, root, inode);
		if (ret && !err)
			err = ret;
6889

6890
		nr = trans->blocks_used;
6891
		ret = btrfs_end_transaction(trans, root);
6892 6893
		btrfs_btree_balance_dirty(root, nr);
	}
6894 6895 6896 6897

out:
	btrfs_free_block_rsv(root, rsv);

6898 6899
	if (ret && !err)
		err = ret;
6900

6901
	return err;
C
Chris Mason 已提交
6902 6903
}

C
Chris Mason 已提交
6904 6905 6906
/*
 * create a new subvolume directory/inode (helper for the ioctl).
 */
6907
int btrfs_create_subvol_root(struct btrfs_trans_handle *trans,
6908
			     struct btrfs_root *new_root, u64 new_dirid)
C
Chris Mason 已提交
6909 6910
{
	struct inode *inode;
6911
	int err;
6912
	u64 index = 0;
C
Chris Mason 已提交
6913

6914 6915 6916 6917
	inode = btrfs_new_inode(trans, new_root, NULL, "..", 2,
				new_dirid, new_dirid,
				S_IFDIR | (~current_umask() & S_IRWXUGO),
				&index);
6918
	if (IS_ERR(inode))
C
Christoph Hellwig 已提交
6919
		return PTR_ERR(inode);
C
Chris Mason 已提交
6920 6921 6922
	inode->i_op = &btrfs_dir_inode_operations;
	inode->i_fop = &btrfs_dir_file_operations;

M
Miklos Szeredi 已提交
6923
	set_nlink(inode, 1);
6924
	btrfs_i_size_write(inode, 0);
6925

6926
	err = btrfs_update_inode(trans, new_root, inode);
6927

6928
	iput(inode);
6929
	return err;
C
Chris Mason 已提交
6930 6931 6932 6933 6934
}

struct inode *btrfs_alloc_inode(struct super_block *sb)
{
	struct btrfs_inode *ei;
Y
Yan, Zheng 已提交
6935
	struct inode *inode;
C
Chris Mason 已提交
6936 6937 6938 6939

	ei = kmem_cache_alloc(btrfs_inode_cachep, GFP_NOFS);
	if (!ei)
		return NULL;
Y
Yan, Zheng 已提交
6940 6941 6942 6943

	ei->root = NULL;
	ei->space_info = NULL;
	ei->generation = 0;
6944
	ei->last_trans = 0;
6945
	ei->last_sub_trans = 0;
6946
	ei->logged_trans = 0;
Y
Yan, Zheng 已提交
6947 6948 6949
	ei->delalloc_bytes = 0;
	ei->disk_i_size = 0;
	ei->flags = 0;
6950
	ei->csum_bytes = 0;
Y
Yan, Zheng 已提交
6951 6952 6953
	ei->index_cnt = (u64)-1;
	ei->last_unlink_trans = 0;

6954 6955 6956
	spin_lock_init(&ei->lock);
	ei->outstanding_extents = 0;
	ei->reserved_extents = 0;
Y
Yan, Zheng 已提交
6957

6958
	ei->runtime_flags = 0;
6959
	ei->force_compress = BTRFS_COMPRESS_NONE;
Y
Yan, Zheng 已提交
6960

6961 6962
	ei->delayed_node = NULL;

Y
Yan, Zheng 已提交
6963
	inode = &ei->vfs_inode;
6964
	extent_map_tree_init(&ei->extent_tree);
6965 6966
	extent_io_tree_init(&ei->io_tree, &inode->i_data);
	extent_io_tree_init(&ei->io_failure_tree, &inode->i_data);
6967 6968
	ei->io_tree.track_uptodate = 1;
	ei->io_failure_tree.track_uptodate = 1;
Y
Yan, Zheng 已提交
6969
	mutex_init(&ei->log_mutex);
6970
	mutex_init(&ei->delalloc_mutex);
6971
	btrfs_ordered_inode_tree_init(&ei->ordered_tree);
Y
Yan, Zheng 已提交
6972
	INIT_LIST_HEAD(&ei->delalloc_inodes);
6973
	INIT_LIST_HEAD(&ei->ordered_operations);
Y
Yan, Zheng 已提交
6974 6975 6976
	RB_CLEAR_NODE(&ei->rb_node);

	return inode;
C
Chris Mason 已提交
6977 6978
}

N
Nick Piggin 已提交
6979 6980 6981 6982 6983 6984
static void btrfs_i_callback(struct rcu_head *head)
{
	struct inode *inode = container_of(head, struct inode, i_rcu);
	kmem_cache_free(btrfs_inode_cachep, BTRFS_I(inode));
}

C
Chris Mason 已提交
6985 6986
void btrfs_destroy_inode(struct inode *inode)
{
6987
	struct btrfs_ordered_extent *ordered;
6988 6989
	struct btrfs_root *root = BTRFS_I(inode)->root;

A
Al Viro 已提交
6990
	WARN_ON(!hlist_empty(&inode->i_dentry));
C
Chris Mason 已提交
6991
	WARN_ON(inode->i_data.nrpages);
6992 6993
	WARN_ON(BTRFS_I(inode)->outstanding_extents);
	WARN_ON(BTRFS_I(inode)->reserved_extents);
6994 6995
	WARN_ON(BTRFS_I(inode)->delalloc_bytes);
	WARN_ON(BTRFS_I(inode)->csum_bytes);
C
Chris Mason 已提交
6996

6997 6998 6999 7000 7001 7002 7003 7004
	/*
	 * This can happen where we create an inode, but somebody else also
	 * created the same inode and we need to destroy the one we already
	 * created.
	 */
	if (!root)
		goto free;

7005 7006 7007 7008 7009 7010 7011 7012 7013 7014 7015
	/*
	 * Make sure we're properly removed from the ordered operation
	 * lists.
	 */
	smp_mb();
	if (!list_empty(&BTRFS_I(inode)->ordered_operations)) {
		spin_lock(&root->fs_info->ordered_extent_lock);
		list_del_init(&BTRFS_I(inode)->ordered_operations);
		spin_unlock(&root->fs_info->ordered_extent_lock);
	}

7016 7017
	if (test_bit(BTRFS_INODE_HAS_ORPHAN_ITEM,
		     &BTRFS_I(inode)->runtime_flags)) {
L
Li Zefan 已提交
7018 7019
		printk(KERN_INFO "BTRFS: inode %llu still on the orphan list\n",
		       (unsigned long long)btrfs_ino(inode));
7020
		atomic_dec(&root->orphan_inodes);
7021 7022
	}

C
Chris Mason 已提交
7023
	while (1) {
7024 7025 7026 7027
		ordered = btrfs_lookup_first_ordered_extent(inode, (u64)-1);
		if (!ordered)
			break;
		else {
C
Chris Mason 已提交
7028 7029 7030 7031
			printk(KERN_ERR "btrfs found ordered "
			       "extent %llu %llu on inode cleanup\n",
			       (unsigned long long)ordered->file_offset,
			       (unsigned long long)ordered->len);
7032 7033 7034 7035 7036
			btrfs_remove_ordered_extent(inode, ordered);
			btrfs_put_ordered_extent(ordered);
			btrfs_put_ordered_extent(ordered);
		}
	}
7037
	inode_tree_del(inode);
7038
	btrfs_drop_extent_cache(inode, 0, (u64)-1, 0);
7039
free:
7040
	btrfs_remove_delayed_node(inode);
N
Nick Piggin 已提交
7041
	call_rcu(&inode->i_rcu, btrfs_i_callback);
C
Chris Mason 已提交
7042 7043
}

7044
int btrfs_drop_inode(struct inode *inode)
7045 7046
{
	struct btrfs_root *root = BTRFS_I(inode)->root;
7047

7048
	if (btrfs_root_refs(&root->root_item) == 0 &&
7049
	    !btrfs_is_free_space_inode(root, inode))
7050
		return 1;
7051
	else
7052
		return generic_drop_inode(inode);
7053 7054
}

7055
static void init_once(void *foo)
C
Chris Mason 已提交
7056 7057 7058 7059 7060 7061 7062 7063 7064 7065 7066 7067 7068 7069 7070 7071
{
	struct btrfs_inode *ei = (struct btrfs_inode *) foo;

	inode_init_once(&ei->vfs_inode);
}

void btrfs_destroy_cachep(void)
{
	if (btrfs_inode_cachep)
		kmem_cache_destroy(btrfs_inode_cachep);
	if (btrfs_trans_handle_cachep)
		kmem_cache_destroy(btrfs_trans_handle_cachep);
	if (btrfs_transaction_cachep)
		kmem_cache_destroy(btrfs_transaction_cachep);
	if (btrfs_path_cachep)
		kmem_cache_destroy(btrfs_path_cachep);
7072 7073
	if (btrfs_free_space_cachep)
		kmem_cache_destroy(btrfs_free_space_cachep);
C
Chris Mason 已提交
7074 7075 7076 7077
}

int btrfs_init_cachep(void)
{
7078 7079 7080
	btrfs_inode_cachep = kmem_cache_create("btrfs_inode_cache",
			sizeof(struct btrfs_inode), 0,
			SLAB_RECLAIM_ACCOUNT | SLAB_MEM_SPREAD, init_once);
C
Chris Mason 已提交
7081 7082
	if (!btrfs_inode_cachep)
		goto fail;
7083 7084 7085 7086

	btrfs_trans_handle_cachep = kmem_cache_create("btrfs_trans_handle_cache",
			sizeof(struct btrfs_trans_handle), 0,
			SLAB_RECLAIM_ACCOUNT | SLAB_MEM_SPREAD, NULL);
C
Chris Mason 已提交
7087 7088
	if (!btrfs_trans_handle_cachep)
		goto fail;
7089 7090 7091 7092

	btrfs_transaction_cachep = kmem_cache_create("btrfs_transaction_cache",
			sizeof(struct btrfs_transaction), 0,
			SLAB_RECLAIM_ACCOUNT | SLAB_MEM_SPREAD, NULL);
C
Chris Mason 已提交
7093 7094
	if (!btrfs_transaction_cachep)
		goto fail;
7095 7096 7097 7098

	btrfs_path_cachep = kmem_cache_create("btrfs_path_cache",
			sizeof(struct btrfs_path), 0,
			SLAB_RECLAIM_ACCOUNT | SLAB_MEM_SPREAD, NULL);
C
Chris Mason 已提交
7099 7100
	if (!btrfs_path_cachep)
		goto fail;
7101

7102 7103 7104 7105 7106 7107
	btrfs_free_space_cachep = kmem_cache_create("btrfs_free_space_cache",
			sizeof(struct btrfs_free_space), 0,
			SLAB_RECLAIM_ACCOUNT | SLAB_MEM_SPREAD, NULL);
	if (!btrfs_free_space_cachep)
		goto fail;

C
Chris Mason 已提交
7108 7109 7110 7111 7112 7113 7114 7115 7116 7117
	return 0;
fail:
	btrfs_destroy_cachep();
	return -ENOMEM;
}

static int btrfs_getattr(struct vfsmount *mnt,
			 struct dentry *dentry, struct kstat *stat)
{
	struct inode *inode = dentry->d_inode;
D
David Sterba 已提交
7118 7119
	u32 blocksize = inode->i_sb->s_blocksize;

C
Chris Mason 已提交
7120
	generic_fillattr(inode, stat);
7121
	stat->dev = BTRFS_I(inode)->root->anon_dev;
C
Chris Mason 已提交
7122
	stat->blksize = PAGE_CACHE_SIZE;
D
David Sterba 已提交
7123 7124
	stat->blocks = (ALIGN(inode_get_bytes(inode), blocksize) +
		ALIGN(BTRFS_I(inode)->delalloc_bytes, blocksize)) >> 9;
C
Chris Mason 已提交
7125 7126 7127
	return 0;
}

7128 7129 7130 7131 7132 7133 7134 7135 7136 7137 7138 7139 7140 7141
/*
 * If a file is moved, it will inherit the cow and compression flags of the new
 * directory.
 */
static void fixup_inode_flags(struct inode *dir, struct inode *inode)
{
	struct btrfs_inode *b_dir = BTRFS_I(dir);
	struct btrfs_inode *b_inode = BTRFS_I(inode);

	if (b_dir->flags & BTRFS_INODE_NODATACOW)
		b_inode->flags |= BTRFS_INODE_NODATACOW;
	else
		b_inode->flags &= ~BTRFS_INODE_NODATACOW;

7142
	if (b_dir->flags & BTRFS_INODE_COMPRESS) {
7143
		b_inode->flags |= BTRFS_INODE_COMPRESS;
7144 7145 7146 7147 7148
		b_inode->flags &= ~BTRFS_INODE_NOCOMPRESS;
	} else {
		b_inode->flags &= ~(BTRFS_INODE_COMPRESS |
				    BTRFS_INODE_NOCOMPRESS);
	}
7149 7150
}

C
Chris Mason 已提交
7151 7152
static int btrfs_rename(struct inode *old_dir, struct dentry *old_dentry,
			   struct inode *new_dir, struct dentry *new_dentry)
C
Chris Mason 已提交
7153 7154 7155
{
	struct btrfs_trans_handle *trans;
	struct btrfs_root *root = BTRFS_I(old_dir)->root;
7156
	struct btrfs_root *dest = BTRFS_I(new_dir)->root;
C
Chris Mason 已提交
7157 7158 7159
	struct inode *new_inode = new_dentry->d_inode;
	struct inode *old_inode = old_dentry->d_inode;
	struct timespec ctime = CURRENT_TIME;
7160
	u64 index = 0;
7161
	u64 root_objectid;
C
Chris Mason 已提交
7162
	int ret;
L
Li Zefan 已提交
7163
	u64 old_ino = btrfs_ino(old_inode);
C
Chris Mason 已提交
7164

L
Li Zefan 已提交
7165
	if (btrfs_ino(new_dir) == BTRFS_EMPTY_SUBVOL_DIR_OBJECTID)
7166 7167
		return -EPERM;

7168
	/* we only allow rename subvolume link between subvolumes */
L
Li Zefan 已提交
7169
	if (old_ino != BTRFS_FIRST_FREE_OBJECTID && root != dest)
7170 7171
		return -EXDEV;

L
Li Zefan 已提交
7172 7173
	if (old_ino == BTRFS_EMPTY_SUBVOL_DIR_OBJECTID ||
	    (new_inode && btrfs_ino(new_inode) == BTRFS_FIRST_FREE_OBJECTID))
C
Chris Mason 已提交
7174
		return -ENOTEMPTY;
7175

7176 7177 7178
	if (S_ISDIR(old_inode->i_mode) && new_inode &&
	    new_inode->i_size > BTRFS_EMPTY_DIR_SIZE)
		return -ENOTEMPTY;
7179 7180 7181 7182 7183
	/*
	 * we're using rename to replace one file with another.
	 * and the replacement file is large.  Start IO on it now so
	 * we don't add too much work to the end of the transaction
	 */
7184
	if (new_inode && S_ISREG(old_inode->i_mode) && new_inode->i_size &&
7185 7186 7187
	    old_inode->i_size > BTRFS_ORDERED_OPERATIONS_FLUSH_LIMIT)
		filemap_flush(old_inode->i_mapping);

7188
	/* close the racy window with snapshot create/destroy ioctl */
L
Li Zefan 已提交
7189
	if (old_ino == BTRFS_FIRST_FREE_OBJECTID)
7190
		down_read(&root->fs_info->subvol_sem);
7191 7192 7193 7194 7195 7196 7197 7198 7199
	/*
	 * We want to reserve the absolute worst case amount of items.  So if
	 * both inodes are subvols and we need to unlink them then that would
	 * require 4 item modifications, but if they are both normal inodes it
	 * would require 5 item modifications, so we'll assume their normal
	 * inodes.  So 5 * 2 is 10, plus 1 for the new link, so 11 total items
	 * should cover the worst case number of items we'll modify.
	 */
	trans = btrfs_start_transaction(root, 20);
7200 7201 7202 7203
	if (IS_ERR(trans)) {
                ret = PTR_ERR(trans);
                goto out_notrans;
        }
7204

7205 7206
	if (dest != root)
		btrfs_record_root_in_trans(trans, dest);
7207

7208 7209 7210
	ret = btrfs_set_inode_index(new_dir, &index);
	if (ret)
		goto out_fail;
7211

L
Li Zefan 已提交
7212
	if (unlikely(old_ino == BTRFS_FIRST_FREE_OBJECTID)) {
7213 7214 7215
		/* force full log commit if subvolume involved. */
		root->fs_info->last_trans_log_full_commit = trans->transid;
	} else {
7216 7217 7218
		ret = btrfs_insert_inode_ref(trans, dest,
					     new_dentry->d_name.name,
					     new_dentry->d_name.len,
L
Li Zefan 已提交
7219 7220
					     old_ino,
					     btrfs_ino(new_dir), index);
7221 7222
		if (ret)
			goto out_fail;
7223 7224 7225 7226 7227 7228 7229 7230 7231
		/*
		 * this is an ugly little race, but the rename is required
		 * to make sure that if we crash, the inode is either at the
		 * old name or the new one.  pinning the log transaction lets
		 * us make sure we don't allow a log commit to come in after
		 * we unlink the name but before we add the new name back in.
		 */
		btrfs_pin_log_trans(root);
	}
7232 7233 7234 7235
	/*
	 * make sure the inode gets flushed if it is replacing
	 * something.
	 */
L
Li Zefan 已提交
7236
	if (new_inode && new_inode->i_size && S_ISREG(old_inode->i_mode))
7237 7238
		btrfs_add_ordered_operation(trans, root, old_inode);

7239 7240 7241
	inode_inc_iversion(old_dir);
	inode_inc_iversion(new_dir);
	inode_inc_iversion(old_inode);
C
Chris Mason 已提交
7242 7243 7244
	old_dir->i_ctime = old_dir->i_mtime = ctime;
	new_dir->i_ctime = new_dir->i_mtime = ctime;
	old_inode->i_ctime = ctime;
7245

7246 7247 7248
	if (old_dentry->d_parent != new_dentry->d_parent)
		btrfs_record_unlink_dir(trans, old_dir, old_inode, 1);

L
Li Zefan 已提交
7249
	if (unlikely(old_ino == BTRFS_FIRST_FREE_OBJECTID)) {
7250 7251 7252 7253 7254
		root_objectid = BTRFS_I(old_inode)->root->root_key.objectid;
		ret = btrfs_unlink_subvol(trans, root, old_dir, root_objectid,
					old_dentry->d_name.name,
					old_dentry->d_name.len);
	} else {
7255 7256 7257 7258 7259 7260
		ret = __btrfs_unlink_inode(trans, root, old_dir,
					old_dentry->d_inode,
					old_dentry->d_name.name,
					old_dentry->d_name.len);
		if (!ret)
			ret = btrfs_update_inode(trans, root, old_inode);
7261
	}
7262 7263 7264 7265
	if (ret) {
		btrfs_abort_transaction(trans, root, ret);
		goto out_fail;
	}
C
Chris Mason 已提交
7266 7267

	if (new_inode) {
7268
		inode_inc_iversion(new_inode);
C
Chris Mason 已提交
7269
		new_inode->i_ctime = CURRENT_TIME;
L
Li Zefan 已提交
7270
		if (unlikely(btrfs_ino(new_inode) ==
7271 7272 7273 7274 7275 7276 7277 7278 7279 7280 7281 7282 7283
			     BTRFS_EMPTY_SUBVOL_DIR_OBJECTID)) {
			root_objectid = BTRFS_I(new_inode)->location.objectid;
			ret = btrfs_unlink_subvol(trans, dest, new_dir,
						root_objectid,
						new_dentry->d_name.name,
						new_dentry->d_name.len);
			BUG_ON(new_inode->i_nlink == 0);
		} else {
			ret = btrfs_unlink_inode(trans, dest, new_dir,
						 new_dentry->d_inode,
						 new_dentry->d_name.name,
						 new_dentry->d_name.len);
		}
7284
		if (!ret && new_inode->i_nlink == 0) {
7285
			ret = btrfs_orphan_add(trans, new_dentry->d_inode);
7286
			BUG_ON(ret);
7287
		}
7288 7289 7290 7291
		if (ret) {
			btrfs_abort_transaction(trans, root, ret);
			goto out_fail;
		}
C
Chris Mason 已提交
7292
	}
7293

7294 7295
	fixup_inode_flags(new_dir, old_inode);

7296 7297
	ret = btrfs_add_link(trans, new_dir, old_inode,
			     new_dentry->d_name.name,
7298
			     new_dentry->d_name.len, 0, index);
7299 7300 7301 7302
	if (ret) {
		btrfs_abort_transaction(trans, root, ret);
		goto out_fail;
	}
C
Chris Mason 已提交
7303

L
Li Zefan 已提交
7304
	if (old_ino != BTRFS_FIRST_FREE_OBJECTID) {
7305
		struct dentry *parent = new_dentry->d_parent;
7306
		btrfs_log_new_name(trans, old_inode, old_dir, parent);
7307 7308
		btrfs_end_log_trans(root);
	}
C
Chris Mason 已提交
7309
out_fail:
7310
	btrfs_end_transaction(trans, root);
7311
out_notrans:
L
Li Zefan 已提交
7312
	if (old_ino == BTRFS_FIRST_FREE_OBJECTID)
7313
		up_read(&root->fs_info->subvol_sem);
J
Josef Bacik 已提交
7314

C
Chris Mason 已提交
7315 7316 7317
	return ret;
}

C
Chris Mason 已提交
7318 7319 7320 7321
/*
 * some fairly slow code that needs optimization. This walks the list
 * of all the inodes with pending delalloc and forces them to disk.
 */
Y
Yan, Zheng 已提交
7322
int btrfs_start_delalloc_inodes(struct btrfs_root *root, int delay_iput)
7323 7324 7325
{
	struct list_head *head = &root->fs_info->delalloc_inodes;
	struct btrfs_inode *binode;
7326
	struct inode *inode;
7327

Y
Yan Zheng 已提交
7328 7329 7330
	if (root->fs_info->sb->s_flags & MS_RDONLY)
		return -EROFS;

7331
	spin_lock(&root->fs_info->delalloc_lock);
C
Chris Mason 已提交
7332
	while (!list_empty(head)) {
7333 7334
		binode = list_entry(head->next, struct btrfs_inode,
				    delalloc_inodes);
7335 7336 7337
		inode = igrab(&binode->vfs_inode);
		if (!inode)
			list_del_init(&binode->delalloc_inodes);
7338
		spin_unlock(&root->fs_info->delalloc_lock);
7339
		if (inode) {
7340
			filemap_flush(inode->i_mapping);
Y
Yan, Zheng 已提交
7341 7342 7343 7344
			if (delay_iput)
				btrfs_add_delayed_iput(inode);
			else
				iput(inode);
7345 7346
		}
		cond_resched();
7347
		spin_lock(&root->fs_info->delalloc_lock);
7348
	}
7349
	spin_unlock(&root->fs_info->delalloc_lock);
7350 7351 7352 7353 7354 7355

	/* the filemap_flush will queue IO into the worker threads, but
	 * we have to make sure the IO is actually started and that
	 * ordered extents get created before we return
	 */
	atomic_inc(&root->fs_info->async_submit_draining);
C
Chris Mason 已提交
7356
	while (atomic_read(&root->fs_info->nr_async_submits) ||
7357
	      atomic_read(&root->fs_info->async_delalloc_pages)) {
7358
		wait_event(root->fs_info->async_submit_wait,
7359 7360
		   (atomic_read(&root->fs_info->nr_async_submits) == 0 &&
		    atomic_read(&root->fs_info->async_delalloc_pages) == 0));
7361 7362
	}
	atomic_dec(&root->fs_info->async_submit_draining);
7363 7364 7365
	return 0;
}

C
Chris Mason 已提交
7366 7367 7368 7369 7370 7371 7372
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;
7373
	struct inode *inode = NULL;
C
Chris Mason 已提交
7374 7375 7376
	int err;
	int drop_inode = 0;
	u64 objectid;
7377
	u64 index = 0 ;
C
Chris Mason 已提交
7378 7379
	int name_len;
	int datasize;
7380
	unsigned long ptr;
C
Chris Mason 已提交
7381
	struct btrfs_file_extent_item *ei;
7382
	struct extent_buffer *leaf;
7383
	unsigned long nr = 0;
C
Chris Mason 已提交
7384 7385 7386 7387

	name_len = strlen(symname) + 1;
	if (name_len > BTRFS_MAX_INLINE_DATA_SIZE(root))
		return -ENAMETOOLONG;
7388

J
Josef Bacik 已提交
7389 7390 7391 7392 7393
	/*
	 * 2 items for inode item and ref
	 * 2 items for dir items
	 * 1 item for xattr if selinux is on
	 */
7394 7395 7396
	trans = btrfs_start_transaction(root, 5);
	if (IS_ERR(trans))
		return PTR_ERR(trans);
7397

7398 7399 7400 7401
	err = btrfs_find_free_ino(root, &objectid);
	if (err)
		goto out_unlock;

7402
	inode = btrfs_new_inode(trans, root, dir, dentry->d_name.name,
L
Li Zefan 已提交
7403
				dentry->d_name.len, btrfs_ino(dir), objectid,
7404
				S_IFLNK|S_IRWXUGO, &index);
7405 7406
	if (IS_ERR(inode)) {
		err = PTR_ERR(inode);
C
Chris Mason 已提交
7407
		goto out_unlock;
7408
	}
C
Chris Mason 已提交
7409

7410
	err = btrfs_init_inode_security(trans, inode, dir, &dentry->d_name);
J
Josef Bacik 已提交
7411 7412 7413 7414 7415
	if (err) {
		drop_inode = 1;
		goto out_unlock;
	}

7416 7417 7418 7419 7420 7421 7422 7423 7424
	/*
	* If the active LSM wants to access the inode during
	* d_instantiate it needs these. Smack checks to see
	* if the filesystem supports xattrs by looking at the
	* ops vector.
	*/
	inode->i_fop = &btrfs_file_operations;
	inode->i_op = &btrfs_file_inode_operations;

7425
	err = btrfs_add_nondir(trans, dir, dentry, inode, 0, index);
C
Chris Mason 已提交
7426 7427 7428 7429
	if (err)
		drop_inode = 1;
	else {
		inode->i_mapping->a_ops = &btrfs_aops;
C
Chris Mason 已提交
7430
		inode->i_mapping->backing_dev_info = &root->fs_info->bdi;
7431
		BTRFS_I(inode)->io_tree.ops = &btrfs_extent_io_ops;
C
Chris Mason 已提交
7432 7433 7434 7435 7436
	}
	if (drop_inode)
		goto out_unlock;

	path = btrfs_alloc_path();
7437 7438 7439 7440 7441
	if (!path) {
		err = -ENOMEM;
		drop_inode = 1;
		goto out_unlock;
	}
L
Li Zefan 已提交
7442
	key.objectid = btrfs_ino(inode);
C
Chris Mason 已提交
7443 7444 7445 7446 7447
	key.offset = 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);
7448 7449
	if (err) {
		drop_inode = 1;
7450
		btrfs_free_path(path);
7451 7452
		goto out_unlock;
	}
7453 7454 7455 7456 7457
	leaf = path->nodes[0];
	ei = btrfs_item_ptr(leaf, path->slots[0],
			    struct btrfs_file_extent_item);
	btrfs_set_file_extent_generation(leaf, ei, trans->transid);
	btrfs_set_file_extent_type(leaf, ei,
C
Chris Mason 已提交
7458
				   BTRFS_FILE_EXTENT_INLINE);
C
Chris Mason 已提交
7459 7460 7461 7462 7463
	btrfs_set_file_extent_encryption(leaf, ei, 0);
	btrfs_set_file_extent_compression(leaf, ei, 0);
	btrfs_set_file_extent_other_encoding(leaf, ei, 0);
	btrfs_set_file_extent_ram_bytes(leaf, ei, name_len);

C
Chris Mason 已提交
7464
	ptr = btrfs_file_extent_inline_start(ei);
7465 7466
	write_extent_buffer(leaf, symname, ptr, name_len);
	btrfs_mark_buffer_dirty(leaf);
C
Chris Mason 已提交
7467
	btrfs_free_path(path);
7468

C
Chris Mason 已提交
7469 7470
	inode->i_op = &btrfs_symlink_inode_operations;
	inode->i_mapping->a_ops = &btrfs_symlink_aops;
C
Chris Mason 已提交
7471
	inode->i_mapping->backing_dev_info = &root->fs_info->bdi;
Y
Yan Zheng 已提交
7472
	inode_set_bytes(inode, name_len);
7473
	btrfs_i_size_write(inode, name_len - 1);
7474 7475 7476
	err = btrfs_update_inode(trans, root, inode);
	if (err)
		drop_inode = 1;
C
Chris Mason 已提交
7477 7478

out_unlock:
7479 7480
	if (!err)
		d_instantiate(dentry, inode);
7481
	nr = trans->blocks_used;
7482
	btrfs_end_transaction(trans, root);
C
Chris Mason 已提交
7483 7484 7485 7486
	if (drop_inode) {
		inode_dec_link_count(inode);
		iput(inode);
	}
7487
	btrfs_btree_balance_dirty(root, nr);
C
Chris Mason 已提交
7488 7489
	return err;
}
7490

7491 7492 7493 7494
static int __btrfs_prealloc_file_range(struct inode *inode, int mode,
				       u64 start, u64 num_bytes, u64 min_size,
				       loff_t actual_len, u64 *alloc_hint,
				       struct btrfs_trans_handle *trans)
Y
Yan Zheng 已提交
7495 7496 7497 7498
{
	struct btrfs_root *root = BTRFS_I(inode)->root;
	struct btrfs_key ins;
	u64 cur_offset = start;
7499
	u64 i_size;
Y
Yan Zheng 已提交
7500
	int ret = 0;
7501
	bool own_trans = true;
Y
Yan Zheng 已提交
7502

7503 7504
	if (trans)
		own_trans = false;
Y
Yan Zheng 已提交
7505
	while (num_bytes > 0) {
7506 7507 7508 7509 7510 7511
		if (own_trans) {
			trans = btrfs_start_transaction(root, 3);
			if (IS_ERR(trans)) {
				ret = PTR_ERR(trans);
				break;
			}
7512 7513
		}

7514
		ret = btrfs_reserve_extent(trans, root, num_bytes, min_size,
7515
					   0, *alloc_hint, &ins, 1);
7516
		if (ret) {
7517 7518
			if (own_trans)
				btrfs_end_transaction(trans, root);
7519
			break;
Y
Yan Zheng 已提交
7520
		}
7521

Y
Yan Zheng 已提交
7522 7523 7524
		ret = insert_reserved_file_extent(trans, inode,
						  cur_offset, ins.objectid,
						  ins.offset, ins.offset,
Y
Yan, Zheng 已提交
7525
						  ins.offset, 0, 0, 0,
Y
Yan Zheng 已提交
7526
						  BTRFS_FILE_EXTENT_PREALLOC);
7527 7528 7529 7530 7531 7532
		if (ret) {
			btrfs_abort_transaction(trans, root, ret);
			if (own_trans)
				btrfs_end_transaction(trans, root);
			break;
		}
C
Chris Mason 已提交
7533 7534
		btrfs_drop_extent_cache(inode, cur_offset,
					cur_offset + ins.offset -1, 0);
7535

Y
Yan Zheng 已提交
7536 7537
		num_bytes -= ins.offset;
		cur_offset += ins.offset;
7538
		*alloc_hint = ins.objectid + ins.offset;
7539

7540
		inode_inc_iversion(inode);
Y
Yan Zheng 已提交
7541
		inode->i_ctime = CURRENT_TIME;
7542
		BTRFS_I(inode)->flags |= BTRFS_INODE_PREALLOC;
Y
Yan Zheng 已提交
7543
		if (!(mode & FALLOC_FL_KEEP_SIZE) &&
7544 7545
		    (actual_len > inode->i_size) &&
		    (cur_offset > inode->i_size)) {
7546
			if (cur_offset > actual_len)
7547
				i_size = actual_len;
7548
			else
7549 7550 7551
				i_size = cur_offset;
			i_size_write(inode, i_size);
			btrfs_ordered_update_i_size(inode, i_size, NULL);
7552 7553
		}

Y
Yan Zheng 已提交
7554
		ret = btrfs_update_inode(trans, root, inode);
7555 7556 7557 7558 7559 7560 7561

		if (ret) {
			btrfs_abort_transaction(trans, root, ret);
			if (own_trans)
				btrfs_end_transaction(trans, root);
			break;
		}
Y
Yan Zheng 已提交
7562

7563 7564
		if (own_trans)
			btrfs_end_transaction(trans, root);
7565
	}
Y
Yan Zheng 已提交
7566 7567 7568
	return ret;
}

7569 7570 7571 7572 7573 7574 7575 7576 7577 7578 7579 7580 7581 7582 7583 7584 7585 7586
int btrfs_prealloc_file_range(struct inode *inode, int mode,
			      u64 start, u64 num_bytes, u64 min_size,
			      loff_t actual_len, u64 *alloc_hint)
{
	return __btrfs_prealloc_file_range(inode, mode, start, num_bytes,
					   min_size, actual_len, alloc_hint,
					   NULL);
}

int btrfs_prealloc_file_range_trans(struct inode *inode,
				    struct btrfs_trans_handle *trans, int mode,
				    u64 start, u64 num_bytes, u64 min_size,
				    loff_t actual_len, u64 *alloc_hint)
{
	return __btrfs_prealloc_file_range(inode, mode, start, num_bytes,
					   min_size, actual_len, alloc_hint, trans);
}

7587 7588 7589 7590 7591
static int btrfs_set_page_dirty(struct page *page)
{
	return __set_page_dirty_nobuffers(page);
}

7592
static int btrfs_permission(struct inode *inode, int mask)
Y
Yan 已提交
7593
{
L
Li Zefan 已提交
7594
	struct btrfs_root *root = BTRFS_I(inode)->root;
7595
	umode_t mode = inode->i_mode;
L
Li Zefan 已提交
7596

7597 7598 7599 7600 7601 7602 7603
	if (mask & MAY_WRITE &&
	    (S_ISREG(mode) || S_ISDIR(mode) || S_ISLNK(mode))) {
		if (btrfs_root_readonly(root))
			return -EROFS;
		if (BTRFS_I(inode)->flags & BTRFS_INODE_READONLY)
			return -EACCES;
	}
7604
	return generic_permission(inode, mask);
Y
Yan 已提交
7605
}
C
Chris Mason 已提交
7606

7607
static const struct inode_operations btrfs_dir_inode_operations = {
7608
	.getattr	= btrfs_getattr,
C
Chris Mason 已提交
7609 7610 7611 7612 7613 7614 7615 7616 7617
	.lookup		= btrfs_lookup,
	.create		= btrfs_create,
	.unlink		= btrfs_unlink,
	.link		= btrfs_link,
	.mkdir		= btrfs_mkdir,
	.rmdir		= btrfs_rmdir,
	.rename		= btrfs_rename,
	.symlink	= btrfs_symlink,
	.setattr	= btrfs_setattr,
J
Josef Bacik 已提交
7618
	.mknod		= btrfs_mknod,
7619 7620
	.setxattr	= btrfs_setxattr,
	.getxattr	= btrfs_getxattr,
J
Josef Bacik 已提交
7621
	.listxattr	= btrfs_listxattr,
7622
	.removexattr	= btrfs_removexattr,
Y
Yan 已提交
7623
	.permission	= btrfs_permission,
7624
	.get_acl	= btrfs_get_acl,
C
Chris Mason 已提交
7625
};
7626
static const struct inode_operations btrfs_dir_ro_inode_operations = {
C
Chris Mason 已提交
7627
	.lookup		= btrfs_lookup,
Y
Yan 已提交
7628
	.permission	= btrfs_permission,
7629
	.get_acl	= btrfs_get_acl,
C
Chris Mason 已提交
7630
};
7631

7632
static const struct file_operations btrfs_dir_file_operations = {
C
Chris Mason 已提交
7633 7634
	.llseek		= generic_file_llseek,
	.read		= generic_read_dir,
7635
	.readdir	= btrfs_real_readdir,
C
Christoph Hellwig 已提交
7636
	.unlocked_ioctl	= btrfs_ioctl,
C
Chris Mason 已提交
7637
#ifdef CONFIG_COMPAT
C
Christoph Hellwig 已提交
7638
	.compat_ioctl	= btrfs_ioctl,
C
Chris Mason 已提交
7639
#endif
S
Sage Weil 已提交
7640
	.release        = btrfs_release_file,
7641
	.fsync		= btrfs_sync_file,
C
Chris Mason 已提交
7642 7643
};

7644
static struct extent_io_ops btrfs_extent_io_ops = {
7645
	.fill_delalloc = run_delalloc_range,
7646
	.submit_bio_hook = btrfs_submit_bio_hook,
7647
	.merge_bio_hook = btrfs_merge_bio_hook,
7648
	.readpage_end_io_hook = btrfs_readpage_end_io_hook,
7649
	.writepage_end_io_hook = btrfs_writepage_end_io_hook,
7650
	.writepage_start_hook = btrfs_writepage_start_hook,
C
Chris Mason 已提交
7651 7652
	.set_bit_hook = btrfs_set_bit_hook,
	.clear_bit_hook = btrfs_clear_bit_hook,
J
Josef Bacik 已提交
7653 7654
	.merge_extent_hook = btrfs_merge_extent_hook,
	.split_extent_hook = btrfs_split_extent_hook,
7655 7656
};

7657 7658 7659 7660 7661 7662 7663 7664 7665 7666 7667 7668
/*
 * btrfs doesn't support the bmap operation because swapfiles
 * use bmap to make a mapping of extents in the file.  They assume
 * these extents won't change over the life of the file and they
 * use the bmap result to do IO directly to the drive.
 *
 * the btrfs bmap call would return logical addresses that aren't
 * suitable for IO and they also will change frequently as COW
 * operations happen.  So, swapfile + btrfs == corruption.
 *
 * For now we're avoiding this by dropping bmap.
 */
7669
static const struct address_space_operations btrfs_aops = {
C
Chris Mason 已提交
7670 7671
	.readpage	= btrfs_readpage,
	.writepage	= btrfs_writepage,
C
Chris Mason 已提交
7672
	.writepages	= btrfs_writepages,
C
Chris Mason 已提交
7673
	.readpages	= btrfs_readpages,
7674
	.direct_IO	= btrfs_direct_IO,
7675 7676
	.invalidatepage = btrfs_invalidatepage,
	.releasepage	= btrfs_releasepage,
7677
	.set_page_dirty	= btrfs_set_page_dirty,
7678
	.error_remove_page = generic_error_remove_page,
C
Chris Mason 已提交
7679 7680
};

7681
static const struct address_space_operations btrfs_symlink_aops = {
C
Chris Mason 已提交
7682 7683
	.readpage	= btrfs_readpage,
	.writepage	= btrfs_writepage,
C
Chris Mason 已提交
7684 7685
	.invalidatepage = btrfs_invalidatepage,
	.releasepage	= btrfs_releasepage,
C
Chris Mason 已提交
7686 7687
};

7688
static const struct inode_operations btrfs_file_inode_operations = {
C
Chris Mason 已提交
7689 7690
	.getattr	= btrfs_getattr,
	.setattr	= btrfs_setattr,
7691 7692
	.setxattr	= btrfs_setxattr,
	.getxattr	= btrfs_getxattr,
J
Josef Bacik 已提交
7693
	.listxattr      = btrfs_listxattr,
7694
	.removexattr	= btrfs_removexattr,
Y
Yan 已提交
7695
	.permission	= btrfs_permission,
Y
Yehuda Sadeh 已提交
7696
	.fiemap		= btrfs_fiemap,
7697
	.get_acl	= btrfs_get_acl,
7698
	.update_time	= btrfs_update_time,
C
Chris Mason 已提交
7699
};
7700
static const struct inode_operations btrfs_special_inode_operations = {
J
Josef Bacik 已提交
7701 7702
	.getattr	= btrfs_getattr,
	.setattr	= btrfs_setattr,
Y
Yan 已提交
7703
	.permission	= btrfs_permission,
7704 7705
	.setxattr	= btrfs_setxattr,
	.getxattr	= btrfs_getxattr,
J
Josef Bacik 已提交
7706
	.listxattr	= btrfs_listxattr,
7707
	.removexattr	= btrfs_removexattr,
7708
	.get_acl	= btrfs_get_acl,
7709
	.update_time	= btrfs_update_time,
J
Josef Bacik 已提交
7710
};
7711
static const struct inode_operations btrfs_symlink_inode_operations = {
C
Chris Mason 已提交
7712 7713 7714
	.readlink	= generic_readlink,
	.follow_link	= page_follow_link_light,
	.put_link	= page_put_link,
7715
	.getattr	= btrfs_getattr,
7716
	.setattr	= btrfs_setattr,
Y
Yan 已提交
7717
	.permission	= btrfs_permission,
J
Jim Owens 已提交
7718 7719 7720 7721
	.setxattr	= btrfs_setxattr,
	.getxattr	= btrfs_getxattr,
	.listxattr	= btrfs_listxattr,
	.removexattr	= btrfs_removexattr,
7722
	.get_acl	= btrfs_get_acl,
7723
	.update_time	= btrfs_update_time,
C
Chris Mason 已提交
7724
};
7725

7726
const struct dentry_operations btrfs_dentry_operations = {
7727
	.d_delete	= btrfs_dentry_delete,
7728
	.d_release	= btrfs_dentry_release,
7729
};