inode.c 150.9 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>
C
Chris Mason 已提交
39
#include "compat.h"
C
Chris Mason 已提交
40 41 42 43 44 45
#include "ctree.h"
#include "disk-io.h"
#include "transaction.h"
#include "btrfs_inode.h"
#include "ioctl.h"
#include "print-tree.h"
46
#include "volumes.h"
47
#include "ordered-data.h"
48
#include "xattr.h"
49
#include "tree-log.h"
C
Chris Mason 已提交
50
#include "compression.h"
51
#include "locking.h"
C
Chris Mason 已提交
52 53 54 55 56 57 58 59 60

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

static struct inode_operations btrfs_dir_inode_operations;
static struct inode_operations btrfs_symlink_inode_operations;
static struct inode_operations btrfs_dir_ro_inode_operations;
J
Josef Bacik 已提交
61
static struct inode_operations btrfs_special_inode_operations;
C
Chris Mason 已提交
62 63 64 65
static struct inode_operations btrfs_file_inode_operations;
static struct address_space_operations btrfs_aops;
static struct address_space_operations btrfs_symlink_aops;
static struct file_operations btrfs_dir_file_operations;
66
static struct extent_io_ops btrfs_extent_io_ops;
C
Chris Mason 已提交
67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83

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;

#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,
};

84
static void btrfs_truncate(struct inode *inode);
C
Chris Mason 已提交
85
static int btrfs_finish_ordered_io(struct inode *inode, u64 start, u64 end);
86 87 88 89
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);
90

J
Jim Owens 已提交
91 92 93 94 95 96 97 98 99 100
static int btrfs_init_inode_security(struct inode *inode,  struct inode *dir)
{
	int err;

	err = btrfs_init_acl(inode, dir);
	if (!err)
		err = btrfs_xattr_security_init(inode, dir);
	return err;
}

C
Chris Mason 已提交
101 102 103 104 105
/*
 * 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
106
static noinline int insert_inline_extent(struct btrfs_trans_handle *trans,
C
Chris Mason 已提交
107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129
				struct btrfs_root *root, struct inode *inode,
				u64 start, size_t size, size_t compressed_size,
				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;
	int use_compress = 0;

	if (compressed_size && compressed_pages) {
		use_compress = 1;
		cur_size = compressed_size;
	}

C
Chris Mason 已提交
130 131
	path = btrfs_alloc_path();
	if (!path)
C
Chris Mason 已提交
132 133
		return -ENOMEM;

134
	path->leave_spinning = 1;
C
Chris Mason 已提交
135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162
	btrfs_set_trans_block_group(trans, inode);

	key.objectid = inode->i_ino;
	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);
	BUG_ON(ret);
	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);

	if (use_compress) {
		struct page *cpage;
		int i = 0;
C
Chris Mason 已提交
163
		while (compressed_size > 0) {
C
Chris Mason 已提交
164
			cpage = compressed_pages[i];
165
			cur_size = min_t(unsigned long, compressed_size,
C
Chris Mason 已提交
166 167
				       PAGE_CACHE_SIZE);

168
			kaddr = kmap_atomic(cpage, KM_USER0);
C
Chris Mason 已提交
169
			write_extent_buffer(leaf, kaddr, ptr, cur_size);
170
			kunmap_atomic(kaddr, KM_USER0);
C
Chris Mason 已提交
171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204

			i++;
			ptr += cur_size;
			compressed_size -= cur_size;
		}
		btrfs_set_file_extent_compression(leaf, ei,
						  BTRFS_COMPRESS_ZLIB);
	} else {
		page = find_get_page(inode->i_mapping,
				     start >> PAGE_CACHE_SHIFT);
		btrfs_set_file_extent_compression(leaf, ei, 0);
		kaddr = kmap_atomic(page, KM_USER0);
		offset = start & (PAGE_CACHE_SIZE - 1);
		write_extent_buffer(leaf, kaddr + offset, ptr, size);
		kunmap_atomic(kaddr, KM_USER0);
		page_cache_release(page);
	}
	btrfs_mark_buffer_dirty(leaf);
	btrfs_free_path(path);

	BTRFS_I(inode)->disk_i_size = inode->i_size;
	btrfs_update_inode(trans, root, inode);
	return 0;
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.
 */
205
static noinline int cow_file_range_inline(struct btrfs_trans_handle *trans,
C
Chris Mason 已提交
206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223
				 struct btrfs_root *root,
				 struct inode *inode, u64 start, u64 end,
				 size_t compressed_size,
				 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
224
	    actual_end >= PAGE_CACHE_SIZE ||
C
Chris Mason 已提交
225 226 227 228 229 230 231 232 233
	    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;
	}

	ret = btrfs_drop_extents(trans, root, inode, start,
C
Chris Mason 已提交
234 235
				 aligned_end, aligned_end, start,
				 &hint_byte, 1);
C
Chris Mason 已提交
236 237 238 239 240 241 242 243
	BUG_ON(ret);

	if (isize > actual_end)
		inline_len = min_t(u64, isize, actual_end);
	ret = insert_inline_extent(trans, root, inode, start,
				   inline_len, compressed_size,
				   compressed_pages);
	BUG_ON(ret);
C
Chris Mason 已提交
244
	btrfs_drop_extent_cache(inode, start, aligned_end - 1, 0);
C
Chris Mason 已提交
245 246 247
	return 0;
}

248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284
struct async_extent {
	u64 start;
	u64 ram_size;
	u64 compressed_size;
	struct page **pages;
	unsigned long nr_pages;
	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,
				     unsigned long nr_pages)
{
	struct async_extent *async_extent;

	async_extent = kmalloc(sizeof(*async_extent), GFP_NOFS);
	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;
	list_add_tail(&async_extent->list, &cow->extents);
	return 0;
}

C
Chris Mason 已提交
285
/*
286 287 288
 * 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
289
 *
290 291 292 293 294
 * 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
295
 *
296 297 298 299
 * 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
300
 */
301 302 303 304 305
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)
306 307 308
{
	struct btrfs_root *root = BTRFS_I(inode)->root;
	struct btrfs_trans_handle *trans;
309
	u64 num_bytes;
C
Chris Mason 已提交
310 311
	u64 orig_start;
	u64 disk_num_bytes;
312
	u64 blocksize = root->sectorsize;
C
Chris Mason 已提交
313
	u64 actual_end;
314
	u64 isize = i_size_read(inode);
315
	int ret = 0;
C
Chris Mason 已提交
316 317 318 319 320 321
	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;
322
	unsigned long max_uncompressed = 128 * 1024;
C
Chris Mason 已提交
323 324
	int i;
	int will_compress;
325

C
Chris Mason 已提交
326 327
	orig_start = start;

328
	actual_end = min_t(u64, isize, end + 1);
C
Chris Mason 已提交
329 330 331 332
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);
333

334 335 336 337 338 339 340 341 342 343 344 345 346
	/*
	 * 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
347 348 349 350
	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
351 352 353 354 355 356 357
	 * 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
358 359
	 */
	total_compressed = min(total_compressed, max_uncompressed);
360
	num_bytes = (end - start + blocksize) & ~(blocksize - 1);
361
	num_bytes = max(blocksize,  num_bytes);
C
Chris Mason 已提交
362 363 364
	disk_num_bytes = num_bytes;
	total_in = 0;
	ret = 0;
365

366 367 368 369
	/*
	 * 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
370
	 */
371
	if (!(BTRFS_I(inode)->flags & BTRFS_INODE_NOCOMPRESS) &&
C
Chris Mason 已提交
372 373
	    btrfs_test_opt(root, COMPRESS)) {
		WARN_ON(pages);
374
		pages = kzalloc(sizeof(struct page *) * nr_pages, GFP_NOFS);
C
Chris Mason 已提交
375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401

		ret = btrfs_zlib_compress_pages(inode->i_mapping, start,
						total_compressed, pages,
						nr_pages, &nr_pages_ret,
						&total_in,
						&total_compressed,
						max_compressed);

		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) {
				kaddr = kmap_atomic(page, KM_USER0);
				memset(kaddr + offset, 0,
				       PAGE_CACHE_SIZE - offset);
				kunmap_atomic(kaddr, KM_USER0);
			}
			will_compress = 1;
		}
	}
	if (start == 0) {
402 403 404 405
		trans = btrfs_join_transaction(root, 1);
		BUG_ON(!trans);
		btrfs_set_trans_block_group(trans, inode);

C
Chris Mason 已提交
406
		/* lets try to make an inline extent */
407
		if (ret || total_in < (actual_end - start)) {
C
Chris Mason 已提交
408
			/* we didn't compress the entire range, try
409
			 * to make an uncompressed inline extent.
C
Chris Mason 已提交
410 411 412 413
			 */
			ret = cow_file_range_inline(trans, root, inode,
						    start, end, 0, NULL);
		} else {
414
			/* try making a compressed inline extent */
C
Chris Mason 已提交
415 416 417 418
			ret = cow_file_range_inline(trans, root, inode,
						    start, end,
						    total_compressed, pages);
		}
419
		btrfs_end_transaction(trans, root);
C
Chris Mason 已提交
420
		if (ret == 0) {
421 422 423 424 425
			/*
			 * inline extent creation worked, we don't need
			 * to create any more async work items.  Unlock
			 * and free up our temp pages.
			 */
C
Chris Mason 已提交
426 427
			extent_clear_unlock_delalloc(inode,
						     &BTRFS_I(inode)->io_tree,
428
						     start, end, NULL, 1, 0,
429
						     0, 1, 1, 1, 0);
C
Chris Mason 已提交
430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462
			ret = 0;
			goto free_pages_out;
		}
	}

	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 {
			disk_num_bytes = total_compressed;
			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
463
			WARN_ON(pages[i]->mapping);
C
Chris Mason 已提交
464 465 466 467 468 469 470 471
			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 */
472
		BTRFS_I(inode)->flags |= BTRFS_INODE_NOCOMPRESS;
C
Chris Mason 已提交
473
	}
474 475
	if (will_compress) {
		*num_added += 1;
C
Chris Mason 已提交
476

477 478 479 480 481 482
		/* 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,
				 total_compressed, pages, nr_pages_ret);
483

484
		if (start + num_bytes < end && start + num_bytes < actual_end) {
485 486 487 488 489 490
			start += num_bytes;
			pages = NULL;
			cond_resched();
			goto again;
		}
	} else {
491
cleanup_and_bail_uncompressed:
492 493 494 495 496 497 498 499 500 501 502 503 504 505 506
		/*
		 * 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 */
		}
		add_async_extent(async_cow, start, end - start + 1, 0, NULL, 0);
		*num_added += 1;
	}
507

508 509 510 511 512 513 514 515
out:
	return 0;

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
516
	kfree(pages);
517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544

	goto out;
}

/*
 * 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;
	int ret;

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

	trans = btrfs_join_transaction(root, 1);

C
Chris Mason 已提交
545
	while (!list_empty(&async_cow->extents)) {
546 547 548
		async_extent = list_entry(async_cow->extents.next,
					  struct async_extent, list);
		list_del(&async_extent->list);
C
Chris Mason 已提交
549

550 551 552 553 554 555 556 557
		io_tree = &BTRFS_I(inode)->io_tree;

		/* 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,
C
Chris Mason 已提交
558 559
				    async_extent->start +
				    async_extent->ram_size - 1, GFP_NOFS);
560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576

			/* allocate blocks */
			cow_file_range(inode, async_cow->locked_page,
				       async_extent->start,
				       async_extent->start +
				       async_extent->ram_size - 1,
				       &page_started, &nr_written, 0);

			/*
			 * 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.
			 */
			if (!page_started)
				extent_write_locked_range(io_tree,
						  inode, async_extent->start,
C
Chris Mason 已提交
577
						  async_extent->start +
578 579 580 581 582 583 584 585 586 587 588
						  async_extent->ram_size - 1,
						  btrfs_get_extent,
						  WB_SYNC_ALL);
			kfree(async_extent);
			cond_resched();
			continue;
		}

		lock_extent(io_tree, async_extent->start,
			    async_extent->start + async_extent->ram_size - 1,
			    GFP_NOFS);
C
Chris Mason 已提交
589
		/*
590 591
		 * here we're doing allocation and writeback of the
		 * compressed pages
C
Chris Mason 已提交
592
		 */
593 594 595 596 597 598 599 600 601 602 603 604 605
		btrfs_drop_extent_cache(inode, async_extent->start,
					async_extent->start +
					async_extent->ram_size - 1, 0);

		ret = btrfs_reserve_extent(trans, root,
					   async_extent->compressed_size,
					   async_extent->compressed_size,
					   0, alloc_hint,
					   (u64)-1, &ins, 1);
		BUG_ON(ret);
		em = alloc_extent_map(GFP_NOFS);
		em->start = async_extent->start;
		em->len = async_extent->ram_size;
606
		em->orig_start = em->start;
C
Chris Mason 已提交
607

608 609 610 611 612 613
		em->block_start = ins.objectid;
		em->block_len = ins.offset;
		em->bdev = root->fs_info->fs_devices->latest_bdev;
		set_bit(EXTENT_FLAG_PINNED, &em->flags);
		set_bit(EXTENT_FLAG_COMPRESSED, &em->flags);

C
Chris Mason 已提交
614
		while (1) {
615
			write_lock(&em_tree->lock);
616
			ret = add_extent_mapping(em_tree, em);
617
			write_unlock(&em_tree->lock);
618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643
			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);
		}

		ret = btrfs_add_ordered_extent(inode, async_extent->start,
					       ins.objectid,
					       async_extent->ram_size,
					       ins.offset,
					       BTRFS_ORDERED_COMPRESSED);
		BUG_ON(ret);

		btrfs_end_transaction(trans, root);

		/*
		 * clear dirty, set writeback and unlock the pages.
		 */
		extent_clear_unlock_delalloc(inode,
					     &BTRFS_I(inode)->io_tree,
					     async_extent->start,
					     async_extent->start +
					     async_extent->ram_size - 1,
644
					     NULL, 1, 1, 0, 1, 1, 0, 0);
645 646

		ret = btrfs_submit_compressed_write(inode,
C
Chris Mason 已提交
647 648 649 650 651
				    async_extent->start,
				    async_extent->ram_size,
				    ins.objectid,
				    ins.offset, async_extent->pages,
				    async_extent->nr_pages);
652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691

		BUG_ON(ret);
		trans = btrfs_join_transaction(root, 1);
		alloc_hint = ins.objectid + ins.offset;
		kfree(async_extent);
		cond_resched();
	}

	btrfs_end_transaction(trans, root);
	return 0;
}

/*
 * 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;
	u64 actual_end;
692
	u64 isize = i_size_read(inode);
693 694 695 696 697 698 699 700 701
	struct btrfs_key ins;
	struct extent_map *em;
	struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
	int ret = 0;

	trans = btrfs_join_transaction(root, 1);
	BUG_ON(!trans);
	btrfs_set_trans_block_group(trans, inode);

702
	actual_end = min_t(u64, isize, end + 1);
703 704 705 706 707 708 709 710 711 712 713 714 715 716

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

	if (start == 0) {
		/* lets try to make an inline extent */
		ret = cow_file_range_inline(trans, root, inode,
					    start, end, 0, NULL);
		if (ret == 0) {
			extent_clear_unlock_delalloc(inode,
						     &BTRFS_I(inode)->io_tree,
						     start, end, NULL, 1, 1,
717
						     1, 1, 1, 1, 0);
718 719 720 721 722 723 724 725 726 727 728
			*nr_written = *nr_written +
			     (end - start + PAGE_CACHE_SIZE) / PAGE_CACHE_SIZE;
			*page_started = 1;
			ret = 0;
			goto out;
		}
	}

	BUG_ON(disk_num_bytes >
	       btrfs_super_total_bytes(&root->fs_info->super_copy));

729 730 731 732 733 734 735 736 737

	read_lock(&BTRFS_I(inode)->extent_tree.lock);
	em = search_extent_mapping(&BTRFS_I(inode)->extent_tree,
				   start, num_bytes);
	if (em) {
		alloc_hint = em->block_start;
		free_extent_map(em);
	}
	read_unlock(&BTRFS_I(inode)->extent_tree.lock);
738 739
	btrfs_drop_extent_cache(inode, start, start + num_bytes - 1, 0);

C
Chris Mason 已提交
740
	while (disk_num_bytes > 0) {
C
Chris Mason 已提交
741
		cur_alloc_size = min(disk_num_bytes, root->fs_info->max_extent);
742
		ret = btrfs_reserve_extent(trans, root, cur_alloc_size,
743
					   root->sectorsize, 0, alloc_hint,
744
					   (u64)-1, &ins, 1);
C
Chris Mason 已提交
745 746
		BUG_ON(ret);

747 748
		em = alloc_extent_map(GFP_NOFS);
		em->start = start;
749
		em->orig_start = em->start;
750 751
		ram_size = ins.offset;
		em->len = ins.offset;
C
Chris Mason 已提交
752

753
		em->block_start = ins.objectid;
C
Chris Mason 已提交
754
		em->block_len = ins.offset;
755
		em->bdev = root->fs_info->fs_devices->latest_bdev;
756
		set_bit(EXTENT_FLAG_PINNED, &em->flags);
C
Chris Mason 已提交
757

C
Chris Mason 已提交
758
		while (1) {
759
			write_lock(&em_tree->lock);
760
			ret = add_extent_mapping(em_tree, em);
761
			write_unlock(&em_tree->lock);
762 763 764 765 766
			if (ret != -EEXIST) {
				free_extent_map(em);
				break;
			}
			btrfs_drop_extent_cache(inode, start,
C
Chris Mason 已提交
767
						start + ram_size - 1, 0);
768 769
		}

770
		cur_alloc_size = ins.offset;
771
		ret = btrfs_add_ordered_extent(inode, start, ins.objectid,
772
					       ram_size, cur_alloc_size, 0);
773
		BUG_ON(ret);
C
Chris Mason 已提交
774

775 776 777 778 779 780 781
		if (root->root_key.objectid ==
		    BTRFS_DATA_RELOC_TREE_OBJECTID) {
			ret = btrfs_reloc_clone_csums(inode, start,
						      cur_alloc_size);
			BUG_ON(ret);
		}

C
Chris Mason 已提交
782
		if (disk_num_bytes < cur_alloc_size)
783
			break;
C
Chris Mason 已提交
784

C
Chris Mason 已提交
785 786 787
		/* 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
788 789 790
		 *
		 * Do set the Private2 bit so we know this page was properly
		 * setup for writepage
C
Chris Mason 已提交
791 792 793
		 */
		extent_clear_unlock_delalloc(inode, &BTRFS_I(inode)->io_tree,
					     start, start + ram_size - 1,
794
					     locked_page, unlock, 1,
795
					     1, 0, 0, 0, 1);
C
Chris Mason 已提交
796
		disk_num_bytes -= cur_alloc_size;
797 798 799
		num_bytes -= cur_alloc_size;
		alloc_hint = ins.objectid + ins.offset;
		start += cur_alloc_size;
800 801
	}
out:
802
	ret = 0;
803
	btrfs_end_transaction(trans, root);
C
Chris Mason 已提交
804

805
	return ret;
806
}
C
Chris Mason 已提交
807

808 809 810 811 812 813 814 815 816 817 818 819 820 821 822 823 824 825 826 827 828 829 830 831 832 833 834 835 836 837 838 839 840 841 842 843 844 845
/*
 * 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);
	if (num_added == 0)
		async_cow->inode = NULL;
}

/*
 * 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
846
	if (async_cow->inode)
847 848
		submit_compressed_extents(async_cow->inode, async_cow);
}
C
Chris Mason 已提交
849

850 851 852 853 854 855 856 857 858 859 860 861 862 863 864 865 866 867
static noinline void async_cow_free(struct btrfs_work *work)
{
	struct async_cow *async_cow;
	async_cow = container_of(work, struct async_cow, work);
	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;

	clear_extent_bit(&BTRFS_I(inode)->io_tree, start, end, EXTENT_LOCKED |
868
			 EXTENT_DELALLOC, 1, 0, NULL, GFP_NOFS);
C
Chris Mason 已提交
869
	while (start < end) {
870 871 872 873 874 875
		async_cow = kmalloc(sizeof(*async_cow), GFP_NOFS);
		async_cow->inode = inode;
		async_cow->root = root;
		async_cow->locked_page = locked_page;
		async_cow->start = start;

876
		if (BTRFS_I(inode)->flags & BTRFS_INODE_NOCOMPRESS)
877 878 879 880 881 882 883 884 885 886 887 888 889 890 891 892 893 894 895 896 897 898 899 900 901
			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
902
		while (atomic_read(&root->fs_info->async_submit_draining) &&
903 904 905 906 907 908 909 910 911 912 913
		      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;
914 915
}

C
Chris Mason 已提交
916
static noinline int csum_exist_in_range(struct btrfs_root *root,
917 918 919 920 921 922
					u64 bytenr, u64 num_bytes)
{
	int ret;
	struct btrfs_ordered_sum *sums;
	LIST_HEAD(list);

Y
Yan Zheng 已提交
923 924
	ret = btrfs_lookup_csums_range(root->fs_info->csum_root, bytenr,
				       bytenr + num_bytes - 1, &list);
925 926 927 928 929 930 931 932 933 934 935
	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
936 937 938 939 940 941 942
/*
 * 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
 */
943 944
static noinline int run_delalloc_nocow(struct inode *inode,
				       struct page *locked_page,
945 946
			      u64 start, u64 end, int *page_started, int force,
			      unsigned long *nr_written)
947 948
{
	struct btrfs_root *root = BTRFS_I(inode)->root;
949
	struct btrfs_trans_handle *trans;
950 951
	struct extent_buffer *leaf;
	struct btrfs_path *path;
Y
Yan Zheng 已提交
952
	struct btrfs_file_extent_item *fi;
953
	struct btrfs_key found_key;
Y
Yan Zheng 已提交
954 955 956
	u64 cow_start;
	u64 cur_offset;
	u64 extent_end;
957
	u64 extent_offset;
Y
Yan Zheng 已提交
958 959 960 961
	u64 disk_bytenr;
	u64 num_bytes;
	int extent_type;
	int ret;
Y
Yan Zheng 已提交
962
	int type;
Y
Yan Zheng 已提交
963 964
	int nocow;
	int check_prev = 1;
965 966 967

	path = btrfs_alloc_path();
	BUG_ON(!path);
968 969
	trans = btrfs_join_transaction(root, 1);
	BUG_ON(!trans);
970

Y
Yan Zheng 已提交
971 972 973 974 975 976 977 978 979 980 981 982 983 984 985 986 987 988 989 990 991 992 993 994 995
	cow_start = (u64)-1;
	cur_offset = start;
	while (1) {
		ret = btrfs_lookup_file_extent(trans, root, path, inode->i_ino,
					       cur_offset, 0);
		BUG_ON(ret < 0);
		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);
			if (found_key.objectid == inode->i_ino &&
			    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);
			if (ret < 0)
				BUG_ON(1);
			if (ret > 0)
				break;
			leaf = path->nodes[0];
		}
996

Y
Yan Zheng 已提交
997 998
		nocow = 0;
		disk_bytenr = 0;
999
		num_bytes = 0;
Y
Yan Zheng 已提交
1000 1001 1002 1003 1004 1005 1006 1007 1008 1009 1010 1011 1012 1013 1014 1015
		btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);

		if (found_key.objectid > inode->i_ino ||
		    found_key.type > BTRFS_EXTENT_DATA_KEY ||
		    found_key.offset > end)
			break;

		if (found_key.offset > cur_offset) {
			extent_end = found_key.offset;
			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
1016 1017
		if (extent_type == BTRFS_FILE_EXTENT_REG ||
		    extent_type == BTRFS_FILE_EXTENT_PREALLOC) {
Y
Yan Zheng 已提交
1018
			disk_bytenr = btrfs_file_extent_disk_bytenr(leaf, fi);
1019
			extent_offset = btrfs_file_extent_offset(leaf, fi);
Y
Yan Zheng 已提交
1020 1021 1022 1023 1024 1025
			extent_end = found_key.offset +
				btrfs_file_extent_num_bytes(leaf, fi);
			if (extent_end <= start) {
				path->slots[0]++;
				goto next_slot;
			}
1026 1027
			if (disk_bytenr == 0)
				goto out_check;
Y
Yan Zheng 已提交
1028 1029 1030 1031
			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
1032 1033
			if (extent_type == BTRFS_FILE_EXTENT_REG && !force)
				goto out_check;
1034
			if (btrfs_extent_readonly(root, disk_bytenr))
Y
Yan Zheng 已提交
1035
				goto out_check;
1036
			if (btrfs_cross_ref_exist(trans, root, inode->i_ino,
1037 1038
						  found_key.offset -
						  extent_offset, disk_bytenr))
1039
				goto out_check;
1040
			disk_bytenr += extent_offset;
1041 1042 1043 1044 1045 1046 1047 1048 1049
			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
1050 1051 1052 1053 1054 1055 1056 1057 1058 1059 1060 1061 1062 1063 1064 1065 1066 1067 1068 1069 1070
			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;
1071 1072 1073
		}

		btrfs_release_path(root, path);
Y
Yan Zheng 已提交
1074 1075
		if (cow_start != (u64)-1) {
			ret = cow_file_range(inode, locked_page, cow_start,
1076 1077
					found_key.offset - 1, page_started,
					nr_written, 1);
Y
Yan Zheng 已提交
1078 1079
			BUG_ON(ret);
			cow_start = (u64)-1;
1080
		}
Y
Yan Zheng 已提交
1081

Y
Yan Zheng 已提交
1082 1083 1084 1085 1086 1087
		if (extent_type == BTRFS_FILE_EXTENT_PREALLOC) {
			struct extent_map *em;
			struct extent_map_tree *em_tree;
			em_tree = &BTRFS_I(inode)->extent_tree;
			em = alloc_extent_map(GFP_NOFS);
			em->start = cur_offset;
1088
			em->orig_start = em->start;
Y
Yan Zheng 已提交
1089 1090 1091 1092 1093 1094
			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) {
1095
				write_lock(&em_tree->lock);
Y
Yan Zheng 已提交
1096
				ret = add_extent_mapping(em_tree, em);
1097
				write_unlock(&em_tree->lock);
Y
Yan Zheng 已提交
1098 1099 1100 1101 1102 1103 1104 1105 1106 1107 1108
				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
1109 1110

		ret = btrfs_add_ordered_extent(inode, cur_offset, disk_bytenr,
Y
Yan Zheng 已提交
1111 1112
					       num_bytes, num_bytes, type);
		BUG_ON(ret);
1113

Y
Yan Zheng 已提交
1114 1115
		extent_clear_unlock_delalloc(inode, &BTRFS_I(inode)->io_tree,
					cur_offset, cur_offset + num_bytes - 1,
1116
					locked_page, 1, 1, 1, 0, 0, 0, 1);
Y
Yan Zheng 已提交
1117 1118 1119
		cur_offset = extent_end;
		if (cur_offset > end)
			break;
1120
	}
Y
Yan Zheng 已提交
1121 1122 1123 1124 1125 1126
	btrfs_release_path(root, path);

	if (cur_offset <= end && cow_start == (u64)-1)
		cow_start = cur_offset;
	if (cow_start != (u64)-1) {
		ret = cow_file_range(inode, locked_page, cow_start, end,
1127
				     page_started, nr_written, 1);
Y
Yan Zheng 已提交
1128 1129 1130 1131 1132
		BUG_ON(ret);
	}

	ret = btrfs_end_transaction(trans, root);
	BUG_ON(ret);
1133
	btrfs_free_path(path);
Y
Yan Zheng 已提交
1134
	return 0;
1135 1136
}

C
Chris Mason 已提交
1137 1138 1139
/*
 * extent_io.c call back to do delayed allocation processing
 */
C
Chris Mason 已提交
1140
static int run_delalloc_range(struct inode *inode, struct page *locked_page,
1141 1142
			      u64 start, u64 end, int *page_started,
			      unsigned long *nr_written)
1143 1144
{
	int ret;
1145
	struct btrfs_root *root = BTRFS_I(inode)->root;
1146

1147
	if (BTRFS_I(inode)->flags & BTRFS_INODE_NODATACOW)
C
Chris Mason 已提交
1148
		ret = run_delalloc_nocow(inode, locked_page, start, end,
C
Chris Mason 已提交
1149
					 page_started, 1, nr_written);
1150
	else if (BTRFS_I(inode)->flags & BTRFS_INODE_PREALLOC)
Y
Yan Zheng 已提交
1151
		ret = run_delalloc_nocow(inode, locked_page, start, end,
C
Chris Mason 已提交
1152
					 page_started, 0, nr_written);
1153 1154 1155
	else if (!btrfs_test_opt(root, COMPRESS))
		ret = cow_file_range(inode, locked_page, start, end,
				      page_started, nr_written, 1);
1156
	else
1157
		ret = cow_file_range_async(inode, locked_page, start, end,
C
Chris Mason 已提交
1158
					   page_started, nr_written);
1159 1160 1161
	return ret;
}

C
Chris Mason 已提交
1162 1163 1164 1165 1166
/*
 * 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.
 */
1167
static int btrfs_set_bit_hook(struct inode *inode, u64 start, u64 end,
C
Chris Mason 已提交
1168
		       unsigned long old, unsigned long bits)
1169
{
1170 1171 1172 1173 1174
	/*
	 * set_bit and clear bit hooks normally require _irqsave/restore
	 * but in this case, we are only testeing for the DELALLOC
	 * bit, which is only set or cleared with irqs on
	 */
C
Chris Mason 已提交
1175
	if (!(old & EXTENT_DELALLOC) && (bits & EXTENT_DELALLOC)) {
1176
		struct btrfs_root *root = BTRFS_I(inode)->root;
J
Josef Bacik 已提交
1177
		btrfs_delalloc_reserve_space(root, inode, end - start + 1);
1178
		spin_lock(&root->fs_info->delalloc_lock);
C
Chris Mason 已提交
1179
		BTRFS_I(inode)->delalloc_bytes += end - start + 1;
1180
		root->fs_info->delalloc_bytes += end - start + 1;
1181 1182 1183 1184
		if (list_empty(&BTRFS_I(inode)->delalloc_inodes)) {
			list_add_tail(&BTRFS_I(inode)->delalloc_inodes,
				      &root->fs_info->delalloc_inodes);
		}
1185
		spin_unlock(&root->fs_info->delalloc_lock);
1186 1187 1188 1189
	}
	return 0;
}

C
Chris Mason 已提交
1190 1191 1192
/*
 * extent_io.c clear_bit_hook, see set_bit_hook for why
 */
1193
static int btrfs_clear_bit_hook(struct inode *inode, u64 start, u64 end,
C
Chris Mason 已提交
1194
			 unsigned long old, unsigned long bits)
1195
{
1196 1197 1198 1199 1200
	/*
	 * set_bit and clear bit hooks normally require _irqsave/restore
	 * but in this case, we are only testeing for the DELALLOC
	 * bit, which is only set or cleared with irqs on
	 */
C
Chris Mason 已提交
1201
	if ((old & EXTENT_DELALLOC) && (bits & EXTENT_DELALLOC)) {
1202
		struct btrfs_root *root = BTRFS_I(inode)->root;
1203

1204
		spin_lock(&root->fs_info->delalloc_lock);
C
Chris Mason 已提交
1205
		if (end - start + 1 > root->fs_info->delalloc_bytes) {
C
Chris Mason 已提交
1206 1207 1208 1209 1210
			printk(KERN_INFO "btrfs warning: delalloc account "
			       "%llu %llu\n",
			       (unsigned long long)end - start + 1,
			       (unsigned long long)
			       root->fs_info->delalloc_bytes);
J
Josef Bacik 已提交
1211
			btrfs_delalloc_free_space(root, inode, (u64)-1);
C
Chris Mason 已提交
1212
			root->fs_info->delalloc_bytes = 0;
C
Chris Mason 已提交
1213
			BTRFS_I(inode)->delalloc_bytes = 0;
C
Chris Mason 已提交
1214
		} else {
J
Josef Bacik 已提交
1215 1216
			btrfs_delalloc_free_space(root, inode,
						  end - start + 1);
C
Chris Mason 已提交
1217
			root->fs_info->delalloc_bytes -= end - start + 1;
C
Chris Mason 已提交
1218
			BTRFS_I(inode)->delalloc_bytes -= end - start + 1;
C
Chris Mason 已提交
1219
		}
1220 1221 1222 1223
		if (BTRFS_I(inode)->delalloc_bytes == 0 &&
		    !list_empty(&BTRFS_I(inode)->delalloc_inodes)) {
			list_del_init(&BTRFS_I(inode)->delalloc_inodes);
		}
1224
		spin_unlock(&root->fs_info->delalloc_lock);
1225 1226 1227 1228
	}
	return 0;
}

C
Chris Mason 已提交
1229 1230 1231 1232
/*
 * 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
 */
1233
int btrfs_merge_bio_hook(struct page *page, unsigned long offset,
C
Chris Mason 已提交
1234 1235
			 size_t size, struct bio *bio,
			 unsigned long bio_flags)
1236 1237 1238
{
	struct btrfs_root *root = BTRFS_I(page->mapping->host)->root;
	struct btrfs_mapping_tree *map_tree;
1239
	u64 logical = (u64)bio->bi_sector << 9;
1240 1241 1242 1243
	u64 length = 0;
	u64 map_length;
	int ret;

1244 1245 1246
	if (bio_flags & EXTENT_BIO_COMPRESSED)
		return 0;

1247
	length = bio->bi_size;
1248 1249
	map_tree = &root->fs_info->mapping_tree;
	map_length = length;
1250
	ret = btrfs_map_block(map_tree, READ, logical,
1251
			      &map_length, NULL, 0);
1252

C
Chris Mason 已提交
1253
	if (map_length < length + size)
1254 1255 1256 1257
		return 1;
	return 0;
}

C
Chris Mason 已提交
1258 1259 1260 1261 1262 1263 1264 1265
/*
 * 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
1266 1267 1268
static int __btrfs_submit_bio_start(struct inode *inode, int rw,
				    struct bio *bio, int mirror_num,
				    unsigned long bio_flags)
1269 1270 1271
{
	struct btrfs_root *root = BTRFS_I(inode)->root;
	int ret = 0;
1272

1273
	ret = btrfs_csum_one_bio(root, inode, bio, 0, 0);
1274
	BUG_ON(ret);
C
Chris Mason 已提交
1275 1276
	return 0;
}
1277

C
Chris Mason 已提交
1278 1279 1280 1281 1282 1283 1284 1285
/*
 * 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
 */
1286
static int __btrfs_submit_bio_done(struct inode *inode, int rw, struct bio *bio,
C
Chris Mason 已提交
1287 1288 1289
			  int mirror_num, unsigned long bio_flags)
{
	struct btrfs_root *root = BTRFS_I(inode)->root;
1290
	return btrfs_map_bio(root, rw, bio, mirror_num, 1);
1291 1292
}

C
Chris Mason 已提交
1293
/*
1294 1295
 * 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
1296
 */
1297
static int btrfs_submit_bio_hook(struct inode *inode, int rw, struct bio *bio,
C
Chris Mason 已提交
1298
			  int mirror_num, unsigned long bio_flags)
1299 1300 1301
{
	struct btrfs_root *root = BTRFS_I(inode)->root;
	int ret = 0;
1302
	int skip_sum;
1303

1304
	skip_sum = BTRFS_I(inode)->flags & BTRFS_INODE_NODATASUM;
1305

1306 1307
	ret = btrfs_bio_wq_end_io(root->fs_info, bio, 0);
	BUG_ON(ret);
1308

1309
	if (!(rw & (1 << BIO_RW))) {
1310
		if (bio_flags & EXTENT_BIO_COMPRESSED) {
C
Chris Mason 已提交
1311 1312
			return btrfs_submit_compressed_read(inode, bio,
						    mirror_num, bio_flags);
1313 1314
		} else if (!skip_sum)
			btrfs_lookup_bio_sums(root, inode, bio, NULL);
1315
		goto mapit;
1316
	} else if (!skip_sum) {
1317 1318 1319
		/* csum items have already been cloned */
		if (root->root_key.objectid == BTRFS_DATA_RELOC_TREE_OBJECTID)
			goto mapit;
1320 1321
		/* we're doing a write, do the async checksumming */
		return btrfs_wq_submit_bio(BTRFS_I(inode)->root->fs_info,
1322
				   inode, rw, bio, mirror_num,
C
Chris Mason 已提交
1323 1324
				   bio_flags, __btrfs_submit_bio_start,
				   __btrfs_submit_bio_done);
1325 1326
	}

1327
mapit:
1328
	return btrfs_map_bio(root, rw, bio, mirror_num, 0);
1329
}
C
Chris Mason 已提交
1330

C
Chris Mason 已提交
1331 1332 1333 1334
/*
 * 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.
 */
1335
static noinline int add_pending_csums(struct btrfs_trans_handle *trans,
1336 1337 1338 1339 1340 1341
			     struct inode *inode, u64 file_offset,
			     struct list_head *list)
{
	struct btrfs_ordered_sum *sum;

	btrfs_set_trans_block_group(trans, inode);
Q
Qinghuang Feng 已提交
1342 1343

	list_for_each_entry(sum, list, list) {
1344 1345
		btrfs_csum_file_blocks(trans,
		       BTRFS_I(inode)->root->fs_info->csum_root, sum);
1346 1347 1348 1349
	}
	return 0;
}

1350 1351
int btrfs_set_extent_delalloc(struct inode *inode, u64 start, u64 end)
{
C
Chris Mason 已提交
1352
	if ((end & (PAGE_CACHE_SIZE - 1)) == 0)
1353
		WARN_ON(1);
1354 1355 1356 1357
	return set_extent_delalloc(&BTRFS_I(inode)->io_tree, start, end,
				   GFP_NOFS);
}

C
Chris Mason 已提交
1358
/* see btrfs_writepage_start_hook for details on why this is required */
1359 1360 1361 1362 1363
struct btrfs_writepage_fixup {
	struct page *page;
	struct btrfs_work work;
};

1364
static void btrfs_writepage_fixup_worker(struct btrfs_work *work)
1365 1366 1367 1368 1369 1370 1371 1372 1373 1374
{
	struct btrfs_writepage_fixup *fixup;
	struct btrfs_ordered_extent *ordered;
	struct page *page;
	struct inode *inode;
	u64 page_start;
	u64 page_end;

	fixup = container_of(work, struct btrfs_writepage_fixup, work);
	page = fixup->page;
C
Chris Mason 已提交
1375
again:
1376 1377 1378 1379 1380 1381 1382 1383 1384 1385 1386
	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;

	lock_extent(&BTRFS_I(inode)->io_tree, page_start, page_end, GFP_NOFS);
C
Chris Mason 已提交
1387 1388

	/* already ordered? We're done */
1389
	if (PagePrivate2(page))
1390
		goto out;
C
Chris Mason 已提交
1391 1392 1393 1394 1395 1396 1397 1398 1399

	ordered = btrfs_lookup_ordered_extent(inode, page_start);
	if (ordered) {
		unlock_extent(&BTRFS_I(inode)->io_tree, page_start,
			      page_end, GFP_NOFS);
		unlock_page(page);
		btrfs_start_ordered_extent(inode, ordered, 1);
		goto again;
	}
1400

1401
	btrfs_set_extent_delalloc(inode, page_start, page_end);
1402 1403 1404 1405 1406 1407 1408 1409 1410 1411 1412 1413 1414 1415
	ClearPageChecked(page);
out:
	unlock_extent(&BTRFS_I(inode)->io_tree, page_start, page_end, GFP_NOFS);
out_page:
	unlock_page(page);
	page_cache_release(page);
}

/*
 * 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
1416
 * In our case any range that doesn't have the ORDERED bit set
1417 1418 1419 1420
 * 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.
 */
1421
static int btrfs_writepage_start_hook(struct page *page, u64 start, u64 end)
1422 1423 1424 1425 1426
{
	struct inode *inode = page->mapping->host;
	struct btrfs_writepage_fixup *fixup;
	struct btrfs_root *root = BTRFS_I(inode)->root;

1427 1428
	/* this page is properly in the ordered list */
	if (TestClearPagePrivate2(page))
1429 1430 1431 1432 1433 1434 1435 1436
		return 0;

	if (PageChecked(page))
		return -EAGAIN;

	fixup = kzalloc(sizeof(*fixup), GFP_NOFS);
	if (!fixup)
		return -EAGAIN;
1437

1438 1439 1440 1441 1442 1443 1444 1445
	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);
	return -EAGAIN;
}

Y
Yan Zheng 已提交
1446 1447 1448 1449
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,
1450
				       u64 locked_end,
Y
Yan Zheng 已提交
1451 1452 1453 1454 1455 1456 1457 1458 1459 1460 1461 1462 1463 1464
				       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();
	BUG_ON(!path);

1465
	path->leave_spinning = 1;
C
Chris Mason 已提交
1466 1467 1468 1469 1470 1471 1472 1473 1474 1475

	/*
	 * 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
1476
	ret = btrfs_drop_extents(trans, root, inode, file_pos,
1477
				 file_pos + num_bytes, locked_end,
C
Chris Mason 已提交
1478
				 file_pos, &hint, 0);
Y
Yan Zheng 已提交
1479 1480 1481 1482 1483 1484 1485 1486 1487 1488 1489 1490 1491 1492 1493 1494 1495 1496 1497 1498
	BUG_ON(ret);

	ins.objectid = inode->i_ino;
	ins.offset = file_pos;
	ins.type = BTRFS_EXTENT_DATA_KEY;
	ret = btrfs_insert_empty_item(trans, root, path, &ins, sizeof(*fi));
	BUG_ON(ret);
	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);
1499 1500 1501 1502

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

Y
Yan Zheng 已提交
1503 1504 1505 1506 1507 1508 1509
	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;
1510 1511 1512
	ret = btrfs_alloc_reserved_file_extent(trans, root,
					root->root_key.objectid,
					inode->i_ino, file_pos, &ins);
Y
Yan Zheng 已提交
1513 1514
	BUG_ON(ret);
	btrfs_free_path(path);
1515

Y
Yan Zheng 已提交
1516 1517 1518
	return 0;
}

1519 1520 1521 1522 1523 1524 1525 1526 1527 1528 1529 1530 1531 1532 1533 1534 1535 1536 1537 1538 1539 1540 1541 1542
/*
 * 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.
 */
static noinline void reada_csum(struct btrfs_root *root,
				struct btrfs_path *path,
				struct btrfs_ordered_extent *ordered_extent)
{
	struct btrfs_ordered_sum *sum;
	u64 bytenr;

	sum = list_entry(ordered_extent->list.next, struct btrfs_ordered_sum,
			 list);
	bytenr = sum->sums[0].bytenr;

	/*
	 * we don't care about the results, the point of this search is
	 * just to get the btree leaves into ram
	 */
	btrfs_lookup_csum(NULL, root->fs_info->csum_root, path, bytenr, 0);
}

C
Chris Mason 已提交
1543 1544 1545 1546
/* 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.
 */
1547
static int btrfs_finish_ordered_io(struct inode *inode, u64 start, u64 end)
1548 1549 1550
{
	struct btrfs_root *root = BTRFS_I(inode)->root;
	struct btrfs_trans_handle *trans;
1551
	struct btrfs_ordered_extent *ordered_extent = NULL;
1552
	struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
1553
	struct btrfs_path *path;
Y
Yan Zheng 已提交
1554
	int compressed = 0;
1555 1556 1557
	int ret;

	ret = btrfs_dec_test_ordered_pending(inode, start, end - start + 1);
1558
	if (!ret)
1559 1560
		return 0;

1561 1562 1563 1564 1565 1566 1567 1568 1569 1570 1571 1572 1573
	/*
	 * before we join the transaction, try to do some of our IO.
	 * This will limit the amount of IO that we have to do with
	 * the transaction running.  We're unlikely to need to do any
	 * IO if the file extents are new, the disk_i_size checks
	 * covers the most common case.
	 */
	if (start < BTRFS_I(inode)->disk_i_size) {
		path = btrfs_alloc_path();
		if (path) {
			ret = btrfs_lookup_file_extent(NULL, root, path,
						       inode->i_ino,
						       start, 0);
1574 1575 1576 1577 1578 1579
			ordered_extent = btrfs_lookup_ordered_extent(inode,
								     start);
			if (!list_empty(&ordered_extent->list)) {
				btrfs_release_path(root, path);
				reada_csum(root, path, ordered_extent);
			}
1580 1581 1582 1583
			btrfs_free_path(path);
		}
	}

1584
	trans = btrfs_join_transaction(root, 1);
1585

1586 1587
	if (!ordered_extent)
		ordered_extent = btrfs_lookup_ordered_extent(inode, start);
1588
	BUG_ON(!ordered_extent);
1589 1590
	if (test_bit(BTRFS_ORDERED_NOCOW, &ordered_extent->flags))
		goto nocow;
1591 1592 1593 1594 1595

	lock_extent(io_tree, ordered_extent->file_offset,
		    ordered_extent->file_offset + ordered_extent->len - 1,
		    GFP_NOFS);

C
Chris Mason 已提交
1596
	if (test_bit(BTRFS_ORDERED_COMPRESSED, &ordered_extent->flags))
Y
Yan Zheng 已提交
1597 1598 1599 1600 1601 1602 1603 1604 1605 1606 1607 1608 1609 1610 1611
		compressed = 1;
	if (test_bit(BTRFS_ORDERED_PREALLOC, &ordered_extent->flags)) {
		BUG_ON(compressed);
		ret = btrfs_mark_extent_written(trans, root, inode,
						ordered_extent->file_offset,
						ordered_extent->file_offset +
						ordered_extent->len);
		BUG_ON(ret);
	} else {
		ret = insert_reserved_file_extent(trans, inode,
						ordered_extent->file_offset,
						ordered_extent->start,
						ordered_extent->disk_len,
						ordered_extent->len,
						ordered_extent->len,
1612 1613
						ordered_extent->file_offset +
						ordered_extent->len,
Y
Yan Zheng 已提交
1614 1615
						compressed, 0, 0,
						BTRFS_FILE_EXTENT_REG);
C
Chris Mason 已提交
1616 1617 1618
		unpin_extent_cache(&BTRFS_I(inode)->extent_tree,
				   ordered_extent->file_offset,
				   ordered_extent->len);
Y
Yan Zheng 已提交
1619 1620
		BUG_ON(ret);
	}
1621 1622 1623
	unlock_extent(io_tree, ordered_extent->file_offset,
		    ordered_extent->file_offset + ordered_extent->len - 1,
		    GFP_NOFS);
1624
nocow:
1625 1626 1627
	add_pending_csums(trans, inode, ordered_extent->file_offset,
			  &ordered_extent->list);

1628
	mutex_lock(&BTRFS_I(inode)->extent_mutex);
1629
	btrfs_ordered_update_i_size(inode, ordered_extent);
1630
	btrfs_update_inode(trans, root, inode);
1631
	btrfs_remove_ordered_extent(inode, ordered_extent);
1632
	mutex_unlock(&BTRFS_I(inode)->extent_mutex);
1633

1634 1635 1636 1637 1638 1639 1640 1641 1642
	/* once for us */
	btrfs_put_ordered_extent(ordered_extent);
	/* once for the tree */
	btrfs_put_ordered_extent(ordered_extent);

	btrfs_end_transaction(trans, root);
	return 0;
}

1643
static int btrfs_writepage_end_io_hook(struct page *page, u64 start, u64 end,
1644 1645
				struct extent_state *state, int uptodate)
{
1646
	ClearPagePrivate2(page);
1647 1648 1649
	return btrfs_finish_ordered_io(page->mapping->host, start, end);
}

C
Chris Mason 已提交
1650 1651 1652 1653 1654 1655 1656 1657
/*
 * When IO fails, either with EIO or csum verification fails, we
 * try other mirrors that might have a good copy of the data.  This
 * io_failure_record is used to record state as we go through all the
 * mirrors.  If another mirror has good data, the page is set up to date
 * and things continue.  If a good mirror can't be found, the original
 * bio end_io callback is called to indicate things have failed.
 */
1658 1659 1660 1661 1662
struct io_failure_record {
	struct page *page;
	u64 start;
	u64 len;
	u64 logical;
1663
	unsigned long bio_flags;
1664 1665 1666
	int last_mirror;
};

1667
static int btrfs_io_failed_hook(struct bio *failed_bio,
1668 1669
			 struct page *page, u64 start, u64 end,
			 struct extent_state *state)
1670 1671 1672 1673 1674 1675
{
	struct io_failure_record *failrec = NULL;
	u64 private;
	struct extent_map *em;
	struct inode *inode = page->mapping->host;
	struct extent_io_tree *failure_tree = &BTRFS_I(inode)->io_failure_tree;
1676
	struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
1677 1678 1679
	struct bio *bio;
	int num_copies;
	int ret;
1680
	int rw;
1681 1682 1683 1684 1685 1686 1687 1688 1689 1690
	u64 logical;

	ret = get_state_private(failure_tree, start, &private);
	if (ret) {
		failrec = kmalloc(sizeof(*failrec), GFP_NOFS);
		if (!failrec)
			return -ENOMEM;
		failrec->start = start;
		failrec->len = end - start + 1;
		failrec->last_mirror = 0;
1691
		failrec->bio_flags = 0;
1692

1693
		read_lock(&em_tree->lock);
1694 1695 1696 1697 1698
		em = lookup_extent_mapping(em_tree, start, failrec->len);
		if (em->start > start || em->start + em->len < start) {
			free_extent_map(em);
			em = NULL;
		}
1699
		read_unlock(&em_tree->lock);
1700 1701 1702 1703 1704 1705 1706

		if (!em || IS_ERR(em)) {
			kfree(failrec);
			return -EIO;
		}
		logical = start - em->start;
		logical = em->block_start + logical;
1707 1708 1709 1710
		if (test_bit(EXTENT_FLAG_COMPRESSED, &em->flags)) {
			logical = em->block_start;
			failrec->bio_flags = EXTENT_BIO_COMPRESSED;
		}
1711 1712 1713 1714
		failrec->logical = logical;
		free_extent_map(em);
		set_extent_bits(failure_tree, start, end, EXTENT_LOCKED |
				EXTENT_DIRTY, GFP_NOFS);
1715 1716
		set_state_private(failure_tree, start,
				 (u64)(unsigned long)failrec);
1717
	} else {
1718
		failrec = (struct io_failure_record *)(unsigned long)private;
1719 1720 1721 1722 1723 1724
	}
	num_copies = btrfs_num_copies(
			      &BTRFS_I(inode)->root->fs_info->mapping_tree,
			      failrec->logical, failrec->len);
	failrec->last_mirror++;
	if (!state) {
1725
		spin_lock(&BTRFS_I(inode)->io_tree.lock);
1726 1727 1728 1729 1730
		state = find_first_extent_bit_state(&BTRFS_I(inode)->io_tree,
						    failrec->start,
						    EXTENT_LOCKED);
		if (state && state->start != failrec->start)
			state = NULL;
1731
		spin_unlock(&BTRFS_I(inode)->io_tree.lock);
1732 1733 1734 1735 1736 1737 1738 1739 1740 1741 1742 1743 1744 1745
	}
	if (!state || failrec->last_mirror > num_copies) {
		set_state_private(failure_tree, failrec->start, 0);
		clear_extent_bits(failure_tree, failrec->start,
				  failrec->start + failrec->len - 1,
				  EXTENT_LOCKED | EXTENT_DIRTY, GFP_NOFS);
		kfree(failrec);
		return -EIO;
	}
	bio = bio_alloc(GFP_NOFS, 1);
	bio->bi_private = state;
	bio->bi_end_io = failed_bio->bi_end_io;
	bio->bi_sector = failrec->logical >> 9;
	bio->bi_bdev = failed_bio->bi_bdev;
1746
	bio->bi_size = 0;
1747

1748
	bio_add_page(bio, page, failrec->len, start - page_offset(page));
1749 1750 1751 1752 1753 1754
	if (failed_bio->bi_rw & (1 << BIO_RW))
		rw = WRITE;
	else
		rw = READ;

	BTRFS_I(inode)->io_tree.ops->submit_bio_hook(inode, rw, bio,
C
Chris Mason 已提交
1755
						      failrec->last_mirror,
1756
						      failrec->bio_flags);
1757 1758 1759
	return 0;
}

C
Chris Mason 已提交
1760 1761 1762 1763
/*
 * each time an IO finishes, we do a fast check in the IO failure tree
 * to see if we need to process or clean up an io_failure_record
 */
1764
static int btrfs_clean_io_failures(struct inode *inode, u64 start)
1765 1766 1767 1768 1769 1770 1771 1772 1773 1774 1775 1776 1777 1778 1779 1780 1781 1782 1783 1784 1785 1786 1787 1788
{
	u64 private;
	u64 private_failure;
	struct io_failure_record *failure;
	int ret;

	private = 0;
	if (count_range_bits(&BTRFS_I(inode)->io_failure_tree, &private,
			     (u64)-1, 1, EXTENT_DIRTY)) {
		ret = get_state_private(&BTRFS_I(inode)->io_failure_tree,
					start, &private_failure);
		if (ret == 0) {
			failure = (struct io_failure_record *)(unsigned long)
				   private_failure;
			set_state_private(&BTRFS_I(inode)->io_failure_tree,
					  failure->start, 0);
			clear_extent_bits(&BTRFS_I(inode)->io_failure_tree,
					  failure->start,
					  failure->start + failure->len - 1,
					  EXTENT_DIRTY | EXTENT_LOCKED,
					  GFP_NOFS);
			kfree(failure);
		}
	}
1789 1790 1791
	return 0;
}

C
Chris Mason 已提交
1792 1793 1794 1795 1796
/*
 * when reads are done, we need to check csums to verify the data is correct
 * if there's a match, we allow the bio to finish.  If not, we go through
 * the io_failure_record routines to find good copies
 */
1797
static int btrfs_readpage_end_io_hook(struct page *page, u64 start, u64 end,
1798
			       struct extent_state *state)
1799
{
1800
	size_t offset = start - ((u64)page->index << PAGE_CACHE_SHIFT);
1801
	struct inode *inode = page->mapping->host;
1802
	struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
1803
	char *kaddr;
1804
	u64 private = ~(u32)0;
1805
	int ret;
1806 1807
	struct btrfs_root *root = BTRFS_I(inode)->root;
	u32 csum = ~(u32)0;
1808

1809 1810 1811 1812
	if (PageChecked(page)) {
		ClearPageChecked(page);
		goto good;
	}
1813 1814

	if (BTRFS_I(inode)->flags & BTRFS_INODE_NODATASUM)
1815 1816 1817
		return 0;

	if (root->root_key.objectid == BTRFS_DATA_RELOC_TREE_OBJECTID &&
1818
	    test_range_bit(io_tree, start, end, EXTENT_NODATASUM, 1, NULL)) {
1819 1820
		clear_extent_bits(io_tree, start, end, EXTENT_NODATASUM,
				  GFP_NOFS);
1821
		return 0;
1822
	}
1823

Y
Yan 已提交
1824
	if (state && state->start == start) {
1825 1826 1827 1828 1829
		private = state->private;
		ret = 0;
	} else {
		ret = get_state_private(io_tree, start, &private);
	}
1830
	kaddr = kmap_atomic(page, KM_USER0);
C
Chris Mason 已提交
1831
	if (ret)
1832
		goto zeroit;
C
Chris Mason 已提交
1833

1834 1835
	csum = btrfs_csum_data(root, kaddr + offset, csum,  end - start + 1);
	btrfs_csum_final(csum, (char *)&csum);
C
Chris Mason 已提交
1836
	if (csum != private)
1837
		goto zeroit;
C
Chris Mason 已提交
1838

1839
	kunmap_atomic(kaddr, KM_USER0);
1840
good:
1841 1842 1843
	/* if the io failure tree for this inode is non-empty,
	 * check to see if we've recovered from a failed IO
	 */
1844
	btrfs_clean_io_failures(inode, start);
1845 1846 1847
	return 0;

zeroit:
C
Chris Mason 已提交
1848 1849 1850 1851 1852 1853
	if (printk_ratelimit()) {
		printk(KERN_INFO "btrfs csum failed ino %lu off %llu csum %u "
		       "private %llu\n", page->mapping->host->i_ino,
		       (unsigned long long)start, csum,
		       (unsigned long long)private);
	}
1854 1855
	memset(kaddr + offset, 1, end - start + 1);
	flush_dcache_page(page);
1856
	kunmap_atomic(kaddr, KM_USER0);
1857 1858
	if (private == 0)
		return 0;
1859
	return -EIO;
1860
}
1861

1862 1863 1864 1865 1866 1867 1868 1869 1870
/*
 * This creates an orphan entry for the given inode in case something goes
 * wrong in the middle of an unlink/truncate.
 */
int btrfs_orphan_add(struct btrfs_trans_handle *trans, struct inode *inode)
{
	struct btrfs_root *root = BTRFS_I(inode)->root;
	int ret = 0;

1871
	spin_lock(&root->list_lock);
1872 1873 1874

	/* already on the orphan list, we're good */
	if (!list_empty(&BTRFS_I(inode)->i_orphan)) {
1875
		spin_unlock(&root->list_lock);
1876 1877 1878 1879 1880
		return 0;
	}

	list_add(&BTRFS_I(inode)->i_orphan, &root->orphan_list);

1881
	spin_unlock(&root->list_lock);
1882 1883 1884 1885 1886 1887 1888 1889 1890 1891 1892 1893 1894 1895 1896 1897 1898 1899

	/*
	 * insert an orphan item to track this unlinked/truncated file
	 */
	ret = btrfs_insert_orphan_item(trans, root, inode->i_ino);

	return ret;
}

/*
 * 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;
	int ret = 0;

1900
	spin_lock(&root->list_lock);
1901 1902

	if (list_empty(&BTRFS_I(inode)->i_orphan)) {
1903
		spin_unlock(&root->list_lock);
1904 1905 1906 1907 1908
		return 0;
	}

	list_del_init(&BTRFS_I(inode)->i_orphan);
	if (!trans) {
1909
		spin_unlock(&root->list_lock);
1910 1911 1912
		return 0;
	}

1913
	spin_unlock(&root->list_lock);
1914 1915 1916 1917 1918 1919 1920 1921 1922 1923 1924 1925 1926 1927 1928 1929 1930 1931 1932 1933 1934 1935 1936 1937 1938 1939 1940 1941 1942 1943 1944 1945 1946 1947 1948 1949 1950 1951 1952 1953 1954 1955 1956 1957 1958 1959 1960 1961 1962 1963 1964 1965 1966 1967 1968 1969 1970 1971 1972 1973 1974 1975 1976 1977 1978 1979 1980 1981

	ret = btrfs_del_orphan_item(trans, root, inode->i_ino);

	return ret;
}

/*
 * this cleans up any orphans that may be left on the list from the last use
 * of this root.
 */
void btrfs_orphan_cleanup(struct btrfs_root *root)
{
	struct btrfs_path *path;
	struct extent_buffer *leaf;
	struct btrfs_item *item;
	struct btrfs_key key, found_key;
	struct btrfs_trans_handle *trans;
	struct inode *inode;
	int ret = 0, nr_unlink = 0, nr_truncate = 0;

	path = btrfs_alloc_path();
	if (!path)
		return;
	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);
		if (ret < 0) {
			printk(KERN_ERR "Error searching slot for orphan: %d"
			       "\n", ret);
			break;
		}

		/*
		 * if ret == 0 means we found what we were searching for, which
		 * is weird, but possible, so only screw with path if we didnt
		 * find the key and see if we have stuff that matches
		 */
		if (ret > 0) {
			if (path->slots[0] == 0)
				break;
			path->slots[0]--;
		}

		/* pull out the item */
		leaf = path->nodes[0];
		item = btrfs_item_nr(leaf, path->slots[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 */
		btrfs_release_path(root, path);

		/*
		 * 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.
		 */
1982 1983 1984 1985 1986
		found_key.objectid = found_key.offset;
		found_key.type = BTRFS_INODE_ITEM_KEY;
		found_key.offset = 0;
		inode = btrfs_iget(root->fs_info->sb, &found_key, root);
		if (IS_ERR(inode))
1987 1988 1989 1990 1991 1992
			break;

		/*
		 * add this inode to the orphan list so btrfs_orphan_del does
		 * the proper thing when we hit it
		 */
1993
		spin_lock(&root->list_lock);
1994
		list_add(&BTRFS_I(inode)->i_orphan, &root->orphan_list);
1995
		spin_unlock(&root->list_lock);
1996 1997 1998 1999 2000 2001 2002 2003

		/*
		 * if this is a bad inode, means we actually succeeded in
		 * removing the inode, but not the orphan record, which means
		 * we need to manually delete the orphan since iput will just
		 * do a destroy_inode
		 */
		if (is_bad_inode(inode)) {
2004
			trans = btrfs_start_transaction(root, 1);
2005
			btrfs_orphan_del(trans, inode);
2006
			btrfs_end_transaction(trans, root);
2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 2020 2021 2022 2023 2024 2025 2026 2027 2028 2029 2030
			iput(inode);
			continue;
		}

		/* if we have links, this was a truncate, lets do that */
		if (inode->i_nlink) {
			nr_truncate++;
			btrfs_truncate(inode);
		} else {
			nr_unlink++;
		}

		/* this will do delete_inode and everything for us */
		iput(inode);
	}

	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);

	btrfs_free_path(path);
}

2031 2032 2033 2034 2035 2036 2037 2038 2039 2040 2041 2042 2043 2044 2045 2046 2047 2048 2049 2050 2051 2052 2053 2054 2055 2056 2057 2058 2059 2060 2061 2062 2063 2064 2065 2066 2067 2068 2069 2070 2071 2072 2073 2074 2075 2076 2077 2078 2079 2080 2081
/*
 * 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
2082 2083 2084
/*
 * read an inode from the btree into the in-memory inode
 */
2085
static void btrfs_read_locked_inode(struct inode *inode)
C
Chris Mason 已提交
2086 2087
{
	struct btrfs_path *path;
2088
	struct extent_buffer *leaf;
C
Chris Mason 已提交
2089
	struct btrfs_inode_item *inode_item;
2090
	struct btrfs_timespec *tspec;
C
Chris Mason 已提交
2091 2092
	struct btrfs_root *root = BTRFS_I(inode)->root;
	struct btrfs_key location;
2093
	int maybe_acls;
C
Chris Mason 已提交
2094
	u64 alloc_group_block;
J
Josef Bacik 已提交
2095
	u32 rdev;
C
Chris Mason 已提交
2096 2097 2098 2099 2100
	int ret;

	path = btrfs_alloc_path();
	BUG_ON(!path);
	memcpy(&location, &BTRFS_I(inode)->location, sizeof(location));
C
Chris Mason 已提交
2101

C
Chris Mason 已提交
2102
	ret = btrfs_lookup_inode(NULL, root, path, &location, 0);
2103
	if (ret)
C
Chris Mason 已提交
2104 2105
		goto make_bad;

2106 2107 2108 2109 2110 2111 2112 2113
	leaf = path->nodes[0];
	inode_item = btrfs_item_ptr(leaf, path->slots[0],
				    struct btrfs_inode_item);

	inode->i_mode = btrfs_inode_mode(leaf, inode_item);
	inode->i_nlink = btrfs_inode_nlink(leaf, inode_item);
	inode->i_uid = btrfs_inode_uid(leaf, inode_item);
	inode->i_gid = btrfs_inode_gid(leaf, inode_item);
2114
	btrfs_i_size_write(inode, btrfs_inode_size(leaf, inode_item));
2115 2116 2117 2118 2119 2120 2121 2122 2123 2124 2125 2126 2127

	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);

2128
	inode_set_bytes(inode, btrfs_inode_nbytes(leaf, inode_item));
2129
	BTRFS_I(inode)->generation = btrfs_inode_generation(leaf, inode_item);
2130
	BTRFS_I(inode)->sequence = btrfs_inode_sequence(leaf, inode_item);
2131
	inode->i_generation = BTRFS_I(inode)->generation;
J
Josef Bacik 已提交
2132
	inode->i_rdev = 0;
2133 2134
	rdev = btrfs_inode_rdev(leaf, inode_item);

2135
	BTRFS_I(inode)->index_cnt = (u64)-1;
2136
	BTRFS_I(inode)->flags = btrfs_inode_flags(leaf, inode_item);
2137

2138
	alloc_group_block = btrfs_inode_block_group(leaf, inode_item);
2139

2140 2141 2142 2143 2144
	/*
	 * try to precache a NULL acl entry for files that don't have
	 * any xattrs or acls
	 */
	maybe_acls = acls_after_inode_item(leaf, path->slots[0], inode->i_ino);
2145 2146
	if (!maybe_acls)
		cache_no_acl(inode);
2147

2148 2149
	BTRFS_I(inode)->block_group = btrfs_find_block_group(root, 0,
						alloc_group_block, 0);
C
Chris Mason 已提交
2150 2151 2152 2153 2154 2155
	btrfs_free_path(path);
	inode_item = NULL;

	switch (inode->i_mode & S_IFMT) {
	case S_IFREG:
		inode->i_mapping->a_ops = &btrfs_aops;
C
Chris Mason 已提交
2156
		inode->i_mapping->backing_dev_info = &root->fs_info->bdi;
2157
		BTRFS_I(inode)->io_tree.ops = &btrfs_extent_io_ops;
C
Chris Mason 已提交
2158 2159 2160 2161 2162 2163 2164 2165 2166 2167 2168 2169 2170
		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
2171
		inode->i_mapping->backing_dev_info = &root->fs_info->bdi;
C
Chris Mason 已提交
2172
		break;
J
Josef Bacik 已提交
2173
	default:
J
Jim Owens 已提交
2174
		inode->i_op = &btrfs_special_inode_operations;
J
Josef Bacik 已提交
2175 2176
		init_special_inode(inode, inode->i_mode, rdev);
		break;
C
Chris Mason 已提交
2177
	}
2178 2179

	btrfs_update_iflags(inode);
C
Chris Mason 已提交
2180 2181 2182 2183 2184 2185 2186
	return;

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

C
Chris Mason 已提交
2187 2188 2189
/*
 * given a leaf and an inode, copy the inode fields into the leaf
 */
2190 2191
static void fill_inode_item(struct btrfs_trans_handle *trans,
			    struct extent_buffer *leaf,
2192
			    struct btrfs_inode_item *item,
C
Chris Mason 已提交
2193 2194
			    struct inode *inode)
{
2195 2196
	btrfs_set_inode_uid(leaf, item, inode->i_uid);
	btrfs_set_inode_gid(leaf, item, inode->i_gid);
2197
	btrfs_set_inode_size(leaf, item, BTRFS_I(inode)->disk_i_size);
2198 2199 2200 2201 2202 2203 2204 2205 2206 2207 2208 2209 2210 2211 2212 2213 2214 2215
	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);

2216
	btrfs_set_inode_nbytes(leaf, item, inode_get_bytes(inode));
2217
	btrfs_set_inode_generation(leaf, item, BTRFS_I(inode)->generation);
2218
	btrfs_set_inode_sequence(leaf, item, BTRFS_I(inode)->sequence);
2219
	btrfs_set_inode_transid(leaf, item, trans->transid);
2220
	btrfs_set_inode_rdev(leaf, item, inode->i_rdev);
Y
Yan 已提交
2221
	btrfs_set_inode_flags(leaf, item, BTRFS_I(inode)->flags);
2222
	btrfs_set_inode_block_group(leaf, item, BTRFS_I(inode)->block_group);
C
Chris Mason 已提交
2223 2224
}

C
Chris Mason 已提交
2225 2226 2227
/*
 * copy everything in the in-memory inode into the btree.
 */
C
Chris Mason 已提交
2228 2229
noinline int btrfs_update_inode(struct btrfs_trans_handle *trans,
				struct btrfs_root *root, struct inode *inode)
C
Chris Mason 已提交
2230 2231 2232
{
	struct btrfs_inode_item *inode_item;
	struct btrfs_path *path;
2233
	struct extent_buffer *leaf;
C
Chris Mason 已提交
2234 2235 2236 2237
	int ret;

	path = btrfs_alloc_path();
	BUG_ON(!path);
2238
	path->leave_spinning = 1;
C
Chris Mason 已提交
2239 2240 2241 2242 2243 2244 2245 2246
	ret = btrfs_lookup_inode(trans, root, path,
				 &BTRFS_I(inode)->location, 1);
	if (ret) {
		if (ret > 0)
			ret = -ENOENT;
		goto failed;
	}

2247
	btrfs_unlock_up_safe(path, 1);
2248 2249
	leaf = path->nodes[0];
	inode_item = btrfs_item_ptr(leaf, path->slots[0],
C
Chris Mason 已提交
2250 2251
				  struct btrfs_inode_item);

2252
	fill_inode_item(trans, leaf, inode_item, inode);
2253
	btrfs_mark_buffer_dirty(leaf);
2254
	btrfs_set_inode_last_trans(trans, inode);
C
Chris Mason 已提交
2255 2256 2257 2258 2259 2260 2261
	ret = 0;
failed:
	btrfs_free_path(path);
	return ret;
}


C
Chris Mason 已提交
2262 2263 2264 2265 2266
/*
 * 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
 */
2267 2268 2269 2270
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
2271 2272 2273
{
	struct btrfs_path *path;
	int ret = 0;
2274
	struct extent_buffer *leaf;
C
Chris Mason 已提交
2275
	struct btrfs_dir_item *di;
2276
	struct btrfs_key key;
2277
	u64 index;
C
Chris Mason 已提交
2278 2279

	path = btrfs_alloc_path();
2280 2281 2282 2283 2284
	if (!path) {
		ret = -ENOMEM;
		goto err;
	}

2285
	path->leave_spinning = 1;
C
Chris Mason 已提交
2286 2287 2288 2289 2290 2291 2292 2293 2294 2295
	di = btrfs_lookup_dir_item(trans, root, path, dir->i_ino,
				    name, name_len, -1);
	if (IS_ERR(di)) {
		ret = PTR_ERR(di);
		goto err;
	}
	if (!di) {
		ret = -ENOENT;
		goto err;
	}
2296 2297
	leaf = path->nodes[0];
	btrfs_dir_item_key_to_cpu(leaf, di, &key);
C
Chris Mason 已提交
2298
	ret = btrfs_delete_one_dir_name(trans, root, path, di);
2299 2300
	if (ret)
		goto err;
C
Chris Mason 已提交
2301 2302
	btrfs_release_path(root, path);

2303
	ret = btrfs_del_inode_ref(trans, root, name, name_len,
2304 2305
				  inode->i_ino,
				  dir->i_ino, &index);
2306
	if (ret) {
C
Chris Mason 已提交
2307
		printk(KERN_INFO "btrfs failed to delete reference to %.*s, "
2308
		       "inode %lu parent %lu\n", name_len, name,
2309
		       inode->i_ino, dir->i_ino);
2310 2311 2312
		goto err;
	}

C
Chris Mason 已提交
2313
	di = btrfs_lookup_dir_index_item(trans, root, path, dir->i_ino,
2314
					 index, name, name_len, -1);
C
Chris Mason 已提交
2315 2316 2317 2318 2319 2320 2321 2322 2323
	if (IS_ERR(di)) {
		ret = PTR_ERR(di);
		goto err;
	}
	if (!di) {
		ret = -ENOENT;
		goto err;
	}
	ret = btrfs_delete_one_dir_name(trans, root, path, di);
2324
	btrfs_release_path(root, path);
C
Chris Mason 已提交
2325

2326 2327
	ret = btrfs_del_inode_ref_in_log(trans, root, name, name_len,
					 inode, dir->i_ino);
C
Chris Mason 已提交
2328
	BUG_ON(ret != 0 && ret != -ENOENT);
2329 2330 2331 2332

	ret = btrfs_del_dir_entries_in_log(trans, root, name, name_len,
					   dir, index);
	BUG_ON(ret);
C
Chris Mason 已提交
2333 2334
err:
	btrfs_free_path(path);
2335 2336 2337 2338 2339 2340 2341 2342 2343
	if (ret)
		goto out;

	btrfs_i_size_write(dir, dir->i_size - name_len * 2);
	inode->i_ctime = dir->i_mtime = dir->i_ctime = CURRENT_TIME;
	btrfs_update_inode(trans, root, dir);
	btrfs_drop_nlink(inode);
	ret = btrfs_update_inode(trans, root, inode);
out:
C
Chris Mason 已提交
2344 2345 2346 2347 2348 2349 2350
	return ret;
}

static int btrfs_unlink(struct inode *dir, struct dentry *dentry)
{
	struct btrfs_root *root;
	struct btrfs_trans_handle *trans;
2351
	struct inode *inode = dentry->d_inode;
C
Chris Mason 已提交
2352
	int ret;
2353
	unsigned long nr = 0;
C
Chris Mason 已提交
2354 2355

	root = BTRFS_I(dir)->root;
2356

C
Chris Mason 已提交
2357
	trans = btrfs_start_transaction(root, 1);
2358

C
Chris Mason 已提交
2359
	btrfs_set_trans_block_group(trans, dir);
2360 2361 2362

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

2363 2364
	ret = btrfs_unlink_inode(trans, root, dir, dentry->d_inode,
				 dentry->d_name.name, dentry->d_name.len);
2365 2366 2367 2368

	if (inode->i_nlink == 0)
		ret = btrfs_orphan_add(trans, inode);

2369
	nr = trans->blocks_used;
2370

2371
	btrfs_end_transaction_throttle(trans, root);
2372
	btrfs_btree_balance_dirty(root, nr);
C
Chris Mason 已提交
2373 2374 2375
	return ret;
}

2376 2377 2378 2379 2380 2381 2382 2383 2384 2385 2386 2387 2388 2389 2390 2391 2392 2393 2394 2395 2396 2397 2398 2399 2400 2401 2402 2403 2404 2405 2406 2407 2408 2409 2410 2411 2412 2413 2414 2415 2416 2417 2418 2419 2420 2421 2422 2423 2424 2425 2426 2427 2428 2429 2430 2431 2432 2433 2434 2435 2436 2437 2438
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;

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

	di = btrfs_lookup_dir_item(trans, root, path, dir->i_ino,
				   name, name_len, -1);
	BUG_ON(!di || IS_ERR(di));

	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);
	BUG_ON(ret);
	btrfs_release_path(root, path);

	ret = btrfs_del_root_ref(trans, root->fs_info->tree_root,
				 objectid, root->root_key.objectid,
				 dir->i_ino, &index, name, name_len);
	if (ret < 0) {
		BUG_ON(ret != -ENOENT);
		di = btrfs_search_dir_index_item(root, path, dir->i_ino,
						 name, name_len);
		BUG_ON(!di || IS_ERR(di));

		leaf = path->nodes[0];
		btrfs_item_key_to_cpu(leaf, &key, path->slots[0]);
		btrfs_release_path(root, path);
		index = key.offset;
	}

	di = btrfs_lookup_dir_index_item(trans, root, path, dir->i_ino,
					 index, name, name_len, -1);
	BUG_ON(!di || IS_ERR(di));

	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);
	BUG_ON(ret);
	btrfs_release_path(root, path);

	btrfs_i_size_write(dir, dir->i_size - name_len * 2);
	dir->i_mtime = dir->i_ctime = CURRENT_TIME;
	ret = btrfs_update_inode(trans, root, dir);
	BUG_ON(ret);
	dir->i_sb->s_dirt = 1;

	btrfs_free_path(path);
	return 0;
}

C
Chris Mason 已提交
2439 2440 2441
static int btrfs_rmdir(struct inode *dir, struct dentry *dentry)
{
	struct inode *inode = dentry->d_inode;
2442
	int err = 0;
C
Chris Mason 已提交
2443 2444 2445
	int ret;
	struct btrfs_root *root = BTRFS_I(dir)->root;
	struct btrfs_trans_handle *trans;
2446
	unsigned long nr = 0;
C
Chris Mason 已提交
2447

2448
	if (inode->i_size > BTRFS_EMPTY_DIR_SIZE ||
2449
	    inode->i_ino == BTRFS_FIRST_FREE_OBJECTID)
Y
Yan 已提交
2450 2451
		return -ENOTEMPTY;

C
Chris Mason 已提交
2452 2453 2454
	trans = btrfs_start_transaction(root, 1);
	btrfs_set_trans_block_group(trans, dir);

2455 2456 2457 2458 2459 2460 2461 2462
	if (unlikely(inode->i_ino == BTRFS_EMPTY_SUBVOL_DIR_OBJECTID)) {
		err = btrfs_unlink_subvol(trans, root, dir,
					  BTRFS_I(inode)->location.objectid,
					  dentry->d_name.name,
					  dentry->d_name.len);
		goto out;
	}

2463 2464
	err = btrfs_orphan_add(trans, inode);
	if (err)
2465
		goto out;
2466

C
Chris Mason 已提交
2467
	/* now the directory is empty */
2468 2469
	err = btrfs_unlink_inode(trans, root, dir, dentry->d_inode,
				 dentry->d_name.name, dentry->d_name.len);
C
Chris Mason 已提交
2470
	if (!err)
2471
		btrfs_i_size_write(inode, 0);
2472
out:
2473
	nr = trans->blocks_used;
2474
	ret = btrfs_end_transaction_throttle(trans, root);
2475
	btrfs_btree_balance_dirty(root, nr);
2476

C
Chris Mason 已提交
2477 2478 2479 2480 2481
	if (ret && !err)
		err = ret;
	return err;
}

2482
#if 0
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
/*
 * when truncating bytes in a file, it is possible to avoid reading
 * the leaves that contain only checksum items.  This can be the
 * majority of the IO required to delete a large file, but it must
 * be done carefully.
 *
 * The keys in the level just above the leaves are checked to make sure
 * the lowest key in a given leaf is a csum key, and starts at an offset
 * after the new  size.
 *
 * Then the key for the next leaf is checked to make sure it also has
 * a checksum item for the same file.  If it does, we know our target leaf
 * contains only checksum items, and it can be safely freed without reading
 * it.
 *
 * This is just an optimization targeted at large files.  It may do
 * nothing.  It will return 0 unless things went badly.
 */
static noinline int drop_csum_leaves(struct btrfs_trans_handle *trans,
				     struct btrfs_root *root,
				     struct btrfs_path *path,
				     struct inode *inode, u64 new_size)
{
	struct btrfs_key key;
	int ret;
	int nritems;
	struct btrfs_key found_key;
	struct btrfs_key other_key;
Y
Yan Zheng 已提交
2511 2512 2513
	struct btrfs_leaf_ref *ref;
	u64 leaf_gen;
	u64 leaf_start;
2514 2515 2516 2517 2518 2519 2520 2521 2522 2523 2524 2525 2526 2527 2528 2529 2530 2531 2532 2533 2534 2535 2536 2537 2538 2539 2540 2541 2542 2543 2544 2545 2546 2547 2548 2549 2550 2551 2552 2553 2554 2555 2556 2557 2558 2559 2560 2561 2562 2563 2564 2565 2566 2567 2568 2569 2570 2571 2572 2573 2574 2575 2576 2577 2578 2579 2580 2581 2582 2583 2584 2585 2586 2587 2588 2589 2590 2591 2592 2593 2594 2595 2596 2597 2598 2599 2600 2601 2602 2603 2604 2605 2606

	path->lowest_level = 1;
	key.objectid = inode->i_ino;
	key.type = BTRFS_CSUM_ITEM_KEY;
	key.offset = new_size;
again:
	ret = btrfs_search_slot(trans, root, &key, path, -1, 1);
	if (ret < 0)
		goto out;

	if (path->nodes[1] == NULL) {
		ret = 0;
		goto out;
	}
	ret = 0;
	btrfs_node_key_to_cpu(path->nodes[1], &found_key, path->slots[1]);
	nritems = btrfs_header_nritems(path->nodes[1]);

	if (!nritems)
		goto out;

	if (path->slots[1] >= nritems)
		goto next_node;

	/* did we find a key greater than anything we want to delete? */
	if (found_key.objectid > inode->i_ino ||
	   (found_key.objectid == inode->i_ino && found_key.type > key.type))
		goto out;

	/* we check the next key in the node to make sure the leave contains
	 * only checksum items.  This comparison doesn't work if our
	 * leaf is the last one in the node
	 */
	if (path->slots[1] + 1 >= nritems) {
next_node:
		/* search forward from the last key in the node, this
		 * will bring us into the next node in the tree
		 */
		btrfs_node_key_to_cpu(path->nodes[1], &found_key, nritems - 1);

		/* unlikely, but we inc below, so check to be safe */
		if (found_key.offset == (u64)-1)
			goto out;

		/* search_forward needs a path with locks held, do the
		 * search again for the original key.  It is possible
		 * this will race with a balance and return a path that
		 * we could modify, but this drop is just an optimization
		 * and is allowed to miss some leaves.
		 */
		btrfs_release_path(root, path);
		found_key.offset++;

		/* setup a max key for search_forward */
		other_key.offset = (u64)-1;
		other_key.type = key.type;
		other_key.objectid = key.objectid;

		path->keep_locks = 1;
		ret = btrfs_search_forward(root, &found_key, &other_key,
					   path, 0, 0);
		path->keep_locks = 0;
		if (ret || found_key.objectid != key.objectid ||
		    found_key.type != key.type) {
			ret = 0;
			goto out;
		}

		key.offset = found_key.offset;
		btrfs_release_path(root, path);
		cond_resched();
		goto again;
	}

	/* we know there's one more slot after us in the tree,
	 * read that key so we can verify it is also a checksum item
	 */
	btrfs_node_key_to_cpu(path->nodes[1], &other_key, path->slots[1] + 1);

	if (found_key.objectid < inode->i_ino)
		goto next_key;

	if (found_key.type != key.type || found_key.offset < new_size)
		goto next_key;

	/*
	 * if the key for the next leaf isn't a csum key from this objectid,
	 * we can't be sure there aren't good items inside this leaf.
	 * Bail out
	 */
	if (other_key.objectid != inode->i_ino || other_key.type != key.type)
		goto out;

Y
Yan Zheng 已提交
2607 2608
	leaf_start = btrfs_node_blockptr(path->nodes[1], path->slots[1]);
	leaf_gen = btrfs_node_ptr_generation(path->nodes[1], path->slots[1]);
2609 2610 2611 2612
	/*
	 * it is safe to delete this leaf, it contains only
	 * csum items from this inode at an offset >= new_size
	 */
Y
Yan Zheng 已提交
2613
	ret = btrfs_del_leaf(trans, root, path, leaf_start);
2614 2615
	BUG_ON(ret);

Y
Yan Zheng 已提交
2616 2617 2618 2619 2620 2621 2622 2623 2624
	if (root->ref_cows && leaf_gen < trans->transid) {
		ref = btrfs_alloc_leaf_ref(root, 0);
		if (ref) {
			ref->root_gen = root->root_key.offset;
			ref->bytenr = leaf_start;
			ref->owner = 0;
			ref->generation = leaf_gen;
			ref->nritems = 0;

2625 2626
			btrfs_sort_leaf_ref(ref);

Y
Yan Zheng 已提交
2627 2628 2629 2630 2631 2632 2633
			ret = btrfs_add_leaf_ref(root, ref, 0);
			WARN_ON(ret);
			btrfs_free_leaf_ref(root, ref);
		} else {
			WARN_ON(1);
		}
	}
2634 2635 2636 2637 2638 2639 2640 2641 2642 2643 2644 2645 2646 2647 2648 2649 2650 2651
next_key:
	btrfs_release_path(root, path);

	if (other_key.objectid == inode->i_ino &&
	    other_key.type == key.type && other_key.offset > key.offset) {
		key.offset = other_key.offset;
		cond_resched();
		goto again;
	}
	ret = 0;
out:
	/* fixup any changes we've made to the path */
	path->lowest_level = 0;
	path->keep_locks = 0;
	btrfs_release_path(root, path);
	return ret;
}

2652 2653
#endif

C
Chris Mason 已提交
2654 2655 2656
/*
 * 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
2657
 * any higher than new_size
C
Chris Mason 已提交
2658 2659 2660
 *
 * csum items that cross the new i_size are truncated to the new size
 * as well.
2661 2662 2663
 *
 * 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
2664
 */
2665 2666 2667 2668
noinline 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
2669 2670 2671 2672
{
	int ret;
	struct btrfs_path *path;
	struct btrfs_key key;
2673
	struct btrfs_key found_key;
2674
	u32 found_type = (u8)-1;
2675
	struct extent_buffer *leaf;
C
Chris Mason 已提交
2676 2677
	struct btrfs_file_extent_item *fi;
	u64 extent_start = 0;
2678
	u64 extent_num_bytes = 0;
2679
	u64 extent_offset = 0;
C
Chris Mason 已提交
2680 2681 2682
	u64 item_end = 0;
	int found_extent;
	int del_item;
2683 2684
	int pending_del_nr = 0;
	int pending_del_slot = 0;
2685
	int extent_type = -1;
2686
	int encoding;
2687
	u64 mask = root->sectorsize - 1;
C
Chris Mason 已提交
2688

2689
	if (root->ref_cows)
2690
		btrfs_drop_extent_cache(inode, new_size & (~mask), (u64)-1, 0);
C
Chris Mason 已提交
2691 2692
	path = btrfs_alloc_path();
	BUG_ON(!path);
J
Julia Lawall 已提交
2693
	path->reada = -1;
2694

C
Chris Mason 已提交
2695 2696 2697
	/* FIXME, add redo link to tree so we don't leak on crash */
	key.objectid = inode->i_ino;
	key.offset = (u64)-1;
2698 2699
	key.type = (u8)-1;

2700
search_again:
2701
	path->leave_spinning = 1;
2702
	ret = btrfs_search_slot(trans, root, &key, path, -1, 1);
C
Chris Mason 已提交
2703
	if (ret < 0)
2704
		goto error;
C
Chris Mason 已提交
2705

2706
	if (ret > 0) {
2707 2708 2709 2710 2711 2712 2713
		/* there are no items in the tree for us to truncate, we're
		 * done
		 */
		if (path->slots[0] == 0) {
			ret = 0;
			goto error;
		}
2714 2715 2716
		path->slots[0]--;
	}

C
Chris Mason 已提交
2717
	while (1) {
C
Chris Mason 已提交
2718
		fi = NULL;
2719 2720 2721
		leaf = path->nodes[0];
		btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
		found_type = btrfs_key_type(&found_key);
2722
		encoding = 0;
C
Chris Mason 已提交
2723

2724
		if (found_key.objectid != inode->i_ino)
C
Chris Mason 已提交
2725
			break;
2726

2727
		if (found_type < min_type)
C
Chris Mason 已提交
2728 2729
			break;

2730
		item_end = found_key.offset;
C
Chris Mason 已提交
2731
		if (found_type == BTRFS_EXTENT_DATA_KEY) {
2732
			fi = btrfs_item_ptr(leaf, path->slots[0],
C
Chris Mason 已提交
2733
					    struct btrfs_file_extent_item);
2734
			extent_type = btrfs_file_extent_type(leaf, fi);
2735 2736 2737 2738
			encoding = btrfs_file_extent_compression(leaf, fi);
			encoding |= btrfs_file_extent_encryption(leaf, fi);
			encoding |= btrfs_file_extent_other_encoding(leaf, fi);

2739
			if (extent_type != BTRFS_FILE_EXTENT_INLINE) {
2740
				item_end +=
2741
				    btrfs_file_extent_num_bytes(leaf, fi);
2742 2743
			} else if (extent_type == BTRFS_FILE_EXTENT_INLINE) {
				item_end += btrfs_file_extent_inline_len(leaf,
C
Chris Mason 已提交
2744
									 fi);
C
Chris Mason 已提交
2745
			}
2746
			item_end--;
C
Chris Mason 已提交
2747
		}
2748
		if (item_end < new_size) {
C
Chris Mason 已提交
2749
			if (found_type == BTRFS_DIR_ITEM_KEY)
2750
				found_type = BTRFS_INODE_ITEM_KEY;
C
Chris Mason 已提交
2751
			else if (found_type == BTRFS_EXTENT_ITEM_KEY)
2752
				found_type = BTRFS_EXTENT_DATA_KEY;
C
Chris Mason 已提交
2753
			else if (found_type == BTRFS_EXTENT_DATA_KEY)
2754
				found_type = BTRFS_XATTR_ITEM_KEY;
C
Chris Mason 已提交
2755
			else if (found_type == BTRFS_XATTR_ITEM_KEY)
2756
				found_type = BTRFS_INODE_REF_KEY;
C
Chris Mason 已提交
2757
			else if (found_type)
2758
				found_type--;
C
Chris Mason 已提交
2759
			else
2760
				break;
2761
			btrfs_set_key_type(&key, found_type);
2762
			goto next;
C
Chris Mason 已提交
2763
		}
2764
		if (found_key.offset >= new_size)
C
Chris Mason 已提交
2765 2766 2767 2768 2769 2770
			del_item = 1;
		else
			del_item = 0;
		found_extent = 0;

		/* FIXME, shrink the extent if the ref count is only 1 */
2771 2772 2773 2774
		if (found_type != BTRFS_EXTENT_DATA_KEY)
			goto delete;

		if (extent_type != BTRFS_FILE_EXTENT_INLINE) {
C
Chris Mason 已提交
2775
			u64 num_dec;
2776
			extent_start = btrfs_file_extent_disk_bytenr(leaf, fi);
2777
			if (!del_item && !encoding) {
2778 2779
				u64 orig_num_bytes =
					btrfs_file_extent_num_bytes(leaf, fi);
2780
				extent_num_bytes = new_size -
2781
					found_key.offset + root->sectorsize - 1;
2782 2783
				extent_num_bytes = extent_num_bytes &
					~((u64)root->sectorsize - 1);
2784 2785 2786
				btrfs_set_file_extent_num_bytes(leaf, fi,
							 extent_num_bytes);
				num_dec = (orig_num_bytes -
C
Chris Mason 已提交
2787
					   extent_num_bytes);
2788
				if (root->ref_cows && extent_start != 0)
2789
					inode_sub_bytes(inode, num_dec);
2790
				btrfs_mark_buffer_dirty(leaf);
C
Chris Mason 已提交
2791
			} else {
2792 2793 2794
				extent_num_bytes =
					btrfs_file_extent_disk_num_bytes(leaf,
									 fi);
2795 2796 2797
				extent_offset = found_key.offset -
					btrfs_file_extent_offset(leaf, fi);

C
Chris Mason 已提交
2798
				/* FIXME blocksize != 4096 */
C
Chris Mason 已提交
2799
				num_dec = btrfs_file_extent_num_bytes(leaf, fi);
C
Chris Mason 已提交
2800 2801
				if (extent_start != 0) {
					found_extent = 1;
2802
					if (root->ref_cows)
2803
						inode_sub_bytes(inode, num_dec);
2804
				}
C
Chris Mason 已提交
2805
			}
C
Chris Mason 已提交
2806
		} else if (extent_type == BTRFS_FILE_EXTENT_INLINE) {
C
Chris Mason 已提交
2807 2808 2809 2810 2811 2812 2813 2814
			/*
			 * 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) {
2815 2816 2817
				u32 size = new_size - found_key.offset;

				if (root->ref_cows) {
2818 2819
					inode_sub_bytes(inode, item_end + 1 -
							new_size);
2820 2821 2822
				}
				size =
				    btrfs_file_extent_calc_inline_size(size);
C
Chris Mason 已提交
2823
				ret = btrfs_truncate_item(trans, root, path,
2824
							  size, 1);
C
Chris Mason 已提交
2825
				BUG_ON(ret);
2826
			} else if (root->ref_cows) {
2827 2828
				inode_sub_bytes(inode, item_end + 1 -
						found_key.offset);
C
Chris Mason 已提交
2829
			}
C
Chris Mason 已提交
2830
		}
2831
delete:
C
Chris Mason 已提交
2832
		if (del_item) {
2833 2834 2835 2836 2837 2838 2839 2840 2841 2842
			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
2843
				BUG();
2844
			}
C
Chris Mason 已提交
2845 2846 2847
		} else {
			break;
		}
2848
		if (found_extent && root->ref_cows) {
2849
			btrfs_set_path_blocking(path);
C
Chris Mason 已提交
2850
			ret = btrfs_free_extent(trans, root, extent_start,
2851 2852 2853
						extent_num_bytes, 0,
						btrfs_header_owner(leaf),
						inode->i_ino, extent_offset);
C
Chris Mason 已提交
2854 2855
			BUG_ON(ret);
		}
2856 2857 2858 2859 2860
next:
		if (path->slots[0] == 0) {
			if (pending_del_nr)
				goto del_pending;
			btrfs_release_path(root, path);
2861 2862
			if (found_type == BTRFS_INODE_ITEM_KEY)
				break;
2863 2864 2865 2866 2867 2868 2869 2870 2871 2872 2873 2874 2875 2876 2877 2878
			goto search_again;
		}

		path->slots[0]--;
		if (pending_del_nr &&
		    path->slots[0] + 1 != pending_del_slot) {
			struct btrfs_key debug;
del_pending:
			btrfs_item_key_to_cpu(path->nodes[0], &debug,
					      pending_del_slot);
			ret = btrfs_del_items(trans, root, path,
					      pending_del_slot,
					      pending_del_nr);
			BUG_ON(ret);
			pending_del_nr = 0;
			btrfs_release_path(root, path);
2879 2880
			if (found_type == BTRFS_INODE_ITEM_KEY)
				break;
2881 2882
			goto search_again;
		}
C
Chris Mason 已提交
2883 2884 2885
	}
	ret = 0;
error:
2886 2887 2888 2889
	if (pending_del_nr) {
		ret = btrfs_del_items(trans, root, path, pending_del_slot,
				      pending_del_nr);
	}
C
Chris Mason 已提交
2890 2891 2892 2893 2894 2895 2896 2897 2898 2899 2900
	btrfs_free_path(path);
	return ret;
}

/*
 * 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;
2901
	struct btrfs_root *root = BTRFS_I(inode)->root;
2902 2903 2904
	struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
	struct btrfs_ordered_extent *ordered;
	char *kaddr;
2905
	u32 blocksize = root->sectorsize;
C
Chris Mason 已提交
2906 2907 2908 2909
	pgoff_t index = from >> PAGE_CACHE_SHIFT;
	unsigned offset = from & (PAGE_CACHE_SIZE-1);
	struct page *page;
	int ret = 0;
2910
	u64 page_start;
2911
	u64 page_end;
C
Chris Mason 已提交
2912 2913 2914 2915 2916

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

	ret = -ENOMEM;
2917
again:
C
Chris Mason 已提交
2918 2919 2920
	page = grab_cache_page(mapping, index);
	if (!page)
		goto out;
2921 2922 2923 2924

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

C
Chris Mason 已提交
2925
	if (!PageUptodate(page)) {
C
Chris Mason 已提交
2926
		ret = btrfs_readpage(NULL, page);
C
Chris Mason 已提交
2927
		lock_page(page);
2928 2929 2930 2931 2932
		if (page->mapping != mapping) {
			unlock_page(page);
			page_cache_release(page);
			goto again;
		}
C
Chris Mason 已提交
2933 2934
		if (!PageUptodate(page)) {
			ret = -EIO;
2935
			goto out_unlock;
C
Chris Mason 已提交
2936 2937
		}
	}
2938
	wait_on_page_writeback(page);
2939 2940 2941 2942 2943 2944 2945 2946 2947

	lock_extent(io_tree, page_start, page_end, GFP_NOFS);
	set_page_extent_mapped(page);

	ordered = btrfs_lookup_ordered_extent(inode, page_start);
	if (ordered) {
		unlock_extent(io_tree, page_start, page_end, GFP_NOFS);
		unlock_page(page);
		page_cache_release(page);
2948
		btrfs_start_ordered_extent(inode, ordered, 1);
2949 2950 2951 2952
		btrfs_put_ordered_extent(ordered);
		goto again;
	}

2953
	btrfs_set_extent_delalloc(inode, page_start, page_end);
2954 2955 2956 2957 2958 2959 2960
	ret = 0;
	if (offset != PAGE_CACHE_SIZE) {
		kaddr = kmap(page);
		memset(kaddr + offset, 0, PAGE_CACHE_SIZE - offset);
		flush_dcache_page(page);
		kunmap(page);
	}
2961
	ClearPageChecked(page);
2962 2963
	set_page_dirty(page);
	unlock_extent(io_tree, page_start, page_end, GFP_NOFS);
C
Chris Mason 已提交
2964

2965
out_unlock:
C
Chris Mason 已提交
2966 2967 2968 2969 2970 2971
	unlock_page(page);
	page_cache_release(page);
out:
	return ret;
}

Y
Yan Zheng 已提交
2972
int btrfs_cont_expand(struct inode *inode, loff_t size)
C
Chris Mason 已提交
2973
{
Y
Yan Zheng 已提交
2974 2975 2976 2977 2978 2979 2980 2981 2982 2983
	struct btrfs_trans_handle *trans;
	struct btrfs_root *root = BTRFS_I(inode)->root;
	struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
	struct extent_map *em;
	u64 mask = root->sectorsize - 1;
	u64 hole_start = (inode->i_size + mask) & ~mask;
	u64 block_end = (size + mask) & ~mask;
	u64 last_byte;
	u64 cur_offset;
	u64 hole_size;
C
Chris Mason 已提交
2984 2985
	int err;

Y
Yan Zheng 已提交
2986 2987 2988
	if (size <= hole_start)
		return 0;

J
Josef Bacik 已提交
2989
	err = btrfs_check_metadata_free_space(root);
C
Chris Mason 已提交
2990 2991 2992
	if (err)
		return err;

Y
Yan Zheng 已提交
2993
	btrfs_truncate_page(inode->i_mapping, inode->i_size);
C
Chris Mason 已提交
2994

Y
Yan Zheng 已提交
2995 2996 2997 2998 2999 3000 3001 3002 3003 3004 3005
	while (1) {
		struct btrfs_ordered_extent *ordered;
		btrfs_wait_ordered_range(inode, hole_start,
					 block_end - hole_start);
		lock_extent(io_tree, hole_start, block_end - 1, GFP_NOFS);
		ordered = btrfs_lookup_ordered_extent(inode, hole_start);
		if (!ordered)
			break;
		unlock_extent(io_tree, hole_start, block_end - 1, GFP_NOFS);
		btrfs_put_ordered_extent(ordered);
	}
C
Chris Mason 已提交
3006

Y
Yan Zheng 已提交
3007 3008
	trans = btrfs_start_transaction(root, 1);
	btrfs_set_trans_block_group(trans, inode);
C
Chris Mason 已提交
3009

Y
Yan Zheng 已提交
3010 3011 3012 3013 3014 3015 3016 3017
	cur_offset = hole_start;
	while (1) {
		em = btrfs_get_extent(inode, NULL, 0, cur_offset,
				block_end - cur_offset, 0);
		BUG_ON(IS_ERR(em) || !em);
		last_byte = min(extent_map_end(em), block_end);
		last_byte = (last_byte + mask) & ~mask;
		if (test_bit(EXTENT_FLAG_VACANCY, &em->flags)) {
3018
			u64 hint_byte = 0;
Y
Yan Zheng 已提交
3019
			hole_size = last_byte - cur_offset;
3020 3021 3022
			err = btrfs_drop_extents(trans, root, inode,
						 cur_offset,
						 cur_offset + hole_size,
3023
						 block_end,
C
Chris Mason 已提交
3024
						 cur_offset, &hint_byte, 1);
3025 3026
			if (err)
				break;
Y
Yan Zheng 已提交
3027 3028 3029 3030 3031 3032 3033 3034 3035 3036 3037 3038
			err = btrfs_insert_file_extent(trans, root,
					inode->i_ino, cur_offset, 0,
					0, hole_size, 0, hole_size,
					0, 0, 0);
			btrfs_drop_extent_cache(inode, hole_start,
					last_byte - 1, 0);
		}
		free_extent_map(em);
		cur_offset = last_byte;
		if (err || cur_offset >= block_end)
			break;
	}
3039

Y
Yan Zheng 已提交
3040 3041 3042 3043
	btrfs_end_transaction(trans, root);
	unlock_extent(io_tree, hole_start, block_end - 1, GFP_NOFS);
	return err;
}
C
Chris Mason 已提交
3044

Y
Yan Zheng 已提交
3045 3046 3047 3048
static int btrfs_setattr(struct dentry *dentry, struct iattr *attr)
{
	struct inode *inode = dentry->d_inode;
	int err;
C
Chris Mason 已提交
3049

Y
Yan Zheng 已提交
3050 3051 3052
	err = inode_change_ok(inode, attr);
	if (err)
		return err;
C
Chris Mason 已提交
3053

3054 3055 3056 3057 3058 3059 3060 3061 3062 3063 3064 3065 3066 3067 3068
	if (S_ISREG(inode->i_mode) && (attr->ia_valid & ATTR_SIZE)) {
		if (attr->ia_size > inode->i_size) {
			err = btrfs_cont_expand(inode, attr->ia_size);
			if (err)
				return err;
		} else if (inode->i_size > 0 &&
			   attr->ia_size == 0) {

			/* 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.
			 */
			BTRFS_I(inode)->ordered_data_close = 1;
		}
C
Chris Mason 已提交
3069
	}
Y
Yan Zheng 已提交
3070

C
Chris Mason 已提交
3071
	err = inode_setattr(inode, attr);
J
Josef Bacik 已提交
3072 3073 3074

	if (!err && ((attr->ia_valid & ATTR_MODE)))
		err = btrfs_acl_chmod(inode);
C
Chris Mason 已提交
3075 3076
	return err;
}
3077

C
Chris Mason 已提交
3078 3079 3080 3081
void btrfs_delete_inode(struct inode *inode)
{
	struct btrfs_trans_handle *trans;
	struct btrfs_root *root = BTRFS_I(inode)->root;
3082
	unsigned long nr;
C
Chris Mason 已提交
3083 3084 3085 3086
	int ret;

	truncate_inode_pages(&inode->i_data, 0);
	if (is_bad_inode(inode)) {
3087
		btrfs_orphan_del(NULL, inode);
C
Chris Mason 已提交
3088 3089
		goto no_delete;
	}
C
Chris Mason 已提交
3090
	btrfs_wait_ordered_range(inode, 0, (u64)-1);
3091

3092 3093 3094 3095 3096
	if (inode->i_nlink > 0) {
		BUG_ON(btrfs_root_refs(&root->root_item) != 0);
		goto no_delete;
	}

3097
	btrfs_i_size_write(inode, 0);
3098
	trans = btrfs_join_transaction(root, 1);
3099

C
Chris Mason 已提交
3100
	btrfs_set_trans_block_group(trans, inode);
3101
	ret = btrfs_truncate_inode_items(trans, root, inode, inode->i_size, 0);
3102 3103
	if (ret) {
		btrfs_orphan_del(NULL, inode);
3104
		goto no_delete_lock;
3105 3106 3107
	}

	btrfs_orphan_del(trans, inode);
3108

3109
	nr = trans->blocks_used;
3110
	clear_inode(inode);
3111

C
Chris Mason 已提交
3112
	btrfs_end_transaction(trans, root);
3113
	btrfs_btree_balance_dirty(root, nr);
C
Chris Mason 已提交
3114
	return;
3115 3116

no_delete_lock:
3117
	nr = trans->blocks_used;
3118
	btrfs_end_transaction(trans, root);
3119
	btrfs_btree_balance_dirty(root, nr);
C
Chris Mason 已提交
3120 3121 3122 3123 3124 3125 3126 3127 3128 3129 3130 3131 3132 3133 3134 3135
no_delete:
	clear_inode(inode);
}

/*
 * 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
3136
	int ret = 0;
C
Chris Mason 已提交
3137 3138 3139

	path = btrfs_alloc_path();
	BUG_ON(!path);
3140

C
Chris Mason 已提交
3141 3142
	di = btrfs_lookup_dir_item(NULL, root, path, dir->i_ino, name,
				    namelen, 0);
Y
Yan 已提交
3143 3144
	if (IS_ERR(di))
		ret = PTR_ERR(di);
C
Chris Mason 已提交
3145 3146

	if (!di || IS_ERR(di))
3147
		goto out_err;
C
Chris Mason 已提交
3148

3149
	btrfs_dir_item_key_to_cpu(path->nodes[0], di, location);
C
Chris Mason 已提交
3150 3151 3152
out:
	btrfs_free_path(path);
	return ret;
3153 3154 3155
out_err:
	location->objectid = 0;
	goto out;
C
Chris Mason 已提交
3156 3157 3158 3159 3160 3161 3162 3163
}

/*
 * 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,
3164 3165 3166 3167
				    struct inode *dir,
				    struct dentry *dentry,
				    struct btrfs_key *location,
				    struct btrfs_root **sub_root)
C
Chris Mason 已提交
3168
{
3169 3170 3171 3172 3173 3174
	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
3175

3176 3177 3178 3179 3180
	path = btrfs_alloc_path();
	if (!path) {
		err = -ENOMEM;
		goto out;
	}
C
Chris Mason 已提交
3181

3182 3183 3184 3185 3186 3187 3188 3189 3190
	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
3191

3192 3193 3194 3195 3196
	leaf = path->nodes[0];
	ref = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_root_ref);
	if (btrfs_root_ref_dirid(leaf, ref) != dir->i_ino ||
	    btrfs_root_ref_name_len(leaf, ref) != dentry->d_name.len)
		goto out;
C
Chris Mason 已提交
3197

3198 3199 3200 3201 3202 3203 3204 3205 3206 3207 3208 3209 3210 3211 3212 3213 3214 3215 3216 3217 3218 3219 3220 3221 3222 3223 3224
	ret = memcmp_extent_buffer(leaf, dentry->d_name.name,
				   (unsigned long)(ref + 1),
				   dentry->d_name.len);
	if (ret)
		goto out;

	btrfs_release_path(root->fs_info->tree_root, path);

	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
3225 3226
}

3227 3228 3229 3230
static void inode_tree_add(struct inode *inode)
{
	struct btrfs_root *root = BTRFS_I(inode)->root;
	struct btrfs_inode *entry;
3231 3232 3233 3234 3235
	struct rb_node **p;
	struct rb_node *parent;
again:
	p = &root->inode_tree.rb_node;
	parent = NULL;
3236

3237 3238 3239
	if (hlist_unhashed(&inode->i_hash))
		return;

3240 3241 3242 3243 3244 3245
	spin_lock(&root->inode_lock);
	while (*p) {
		parent = *p;
		entry = rb_entry(parent, struct btrfs_inode, rb_node);

		if (inode->i_ino < entry->vfs_inode.i_ino)
3246
			p = &parent->rb_left;
3247
		else if (inode->i_ino > entry->vfs_inode.i_ino)
3248
			p = &parent->rb_right;
3249 3250 3251
		else {
			WARN_ON(!(entry->vfs_inode.i_state &
				  (I_WILL_FREE | I_FREEING | I_CLEAR)));
3252 3253 3254 3255
			rb_erase(parent, &root->inode_tree);
			RB_CLEAR_NODE(parent);
			spin_unlock(&root->inode_lock);
			goto again;
3256 3257 3258 3259 3260 3261 3262 3263 3264 3265
		}
	}
	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;
3266
	int empty = 0;
3267

3268
	spin_lock(&root->inode_lock);
3269 3270 3271
	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);
3272
		empty = RB_EMPTY_ROOT(&root->inode_tree);
3273
	}
3274
	spin_unlock(&root->inode_lock);
3275 3276 3277 3278 3279 3280 3281 3282 3283 3284 3285 3286 3287 3288 3289 3290 3291 3292 3293 3294 3295 3296 3297 3298 3299 3300 3301 3302 3303 3304 3305 3306 3307 3308 3309 3310 3311 3312 3313 3314 3315 3316 3317 3318 3319 3320 3321 3322 3323 3324 3325 3326 3327 3328 3329 3330 3331 3332 3333 3334 3335 3336 3337 3338 3339 3340 3341 3342 3343 3344 3345 3346

	if (empty && btrfs_root_refs(&root->root_item) == 0) {
		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);
	}
}

int btrfs_invalidate_inodes(struct btrfs_root *root)
{
	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);

		if (objectid < entry->vfs_inode.i_ino)
			node = node->rb_left;
		else if (objectid > entry->vfs_inode.i_ino)
			node = node->rb_right;
		else
			break;
	}
	if (!node) {
		while (prev) {
			entry = rb_entry(prev, struct btrfs_inode, rb_node);
			if (objectid <= entry->vfs_inode.i_ino) {
				node = prev;
				break;
			}
			prev = rb_next(prev);
		}
	}
	while (node) {
		entry = rb_entry(node, struct btrfs_inode, rb_node);
		objectid = entry->vfs_inode.i_ino + 1;
		inode = igrab(&entry->vfs_inode);
		if (inode) {
			spin_unlock(&root->inode_lock);
			if (atomic_read(&inode->i_count) > 1)
				d_prune_aliases(inode);
			/*
			 * btrfs_drop_inode will remove it from
			 * 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);
	return 0;
3347 3348
}

3349
static noinline void init_btrfs_i(struct inode *inode)
C
Chris Mason 已提交
3350
{
3351 3352 3353
	struct btrfs_inode *bi = BTRFS_I(inode);

	bi->generation = 0;
3354
	bi->sequence = 0;
3355 3356 3357
	bi->last_trans = 0;
	bi->logged_trans = 0;
	bi->delalloc_bytes = 0;
J
Josef Bacik 已提交
3358
	bi->reserved_bytes = 0;
3359 3360 3361
	bi->disk_i_size = 0;
	bi->flags = 0;
	bi->index_cnt = (u64)-1;
3362
	bi->last_unlink_trans = 0;
3363
	bi->ordered_data_close = 0;
3364 3365
	extent_map_tree_init(&BTRFS_I(inode)->extent_tree, GFP_NOFS);
	extent_io_tree_init(&BTRFS_I(inode)->io_tree,
3366
			     inode->i_mapping, GFP_NOFS);
3367 3368
	extent_io_tree_init(&BTRFS_I(inode)->io_failure_tree,
			     inode->i_mapping, GFP_NOFS);
3369
	INIT_LIST_HEAD(&BTRFS_I(inode)->delalloc_inodes);
3370
	INIT_LIST_HEAD(&BTRFS_I(inode)->ordered_operations);
3371
	RB_CLEAR_NODE(&BTRFS_I(inode)->rb_node);
3372
	btrfs_ordered_inode_tree_init(&BTRFS_I(inode)->ordered_tree);
3373
	mutex_init(&BTRFS_I(inode)->extent_mutex);
3374 3375 3376 3377 3378 3379 3380 3381 3382
	mutex_init(&BTRFS_I(inode)->log_mutex);
}

static int btrfs_init_locked_inode(struct inode *inode, void *p)
{
	struct btrfs_iget_args *args = p;
	inode->i_ino = args->ino;
	init_btrfs_i(inode);
	BTRFS_I(inode)->root = args->root;
J
Josef Bacik 已提交
3383
	btrfs_set_inode_space_info(args->root, inode);
C
Chris Mason 已提交
3384 3385 3386 3387 3388 3389
	return 0;
}

static int btrfs_find_actor(struct inode *inode, void *opaque)
{
	struct btrfs_iget_args *args = opaque;
C
Chris Mason 已提交
3390 3391
	return args->ino == inode->i_ino &&
		args->root == BTRFS_I(inode)->root;
C
Chris Mason 已提交
3392 3393
}

3394 3395 3396
static struct inode *btrfs_iget_locked(struct super_block *s,
				       u64 objectid,
				       struct btrfs_root *root)
C
Chris Mason 已提交
3397 3398 3399 3400 3401 3402 3403 3404 3405 3406 3407 3408
{
	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
3409 3410 3411 3412
/* 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,
3413
			 struct btrfs_root *root)
B
Balaji Rao 已提交
3414 3415 3416 3417 3418
{
	struct inode *inode;

	inode = btrfs_iget_locked(s, location->objectid, root);
	if (!inode)
3419
		return ERR_PTR(-ENOMEM);
B
Balaji Rao 已提交
3420 3421 3422 3423 3424

	if (inode->i_state & I_NEW) {
		BTRFS_I(inode)->root = root;
		memcpy(&BTRFS_I(inode)->location, location, sizeof(*location));
		btrfs_read_locked_inode(inode);
3425 3426

		inode_tree_add(inode);
B
Balaji Rao 已提交
3427 3428 3429 3430 3431 3432
		unlock_new_inode(inode);
	}

	return inode;
}

3433 3434 3435 3436 3437 3438 3439 3440 3441 3442 3443 3444 3445 3446 3447 3448 3449 3450 3451 3452 3453 3454 3455 3456
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);

	init_btrfs_i(inode);

	BTRFS_I(inode)->root = root;
	memcpy(&BTRFS_I(inode)->location, key, sizeof(*key));
	BTRFS_I(inode)->dummy_inode = 1;

	inode->i_ino = BTRFS_EMPTY_SUBVOL_DIR_OBJECTID;
	inode->i_op = &simple_dir_inode_operations;
	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;
}

3457
struct inode *btrfs_lookup_dentry(struct inode *dir, struct dentry *dentry)
C
Chris Mason 已提交
3458
{
C
Chris Mason 已提交
3459
	struct inode *inode;
3460
	struct btrfs_root *root = BTRFS_I(dir)->root;
C
Chris Mason 已提交
3461 3462
	struct btrfs_root *sub_root = root;
	struct btrfs_key location;
3463
	int index;
3464
	int ret;
C
Chris Mason 已提交
3465

3466 3467
	dentry->d_op = &btrfs_dentry_operations;

C
Chris Mason 已提交
3468 3469
	if (dentry->d_name.len > BTRFS_NAME_LEN)
		return ERR_PTR(-ENAMETOOLONG);
3470

C
Chris Mason 已提交
3471
	ret = btrfs_inode_by_name(dir, dentry, &location);
3472

C
Chris Mason 已提交
3473 3474
	if (ret < 0)
		return ERR_PTR(ret);
3475

3476 3477 3478 3479 3480 3481 3482 3483 3484 3485
	if (location.objectid == 0)
		return NULL;

	if (location.type == BTRFS_INODE_ITEM_KEY) {
		inode = btrfs_iget(dir->i_sb, &location, root);
		return inode;
	}

	BUG_ON(location.type != BTRFS_ROOT_ITEM_KEY);

3486
	index = srcu_read_lock(&root->fs_info->subvol_srcu);
3487 3488 3489 3490 3491 3492 3493 3494
	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 {
3495
		inode = btrfs_iget(dir->i_sb, &location, sub_root);
C
Chris Mason 已提交
3496
	}
3497 3498
	srcu_read_unlock(&root->fs_info->subvol_srcu, index);

3499 3500 3501
	return inode;
}

3502 3503 3504 3505 3506 3507 3508 3509 3510 3511 3512 3513 3514
static int btrfs_dentry_delete(struct dentry *dentry)
{
	struct btrfs_root *root;

	if (!dentry->d_inode)
		return 0;

	root = BTRFS_I(dentry->d_inode)->root;
	if (btrfs_root_refs(&root->root_item) == 0)
		return 1;
	return 0;
}

3515 3516 3517 3518 3519 3520 3521 3522
static struct dentry *btrfs_lookup(struct inode *dir, struct dentry *dentry,
				   struct nameidata *nd)
{
	struct inode *inode;

	inode = btrfs_lookup_dentry(dir, dentry);
	if (IS_ERR(inode))
		return ERR_CAST(inode);
3523

C
Chris Mason 已提交
3524 3525 3526 3527 3528 3529 3530
	return d_splice_alias(inode, dentry);
}

static unsigned char btrfs_filetype_table[] = {
	DT_UNKNOWN, DT_REG, DT_DIR, DT_CHR, DT_BLK, DT_FIFO, DT_SOCK, DT_LNK
};

3531 3532
static int btrfs_real_readdir(struct file *filp, void *dirent,
			      filldir_t filldir)
C
Chris Mason 已提交
3533
{
3534
	struct inode *inode = filp->f_dentry->d_inode;
C
Chris Mason 已提交
3535 3536 3537 3538
	struct btrfs_root *root = BTRFS_I(inode)->root;
	struct btrfs_item *item;
	struct btrfs_dir_item *di;
	struct btrfs_key key;
3539
	struct btrfs_key found_key;
C
Chris Mason 已提交
3540 3541 3542
	struct btrfs_path *path;
	int ret;
	u32 nritems;
3543
	struct extent_buffer *leaf;
C
Chris Mason 已提交
3544 3545 3546 3547 3548 3549 3550 3551
	int slot;
	int advance;
	unsigned char d_type;
	int over = 0;
	u32 di_cur;
	u32 di_total;
	u32 di_len;
	int key_type = BTRFS_DIR_INDEX_KEY;
3552 3553 3554
	char tmp_name[32];
	char *name_ptr;
	int name_len;
C
Chris Mason 已提交
3555 3556 3557 3558

	/* FIXME, use a real flag for deciding about the key type */
	if (root->fs_info->tree_root == root)
		key_type = BTRFS_DIR_ITEM_KEY;
3559

3560 3561 3562 3563 3564 3565 3566 3567 3568 3569 3570
	/* special case for "." */
	if (filp->f_pos == 0) {
		over = filldir(dirent, ".", 1,
			       1, inode->i_ino,
			       DT_DIR);
		if (over)
			return 0;
		filp->f_pos = 1;
	}
	/* special case for .., just use the back ref */
	if (filp->f_pos == 1) {
3571
		u64 pino = parent_ino(filp->f_path.dentry);
3572
		over = filldir(dirent, "..", 2,
3573
			       2, pino, DT_DIR);
3574
		if (over)
3575
			return 0;
3576 3577
		filp->f_pos = 2;
	}
3578 3579 3580
	path = btrfs_alloc_path();
	path->reada = 2;

C
Chris Mason 已提交
3581 3582
	btrfs_set_key_type(&key, key_type);
	key.offset = filp->f_pos;
3583
	key.objectid = inode->i_ino;
3584

C
Chris Mason 已提交
3585 3586 3587 3588
	ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
	if (ret < 0)
		goto err;
	advance = 0;
3589 3590

	while (1) {
3591 3592
		leaf = path->nodes[0];
		nritems = btrfs_header_nritems(leaf);
C
Chris Mason 已提交
3593 3594
		slot = path->slots[0];
		if (advance || slot >= nritems) {
3595
			if (slot >= nritems - 1) {
C
Chris Mason 已提交
3596 3597 3598
				ret = btrfs_next_leaf(root, path);
				if (ret)
					break;
3599 3600
				leaf = path->nodes[0];
				nritems = btrfs_header_nritems(leaf);
C
Chris Mason 已提交
3601 3602 3603 3604 3605 3606
				slot = path->slots[0];
			} else {
				slot++;
				path->slots[0]++;
			}
		}
3607

C
Chris Mason 已提交
3608
		advance = 1;
3609 3610 3611 3612
		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
3613
			break;
3614
		if (btrfs_key_type(&found_key) != key_type)
C
Chris Mason 已提交
3615
			break;
3616
		if (found_key.offset < filp->f_pos)
C
Chris Mason 已提交
3617
			continue;
3618 3619

		filp->f_pos = found_key.offset;
3620

C
Chris Mason 已提交
3621 3622
		di = btrfs_item_ptr(leaf, slot, struct btrfs_dir_item);
		di_cur = 0;
3623
		di_total = btrfs_item_size(leaf, item);
3624 3625

		while (di_cur < di_total) {
3626 3627 3628
			struct btrfs_key location;

			name_len = btrfs_dir_name_len(leaf, di);
3629
			if (name_len <= sizeof(tmp_name)) {
3630 3631 3632
				name_ptr = tmp_name;
			} else {
				name_ptr = kmalloc(name_len, GFP_NOFS);
3633 3634 3635 3636
				if (!name_ptr) {
					ret = -ENOMEM;
					goto err;
				}
3637 3638 3639 3640 3641 3642
			}
			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);
3643 3644 3645 3646 3647 3648 3649 3650 3651

			/* is this a reference to our own snapshot? If so
			 * skip it
			 */
			if (location.type == BTRFS_ROOT_ITEM_KEY &&
			    location.objectid == root->root_key.objectid) {
				over = 0;
				goto skip;
			}
3652
			over = filldir(dirent, name_ptr, name_len,
3653
				       found_key.offset, location.objectid,
C
Chris Mason 已提交
3654
				       d_type);
3655

3656
skip:
3657 3658 3659
			if (name_ptr != tmp_name)
				kfree(name_ptr);

C
Chris Mason 已提交
3660 3661
			if (over)
				goto nopos;
J
Josef Bacik 已提交
3662
			di_len = btrfs_dir_name_len(leaf, di) +
3663
				 btrfs_dir_data_len(leaf, di) + sizeof(*di);
C
Chris Mason 已提交
3664 3665 3666 3667
			di_cur += di_len;
			di = (struct btrfs_dir_item *)((char *)di + di_len);
		}
	}
3668 3669

	/* Reached end of directory/root. Bump pos past the last item. */
3670
	if (key_type == BTRFS_DIR_INDEX_KEY)
3671
		filp->f_pos = INT_LIMIT(off_t);
3672 3673
	else
		filp->f_pos++;
C
Chris Mason 已提交
3674 3675 3676 3677 3678 3679 3680 3681 3682 3683 3684 3685 3686
nopos:
	ret = 0;
err:
	btrfs_free_path(path);
	return ret;
}

int btrfs_write_inode(struct inode *inode, int wait)
{
	struct btrfs_root *root = BTRFS_I(inode)->root;
	struct btrfs_trans_handle *trans;
	int ret = 0;

Y
Yan Zheng 已提交
3687
	if (root->fs_info->btree_inode == inode)
3688 3689
		return 0;

C
Chris Mason 已提交
3690
	if (wait) {
3691
		trans = btrfs_join_transaction(root, 1);
C
Chris Mason 已提交
3692 3693 3694 3695 3696 3697 3698
		btrfs_set_trans_block_group(trans, inode);
		ret = btrfs_commit_transaction(trans, root);
	}
	return ret;
}

/*
3699
 * This is somewhat expensive, updating the tree every time the
C
Chris Mason 已提交
3700 3701 3702 3703 3704 3705 3706 3707 3708
 * 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.
 */
void btrfs_dirty_inode(struct inode *inode)
{
	struct btrfs_root *root = BTRFS_I(inode)->root;
	struct btrfs_trans_handle *trans;

3709
	trans = btrfs_join_transaction(root, 1);
C
Chris Mason 已提交
3710 3711 3712 3713 3714
	btrfs_set_trans_block_group(trans, inode);
	btrfs_update_inode(trans, root, inode);
	btrfs_end_transaction(trans, root);
}

C
Chris Mason 已提交
3715 3716 3717 3718 3719
/*
 * find the highest existing sequence number in a directory
 * and then set the in-memory index_cnt variable to reflect
 * free sequence numbers
 */
3720 3721 3722 3723 3724 3725 3726 3727 3728 3729 3730 3731 3732 3733 3734 3735 3736 3737 3738 3739 3740 3741 3742 3743 3744 3745 3746 3747 3748 3749 3750 3751 3752 3753 3754 3755 3756 3757 3758 3759 3760 3761 3762 3763 3764 3765 3766 3767 3768 3769 3770 3771
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;

	key.objectid = inode->i_ino;
	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]);

	if (found_key.objectid != inode->i_ino ||
	    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
3772 3773 3774 3775
/*
 * 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
 */
3776
int btrfs_set_inode_index(struct inode *dir, u64 *index)
3777 3778 3779 3780 3781
{
	int ret = 0;

	if (BTRFS_I(dir)->index_cnt == (u64)-1) {
		ret = btrfs_set_inode_index_count(dir);
C
Chris Mason 已提交
3782
		if (ret)
3783 3784 3785
			return ret;
	}

3786
	*index = BTRFS_I(dir)->index_cnt;
3787 3788 3789 3790 3791
	BTRFS_I(dir)->index_cnt++;

	return ret;
}

C
Chris Mason 已提交
3792 3793
static struct inode *btrfs_new_inode(struct btrfs_trans_handle *trans,
				     struct btrfs_root *root,
3794
				     struct inode *dir,
3795
				     const char *name, int name_len,
3796 3797
				     u64 ref_objectid, u64 objectid,
				     u64 alloc_hint, int mode, u64 *index)
C
Chris Mason 已提交
3798 3799
{
	struct inode *inode;
3800
	struct btrfs_inode_item *inode_item;
C
Chris Mason 已提交
3801
	struct btrfs_key *location;
3802
	struct btrfs_path *path;
3803 3804 3805 3806
	struct btrfs_inode_ref *ref;
	struct btrfs_key key[2];
	u32 sizes[2];
	unsigned long ptr;
C
Chris Mason 已提交
3807 3808 3809
	int ret;
	int owner;

3810 3811 3812
	path = btrfs_alloc_path();
	BUG_ON(!path);

C
Chris Mason 已提交
3813 3814 3815 3816
	inode = new_inode(root->fs_info->sb);
	if (!inode)
		return ERR_PTR(-ENOMEM);

3817
	if (dir) {
3818
		ret = btrfs_set_inode_index(dir, index);
3819 3820
		if (ret) {
			iput(inode);
3821
			return ERR_PTR(ret);
3822
		}
3823 3824 3825 3826 3827 3828
	}
	/*
	 * index_cnt is ignored for everything but a dir,
	 * btrfs_get_inode_index_count has an explanation for the magic
	 * number
	 */
3829
	init_btrfs_i(inode);
3830
	BTRFS_I(inode)->index_cnt = 2;
C
Chris Mason 已提交
3831
	BTRFS_I(inode)->root = root;
3832
	BTRFS_I(inode)->generation = trans->transid;
J
Josef Bacik 已提交
3833
	btrfs_set_inode_space_info(root, inode);
3834

C
Chris Mason 已提交
3835 3836 3837 3838
	if (mode & S_IFDIR)
		owner = 0;
	else
		owner = 1;
3839 3840
	BTRFS_I(inode)->block_group =
			btrfs_find_block_group(root, 0, alloc_hint, owner);
3841 3842 3843 3844 3845 3846 3847 3848 3849 3850 3851 3852

	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);

3853
	path->leave_spinning = 1;
3854 3855
	ret = btrfs_insert_empty_items(trans, root, path, key, sizes, 2);
	if (ret != 0)
3856 3857
		goto fail;

C
Chris Mason 已提交
3858
	inode->i_uid = current_fsuid();
3859

3860
	if (dir && (dir->i_mode & S_ISGID)) {
3861 3862 3863 3864 3865 3866
		inode->i_gid = dir->i_gid;
		if (S_ISDIR(mode))
			mode |= S_ISGID;
	} else
		inode->i_gid = current_fsgid();

C
Chris Mason 已提交
3867 3868
	inode->i_mode = mode;
	inode->i_ino = objectid;
3869
	inode_set_bytes(inode, 0);
C
Chris Mason 已提交
3870
	inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
3871 3872
	inode_item = btrfs_item_ptr(path->nodes[0], path->slots[0],
				  struct btrfs_inode_item);
3873
	fill_inode_item(trans, path->nodes[0], inode_item, inode);
3874 3875 3876 3877

	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);
3878
	btrfs_set_inode_ref_index(path->nodes[0], ref, *index);
3879 3880 3881
	ptr = (unsigned long)(ref + 1);
	write_extent_buffer(path->nodes[0], name, ptr, name_len);

3882 3883 3884
	btrfs_mark_buffer_dirty(path->nodes[0]);
	btrfs_free_path(path);

C
Chris Mason 已提交
3885 3886 3887 3888 3889
	location = &BTRFS_I(inode)->location;
	location->objectid = objectid;
	location->offset = 0;
	btrfs_set_key_type(location, BTRFS_INODE_ITEM_KEY);

3890 3891
	btrfs_inherit_iflags(inode, dir);

3892 3893 3894 3895 3896 3897 3898
	if ((mode & S_IFREG)) {
		if (btrfs_test_opt(root, NODATASUM))
			BTRFS_I(inode)->flags |= BTRFS_INODE_NODATASUM;
		if (btrfs_test_opt(root, NODATACOW))
			BTRFS_I(inode)->flags |= BTRFS_INODE_NODATACOW;
	}

C
Chris Mason 已提交
3899
	insert_inode_hash(inode);
3900
	inode_tree_add(inode);
C
Chris Mason 已提交
3901
	return inode;
3902
fail:
3903 3904
	if (dir)
		BTRFS_I(dir)->index_cnt--;
3905
	btrfs_free_path(path);
3906
	iput(inode);
3907
	return ERR_PTR(ret);
C
Chris Mason 已提交
3908 3909 3910 3911 3912 3913 3914
}

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
3915 3916 3917 3918 3919 3920
/*
 * 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.
 */
3921 3922 3923
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
3924
{
3925
	int ret = 0;
C
Chris Mason 已提交
3926
	struct btrfs_key key;
3927
	struct btrfs_root *root = BTRFS_I(parent_inode)->root;
3928

3929 3930 3931 3932 3933 3934 3935 3936 3937 3938 3939 3940 3941 3942 3943 3944 3945 3946
	if (unlikely(inode->i_ino == BTRFS_FIRST_FREE_OBJECTID)) {
		memcpy(&key, &BTRFS_I(inode)->root->root_key, sizeof(key));
	} else {
		key.objectid = inode->i_ino;
		btrfs_set_key_type(&key, BTRFS_INODE_ITEM_KEY);
		key.offset = 0;
	}

	if (unlikely(inode->i_ino == BTRFS_FIRST_FREE_OBJECTID)) {
		ret = btrfs_add_root_ref(trans, root->fs_info->tree_root,
					 key.objectid, root->root_key.objectid,
					 parent_inode->i_ino,
					 index, name, name_len);
	} else if (add_backref) {
		ret = btrfs_insert_inode_ref(trans, root,
					     name, name_len, inode->i_ino,
					     parent_inode->i_ino, index);
	}
C
Chris Mason 已提交
3947 3948

	if (ret == 0) {
3949 3950 3951 3952 3953
		ret = btrfs_insert_dir_item(trans, root, name, name_len,
					    parent_inode->i_ino, &key,
					    btrfs_inode_type(inode), index);
		BUG_ON(ret);

3954
		btrfs_i_size_write(parent_inode, parent_inode->i_size +
3955
				   name_len * 2);
3956
		parent_inode->i_mtime = parent_inode->i_ctime = CURRENT_TIME;
3957
		ret = btrfs_update_inode(trans, root, parent_inode);
C
Chris Mason 已提交
3958 3959 3960 3961 3962
	}
	return ret;
}

static int btrfs_add_nondir(struct btrfs_trans_handle *trans,
3963
			    struct dentry *dentry, struct inode *inode,
3964
			    int backref, u64 index)
C
Chris Mason 已提交
3965
{
3966 3967 3968
	int err = btrfs_add_link(trans, dentry->d_parent->d_inode,
				 inode, dentry->d_name.name,
				 dentry->d_name.len, backref, index);
C
Chris Mason 已提交
3969 3970 3971 3972 3973 3974 3975 3976 3977
	if (!err) {
		d_instantiate(dentry, inode);
		return 0;
	}
	if (err > 0)
		err = -EEXIST;
	return err;
}

J
Josef Bacik 已提交
3978 3979 3980 3981 3982
static int btrfs_mknod(struct inode *dir, struct dentry *dentry,
			int mode, dev_t rdev)
{
	struct btrfs_trans_handle *trans;
	struct btrfs_root *root = BTRFS_I(dir)->root;
3983
	struct inode *inode = NULL;
J
Josef Bacik 已提交
3984 3985 3986
	int err;
	int drop_inode = 0;
	u64 objectid;
3987
	unsigned long nr = 0;
3988
	u64 index = 0;
J
Josef Bacik 已提交
3989 3990 3991 3992

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

J
Josef Bacik 已提交
3993
	err = btrfs_check_metadata_free_space(root);
3994 3995 3996
	if (err)
		goto fail;

J
Josef Bacik 已提交
3997 3998 3999 4000 4001 4002 4003 4004 4005
	trans = btrfs_start_transaction(root, 1);
	btrfs_set_trans_block_group(trans, dir);

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

4006
	inode = btrfs_new_inode(trans, root, dir, dentry->d_name.name,
4007 4008
				dentry->d_name.len,
				dentry->d_parent->d_inode->i_ino, objectid,
4009
				BTRFS_I(dir)->block_group, mode, &index);
J
Josef Bacik 已提交
4010 4011 4012 4013
	err = PTR_ERR(inode);
	if (IS_ERR(inode))
		goto out_unlock;

J
Jim Owens 已提交
4014
	err = btrfs_init_inode_security(inode, dir);
J
Josef Bacik 已提交
4015 4016 4017 4018 4019
	if (err) {
		drop_inode = 1;
		goto out_unlock;
	}

J
Josef Bacik 已提交
4020
	btrfs_set_trans_block_group(trans, inode);
4021
	err = btrfs_add_nondir(trans, dentry, inode, 0, index);
J
Josef Bacik 已提交
4022 4023 4024 4025 4026
	if (err)
		drop_inode = 1;
	else {
		inode->i_op = &btrfs_special_inode_operations;
		init_special_inode(inode, inode->i_mode, rdev);
4027
		btrfs_update_inode(trans, root, inode);
J
Josef Bacik 已提交
4028 4029 4030 4031
	}
	btrfs_update_inode_block_group(trans, inode);
	btrfs_update_inode_block_group(trans, dir);
out_unlock:
4032
	nr = trans->blocks_used;
4033
	btrfs_end_transaction_throttle(trans, root);
4034
fail:
J
Josef Bacik 已提交
4035 4036 4037 4038
	if (drop_inode) {
		inode_dec_link_count(inode);
		iput(inode);
	}
4039
	btrfs_btree_balance_dirty(root, nr);
J
Josef Bacik 已提交
4040 4041 4042
	return err;
}

C
Chris Mason 已提交
4043 4044 4045 4046 4047
static int btrfs_create(struct inode *dir, struct dentry *dentry,
			int mode, struct nameidata *nd)
{
	struct btrfs_trans_handle *trans;
	struct btrfs_root *root = BTRFS_I(dir)->root;
4048
	struct inode *inode = NULL;
C
Chris Mason 已提交
4049 4050
	int err;
	int drop_inode = 0;
4051
	unsigned long nr = 0;
C
Chris Mason 已提交
4052
	u64 objectid;
4053
	u64 index = 0;
C
Chris Mason 已提交
4054

J
Josef Bacik 已提交
4055
	err = btrfs_check_metadata_free_space(root);
4056 4057
	if (err)
		goto fail;
C
Chris Mason 已提交
4058 4059 4060 4061 4062 4063 4064 4065 4066
	trans = btrfs_start_transaction(root, 1);
	btrfs_set_trans_block_group(trans, dir);

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

4067
	inode = btrfs_new_inode(trans, root, dir, dentry->d_name.name,
4068 4069
				dentry->d_name.len,
				dentry->d_parent->d_inode->i_ino,
4070 4071
				objectid, BTRFS_I(dir)->block_group, mode,
				&index);
C
Chris Mason 已提交
4072 4073 4074 4075
	err = PTR_ERR(inode);
	if (IS_ERR(inode))
		goto out_unlock;

J
Jim Owens 已提交
4076
	err = btrfs_init_inode_security(inode, dir);
J
Josef Bacik 已提交
4077 4078 4079 4080 4081
	if (err) {
		drop_inode = 1;
		goto out_unlock;
	}

C
Chris Mason 已提交
4082
	btrfs_set_trans_block_group(trans, inode);
4083
	err = btrfs_add_nondir(trans, dentry, inode, 0, index);
C
Chris Mason 已提交
4084 4085 4086 4087
	if (err)
		drop_inode = 1;
	else {
		inode->i_mapping->a_ops = &btrfs_aops;
C
Chris Mason 已提交
4088
		inode->i_mapping->backing_dev_info = &root->fs_info->bdi;
C
Chris Mason 已提交
4089 4090
		inode->i_fop = &btrfs_file_operations;
		inode->i_op = &btrfs_file_inode_operations;
4091
		BTRFS_I(inode)->io_tree.ops = &btrfs_extent_io_ops;
C
Chris Mason 已提交
4092 4093 4094 4095
	}
	btrfs_update_inode_block_group(trans, inode);
	btrfs_update_inode_block_group(trans, dir);
out_unlock:
4096
	nr = trans->blocks_used;
4097
	btrfs_end_transaction_throttle(trans, root);
4098
fail:
C
Chris Mason 已提交
4099 4100 4101 4102
	if (drop_inode) {
		inode_dec_link_count(inode);
		iput(inode);
	}
4103
	btrfs_btree_balance_dirty(root, nr);
C
Chris Mason 已提交
4104 4105 4106 4107 4108 4109 4110 4111 4112
	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;
4113
	u64 index;
4114
	unsigned long nr = 0;
C
Chris Mason 已提交
4115 4116 4117 4118 4119 4120
	int err;
	int drop_inode = 0;

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

4121
	btrfs_inc_nlink(inode);
J
Josef Bacik 已提交
4122
	err = btrfs_check_metadata_free_space(root);
4123 4124
	if (err)
		goto fail;
4125
	err = btrfs_set_inode_index(dir, &index);
4126 4127 4128
	if (err)
		goto fail;

C
Chris Mason 已提交
4129
	trans = btrfs_start_transaction(root, 1);
4130

C
Chris Mason 已提交
4131 4132
	btrfs_set_trans_block_group(trans, dir);
	atomic_inc(&inode->i_count);
4133

4134
	err = btrfs_add_nondir(trans, dentry, inode, 1, index);
4135

4136
	if (err) {
4137
		drop_inode = 1;
4138 4139 4140 4141 4142 4143
	} else {
		btrfs_update_inode_block_group(trans, dir);
		err = btrfs_update_inode(trans, root, inode);
		BUG_ON(err);
		btrfs_log_new_name(trans, inode, NULL, dentry->d_parent);
	}
C
Chris Mason 已提交
4144

4145
	nr = trans->blocks_used;
4146
	btrfs_end_transaction_throttle(trans, root);
4147
fail:
C
Chris Mason 已提交
4148 4149 4150 4151
	if (drop_inode) {
		inode_dec_link_count(inode);
		iput(inode);
	}
4152
	btrfs_btree_balance_dirty(root, nr);
C
Chris Mason 已提交
4153 4154 4155 4156 4157
	return err;
}

static int btrfs_mkdir(struct inode *dir, struct dentry *dentry, int mode)
{
4158
	struct inode *inode = NULL;
C
Chris Mason 已提交
4159 4160 4161 4162
	struct btrfs_trans_handle *trans;
	struct btrfs_root *root = BTRFS_I(dir)->root;
	int err = 0;
	int drop_on_err = 0;
4163
	u64 objectid = 0;
4164
	u64 index = 0;
4165
	unsigned long nr = 1;
C
Chris Mason 已提交
4166

J
Josef Bacik 已提交
4167
	err = btrfs_check_metadata_free_space(root);
4168 4169 4170
	if (err)
		goto out_unlock;

C
Chris Mason 已提交
4171 4172
	trans = btrfs_start_transaction(root, 1);
	btrfs_set_trans_block_group(trans, dir);
4173

C
Chris Mason 已提交
4174 4175 4176 4177 4178 4179 4180 4181 4182 4183 4184
	if (IS_ERR(trans)) {
		err = PTR_ERR(trans);
		goto out_unlock;
	}

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

4185
	inode = btrfs_new_inode(trans, root, dir, dentry->d_name.name,
4186 4187
				dentry->d_name.len,
				dentry->d_parent->d_inode->i_ino, objectid,
4188 4189
				BTRFS_I(dir)->block_group, S_IFDIR | mode,
				&index);
C
Chris Mason 已提交
4190 4191 4192 4193
	if (IS_ERR(inode)) {
		err = PTR_ERR(inode);
		goto out_fail;
	}
4194

C
Chris Mason 已提交
4195
	drop_on_err = 1;
J
Josef Bacik 已提交
4196

J
Jim Owens 已提交
4197
	err = btrfs_init_inode_security(inode, dir);
J
Josef Bacik 已提交
4198 4199 4200
	if (err)
		goto out_fail;

C
Chris Mason 已提交
4201 4202 4203 4204
	inode->i_op = &btrfs_dir_inode_operations;
	inode->i_fop = &btrfs_dir_file_operations;
	btrfs_set_trans_block_group(trans, inode);

4205
	btrfs_i_size_write(inode, 0);
C
Chris Mason 已提交
4206 4207 4208
	err = btrfs_update_inode(trans, root, inode);
	if (err)
		goto out_fail;
4209

4210 4211 4212
	err = btrfs_add_link(trans, dentry->d_parent->d_inode,
				 inode, dentry->d_name.name,
				 dentry->d_name.len, 0, index);
C
Chris Mason 已提交
4213 4214
	if (err)
		goto out_fail;
4215

C
Chris Mason 已提交
4216 4217 4218 4219 4220 4221
	d_instantiate(dentry, inode);
	drop_on_err = 0;
	btrfs_update_inode_block_group(trans, inode);
	btrfs_update_inode_block_group(trans, dir);

out_fail:
4222
	nr = trans->blocks_used;
4223
	btrfs_end_transaction_throttle(trans, root);
4224

C
Chris Mason 已提交
4225 4226 4227
out_unlock:
	if (drop_on_err)
		iput(inode);
4228
	btrfs_btree_balance_dirty(root, nr);
C
Chris Mason 已提交
4229 4230 4231
	return err;
}

C
Chris Mason 已提交
4232 4233 4234 4235
/* 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.
 */
4236 4237
static int merge_extent_mapping(struct extent_map_tree *em_tree,
				struct extent_map *existing,
4238 4239
				struct extent_map *em,
				u64 map_start, u64 map_len)
4240 4241 4242
{
	u64 start_diff;

4243 4244 4245 4246
	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
4247 4248
	if (em->block_start < EXTENT_MAP_LAST_BYTE &&
	    !test_bit(EXTENT_FLAG_COMPRESSED, &em->flags)) {
4249
		em->block_start += start_diff;
C
Chris Mason 已提交
4250 4251
		em->block_len -= start_diff;
	}
4252
	return add_extent_mapping(em_tree, em);
4253 4254
}

C
Chris Mason 已提交
4255 4256 4257 4258 4259 4260 4261 4262 4263 4264 4265 4266 4267 4268 4269 4270 4271 4272 4273 4274 4275
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;

	WARN_ON(pg_offset != 0);
	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);
	ptr = btrfs_file_extent_inline_start(item);

	read_extent_buffer(leaf, tmp, ptr, inline_size);

4276
	max_size = min_t(unsigned long, PAGE_CACHE_SIZE, max_size);
C
Chris Mason 已提交
4277 4278 4279 4280 4281 4282 4283 4284 4285 4286 4287 4288 4289 4290
	ret = btrfs_zlib_decompress(tmp, page, extent_offset,
				    inline_size, max_size);
	if (ret) {
		char *kaddr = kmap_atomic(page, KM_USER0);
		unsigned long copy_size = min_t(u64,
				  PAGE_CACHE_SIZE - pg_offset,
				  max_size - extent_offset);
		memset(kaddr + pg_offset, 0, copy_size);
		kunmap_atomic(kaddr, KM_USER0);
	}
	kfree(tmp);
	return 0;
}

C
Chris Mason 已提交
4291 4292
/*
 * a bit scary, this does extent mapping from logical file offset to the disk.
C
Chris Mason 已提交
4293 4294
 * 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
4295 4296 4297 4298
 * 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
4299

4300
struct extent_map *btrfs_get_extent(struct inode *inode, struct page *page,
4301
				    size_t pg_offset, u64 start, u64 len,
4302 4303 4304 4305
				    int create)
{
	int ret;
	int err = 0;
4306
	u64 bytenr;
4307 4308 4309 4310
	u64 extent_start = 0;
	u64 extent_end = 0;
	u64 objectid = inode->i_ino;
	u32 found_type;
4311
	struct btrfs_path *path = NULL;
4312 4313
	struct btrfs_root *root = BTRFS_I(inode)->root;
	struct btrfs_file_extent_item *item;
4314 4315
	struct extent_buffer *leaf;
	struct btrfs_key found_key;
4316 4317
	struct extent_map *em = NULL;
	struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
4318
	struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
4319
	struct btrfs_trans_handle *trans = NULL;
C
Chris Mason 已提交
4320
	int compressed;
4321 4322

again:
4323
	read_lock(&em_tree->lock);
4324
	em = lookup_extent_mapping(em_tree, start, len);
4325 4326
	if (em)
		em->bdev = root->fs_info->fs_devices->latest_bdev;
4327
	read_unlock(&em_tree->lock);
4328

4329
	if (em) {
4330 4331 4332
		if (em->start > start || em->start + em->len <= start)
			free_extent_map(em);
		else if (em->block_start == EXTENT_MAP_INLINE && page)
4333 4334 4335
			free_extent_map(em);
		else
			goto out;
4336
	}
4337
	em = alloc_extent_map(GFP_NOFS);
4338
	if (!em) {
4339 4340
		err = -ENOMEM;
		goto out;
4341
	}
4342
	em->bdev = root->fs_info->fs_devices->latest_bdev;
4343
	em->start = EXTENT_MAP_HOLE;
4344
	em->orig_start = EXTENT_MAP_HOLE;
4345
	em->len = (u64)-1;
C
Chris Mason 已提交
4346
	em->block_len = (u64)-1;
4347 4348 4349 4350 4351 4352

	if (!path) {
		path = btrfs_alloc_path();
		BUG_ON(!path);
	}

4353 4354
	ret = btrfs_lookup_file_extent(trans, root, path,
				       objectid, start, trans != NULL);
4355 4356 4357 4358 4359 4360 4361 4362 4363 4364 4365
	if (ret < 0) {
		err = ret;
		goto out;
	}

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

4366 4367
	leaf = path->nodes[0];
	item = btrfs_item_ptr(leaf, path->slots[0],
4368 4369
			      struct btrfs_file_extent_item);
	/* are we inside the extent that was found? */
4370 4371 4372
	btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
	found_type = btrfs_key_type(&found_key);
	if (found_key.objectid != objectid ||
4373 4374 4375 4376
	    found_type != BTRFS_EXTENT_DATA_KEY) {
		goto not_found;
	}

4377 4378
	found_type = btrfs_file_extent_type(leaf, item);
	extent_start = found_key.offset;
C
Chris Mason 已提交
4379
	compressed = btrfs_file_extent_compression(leaf, item);
Y
Yan Zheng 已提交
4380 4381
	if (found_type == BTRFS_FILE_EXTENT_REG ||
	    found_type == BTRFS_FILE_EXTENT_PREALLOC) {
4382
		extent_end = extent_start +
4383
		       btrfs_file_extent_num_bytes(leaf, item);
Y
Yan Zheng 已提交
4384 4385 4386 4387 4388 4389 4390 4391 4392 4393 4394 4395 4396 4397
	} 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;
4398
			}
Y
Yan Zheng 已提交
4399 4400 4401
			if (ret > 0)
				goto not_found;
			leaf = path->nodes[0];
4402
		}
Y
Yan Zheng 已提交
4403 4404 4405 4406 4407 4408 4409 4410 4411 4412 4413
		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
4414 4415
	if (found_type == BTRFS_FILE_EXTENT_REG ||
	    found_type == BTRFS_FILE_EXTENT_PREALLOC) {
Y
Yan Zheng 已提交
4416 4417
		em->start = extent_start;
		em->len = extent_end - extent_start;
4418 4419
		em->orig_start = extent_start -
				 btrfs_file_extent_offset(leaf, item);
4420 4421
		bytenr = btrfs_file_extent_disk_bytenr(leaf, item);
		if (bytenr == 0) {
4422
			em->block_start = EXTENT_MAP_HOLE;
4423 4424
			goto insert;
		}
C
Chris Mason 已提交
4425 4426 4427 4428 4429 4430 4431 4432 4433
		if (compressed) {
			set_bit(EXTENT_FLAG_COMPRESSED, &em->flags);
			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
4434 4435
			if (found_type == BTRFS_FILE_EXTENT_PREALLOC)
				set_bit(EXTENT_FLAG_PREALLOC, &em->flags);
C
Chris Mason 已提交
4436
		}
4437 4438
		goto insert;
	} else if (found_type == BTRFS_FILE_EXTENT_INLINE) {
4439
		unsigned long ptr;
4440
		char *map;
4441 4442 4443
		size_t size;
		size_t extent_offset;
		size_t copy_size;
4444

4445
		em->block_start = EXTENT_MAP_INLINE;
C
Chris Mason 已提交
4446
		if (!page || create) {
4447
			em->start = extent_start;
Y
Yan Zheng 已提交
4448
			em->len = extent_end - extent_start;
4449 4450
			goto out;
		}
4451

Y
Yan Zheng 已提交
4452 4453
		size = btrfs_file_extent_inline_len(leaf, item);
		extent_offset = page_offset(page) + pg_offset - extent_start;
4454
		copy_size = min_t(u64, PAGE_CACHE_SIZE - pg_offset,
4455 4456
				size - extent_offset);
		em->start = extent_start + extent_offset;
4457 4458
		em->len = (copy_size + root->sectorsize - 1) &
			~((u64)root->sectorsize - 1);
4459
		em->orig_start = EXTENT_MAP_INLINE;
C
Chris Mason 已提交
4460 4461
		if (compressed)
			set_bit(EXTENT_FLAG_COMPRESSED, &em->flags);
4462
		ptr = btrfs_file_extent_inline_start(item) + extent_offset;
4463
		if (create == 0 && !PageUptodate(page)) {
C
Chris Mason 已提交
4464 4465 4466 4467 4468 4469 4470 4471 4472 4473
			if (btrfs_file_extent_compression(leaf, item) ==
			    BTRFS_COMPRESS_ZLIB) {
				ret = uncompress_inline(path, inode, page,
							pg_offset,
							extent_offset, item);
				BUG_ON(ret);
			} else {
				map = kmap(page);
				read_extent_buffer(leaf, map + pg_offset, ptr,
						   copy_size);
4474 4475 4476 4477 4478
				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
4479 4480
				kunmap(page);
			}
4481 4482 4483 4484 4485 4486 4487
			flush_dcache_page(page);
		} else if (create && PageUptodate(page)) {
			if (!trans) {
				kunmap(page);
				free_extent_map(em);
				em = NULL;
				btrfs_release_path(root, path);
4488
				trans = btrfs_join_transaction(root, 1);
4489 4490
				goto again;
			}
C
Chris Mason 已提交
4491
			map = kmap(page);
4492
			write_extent_buffer(leaf, map + pg_offset, ptr,
4493
					    copy_size);
C
Chris Mason 已提交
4494
			kunmap(page);
4495
			btrfs_mark_buffer_dirty(leaf);
4496
		}
4497 4498
		set_extent_uptodate(io_tree, em->start,
				    extent_map_end(em) - 1, GFP_NOFS);
4499 4500
		goto insert;
	} else {
C
Chris Mason 已提交
4501
		printk(KERN_ERR "btrfs unknown found_type %d\n", found_type);
4502 4503 4504 4505
		WARN_ON(1);
	}
not_found:
	em->start = start;
4506
	em->len = len;
4507
not_found_em:
4508
	em->block_start = EXTENT_MAP_HOLE;
Y
Yan Zheng 已提交
4509
	set_bit(EXTENT_FLAG_VACANCY, &em->flags);
4510 4511
insert:
	btrfs_release_path(root, path);
4512
	if (em->start > start || extent_map_end(em) <= start) {
C
Chris Mason 已提交
4513 4514 4515 4516 4517
		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);
4518 4519 4520
		err = -EIO;
		goto out;
	}
4521 4522

	err = 0;
4523
	write_lock(&em_tree->lock);
4524
	ret = add_extent_mapping(em_tree, em);
4525 4526 4527 4528
	/* 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
	 */
4529
	if (ret == -EEXIST) {
4530
		struct extent_map *existing;
4531 4532 4533

		ret = 0;

4534
		existing = lookup_extent_mapping(em_tree, start, len);
4535 4536 4537 4538 4539
		if (existing && (existing->start > start ||
		    existing->start + existing->len <= start)) {
			free_extent_map(existing);
			existing = NULL;
		}
4540 4541 4542 4543 4544
		if (!existing) {
			existing = lookup_extent_mapping(em_tree, em->start,
							 em->len);
			if (existing) {
				err = merge_extent_mapping(em_tree, existing,
4545 4546
							   em, start,
							   root->sectorsize);
4547 4548 4549 4550 4551 4552 4553 4554 4555 4556 4557 4558 4559
				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;
4560
			err = 0;
4561 4562
		}
	}
4563
	write_unlock(&em_tree->lock);
4564
out:
4565 4566
	if (path)
		btrfs_free_path(path);
4567 4568
	if (trans) {
		ret = btrfs_end_transaction(trans, root);
C
Chris Mason 已提交
4569
		if (!err)
4570 4571 4572 4573 4574 4575 4576 4577 4578
			err = ret;
	}
	if (err) {
		free_extent_map(em);
		return ERR_PTR(err);
	}
	return em;
}

4579 4580 4581 4582
static ssize_t btrfs_direct_IO(int rw, struct kiocb *iocb,
			const struct iovec *iov, loff_t offset,
			unsigned long nr_segs)
{
4583
	return -EINVAL;
4584 4585
}

Y
Yehuda Sadeh 已提交
4586 4587 4588 4589 4590 4591
static int btrfs_fiemap(struct inode *inode, struct fiemap_extent_info *fieinfo,
		__u64 start, __u64 len)
{
	return extent_fiemap(inode, fieinfo, start, len, btrfs_get_extent);
}

4592
int btrfs_readpage(struct file *file, struct page *page)
C
Chris Mason 已提交
4593
{
4594 4595
	struct extent_io_tree *tree;
	tree = &BTRFS_I(page->mapping->host)->io_tree;
4596
	return extent_read_full_page(tree, page, btrfs_get_extent);
C
Chris Mason 已提交
4597
}
4598

4599
static int btrfs_writepage(struct page *page, struct writeback_control *wbc)
C
Chris Mason 已提交
4600
{
4601
	struct extent_io_tree *tree;
4602 4603 4604 4605 4606 4607 4608


	if (current->flags & PF_MEMALLOC) {
		redirty_page_for_writepage(wbc, page);
		unlock_page(page);
		return 0;
	}
4609
	tree = &BTRFS_I(page->mapping->host)->io_tree;
4610
	return extent_write_full_page(tree, page, btrfs_get_extent, wbc);
C
Chris Mason 已提交
4611 4612
}

4613 4614
int btrfs_writepages(struct address_space *mapping,
		     struct writeback_control *wbc)
C
Chris Mason 已提交
4615
{
4616
	struct extent_io_tree *tree;
4617

4618
	tree = &BTRFS_I(mapping->host)->io_tree;
C
Chris Mason 已提交
4619 4620 4621
	return extent_writepages(tree, mapping, btrfs_get_extent, wbc);
}

C
Chris Mason 已提交
4622 4623 4624 4625
static int
btrfs_readpages(struct file *file, struct address_space *mapping,
		struct list_head *pages, unsigned nr_pages)
{
4626 4627
	struct extent_io_tree *tree;
	tree = &BTRFS_I(mapping->host)->io_tree;
C
Chris Mason 已提交
4628 4629 4630
	return extent_readpages(tree, mapping, pages, nr_pages,
				btrfs_get_extent);
}
4631
static int __btrfs_releasepage(struct page *page, gfp_t gfp_flags)
C
Chris Mason 已提交
4632
{
4633 4634
	struct extent_io_tree *tree;
	struct extent_map_tree *map;
4635
	int ret;
4636

4637 4638
	tree = &BTRFS_I(page->mapping->host)->io_tree;
	map = &BTRFS_I(page->mapping->host)->extent_tree;
4639
	ret = try_release_extent_mapping(map, tree, page, gfp_flags);
4640 4641 4642 4643
	if (ret == 1) {
		ClearPagePrivate(page);
		set_page_private(page, 0);
		page_cache_release(page);
C
Chris Mason 已提交
4644
	}
4645
	return ret;
C
Chris Mason 已提交
4646 4647
}

4648 4649
static int btrfs_releasepage(struct page *page, gfp_t gfp_flags)
{
4650 4651
	if (PageWriteback(page) || PageDirty(page))
		return 0;
4652
	return __btrfs_releasepage(page, gfp_flags & GFP_NOFS);
4653 4654
}

4655
static void btrfs_invalidatepage(struct page *page, unsigned long offset)
C
Chris Mason 已提交
4656
{
4657
	struct extent_io_tree *tree;
4658 4659 4660
	struct btrfs_ordered_extent *ordered;
	u64 page_start = page_offset(page);
	u64 page_end = page_start + PAGE_CACHE_SIZE - 1;
C
Chris Mason 已提交
4661

4662 4663 4664 4665 4666 4667 4668 4669

	/*
	 * 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
	 */
4670
	wait_on_page_writeback(page);
4671

4672
	tree = &BTRFS_I(page->mapping->host)->io_tree;
4673 4674 4675 4676 4677 4678 4679 4680
	if (offset) {
		btrfs_releasepage(page, GFP_NOFS);
		return;
	}
	lock_extent(tree, page_start, page_end, GFP_NOFS);
	ordered = btrfs_lookup_ordered_extent(page->mapping->host,
					   page_offset(page));
	if (ordered) {
4681 4682 4683 4684
		/*
		 * IO on this page will never be started, so we need
		 * to account for any ordered extents now
		 */
4685 4686
		clear_extent_bit(tree, page_start, page_end,
				 EXTENT_DIRTY | EXTENT_DELALLOC |
4687
				 EXTENT_LOCKED, 1, 0, NULL, GFP_NOFS);
4688 4689 4690 4691 4692 4693 4694 4695
		/*
		 * whoever cleared the private bit is responsible
		 * for the finish_ordered_io
		 */
		if (TestClearPagePrivate2(page)) {
			btrfs_finish_ordered_io(page->mapping->host,
						page_start, page_end);
		}
4696 4697 4698 4699
		btrfs_put_ordered_extent(ordered);
		lock_extent(tree, page_start, page_end, GFP_NOFS);
	}
	clear_extent_bit(tree, page_start, page_end,
4700
		 EXTENT_LOCKED | EXTENT_DIRTY | EXTENT_DELALLOC,
4701
		 1, 1, NULL, GFP_NOFS);
4702 4703
	__btrfs_releasepage(page, GFP_NOFS);

C
Chris Mason 已提交
4704
	ClearPageChecked(page);
4705 4706 4707 4708 4709
	if (PagePrivate(page)) {
		ClearPagePrivate(page);
		set_page_private(page, 0);
		page_cache_release(page);
	}
C
Chris Mason 已提交
4710 4711
}

C
Chris Mason 已提交
4712 4713 4714 4715 4716 4717 4718 4719 4720 4721 4722 4723 4724 4725 4726
/*
 * 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.
 */
4727
int btrfs_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf)
C
Chris Mason 已提交
4728
{
4729
	struct page *page = vmf->page;
4730
	struct inode *inode = fdentry(vma->vm_file)->d_inode;
4731
	struct btrfs_root *root = BTRFS_I(inode)->root;
4732 4733 4734 4735
	struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
	struct btrfs_ordered_extent *ordered;
	char *kaddr;
	unsigned long zero_start;
C
Chris Mason 已提交
4736
	loff_t size;
4737
	int ret;
4738
	u64 page_start;
4739
	u64 page_end;
C
Chris Mason 已提交
4740

J
Josef Bacik 已提交
4741
	ret = btrfs_check_data_free_space(root, inode, PAGE_CACHE_SIZE);
4742 4743 4744 4745 4746
	if (ret) {
		if (ret == -ENOMEM)
			ret = VM_FAULT_OOM;
		else /* -ENOSPC, -EIO, etc */
			ret = VM_FAULT_SIGBUS;
4747
		goto out;
4748
	}
4749

4750
	ret = VM_FAULT_NOPAGE; /* make the VM retry the fault */
4751
again:
C
Chris Mason 已提交
4752 4753
	lock_page(page);
	size = i_size_read(inode);
4754 4755
	page_start = page_offset(page);
	page_end = page_start + PAGE_CACHE_SIZE - 1;
4756

C
Chris Mason 已提交
4757
	if ((page->mapping != inode->i_mapping) ||
4758
	    (page_start >= size)) {
J
Josef Bacik 已提交
4759
		btrfs_free_reserved_data_space(root, inode, PAGE_CACHE_SIZE);
C
Chris Mason 已提交
4760 4761 4762
		/* page got truncated out from underneath us */
		goto out_unlock;
	}
4763 4764 4765 4766 4767
	wait_on_page_writeback(page);

	lock_extent(io_tree, page_start, page_end, GFP_NOFS);
	set_page_extent_mapped(page);

4768 4769 4770 4771
	/*
	 * we can't set the delalloc bits if there are pending ordered
	 * extents.  Drop our locks and wait for them to finish
	 */
4772 4773 4774 4775
	ordered = btrfs_lookup_ordered_extent(inode, page_start);
	if (ordered) {
		unlock_extent(io_tree, page_start, page_end, GFP_NOFS);
		unlock_page(page);
4776
		btrfs_start_ordered_extent(inode, ordered, 1);
4777 4778 4779 4780
		btrfs_put_ordered_extent(ordered);
		goto again;
	}

4781
	btrfs_set_extent_delalloc(inode, page_start, page_end);
4782
	ret = 0;
C
Chris Mason 已提交
4783 4784

	/* page is wholly or partially inside EOF */
4785
	if (page_start + PAGE_CACHE_SIZE > size)
4786
		zero_start = size & ~PAGE_CACHE_MASK;
C
Chris Mason 已提交
4787
	else
4788
		zero_start = PAGE_CACHE_SIZE;
C
Chris Mason 已提交
4789

4790 4791 4792 4793 4794 4795
	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);
	}
4796
	ClearPageChecked(page);
4797
	set_page_dirty(page);
4798
	SetPageUptodate(page);
4799 4800

	BTRFS_I(inode)->last_trans = root->fs_info->generation + 1;
4801
	unlock_extent(io_tree, page_start, page_end, GFP_NOFS);
C
Chris Mason 已提交
4802 4803

out_unlock:
4804 4805
	if (!ret)
		return VM_FAULT_LOCKED;
C
Chris Mason 已提交
4806
	unlock_page(page);
4807
out:
C
Chris Mason 已提交
4808 4809 4810
	return ret;
}

C
Chris Mason 已提交
4811 4812 4813 4814 4815
static void btrfs_truncate(struct inode *inode)
{
	struct btrfs_root *root = BTRFS_I(inode)->root;
	int ret;
	struct btrfs_trans_handle *trans;
4816
	unsigned long nr;
4817
	u64 mask = root->sectorsize - 1;
C
Chris Mason 已提交
4818 4819 4820 4821 4822 4823 4824

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

	btrfs_truncate_page(inode->i_mapping, inode->i_size);
C
Chris Mason 已提交
4825
	btrfs_wait_ordered_range(inode, inode->i_size & (~mask), (u64)-1);
C
Chris Mason 已提交
4826 4827

	trans = btrfs_start_transaction(root, 1);
4828 4829 4830 4831 4832 4833 4834 4835 4836 4837 4838 4839 4840 4841 4842 4843 4844 4845 4846 4847 4848

	/*
	 * 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.
	 */
	if (inode->i_size == 0 && BTRFS_I(inode)->ordered_data_close)
		btrfs_add_ordered_operation(trans, root, inode);

C
Chris Mason 已提交
4849
	btrfs_set_trans_block_group(trans, inode);
4850
	btrfs_i_size_write(inode, inode->i_size);
C
Chris Mason 已提交
4851

4852 4853 4854
	ret = btrfs_orphan_add(trans, inode);
	if (ret)
		goto out;
C
Chris Mason 已提交
4855
	/* FIXME, add redo link to tree so we don't leak on crash */
4856
	ret = btrfs_truncate_inode_items(trans, root, inode, inode->i_size,
4857
				      BTRFS_EXTENT_DATA_KEY);
C
Chris Mason 已提交
4858
	btrfs_update_inode(trans, root, inode);
4859

4860 4861 4862 4863 4864
	ret = btrfs_orphan_del(trans, inode);
	BUG_ON(ret);

out:
	nr = trans->blocks_used;
4865
	ret = btrfs_end_transaction_throttle(trans, root);
C
Chris Mason 已提交
4866
	BUG_ON(ret);
4867
	btrfs_btree_balance_dirty(root, nr);
C
Chris Mason 已提交
4868 4869
}

C
Chris Mason 已提交
4870 4871 4872
/*
 * create a new subvolume directory/inode (helper for the ioctl).
 */
4873
int btrfs_create_subvol_root(struct btrfs_trans_handle *trans,
4874
			     struct btrfs_root *new_root,
4875
			     u64 new_dirid, u64 alloc_hint)
C
Chris Mason 已提交
4876 4877
{
	struct inode *inode;
4878
	int err;
4879
	u64 index = 0;
C
Chris Mason 已提交
4880

4881
	inode = btrfs_new_inode(trans, new_root, NULL, "..", 2, new_dirid,
4882
				new_dirid, alloc_hint, S_IFDIR | 0700, &index);
4883
	if (IS_ERR(inode))
C
Christoph Hellwig 已提交
4884
		return PTR_ERR(inode);
C
Chris Mason 已提交
4885 4886 4887 4888
	inode->i_op = &btrfs_dir_inode_operations;
	inode->i_fop = &btrfs_dir_file_operations;

	inode->i_nlink = 1;
4889
	btrfs_i_size_write(inode, 0);
4890

4891 4892
	err = btrfs_update_inode(trans, new_root, inode);
	BUG_ON(err);
4893

4894
	iput(inode);
4895
	return 0;
C
Chris Mason 已提交
4896 4897
}

C
Chris Mason 已提交
4898 4899 4900
/* helper function for file defrag and space balancing.  This
 * forces readahead on a given range of bytes in an inode
 */
4901
unsigned long btrfs_force_ra(struct address_space *mapping,
4902 4903 4904
			      struct file_ra_state *ra, struct file *file,
			      pgoff_t offset, pgoff_t last_index)
{
4905
	pgoff_t req_size = last_index - offset + 1;
4906 4907 4908 4909 4910

	page_cache_sync_readahead(mapping, ra, file, offset, req_size);
	return offset + req_size;
}

C
Chris Mason 已提交
4911 4912 4913 4914 4915 4916 4917
struct inode *btrfs_alloc_inode(struct super_block *sb)
{
	struct btrfs_inode *ei;

	ei = kmem_cache_alloc(btrfs_inode_cachep, GFP_NOFS);
	if (!ei)
		return NULL;
4918
	ei->last_trans = 0;
4919
	ei->logged_trans = 0;
4920
	btrfs_ordered_inode_tree_init(&ei->ordered_tree);
4921
	INIT_LIST_HEAD(&ei->i_orphan);
4922
	INIT_LIST_HEAD(&ei->ordered_operations);
C
Chris Mason 已提交
4923 4924 4925 4926 4927
	return &ei->vfs_inode;
}

void btrfs_destroy_inode(struct inode *inode)
{
4928
	struct btrfs_ordered_extent *ordered;
4929 4930
	struct btrfs_root *root = BTRFS_I(inode)->root;

C
Chris Mason 已提交
4931 4932 4933
	WARN_ON(!list_empty(&inode->i_dentry));
	WARN_ON(inode->i_data.nrpages);

4934 4935 4936 4937 4938 4939 4940 4941 4942 4943 4944 4945
	/*
	 * 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);
	}

	spin_lock(&root->list_lock);
4946 4947 4948 4949 4950
	if (!list_empty(&BTRFS_I(inode)->i_orphan)) {
		printk(KERN_ERR "BTRFS: inode %lu: inode still on the orphan"
		       " list\n", inode->i_ino);
		dump_stack();
	}
4951
	spin_unlock(&root->list_lock);
4952

C
Chris Mason 已提交
4953
	while (1) {
4954 4955 4956 4957
		ordered = btrfs_lookup_first_ordered_extent(inode, (u64)-1);
		if (!ordered)
			break;
		else {
C
Chris Mason 已提交
4958 4959 4960 4961
			printk(KERN_ERR "btrfs found ordered "
			       "extent %llu %llu on inode cleanup\n",
			       (unsigned long long)ordered->file_offset,
			       (unsigned long long)ordered->len);
4962 4963 4964 4965 4966
			btrfs_remove_ordered_extent(inode, ordered);
			btrfs_put_ordered_extent(ordered);
			btrfs_put_ordered_extent(ordered);
		}
	}
4967
	inode_tree_del(inode);
4968
	btrfs_drop_extent_cache(inode, 0, (u64)-1, 0);
C
Chris Mason 已提交
4969 4970 4971
	kmem_cache_free(btrfs_inode_cachep, BTRFS_I(inode));
}

4972 4973 4974 4975 4976 4977 4978 4979 4980 4981
void btrfs_drop_inode(struct inode *inode)
{
	struct btrfs_root *root = BTRFS_I(inode)->root;

	if (inode->i_nlink > 0 && btrfs_root_refs(&root->root_item) == 0)
		generic_delete_inode(inode);
	else
		generic_drop_inode(inode);
}

4982
static void init_once(void *foo)
C
Chris Mason 已提交
4983 4984 4985 4986 4987 4988 4989 4990 4991 4992 4993 4994 4995 4996 4997 4998 4999 5000 5001 5002
{
	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);
}

int btrfs_init_cachep(void)
{
5003 5004 5005
	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
5006 5007
	if (!btrfs_inode_cachep)
		goto fail;
5008 5009 5010 5011

	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
5012 5013
	if (!btrfs_trans_handle_cachep)
		goto fail;
5014 5015 5016 5017

	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
5018 5019
	if (!btrfs_transaction_cachep)
		goto fail;
5020 5021 5022 5023

	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
5024 5025
	if (!btrfs_path_cachep)
		goto fail;
5026

C
Chris Mason 已提交
5027 5028 5029 5030 5031 5032 5033 5034 5035 5036 5037
	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;
	generic_fillattr(inode, stat);
5038
	stat->dev = BTRFS_I(inode)->root->anon_super.s_dev;
C
Chris Mason 已提交
5039
	stat->blksize = PAGE_CACHE_SIZE;
5040 5041
	stat->blocks = (inode_get_bytes(inode) +
			BTRFS_I(inode)->delalloc_bytes) >> 9;
C
Chris Mason 已提交
5042 5043 5044
	return 0;
}

C
Chris Mason 已提交
5045 5046
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
5047 5048 5049
{
	struct btrfs_trans_handle *trans;
	struct btrfs_root *root = BTRFS_I(old_dir)->root;
5050
	struct btrfs_root *dest = BTRFS_I(new_dir)->root;
C
Chris Mason 已提交
5051 5052 5053
	struct inode *new_inode = new_dentry->d_inode;
	struct inode *old_inode = old_dentry->d_inode;
	struct timespec ctime = CURRENT_TIME;
5054
	u64 index = 0;
5055
	u64 root_objectid;
C
Chris Mason 已提交
5056 5057
	int ret;

5058 5059
	/* we only allow rename subvolume link between subvolumes */
	if (old_inode->i_ino != BTRFS_FIRST_FREE_OBJECTID && root != dest)
5060 5061
		return -EXDEV;

5062 5063
	if (old_inode->i_ino == BTRFS_EMPTY_SUBVOL_DIR_OBJECTID ||
	    (new_inode && new_inode->i_ino == BTRFS_FIRST_FREE_OBJECTID))
C
Chris Mason 已提交
5064
		return -ENOTEMPTY;
5065

5066 5067 5068
	if (S_ISDIR(old_inode->i_mode) && new_inode &&
	    new_inode->i_size > BTRFS_EMPTY_DIR_SIZE)
		return -ENOTEMPTY;
5069

J
Josef Bacik 已提交
5070
	ret = btrfs_check_metadata_free_space(root);
5071
	if (ret)
5072
		return ret;
5073

5074 5075 5076 5077 5078
	/*
	 * 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
	 */
5079
	if (new_inode && S_ISREG(old_inode->i_mode) && new_inode->i_size &&
5080 5081 5082
	    old_inode->i_size > BTRFS_ORDERED_OPERATIONS_FLUSH_LIMIT)
		filemap_flush(old_inode->i_mapping);

5083 5084 5085 5086
	/* close the racy window with snapshot create/destroy ioctl */
	if (old_inode->i_ino == BTRFS_FIRST_FREE_OBJECTID)
		down_read(&root->fs_info->subvol_sem);

C
Chris Mason 已提交
5087
	trans = btrfs_start_transaction(root, 1);
5088
	btrfs_set_trans_block_group(trans, new_dir);
5089

5090 5091 5092
	if (dest != root)
		btrfs_record_root_in_trans(trans, dest);

5093 5094 5095
	ret = btrfs_set_inode_index(new_dir, &index);
	if (ret)
		goto out_fail;
5096

5097
	if (unlikely(old_inode->i_ino == BTRFS_FIRST_FREE_OBJECTID)) {
5098 5099 5100
		/* force full log commit if subvolume involved. */
		root->fs_info->last_trans_log_full_commit = trans->transid;
	} else {
5101 5102 5103 5104 5105 5106 5107
		ret = btrfs_insert_inode_ref(trans, dest,
					     new_dentry->d_name.name,
					     new_dentry->d_name.len,
					     old_inode->i_ino,
					     new_dir->i_ino, index);
		if (ret)
			goto out_fail;
5108 5109 5110 5111 5112 5113 5114 5115 5116
		/*
		 * 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);
	}
5117 5118 5119 5120 5121 5122 5123 5124
	/*
	 * make sure the inode gets flushed if it is replacing
	 * something.
	 */
	if (new_inode && new_inode->i_size &&
	    old_inode && S_ISREG(old_inode->i_mode)) {
		btrfs_add_ordered_operation(trans, root, old_inode);
	}
C
Chris Mason 已提交
5125 5126 5127 5128

	old_dir->i_ctime = old_dir->i_mtime = ctime;
	new_dir->i_ctime = new_dir->i_mtime = ctime;
	old_inode->i_ctime = ctime;
5129

5130 5131 5132
	if (old_dentry->d_parent != new_dentry->d_parent)
		btrfs_record_unlink_dir(trans, old_dir, old_inode, 1);

5133 5134 5135 5136 5137 5138 5139 5140 5141 5142 5143 5144 5145
	if (unlikely(old_inode->i_ino == BTRFS_FIRST_FREE_OBJECTID)) {
		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 {
		btrfs_inc_nlink(old_dentry->d_inode);
		ret = btrfs_unlink_inode(trans, root, old_dir,
					 old_dentry->d_inode,
					 old_dentry->d_name.name,
					 old_dentry->d_name.len);
	}
	BUG_ON(ret);
C
Chris Mason 已提交
5146 5147 5148

	if (new_inode) {
		new_inode->i_ctime = CURRENT_TIME;
5149 5150 5151 5152 5153 5154 5155 5156 5157 5158 5159 5160 5161 5162 5163
		if (unlikely(new_inode->i_ino ==
			     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);
		}
		BUG_ON(ret);
5164
		if (new_inode->i_nlink == 0) {
5165
			ret = btrfs_orphan_add(trans, new_dentry->d_inode);
5166
			BUG_ON(ret);
5167
		}
C
Chris Mason 已提交
5168
	}
5169

5170 5171
	ret = btrfs_add_link(trans, new_dir, old_inode,
			     new_dentry->d_name.name,
5172
			     new_dentry->d_name.len, 0, index);
5173
	BUG_ON(ret);
C
Chris Mason 已提交
5174

5175 5176 5177 5178 5179
	if (old_inode->i_ino != BTRFS_FIRST_FREE_OBJECTID) {
		btrfs_log_new_name(trans, old_inode, old_dir,
				   new_dentry->d_parent);
		btrfs_end_log_trans(root);
	}
5180
out_fail:
5181
	btrfs_end_transaction_throttle(trans, root);
5182

5183 5184
	if (old_inode->i_ino == BTRFS_FIRST_FREE_OBJECTID)
		up_read(&root->fs_info->subvol_sem);
C
Chris Mason 已提交
5185 5186 5187
	return ret;
}

C
Chris Mason 已提交
5188 5189 5190 5191
/*
 * some fairly slow code that needs optimization. This walks the list
 * of all the inodes with pending delalloc and forces them to disk.
 */
5192 5193 5194 5195
int btrfs_start_delalloc_inodes(struct btrfs_root *root)
{
	struct list_head *head = &root->fs_info->delalloc_inodes;
	struct btrfs_inode *binode;
5196
	struct inode *inode;
5197

Y
Yan Zheng 已提交
5198 5199 5200
	if (root->fs_info->sb->s_flags & MS_RDONLY)
		return -EROFS;

5201
	spin_lock(&root->fs_info->delalloc_lock);
C
Chris Mason 已提交
5202
	while (!list_empty(head)) {
5203 5204
		binode = list_entry(head->next, struct btrfs_inode,
				    delalloc_inodes);
5205 5206 5207
		inode = igrab(&binode->vfs_inode);
		if (!inode)
			list_del_init(&binode->delalloc_inodes);
5208
		spin_unlock(&root->fs_info->delalloc_lock);
5209
		if (inode) {
5210
			filemap_flush(inode->i_mapping);
5211 5212 5213
			iput(inode);
		}
		cond_resched();
5214
		spin_lock(&root->fs_info->delalloc_lock);
5215
	}
5216
	spin_unlock(&root->fs_info->delalloc_lock);
5217 5218 5219 5220 5221 5222

	/* 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
5223
	while (atomic_read(&root->fs_info->nr_async_submits) ||
5224
	      atomic_read(&root->fs_info->async_delalloc_pages)) {
5225
		wait_event(root->fs_info->async_submit_wait,
5226 5227
		   (atomic_read(&root->fs_info->nr_async_submits) == 0 &&
		    atomic_read(&root->fs_info->async_delalloc_pages) == 0));
5228 5229
	}
	atomic_dec(&root->fs_info->async_submit_draining);
5230 5231 5232
	return 0;
}

C
Chris Mason 已提交
5233 5234 5235 5236 5237 5238 5239
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;
5240
	struct inode *inode = NULL;
C
Chris Mason 已提交
5241 5242 5243
	int err;
	int drop_inode = 0;
	u64 objectid;
5244
	u64 index = 0 ;
C
Chris Mason 已提交
5245 5246
	int name_len;
	int datasize;
5247
	unsigned long ptr;
C
Chris Mason 已提交
5248
	struct btrfs_file_extent_item *ei;
5249
	struct extent_buffer *leaf;
5250
	unsigned long nr = 0;
C
Chris Mason 已提交
5251 5252 5253 5254

	name_len = strlen(symname) + 1;
	if (name_len > BTRFS_MAX_INLINE_DATA_SIZE(root))
		return -ENAMETOOLONG;
5255

J
Josef Bacik 已提交
5256
	err = btrfs_check_metadata_free_space(root);
5257 5258 5259
	if (err)
		goto out_fail;

C
Chris Mason 已提交
5260 5261 5262 5263 5264 5265 5266 5267 5268
	trans = btrfs_start_transaction(root, 1);
	btrfs_set_trans_block_group(trans, dir);

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

5269
	inode = btrfs_new_inode(trans, root, dir, dentry->d_name.name,
5270 5271
				dentry->d_name.len,
				dentry->d_parent->d_inode->i_ino, objectid,
5272 5273
				BTRFS_I(dir)->block_group, S_IFLNK|S_IRWXUGO,
				&index);
C
Chris Mason 已提交
5274 5275 5276 5277
	err = PTR_ERR(inode);
	if (IS_ERR(inode))
		goto out_unlock;

J
Jim Owens 已提交
5278
	err = btrfs_init_inode_security(inode, dir);
J
Josef Bacik 已提交
5279 5280 5281 5282 5283
	if (err) {
		drop_inode = 1;
		goto out_unlock;
	}

C
Chris Mason 已提交
5284
	btrfs_set_trans_block_group(trans, inode);
5285
	err = btrfs_add_nondir(trans, dentry, inode, 0, index);
C
Chris Mason 已提交
5286 5287 5288 5289
	if (err)
		drop_inode = 1;
	else {
		inode->i_mapping->a_ops = &btrfs_aops;
C
Chris Mason 已提交
5290
		inode->i_mapping->backing_dev_info = &root->fs_info->bdi;
C
Chris Mason 已提交
5291 5292
		inode->i_fop = &btrfs_file_operations;
		inode->i_op = &btrfs_file_inode_operations;
5293
		BTRFS_I(inode)->io_tree.ops = &btrfs_extent_io_ops;
C
Chris Mason 已提交
5294 5295 5296 5297 5298 5299 5300 5301 5302 5303 5304 5305 5306 5307
	}
	btrfs_update_inode_block_group(trans, inode);
	btrfs_update_inode_block_group(trans, dir);
	if (drop_inode)
		goto out_unlock;

	path = btrfs_alloc_path();
	BUG_ON(!path);
	key.objectid = inode->i_ino;
	key.offset = 0;
	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);
5308 5309 5310 5311
	if (err) {
		drop_inode = 1;
		goto out_unlock;
	}
5312 5313 5314 5315 5316
	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
5317
				   BTRFS_FILE_EXTENT_INLINE);
C
Chris Mason 已提交
5318 5319 5320 5321 5322
	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
5323
	ptr = btrfs_file_extent_inline_start(ei);
5324 5325
	write_extent_buffer(leaf, symname, ptr, name_len);
	btrfs_mark_buffer_dirty(leaf);
C
Chris Mason 已提交
5326
	btrfs_free_path(path);
5327

C
Chris Mason 已提交
5328 5329
	inode->i_op = &btrfs_symlink_inode_operations;
	inode->i_mapping->a_ops = &btrfs_symlink_aops;
C
Chris Mason 已提交
5330
	inode->i_mapping->backing_dev_info = &root->fs_info->bdi;
Y
Yan Zheng 已提交
5331
	inode_set_bytes(inode, name_len);
5332
	btrfs_i_size_write(inode, name_len - 1);
5333 5334 5335
	err = btrfs_update_inode(trans, root, inode);
	if (err)
		drop_inode = 1;
C
Chris Mason 已提交
5336 5337

out_unlock:
5338
	nr = trans->blocks_used;
5339
	btrfs_end_transaction_throttle(trans, root);
5340
out_fail:
C
Chris Mason 已提交
5341 5342 5343 5344
	if (drop_inode) {
		inode_dec_link_count(inode);
		iput(inode);
	}
5345
	btrfs_btree_balance_dirty(root, nr);
C
Chris Mason 已提交
5346 5347
	return err;
}
5348

5349 5350
static int prealloc_file_range(struct btrfs_trans_handle *trans,
			       struct inode *inode, u64 start, u64 end,
5351
			       u64 locked_end, u64 alloc_hint, int mode)
Y
Yan Zheng 已提交
5352 5353 5354 5355 5356 5357 5358 5359 5360 5361 5362 5363 5364 5365 5366 5367 5368 5369 5370 5371
{
	struct btrfs_root *root = BTRFS_I(inode)->root;
	struct btrfs_key ins;
	u64 alloc_size;
	u64 cur_offset = start;
	u64 num_bytes = end - start;
	int ret = 0;

	while (num_bytes > 0) {
		alloc_size = min(num_bytes, root->fs_info->max_extent);
		ret = btrfs_reserve_extent(trans, root, alloc_size,
					   root->sectorsize, 0, alloc_hint,
					   (u64)-1, &ins, 1);
		if (ret) {
			WARN_ON(1);
			goto out;
		}
		ret = insert_reserved_file_extent(trans, inode,
						  cur_offset, ins.objectid,
						  ins.offset, ins.offset,
5372 5373
						  ins.offset, locked_end,
						  0, 0, 0,
Y
Yan Zheng 已提交
5374 5375
						  BTRFS_FILE_EXTENT_PREALLOC);
		BUG_ON(ret);
C
Chris Mason 已提交
5376 5377
		btrfs_drop_extent_cache(inode, cur_offset,
					cur_offset + ins.offset -1, 0);
Y
Yan Zheng 已提交
5378 5379 5380 5381 5382 5383 5384
		num_bytes -= ins.offset;
		cur_offset += ins.offset;
		alloc_hint = ins.objectid + ins.offset;
	}
out:
	if (cur_offset > start) {
		inode->i_ctime = CURRENT_TIME;
5385
		BTRFS_I(inode)->flags |= BTRFS_INODE_PREALLOC;
Y
Yan Zheng 已提交
5386 5387 5388 5389 5390 5391 5392 5393 5394 5395 5396 5397 5398 5399 5400 5401 5402 5403
		if (!(mode & FALLOC_FL_KEEP_SIZE) &&
		    cur_offset > i_size_read(inode))
			btrfs_i_size_write(inode, cur_offset);
		ret = btrfs_update_inode(trans, root, inode);
		BUG_ON(ret);
	}

	return ret;
}

static long btrfs_fallocate(struct inode *inode, int mode,
			    loff_t offset, loff_t len)
{
	u64 cur_offset;
	u64 last_byte;
	u64 alloc_start;
	u64 alloc_end;
	u64 alloc_hint = 0;
5404
	u64 locked_end;
Y
Yan Zheng 已提交
5405 5406
	u64 mask = BTRFS_I(inode)->root->sectorsize - 1;
	struct extent_map *em;
5407
	struct btrfs_trans_handle *trans;
5408
	struct btrfs_root *root;
Y
Yan Zheng 已提交
5409 5410 5411 5412 5413
	int ret;

	alloc_start = offset & ~mask;
	alloc_end =  (offset + len + mask) & ~mask;

5414 5415 5416 5417 5418 5419
	/*
	 * wait for ordered IO before we have any locks.  We'll loop again
	 * below with the locks held.
	 */
	btrfs_wait_ordered_range(inode, alloc_start, alloc_end - alloc_start);

Y
Yan Zheng 已提交
5420 5421 5422 5423 5424 5425 5426
	mutex_lock(&inode->i_mutex);
	if (alloc_start > inode->i_size) {
		ret = btrfs_cont_expand(inode, alloc_start);
		if (ret)
			goto out;
	}

5427 5428 5429 5430 5431 5432 5433
	root = BTRFS_I(inode)->root;

	ret = btrfs_check_data_free_space(root, inode,
					  alloc_end - alloc_start);
	if (ret)
		goto out;

5434
	locked_end = alloc_end - 1;
Y
Yan Zheng 已提交
5435 5436
	while (1) {
		struct btrfs_ordered_extent *ordered;
5437 5438 5439 5440

		trans = btrfs_start_transaction(BTRFS_I(inode)->root, 1);
		if (!trans) {
			ret = -EIO;
5441
			goto out_free;
5442 5443 5444 5445 5446
		}

		/* the extent lock is ordered inside the running
		 * transaction
		 */
5447 5448
		lock_extent(&BTRFS_I(inode)->io_tree, alloc_start, locked_end,
			    GFP_NOFS);
Y
Yan Zheng 已提交
5449 5450 5451 5452 5453 5454 5455
		ordered = btrfs_lookup_first_ordered_extent(inode,
							    alloc_end - 1);
		if (ordered &&
		    ordered->file_offset + ordered->len > alloc_start &&
		    ordered->file_offset < alloc_end) {
			btrfs_put_ordered_extent(ordered);
			unlock_extent(&BTRFS_I(inode)->io_tree,
5456
				      alloc_start, locked_end, GFP_NOFS);
5457 5458 5459 5460 5461 5462
			btrfs_end_transaction(trans, BTRFS_I(inode)->root);

			/*
			 * we can't wait on the range with the transaction
			 * running or with the extent lock held
			 */
Y
Yan Zheng 已提交
5463 5464 5465 5466 5467 5468 5469 5470 5471 5472 5473 5474 5475 5476 5477 5478 5479
			btrfs_wait_ordered_range(inode, alloc_start,
						 alloc_end - alloc_start);
		} else {
			if (ordered)
				btrfs_put_ordered_extent(ordered);
			break;
		}
	}

	cur_offset = alloc_start;
	while (1) {
		em = btrfs_get_extent(inode, NULL, 0, cur_offset,
				      alloc_end - cur_offset, 0);
		BUG_ON(IS_ERR(em) || !em);
		last_byte = min(extent_map_end(em), alloc_end);
		last_byte = (last_byte + mask) & ~mask;
		if (em->block_start == EXTENT_MAP_HOLE) {
5480
			ret = prealloc_file_range(trans, inode, cur_offset,
5481 5482
					last_byte, locked_end + 1,
					alloc_hint, mode);
Y
Yan Zheng 已提交
5483 5484 5485 5486 5487 5488 5489 5490 5491 5492 5493 5494 5495 5496 5497
			if (ret < 0) {
				free_extent_map(em);
				break;
			}
		}
		if (em->block_start <= EXTENT_MAP_LAST_BYTE)
			alloc_hint = em->block_start;
		free_extent_map(em);

		cur_offset = last_byte;
		if (cur_offset >= alloc_end) {
			ret = 0;
			break;
		}
	}
5498
	unlock_extent(&BTRFS_I(inode)->io_tree, alloc_start, locked_end,
Y
Yan Zheng 已提交
5499
		      GFP_NOFS);
5500 5501

	btrfs_end_transaction(trans, BTRFS_I(inode)->root);
5502 5503
out_free:
	btrfs_free_reserved_data_space(root, inode, alloc_end - alloc_start);
Y
Yan Zheng 已提交
5504 5505 5506 5507 5508
out:
	mutex_unlock(&inode->i_mutex);
	return ret;
}

5509 5510 5511 5512 5513
static int btrfs_set_page_dirty(struct page *page)
{
	return __set_page_dirty_nobuffers(page);
}

5514
static int btrfs_permission(struct inode *inode, int mask)
Y
Yan 已提交
5515
{
5516
	if ((BTRFS_I(inode)->flags & BTRFS_INODE_READONLY) && (mask & MAY_WRITE))
Y
Yan 已提交
5517
		return -EACCES;
J
Josef Bacik 已提交
5518
	return generic_permission(inode, mask, btrfs_check_acl);
Y
Yan 已提交
5519
}
C
Chris Mason 已提交
5520 5521

static struct inode_operations btrfs_dir_inode_operations = {
5522
	.getattr	= btrfs_getattr,
C
Chris Mason 已提交
5523 5524 5525 5526 5527 5528 5529 5530 5531
	.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
5532
	.mknod		= btrfs_mknod,
5533 5534
	.setxattr	= btrfs_setxattr,
	.getxattr	= btrfs_getxattr,
J
Josef Bacik 已提交
5535
	.listxattr	= btrfs_listxattr,
5536
	.removexattr	= btrfs_removexattr,
Y
Yan 已提交
5537
	.permission	= btrfs_permission,
C
Chris Mason 已提交
5538 5539 5540
};
static struct inode_operations btrfs_dir_ro_inode_operations = {
	.lookup		= btrfs_lookup,
Y
Yan 已提交
5541
	.permission	= btrfs_permission,
C
Chris Mason 已提交
5542
};
5543

C
Chris Mason 已提交
5544 5545 5546
static struct file_operations btrfs_dir_file_operations = {
	.llseek		= generic_file_llseek,
	.read		= generic_read_dir,
5547
	.readdir	= btrfs_real_readdir,
C
Christoph Hellwig 已提交
5548
	.unlocked_ioctl	= btrfs_ioctl,
C
Chris Mason 已提交
5549
#ifdef CONFIG_COMPAT
C
Christoph Hellwig 已提交
5550
	.compat_ioctl	= btrfs_ioctl,
C
Chris Mason 已提交
5551
#endif
S
Sage Weil 已提交
5552
	.release        = btrfs_release_file,
5553
	.fsync		= btrfs_sync_file,
C
Chris Mason 已提交
5554 5555
};

5556
static struct extent_io_ops btrfs_extent_io_ops = {
5557
	.fill_delalloc = run_delalloc_range,
5558
	.submit_bio_hook = btrfs_submit_bio_hook,
5559
	.merge_bio_hook = btrfs_merge_bio_hook,
5560
	.readpage_end_io_hook = btrfs_readpage_end_io_hook,
5561
	.writepage_end_io_hook = btrfs_writepage_end_io_hook,
5562
	.writepage_start_hook = btrfs_writepage_start_hook,
5563
	.readpage_io_failed_hook = btrfs_io_failed_hook,
C
Chris Mason 已提交
5564 5565
	.set_bit_hook = btrfs_set_bit_hook,
	.clear_bit_hook = btrfs_clear_bit_hook,
5566 5567
};

5568 5569 5570 5571 5572 5573 5574 5575 5576 5577 5578 5579
/*
 * 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.
 */
C
Chris Mason 已提交
5580 5581 5582
static struct address_space_operations btrfs_aops = {
	.readpage	= btrfs_readpage,
	.writepage	= btrfs_writepage,
C
Chris Mason 已提交
5583
	.writepages	= btrfs_writepages,
C
Chris Mason 已提交
5584
	.readpages	= btrfs_readpages,
C
Chris Mason 已提交
5585
	.sync_page	= block_sync_page,
5586
	.direct_IO	= btrfs_direct_IO,
5587 5588
	.invalidatepage = btrfs_invalidatepage,
	.releasepage	= btrfs_releasepage,
5589
	.set_page_dirty	= btrfs_set_page_dirty,
C
Chris Mason 已提交
5590 5591 5592 5593 5594
};

static struct address_space_operations btrfs_symlink_aops = {
	.readpage	= btrfs_readpage,
	.writepage	= btrfs_writepage,
C
Chris Mason 已提交
5595 5596
	.invalidatepage = btrfs_invalidatepage,
	.releasepage	= btrfs_releasepage,
C
Chris Mason 已提交
5597 5598 5599 5600 5601 5602
};

static struct inode_operations btrfs_file_inode_operations = {
	.truncate	= btrfs_truncate,
	.getattr	= btrfs_getattr,
	.setattr	= btrfs_setattr,
5603 5604
	.setxattr	= btrfs_setxattr,
	.getxattr	= btrfs_getxattr,
J
Josef Bacik 已提交
5605
	.listxattr      = btrfs_listxattr,
5606
	.removexattr	= btrfs_removexattr,
Y
Yan 已提交
5607
	.permission	= btrfs_permission,
Y
Yan Zheng 已提交
5608
	.fallocate	= btrfs_fallocate,
Y
Yehuda Sadeh 已提交
5609
	.fiemap		= btrfs_fiemap,
C
Chris Mason 已提交
5610
};
J
Josef Bacik 已提交
5611 5612 5613
static struct inode_operations btrfs_special_inode_operations = {
	.getattr	= btrfs_getattr,
	.setattr	= btrfs_setattr,
Y
Yan 已提交
5614
	.permission	= btrfs_permission,
5615 5616
	.setxattr	= btrfs_setxattr,
	.getxattr	= btrfs_getxattr,
J
Josef Bacik 已提交
5617
	.listxattr	= btrfs_listxattr,
5618
	.removexattr	= btrfs_removexattr,
J
Josef Bacik 已提交
5619
};
C
Chris Mason 已提交
5620 5621 5622 5623
static struct inode_operations btrfs_symlink_inode_operations = {
	.readlink	= generic_readlink,
	.follow_link	= page_follow_link_light,
	.put_link	= page_put_link,
Y
Yan 已提交
5624
	.permission	= btrfs_permission,
J
Jim Owens 已提交
5625 5626 5627 5628
	.setxattr	= btrfs_setxattr,
	.getxattr	= btrfs_getxattr,
	.listxattr	= btrfs_listxattr,
	.removexattr	= btrfs_removexattr,
C
Chris Mason 已提交
5629
};
5630 5631 5632 5633

struct dentry_operations btrfs_dentry_operations = {
	.d_delete	= btrfs_dentry_delete,
};