inode.c 214.7 KB
Newer Older
C
Chris Mason 已提交
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18
/*
 * Copyright (C) 2007 Oracle.  All rights reserved.
 *
 * This program is free software; you can redistribute it and/or
 * modify it under the terms of the GNU General Public
 * License v2 as published by the Free Software Foundation.
 *
 * This program is distributed in the hope that it will be useful,
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
 * General Public License for more details.
 *
 * You should have received a copy of the GNU General Public
 * License along with this program; if not, write to the
 * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
 * Boston, MA 021110-1307, USA.
 */

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

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

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

static struct kmem_cache *btrfs_inode_cachep;
74
static struct kmem_cache *btrfs_delalloc_work_cachep;
C
Chris Mason 已提交
75 76 77
struct kmem_cache *btrfs_trans_handle_cachep;
struct kmem_cache *btrfs_transaction_cachep;
struct kmem_cache *btrfs_path_cachep;
78
struct kmem_cache *btrfs_free_space_cachep;
C
Chris Mason 已提交
79 80 81 82 83 84 85 86 87 88 89 90

#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,
};

91
static int btrfs_setsize(struct inode *inode, struct iattr *attr);
92
static int btrfs_truncate(struct inode *inode);
93
static int btrfs_finish_ordered_io(struct btrfs_ordered_extent *ordered_extent);
94 95 96 97
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);
98 99 100 101
static struct extent_map *create_pinned_em(struct inode *inode, u64 start,
					   u64 len, u64 orig_start,
					   u64 block_start, u64 block_len,
					   u64 orig_block_len, int type);
102

103
static int btrfs_init_inode_security(struct btrfs_trans_handle *trans,
104 105
				     struct inode *inode,  struct inode *dir,
				     const struct qstr *qstr)
J
Jim Owens 已提交
106 107 108
{
	int err;

109
	err = btrfs_init_acl(trans, inode, dir);
J
Jim Owens 已提交
110
	if (!err)
111
		err = btrfs_xattr_security_init(trans, inode, dir, qstr);
J
Jim Owens 已提交
112 113 114
	return err;
}

C
Chris Mason 已提交
115 116 117 118 119
/*
 * 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
120
static noinline int insert_inline_extent(struct btrfs_trans_handle *trans,
C
Chris Mason 已提交
121 122
				struct btrfs_root *root, struct inode *inode,
				u64 start, size_t size, size_t compressed_size,
123
				int compress_type,
C
Chris Mason 已提交
124 125 126 127 128 129 130 131 132 133 134 135 136 137 138
				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;

139
	if (compressed_size && compressed_pages)
C
Chris Mason 已提交
140 141
		cur_size = compressed_size;

C
Chris Mason 已提交
142 143
	path = btrfs_alloc_path();
	if (!path)
C
Chris Mason 已提交
144 145
		return -ENOMEM;

146
	path->leave_spinning = 1;
C
Chris Mason 已提交
147

L
Li Zefan 已提交
148
	key.objectid = btrfs_ino(inode);
C
Chris Mason 已提交
149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169
	key.offset = start;
	btrfs_set_key_type(&key, BTRFS_EXTENT_DATA_KEY);
	datasize = btrfs_file_extent_calc_inline_size(cur_size);

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

170
	if (compress_type != BTRFS_COMPRESS_NONE) {
C
Chris Mason 已提交
171 172
		struct page *cpage;
		int i = 0;
C
Chris Mason 已提交
173
		while (compressed_size > 0) {
C
Chris Mason 已提交
174
			cpage = compressed_pages[i];
175
			cur_size = min_t(unsigned long, compressed_size,
C
Chris Mason 已提交
176 177
				       PAGE_CACHE_SIZE);

178
			kaddr = kmap_atomic(cpage);
C
Chris Mason 已提交
179
			write_extent_buffer(leaf, kaddr, ptr, cur_size);
180
			kunmap_atomic(kaddr);
C
Chris Mason 已提交
181 182 183 184 185 186

			i++;
			ptr += cur_size;
			compressed_size -= cur_size;
		}
		btrfs_set_file_extent_compression(leaf, ei,
187
						  compress_type);
C
Chris Mason 已提交
188 189 190 191
	} else {
		page = find_get_page(inode->i_mapping,
				     start >> PAGE_CACHE_SHIFT);
		btrfs_set_file_extent_compression(leaf, ei, 0);
192
		kaddr = kmap_atomic(page);
C
Chris Mason 已提交
193 194
		offset = start & (PAGE_CACHE_SIZE - 1);
		write_extent_buffer(leaf, kaddr + offset, ptr, size);
195
		kunmap_atomic(kaddr);
C
Chris Mason 已提交
196 197 198 199 200
		page_cache_release(page);
	}
	btrfs_mark_buffer_dirty(leaf);
	btrfs_free_path(path);

201 202 203 204 205 206 207 208 209
	/*
	 * we're an inline extent, so nobody can
	 * extend the file past i_size without locking
	 * a page we already have locked.
	 *
	 * We must do any isize and inode updates
	 * before we unlock the pages.  Otherwise we
	 * could end up racing with unlink.
	 */
C
Chris Mason 已提交
210
	BTRFS_I(inode)->disk_i_size = inode->i_size;
211
	ret = btrfs_update_inode(trans, root, inode);
212

213
	return ret;
C
Chris Mason 已提交
214 215 216 217 218 219 220 221 222 223 224
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.
 */
225
static noinline int cow_file_range_inline(struct btrfs_trans_handle *trans,
C
Chris Mason 已提交
226 227
				 struct btrfs_root *root,
				 struct inode *inode, u64 start, u64 end,
228
				 size_t compressed_size, int compress_type,
C
Chris Mason 已提交
229 230 231 232 233 234 235 236 237 238 239 240 241 242
				 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 data_len = inline_len;
	int ret;

	if (compressed_size)
		data_len = compressed_size;

	if (start > 0 ||
C
Chris Mason 已提交
243
	    actual_end >= PAGE_CACHE_SIZE ||
C
Chris Mason 已提交
244 245 246 247 248 249 250 251
	    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;
	}

252
	ret = btrfs_drop_extents(trans, root, inode, start, aligned_end, 1);
253 254
	if (ret)
		return ret;
C
Chris Mason 已提交
255 256 257 258 259

	if (isize > actual_end)
		inline_len = min_t(u64, isize, actual_end);
	ret = insert_inline_extent(trans, root, inode, start,
				   inline_len, compressed_size,
260
				   compress_type, compressed_pages);
261
	if (ret && ret != -ENOSPC) {
262 263
		btrfs_abort_transaction(trans, root, ret);
		return ret;
264 265
	} else if (ret == -ENOSPC) {
		return 1;
266
	}
267

268
	btrfs_delalloc_release_metadata(inode, end + 1 - start);
C
Chris Mason 已提交
269
	btrfs_drop_extent_cache(inode, start, aligned_end - 1, 0);
C
Chris Mason 已提交
270 271 272
	return 0;
}

273 274 275 276 277 278
struct async_extent {
	u64 start;
	u64 ram_size;
	u64 compressed_size;
	struct page **pages;
	unsigned long nr_pages;
279
	int compress_type;
280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296
	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,
297 298
				     unsigned long nr_pages,
				     int compress_type)
299 300 301 302
{
	struct async_extent *async_extent;

	async_extent = kmalloc(sizeof(*async_extent), GFP_NOFS);
303
	BUG_ON(!async_extent); /* -ENOMEM */
304 305 306 307 308
	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;
309
	async_extent->compress_type = compress_type;
310 311 312 313
	list_add_tail(&async_extent->list, &cow->extents);
	return 0;
}

C
Chris Mason 已提交
314
/*
315 316 317
 * 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
318
 *
319 320 321 322 323
 * 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
324
 *
325 326 327
 * If this code finds it can't get good compression, it puts an
 * entry onto the work queue to write the uncompressed bytes.  This
 * makes sure that both compressed inodes and uncompressed inodes
328 329
 * are written in the same order that the flusher thread sent them
 * down.
C
Chris Mason 已提交
330
 */
331 332 333 334 335
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)
336 337 338
{
	struct btrfs_root *root = BTRFS_I(inode)->root;
	struct btrfs_trans_handle *trans;
339 340
	u64 num_bytes;
	u64 blocksize = root->sectorsize;
C
Chris Mason 已提交
341
	u64 actual_end;
342
	u64 isize = i_size_read(inode);
343
	int ret = 0;
C
Chris Mason 已提交
344 345 346 347 348 349
	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;
350
	unsigned long max_uncompressed = 128 * 1024;
C
Chris Mason 已提交
351 352
	int i;
	int will_compress;
353
	int compress_type = root->fs_info->compress_type;
354

355 356 357
	/* if this is a small write inside eof, kick off a defrag */
	if ((end - start + 1) < 16 * 1024 &&
	    (start > 0 || end + 1 < BTRFS_I(inode)->disk_i_size))
C
Chris Mason 已提交
358 359
		btrfs_add_inode_defrag(NULL, inode);

360
	actual_end = min_t(u64, isize, end + 1);
C
Chris Mason 已提交
361 362 363 364
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);
365

366 367 368 369 370 371 372 373 374 375 376 377 378
	/*
	 * 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
379 380 381 382
	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
383 384 385 386 387 388 389
	 * 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
390 391
	 */
	total_compressed = min(total_compressed, max_uncompressed);
392
	num_bytes = (end - start + blocksize) & ~(blocksize - 1);
393
	num_bytes = max(blocksize,  num_bytes);
C
Chris Mason 已提交
394 395
	total_in = 0;
	ret = 0;
396

397 398 399 400
	/*
	 * 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
401
	 */
402
	if (!(BTRFS_I(inode)->flags & BTRFS_INODE_NOCOMPRESS) &&
C
Chris Mason 已提交
403
	    (btrfs_test_opt(root, COMPRESS) ||
404 405
	     (BTRFS_I(inode)->force_compress) ||
	     (BTRFS_I(inode)->flags & BTRFS_INODE_COMPRESS))) {
C
Chris Mason 已提交
406
		WARN_ON(pages);
407
		pages = kzalloc(sizeof(struct page *) * nr_pages, GFP_NOFS);
408 409 410 411
		if (!pages) {
			/* just bail out to the uncompressed code */
			goto cont;
		}
C
Chris Mason 已提交
412

413 414 415 416 417 418 419 420 421 422
		if (BTRFS_I(inode)->force_compress)
			compress_type = BTRFS_I(inode)->force_compress;

		ret = btrfs_compress_pages(compress_type,
					   inode->i_mapping, start,
					   total_compressed, pages,
					   nr_pages, &nr_pages_ret,
					   &total_in,
					   &total_compressed,
					   max_compressed);
C
Chris Mason 已提交
423 424 425 426 427 428 429 430 431 432 433

		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) {
434
				kaddr = kmap_atomic(page);
C
Chris Mason 已提交
435 436
				memset(kaddr + offset, 0,
				       PAGE_CACHE_SIZE - offset);
437
				kunmap_atomic(kaddr);
C
Chris Mason 已提交
438 439 440 441
			}
			will_compress = 1;
		}
	}
442
cont:
C
Chris Mason 已提交
443
	if (start == 0) {
444
		trans = btrfs_join_transaction(root);
445 446 447 448 449
		if (IS_ERR(trans)) {
			ret = PTR_ERR(trans);
			trans = NULL;
			goto cleanup_and_out;
		}
450
		trans->block_rsv = &root->fs_info->delalloc_block_rsv;
451

C
Chris Mason 已提交
452
		/* lets try to make an inline extent */
453
		if (ret || total_in < (actual_end - start)) {
C
Chris Mason 已提交
454
			/* we didn't compress the entire range, try
455
			 * to make an uncompressed inline extent.
C
Chris Mason 已提交
456 457
			 */
			ret = cow_file_range_inline(trans, root, inode,
458
						    start, end, 0, 0, NULL);
C
Chris Mason 已提交
459
		} else {
460
			/* try making a compressed inline extent */
C
Chris Mason 已提交
461 462
			ret = cow_file_range_inline(trans, root, inode,
						    start, end,
463 464
						    total_compressed,
						    compress_type, pages);
C
Chris Mason 已提交
465
		}
466
		if (ret <= 0) {
467
			/*
468 469 470
			 * inline extent creation worked or returned error,
			 * we don't need to create any more async work items.
			 * Unlock and free up our temp pages.
471
			 */
C
Chris Mason 已提交
472
			extent_clear_unlock_delalloc(inode,
473 474 475
			     &BTRFS_I(inode)->io_tree,
			     start, end, NULL,
			     EXTENT_CLEAR_UNLOCK_PAGE | EXTENT_CLEAR_DIRTY |
476
			     EXTENT_CLEAR_DELALLOC |
477
			     EXTENT_SET_WRITEBACK | EXTENT_END_WRITEBACK);
478 479

			btrfs_end_transaction(trans, root);
C
Chris Mason 已提交
480 481
			goto free_pages_out;
		}
482
		btrfs_end_transaction(trans, root);
C
Chris Mason 已提交
483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511
	}

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

		/*
		 * one last check to make sure the compression is really a
		 * win, compare the page count read with the blocks on disk
		 */
		total_in = (total_in + PAGE_CACHE_SIZE - 1) &
			~(PAGE_CACHE_SIZE - 1);
		if (total_compressed >= total_in) {
			will_compress = 0;
		} else {
			num_bytes = total_in;
		}
	}
	if (!will_compress && pages) {
		/*
		 * the compression code ran but failed to make things smaller,
		 * free any pages it allocated and our page pointer array
		 */
		for (i = 0; i < nr_pages_ret; i++) {
C
Chris Mason 已提交
512
			WARN_ON(pages[i]->mapping);
C
Chris Mason 已提交
513 514 515 516 517 518 519 520
			page_cache_release(pages[i]);
		}
		kfree(pages);
		pages = NULL;
		total_compressed = 0;
		nr_pages_ret = 0;

		/* flag the file so we don't compress in the future */
C
Chris Mason 已提交
521 522
		if (!btrfs_test_opt(root, FORCE_COMPRESS) &&
		    !(BTRFS_I(inode)->force_compress)) {
C
Chris Mason 已提交
523
			BTRFS_I(inode)->flags |= BTRFS_INODE_NOCOMPRESS;
C
Chris Mason 已提交
524
		}
C
Chris Mason 已提交
525
	}
526 527
	if (will_compress) {
		*num_added += 1;
C
Chris Mason 已提交
528

529 530 531 532 533
		/* 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,
534 535
				 total_compressed, pages, nr_pages_ret,
				 compress_type);
536

537
		if (start + num_bytes < end) {
538 539 540 541 542 543
			start += num_bytes;
			pages = NULL;
			cond_resched();
			goto again;
		}
	} else {
544
cleanup_and_bail_uncompressed:
545 546 547 548 549 550 551 552 553 554 555 556
		/*
		 * 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 */
		}
557 558
		add_async_extent(async_cow, start, end - start + 1,
				 0, NULL, 0, BTRFS_COMPRESS_NONE);
559 560
		*num_added += 1;
	}
561

562
out:
563
	return ret;
564 565 566 567 568 569

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
570
	kfree(pages);
571 572

	goto out;
573 574 575 576 577 578 579 580 581 582 583 584 585 586

cleanup_and_out:
	extent_clear_unlock_delalloc(inode, &BTRFS_I(inode)->io_tree,
				     start, end, NULL,
				     EXTENT_CLEAR_UNLOCK_PAGE |
				     EXTENT_CLEAR_DIRTY |
				     EXTENT_CLEAR_DELALLOC |
				     EXTENT_SET_WRITEBACK |
				     EXTENT_END_WRITEBACK);
	if (!trans || IS_ERR(trans))
		btrfs_error(root->fs_info, ret, "Failed to join transaction");
	else
		btrfs_abort_transaction(trans, root, ret);
	goto free_pages_out;
587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605
}

/*
 * 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;
606
	int ret = 0;
607 608 609 610

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

611
again:
C
Chris Mason 已提交
612
	while (!list_empty(&async_cow->extents)) {
613 614 615
		async_extent = list_entry(async_cow->extents.next,
					  struct async_extent, list);
		list_del(&async_extent->list);
C
Chris Mason 已提交
616

617 618
		io_tree = &BTRFS_I(inode)->io_tree;

619
retry:
620 621 622 623 624 625
		/* 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,
626
					 async_extent->start +
627
					 async_extent->ram_size - 1);
628 629

			/* allocate blocks */
630 631 632 633 634
			ret = cow_file_range(inode, async_cow->locked_page,
					     async_extent->start,
					     async_extent->start +
					     async_extent->ram_size - 1,
					     &page_started, &nr_written, 0);
635

636 637
			/* JDM XXX */

638 639 640 641 642 643
			/*
			 * 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.
			 */
644
			if (!page_started && !ret)
645 646
				extent_write_locked_range(io_tree,
						  inode, async_extent->start,
C
Chris Mason 已提交
647
						  async_extent->start +
648 649 650
						  async_extent->ram_size - 1,
						  btrfs_get_extent,
						  WB_SYNC_ALL);
651 652
			else if (ret)
				unlock_page(async_cow->locked_page);
653 654 655 656 657 658
			kfree(async_extent);
			cond_resched();
			continue;
		}

		lock_extent(io_tree, async_extent->start,
659
			    async_extent->start + async_extent->ram_size - 1);
660

661
		trans = btrfs_join_transaction(root);
662 663 664 665 666
		if (IS_ERR(trans)) {
			ret = PTR_ERR(trans);
		} else {
			trans->block_rsv = &root->fs_info->delalloc_block_rsv;
			ret = btrfs_reserve_extent(trans, root,
667 668
					   async_extent->compressed_size,
					   async_extent->compressed_size,
669
					   0, alloc_hint, &ins, 1);
670
			if (ret && ret != -ENOSPC)
671 672 673
				btrfs_abort_transaction(trans, root, ret);
			btrfs_end_transaction(trans, root);
		}
674

675 676
		if (ret) {
			int i;
677

678 679 680 681 682 683 684
			for (i = 0; i < async_extent->nr_pages; i++) {
				WARN_ON(async_extent->pages[i]->mapping);
				page_cache_release(async_extent->pages[i]);
			}
			kfree(async_extent->pages);
			async_extent->nr_pages = 0;
			async_extent->pages = NULL;
685

686 687
			if (ret == -ENOSPC)
				goto retry;
688
			goto out_free;
689 690
		}

691 692 693 694 695 696 697 698
		/*
		 * here we're doing allocation and writeback of the
		 * compressed pages
		 */
		btrfs_drop_extent_cache(inode, async_extent->start,
					async_extent->start +
					async_extent->ram_size - 1, 0);

699
		em = alloc_extent_map();
700 701
		if (!em)
			goto out_free_reserve;
702 703
		em->start = async_extent->start;
		em->len = async_extent->ram_size;
704
		em->orig_start = em->start;
705 706
		em->mod_start = em->start;
		em->mod_len = em->len;
C
Chris Mason 已提交
707

708 709
		em->block_start = ins.objectid;
		em->block_len = ins.offset;
710
		em->orig_block_len = ins.offset;
711
		em->bdev = root->fs_info->fs_devices->latest_bdev;
712
		em->compress_type = async_extent->compress_type;
713 714
		set_bit(EXTENT_FLAG_PINNED, &em->flags);
		set_bit(EXTENT_FLAG_COMPRESSED, &em->flags);
715
		em->generation = -1;
716

C
Chris Mason 已提交
717
		while (1) {
718
			write_lock(&em_tree->lock);
719
			ret = add_extent_mapping(em_tree, em);
720 721 722
			if (!ret)
				list_move(&em->list,
					  &em_tree->modified_extents);
723
			write_unlock(&em_tree->lock);
724 725 726 727 728 729 730 731 732
			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);
		}

733 734 735
		if (ret)
			goto out_free_reserve;

736 737 738 739 740 741 742
		ret = btrfs_add_ordered_extent_compress(inode,
						async_extent->start,
						ins.objectid,
						async_extent->ram_size,
						ins.offset,
						BTRFS_ORDERED_COMPRESSED,
						async_extent->compress_type);
743 744
		if (ret)
			goto out_free_reserve;
745 746 747 748 749

		/*
		 * clear dirty, set writeback and unlock the pages.
		 */
		extent_clear_unlock_delalloc(inode,
750 751 752 753 754 755
				&BTRFS_I(inode)->io_tree,
				async_extent->start,
				async_extent->start +
				async_extent->ram_size - 1,
				NULL, EXTENT_CLEAR_UNLOCK_PAGE |
				EXTENT_CLEAR_UNLOCK |
756
				EXTENT_CLEAR_DELALLOC |
757
				EXTENT_CLEAR_DIRTY | EXTENT_SET_WRITEBACK);
758 759

		ret = btrfs_submit_compressed_write(inode,
C
Chris Mason 已提交
760 761 762 763 764
				    async_extent->start,
				    async_extent->ram_size,
				    ins.objectid,
				    ins.offset, async_extent->pages,
				    async_extent->nr_pages);
765 766
		alloc_hint = ins.objectid + ins.offset;
		kfree(async_extent);
767 768
		if (ret)
			goto out;
769 770
		cond_resched();
	}
771 772 773
	ret = 0;
out:
	return ret;
774 775
out_free_reserve:
	btrfs_free_reserved_extent(root, ins.objectid, ins.offset);
776
out_free:
777 778 779 780 781 782 783 784 785 786
	extent_clear_unlock_delalloc(inode, &BTRFS_I(inode)->io_tree,
				     async_extent->start,
				     async_extent->start +
				     async_extent->ram_size - 1,
				     NULL, EXTENT_CLEAR_UNLOCK_PAGE |
				     EXTENT_CLEAR_UNLOCK |
				     EXTENT_CLEAR_DELALLOC |
				     EXTENT_CLEAR_DIRTY |
				     EXTENT_SET_WRITEBACK |
				     EXTENT_END_WRITEBACK);
787
	kfree(async_extent);
788
	goto again;
789 790
}

791 792 793 794 795 796 797 798 799 800 801 802 803 804 805 806 807 808 809 810 811 812 813 814 815 816 817 818 819 820 821 822
static u64 get_extent_allocation_hint(struct inode *inode, u64 start,
				      u64 num_bytes)
{
	struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
	struct extent_map *em;
	u64 alloc_hint = 0;

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

	return alloc_hint;
}

823 824 825 826 827 828 829 830 831 832 833 834 835
/*
 * 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.
 */
836 837 838 839 840 841 842
static noinline int __cow_file_range(struct btrfs_trans_handle *trans,
				     struct inode *inode,
				     struct btrfs_root *root,
				     struct page *locked_page,
				     u64 start, u64 end, int *page_started,
				     unsigned long *nr_written,
				     int unlock)
843 844 845 846 847 848 849 850 851 852 853 854
{
	u64 alloc_hint = 0;
	u64 num_bytes;
	unsigned long ram_size;
	u64 disk_num_bytes;
	u64 cur_alloc_size;
	u64 blocksize = root->sectorsize;
	struct btrfs_key ins;
	struct extent_map *em;
	struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
	int ret = 0;

855
	BUG_ON(btrfs_is_free_space_inode(inode));
856 857 858 859 860

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

C
Chris Mason 已提交
861
	/* if this is a small write inside eof, kick off defrag */
862 863
	if (num_bytes < 64 * 1024 &&
	    (start > 0 || end + 1 < BTRFS_I(inode)->disk_i_size))
C
Chris Mason 已提交
864 865
		btrfs_add_inode_defrag(trans, inode);

866 867 868
	if (start == 0) {
		/* lets try to make an inline extent */
		ret = cow_file_range_inline(trans, root, inode,
869
					    start, end, 0, 0, NULL);
870 871
		if (ret == 0) {
			extent_clear_unlock_delalloc(inode,
872 873 874 875 876 877 878 879
				     &BTRFS_I(inode)->io_tree,
				     start, end, NULL,
				     EXTENT_CLEAR_UNLOCK_PAGE |
				     EXTENT_CLEAR_UNLOCK |
				     EXTENT_CLEAR_DELALLOC |
				     EXTENT_CLEAR_DIRTY |
				     EXTENT_SET_WRITEBACK |
				     EXTENT_END_WRITEBACK);
880

881 882 883 884
			*nr_written = *nr_written +
			     (end - start + PAGE_CACHE_SIZE) / PAGE_CACHE_SIZE;
			*page_started = 1;
			goto out;
885 886 887
		} else if (ret < 0) {
			btrfs_abort_transaction(trans, root, ret);
			goto out_unlock;
888 889 890 891
		}
	}

	BUG_ON(disk_num_bytes >
892
	       btrfs_super_total_bytes(root->fs_info->super_copy));
893

894
	alloc_hint = get_extent_allocation_hint(inode, start, num_bytes);
895 896
	btrfs_drop_extent_cache(inode, start, start + num_bytes - 1, 0);

C
Chris Mason 已提交
897
	while (disk_num_bytes > 0) {
898 899
		unsigned long op;

900
		cur_alloc_size = disk_num_bytes;
901
		ret = btrfs_reserve_extent(trans, root, cur_alloc_size,
902
					   root->sectorsize, 0, alloc_hint,
903
					   &ins, 1);
904 905 906 907
		if (ret < 0) {
			btrfs_abort_transaction(trans, root, ret);
			goto out_unlock;
		}
C
Chris Mason 已提交
908

909
		em = alloc_extent_map();
910
		BUG_ON(!em); /* -ENOMEM */
911
		em->start = start;
912
		em->orig_start = em->start;
913 914
		ram_size = ins.offset;
		em->len = ins.offset;
915 916
		em->mod_start = em->start;
		em->mod_len = em->len;
C
Chris Mason 已提交
917

918
		em->block_start = ins.objectid;
C
Chris Mason 已提交
919
		em->block_len = ins.offset;
920
		em->orig_block_len = ins.offset;
921
		em->bdev = root->fs_info->fs_devices->latest_bdev;
922
		set_bit(EXTENT_FLAG_PINNED, &em->flags);
923
		em->generation = -1;
C
Chris Mason 已提交
924

C
Chris Mason 已提交
925
		while (1) {
926
			write_lock(&em_tree->lock);
927
			ret = add_extent_mapping(em_tree, em);
928 929 930
			if (!ret)
				list_move(&em->list,
					  &em_tree->modified_extents);
931
			write_unlock(&em_tree->lock);
932 933 934 935 936
			if (ret != -EEXIST) {
				free_extent_map(em);
				break;
			}
			btrfs_drop_extent_cache(inode, start,
C
Chris Mason 已提交
937
						start + ram_size - 1, 0);
938 939
		}

940
		cur_alloc_size = ins.offset;
941
		ret = btrfs_add_ordered_extent(inode, start, ins.objectid,
942
					       ram_size, cur_alloc_size, 0);
943
		BUG_ON(ret); /* -ENOMEM */
C
Chris Mason 已提交
944

945 946 947 948
		if (root->root_key.objectid ==
		    BTRFS_DATA_RELOC_TREE_OBJECTID) {
			ret = btrfs_reloc_clone_csums(inode, start,
						      cur_alloc_size);
949 950 951 952
			if (ret) {
				btrfs_abort_transaction(trans, root, ret);
				goto out_unlock;
			}
953 954
		}

C
Chris Mason 已提交
955
		if (disk_num_bytes < cur_alloc_size)
956
			break;
C
Chris Mason 已提交
957

C
Chris Mason 已提交
958 959 960
		/* 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
961 962 963
		 *
		 * Do set the Private2 bit so we know this page was properly
		 * setup for writepage
C
Chris Mason 已提交
964
		 */
965 966 967 968
		op = unlock ? EXTENT_CLEAR_UNLOCK_PAGE : 0;
		op |= EXTENT_CLEAR_UNLOCK | EXTENT_CLEAR_DELALLOC |
			EXTENT_SET_PRIVATE2;

C
Chris Mason 已提交
969 970
		extent_clear_unlock_delalloc(inode, &BTRFS_I(inode)->io_tree,
					     start, start + ram_size - 1,
971
					     locked_page, op);
C
Chris Mason 已提交
972
		disk_num_bytes -= cur_alloc_size;
973 974 975
		num_bytes -= cur_alloc_size;
		alloc_hint = ins.objectid + ins.offset;
		start += cur_alloc_size;
976
	}
977
out:
978
	return ret;
979

980 981 982
out_unlock:
	extent_clear_unlock_delalloc(inode,
		     &BTRFS_I(inode)->io_tree,
983
		     start, end, locked_page,
984 985 986 987 988 989 990 991
		     EXTENT_CLEAR_UNLOCK_PAGE |
		     EXTENT_CLEAR_UNLOCK |
		     EXTENT_CLEAR_DELALLOC |
		     EXTENT_CLEAR_DIRTY |
		     EXTENT_SET_WRITEBACK |
		     EXTENT_END_WRITEBACK);

	goto out;
992
}
C
Chris Mason 已提交
993

994 995 996 997 998 999 1000 1001 1002 1003 1004 1005 1006 1007 1008 1009 1010 1011 1012 1013 1014 1015 1016 1017 1018 1019 1020 1021 1022 1023 1024 1025 1026
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_trans_handle *trans;
	struct btrfs_root *root = BTRFS_I(inode)->root;
	int ret;

	trans = btrfs_join_transaction(root);
	if (IS_ERR(trans)) {
		extent_clear_unlock_delalloc(inode,
			     &BTRFS_I(inode)->io_tree,
			     start, end, locked_page,
			     EXTENT_CLEAR_UNLOCK_PAGE |
			     EXTENT_CLEAR_UNLOCK |
			     EXTENT_CLEAR_DELALLOC |
			     EXTENT_CLEAR_DIRTY |
			     EXTENT_SET_WRITEBACK |
			     EXTENT_END_WRITEBACK);
		return PTR_ERR(trans);
	}
	trans->block_rsv = &root->fs_info->delalloc_block_rsv;

	ret = __cow_file_range(trans, inode, root, locked_page, start, end,
			       page_started, nr_written, unlock);

	btrfs_end_transaction(trans, root);

	return ret;
}

1027 1028 1029 1030 1031 1032 1033 1034 1035 1036 1037 1038
/*
 * 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);
1039
	if (num_added == 0) {
1040
		btrfs_add_delayed_iput(async_cow->inode);
1041
		async_cow->inode = NULL;
1042
	}
1043 1044 1045 1046 1047 1048 1049 1050 1051 1052 1053 1054 1055 1056 1057 1058 1059
}

/*
 * 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;

1060
	if (atomic_sub_return(nr_pages, &root->fs_info->async_delalloc_pages) <
1061
	    5 * 1024 * 1024 &&
1062 1063 1064
	    waitqueue_active(&root->fs_info->async_submit_wait))
		wake_up(&root->fs_info->async_submit_wait);

C
Chris Mason 已提交
1065
	if (async_cow->inode)
1066 1067
		submit_compressed_extents(async_cow->inode, async_cow);
}
C
Chris Mason 已提交
1068

1069 1070 1071 1072
static noinline void async_cow_free(struct btrfs_work *work)
{
	struct async_cow *async_cow;
	async_cow = container_of(work, struct async_cow, work);
1073
	if (async_cow->inode)
1074
		btrfs_add_delayed_iput(async_cow->inode);
1075 1076 1077 1078 1079 1080 1081 1082 1083 1084 1085
	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;
1086
	int limit = 10 * 1024 * 1024;
1087

1088 1089
	clear_extent_bit(&BTRFS_I(inode)->io_tree, start, end, EXTENT_LOCKED,
			 1, 0, NULL, GFP_NOFS);
C
Chris Mason 已提交
1090
	while (start < end) {
1091
		async_cow = kmalloc(sizeof(*async_cow), GFP_NOFS);
1092
		BUG_ON(!async_cow); /* -ENOMEM */
1093
		async_cow->inode = igrab(inode);
1094 1095 1096 1097
		async_cow->root = root;
		async_cow->locked_page = locked_page;
		async_cow->start = start;

1098
		if (BTRFS_I(inode)->flags & BTRFS_INODE_NOCOMPRESS)
1099 1100 1101 1102 1103 1104 1105 1106 1107 1108 1109 1110 1111 1112 1113 1114 1115 1116 1117 1118 1119 1120 1121 1122 1123
			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
1124
		while (atomic_read(&root->fs_info->async_submit_draining) &&
1125 1126 1127 1128 1129 1130 1131 1132 1133 1134 1135
		      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;
1136 1137
}

C
Chris Mason 已提交
1138
static noinline int csum_exist_in_range(struct btrfs_root *root,
1139 1140 1141 1142 1143 1144
					u64 bytenr, u64 num_bytes)
{
	int ret;
	struct btrfs_ordered_sum *sums;
	LIST_HEAD(list);

Y
Yan Zheng 已提交
1145
	ret = btrfs_lookup_csums_range(root->fs_info->csum_root, bytenr,
A
Arne Jansen 已提交
1146
				       bytenr + num_bytes - 1, &list, 0);
1147 1148 1149 1150 1151 1152 1153 1154 1155 1156 1157
	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
1158 1159 1160 1161 1162 1163 1164
/*
 * 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
 */
1165 1166
static noinline int run_delalloc_nocow(struct inode *inode,
				       struct page *locked_page,
1167 1168
			      u64 start, u64 end, int *page_started, int force,
			      unsigned long *nr_written)
1169 1170
{
	struct btrfs_root *root = BTRFS_I(inode)->root;
1171
	struct btrfs_trans_handle *trans;
1172 1173
	struct extent_buffer *leaf;
	struct btrfs_path *path;
Y
Yan Zheng 已提交
1174
	struct btrfs_file_extent_item *fi;
1175
	struct btrfs_key found_key;
Y
Yan Zheng 已提交
1176 1177 1178
	u64 cow_start;
	u64 cur_offset;
	u64 extent_end;
1179
	u64 extent_offset;
Y
Yan Zheng 已提交
1180 1181
	u64 disk_bytenr;
	u64 num_bytes;
1182
	u64 disk_num_bytes;
Y
Yan Zheng 已提交
1183
	int extent_type;
1184
	int ret, err;
Y
Yan Zheng 已提交
1185
	int type;
Y
Yan Zheng 已提交
1186 1187
	int nocow;
	int check_prev = 1;
1188
	bool nolock;
L
Li Zefan 已提交
1189
	u64 ino = btrfs_ino(inode);
1190 1191

	path = btrfs_alloc_path();
1192 1193 1194 1195 1196 1197 1198 1199 1200 1201
	if (!path) {
		extent_clear_unlock_delalloc(inode,
			     &BTRFS_I(inode)->io_tree,
			     start, end, locked_page,
			     EXTENT_CLEAR_UNLOCK_PAGE |
			     EXTENT_CLEAR_UNLOCK |
			     EXTENT_CLEAR_DELALLOC |
			     EXTENT_CLEAR_DIRTY |
			     EXTENT_SET_WRITEBACK |
			     EXTENT_END_WRITEBACK);
1202
		return -ENOMEM;
1203
	}
1204

1205
	nolock = btrfs_is_free_space_inode(inode);
1206 1207

	if (nolock)
1208
		trans = btrfs_join_transaction_nolock(root);
1209
	else
1210
		trans = btrfs_join_transaction(root);
C
Chris Mason 已提交
1211

1212
	if (IS_ERR(trans)) {
1213 1214 1215 1216 1217 1218 1219 1220 1221
		extent_clear_unlock_delalloc(inode,
			     &BTRFS_I(inode)->io_tree,
			     start, end, locked_page,
			     EXTENT_CLEAR_UNLOCK_PAGE |
			     EXTENT_CLEAR_UNLOCK |
			     EXTENT_CLEAR_DELALLOC |
			     EXTENT_CLEAR_DIRTY |
			     EXTENT_SET_WRITEBACK |
			     EXTENT_END_WRITEBACK);
1222 1223 1224 1225
		btrfs_free_path(path);
		return PTR_ERR(trans);
	}

1226
	trans->block_rsv = &root->fs_info->delalloc_block_rsv;
1227

Y
Yan Zheng 已提交
1228 1229 1230
	cow_start = (u64)-1;
	cur_offset = start;
	while (1) {
L
Li Zefan 已提交
1231
		ret = btrfs_lookup_file_extent(trans, root, path, ino,
Y
Yan Zheng 已提交
1232
					       cur_offset, 0);
1233 1234 1235 1236
		if (ret < 0) {
			btrfs_abort_transaction(trans, root, ret);
			goto error;
		}
Y
Yan Zheng 已提交
1237 1238 1239 1240
		if (ret > 0 && path->slots[0] > 0 && check_prev) {
			leaf = path->nodes[0];
			btrfs_item_key_to_cpu(leaf, &found_key,
					      path->slots[0] - 1);
L
Li Zefan 已提交
1241
			if (found_key.objectid == ino &&
Y
Yan Zheng 已提交
1242 1243 1244 1245 1246 1247 1248 1249
			    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);
1250 1251 1252 1253
			if (ret < 0) {
				btrfs_abort_transaction(trans, root, ret);
				goto error;
			}
Y
Yan Zheng 已提交
1254 1255 1256 1257
			if (ret > 0)
				break;
			leaf = path->nodes[0];
		}
1258

Y
Yan Zheng 已提交
1259 1260
		nocow = 0;
		disk_bytenr = 0;
1261
		num_bytes = 0;
Y
Yan Zheng 已提交
1262 1263
		btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);

L
Li Zefan 已提交
1264
		if (found_key.objectid > ino ||
Y
Yan Zheng 已提交
1265 1266 1267 1268 1269 1270
		    found_key.type > BTRFS_EXTENT_DATA_KEY ||
		    found_key.offset > end)
			break;

		if (found_key.offset > cur_offset) {
			extent_end = found_key.offset;
1271
			extent_type = 0;
Y
Yan Zheng 已提交
1272 1273 1274 1275 1276 1277 1278
			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
1279 1280
		if (extent_type == BTRFS_FILE_EXTENT_REG ||
		    extent_type == BTRFS_FILE_EXTENT_PREALLOC) {
Y
Yan Zheng 已提交
1281
			disk_bytenr = btrfs_file_extent_disk_bytenr(leaf, fi);
1282
			extent_offset = btrfs_file_extent_offset(leaf, fi);
Y
Yan Zheng 已提交
1283 1284
			extent_end = found_key.offset +
				btrfs_file_extent_num_bytes(leaf, fi);
1285 1286
			disk_num_bytes =
				btrfs_file_extent_disk_num_bytes(leaf, fi);
Y
Yan Zheng 已提交
1287 1288 1289 1290
			if (extent_end <= start) {
				path->slots[0]++;
				goto next_slot;
			}
1291 1292
			if (disk_bytenr == 0)
				goto out_check;
Y
Yan Zheng 已提交
1293 1294 1295 1296
			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
1297 1298
			if (extent_type == BTRFS_FILE_EXTENT_REG && !force)
				goto out_check;
1299
			if (btrfs_extent_readonly(root, disk_bytenr))
Y
Yan Zheng 已提交
1300
				goto out_check;
L
Li Zefan 已提交
1301
			if (btrfs_cross_ref_exist(trans, root, ino,
1302 1303
						  found_key.offset -
						  extent_offset, disk_bytenr))
1304
				goto out_check;
1305
			disk_bytenr += extent_offset;
1306 1307 1308 1309 1310 1311 1312 1313 1314
			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
1315 1316 1317 1318 1319 1320 1321 1322 1323 1324 1325 1326 1327 1328 1329 1330 1331 1332 1333 1334 1335
			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;
1336 1337
		}

1338
		btrfs_release_path(path);
Y
Yan Zheng 已提交
1339
		if (cow_start != (u64)-1) {
1340 1341 1342
			ret = __cow_file_range(trans, inode, root, locked_page,
					       cow_start, found_key.offset - 1,
					       page_started, nr_written, 1);
1343 1344 1345 1346
			if (ret) {
				btrfs_abort_transaction(trans, root, ret);
				goto error;
			}
Y
Yan Zheng 已提交
1347
			cow_start = (u64)-1;
1348
		}
Y
Yan Zheng 已提交
1349

Y
Yan Zheng 已提交
1350 1351 1352 1353
		if (extent_type == BTRFS_FILE_EXTENT_PREALLOC) {
			struct extent_map *em;
			struct extent_map_tree *em_tree;
			em_tree = &BTRFS_I(inode)->extent_tree;
1354
			em = alloc_extent_map();
1355
			BUG_ON(!em); /* -ENOMEM */
Y
Yan Zheng 已提交
1356
			em->start = cur_offset;
1357
			em->orig_start = found_key.offset - extent_offset;
Y
Yan Zheng 已提交
1358 1359 1360
			em->len = num_bytes;
			em->block_len = num_bytes;
			em->block_start = disk_bytenr;
1361
			em->orig_block_len = disk_num_bytes;
Y
Yan Zheng 已提交
1362
			em->bdev = root->fs_info->fs_devices->latest_bdev;
1363 1364
			em->mod_start = em->start;
			em->mod_len = em->len;
Y
Yan Zheng 已提交
1365
			set_bit(EXTENT_FLAG_PINNED, &em->flags);
1366
			set_bit(EXTENT_FLAG_FILLING, &em->flags);
1367
			em->generation = -1;
Y
Yan Zheng 已提交
1368
			while (1) {
1369
				write_lock(&em_tree->lock);
Y
Yan Zheng 已提交
1370
				ret = add_extent_mapping(em_tree, em);
1371 1372 1373
				if (!ret)
					list_move(&em->list,
						  &em_tree->modified_extents);
1374
				write_unlock(&em_tree->lock);
Y
Yan Zheng 已提交
1375 1376 1377 1378 1379 1380 1381 1382 1383 1384 1385
				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
1386 1387

		ret = btrfs_add_ordered_extent(inode, cur_offset, disk_bytenr,
Y
Yan Zheng 已提交
1388
					       num_bytes, num_bytes, type);
1389
		BUG_ON(ret); /* -ENOMEM */
1390

1391 1392 1393 1394
		if (root->root_key.objectid ==
		    BTRFS_DATA_RELOC_TREE_OBJECTID) {
			ret = btrfs_reloc_clone_csums(inode, cur_offset,
						      num_bytes);
1395 1396 1397 1398
			if (ret) {
				btrfs_abort_transaction(trans, root, ret);
				goto error;
			}
1399 1400
		}

Y
Yan Zheng 已提交
1401
		extent_clear_unlock_delalloc(inode, &BTRFS_I(inode)->io_tree,
1402 1403 1404 1405
				cur_offset, cur_offset + num_bytes - 1,
				locked_page, EXTENT_CLEAR_UNLOCK_PAGE |
				EXTENT_CLEAR_UNLOCK | EXTENT_CLEAR_DELALLOC |
				EXTENT_SET_PRIVATE2);
Y
Yan Zheng 已提交
1406 1407 1408
		cur_offset = extent_end;
		if (cur_offset > end)
			break;
1409
	}
1410
	btrfs_release_path(path);
Y
Yan Zheng 已提交
1411

1412
	if (cur_offset <= end && cow_start == (u64)-1) {
Y
Yan Zheng 已提交
1413
		cow_start = cur_offset;
1414 1415 1416
		cur_offset = end;
	}

Y
Yan Zheng 已提交
1417
	if (cow_start != (u64)-1) {
1418 1419 1420
		ret = __cow_file_range(trans, inode, root, locked_page,
				       cow_start, end,
				       page_started, nr_written, 1);
1421 1422 1423 1424
		if (ret) {
			btrfs_abort_transaction(trans, root, ret);
			goto error;
		}
Y
Yan Zheng 已提交
1425 1426
	}

1427
error:
1428
	err = btrfs_end_transaction(trans, root);
1429 1430 1431
	if (!ret)
		ret = err;

1432 1433 1434 1435 1436 1437 1438 1439 1440 1441 1442
	if (ret && cur_offset < end)
		extent_clear_unlock_delalloc(inode,
			     &BTRFS_I(inode)->io_tree,
			     cur_offset, end, locked_page,
			     EXTENT_CLEAR_UNLOCK_PAGE |
			     EXTENT_CLEAR_UNLOCK |
			     EXTENT_CLEAR_DELALLOC |
			     EXTENT_CLEAR_DIRTY |
			     EXTENT_SET_WRITEBACK |
			     EXTENT_END_WRITEBACK);

1443
	btrfs_free_path(path);
1444
	return ret;
1445 1446
}

C
Chris Mason 已提交
1447 1448 1449
/*
 * extent_io.c call back to do delayed allocation processing
 */
C
Chris Mason 已提交
1450
static int run_delalloc_range(struct inode *inode, struct page *locked_page,
1451 1452
			      u64 start, u64 end, int *page_started,
			      unsigned long *nr_written)
1453 1454
{
	int ret;
1455
	struct btrfs_root *root = BTRFS_I(inode)->root;
1456

1457
	if (BTRFS_I(inode)->flags & BTRFS_INODE_NODATACOW) {
C
Chris Mason 已提交
1458
		ret = run_delalloc_nocow(inode, locked_page, start, end,
C
Chris Mason 已提交
1459
					 page_started, 1, nr_written);
1460
	} else if (BTRFS_I(inode)->flags & BTRFS_INODE_PREALLOC) {
Y
Yan Zheng 已提交
1461
		ret = run_delalloc_nocow(inode, locked_page, start, end,
C
Chris Mason 已提交
1462
					 page_started, 0, nr_written);
1463 1464 1465
	} else if (!btrfs_test_opt(root, COMPRESS) &&
		   !(BTRFS_I(inode)->force_compress) &&
		   !(BTRFS_I(inode)->flags & BTRFS_INODE_COMPRESS)) {
1466 1467
		ret = cow_file_range(inode, locked_page, start, end,
				      page_started, nr_written, 1);
1468 1469 1470
	} else {
		set_bit(BTRFS_INODE_HAS_ASYNC_EXTENT,
			&BTRFS_I(inode)->runtime_flags);
1471
		ret = cow_file_range_async(inode, locked_page, start, end,
C
Chris Mason 已提交
1472
					   page_started, nr_written);
1473
	}
1474 1475 1476
	return ret;
}

1477 1478
static void btrfs_split_extent_hook(struct inode *inode,
				    struct extent_state *orig, u64 split)
J
Josef Bacik 已提交
1479
{
1480
	/* not delalloc, ignore it */
J
Josef Bacik 已提交
1481
	if (!(orig->state & EXTENT_DELALLOC))
1482
		return;
J
Josef Bacik 已提交
1483

1484 1485 1486
	spin_lock(&BTRFS_I(inode)->lock);
	BTRFS_I(inode)->outstanding_extents++;
	spin_unlock(&BTRFS_I(inode)->lock);
J
Josef Bacik 已提交
1487 1488 1489 1490 1491 1492 1493 1494
}

/*
 * extent_io.c merge_extent_hook, used to track merged delayed allocation
 * extents so we can keep track of new extents that are just merged onto old
 * extents, such as when we are doing sequential writes, so we can properly
 * account for the metadata space we'll need.
 */
1495 1496 1497
static void btrfs_merge_extent_hook(struct inode *inode,
				    struct extent_state *new,
				    struct extent_state *other)
J
Josef Bacik 已提交
1498 1499 1500
{
	/* not delalloc, ignore it */
	if (!(other->state & EXTENT_DELALLOC))
1501
		return;
J
Josef Bacik 已提交
1502

1503 1504 1505
	spin_lock(&BTRFS_I(inode)->lock);
	BTRFS_I(inode)->outstanding_extents--;
	spin_unlock(&BTRFS_I(inode)->lock);
J
Josef Bacik 已提交
1506 1507
}

C
Chris Mason 已提交
1508 1509 1510 1511 1512
/*
 * 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.
 */
1513 1514
static void btrfs_set_bit_hook(struct inode *inode,
			       struct extent_state *state, int *bits)
1515
{
J
Josef Bacik 已提交
1516

1517 1518
	/*
	 * set_bit and clear bit hooks normally require _irqsave/restore
1519
	 * but in this case, we are only testing for the DELALLOC
1520 1521
	 * bit, which is only set or cleared with irqs on
	 */
1522
	if (!(state->state & EXTENT_DELALLOC) && (*bits & EXTENT_DELALLOC)) {
1523
		struct btrfs_root *root = BTRFS_I(inode)->root;
1524
		u64 len = state->end + 1 - state->start;
1525
		bool do_list = !btrfs_is_free_space_inode(inode);
J
Josef Bacik 已提交
1526

1527
		if (*bits & EXTENT_FIRST_DELALLOC) {
1528
			*bits &= ~EXTENT_FIRST_DELALLOC;
1529 1530 1531 1532 1533
		} else {
			spin_lock(&BTRFS_I(inode)->lock);
			BTRFS_I(inode)->outstanding_extents++;
			spin_unlock(&BTRFS_I(inode)->lock);
		}
1534

1535 1536
		__percpu_counter_add(&root->fs_info->delalloc_bytes, len,
				     root->fs_info->delalloc_batch);
1537 1538 1539 1540 1541 1542 1543 1544 1545 1546 1547 1548
		spin_lock(&BTRFS_I(inode)->lock);
		BTRFS_I(inode)->delalloc_bytes += len;
		if (do_list && !test_bit(BTRFS_INODE_IN_DELALLOC_LIST,
					 &BTRFS_I(inode)->runtime_flags)) {
			spin_lock(&root->fs_info->delalloc_lock);
			if (list_empty(&BTRFS_I(inode)->delalloc_inodes)) {
				list_add_tail(&BTRFS_I(inode)->delalloc_inodes,
					      &root->fs_info->delalloc_inodes);
				set_bit(BTRFS_INODE_IN_DELALLOC_LIST,
					&BTRFS_I(inode)->runtime_flags);
			}
			spin_unlock(&root->fs_info->delalloc_lock);
1549
		}
1550
		spin_unlock(&BTRFS_I(inode)->lock);
1551 1552 1553
	}
}

C
Chris Mason 已提交
1554 1555 1556
/*
 * extent_io.c clear_bit_hook, see set_bit_hook for why
 */
1557 1558
static void btrfs_clear_bit_hook(struct inode *inode,
				 struct extent_state *state, int *bits)
1559
{
1560 1561
	/*
	 * set_bit and clear bit hooks normally require _irqsave/restore
1562
	 * but in this case, we are only testing for the DELALLOC
1563 1564
	 * bit, which is only set or cleared with irqs on
	 */
1565
	if ((state->state & EXTENT_DELALLOC) && (*bits & EXTENT_DELALLOC)) {
1566
		struct btrfs_root *root = BTRFS_I(inode)->root;
1567
		u64 len = state->end + 1 - state->start;
1568
		bool do_list = !btrfs_is_free_space_inode(inode);
1569

1570
		if (*bits & EXTENT_FIRST_DELALLOC) {
1571
			*bits &= ~EXTENT_FIRST_DELALLOC;
1572 1573 1574 1575 1576
		} else if (!(*bits & EXTENT_DO_ACCOUNTING)) {
			spin_lock(&BTRFS_I(inode)->lock);
			BTRFS_I(inode)->outstanding_extents--;
			spin_unlock(&BTRFS_I(inode)->lock);
		}
1577 1578 1579 1580

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

J
Josef Bacik 已提交
1581 1582
		if (root->root_key.objectid != BTRFS_DATA_RELOC_TREE_OBJECTID
		    && do_list)
1583
			btrfs_free_reserved_data_space(inode, len);
J
Josef Bacik 已提交
1584

1585 1586
		__percpu_counter_add(&root->fs_info->delalloc_bytes, -len,
				     root->fs_info->delalloc_batch);
1587
		spin_lock(&BTRFS_I(inode)->lock);
1588
		BTRFS_I(inode)->delalloc_bytes -= len;
J
Josef Bacik 已提交
1589
		if (do_list && BTRFS_I(inode)->delalloc_bytes == 0 &&
1590 1591 1592 1593 1594 1595 1596 1597 1598
		    test_bit(BTRFS_INODE_IN_DELALLOC_LIST,
			     &BTRFS_I(inode)->runtime_flags)) {
			spin_lock(&root->fs_info->delalloc_lock);
			if (!list_empty(&BTRFS_I(inode)->delalloc_inodes)) {
				list_del_init(&BTRFS_I(inode)->delalloc_inodes);
				clear_bit(BTRFS_INODE_IN_DELALLOC_LIST,
					  &BTRFS_I(inode)->runtime_flags);
			}
			spin_unlock(&root->fs_info->delalloc_lock);
1599
		}
1600
		spin_unlock(&BTRFS_I(inode)->lock);
1601 1602 1603
	}
}

C
Chris Mason 已提交
1604 1605 1606 1607
/*
 * 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
 */
1608
int btrfs_merge_bio_hook(struct page *page, unsigned long offset,
C
Chris Mason 已提交
1609 1610
			 size_t size, struct bio *bio,
			 unsigned long bio_flags)
1611 1612
{
	struct btrfs_root *root = BTRFS_I(page->mapping->host)->root;
1613
	u64 logical = (u64)bio->bi_sector << 9;
1614 1615 1616 1617
	u64 length = 0;
	u64 map_length;
	int ret;

1618 1619 1620
	if (bio_flags & EXTENT_BIO_COMPRESSED)
		return 0;

1621
	length = bio->bi_size;
1622
	map_length = length;
1623
	ret = btrfs_map_block(root->fs_info, READ, logical,
1624
			      &map_length, NULL, 0);
1625
	/* Will always return 0 with map_multi == NULL */
1626
	BUG_ON(ret < 0);
C
Chris Mason 已提交
1627
	if (map_length < length + size)
1628
		return 1;
1629
	return 0;
1630 1631
}

C
Chris Mason 已提交
1632 1633 1634 1635 1636 1637 1638 1639
/*
 * 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
1640 1641
static int __btrfs_submit_bio_start(struct inode *inode, int rw,
				    struct bio *bio, int mirror_num,
1642 1643
				    unsigned long bio_flags,
				    u64 bio_offset)
1644 1645 1646
{
	struct btrfs_root *root = BTRFS_I(inode)->root;
	int ret = 0;
1647

1648
	ret = btrfs_csum_one_bio(root, inode, bio, 0, 0);
1649
	BUG_ON(ret); /* -ENOMEM */
C
Chris Mason 已提交
1650 1651
	return 0;
}
1652

C
Chris Mason 已提交
1653 1654 1655 1656 1657 1658 1659 1660
/*
 * 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
 */
1661
static int __btrfs_submit_bio_done(struct inode *inode, int rw, struct bio *bio,
1662 1663
			  int mirror_num, unsigned long bio_flags,
			  u64 bio_offset)
C
Chris Mason 已提交
1664 1665
{
	struct btrfs_root *root = BTRFS_I(inode)->root;
1666 1667 1668 1669 1670 1671
	int ret;

	ret = btrfs_map_bio(root, rw, bio, mirror_num, 1);
	if (ret)
		bio_endio(bio, ret);
	return ret;
1672 1673
}

C
Chris Mason 已提交
1674
/*
1675 1676
 * 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
1677
 */
1678
static int btrfs_submit_bio_hook(struct inode *inode, int rw, struct bio *bio,
1679 1680
			  int mirror_num, unsigned long bio_flags,
			  u64 bio_offset)
1681 1682 1683
{
	struct btrfs_root *root = BTRFS_I(inode)->root;
	int ret = 0;
1684
	int skip_sum;
1685
	int metadata = 0;
1686
	int async = !atomic_read(&BTRFS_I(inode)->sync_writers);
1687

1688
	skip_sum = BTRFS_I(inode)->flags & BTRFS_INODE_NODATASUM;
1689

1690
	if (btrfs_is_free_space_inode(inode))
1691 1692
		metadata = 2;

1693
	if (!(rw & REQ_WRITE)) {
1694 1695
		ret = btrfs_bio_wq_end_io(root->fs_info, bio, metadata);
		if (ret)
1696
			goto out;
1697

1698
		if (bio_flags & EXTENT_BIO_COMPRESSED) {
1699 1700 1701 1702
			ret = btrfs_submit_compressed_read(inode, bio,
							   mirror_num,
							   bio_flags);
			goto out;
1703 1704 1705
		} else if (!skip_sum) {
			ret = btrfs_lookup_bio_sums(root, inode, bio, NULL);
			if (ret)
1706
				goto out;
1707
		}
1708
		goto mapit;
1709
	} else if (async && !skip_sum) {
1710 1711 1712
		/* csum items have already been cloned */
		if (root->root_key.objectid == BTRFS_DATA_RELOC_TREE_OBJECTID)
			goto mapit;
1713
		/* we're doing a write, do the async checksumming */
1714
		ret = btrfs_wq_submit_bio(BTRFS_I(inode)->root->fs_info,
1715
				   inode, rw, bio, mirror_num,
1716 1717
				   bio_flags, bio_offset,
				   __btrfs_submit_bio_start,
C
Chris Mason 已提交
1718
				   __btrfs_submit_bio_done);
1719
		goto out;
1720 1721 1722 1723
	} else if (!skip_sum) {
		ret = btrfs_csum_one_bio(root, inode, bio, 0, 0);
		if (ret)
			goto out;
1724 1725
	}

1726
mapit:
1727 1728 1729 1730 1731 1732
	ret = btrfs_map_bio(root, rw, bio, mirror_num, 0);

out:
	if (ret < 0)
		bio_endio(bio, ret);
	return ret;
1733
}
C
Chris Mason 已提交
1734

C
Chris Mason 已提交
1735 1736 1737 1738
/*
 * 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.
 */
1739
static noinline int add_pending_csums(struct btrfs_trans_handle *trans,
1740 1741 1742 1743 1744
			     struct inode *inode, u64 file_offset,
			     struct list_head *list)
{
	struct btrfs_ordered_sum *sum;

Q
Qinghuang Feng 已提交
1745
	list_for_each_entry(sum, list, list) {
1746 1747
		btrfs_csum_file_blocks(trans,
		       BTRFS_I(inode)->root->fs_info->csum_root, sum);
1748 1749 1750 1751
	}
	return 0;
}

1752 1753
int btrfs_set_extent_delalloc(struct inode *inode, u64 start, u64 end,
			      struct extent_state **cached_state)
1754
{
J
Julia Lawall 已提交
1755
	WARN_ON((end & (PAGE_CACHE_SIZE - 1)) == 0);
1756
	return set_extent_delalloc(&BTRFS_I(inode)->io_tree, start, end,
1757
				   cached_state, GFP_NOFS);
1758 1759
}

C
Chris Mason 已提交
1760
/* see btrfs_writepage_start_hook for details on why this is required */
1761 1762 1763 1764 1765
struct btrfs_writepage_fixup {
	struct page *page;
	struct btrfs_work work;
};

1766
static void btrfs_writepage_fixup_worker(struct btrfs_work *work)
1767 1768 1769
{
	struct btrfs_writepage_fixup *fixup;
	struct btrfs_ordered_extent *ordered;
1770
	struct extent_state *cached_state = NULL;
1771 1772 1773 1774
	struct page *page;
	struct inode *inode;
	u64 page_start;
	u64 page_end;
1775
	int ret;
1776 1777 1778

	fixup = container_of(work, struct btrfs_writepage_fixup, work);
	page = fixup->page;
C
Chris Mason 已提交
1779
again:
1780 1781 1782 1783 1784 1785 1786 1787 1788 1789
	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;

1790
	lock_extent_bits(&BTRFS_I(inode)->io_tree, page_start, page_end, 0,
1791
			 &cached_state);
C
Chris Mason 已提交
1792 1793

	/* already ordered? We're done */
1794
	if (PagePrivate2(page))
1795
		goto out;
C
Chris Mason 已提交
1796 1797 1798

	ordered = btrfs_lookup_ordered_extent(inode, page_start);
	if (ordered) {
1799 1800
		unlock_extent_cached(&BTRFS_I(inode)->io_tree, page_start,
				     page_end, &cached_state, GFP_NOFS);
C
Chris Mason 已提交
1801 1802
		unlock_page(page);
		btrfs_start_ordered_extent(inode, ordered, 1);
1803
		btrfs_put_ordered_extent(ordered);
C
Chris Mason 已提交
1804 1805
		goto again;
	}
1806

1807 1808 1809 1810 1811 1812 1813 1814
	ret = btrfs_delalloc_reserve_space(inode, PAGE_CACHE_SIZE);
	if (ret) {
		mapping_set_error(page->mapping, ret);
		end_extent_writepage(page, ret, page_start, page_end);
		ClearPageChecked(page);
		goto out;
	 }

1815
	btrfs_set_extent_delalloc(inode, page_start, page_end, &cached_state);
1816
	ClearPageChecked(page);
1817
	set_page_dirty(page);
1818
out:
1819 1820
	unlock_extent_cached(&BTRFS_I(inode)->io_tree, page_start, page_end,
			     &cached_state, GFP_NOFS);
1821 1822 1823
out_page:
	unlock_page(page);
	page_cache_release(page);
1824
	kfree(fixup);
1825 1826 1827 1828 1829 1830 1831 1832
}

/*
 * 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
1833
 * In our case any range that doesn't have the ORDERED bit set
1834 1835 1836 1837
 * 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.
 */
1838
static int btrfs_writepage_start_hook(struct page *page, u64 start, u64 end)
1839 1840 1841 1842 1843
{
	struct inode *inode = page->mapping->host;
	struct btrfs_writepage_fixup *fixup;
	struct btrfs_root *root = BTRFS_I(inode)->root;

1844 1845
	/* this page is properly in the ordered list */
	if (TestClearPagePrivate2(page))
1846 1847 1848 1849 1850 1851 1852 1853
		return 0;

	if (PageChecked(page))
		return -EAGAIN;

	fixup = kzalloc(sizeof(*fixup), GFP_NOFS);
	if (!fixup)
		return -EAGAIN;
1854

1855 1856 1857 1858 1859
	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);
1860
	return -EBUSY;
1861 1862
}

Y
Yan Zheng 已提交
1863 1864 1865 1866 1867 1868 1869 1870 1871 1872 1873 1874 1875 1876 1877
static int insert_reserved_file_extent(struct btrfs_trans_handle *trans,
				       struct inode *inode, u64 file_pos,
				       u64 disk_bytenr, u64 disk_num_bytes,
				       u64 num_bytes, u64 ram_bytes,
				       u8 compression, u8 encryption,
				       u16 other_encoding, int extent_type)
{
	struct btrfs_root *root = BTRFS_I(inode)->root;
	struct btrfs_file_extent_item *fi;
	struct btrfs_path *path;
	struct extent_buffer *leaf;
	struct btrfs_key ins;
	int ret;

	path = btrfs_alloc_path();
1878 1879
	if (!path)
		return -ENOMEM;
Y
Yan Zheng 已提交
1880

1881
	path->leave_spinning = 1;
C
Chris Mason 已提交
1882 1883 1884 1885 1886 1887 1888 1889 1890 1891

	/*
	 * 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.
	 */
J
Josef Bacik 已提交
1892
	ret = btrfs_drop_extents(trans, root, inode, file_pos,
1893
				 file_pos + num_bytes, 0);
1894 1895
	if (ret)
		goto out;
Y
Yan Zheng 已提交
1896

L
Li Zefan 已提交
1897
	ins.objectid = btrfs_ino(inode);
Y
Yan Zheng 已提交
1898 1899 1900
	ins.offset = file_pos;
	ins.type = BTRFS_EXTENT_DATA_KEY;
	ret = btrfs_insert_empty_item(trans, root, path, &ins, sizeof(*fi));
1901 1902
	if (ret)
		goto out;
Y
Yan Zheng 已提交
1903 1904 1905 1906 1907 1908 1909 1910 1911 1912 1913 1914 1915
	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);
1916

Y
Yan Zheng 已提交
1917
	btrfs_mark_buffer_dirty(leaf);
1918
	btrfs_release_path(path);
Y
Yan Zheng 已提交
1919 1920 1921 1922 1923 1924

	inode_add_bytes(inode, num_bytes);

	ins.objectid = disk_bytenr;
	ins.offset = disk_num_bytes;
	ins.type = BTRFS_EXTENT_ITEM_KEY;
1925 1926
	ret = btrfs_alloc_reserved_file_extent(trans, root,
					root->root_key.objectid,
L
Li Zefan 已提交
1927
					btrfs_ino(inode), file_pos, &ins);
1928
out:
Y
Yan Zheng 已提交
1929
	btrfs_free_path(path);
1930

1931
	return ret;
Y
Yan Zheng 已提交
1932 1933
}

1934 1935 1936 1937 1938 1939
/*
 * helper function for btrfs_finish_ordered_io, this
 * just reads in some of the csum leaves to prime them into ram
 * before we start the transaction.  It limits the amount of btree
 * reads required while inside the transaction.
 */
C
Chris Mason 已提交
1940 1941 1942 1943
/* 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.
 */
1944
static int btrfs_finish_ordered_io(struct btrfs_ordered_extent *ordered_extent)
1945
{
1946
	struct inode *inode = ordered_extent->inode;
1947
	struct btrfs_root *root = BTRFS_I(inode)->root;
1948
	struct btrfs_trans_handle *trans = NULL;
1949
	struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
1950
	struct extent_state *cached_state = NULL;
1951
	int compress_type = 0;
1952
	int ret;
1953
	bool nolock;
1954

1955
	nolock = btrfs_is_free_space_inode(inode);
J
Josef Bacik 已提交
1956

1957 1958 1959 1960 1961
	if (test_bit(BTRFS_ORDERED_IOERR, &ordered_extent->flags)) {
		ret = -EIO;
		goto out;
	}

1962
	if (test_bit(BTRFS_ORDERED_NOCOW, &ordered_extent->flags)) {
1963
		BUG_ON(!list_empty(&ordered_extent->list)); /* Logic error */
1964 1965 1966 1967 1968 1969 1970 1971 1972
		btrfs_ordered_update_i_size(inode, 0, ordered_extent);
		if (nolock)
			trans = btrfs_join_transaction_nolock(root);
		else
			trans = btrfs_join_transaction(root);
		if (IS_ERR(trans)) {
			ret = PTR_ERR(trans);
			trans = NULL;
			goto out;
1973
		}
1974 1975 1976 1977
		trans->block_rsv = &root->fs_info->delalloc_block_rsv;
		ret = btrfs_update_inode_fallback(trans, root, inode);
		if (ret) /* -ENOMEM or corruption */
			btrfs_abort_transaction(trans, root, ret);
1978 1979
		goto out;
	}
1980

1981 1982
	lock_extent_bits(io_tree, ordered_extent->file_offset,
			 ordered_extent->file_offset + ordered_extent->len - 1,
1983
			 0, &cached_state);
1984

J
Josef Bacik 已提交
1985
	if (nolock)
1986
		trans = btrfs_join_transaction_nolock(root);
J
Josef Bacik 已提交
1987
	else
1988
		trans = btrfs_join_transaction(root);
1989 1990 1991 1992 1993
	if (IS_ERR(trans)) {
		ret = PTR_ERR(trans);
		trans = NULL;
		goto out_unlock;
	}
1994
	trans->block_rsv = &root->fs_info->delalloc_block_rsv;
1995

C
Chris Mason 已提交
1996
	if (test_bit(BTRFS_ORDERED_COMPRESSED, &ordered_extent->flags))
1997
		compress_type = ordered_extent->compress_type;
Y
Yan Zheng 已提交
1998
	if (test_bit(BTRFS_ORDERED_PREALLOC, &ordered_extent->flags)) {
1999
		BUG_ON(compress_type);
Y
Yan, Zheng 已提交
2000
		ret = btrfs_mark_extent_written(trans, inode,
Y
Yan Zheng 已提交
2001 2002 2003 2004
						ordered_extent->file_offset,
						ordered_extent->file_offset +
						ordered_extent->len);
	} else {
2005
		BUG_ON(root == root->fs_info->tree_root);
Y
Yan Zheng 已提交
2006 2007 2008 2009 2010 2011
		ret = insert_reserved_file_extent(trans, inode,
						ordered_extent->file_offset,
						ordered_extent->start,
						ordered_extent->disk_len,
						ordered_extent->len,
						ordered_extent->len,
2012
						compress_type, 0, 0,
Y
Yan Zheng 已提交
2013 2014
						BTRFS_FILE_EXTENT_REG);
	}
J
Josef Bacik 已提交
2015 2016 2017
	unpin_extent_cache(&BTRFS_I(inode)->extent_tree,
			   ordered_extent->file_offset, ordered_extent->len,
			   trans->transid);
2018 2019
	if (ret < 0) {
		btrfs_abort_transaction(trans, root, ret);
2020
		goto out_unlock;
2021
	}
2022

2023 2024 2025
	add_pending_csums(trans, inode, ordered_extent->file_offset,
			  &ordered_extent->list);

2026 2027 2028 2029 2030
	btrfs_ordered_update_i_size(inode, 0, ordered_extent);
	ret = btrfs_update_inode_fallback(trans, root, inode);
	if (ret) { /* -ENOMEM or corruption */
		btrfs_abort_transaction(trans, root, ret);
		goto out_unlock;
2031 2032
	}
	ret = 0;
2033 2034 2035 2036
out_unlock:
	unlock_extent_cached(io_tree, ordered_extent->file_offset,
			     ordered_extent->file_offset +
			     ordered_extent->len - 1, &cached_state, GFP_NOFS);
2037
out:
2038
	if (root != root->fs_info->tree_root)
J
Josef Bacik 已提交
2039
		btrfs_delalloc_release_metadata(inode, ordered_extent->len);
2040 2041
	if (trans)
		btrfs_end_transaction(trans, root);
J
Josef Bacik 已提交
2042

2043
	if (ret) {
2044 2045 2046 2047
		clear_extent_uptodate(io_tree, ordered_extent->file_offset,
				      ordered_extent->file_offset +
				      ordered_extent->len - 1, NULL, GFP_NOFS);

2048 2049 2050 2051 2052 2053 2054 2055 2056 2057 2058 2059
		/*
		 * If the ordered extent had an IOERR or something else went
		 * wrong we need to return the space for this ordered extent
		 * back to the allocator.
		 */
		if (!test_bit(BTRFS_ORDERED_NOCOW, &ordered_extent->flags) &&
		    !test_bit(BTRFS_ORDERED_PREALLOC, &ordered_extent->flags))
			btrfs_free_reserved_extent(root, ordered_extent->start,
						   ordered_extent->disk_len);
	}


2060
	/*
2061 2062
	 * This needs to be done to make sure anybody waiting knows we are done
	 * updating everything for this ordered extent.
2063 2064 2065
	 */
	btrfs_remove_ordered_extent(inode, ordered_extent);

2066 2067 2068 2069 2070
	/* once for us */
	btrfs_put_ordered_extent(ordered_extent);
	/* once for the tree */
	btrfs_put_ordered_extent(ordered_extent);

2071 2072 2073 2074 2075 2076 2077 2078
	return ret;
}

static void finish_ordered_fn(struct btrfs_work *work)
{
	struct btrfs_ordered_extent *ordered_extent;
	ordered_extent = container_of(work, struct btrfs_ordered_extent, work);
	btrfs_finish_ordered_io(ordered_extent);
2079 2080
}

2081
static int btrfs_writepage_end_io_hook(struct page *page, u64 start, u64 end,
2082 2083
				struct extent_state *state, int uptodate)
{
2084 2085 2086 2087 2088
	struct inode *inode = page->mapping->host;
	struct btrfs_root *root = BTRFS_I(inode)->root;
	struct btrfs_ordered_extent *ordered_extent = NULL;
	struct btrfs_workers *workers;

2089 2090
	trace_btrfs_writepage_end_io_hook(page, start, end, uptodate);

2091
	ClearPagePrivate2(page);
2092 2093 2094 2095 2096 2097 2098
	if (!btrfs_dec_test_ordered_pending(inode, &ordered_extent, start,
					    end - start + 1, uptodate))
		return 0;

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

2099
	if (btrfs_is_free_space_inode(inode))
2100 2101 2102 2103 2104 2105
		workers = &root->fs_info->endio_freespace_worker;
	else
		workers = &root->fs_info->endio_write_workers;
	btrfs_queue_worker(workers, &ordered_extent->work);

	return 0;
2106 2107
}

C
Chris Mason 已提交
2108 2109
/*
 * when reads are done, we need to check csums to verify the data is correct
2110 2111
 * if there's a match, we allow the bio to finish.  If not, the code in
 * extent_io.c will try to find good copies for us.
C
Chris Mason 已提交
2112
 */
2113
static int btrfs_readpage_end_io_hook(struct page *page, u64 start, u64 end,
2114
			       struct extent_state *state, int mirror)
2115
{
M
Miao Xie 已提交
2116
	size_t offset = start - page_offset(page);
2117
	struct inode *inode = page->mapping->host;
2118
	struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
2119
	char *kaddr;
2120
	u64 private = ~(u32)0;
2121
	int ret;
2122 2123
	struct btrfs_root *root = BTRFS_I(inode)->root;
	u32 csum = ~(u32)0;
2124

2125 2126 2127 2128
	if (PageChecked(page)) {
		ClearPageChecked(page);
		goto good;
	}
2129 2130

	if (BTRFS_I(inode)->flags & BTRFS_INODE_NODATASUM)
2131
		goto good;
2132 2133

	if (root->root_key.objectid == BTRFS_DATA_RELOC_TREE_OBJECTID &&
2134
	    test_range_bit(io_tree, start, end, EXTENT_NODATASUM, 1, NULL)) {
2135 2136
		clear_extent_bits(io_tree, start, end, EXTENT_NODATASUM,
				  GFP_NOFS);
2137
		return 0;
2138
	}
2139

Y
Yan 已提交
2140
	if (state && state->start == start) {
2141 2142 2143 2144 2145
		private = state->private;
		ret = 0;
	} else {
		ret = get_state_private(io_tree, start, &private);
	}
2146
	kaddr = kmap_atomic(page);
C
Chris Mason 已提交
2147
	if (ret)
2148
		goto zeroit;
C
Chris Mason 已提交
2149

2150 2151
	csum = btrfs_csum_data(root, kaddr + offset, csum,  end - start + 1);
	btrfs_csum_final(csum, (char *)&csum);
C
Chris Mason 已提交
2152
	if (csum != private)
2153
		goto zeroit;
C
Chris Mason 已提交
2154

2155
	kunmap_atomic(kaddr);
2156
good:
2157 2158 2159
	return 0;

zeroit:
2160
	printk_ratelimited(KERN_INFO "btrfs csum failed ino %llu off %llu csum %u "
L
Li Zefan 已提交
2161 2162
		       "private %llu\n",
		       (unsigned long long)btrfs_ino(page->mapping->host),
C
Chris Mason 已提交
2163 2164
		       (unsigned long long)start, csum,
		       (unsigned long long)private);
2165 2166
	memset(kaddr + offset, 1, end - start + 1);
	flush_dcache_page(page);
2167
	kunmap_atomic(kaddr);
2168 2169
	if (private == 0)
		return 0;
2170
	return -EIO;
2171
}
2172

Y
Yan, Zheng 已提交
2173 2174 2175 2176 2177
struct delayed_iput {
	struct list_head list;
	struct inode *inode;
};

2178 2179
/* JDM: If this is fs-wide, why can't we add a pointer to
 * btrfs_inode instead and avoid the allocation? */
Y
Yan, Zheng 已提交
2180 2181 2182 2183 2184 2185 2186 2187 2188 2189 2190 2191 2192 2193 2194 2195 2196 2197 2198 2199 2200 2201 2202 2203 2204 2205 2206 2207 2208 2209 2210 2211 2212 2213 2214 2215 2216 2217 2218 2219 2220
void btrfs_add_delayed_iput(struct inode *inode)
{
	struct btrfs_fs_info *fs_info = BTRFS_I(inode)->root->fs_info;
	struct delayed_iput *delayed;

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

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

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

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

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

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

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

2221
/*
2222
 * This is called in transaction commit time. If there are no orphan
2223 2224 2225 2226 2227 2228
 * files in the subvolume, it removes orphan item and frees block_rsv
 * structure.
 */
void btrfs_orphan_commit_root(struct btrfs_trans_handle *trans,
			      struct btrfs_root *root)
{
2229
	struct btrfs_block_rsv *block_rsv;
2230 2231
	int ret;

2232
	if (atomic_read(&root->orphan_inodes) ||
2233 2234 2235
	    root->orphan_cleanup_state != ORPHAN_CLEANUP_DONE)
		return;

2236
	spin_lock(&root->orphan_lock);
2237
	if (atomic_read(&root->orphan_inodes)) {
2238 2239 2240 2241 2242 2243 2244 2245 2246 2247 2248 2249 2250
		spin_unlock(&root->orphan_lock);
		return;
	}

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

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

2251 2252 2253 2254 2255 2256 2257 2258
	if (root->orphan_item_inserted &&
	    btrfs_root_refs(&root->root_item) > 0) {
		ret = btrfs_del_orphan_item(trans, root->fs_info->tree_root,
					    root->root_key.objectid);
		BUG_ON(ret);
		root->orphan_item_inserted = 0;
	}

2259 2260 2261
	if (block_rsv) {
		WARN_ON(block_rsv->size > 0);
		btrfs_free_block_rsv(root, block_rsv);
2262 2263 2264
	}
}

2265 2266 2267
/*
 * This creates an orphan entry for the given inode in case something goes
 * wrong in the middle of an unlink/truncate.
2268 2269 2270
 *
 * NOTE: caller of this function should reserve 5 units of metadata for
 *	 this function.
2271 2272 2273 2274
 */
int btrfs_orphan_add(struct btrfs_trans_handle *trans, struct inode *inode)
{
	struct btrfs_root *root = BTRFS_I(inode)->root;
2275 2276 2277 2278
	struct btrfs_block_rsv *block_rsv = NULL;
	int reserve = 0;
	int insert = 0;
	int ret;
2279

2280
	if (!root->orphan_block_rsv) {
2281
		block_rsv = btrfs_alloc_block_rsv(root, BTRFS_BLOCK_RSV_TEMP);
2282 2283
		if (!block_rsv)
			return -ENOMEM;
2284
	}
2285

2286 2287 2288 2289 2290 2291
	spin_lock(&root->orphan_lock);
	if (!root->orphan_block_rsv) {
		root->orphan_block_rsv = block_rsv;
	} else if (block_rsv) {
		btrfs_free_block_rsv(root, block_rsv);
		block_rsv = NULL;
2292 2293
	}

2294 2295
	if (!test_and_set_bit(BTRFS_INODE_HAS_ORPHAN_ITEM,
			      &BTRFS_I(inode)->runtime_flags)) {
2296 2297 2298 2299 2300 2301 2302 2303 2304 2305 2306 2307
#if 0
		/*
		 * For proper ENOSPC handling, we should do orphan
		 * cleanup when mounting. But this introduces backward
		 * compatibility issue.
		 */
		if (!xchg(&root->orphan_item_inserted, 1))
			insert = 2;
		else
			insert = 1;
#endif
		insert = 1;
2308
		atomic_inc(&root->orphan_inodes);
2309 2310
	}

2311 2312
	if (!test_and_set_bit(BTRFS_INODE_ORPHAN_META_RESERVED,
			      &BTRFS_I(inode)->runtime_flags))
2313 2314
		reserve = 1;
	spin_unlock(&root->orphan_lock);
2315

2316 2317 2318
	/* grab metadata reservation from transaction handle */
	if (reserve) {
		ret = btrfs_orphan_reserve_metadata(trans, inode);
2319
		BUG_ON(ret); /* -ENOSPC in reservation; Logic error? JDM */
2320
	}
2321

2322 2323
	/* insert an orphan item to track this unlinked/truncated file */
	if (insert >= 1) {
L
Li Zefan 已提交
2324
		ret = btrfs_insert_orphan_item(trans, root, btrfs_ino(inode));
2325
		if (ret && ret != -EEXIST) {
2326 2327
			clear_bit(BTRFS_INODE_HAS_ORPHAN_ITEM,
				  &BTRFS_I(inode)->runtime_flags);
2328 2329 2330 2331
			btrfs_abort_transaction(trans, root, ret);
			return ret;
		}
		ret = 0;
2332 2333 2334 2335 2336 2337
	}

	/* insert an orphan item to track subvolume contains orphan files */
	if (insert >= 2) {
		ret = btrfs_insert_orphan_item(trans, root->fs_info->tree_root,
					       root->root_key.objectid);
2338 2339 2340 2341
		if (ret && ret != -EEXIST) {
			btrfs_abort_transaction(trans, root, ret);
			return ret;
		}
2342 2343
	}
	return 0;
2344 2345 2346 2347 2348 2349 2350 2351 2352
}

/*
 * 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;
2353 2354
	int delete_item = 0;
	int release_rsv = 0;
2355 2356
	int ret = 0;

2357
	spin_lock(&root->orphan_lock);
2358 2359
	if (test_and_clear_bit(BTRFS_INODE_HAS_ORPHAN_ITEM,
			       &BTRFS_I(inode)->runtime_flags))
2360
		delete_item = 1;
2361

2362 2363
	if (test_and_clear_bit(BTRFS_INODE_ORPHAN_META_RESERVED,
			       &BTRFS_I(inode)->runtime_flags))
2364 2365
		release_rsv = 1;
	spin_unlock(&root->orphan_lock);
2366

2367
	if (trans && delete_item) {
L
Li Zefan 已提交
2368
		ret = btrfs_del_orphan_item(trans, root, btrfs_ino(inode));
2369
		BUG_ON(ret); /* -ENOMEM or corruption (JDM: Recheck) */
2370
	}
2371

2372
	if (release_rsv) {
2373
		btrfs_orphan_release_metadata(inode);
2374 2375
		atomic_dec(&root->orphan_inodes);
	}
2376

2377
	return 0;
2378 2379 2380 2381 2382 2383
}

/*
 * this cleans up any orphans that may be left on the list from the last use
 * of this root.
 */
2384
int btrfs_orphan_cleanup(struct btrfs_root *root)
2385 2386 2387 2388 2389 2390
{
	struct btrfs_path *path;
	struct extent_buffer *leaf;
	struct btrfs_key key, found_key;
	struct btrfs_trans_handle *trans;
	struct inode *inode;
2391
	u64 last_objectid = 0;
2392 2393
	int ret = 0, nr_unlink = 0, nr_truncate = 0;

2394
	if (cmpxchg(&root->orphan_cleanup_state, 0, ORPHAN_CLEANUP_STARTED))
2395
		return 0;
2396 2397

	path = btrfs_alloc_path();
2398 2399 2400 2401
	if (!path) {
		ret = -ENOMEM;
		goto out;
	}
2402 2403 2404 2405 2406 2407 2408 2409
	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);
2410 2411
		if (ret < 0)
			goto out;
2412 2413 2414

		/*
		 * if ret == 0 means we found what we were searching for, which
L
Lucas De Marchi 已提交
2415
		 * is weird, but possible, so only screw with path if we didn't
2416 2417 2418
		 * find the key and see if we have stuff that matches
		 */
		if (ret > 0) {
2419
			ret = 0;
2420 2421 2422 2423 2424 2425 2426 2427 2428 2429 2430 2431 2432 2433 2434 2435
			if (path->slots[0] == 0)
				break;
			path->slots[0]--;
		}

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

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

		/* release the path since we're done with it */
2436
		btrfs_release_path(path);
2437 2438 2439 2440 2441 2442

		/*
		 * 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.
		 */
2443 2444 2445 2446 2447 2448 2449 2450 2451 2452

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

		last_objectid = found_key.offset;

2453 2454 2455
		found_key.objectid = found_key.offset;
		found_key.type = BTRFS_INODE_ITEM_KEY;
		found_key.offset = 0;
2456
		inode = btrfs_iget(root->fs_info->sb, &found_key, root, NULL);
J
Josef Bacik 已提交
2457 2458
		ret = PTR_RET(inode);
		if (ret && ret != -ESTALE)
2459
			goto out;
2460

2461 2462 2463 2464 2465 2466 2467 2468 2469 2470 2471 2472 2473 2474 2475 2476 2477 2478 2479 2480 2481 2482 2483 2484 2485 2486 2487 2488 2489 2490 2491 2492
		if (ret == -ESTALE && root == root->fs_info->tree_root) {
			struct btrfs_root *dead_root;
			struct btrfs_fs_info *fs_info = root->fs_info;
			int is_dead_root = 0;

			/*
			 * this is an orphan in the tree root. Currently these
			 * could come from 2 sources:
			 *  a) a snapshot deletion in progress
			 *  b) a free space cache inode
			 * We need to distinguish those two, as the snapshot
			 * orphan must not get deleted.
			 * find_dead_roots already ran before us, so if this
			 * is a snapshot deletion, we should find the root
			 * in the dead_roots list
			 */
			spin_lock(&fs_info->trans_lock);
			list_for_each_entry(dead_root, &fs_info->dead_roots,
					    root_list) {
				if (dead_root->root_key.objectid ==
				    found_key.objectid) {
					is_dead_root = 1;
					break;
				}
			}
			spin_unlock(&fs_info->trans_lock);
			if (is_dead_root) {
				/* prevent this orphan from being found again */
				key.offset = found_key.objectid - 1;
				continue;
			}
		}
2493
		/*
J
Josef Bacik 已提交
2494 2495
		 * Inode is already gone but the orphan item is still there,
		 * kill the orphan item.
2496
		 */
J
Josef Bacik 已提交
2497 2498
		if (ret == -ESTALE) {
			trans = btrfs_start_transaction(root, 1);
2499 2500 2501 2502
			if (IS_ERR(trans)) {
				ret = PTR_ERR(trans);
				goto out;
			}
2503 2504
			printk(KERN_ERR "auto deleting %Lu\n",
			       found_key.objectid);
J
Josef Bacik 已提交
2505 2506
			ret = btrfs_del_orphan_item(trans, root,
						    found_key.objectid);
2507
			BUG_ON(ret); /* -ENOMEM or corruption (JDM: Recheck) */
2508
			btrfs_end_transaction(trans, root);
2509 2510 2511
			continue;
		}

J
Josef Bacik 已提交
2512 2513 2514 2515
		/*
		 * add this inode to the orphan list so btrfs_orphan_del does
		 * the proper thing when we hit it
		 */
2516 2517
		set_bit(BTRFS_INODE_HAS_ORPHAN_ITEM,
			&BTRFS_I(inode)->runtime_flags);
2518
		atomic_inc(&root->orphan_inodes);
J
Josef Bacik 已提交
2519

2520 2521
		/* if we have links, this was a truncate, lets do that */
		if (inode->i_nlink) {
2522 2523 2524 2525 2526
			if (!S_ISREG(inode->i_mode)) {
				WARN_ON(1);
				iput(inode);
				continue;
			}
2527
			nr_truncate++;
2528 2529 2530 2531 2532 2533 2534 2535 2536 2537 2538 2539

			/* 1 for the orphan item deletion. */
			trans = btrfs_start_transaction(root, 1);
			if (IS_ERR(trans)) {
				ret = PTR_ERR(trans);
				goto out;
			}
			ret = btrfs_orphan_add(trans, inode);
			btrfs_end_transaction(trans, root);
			if (ret)
				goto out;

2540
			ret = btrfs_truncate(inode);
2541 2542
			if (ret)
				btrfs_orphan_del(NULL, inode);
2543 2544 2545 2546 2547 2548
		} else {
			nr_unlink++;
		}

		/* this will do delete_inode and everything for us */
		iput(inode);
2549 2550
		if (ret)
			goto out;
2551
	}
2552 2553 2554
	/* release the path since we're done with it */
	btrfs_release_path(path);

2555 2556 2557 2558 2559 2560 2561
	root->orphan_cleanup_state = ORPHAN_CLEANUP_DONE;

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

	if (root->orphan_block_rsv || root->orphan_item_inserted) {
2562
		trans = btrfs_join_transaction(root);
2563 2564
		if (!IS_ERR(trans))
			btrfs_end_transaction(trans, root);
2565
	}
2566 2567 2568 2569 2570

	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);
2571 2572 2573 2574 2575 2576

out:
	if (ret)
		printk(KERN_CRIT "btrfs: could not do orphan cleanup %d\n", ret);
	btrfs_free_path(path);
	return ret;
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 2607 2608 2609 2610 2611 2612 2613 2614 2615 2616 2617 2618 2619 2620 2621 2622 2623 2624 2625 2626 2627 2628 2629
/*
 * 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
2630 2631 2632
/*
 * read an inode from the btree into the in-memory inode
 */
2633
static void btrfs_read_locked_inode(struct inode *inode)
C
Chris Mason 已提交
2634 2635
{
	struct btrfs_path *path;
2636
	struct extent_buffer *leaf;
C
Chris Mason 已提交
2637
	struct btrfs_inode_item *inode_item;
2638
	struct btrfs_timespec *tspec;
C
Chris Mason 已提交
2639 2640
	struct btrfs_root *root = BTRFS_I(inode)->root;
	struct btrfs_key location;
2641
	int maybe_acls;
J
Josef Bacik 已提交
2642
	u32 rdev;
C
Chris Mason 已提交
2643
	int ret;
2644 2645 2646 2647 2648
	bool filled = false;

	ret = btrfs_fill_inode(inode, &rdev);
	if (!ret)
		filled = true;
C
Chris Mason 已提交
2649 2650

	path = btrfs_alloc_path();
2651 2652 2653
	if (!path)
		goto make_bad;

2654
	path->leave_spinning = 1;
C
Chris Mason 已提交
2655
	memcpy(&location, &BTRFS_I(inode)->location, sizeof(location));
C
Chris Mason 已提交
2656

C
Chris Mason 已提交
2657
	ret = btrfs_lookup_inode(NULL, root, path, &location, 0);
2658
	if (ret)
C
Chris Mason 已提交
2659 2660
		goto make_bad;

2661
	leaf = path->nodes[0];
2662 2663 2664 2665

	if (filled)
		goto cache_acl;

2666 2667 2668
	inode_item = btrfs_item_ptr(leaf, path->slots[0],
				    struct btrfs_inode_item);
	inode->i_mode = btrfs_inode_mode(leaf, inode_item);
M
Miklos Szeredi 已提交
2669
	set_nlink(inode, btrfs_inode_nlink(leaf, inode_item));
2670 2671
	i_uid_write(inode, btrfs_inode_uid(leaf, inode_item));
	i_gid_write(inode, btrfs_inode_gid(leaf, inode_item));
2672
	btrfs_i_size_write(inode, btrfs_inode_size(leaf, inode_item));
2673 2674 2675 2676 2677 2678 2679 2680 2681 2682 2683 2684 2685

	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);

2686
	inode_set_bytes(inode, btrfs_inode_nbytes(leaf, inode_item));
2687
	BTRFS_I(inode)->generation = btrfs_inode_generation(leaf, inode_item);
J
Josef Bacik 已提交
2688 2689 2690 2691 2692 2693 2694 2695 2696 2697 2698 2699
	BTRFS_I(inode)->last_trans = btrfs_inode_transid(leaf, inode_item);

	/*
	 * If we were modified in the current generation and evicted from memory
	 * and then re-read we need to do a full sync since we don't have any
	 * idea about which extents were modified before we were evicted from
	 * cache.
	 */
	if (BTRFS_I(inode)->last_trans == root->fs_info->generation)
		set_bit(BTRFS_INODE_NEEDS_FULL_SYNC,
			&BTRFS_I(inode)->runtime_flags);

2700
	inode->i_version = btrfs_inode_sequence(leaf, inode_item);
2701
	inode->i_generation = BTRFS_I(inode)->generation;
J
Josef Bacik 已提交
2702
	inode->i_rdev = 0;
2703 2704
	rdev = btrfs_inode_rdev(leaf, inode_item);

2705
	BTRFS_I(inode)->index_cnt = (u64)-1;
2706
	BTRFS_I(inode)->flags = btrfs_inode_flags(leaf, inode_item);
2707
cache_acl:
2708 2709 2710 2711
	/*
	 * try to precache a NULL acl entry for files that don't have
	 * any xattrs or acls
	 */
L
Li Zefan 已提交
2712 2713
	maybe_acls = acls_after_inode_item(leaf, path->slots[0],
					   btrfs_ino(inode));
2714 2715
	if (!maybe_acls)
		cache_no_acl(inode);
2716

C
Chris Mason 已提交
2717 2718 2719 2720 2721
	btrfs_free_path(path);

	switch (inode->i_mode & S_IFMT) {
	case S_IFREG:
		inode->i_mapping->a_ops = &btrfs_aops;
C
Chris Mason 已提交
2722
		inode->i_mapping->backing_dev_info = &root->fs_info->bdi;
2723
		BTRFS_I(inode)->io_tree.ops = &btrfs_extent_io_ops;
C
Chris Mason 已提交
2724 2725 2726 2727 2728 2729 2730 2731 2732 2733 2734 2735 2736
		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
2737
		inode->i_mapping->backing_dev_info = &root->fs_info->bdi;
C
Chris Mason 已提交
2738
		break;
J
Josef Bacik 已提交
2739
	default:
J
Jim Owens 已提交
2740
		inode->i_op = &btrfs_special_inode_operations;
J
Josef Bacik 已提交
2741 2742
		init_special_inode(inode, inode->i_mode, rdev);
		break;
C
Chris Mason 已提交
2743
	}
2744 2745

	btrfs_update_iflags(inode);
C
Chris Mason 已提交
2746 2747 2748 2749 2750 2751 2752
	return;

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

C
Chris Mason 已提交
2753 2754 2755
/*
 * given a leaf and an inode, copy the inode fields into the leaf
 */
2756 2757
static void fill_inode_item(struct btrfs_trans_handle *trans,
			    struct extent_buffer *leaf,
2758
			    struct btrfs_inode_item *item,
C
Chris Mason 已提交
2759 2760
			    struct inode *inode)
{
2761 2762 2763 2764 2765 2766 2767 2768 2769 2770 2771 2772 2773 2774 2775 2776 2777 2778 2779 2780 2781 2782 2783 2784 2785 2786 2787 2788 2789 2790 2791 2792 2793 2794 2795
	struct btrfs_map_token token;

	btrfs_init_map_token(&token);

	btrfs_set_token_inode_uid(leaf, item, i_uid_read(inode), &token);
	btrfs_set_token_inode_gid(leaf, item, i_gid_read(inode), &token);
	btrfs_set_token_inode_size(leaf, item, BTRFS_I(inode)->disk_i_size,
				   &token);
	btrfs_set_token_inode_mode(leaf, item, inode->i_mode, &token);
	btrfs_set_token_inode_nlink(leaf, item, inode->i_nlink, &token);

	btrfs_set_token_timespec_sec(leaf, btrfs_inode_atime(item),
				     inode->i_atime.tv_sec, &token);
	btrfs_set_token_timespec_nsec(leaf, btrfs_inode_atime(item),
				      inode->i_atime.tv_nsec, &token);

	btrfs_set_token_timespec_sec(leaf, btrfs_inode_mtime(item),
				     inode->i_mtime.tv_sec, &token);
	btrfs_set_token_timespec_nsec(leaf, btrfs_inode_mtime(item),
				      inode->i_mtime.tv_nsec, &token);

	btrfs_set_token_timespec_sec(leaf, btrfs_inode_ctime(item),
				     inode->i_ctime.tv_sec, &token);
	btrfs_set_token_timespec_nsec(leaf, btrfs_inode_ctime(item),
				      inode->i_ctime.tv_nsec, &token);

	btrfs_set_token_inode_nbytes(leaf, item, inode_get_bytes(inode),
				     &token);
	btrfs_set_token_inode_generation(leaf, item, BTRFS_I(inode)->generation,
					 &token);
	btrfs_set_token_inode_sequence(leaf, item, inode->i_version, &token);
	btrfs_set_token_inode_transid(leaf, item, trans->transid, &token);
	btrfs_set_token_inode_rdev(leaf, item, inode->i_rdev, &token);
	btrfs_set_token_inode_flags(leaf, item, BTRFS_I(inode)->flags, &token);
	btrfs_set_token_inode_block_group(leaf, item, 0, &token);
C
Chris Mason 已提交
2796 2797
}

C
Chris Mason 已提交
2798 2799 2800
/*
 * copy everything in the in-memory inode into the btree.
 */
2801
static noinline int btrfs_update_inode_item(struct btrfs_trans_handle *trans,
C
Chris Mason 已提交
2802
				struct btrfs_root *root, struct inode *inode)
C
Chris Mason 已提交
2803 2804 2805
{
	struct btrfs_inode_item *inode_item;
	struct btrfs_path *path;
2806
	struct extent_buffer *leaf;
C
Chris Mason 已提交
2807 2808 2809
	int ret;

	path = btrfs_alloc_path();
2810 2811 2812
	if (!path)
		return -ENOMEM;

2813
	path->leave_spinning = 1;
2814 2815
	ret = btrfs_lookup_inode(trans, root, path, &BTRFS_I(inode)->location,
				 1);
C
Chris Mason 已提交
2816 2817 2818 2819 2820 2821
	if (ret) {
		if (ret > 0)
			ret = -ENOENT;
		goto failed;
	}

2822
	btrfs_unlock_up_safe(path, 1);
2823 2824
	leaf = path->nodes[0];
	inode_item = btrfs_item_ptr(leaf, path->slots[0],
2825
				    struct btrfs_inode_item);
C
Chris Mason 已提交
2826

2827
	fill_inode_item(trans, leaf, inode_item, inode);
2828
	btrfs_mark_buffer_dirty(leaf);
2829
	btrfs_set_inode_last_trans(trans, inode);
C
Chris Mason 已提交
2830 2831 2832 2833 2834 2835
	ret = 0;
failed:
	btrfs_free_path(path);
	return ret;
}

2836 2837 2838 2839 2840 2841 2842 2843 2844 2845 2846 2847 2848 2849 2850
/*
 * copy everything in the in-memory inode into the btree.
 */
noinline int btrfs_update_inode(struct btrfs_trans_handle *trans,
				struct btrfs_root *root, struct inode *inode)
{
	int ret;

	/*
	 * If the inode is a free space inode, we can deadlock during commit
	 * if we put it into the delayed code.
	 *
	 * The data relocation inode should also be directly updated
	 * without delay
	 */
2851
	if (!btrfs_is_free_space_inode(inode)
2852
	    && root->root_key.objectid != BTRFS_DATA_RELOC_TREE_OBJECTID) {
2853 2854
		btrfs_update_root_times(trans, root);

2855 2856 2857 2858 2859 2860 2861 2862 2863
		ret = btrfs_delayed_update_inode(trans, root, inode);
		if (!ret)
			btrfs_set_inode_last_trans(trans, inode);
		return ret;
	}

	return btrfs_update_inode_item(trans, root, inode);
}

2864 2865 2866
noinline int btrfs_update_inode_fallback(struct btrfs_trans_handle *trans,
					 struct btrfs_root *root,
					 struct inode *inode)
2867 2868 2869 2870 2871 2872 2873 2874 2875
{
	int ret;

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

C
Chris Mason 已提交
2876 2877 2878 2879 2880
/*
 * 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
 */
2881 2882 2883 2884
static int __btrfs_unlink_inode(struct btrfs_trans_handle *trans,
				struct btrfs_root *root,
				struct inode *dir, struct inode *inode,
				const char *name, int name_len)
C
Chris Mason 已提交
2885 2886 2887
{
	struct btrfs_path *path;
	int ret = 0;
2888
	struct extent_buffer *leaf;
C
Chris Mason 已提交
2889
	struct btrfs_dir_item *di;
2890
	struct btrfs_key key;
2891
	u64 index;
L
Li Zefan 已提交
2892 2893
	u64 ino = btrfs_ino(inode);
	u64 dir_ino = btrfs_ino(dir);
C
Chris Mason 已提交
2894 2895

	path = btrfs_alloc_path();
2896 2897
	if (!path) {
		ret = -ENOMEM;
2898
		goto out;
2899 2900
	}

2901
	path->leave_spinning = 1;
L
Li Zefan 已提交
2902
	di = btrfs_lookup_dir_item(trans, root, path, dir_ino,
C
Chris Mason 已提交
2903 2904 2905 2906 2907 2908 2909 2910 2911
				    name, name_len, -1);
	if (IS_ERR(di)) {
		ret = PTR_ERR(di);
		goto err;
	}
	if (!di) {
		ret = -ENOENT;
		goto err;
	}
2912 2913
	leaf = path->nodes[0];
	btrfs_dir_item_key_to_cpu(leaf, di, &key);
C
Chris Mason 已提交
2914
	ret = btrfs_delete_one_dir_name(trans, root, path, di);
2915 2916
	if (ret)
		goto err;
2917
	btrfs_release_path(path);
C
Chris Mason 已提交
2918

L
Li Zefan 已提交
2919 2920
	ret = btrfs_del_inode_ref(trans, root, name, name_len, ino,
				  dir_ino, &index);
2921
	if (ret) {
C
Chris Mason 已提交
2922
		printk(KERN_INFO "btrfs failed to delete reference to %.*s, "
L
Li Zefan 已提交
2923 2924
		       "inode %llu parent %llu\n", name_len, name,
		       (unsigned long long)ino, (unsigned long long)dir_ino);
2925
		btrfs_abort_transaction(trans, root, ret);
2926 2927 2928
		goto err;
	}

2929
	ret = btrfs_delete_delayed_dir_index(trans, root, dir, index);
2930 2931
	if (ret) {
		btrfs_abort_transaction(trans, root, ret);
C
Chris Mason 已提交
2932
		goto err;
2933
	}
C
Chris Mason 已提交
2934

2935
	ret = btrfs_del_inode_ref_in_log(trans, root, name, name_len,
L
Li Zefan 已提交
2936
					 inode, dir_ino);
2937 2938 2939 2940
	if (ret != 0 && ret != -ENOENT) {
		btrfs_abort_transaction(trans, root, ret);
		goto err;
	}
2941 2942 2943

	ret = btrfs_del_dir_entries_in_log(trans, root, name, name_len,
					   dir, index);
2944 2945
	if (ret == -ENOENT)
		ret = 0;
C
Chris Mason 已提交
2946 2947
err:
	btrfs_free_path(path);
2948 2949 2950 2951
	if (ret)
		goto out;

	btrfs_i_size_write(dir, dir->i_size - name_len * 2);
2952 2953
	inode_inc_iversion(inode);
	inode_inc_iversion(dir);
2954
	inode->i_ctime = dir->i_mtime = dir->i_ctime = CURRENT_TIME;
2955
	ret = btrfs_update_inode(trans, root, dir);
2956
out:
C
Chris Mason 已提交
2957 2958 2959
	return ret;
}

2960 2961 2962 2963 2964 2965 2966 2967 2968 2969 2970 2971 2972 2973 2974
int btrfs_unlink_inode(struct btrfs_trans_handle *trans,
		       struct btrfs_root *root,
		       struct inode *dir, struct inode *inode,
		       const char *name, int name_len)
{
	int ret;
	ret = __btrfs_unlink_inode(trans, root, dir, inode, name, name_len);
	if (!ret) {
		btrfs_drop_nlink(inode);
		ret = btrfs_update_inode(trans, root, inode);
	}
	return ret;
}
		

2975 2976 2977
/* helper to check if there is any shared block in the path */
static int check_path_shared(struct btrfs_root *root,
			     struct btrfs_path *path)
C
Chris Mason 已提交
2978
{
2979 2980
	struct extent_buffer *eb;
	int level;
2981
	u64 refs = 1;
2982

2983
	for (level = 0; level < BTRFS_MAX_LEVEL; level++) {
2984 2985
		int ret;

2986 2987 2988 2989 2990 2991 2992 2993 2994
		if (!path->nodes[level])
			break;
		eb = path->nodes[level];
		if (!btrfs_block_can_be_shared(root, eb))
			continue;
		ret = btrfs_lookup_extent_info(NULL, root, eb->start, eb->len,
					       &refs, NULL);
		if (refs > 1)
			return 1;
2995
	}
2996
	return 0;
C
Chris Mason 已提交
2997 2998
}

2999 3000 3001 3002 3003 3004 3005 3006 3007
/*
 * helper to start transaction for unlink and rmdir.
 *
 * unlink and rmdir are special in btrfs, they do not always free space.
 * so in enospc case, we should make sure they will free space before
 * allowing them to use the global metadata reservation.
 */
static struct btrfs_trans_handle *__unlink_start_trans(struct inode *dir,
						       struct dentry *dentry)
3008
{
C
Chris Mason 已提交
3009
	struct btrfs_trans_handle *trans;
3010
	struct btrfs_root *root = BTRFS_I(dir)->root;
3011 3012
	struct btrfs_path *path;
	struct btrfs_dir_item *di;
3013
	struct inode *inode = dentry->d_inode;
3014
	u64 index;
3015 3016
	int check_link = 1;
	int err = -ENOSPC;
3017
	int ret;
L
Li Zefan 已提交
3018 3019
	u64 ino = btrfs_ino(inode);
	u64 dir_ino = btrfs_ino(dir);
3020

3021 3022 3023 3024 3025 3026 3027 3028 3029 3030
	/*
	 * 1 for the possible orphan item
	 * 1 for the dir item
	 * 1 for the dir index
	 * 1 for the inode ref
	 * 1 for the inode ref in the tree log
	 * 2 for the dir entries in the log
	 * 1 for the inode
	 */
	trans = btrfs_start_transaction(root, 8);
3031 3032
	if (!IS_ERR(trans) || PTR_ERR(trans) != -ENOSPC)
		return trans;
3033

L
Li Zefan 已提交
3034
	if (ino == BTRFS_EMPTY_SUBVOL_DIR_OBJECTID)
3035
		return ERR_PTR(-ENOSPC);
3036

3037 3038 3039
	/* check if there is someone else holds reference */
	if (S_ISDIR(inode->i_mode) && atomic_read(&inode->i_count) > 1)
		return ERR_PTR(-ENOSPC);
3040

3041 3042
	if (atomic_read(&inode->i_count) > 2)
		return ERR_PTR(-ENOSPC);
3043

3044 3045 3046 3047 3048 3049 3050
	if (xchg(&root->fs_info->enospc_unlink, 1))
		return ERR_PTR(-ENOSPC);

	path = btrfs_alloc_path();
	if (!path) {
		root->fs_info->enospc_unlink = 0;
		return ERR_PTR(-ENOMEM);
3051 3052
	}

3053 3054
	/* 1 for the orphan item */
	trans = btrfs_start_transaction(root, 1);
3055
	if (IS_ERR(trans)) {
3056 3057 3058 3059
		btrfs_free_path(path);
		root->fs_info->enospc_unlink = 0;
		return trans;
	}
3060

3061 3062
	path->skip_locking = 1;
	path->search_commit_root = 1;
3063

3064 3065 3066 3067 3068 3069 3070 3071 3072 3073 3074
	ret = btrfs_lookup_inode(trans, root, path,
				&BTRFS_I(dir)->location, 0);
	if (ret < 0) {
		err = ret;
		goto out;
	}
	if (ret == 0) {
		if (check_path_shared(root, path))
			goto out;
	} else {
		check_link = 0;
3075
	}
3076
	btrfs_release_path(path);
3077 3078 3079 3080 3081 3082 3083 3084 3085 3086 3087 3088 3089

	ret = btrfs_lookup_inode(trans, root, path,
				&BTRFS_I(inode)->location, 0);
	if (ret < 0) {
		err = ret;
		goto out;
	}
	if (ret == 0) {
		if (check_path_shared(root, path))
			goto out;
	} else {
		check_link = 0;
	}
3090
	btrfs_release_path(path);
3091 3092 3093

	if (ret == 0 && S_ISREG(inode->i_mode)) {
		ret = btrfs_lookup_file_extent(trans, root, path,
L
Li Zefan 已提交
3094
					       ino, (u64)-1, 0);
3095 3096 3097 3098
		if (ret < 0) {
			err = ret;
			goto out;
		}
3099
		BUG_ON(ret == 0); /* Corruption */
3100 3101
		if (check_path_shared(root, path))
			goto out;
3102
		btrfs_release_path(path);
3103 3104 3105 3106 3107 3108 3109
	}

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

L
Li Zefan 已提交
3110
	di = btrfs_lookup_dir_item(trans, root, path, dir_ino,
3111 3112 3113 3114 3115 3116 3117 3118 3119 3120 3121 3122
				dentry->d_name.name, dentry->d_name.len, 0);
	if (IS_ERR(di)) {
		err = PTR_ERR(di);
		goto out;
	}
	if (di) {
		if (check_path_shared(root, path))
			goto out;
	} else {
		err = 0;
		goto out;
	}
3123
	btrfs_release_path(path);
3124

M
Mark Fasheh 已提交
3125 3126 3127 3128 3129
	ret = btrfs_get_inode_ref_index(trans, root, path, dentry->d_name.name,
					dentry->d_name.len, ino, dir_ino, 0,
					&index);
	if (ret) {
		err = ret;
3130 3131
		goto out;
	}
M
Mark Fasheh 已提交
3132

3133 3134
	if (check_path_shared(root, path))
		goto out;
M
Mark Fasheh 已提交
3135

3136
	btrfs_release_path(path);
3137

3138 3139 3140 3141 3142 3143 3144 3145
	/*
	 * This is a commit root search, if we can lookup inode item and other
	 * relative items in the commit root, it means the transaction of
	 * dir/file creation has been committed, and the dir index item that we
	 * delay to insert has also been inserted into the commit root. So
	 * we needn't worry about the delayed insertion of the dir index item
	 * here.
	 */
L
Li Zefan 已提交
3146
	di = btrfs_lookup_dir_index_item(trans, root, path, dir_ino, index,
3147 3148 3149 3150 3151 3152 3153 3154 3155 3156 3157 3158
				dentry->d_name.name, dentry->d_name.len, 0);
	if (IS_ERR(di)) {
		err = PTR_ERR(di);
		goto out;
	}
	BUG_ON(ret == -ENOENT);
	if (check_path_shared(root, path))
		goto out;

	err = 0;
out:
	btrfs_free_path(path);
3159 3160 3161 3162
	/* Migrate the orphan reservation over */
	if (!err)
		err = btrfs_block_rsv_migrate(trans->block_rsv,
				&root->fs_info->global_block_rsv,
3163
				trans->bytes_reserved);
3164

3165 3166 3167 3168 3169 3170 3171 3172 3173 3174 3175 3176 3177
	if (err) {
		btrfs_end_transaction(trans, root);
		root->fs_info->enospc_unlink = 0;
		return ERR_PTR(err);
	}

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

static void __unlink_end_trans(struct btrfs_trans_handle *trans,
			       struct btrfs_root *root)
{
3178
	if (trans->block_rsv->type == BTRFS_BLOCK_RSV_GLOBAL) {
3179 3180 3181
		btrfs_block_rsv_release(root, trans->block_rsv,
					trans->bytes_reserved);
		trans->block_rsv = &root->fs_info->trans_block_rsv;
3182 3183 3184
		BUG_ON(!root->fs_info->enospc_unlink);
		root->fs_info->enospc_unlink = 0;
	}
3185
	btrfs_end_transaction(trans, root);
3186 3187 3188 3189 3190 3191 3192 3193 3194 3195 3196 3197
}

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

	trans = __unlink_start_trans(dir, dentry);
	if (IS_ERR(trans))
		return PTR_ERR(trans);
3198

3199 3200
	btrfs_record_unlink_dir(trans, dir, dentry->d_inode, 0);

3201 3202
	ret = btrfs_unlink_inode(trans, root, dir, dentry->d_inode,
				 dentry->d_name.name, dentry->d_name.len);
3203 3204
	if (ret)
		goto out;
3205

3206
	if (inode->i_nlink == 0) {
3207
		ret = btrfs_orphan_add(trans, inode);
3208 3209
		if (ret)
			goto out;
3210
	}
3211

3212
out:
3213
	__unlink_end_trans(trans, root);
3214
	btrfs_btree_balance_dirty(root);
C
Chris Mason 已提交
3215 3216 3217
	return ret;
}

3218 3219 3220 3221 3222 3223 3224 3225 3226 3227 3228
int btrfs_unlink_subvol(struct btrfs_trans_handle *trans,
			struct btrfs_root *root,
			struct inode *dir, u64 objectid,
			const char *name, int name_len)
{
	struct btrfs_path *path;
	struct extent_buffer *leaf;
	struct btrfs_dir_item *di;
	struct btrfs_key key;
	u64 index;
	int ret;
L
Li Zefan 已提交
3229
	u64 dir_ino = btrfs_ino(dir);
3230 3231 3232 3233 3234

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

L
Li Zefan 已提交
3235
	di = btrfs_lookup_dir_item(trans, root, path, dir_ino,
3236
				   name, name_len, -1);
3237 3238 3239 3240 3241 3242 3243
	if (IS_ERR_OR_NULL(di)) {
		if (!di)
			ret = -ENOENT;
		else
			ret = PTR_ERR(di);
		goto out;
	}
3244 3245 3246 3247 3248

	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);
3249 3250 3251 3252
	if (ret) {
		btrfs_abort_transaction(trans, root, ret);
		goto out;
	}
3253
	btrfs_release_path(path);
3254 3255 3256

	ret = btrfs_del_root_ref(trans, root->fs_info->tree_root,
				 objectid, root->root_key.objectid,
L
Li Zefan 已提交
3257
				 dir_ino, &index, name, name_len);
3258
	if (ret < 0) {
3259 3260 3261 3262
		if (ret != -ENOENT) {
			btrfs_abort_transaction(trans, root, ret);
			goto out;
		}
L
Li Zefan 已提交
3263
		di = btrfs_search_dir_index_item(root, path, dir_ino,
3264
						 name, name_len);
3265 3266 3267 3268 3269 3270 3271 3272
		if (IS_ERR_OR_NULL(di)) {
			if (!di)
				ret = -ENOENT;
			else
				ret = PTR_ERR(di);
			btrfs_abort_transaction(trans, root, ret);
			goto out;
		}
3273 3274 3275

		leaf = path->nodes[0];
		btrfs_item_key_to_cpu(leaf, &key, path->slots[0]);
3276
		btrfs_release_path(path);
3277 3278
		index = key.offset;
	}
3279
	btrfs_release_path(path);
3280

3281
	ret = btrfs_delete_delayed_dir_index(trans, root, dir, index);
3282 3283 3284 3285
	if (ret) {
		btrfs_abort_transaction(trans, root, ret);
		goto out;
	}
3286 3287

	btrfs_i_size_write(dir, dir->i_size - name_len * 2);
3288
	inode_inc_iversion(dir);
3289
	dir->i_mtime = dir->i_ctime = CURRENT_TIME;
3290
	ret = btrfs_update_inode_fallback(trans, root, dir);
3291 3292 3293
	if (ret)
		btrfs_abort_transaction(trans, root, ret);
out:
3294
	btrfs_free_path(path);
3295
	return ret;
3296 3297
}

C
Chris Mason 已提交
3298 3299 3300
static int btrfs_rmdir(struct inode *dir, struct dentry *dentry)
{
	struct inode *inode = dentry->d_inode;
3301
	int err = 0;
C
Chris Mason 已提交
3302 3303 3304
	struct btrfs_root *root = BTRFS_I(dir)->root;
	struct btrfs_trans_handle *trans;

3305
	if (inode->i_size > BTRFS_EMPTY_DIR_SIZE)
Y
Yan 已提交
3306
		return -ENOTEMPTY;
3307 3308
	if (btrfs_ino(inode) == BTRFS_FIRST_FREE_OBJECTID)
		return -EPERM;
Y
Yan 已提交
3309

3310 3311
	trans = __unlink_start_trans(dir, dentry);
	if (IS_ERR(trans))
3312 3313
		return PTR_ERR(trans);

L
Li Zefan 已提交
3314
	if (unlikely(btrfs_ino(inode) == BTRFS_EMPTY_SUBVOL_DIR_OBJECTID)) {
3315 3316 3317 3318 3319 3320 3321
		err = btrfs_unlink_subvol(trans, root, dir,
					  BTRFS_I(inode)->location.objectid,
					  dentry->d_name.name,
					  dentry->d_name.len);
		goto out;
	}

3322 3323
	err = btrfs_orphan_add(trans, inode);
	if (err)
3324
		goto out;
3325

C
Chris Mason 已提交
3326
	/* now the directory is empty */
3327 3328
	err = btrfs_unlink_inode(trans, root, dir, dentry->d_inode,
				 dentry->d_name.name, dentry->d_name.len);
C
Chris Mason 已提交
3329
	if (!err)
3330
		btrfs_i_size_write(inode, 0);
3331
out:
3332
	__unlink_end_trans(trans, root);
3333
	btrfs_btree_balance_dirty(root);
3334

C
Chris Mason 已提交
3335 3336 3337 3338 3339 3340
	return err;
}

/*
 * this can truncate away extent items, csum items and directory items.
 * It starts at a high offset and removes keys until it can't find
C
Chris Mason 已提交
3341
 * any higher than new_size
C
Chris Mason 已提交
3342 3343 3344
 *
 * csum items that cross the new i_size are truncated to the new size
 * as well.
3345 3346 3347
 *
 * 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
3348
 */
3349 3350 3351 3352
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
3353 3354
{
	struct btrfs_path *path;
3355
	struct extent_buffer *leaf;
C
Chris Mason 已提交
3356
	struct btrfs_file_extent_item *fi;
3357 3358
	struct btrfs_key key;
	struct btrfs_key found_key;
C
Chris Mason 已提交
3359
	u64 extent_start = 0;
3360
	u64 extent_num_bytes = 0;
3361
	u64 extent_offset = 0;
C
Chris Mason 已提交
3362
	u64 item_end = 0;
3363 3364
	u64 mask = root->sectorsize - 1;
	u32 found_type = (u8)-1;
C
Chris Mason 已提交
3365 3366
	int found_extent;
	int del_item;
3367 3368
	int pending_del_nr = 0;
	int pending_del_slot = 0;
3369
	int extent_type = -1;
3370 3371
	int ret;
	int err = 0;
L
Li Zefan 已提交
3372
	u64 ino = btrfs_ino(inode);
3373 3374

	BUG_ON(new_size > 0 && min_type != BTRFS_EXTENT_DATA_KEY);
C
Chris Mason 已提交
3375

3376 3377 3378 3379 3380
	path = btrfs_alloc_path();
	if (!path)
		return -ENOMEM;
	path->reada = -1;

J
Josef Bacik 已提交
3381 3382 3383 3384 3385
	/*
	 * We want to drop from the next block forward in case this new size is
	 * not block aligned since we will be keeping the last block of the
	 * extent just the way it is.
	 */
3386
	if (root->ref_cows || root == root->fs_info->tree_root)
J
Josef Bacik 已提交
3387
		btrfs_drop_extent_cache(inode, (new_size + mask) & (~mask), (u64)-1, 0);
3388

3389 3390 3391 3392 3393 3394 3395 3396 3397
	/*
	 * This function is also used to drop the items in the log tree before
	 * we relog the inode, so if root != BTRFS_I(inode)->root, it means
	 * it is used to drop the loged items. So we shouldn't kill the delayed
	 * items.
	 */
	if (min_type == 0 && root == BTRFS_I(inode)->root)
		btrfs_kill_delayed_inode_items(inode);

L
Li Zefan 已提交
3398
	key.objectid = ino;
C
Chris Mason 已提交
3399
	key.offset = (u64)-1;
3400 3401
	key.type = (u8)-1;

3402
search_again:
3403
	path->leave_spinning = 1;
3404
	ret = btrfs_search_slot(trans, root, &key, path, -1, 1);
3405 3406 3407 3408
	if (ret < 0) {
		err = ret;
		goto out;
	}
C
Chris Mason 已提交
3409

3410
	if (ret > 0) {
3411 3412 3413
		/* there are no items in the tree for us to truncate, we're
		 * done
		 */
3414 3415
		if (path->slots[0] == 0)
			goto out;
3416 3417 3418
		path->slots[0]--;
	}

C
Chris Mason 已提交
3419
	while (1) {
C
Chris Mason 已提交
3420
		fi = NULL;
3421 3422 3423
		leaf = path->nodes[0];
		btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
		found_type = btrfs_key_type(&found_key);
C
Chris Mason 已提交
3424

L
Li Zefan 已提交
3425
		if (found_key.objectid != ino)
C
Chris Mason 已提交
3426
			break;
3427

3428
		if (found_type < min_type)
C
Chris Mason 已提交
3429 3430
			break;

3431
		item_end = found_key.offset;
C
Chris Mason 已提交
3432
		if (found_type == BTRFS_EXTENT_DATA_KEY) {
3433
			fi = btrfs_item_ptr(leaf, path->slots[0],
C
Chris Mason 已提交
3434
					    struct btrfs_file_extent_item);
3435 3436
			extent_type = btrfs_file_extent_type(leaf, fi);
			if (extent_type != BTRFS_FILE_EXTENT_INLINE) {
3437
				item_end +=
3438
				    btrfs_file_extent_num_bytes(leaf, fi);
3439 3440
			} else if (extent_type == BTRFS_FILE_EXTENT_INLINE) {
				item_end += btrfs_file_extent_inline_len(leaf,
C
Chris Mason 已提交
3441
									 fi);
C
Chris Mason 已提交
3442
			}
3443
			item_end--;
C
Chris Mason 已提交
3444
		}
3445 3446 3447 3448
		if (found_type > min_type) {
			del_item = 1;
		} else {
			if (item_end < new_size)
3449
				break;
3450 3451 3452 3453
			if (found_key.offset >= new_size)
				del_item = 1;
			else
				del_item = 0;
C
Chris Mason 已提交
3454 3455 3456
		}
		found_extent = 0;
		/* FIXME, shrink the extent if the ref count is only 1 */
3457 3458 3459 3460
		if (found_type != BTRFS_EXTENT_DATA_KEY)
			goto delete;

		if (extent_type != BTRFS_FILE_EXTENT_INLINE) {
C
Chris Mason 已提交
3461
			u64 num_dec;
3462
			extent_start = btrfs_file_extent_disk_bytenr(leaf, fi);
3463
			if (!del_item) {
3464 3465
				u64 orig_num_bytes =
					btrfs_file_extent_num_bytes(leaf, fi);
3466
				extent_num_bytes = new_size -
3467
					found_key.offset + root->sectorsize - 1;
3468 3469
				extent_num_bytes = extent_num_bytes &
					~((u64)root->sectorsize - 1);
3470 3471 3472
				btrfs_set_file_extent_num_bytes(leaf, fi,
							 extent_num_bytes);
				num_dec = (orig_num_bytes -
C
Chris Mason 已提交
3473
					   extent_num_bytes);
3474
				if (root->ref_cows && extent_start != 0)
3475
					inode_sub_bytes(inode, num_dec);
3476
				btrfs_mark_buffer_dirty(leaf);
C
Chris Mason 已提交
3477
			} else {
3478 3479 3480
				extent_num_bytes =
					btrfs_file_extent_disk_num_bytes(leaf,
									 fi);
3481 3482 3483
				extent_offset = found_key.offset -
					btrfs_file_extent_offset(leaf, fi);

C
Chris Mason 已提交
3484
				/* FIXME blocksize != 4096 */
C
Chris Mason 已提交
3485
				num_dec = btrfs_file_extent_num_bytes(leaf, fi);
C
Chris Mason 已提交
3486 3487
				if (extent_start != 0) {
					found_extent = 1;
3488
					if (root->ref_cows)
3489
						inode_sub_bytes(inode, num_dec);
3490
				}
C
Chris Mason 已提交
3491
			}
C
Chris Mason 已提交
3492
		} else if (extent_type == BTRFS_FILE_EXTENT_INLINE) {
C
Chris Mason 已提交
3493 3494 3495 3496 3497 3498 3499 3500
			/*
			 * 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) {
3501 3502 3503
				u32 size = new_size - found_key.offset;

				if (root->ref_cows) {
3504 3505
					inode_sub_bytes(inode, item_end + 1 -
							new_size);
3506 3507 3508
				}
				size =
				    btrfs_file_extent_calc_inline_size(size);
3509 3510
				btrfs_truncate_item(trans, root, path,
						    size, 1);
3511
			} else if (root->ref_cows) {
3512 3513
				inode_sub_bytes(inode, item_end + 1 -
						found_key.offset);
C
Chris Mason 已提交
3514
			}
C
Chris Mason 已提交
3515
		}
3516
delete:
C
Chris Mason 已提交
3517
		if (del_item) {
3518 3519 3520 3521 3522 3523 3524 3525 3526 3527
			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
3528
				BUG();
3529
			}
C
Chris Mason 已提交
3530 3531 3532
		} else {
			break;
		}
3533 3534
		if (found_extent && (root->ref_cows ||
				     root == root->fs_info->tree_root)) {
3535
			btrfs_set_path_blocking(path);
C
Chris Mason 已提交
3536
			ret = btrfs_free_extent(trans, root, extent_start,
3537 3538
						extent_num_bytes, 0,
						btrfs_header_owner(leaf),
A
Arne Jansen 已提交
3539
						ino, extent_offset, 0);
C
Chris Mason 已提交
3540 3541
			BUG_ON(ret);
		}
3542

3543 3544 3545 3546 3547 3548 3549 3550 3551
		if (found_type == BTRFS_INODE_ITEM_KEY)
			break;

		if (path->slots[0] == 0 ||
		    path->slots[0] != pending_del_slot) {
			if (pending_del_nr) {
				ret = btrfs_del_items(trans, root, path,
						pending_del_slot,
						pending_del_nr);
3552 3553 3554 3555 3556
				if (ret) {
					btrfs_abort_transaction(trans,
								root, ret);
					goto error;
				}
3557 3558
				pending_del_nr = 0;
			}
3559
			btrfs_release_path(path);
3560
			goto search_again;
3561 3562
		} else {
			path->slots[0]--;
3563
		}
C
Chris Mason 已提交
3564
	}
3565
out:
3566 3567 3568
	if (pending_del_nr) {
		ret = btrfs_del_items(trans, root, path, pending_del_slot,
				      pending_del_nr);
3569 3570
		if (ret)
			btrfs_abort_transaction(trans, root, ret);
3571
	}
3572
error:
C
Chris Mason 已提交
3573
	btrfs_free_path(path);
3574
	return err;
C
Chris Mason 已提交
3575 3576 3577
}

/*
J
Josef Bacik 已提交
3578 3579 3580 3581 3582 3583 3584 3585 3586
 * btrfs_truncate_page - read, zero a chunk and write a page
 * @inode - inode that we're zeroing
 * @from - the offset to start zeroing
 * @len - the length to zero, 0 to zero the entire range respective to the
 *	offset
 * @front - zero up to the offset instead of from the offset on
 *
 * This will find the page for the "from" offset and cow the page and zero the
 * part we want to zero.  This is used with truncate and hole punching.
C
Chris Mason 已提交
3587
 */
J
Josef Bacik 已提交
3588 3589
int btrfs_truncate_page(struct inode *inode, loff_t from, loff_t len,
			int front)
C
Chris Mason 已提交
3590
{
J
Josef Bacik 已提交
3591
	struct address_space *mapping = inode->i_mapping;
3592
	struct btrfs_root *root = BTRFS_I(inode)->root;
3593 3594
	struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
	struct btrfs_ordered_extent *ordered;
3595
	struct extent_state *cached_state = NULL;
3596
	char *kaddr;
3597
	u32 blocksize = root->sectorsize;
C
Chris Mason 已提交
3598 3599 3600
	pgoff_t index = from >> PAGE_CACHE_SHIFT;
	unsigned offset = from & (PAGE_CACHE_SIZE-1);
	struct page *page;
3601
	gfp_t mask = btrfs_alloc_write_mask(mapping);
C
Chris Mason 已提交
3602
	int ret = 0;
3603
	u64 page_start;
3604
	u64 page_end;
C
Chris Mason 已提交
3605

J
Josef Bacik 已提交
3606 3607
	if ((offset & (blocksize - 1)) == 0 &&
	    (!len || ((len & (blocksize - 1)) == 0)))
C
Chris Mason 已提交
3608
		goto out;
3609
	ret = btrfs_delalloc_reserve_space(inode, PAGE_CACHE_SIZE);
3610 3611
	if (ret)
		goto out;
C
Chris Mason 已提交
3612

3613
again:
3614
	page = find_or_create_page(mapping, index, mask);
3615
	if (!page) {
3616
		btrfs_delalloc_release_space(inode, PAGE_CACHE_SIZE);
3617
		ret = -ENOMEM;
C
Chris Mason 已提交
3618
		goto out;
3619
	}
3620 3621 3622 3623

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

C
Chris Mason 已提交
3624
	if (!PageUptodate(page)) {
C
Chris Mason 已提交
3625
		ret = btrfs_readpage(NULL, page);
C
Chris Mason 已提交
3626
		lock_page(page);
3627 3628 3629 3630 3631
		if (page->mapping != mapping) {
			unlock_page(page);
			page_cache_release(page);
			goto again;
		}
C
Chris Mason 已提交
3632 3633
		if (!PageUptodate(page)) {
			ret = -EIO;
3634
			goto out_unlock;
C
Chris Mason 已提交
3635 3636
		}
	}
3637
	wait_on_page_writeback(page);
3638

3639
	lock_extent_bits(io_tree, page_start, page_end, 0, &cached_state);
3640 3641 3642 3643
	set_page_extent_mapped(page);

	ordered = btrfs_lookup_ordered_extent(inode, page_start);
	if (ordered) {
3644 3645
		unlock_extent_cached(io_tree, page_start, page_end,
				     &cached_state, GFP_NOFS);
3646 3647
		unlock_page(page);
		page_cache_release(page);
3648
		btrfs_start_ordered_extent(inode, ordered, 1);
3649 3650 3651 3652
		btrfs_put_ordered_extent(ordered);
		goto again;
	}

3653
	clear_extent_bit(&BTRFS_I(inode)->io_tree, page_start, page_end,
3654 3655
			  EXTENT_DIRTY | EXTENT_DELALLOC |
			  EXTENT_DO_ACCOUNTING | EXTENT_DEFRAG,
3656
			  0, 0, &cached_state, GFP_NOFS);
3657

3658 3659
	ret = btrfs_set_extent_delalloc(inode, page_start, page_end,
					&cached_state);
J
Josef Bacik 已提交
3660
	if (ret) {
3661 3662
		unlock_extent_cached(io_tree, page_start, page_end,
				     &cached_state, GFP_NOFS);
J
Josef Bacik 已提交
3663 3664 3665
		goto out_unlock;
	}

3666
	if (offset != PAGE_CACHE_SIZE) {
J
Josef Bacik 已提交
3667 3668
		if (!len)
			len = PAGE_CACHE_SIZE - offset;
3669
		kaddr = kmap(page);
J
Josef Bacik 已提交
3670 3671 3672 3673
		if (front)
			memset(kaddr, 0, offset);
		else
			memset(kaddr + offset, 0, len);
3674 3675 3676
		flush_dcache_page(page);
		kunmap(page);
	}
3677
	ClearPageChecked(page);
3678
	set_page_dirty(page);
3679 3680
	unlock_extent_cached(io_tree, page_start, page_end, &cached_state,
			     GFP_NOFS);
C
Chris Mason 已提交
3681

3682
out_unlock:
3683
	if (ret)
3684
		btrfs_delalloc_release_space(inode, PAGE_CACHE_SIZE);
C
Chris Mason 已提交
3685 3686 3687 3688 3689 3690
	unlock_page(page);
	page_cache_release(page);
out:
	return ret;
}

3691 3692 3693 3694 3695 3696
/*
 * This function puts in dummy file extents for the area we're creating a hole
 * for.  So if we are truncating this file to a larger size we need to insert
 * these file extents so that btrfs_get_extent will return a EXTENT_MAP_HOLE for
 * the range between oldsize and size
 */
3697
int btrfs_cont_expand(struct inode *inode, loff_t oldsize, loff_t size)
C
Chris Mason 已提交
3698
{
Y
Yan Zheng 已提交
3699 3700 3701
	struct btrfs_trans_handle *trans;
	struct btrfs_root *root = BTRFS_I(inode)->root;
	struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
3702
	struct extent_map *em = NULL;
3703
	struct extent_state *cached_state = NULL;
J
Josef Bacik 已提交
3704
	struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
Y
Yan Zheng 已提交
3705
	u64 mask = root->sectorsize - 1;
3706
	u64 hole_start = (oldsize + mask) & ~mask;
Y
Yan Zheng 已提交
3707 3708 3709 3710
	u64 block_end = (size + mask) & ~mask;
	u64 last_byte;
	u64 cur_offset;
	u64 hole_size;
J
Josef Bacik 已提交
3711
	int err = 0;
C
Chris Mason 已提交
3712

Y
Yan Zheng 已提交
3713 3714 3715 3716 3717 3718 3719
	if (size <= hole_start)
		return 0;

	while (1) {
		struct btrfs_ordered_extent *ordered;
		btrfs_wait_ordered_range(inode, hole_start,
					 block_end - hole_start);
3720
		lock_extent_bits(io_tree, hole_start, block_end - 1, 0,
3721
				 &cached_state);
Y
Yan Zheng 已提交
3722 3723 3724
		ordered = btrfs_lookup_ordered_extent(inode, hole_start);
		if (!ordered)
			break;
3725 3726
		unlock_extent_cached(io_tree, hole_start, block_end - 1,
				     &cached_state, GFP_NOFS);
Y
Yan Zheng 已提交
3727 3728
		btrfs_put_ordered_extent(ordered);
	}
C
Chris Mason 已提交
3729

Y
Yan Zheng 已提交
3730 3731 3732 3733
	cur_offset = hole_start;
	while (1) {
		em = btrfs_get_extent(inode, NULL, 0, cur_offset,
				block_end - cur_offset, 0);
3734 3735
		if (IS_ERR(em)) {
			err = PTR_ERR(em);
3736
			em = NULL;
3737 3738
			break;
		}
Y
Yan Zheng 已提交
3739 3740
		last_byte = min(extent_map_end(em), block_end);
		last_byte = (last_byte + mask) & ~mask;
3741
		if (!test_bit(EXTENT_FLAG_PREALLOC, &em->flags)) {
J
Josef Bacik 已提交
3742
			struct extent_map *hole_em;
Y
Yan Zheng 已提交
3743
			hole_size = last_byte - cur_offset;
J
Josef Bacik 已提交
3744

3745
			trans = btrfs_start_transaction(root, 3);
3746 3747
			if (IS_ERR(trans)) {
				err = PTR_ERR(trans);
J
Josef Bacik 已提交
3748
				break;
3749
			}
3750

J
Josef Bacik 已提交
3751 3752
			err = btrfs_drop_extents(trans, root, inode,
						 cur_offset,
3753
						 cur_offset + hole_size, 1);
3754
			if (err) {
3755
				btrfs_abort_transaction(trans, root, err);
3756
				btrfs_end_transaction(trans, root);
3757
				break;
3758
			}
3759

Y
Yan Zheng 已提交
3760
			err = btrfs_insert_file_extent(trans, root,
L
Li Zefan 已提交
3761
					btrfs_ino(inode), cur_offset, 0,
Y
Yan Zheng 已提交
3762 3763
					0, hole_size, 0, hole_size,
					0, 0, 0);
3764
			if (err) {
3765
				btrfs_abort_transaction(trans, root, err);
3766
				btrfs_end_transaction(trans, root);
3767
				break;
3768
			}
3769

J
Josef Bacik 已提交
3770 3771 3772 3773 3774 3775 3776 3777 3778 3779 3780
			btrfs_drop_extent_cache(inode, cur_offset,
						cur_offset + hole_size - 1, 0);
			hole_em = alloc_extent_map();
			if (!hole_em) {
				set_bit(BTRFS_INODE_NEEDS_FULL_SYNC,
					&BTRFS_I(inode)->runtime_flags);
				goto next;
			}
			hole_em->start = cur_offset;
			hole_em->len = hole_size;
			hole_em->orig_start = cur_offset;
3781

J
Josef Bacik 已提交
3782 3783
			hole_em->block_start = EXTENT_MAP_HOLE;
			hole_em->block_len = 0;
3784
			hole_em->orig_block_len = 0;
J
Josef Bacik 已提交
3785 3786 3787
			hole_em->bdev = root->fs_info->fs_devices->latest_bdev;
			hole_em->compress_type = BTRFS_COMPRESS_NONE;
			hole_em->generation = trans->transid;
3788

J
Josef Bacik 已提交
3789 3790 3791 3792 3793 3794 3795 3796 3797 3798 3799 3800 3801 3802 3803
			while (1) {
				write_lock(&em_tree->lock);
				err = add_extent_mapping(em_tree, hole_em);
				if (!err)
					list_move(&hole_em->list,
						  &em_tree->modified_extents);
				write_unlock(&em_tree->lock);
				if (err != -EEXIST)
					break;
				btrfs_drop_extent_cache(inode, cur_offset,
							cur_offset +
							hole_size - 1, 0);
			}
			free_extent_map(hole_em);
next:
3804
			btrfs_update_inode(trans, root, inode);
3805
			btrfs_end_transaction(trans, root);
Y
Yan Zheng 已提交
3806 3807
		}
		free_extent_map(em);
3808
		em = NULL;
Y
Yan Zheng 已提交
3809
		cur_offset = last_byte;
3810
		if (cur_offset >= block_end)
Y
Yan Zheng 已提交
3811 3812
			break;
	}
3813

3814
	free_extent_map(em);
3815 3816
	unlock_extent_cached(io_tree, hole_start, block_end - 1, &cached_state,
			     GFP_NOFS);
Y
Yan Zheng 已提交
3817 3818
	return err;
}
C
Chris Mason 已提交
3819

3820
static int btrfs_setsize(struct inode *inode, struct iattr *attr)
3821
{
3822 3823
	struct btrfs_root *root = BTRFS_I(inode)->root;
	struct btrfs_trans_handle *trans;
3824
	loff_t oldsize = i_size_read(inode);
3825 3826
	loff_t newsize = attr->ia_size;
	int mask = attr->ia_valid;
3827 3828
	int ret;

3829
	if (newsize == oldsize)
3830 3831
		return 0;

3832 3833 3834 3835 3836 3837 3838 3839 3840
	/*
	 * The regular truncate() case without ATTR_CTIME and ATTR_MTIME is a
	 * special case where we need to update the times despite not having
	 * these flags set.  For all other operations the VFS set these flags
	 * explicitly if it wants a timestamp update.
	 */
	if (newsize != oldsize && (!(mask & (ATTR_CTIME | ATTR_MTIME))))
		inode->i_ctime = inode->i_mtime = current_fs_time(inode->i_sb);

3841 3842 3843
	if (newsize > oldsize) {
		truncate_pagecache(inode, oldsize, newsize);
		ret = btrfs_cont_expand(inode, oldsize, newsize);
3844
		if (ret)
3845 3846
			return ret;

3847 3848 3849 3850 3851 3852 3853
		trans = btrfs_start_transaction(root, 1);
		if (IS_ERR(trans))
			return PTR_ERR(trans);

		i_size_write(inode, newsize);
		btrfs_ordered_update_i_size(inode, i_size_read(inode), NULL);
		ret = btrfs_update_inode(trans, root, inode);
3854
		btrfs_end_transaction(trans, root);
3855
	} else {
3856

3857 3858 3859 3860 3861 3862
		/*
		 * We're truncating a file that used to have good data down to
		 * zero. Make sure it gets into the ordered flush list so that
		 * any new writes get down to disk quickly.
		 */
		if (newsize == 0)
3863 3864
			set_bit(BTRFS_INODE_ORDERED_DATA_CLOSE,
				&BTRFS_I(inode)->runtime_flags);
3865

3866 3867 3868 3869 3870 3871 3872 3873 3874 3875 3876 3877 3878 3879 3880 3881 3882 3883 3884 3885 3886 3887 3888
		/*
		 * 1 for the orphan item we're going to add
		 * 1 for the orphan item deletion.
		 */
		trans = btrfs_start_transaction(root, 2);
		if (IS_ERR(trans))
			return PTR_ERR(trans);

		/*
		 * We need to do this in case we fail at _any_ point during the
		 * actual truncate.  Once we do the truncate_setsize we could
		 * invalidate pages which forces any outstanding ordered io to
		 * be instantly completed which will give us extents that need
		 * to be truncated.  If we fail to get an orphan inode down we
		 * could have left over extents that were never meant to live,
		 * so we need to garuntee from this point on that everything
		 * will be consistent.
		 */
		ret = btrfs_orphan_add(trans, inode);
		btrfs_end_transaction(trans, root);
		if (ret)
			return ret;

3889 3890 3891
		/* we don't support swapfiles, so vmtruncate shouldn't fail */
		truncate_setsize(inode, newsize);
		ret = btrfs_truncate(inode);
3892 3893
		if (ret && inode->i_nlink)
			btrfs_orphan_del(NULL, inode);
3894 3895
	}

3896
	return ret;
3897 3898
}

Y
Yan Zheng 已提交
3899 3900 3901
static int btrfs_setattr(struct dentry *dentry, struct iattr *attr)
{
	struct inode *inode = dentry->d_inode;
L
Li Zefan 已提交
3902
	struct btrfs_root *root = BTRFS_I(inode)->root;
Y
Yan Zheng 已提交
3903
	int err;
C
Chris Mason 已提交
3904

L
Li Zefan 已提交
3905 3906 3907
	if (btrfs_root_readonly(root))
		return -EROFS;

Y
Yan Zheng 已提交
3908 3909 3910
	err = inode_change_ok(inode, attr);
	if (err)
		return err;
C
Chris Mason 已提交
3911

3912
	if (S_ISREG(inode->i_mode) && (attr->ia_valid & ATTR_SIZE)) {
3913
		err = btrfs_setsize(inode, attr);
3914 3915
		if (err)
			return err;
C
Chris Mason 已提交
3916
	}
Y
Yan Zheng 已提交
3917

C
Christoph Hellwig 已提交
3918 3919
	if (attr->ia_valid) {
		setattr_copy(inode, attr);
3920
		inode_inc_iversion(inode);
3921
		err = btrfs_dirty_inode(inode);
C
Christoph Hellwig 已提交
3922

3923
		if (!err && attr->ia_valid & ATTR_MODE)
C
Christoph Hellwig 已提交
3924 3925
			err = btrfs_acl_chmod(inode);
	}
J
Josef Bacik 已提交
3926

C
Chris Mason 已提交
3927 3928
	return err;
}
3929

A
Al Viro 已提交
3930
void btrfs_evict_inode(struct inode *inode)
C
Chris Mason 已提交
3931 3932 3933
{
	struct btrfs_trans_handle *trans;
	struct btrfs_root *root = BTRFS_I(inode)->root;
3934
	struct btrfs_block_rsv *rsv, *global_rsv;
3935
	u64 min_size = btrfs_calc_trunc_metadata_size(root, 1);
C
Chris Mason 已提交
3936 3937
	int ret;

3938 3939
	trace_btrfs_inode_evict(inode);

C
Chris Mason 已提交
3940
	truncate_inode_pages(&inode->i_data, 0);
3941
	if (inode->i_nlink && (btrfs_root_refs(&root->root_item) != 0 ||
3942
			       btrfs_is_free_space_inode(inode)))
A
Al Viro 已提交
3943 3944
		goto no_delete;

C
Chris Mason 已提交
3945
	if (is_bad_inode(inode)) {
3946
		btrfs_orphan_del(NULL, inode);
C
Chris Mason 已提交
3947 3948
		goto no_delete;
	}
A
Al Viro 已提交
3949
	/* do we really want it for ->i_nlink > 0 and zero btrfs_root_refs? */
C
Chris Mason 已提交
3950
	btrfs_wait_ordered_range(inode, 0, (u64)-1);
3951

3952
	if (root->fs_info->log_root_recovering) {
3953
		BUG_ON(test_bit(BTRFS_INODE_HAS_ORPHAN_ITEM,
3954
				 &BTRFS_I(inode)->runtime_flags));
3955 3956 3957
		goto no_delete;
	}

3958 3959 3960 3961 3962
	if (inode->i_nlink > 0) {
		BUG_ON(btrfs_root_refs(&root->root_item) != 0);
		goto no_delete;
	}

3963 3964 3965 3966 3967 3968
	ret = btrfs_commit_inode_delayed_inode(inode);
	if (ret) {
		btrfs_orphan_del(NULL, inode);
		goto no_delete;
	}

3969
	rsv = btrfs_alloc_block_rsv(root, BTRFS_BLOCK_RSV_TEMP);
3970 3971 3972 3973
	if (!rsv) {
		btrfs_orphan_del(NULL, inode);
		goto no_delete;
	}
J
Josef Bacik 已提交
3974
	rsv->size = min_size;
3975
	rsv->failfast = 1;
3976
	global_rsv = &root->fs_info->global_block_rsv;
3977

3978
	btrfs_i_size_write(inode, 0);
3979

3980
	/*
3981 3982 3983 3984
	 * This is a bit simpler than btrfs_truncate since we've already
	 * reserved our space for our orphan item in the unlink, so we just
	 * need to reserve some slack space in case we add bytes and update
	 * inode item when doing the truncate.
3985
	 */
3986
	while (1) {
M
Miao Xie 已提交
3987 3988
		ret = btrfs_block_rsv_refill(root, rsv, min_size,
					     BTRFS_RESERVE_FLUSH_LIMIT);
3989 3990 3991 3992 3993 3994 3995 3996

		/*
		 * Try and steal from the global reserve since we will
		 * likely not use this space anyway, we want to try as
		 * hard as possible to get this to work.
		 */
		if (ret)
			ret = btrfs_block_rsv_migrate(global_rsv, rsv, min_size);
3997 3998

		if (ret) {
3999
			printk(KERN_WARNING "Could not get space for a "
4000
			       "delete, will truncate on mount %d\n", ret);
4001 4002 4003
			btrfs_orphan_del(NULL, inode);
			btrfs_free_block_rsv(root, rsv);
			goto no_delete;
4004
		}
4005

4006
		trans = btrfs_join_transaction(root);
4007 4008 4009 4010
		if (IS_ERR(trans)) {
			btrfs_orphan_del(NULL, inode);
			btrfs_free_block_rsv(root, rsv);
			goto no_delete;
4011
		}
4012

4013 4014
		trans->block_rsv = rsv;

4015
		ret = btrfs_truncate_inode_items(trans, root, inode, 0, 0);
4016
		if (ret != -ENOSPC)
4017
			break;
4018

4019
		trans->block_rsv = &root->fs_info->trans_block_rsv;
4020 4021
		btrfs_end_transaction(trans, root);
		trans = NULL;
4022
		btrfs_btree_balance_dirty(root);
4023
	}
4024

4025 4026
	btrfs_free_block_rsv(root, rsv);

4027
	if (ret == 0) {
4028
		trans->block_rsv = root->orphan_block_rsv;
4029 4030 4031
		ret = btrfs_orphan_del(trans, inode);
		BUG_ON(ret);
	}
4032

4033
	trans->block_rsv = &root->fs_info->trans_block_rsv;
4034 4035
	if (!(root == root->fs_info->tree_root ||
	      root->root_key.objectid == BTRFS_TREE_RELOC_OBJECTID))
L
Li Zefan 已提交
4036
		btrfs_return_ino(root, btrfs_ino(inode));
4037

4038
	btrfs_end_transaction(trans, root);
4039
	btrfs_btree_balance_dirty(root);
C
Chris Mason 已提交
4040
no_delete:
4041
	clear_inode(inode);
4042
	return;
C
Chris Mason 已提交
4043 4044 4045 4046 4047 4048 4049 4050 4051 4052 4053 4054 4055 4056
}

/*
 * 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
4057
	int ret = 0;
C
Chris Mason 已提交
4058 4059

	path = btrfs_alloc_path();
4060 4061
	if (!path)
		return -ENOMEM;
4062

L
Li Zefan 已提交
4063
	di = btrfs_lookup_dir_item(NULL, root, path, btrfs_ino(dir), name,
C
Chris Mason 已提交
4064
				    namelen, 0);
Y
Yan 已提交
4065 4066
	if (IS_ERR(di))
		ret = PTR_ERR(di);
C
Chris Mason 已提交
4067

4068
	if (IS_ERR_OR_NULL(di))
4069
		goto out_err;
C
Chris Mason 已提交
4070

4071
	btrfs_dir_item_key_to_cpu(path->nodes[0], di, location);
C
Chris Mason 已提交
4072 4073 4074
out:
	btrfs_free_path(path);
	return ret;
4075 4076 4077
out_err:
	location->objectid = 0;
	goto out;
C
Chris Mason 已提交
4078 4079 4080 4081 4082 4083 4084 4085
}

/*
 * 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,
4086 4087 4088 4089
				    struct inode *dir,
				    struct dentry *dentry,
				    struct btrfs_key *location,
				    struct btrfs_root **sub_root)
C
Chris Mason 已提交
4090
{
4091 4092 4093 4094 4095 4096
	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
4097

4098 4099 4100 4101 4102
	path = btrfs_alloc_path();
	if (!path) {
		err = -ENOMEM;
		goto out;
	}
C
Chris Mason 已提交
4103

4104 4105 4106 4107 4108 4109 4110 4111 4112
	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
4113

4114 4115
	leaf = path->nodes[0];
	ref = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_root_ref);
L
Li Zefan 已提交
4116
	if (btrfs_root_ref_dirid(leaf, ref) != btrfs_ino(dir) ||
4117 4118
	    btrfs_root_ref_name_len(leaf, ref) != dentry->d_name.len)
		goto out;
C
Chris Mason 已提交
4119

4120 4121 4122 4123 4124 4125
	ret = memcmp_extent_buffer(leaf, dentry->d_name.name,
				   (unsigned long)(ref + 1),
				   dentry->d_name.len);
	if (ret)
		goto out;

4126
	btrfs_release_path(path);
4127 4128 4129 4130 4131 4132 4133 4134 4135 4136 4137 4138 4139 4140 4141 4142 4143 4144 4145 4146

	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
4147 4148
}

4149 4150 4151 4152
static void inode_tree_add(struct inode *inode)
{
	struct btrfs_root *root = BTRFS_I(inode)->root;
	struct btrfs_inode *entry;
4153 4154
	struct rb_node **p;
	struct rb_node *parent;
L
Li Zefan 已提交
4155
	u64 ino = btrfs_ino(inode);
4156 4157 4158
again:
	p = &root->inode_tree.rb_node;
	parent = NULL;
4159

A
Al Viro 已提交
4160
	if (inode_unhashed(inode))
4161 4162
		return;

4163 4164 4165 4166 4167
	spin_lock(&root->inode_lock);
	while (*p) {
		parent = *p;
		entry = rb_entry(parent, struct btrfs_inode, rb_node);

L
Li Zefan 已提交
4168
		if (ino < btrfs_ino(&entry->vfs_inode))
4169
			p = &parent->rb_left;
L
Li Zefan 已提交
4170
		else if (ino > btrfs_ino(&entry->vfs_inode))
4171
			p = &parent->rb_right;
4172 4173
		else {
			WARN_ON(!(entry->vfs_inode.i_state &
A
Al Viro 已提交
4174
				  (I_WILL_FREE | I_FREEING)));
4175 4176 4177 4178
			rb_erase(parent, &root->inode_tree);
			RB_CLEAR_NODE(parent);
			spin_unlock(&root->inode_lock);
			goto again;
4179 4180 4181 4182 4183 4184 4185 4186 4187 4188
		}
	}
	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;
4189
	int empty = 0;
4190

4191
	spin_lock(&root->inode_lock);
4192 4193 4194
	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);
4195
		empty = RB_EMPTY_ROOT(&root->inode_tree);
4196
	}
4197
	spin_unlock(&root->inode_lock);
4198

4199 4200 4201 4202 4203 4204 4205 4206
	/*
	 * Free space cache has inodes in the tree root, but the tree root has a
	 * root_refs of 0, so this could end up dropping the tree root as a
	 * snapshot, so we need the extra !root->fs_info->tree_root check to
	 * make sure we don't drop it.
	 */
	if (empty && btrfs_root_refs(&root->root_item) == 0 &&
	    root != root->fs_info->tree_root) {
4207 4208 4209 4210 4211 4212 4213 4214 4215
		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);
	}
}

4216
void btrfs_invalidate_inodes(struct btrfs_root *root)
4217 4218 4219 4220 4221 4222 4223 4224 4225 4226 4227 4228 4229 4230 4231 4232 4233
{
	struct rb_node *node;
	struct rb_node *prev;
	struct btrfs_inode *entry;
	struct inode *inode;
	u64 objectid = 0;

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

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

L
Li Zefan 已提交
4234
		if (objectid < btrfs_ino(&entry->vfs_inode))
4235
			node = node->rb_left;
L
Li Zefan 已提交
4236
		else if (objectid > btrfs_ino(&entry->vfs_inode))
4237 4238 4239 4240 4241 4242 4243
			node = node->rb_right;
		else
			break;
	}
	if (!node) {
		while (prev) {
			entry = rb_entry(prev, struct btrfs_inode, rb_node);
L
Li Zefan 已提交
4244
			if (objectid <= btrfs_ino(&entry->vfs_inode)) {
4245 4246 4247 4248 4249 4250 4251 4252
				node = prev;
				break;
			}
			prev = rb_next(prev);
		}
	}
	while (node) {
		entry = rb_entry(node, struct btrfs_inode, rb_node);
L
Li Zefan 已提交
4253
		objectid = btrfs_ino(&entry->vfs_inode) + 1;
4254 4255 4256 4257 4258 4259
		inode = igrab(&entry->vfs_inode);
		if (inode) {
			spin_unlock(&root->inode_lock);
			if (atomic_read(&inode->i_count) > 1)
				d_prune_aliases(inode);
			/*
4260
			 * btrfs_drop_inode will have it removed from
4261 4262 4263 4264 4265 4266 4267 4268 4269 4270 4271 4272 4273 4274 4275
			 * 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);
4276 4277
}

4278 4279 4280 4281 4282
static int btrfs_init_locked_inode(struct inode *inode, void *p)
{
	struct btrfs_iget_args *args = p;
	inode->i_ino = args->ino;
	BTRFS_I(inode)->root = args->root;
C
Chris Mason 已提交
4283 4284 4285 4286 4287 4288
	return 0;
}

static int btrfs_find_actor(struct inode *inode, void *opaque)
{
	struct btrfs_iget_args *args = opaque;
L
Li Zefan 已提交
4289
	return args->ino == btrfs_ino(inode) &&
C
Chris Mason 已提交
4290
		args->root == BTRFS_I(inode)->root;
C
Chris Mason 已提交
4291 4292
}

4293 4294 4295
static struct inode *btrfs_iget_locked(struct super_block *s,
				       u64 objectid,
				       struct btrfs_root *root)
C
Chris Mason 已提交
4296 4297 4298 4299 4300 4301 4302 4303 4304 4305 4306 4307
{
	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
4308 4309 4310 4311
/* 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,
4312
			 struct btrfs_root *root, int *new)
B
Balaji Rao 已提交
4313 4314 4315 4316 4317
{
	struct inode *inode;

	inode = btrfs_iget_locked(s, location->objectid, root);
	if (!inode)
4318
		return ERR_PTR(-ENOMEM);
B
Balaji Rao 已提交
4319 4320 4321 4322 4323

	if (inode->i_state & I_NEW) {
		BTRFS_I(inode)->root = root;
		memcpy(&BTRFS_I(inode)->location, location, sizeof(*location));
		btrfs_read_locked_inode(inode);
4324 4325 4326 4327 4328 4329
		if (!is_bad_inode(inode)) {
			inode_tree_add(inode);
			unlock_new_inode(inode);
			if (new)
				*new = 1;
		} else {
4330 4331 4332
			unlock_new_inode(inode);
			iput(inode);
			inode = ERR_PTR(-ESTALE);
4333 4334 4335
		}
	}

B
Balaji Rao 已提交
4336 4337 4338
	return inode;
}

4339 4340 4341 4342 4343 4344 4345 4346 4347 4348 4349
static struct inode *new_simple_dir(struct super_block *s,
				    struct btrfs_key *key,
				    struct btrfs_root *root)
{
	struct inode *inode = new_inode(s);

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

	BTRFS_I(inode)->root = root;
	memcpy(&BTRFS_I(inode)->location, key, sizeof(*key));
4350
	set_bit(BTRFS_INODE_DUMMY, &BTRFS_I(inode)->runtime_flags);
4351 4352

	inode->i_ino = BTRFS_EMPTY_SUBVOL_DIR_OBJECTID;
L
Li Zefan 已提交
4353
	inode->i_op = &btrfs_dir_ro_inode_operations;
4354 4355 4356 4357 4358 4359 4360
	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;
}

4361
struct inode *btrfs_lookup_dentry(struct inode *dir, struct dentry *dentry)
C
Chris Mason 已提交
4362
{
C
Chris Mason 已提交
4363
	struct inode *inode;
4364
	struct btrfs_root *root = BTRFS_I(dir)->root;
C
Chris Mason 已提交
4365 4366
	struct btrfs_root *sub_root = root;
	struct btrfs_key location;
4367
	int index;
4368
	int ret = 0;
C
Chris Mason 已提交
4369 4370 4371

	if (dentry->d_name.len > BTRFS_NAME_LEN)
		return ERR_PTR(-ENAMETOOLONG);
4372

4373 4374 4375 4376
	if (unlikely(d_need_lookup(dentry))) {
		memcpy(&location, dentry->d_fsdata, sizeof(struct btrfs_key));
		kfree(dentry->d_fsdata);
		dentry->d_fsdata = NULL;
4377 4378
		/* This thing is hashed, drop it for now */
		d_drop(dentry);
4379 4380 4381
	} else {
		ret = btrfs_inode_by_name(dir, dentry, &location);
	}
4382

C
Chris Mason 已提交
4383 4384
	if (ret < 0)
		return ERR_PTR(ret);
4385

4386 4387 4388 4389
	if (location.objectid == 0)
		return NULL;

	if (location.type == BTRFS_INODE_ITEM_KEY) {
4390
		inode = btrfs_iget(dir->i_sb, &location, root, NULL);
4391 4392 4393 4394 4395
		return inode;
	}

	BUG_ON(location.type != BTRFS_ROOT_ITEM_KEY);

4396
	index = srcu_read_lock(&root->fs_info->subvol_srcu);
4397 4398 4399 4400 4401 4402 4403 4404
	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 {
4405
		inode = btrfs_iget(dir->i_sb, &location, sub_root, NULL);
C
Chris Mason 已提交
4406
	}
4407 4408
	srcu_read_unlock(&root->fs_info->subvol_srcu, index);

4409
	if (!IS_ERR(inode) && root != sub_root) {
4410 4411
		down_read(&root->fs_info->cleanup_work_sem);
		if (!(inode->i_sb->s_flags & MS_RDONLY))
4412
			ret = btrfs_orphan_cleanup(sub_root);
4413
		up_read(&root->fs_info->cleanup_work_sem);
4414 4415
		if (ret)
			inode = ERR_PTR(ret);
4416 4417
	}

4418 4419 4420
	return inode;
}

N
Nick Piggin 已提交
4421
static int btrfs_dentry_delete(const struct dentry *dentry)
4422 4423
{
	struct btrfs_root *root;
L
Li Zefan 已提交
4424
	struct inode *inode = dentry->d_inode;
4425

L
Li Zefan 已提交
4426 4427
	if (!inode && !IS_ROOT(dentry))
		inode = dentry->d_parent->d_inode;
4428

L
Li Zefan 已提交
4429 4430
	if (inode) {
		root = BTRFS_I(inode)->root;
4431 4432
		if (btrfs_root_refs(&root->root_item) == 0)
			return 1;
L
Li Zefan 已提交
4433 4434 4435

		if (btrfs_ino(inode) == BTRFS_EMPTY_SUBVOL_DIR_OBJECTID)
			return 1;
4436
	}
4437 4438 4439
	return 0;
}

4440 4441 4442 4443 4444 4445
static void btrfs_dentry_release(struct dentry *dentry)
{
	if (dentry->d_fsdata)
		kfree(dentry->d_fsdata);
}

4446
static struct dentry *btrfs_lookup(struct inode *dir, struct dentry *dentry,
A
Al Viro 已提交
4447
				   unsigned int flags)
4448
{
4449 4450 4451 4452 4453 4454 4455 4456 4457
	struct dentry *ret;

	ret = d_splice_alias(btrfs_lookup_dentry(dir, dentry), dentry);
	if (unlikely(d_need_lookup(dentry))) {
		spin_lock(&dentry->d_lock);
		dentry->d_flags &= ~DCACHE_NEED_LOOKUP;
		spin_unlock(&dentry->d_lock);
	}
	return ret;
C
Chris Mason 已提交
4458 4459
}

4460
unsigned char btrfs_filetype_table[] = {
C
Chris Mason 已提交
4461 4462 4463
	DT_UNKNOWN, DT_REG, DT_DIR, DT_CHR, DT_BLK, DT_FIFO, DT_SOCK, DT_LNK
};

4464 4465
static int btrfs_real_readdir(struct file *filp, void *dirent,
			      filldir_t filldir)
C
Chris Mason 已提交
4466
{
4467
	struct inode *inode = filp->f_dentry->d_inode;
C
Chris Mason 已提交
4468 4469 4470 4471
	struct btrfs_root *root = BTRFS_I(inode)->root;
	struct btrfs_item *item;
	struct btrfs_dir_item *di;
	struct btrfs_key key;
4472
	struct btrfs_key found_key;
C
Chris Mason 已提交
4473
	struct btrfs_path *path;
4474 4475
	struct list_head ins_list;
	struct list_head del_list;
C
Chris Mason 已提交
4476
	int ret;
4477
	struct extent_buffer *leaf;
C
Chris Mason 已提交
4478 4479 4480 4481 4482 4483 4484
	int slot;
	unsigned char d_type;
	int over = 0;
	u32 di_cur;
	u32 di_total;
	u32 di_len;
	int key_type = BTRFS_DIR_INDEX_KEY;
4485 4486 4487
	char tmp_name[32];
	char *name_ptr;
	int name_len;
4488
	int is_curr = 0;	/* filp->f_pos points to the current index? */
C
Chris Mason 已提交
4489 4490 4491 4492

	/* FIXME, use a real flag for deciding about the key type */
	if (root->fs_info->tree_root == root)
		key_type = BTRFS_DIR_ITEM_KEY;
4493

4494 4495
	/* special case for "." */
	if (filp->f_pos == 0) {
4496 4497
		over = filldir(dirent, ".", 1,
			       filp->f_pos, btrfs_ino(inode), DT_DIR);
4498 4499 4500 4501 4502 4503
		if (over)
			return 0;
		filp->f_pos = 1;
	}
	/* special case for .., just use the back ref */
	if (filp->f_pos == 1) {
4504
		u64 pino = parent_ino(filp->f_path.dentry);
4505
		over = filldir(dirent, "..", 2,
4506
			       filp->f_pos, pino, DT_DIR);
4507
		if (over)
4508
			return 0;
4509 4510
		filp->f_pos = 2;
	}
4511
	path = btrfs_alloc_path();
4512 4513
	if (!path)
		return -ENOMEM;
C
Chris Mason 已提交
4514

J
Josef Bacik 已提交
4515
	path->reada = 1;
4516

4517 4518 4519 4520 4521 4522
	if (key_type == BTRFS_DIR_INDEX_KEY) {
		INIT_LIST_HEAD(&ins_list);
		INIT_LIST_HEAD(&del_list);
		btrfs_get_delayed_items(inode, &ins_list, &del_list);
	}

C
Chris Mason 已提交
4523 4524
	btrfs_set_key_type(&key, key_type);
	key.offset = filp->f_pos;
L
Li Zefan 已提交
4525
	key.objectid = btrfs_ino(inode);
4526

C
Chris Mason 已提交
4527 4528 4529
	ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
	if (ret < 0)
		goto err;
4530 4531

	while (1) {
4532
		leaf = path->nodes[0];
C
Chris Mason 已提交
4533
		slot = path->slots[0];
4534 4535 4536 4537 4538 4539 4540
		if (slot >= btrfs_header_nritems(leaf)) {
			ret = btrfs_next_leaf(root, path);
			if (ret < 0)
				goto err;
			else if (ret > 0)
				break;
			continue;
C
Chris Mason 已提交
4541
		}
4542

4543 4544 4545 4546
		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
4547
			break;
4548
		if (btrfs_key_type(&found_key) != key_type)
C
Chris Mason 已提交
4549
			break;
4550
		if (found_key.offset < filp->f_pos)
4551
			goto next;
4552 4553 4554 4555
		if (key_type == BTRFS_DIR_INDEX_KEY &&
		    btrfs_should_delete_dir_index(&del_list,
						  found_key.offset))
			goto next;
4556 4557

		filp->f_pos = found_key.offset;
4558
		is_curr = 1;
4559

C
Chris Mason 已提交
4560 4561
		di = btrfs_item_ptr(leaf, slot, struct btrfs_dir_item);
		di_cur = 0;
4562
		di_total = btrfs_item_size(leaf, item);
4563 4564

		while (di_cur < di_total) {
4565 4566
			struct btrfs_key location;

4567 4568 4569
			if (verify_dir_item(root, leaf, di))
				break;

4570
			name_len = btrfs_dir_name_len(leaf, di);
4571
			if (name_len <= sizeof(tmp_name)) {
4572 4573 4574
				name_ptr = tmp_name;
			} else {
				name_ptr = kmalloc(name_len, GFP_NOFS);
4575 4576 4577 4578
				if (!name_ptr) {
					ret = -ENOMEM;
					goto err;
				}
4579 4580 4581 4582 4583 4584
			}
			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);
4585

4586

4587
			/* is this a reference to our own snapshot? If so
A
Arne Jansen 已提交
4588 4589 4590 4591 4592 4593 4594
			 * skip it.
			 *
			 * In contrast to old kernels, we insert the snapshot's
			 * dir item and dir index after it has been created, so
			 * we won't find a reference to our own snapshot. We
			 * still keep the following code for backward
			 * compatibility.
4595 4596 4597 4598 4599 4600
			 */
			if (location.type == BTRFS_ROOT_ITEM_KEY &&
			    location.objectid == root->root_key.objectid) {
				over = 0;
				goto skip;
			}
4601
			over = filldir(dirent, name_ptr, name_len,
4602
				       found_key.offset, location.objectid,
C
Chris Mason 已提交
4603
				       d_type);
4604

4605
skip:
4606 4607 4608
			if (name_ptr != tmp_name)
				kfree(name_ptr);

C
Chris Mason 已提交
4609 4610
			if (over)
				goto nopos;
J
Josef Bacik 已提交
4611
			di_len = btrfs_dir_name_len(leaf, di) +
4612
				 btrfs_dir_data_len(leaf, di) + sizeof(*di);
C
Chris Mason 已提交
4613 4614 4615
			di_cur += di_len;
			di = (struct btrfs_dir_item *)((char *)di + di_len);
		}
4616 4617
next:
		path->slots[0]++;
C
Chris Mason 已提交
4618
	}
4619

4620 4621 4622 4623 4624 4625 4626 4627 4628
	if (key_type == BTRFS_DIR_INDEX_KEY) {
		if (is_curr)
			filp->f_pos++;
		ret = btrfs_readdir_delayed_dir_index(filp, dirent, filldir,
						      &ins_list);
		if (ret)
			goto nopos;
	}

4629
	/* Reached end of directory/root. Bump pos past the last item. */
4630
	if (key_type == BTRFS_DIR_INDEX_KEY)
4631 4632 4633 4634 4635
		/*
		 * 32-bit glibc will use getdents64, but then strtol -
		 * so the last number we can serve is this.
		 */
		filp->f_pos = 0x7fffffff;
4636 4637
	else
		filp->f_pos++;
C
Chris Mason 已提交
4638 4639 4640
nopos:
	ret = 0;
err:
4641 4642
	if (key_type == BTRFS_DIR_INDEX_KEY)
		btrfs_put_delayed_items(&ins_list, &del_list);
C
Chris Mason 已提交
4643 4644 4645 4646
	btrfs_free_path(path);
	return ret;
}

4647
int btrfs_write_inode(struct inode *inode, struct writeback_control *wbc)
C
Chris Mason 已提交
4648 4649 4650 4651
{
	struct btrfs_root *root = BTRFS_I(inode)->root;
	struct btrfs_trans_handle *trans;
	int ret = 0;
4652
	bool nolock = false;
C
Chris Mason 已提交
4653

4654
	if (test_bit(BTRFS_INODE_DUMMY, &BTRFS_I(inode)->runtime_flags))
4655 4656
		return 0;

4657
	if (btrfs_fs_closing(root->fs_info) && btrfs_is_free_space_inode(inode))
4658
		nolock = true;
4659

4660
	if (wbc->sync_mode == WB_SYNC_ALL) {
4661
		if (nolock)
4662
			trans = btrfs_join_transaction_nolock(root);
4663
		else
4664
			trans = btrfs_join_transaction(root);
4665 4666
		if (IS_ERR(trans))
			return PTR_ERR(trans);
4667
		ret = btrfs_commit_transaction(trans, root);
C
Chris Mason 已提交
4668 4669 4670 4671 4672
	}
	return ret;
}

/*
4673
 * This is somewhat expensive, updating the tree every time the
C
Chris Mason 已提交
4674 4675 4676 4677
 * 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.
 */
4678
int btrfs_dirty_inode(struct inode *inode)
C
Chris Mason 已提交
4679 4680 4681
{
	struct btrfs_root *root = BTRFS_I(inode)->root;
	struct btrfs_trans_handle *trans;
4682 4683
	int ret;

4684
	if (test_bit(BTRFS_INODE_DUMMY, &BTRFS_I(inode)->runtime_flags))
4685
		return 0;
C
Chris Mason 已提交
4686

4687
	trans = btrfs_join_transaction(root);
4688 4689
	if (IS_ERR(trans))
		return PTR_ERR(trans);
4690 4691

	ret = btrfs_update_inode(trans, root, inode);
4692 4693 4694 4695
	if (ret && ret == -ENOSPC) {
		/* whoops, lets try again with the full transaction */
		btrfs_end_transaction(trans, root);
		trans = btrfs_start_transaction(root, 1);
4696 4697
		if (IS_ERR(trans))
			return PTR_ERR(trans);
4698

4699 4700
		ret = btrfs_update_inode(trans, root, inode);
	}
C
Chris Mason 已提交
4701
	btrfs_end_transaction(trans, root);
4702 4703
	if (BTRFS_I(inode)->delayed_node)
		btrfs_balance_delayed_items(root);
4704 4705 4706 4707 4708 4709 4710 4711

	return ret;
}

/*
 * This is a copy of file_update_time.  We need this so we can return error on
 * ENOSPC for updating the inode in the case of file write and mmap writes.
 */
4712 4713
static int btrfs_update_time(struct inode *inode, struct timespec *now,
			     int flags)
4714
{
4715 4716 4717 4718 4719
	struct btrfs_root *root = BTRFS_I(inode)->root;

	if (btrfs_root_readonly(root))
		return -EROFS;

4720
	if (flags & S_VERSION)
4721
		inode_inc_iversion(inode);
4722 4723 4724 4725 4726 4727 4728
	if (flags & S_CTIME)
		inode->i_ctime = *now;
	if (flags & S_MTIME)
		inode->i_mtime = *now;
	if (flags & S_ATIME)
		inode->i_atime = *now;
	return btrfs_dirty_inode(inode);
C
Chris Mason 已提交
4729 4730
}

C
Chris Mason 已提交
4731 4732 4733 4734 4735
/*
 * find the highest existing sequence number in a directory
 * and then set the in-memory index_cnt variable to reflect
 * free sequence numbers
 */
4736 4737 4738 4739 4740 4741 4742 4743
static int btrfs_set_inode_index_count(struct inode *inode)
{
	struct btrfs_root *root = BTRFS_I(inode)->root;
	struct btrfs_key key, found_key;
	struct btrfs_path *path;
	struct extent_buffer *leaf;
	int ret;

L
Li Zefan 已提交
4744
	key.objectid = btrfs_ino(inode);
4745 4746 4747 4748 4749 4750 4751 4752 4753 4754 4755 4756 4757 4758 4759 4760 4761 4762 4763 4764 4765 4766 4767 4768 4769 4770 4771 4772 4773 4774 4775
	btrfs_set_key_type(&key, BTRFS_DIR_INDEX_KEY);
	key.offset = (u64)-1;

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

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

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

	path->slots[0]--;

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

L
Li Zefan 已提交
4776
	if (found_key.objectid != btrfs_ino(inode) ||
4777 4778 4779 4780 4781 4782 4783 4784 4785 4786 4787
	    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
4788 4789 4790 4791
/*
 * 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
 */
4792
int btrfs_set_inode_index(struct inode *dir, u64 *index)
4793 4794 4795 4796
{
	int ret = 0;

	if (BTRFS_I(dir)->index_cnt == (u64)-1) {
4797 4798 4799 4800 4801 4802
		ret = btrfs_inode_delayed_dir_index_count(dir);
		if (ret) {
			ret = btrfs_set_inode_index_count(dir);
			if (ret)
				return ret;
		}
4803 4804
	}

4805
	*index = BTRFS_I(dir)->index_cnt;
4806 4807 4808 4809 4810
	BTRFS_I(dir)->index_cnt++;

	return ret;
}

C
Chris Mason 已提交
4811 4812
static struct inode *btrfs_new_inode(struct btrfs_trans_handle *trans,
				     struct btrfs_root *root,
4813
				     struct inode *dir,
4814
				     const char *name, int name_len,
A
Al Viro 已提交
4815 4816
				     u64 ref_objectid, u64 objectid,
				     umode_t mode, u64 *index)
C
Chris Mason 已提交
4817 4818
{
	struct inode *inode;
4819
	struct btrfs_inode_item *inode_item;
C
Chris Mason 已提交
4820
	struct btrfs_key *location;
4821
	struct btrfs_path *path;
4822 4823 4824 4825
	struct btrfs_inode_ref *ref;
	struct btrfs_key key[2];
	u32 sizes[2];
	unsigned long ptr;
C
Chris Mason 已提交
4826 4827 4828
	int ret;
	int owner;

4829
	path = btrfs_alloc_path();
4830 4831
	if (!path)
		return ERR_PTR(-ENOMEM);
4832

C
Chris Mason 已提交
4833
	inode = new_inode(root->fs_info->sb);
4834 4835
	if (!inode) {
		btrfs_free_path(path);
C
Chris Mason 已提交
4836
		return ERR_PTR(-ENOMEM);
4837
	}
C
Chris Mason 已提交
4838

4839 4840 4841 4842 4843 4844
	/*
	 * we have to initialize this early, so we can reclaim the inode
	 * number if we fail afterwards in this function.
	 */
	inode->i_ino = objectid;

4845
	if (dir) {
4846 4847
		trace_btrfs_inode_request(dir);

4848
		ret = btrfs_set_inode_index(dir, index);
4849
		if (ret) {
4850
			btrfs_free_path(path);
4851
			iput(inode);
4852
			return ERR_PTR(ret);
4853
		}
4854 4855 4856 4857 4858 4859 4860
	}
	/*
	 * index_cnt is ignored for everything but a dir,
	 * btrfs_get_inode_index_count has an explanation for the magic
	 * number
	 */
	BTRFS_I(inode)->index_cnt = 2;
C
Chris Mason 已提交
4861
	BTRFS_I(inode)->root = root;
4862
	BTRFS_I(inode)->generation = trans->transid;
4863
	inode->i_generation = BTRFS_I(inode)->generation;
4864

J
Josef Bacik 已提交
4865 4866 4867 4868 4869 4870 4871 4872
	/*
	 * We could have gotten an inode number from somebody who was fsynced
	 * and then removed in this same transaction, so let's just set full
	 * sync since it will be a full sync anyway and this will blow away the
	 * old info in the log.
	 */
	set_bit(BTRFS_INODE_NEEDS_FULL_SYNC, &BTRFS_I(inode)->runtime_flags);

4873
	if (S_ISDIR(mode))
C
Chris Mason 已提交
4874 4875 4876
		owner = 0;
	else
		owner = 1;
4877 4878 4879 4880 4881

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

M
Mark Fasheh 已提交
4882 4883 4884 4885 4886 4887
	/*
	 * Start new inodes with an inode_ref. This is slightly more
	 * efficient for small numbers of hard links since they will
	 * be packed into one item. Extended refs will kick in if we
	 * add more hard links than can fit in the ref item.
	 */
4888 4889 4890 4891 4892 4893 4894
	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);

4895
	path->leave_spinning = 1;
4896 4897
	ret = btrfs_insert_empty_items(trans, root, path, key, sizes, 2);
	if (ret != 0)
4898 4899
		goto fail;

4900
	inode_init_owner(inode, dir, mode);
4901
	inode_set_bytes(inode, 0);
C
Chris Mason 已提交
4902
	inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
4903 4904
	inode_item = btrfs_item_ptr(path->nodes[0], path->slots[0],
				  struct btrfs_inode_item);
4905 4906
	memset_extent_buffer(path->nodes[0], 0, (unsigned long)inode_item,
			     sizeof(*inode_item));
4907
	fill_inode_item(trans, path->nodes[0], inode_item, inode);
4908 4909 4910 4911

	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);
4912
	btrfs_set_inode_ref_index(path->nodes[0], ref, *index);
4913 4914 4915
	ptr = (unsigned long)(ref + 1);
	write_extent_buffer(path->nodes[0], name, ptr, name_len);

4916 4917 4918
	btrfs_mark_buffer_dirty(path->nodes[0]);
	btrfs_free_path(path);

C
Chris Mason 已提交
4919 4920 4921 4922 4923
	location = &BTRFS_I(inode)->location;
	location->objectid = objectid;
	location->offset = 0;
	btrfs_set_key_type(location, BTRFS_INODE_ITEM_KEY);

4924 4925
	btrfs_inherit_iflags(inode, dir);

4926
	if (S_ISREG(mode)) {
4927 4928
		if (btrfs_test_opt(root, NODATASUM))
			BTRFS_I(inode)->flags |= BTRFS_INODE_NODATASUM;
L
Liu Bo 已提交
4929
		if (btrfs_test_opt(root, NODATACOW))
4930 4931 4932
			BTRFS_I(inode)->flags |= BTRFS_INODE_NODATACOW;
	}

C
Chris Mason 已提交
4933
	insert_inode_hash(inode);
4934
	inode_tree_add(inode);
4935 4936

	trace_btrfs_inode_new(inode);
4937
	btrfs_set_inode_last_trans(trans, inode);
4938

4939 4940
	btrfs_update_root_times(trans, root);

C
Chris Mason 已提交
4941
	return inode;
4942
fail:
4943 4944
	if (dir)
		BTRFS_I(dir)->index_cnt--;
4945
	btrfs_free_path(path);
4946
	iput(inode);
4947
	return ERR_PTR(ret);
C
Chris Mason 已提交
4948 4949 4950 4951 4952 4953 4954
}

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
4955 4956 4957 4958 4959 4960
/*
 * 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.
 */
4961 4962 4963
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
4964
{
4965
	int ret = 0;
C
Chris Mason 已提交
4966
	struct btrfs_key key;
4967
	struct btrfs_root *root = BTRFS_I(parent_inode)->root;
L
Li Zefan 已提交
4968 4969
	u64 ino = btrfs_ino(inode);
	u64 parent_ino = btrfs_ino(parent_inode);
4970

L
Li Zefan 已提交
4971
	if (unlikely(ino == BTRFS_FIRST_FREE_OBJECTID)) {
4972 4973
		memcpy(&key, &BTRFS_I(inode)->root->root_key, sizeof(key));
	} else {
L
Li Zefan 已提交
4974
		key.objectid = ino;
4975 4976 4977 4978
		btrfs_set_key_type(&key, BTRFS_INODE_ITEM_KEY);
		key.offset = 0;
	}

L
Li Zefan 已提交
4979
	if (unlikely(ino == BTRFS_FIRST_FREE_OBJECTID)) {
4980 4981
		ret = btrfs_add_root_ref(trans, root->fs_info->tree_root,
					 key.objectid, root->root_key.objectid,
L
Li Zefan 已提交
4982
					 parent_ino, index, name, name_len);
4983
	} else if (add_backref) {
L
Li Zefan 已提交
4984 4985
		ret = btrfs_insert_inode_ref(trans, root, name, name_len, ino,
					     parent_ino, index);
4986
	}
C
Chris Mason 已提交
4987

4988 4989 4990
	/* Nothing to clean up yet */
	if (ret)
		return ret;
4991

4992 4993 4994
	ret = btrfs_insert_dir_item(trans, root, name, name_len,
				    parent_inode, &key,
				    btrfs_inode_type(inode), index);
C
Chris Mason 已提交
4995
	if (ret == -EEXIST || ret == -EOVERFLOW)
4996 4997 4998 4999
		goto fail_dir_item;
	else if (ret) {
		btrfs_abort_transaction(trans, root, ret);
		return ret;
C
Chris Mason 已提交
5000
	}
5001 5002 5003

	btrfs_i_size_write(parent_inode, parent_inode->i_size +
			   name_len * 2);
5004
	inode_inc_iversion(parent_inode);
5005 5006 5007 5008
	parent_inode->i_mtime = parent_inode->i_ctime = CURRENT_TIME;
	ret = btrfs_update_inode(trans, root, parent_inode);
	if (ret)
		btrfs_abort_transaction(trans, root, ret);
C
Chris Mason 已提交
5009
	return ret;
5010 5011 5012 5013 5014 5015 5016 5017 5018 5019 5020 5021 5022 5023 5024 5025 5026

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

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

		err = btrfs_del_inode_ref(trans, root, name, name_len,
					  ino, parent_ino, &local_index);
	}
	return ret;
C
Chris Mason 已提交
5027 5028 5029
}

static int btrfs_add_nondir(struct btrfs_trans_handle *trans,
5030 5031
			    struct inode *dir, struct dentry *dentry,
			    struct inode *inode, int backref, u64 index)
C
Chris Mason 已提交
5032
{
5033 5034 5035
	int err = btrfs_add_link(trans, dir, inode,
				 dentry->d_name.name, dentry->d_name.len,
				 backref, index);
C
Chris Mason 已提交
5036 5037 5038 5039 5040
	if (err > 0)
		err = -EEXIST;
	return err;
}

J
Josef Bacik 已提交
5041
static int btrfs_mknod(struct inode *dir, struct dentry *dentry,
A
Al Viro 已提交
5042
			umode_t mode, dev_t rdev)
J
Josef Bacik 已提交
5043 5044 5045
{
	struct btrfs_trans_handle *trans;
	struct btrfs_root *root = BTRFS_I(dir)->root;
5046
	struct inode *inode = NULL;
J
Josef Bacik 已提交
5047 5048 5049
	int err;
	int drop_inode = 0;
	u64 objectid;
5050
	u64 index = 0;
J
Josef Bacik 已提交
5051 5052 5053 5054

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

J
Josef Bacik 已提交
5055 5056 5057 5058 5059
	/*
	 * 2 for inode item and ref
	 * 2 for dir items
	 * 1 for xattr if selinux is on
	 */
5060 5061 5062
	trans = btrfs_start_transaction(root, 5);
	if (IS_ERR(trans))
		return PTR_ERR(trans);
5063

5064 5065 5066 5067
	err = btrfs_find_free_ino(root, &objectid);
	if (err)
		goto out_unlock;

5068
	inode = btrfs_new_inode(trans, root, dir, dentry->d_name.name,
L
Li Zefan 已提交
5069
				dentry->d_name.len, btrfs_ino(dir), objectid,
5070
				mode, &index);
5071 5072
	if (IS_ERR(inode)) {
		err = PTR_ERR(inode);
J
Josef Bacik 已提交
5073
		goto out_unlock;
5074
	}
J
Josef Bacik 已提交
5075

5076
	err = btrfs_init_inode_security(trans, inode, dir, &dentry->d_name);
J
Josef Bacik 已提交
5077 5078 5079 5080 5081
	if (err) {
		drop_inode = 1;
		goto out_unlock;
	}

5082 5083 5084 5085 5086 5087 5088 5089
	/*
	* If the active LSM wants to access the inode during
	* d_instantiate it needs these. Smack checks to see
	* if the filesystem supports xattrs by looking at the
	* ops vector.
	*/

	inode->i_op = &btrfs_special_inode_operations;
5090
	err = btrfs_add_nondir(trans, dir, dentry, inode, 0, index);
J
Josef Bacik 已提交
5091 5092 5093 5094
	if (err)
		drop_inode = 1;
	else {
		init_special_inode(inode, inode->i_mode, rdev);
5095
		btrfs_update_inode(trans, root, inode);
5096
		d_instantiate(dentry, inode);
J
Josef Bacik 已提交
5097 5098
	}
out_unlock:
5099
	btrfs_end_transaction(trans, root);
5100
	btrfs_btree_balance_dirty(root);
J
Josef Bacik 已提交
5101 5102 5103 5104 5105 5106 5107
	if (drop_inode) {
		inode_dec_link_count(inode);
		iput(inode);
	}
	return err;
}

C
Chris Mason 已提交
5108
static int btrfs_create(struct inode *dir, struct dentry *dentry,
A
Al Viro 已提交
5109
			umode_t mode, bool excl)
C
Chris Mason 已提交
5110 5111 5112
{
	struct btrfs_trans_handle *trans;
	struct btrfs_root *root = BTRFS_I(dir)->root;
5113
	struct inode *inode = NULL;
5114
	int drop_inode_on_err = 0;
5115
	int err;
C
Chris Mason 已提交
5116
	u64 objectid;
5117
	u64 index = 0;
C
Chris Mason 已提交
5118

J
Josef Bacik 已提交
5119 5120 5121 5122 5123
	/*
	 * 2 for inode item and ref
	 * 2 for dir items
	 * 1 for xattr if selinux is on
	 */
5124 5125 5126
	trans = btrfs_start_transaction(root, 5);
	if (IS_ERR(trans))
		return PTR_ERR(trans);
J
Josef Bacik 已提交
5127

5128 5129 5130 5131
	err = btrfs_find_free_ino(root, &objectid);
	if (err)
		goto out_unlock;

5132
	inode = btrfs_new_inode(trans, root, dir, dentry->d_name.name,
L
Li Zefan 已提交
5133
				dentry->d_name.len, btrfs_ino(dir), objectid,
5134
				mode, &index);
5135 5136
	if (IS_ERR(inode)) {
		err = PTR_ERR(inode);
C
Chris Mason 已提交
5137
		goto out_unlock;
5138
	}
5139
	drop_inode_on_err = 1;
C
Chris Mason 已提交
5140

5141
	err = btrfs_init_inode_security(trans, inode, dir, &dentry->d_name);
5142
	if (err)
J
Josef Bacik 已提交
5143 5144
		goto out_unlock;

5145 5146 5147 5148
	err = btrfs_update_inode(trans, root, inode);
	if (err)
		goto out_unlock;

5149 5150 5151 5152 5153 5154 5155 5156 5157
	/*
	* If the active LSM wants to access the inode during
	* d_instantiate it needs these. Smack checks to see
	* if the filesystem supports xattrs by looking at the
	* ops vector.
	*/
	inode->i_fop = &btrfs_file_operations;
	inode->i_op = &btrfs_file_inode_operations;

5158
	err = btrfs_add_nondir(trans, dir, dentry, inode, 0, index);
C
Chris Mason 已提交
5159
	if (err)
5160 5161 5162 5163 5164 5165 5166
		goto out_unlock;

	inode->i_mapping->a_ops = &btrfs_aops;
	inode->i_mapping->backing_dev_info = &root->fs_info->bdi;
	BTRFS_I(inode)->io_tree.ops = &btrfs_extent_io_ops;
	d_instantiate(dentry, inode);

C
Chris Mason 已提交
5167
out_unlock:
5168
	btrfs_end_transaction(trans, root);
5169
	if (err && drop_inode_on_err) {
C
Chris Mason 已提交
5170 5171 5172
		inode_dec_link_count(inode);
		iput(inode);
	}
5173
	btrfs_btree_balance_dirty(root);
C
Chris Mason 已提交
5174 5175 5176 5177 5178 5179 5180 5181 5182
	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;
5183
	u64 index;
C
Chris Mason 已提交
5184 5185 5186
	int err;
	int drop_inode = 0;

5187 5188
	/* do not allow sys_link's with other subvols of the same device */
	if (root->objectid != BTRFS_I(inode)->root->objectid)
5189
		return -EXDEV;
5190

M
Mark Fasheh 已提交
5191
	if (inode->i_nlink >= BTRFS_LINK_MAX)
5192
		return -EMLINK;
5193

5194
	err = btrfs_set_inode_index(dir, &index);
5195 5196 5197
	if (err)
		goto fail;

5198
	/*
M
Miao Xie 已提交
5199
	 * 2 items for inode and inode ref
5200
	 * 2 items for dir items
M
Miao Xie 已提交
5201
	 * 1 item for parent inode
5202
	 */
M
Miao Xie 已提交
5203
	trans = btrfs_start_transaction(root, 5);
5204 5205 5206 5207
	if (IS_ERR(trans)) {
		err = PTR_ERR(trans);
		goto fail;
	}
5208

5209
	btrfs_inc_nlink(inode);
5210
	inode_inc_iversion(inode);
5211
	inode->i_ctime = CURRENT_TIME;
A
Al Viro 已提交
5212
	ihold(inode);
5213
	set_bit(BTRFS_INODE_COPY_EVERYTHING, &BTRFS_I(inode)->runtime_flags);
5214

5215
	err = btrfs_add_nondir(trans, dir, dentry, inode, 1, index);
5216

5217
	if (err) {
5218
		drop_inode = 1;
5219
	} else {
5220
		struct dentry *parent = dentry->d_parent;
5221
		err = btrfs_update_inode(trans, root, inode);
5222 5223
		if (err)
			goto fail;
5224
		d_instantiate(dentry, inode);
5225
		btrfs_log_new_name(trans, inode, NULL, parent);
5226
	}
C
Chris Mason 已提交
5227

5228
	btrfs_end_transaction(trans, root);
5229
fail:
C
Chris Mason 已提交
5230 5231 5232 5233
	if (drop_inode) {
		inode_dec_link_count(inode);
		iput(inode);
	}
5234
	btrfs_btree_balance_dirty(root);
C
Chris Mason 已提交
5235 5236 5237
	return err;
}

5238
static int btrfs_mkdir(struct inode *dir, struct dentry *dentry, umode_t mode)
C
Chris Mason 已提交
5239
{
5240
	struct inode *inode = NULL;
C
Chris Mason 已提交
5241 5242 5243 5244
	struct btrfs_trans_handle *trans;
	struct btrfs_root *root = BTRFS_I(dir)->root;
	int err = 0;
	int drop_on_err = 0;
5245
	u64 objectid = 0;
5246
	u64 index = 0;
C
Chris Mason 已提交
5247

J
Josef Bacik 已提交
5248 5249 5250 5251 5252
	/*
	 * 2 items for inode and ref
	 * 2 items for dir items
	 * 1 for xattr if selinux is on
	 */
5253 5254 5255
	trans = btrfs_start_transaction(root, 5);
	if (IS_ERR(trans))
		return PTR_ERR(trans);
C
Chris Mason 已提交
5256

5257 5258 5259 5260
	err = btrfs_find_free_ino(root, &objectid);
	if (err)
		goto out_fail;

5261
	inode = btrfs_new_inode(trans, root, dir, dentry->d_name.name,
L
Li Zefan 已提交
5262
				dentry->d_name.len, btrfs_ino(dir), objectid,
5263
				S_IFDIR | mode, &index);
C
Chris Mason 已提交
5264 5265 5266 5267
	if (IS_ERR(inode)) {
		err = PTR_ERR(inode);
		goto out_fail;
	}
5268

C
Chris Mason 已提交
5269
	drop_on_err = 1;
J
Josef Bacik 已提交
5270

5271
	err = btrfs_init_inode_security(trans, inode, dir, &dentry->d_name);
J
Josef Bacik 已提交
5272 5273 5274
	if (err)
		goto out_fail;

C
Chris Mason 已提交
5275 5276 5277
	inode->i_op = &btrfs_dir_inode_operations;
	inode->i_fop = &btrfs_dir_file_operations;

5278
	btrfs_i_size_write(inode, 0);
C
Chris Mason 已提交
5279 5280 5281
	err = btrfs_update_inode(trans, root, inode);
	if (err)
		goto out_fail;
5282

5283 5284
	err = btrfs_add_link(trans, dir, inode, dentry->d_name.name,
			     dentry->d_name.len, 0, index);
C
Chris Mason 已提交
5285 5286
	if (err)
		goto out_fail;
5287

C
Chris Mason 已提交
5288 5289 5290 5291
	d_instantiate(dentry, inode);
	drop_on_err = 0;

out_fail:
5292
	btrfs_end_transaction(trans, root);
C
Chris Mason 已提交
5293 5294
	if (drop_on_err)
		iput(inode);
5295
	btrfs_btree_balance_dirty(root);
C
Chris Mason 已提交
5296 5297 5298
	return err;
}

C
Chris Mason 已提交
5299 5300 5301 5302
/* 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.
 */
5303 5304
static int merge_extent_mapping(struct extent_map_tree *em_tree,
				struct extent_map *existing,
5305 5306
				struct extent_map *em,
				u64 map_start, u64 map_len)
5307 5308 5309
{
	u64 start_diff;

5310 5311 5312 5313
	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
5314 5315
	if (em->block_start < EXTENT_MAP_LAST_BYTE &&
	    !test_bit(EXTENT_FLAG_COMPRESSED, &em->flags)) {
5316
		em->block_start += start_diff;
C
Chris Mason 已提交
5317 5318
		em->block_len -= start_diff;
	}
5319
	return add_extent_mapping(em_tree, em);
5320 5321
}

C
Chris Mason 已提交
5322 5323 5324 5325 5326 5327 5328 5329 5330 5331 5332
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;
5333
	int compress_type;
C
Chris Mason 已提交
5334 5335

	WARN_ON(pg_offset != 0);
5336
	compress_type = btrfs_file_extent_compression(leaf, item);
C
Chris Mason 已提交
5337 5338 5339 5340
	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);
5341 5342
	if (!tmp)
		return -ENOMEM;
C
Chris Mason 已提交
5343 5344 5345 5346
	ptr = btrfs_file_extent_inline_start(item);

	read_extent_buffer(leaf, tmp, ptr, inline_size);

5347
	max_size = min_t(unsigned long, PAGE_CACHE_SIZE, max_size);
5348 5349
	ret = btrfs_decompress(compress_type, tmp, page,
			       extent_offset, inline_size, max_size);
C
Chris Mason 已提交
5350
	if (ret) {
5351
		char *kaddr = kmap_atomic(page);
C
Chris Mason 已提交
5352 5353 5354 5355
		unsigned long copy_size = min_t(u64,
				  PAGE_CACHE_SIZE - pg_offset,
				  max_size - extent_offset);
		memset(kaddr + pg_offset, 0, copy_size);
5356
		kunmap_atomic(kaddr);
C
Chris Mason 已提交
5357 5358 5359 5360 5361
	}
	kfree(tmp);
	return 0;
}

C
Chris Mason 已提交
5362 5363
/*
 * a bit scary, this does extent mapping from logical file offset to the disk.
C
Chris Mason 已提交
5364 5365
 * 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
5366 5367 5368 5369
 * 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
5370

5371
struct extent_map *btrfs_get_extent(struct inode *inode, struct page *page,
5372
				    size_t pg_offset, u64 start, u64 len,
5373 5374 5375 5376
				    int create)
{
	int ret;
	int err = 0;
5377
	u64 bytenr;
5378 5379
	u64 extent_start = 0;
	u64 extent_end = 0;
L
Li Zefan 已提交
5380
	u64 objectid = btrfs_ino(inode);
5381
	u32 found_type;
5382
	struct btrfs_path *path = NULL;
5383 5384
	struct btrfs_root *root = BTRFS_I(inode)->root;
	struct btrfs_file_extent_item *item;
5385 5386
	struct extent_buffer *leaf;
	struct btrfs_key found_key;
5387 5388
	struct extent_map *em = NULL;
	struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
5389
	struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
5390
	struct btrfs_trans_handle *trans = NULL;
5391
	int compress_type;
5392 5393

again:
5394
	read_lock(&em_tree->lock);
5395
	em = lookup_extent_mapping(em_tree, start, len);
5396 5397
	if (em)
		em->bdev = root->fs_info->fs_devices->latest_bdev;
5398
	read_unlock(&em_tree->lock);
5399

5400
	if (em) {
5401 5402 5403
		if (em->start > start || em->start + em->len <= start)
			free_extent_map(em);
		else if (em->block_start == EXTENT_MAP_INLINE && page)
5404 5405 5406
			free_extent_map(em);
		else
			goto out;
5407
	}
5408
	em = alloc_extent_map();
5409
	if (!em) {
5410 5411
		err = -ENOMEM;
		goto out;
5412
	}
5413
	em->bdev = root->fs_info->fs_devices->latest_bdev;
5414
	em->start = EXTENT_MAP_HOLE;
5415
	em->orig_start = EXTENT_MAP_HOLE;
5416
	em->len = (u64)-1;
C
Chris Mason 已提交
5417
	em->block_len = (u64)-1;
5418 5419 5420

	if (!path) {
		path = btrfs_alloc_path();
J
Josef Bacik 已提交
5421 5422 5423 5424 5425 5426 5427 5428 5429
		if (!path) {
			err = -ENOMEM;
			goto out;
		}
		/*
		 * Chances are we'll be called again, so go ahead and do
		 * readahead
		 */
		path->reada = 1;
5430 5431
	}

5432 5433
	ret = btrfs_lookup_file_extent(trans, root, path,
				       objectid, start, trans != NULL);
5434 5435 5436 5437 5438 5439 5440 5441 5442 5443 5444
	if (ret < 0) {
		err = ret;
		goto out;
	}

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

5445 5446
	leaf = path->nodes[0];
	item = btrfs_item_ptr(leaf, path->slots[0],
5447 5448
			      struct btrfs_file_extent_item);
	/* are we inside the extent that was found? */
5449 5450 5451
	btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
	found_type = btrfs_key_type(&found_key);
	if (found_key.objectid != objectid ||
5452 5453 5454 5455
	    found_type != BTRFS_EXTENT_DATA_KEY) {
		goto not_found;
	}

5456 5457
	found_type = btrfs_file_extent_type(leaf, item);
	extent_start = found_key.offset;
5458
	compress_type = btrfs_file_extent_compression(leaf, item);
Y
Yan Zheng 已提交
5459 5460
	if (found_type == BTRFS_FILE_EXTENT_REG ||
	    found_type == BTRFS_FILE_EXTENT_PREALLOC) {
5461
		extent_end = extent_start +
5462
		       btrfs_file_extent_num_bytes(leaf, item);
Y
Yan Zheng 已提交
5463 5464 5465 5466 5467 5468 5469 5470 5471 5472 5473 5474 5475 5476
	} 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;
5477
			}
Y
Yan Zheng 已提交
5478 5479 5480
			if (ret > 0)
				goto not_found;
			leaf = path->nodes[0];
5481
		}
Y
Yan Zheng 已提交
5482 5483 5484 5485 5486 5487 5488
		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;
5489
		em->orig_start = start;
Y
Yan Zheng 已提交
5490 5491 5492 5493
		em->len = found_key.offset - start;
		goto not_found_em;
	}

Y
Yan Zheng 已提交
5494 5495
	if (found_type == BTRFS_FILE_EXTENT_REG ||
	    found_type == BTRFS_FILE_EXTENT_PREALLOC) {
Y
Yan Zheng 已提交
5496 5497
		em->start = extent_start;
		em->len = extent_end - extent_start;
5498 5499
		em->orig_start = extent_start -
				 btrfs_file_extent_offset(leaf, item);
5500 5501
		em->orig_block_len = btrfs_file_extent_disk_num_bytes(leaf,
								      item);
5502 5503
		bytenr = btrfs_file_extent_disk_bytenr(leaf, item);
		if (bytenr == 0) {
5504
			em->block_start = EXTENT_MAP_HOLE;
5505 5506
			goto insert;
		}
5507
		if (compress_type != BTRFS_COMPRESS_NONE) {
C
Chris Mason 已提交
5508
			set_bit(EXTENT_FLAG_COMPRESSED, &em->flags);
5509
			em->compress_type = compress_type;
C
Chris Mason 已提交
5510
			em->block_start = bytenr;
5511
			em->block_len = em->orig_block_len;
C
Chris Mason 已提交
5512 5513 5514 5515
		} else {
			bytenr += btrfs_file_extent_offset(leaf, item);
			em->block_start = bytenr;
			em->block_len = em->len;
Y
Yan Zheng 已提交
5516 5517
			if (found_type == BTRFS_FILE_EXTENT_PREALLOC)
				set_bit(EXTENT_FLAG_PREALLOC, &em->flags);
C
Chris Mason 已提交
5518
		}
5519 5520
		goto insert;
	} else if (found_type == BTRFS_FILE_EXTENT_INLINE) {
5521
		unsigned long ptr;
5522
		char *map;
5523 5524 5525
		size_t size;
		size_t extent_offset;
		size_t copy_size;
5526

5527
		em->block_start = EXTENT_MAP_INLINE;
C
Chris Mason 已提交
5528
		if (!page || create) {
5529
			em->start = extent_start;
Y
Yan Zheng 已提交
5530
			em->len = extent_end - extent_start;
5531 5532
			goto out;
		}
5533

Y
Yan Zheng 已提交
5534 5535
		size = btrfs_file_extent_inline_len(leaf, item);
		extent_offset = page_offset(page) + pg_offset - extent_start;
5536
		copy_size = min_t(u64, PAGE_CACHE_SIZE - pg_offset,
5537 5538
				size - extent_offset);
		em->start = extent_start + extent_offset;
5539 5540
		em->len = (copy_size + root->sectorsize - 1) &
			~((u64)root->sectorsize - 1);
5541
		em->orig_block_len = em->len;
5542
		em->orig_start = em->start;
5543
		if (compress_type) {
C
Chris Mason 已提交
5544
			set_bit(EXTENT_FLAG_COMPRESSED, &em->flags);
5545 5546
			em->compress_type = compress_type;
		}
5547
		ptr = btrfs_file_extent_inline_start(item) + extent_offset;
5548
		if (create == 0 && !PageUptodate(page)) {
5549 5550
			if (btrfs_file_extent_compression(leaf, item) !=
			    BTRFS_COMPRESS_NONE) {
C
Chris Mason 已提交
5551 5552 5553
				ret = uncompress_inline(path, inode, page,
							pg_offset,
							extent_offset, item);
5554
				BUG_ON(ret); /* -ENOMEM */
C
Chris Mason 已提交
5555 5556 5557 5558
			} else {
				map = kmap(page);
				read_extent_buffer(leaf, map + pg_offset, ptr,
						   copy_size);
5559 5560 5561 5562 5563
				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
5564 5565
				kunmap(page);
			}
5566 5567
			flush_dcache_page(page);
		} else if (create && PageUptodate(page)) {
5568
			BUG();
5569 5570 5571 5572
			if (!trans) {
				kunmap(page);
				free_extent_map(em);
				em = NULL;
C
Chris Mason 已提交
5573

5574
				btrfs_release_path(path);
5575
				trans = btrfs_join_transaction(root);
C
Chris Mason 已提交
5576

5577 5578
				if (IS_ERR(trans))
					return ERR_CAST(trans);
5579 5580
				goto again;
			}
C
Chris Mason 已提交
5581
			map = kmap(page);
5582
			write_extent_buffer(leaf, map + pg_offset, ptr,
5583
					    copy_size);
C
Chris Mason 已提交
5584
			kunmap(page);
5585
			btrfs_mark_buffer_dirty(leaf);
5586
		}
5587
		set_extent_uptodate(io_tree, em->start,
5588
				    extent_map_end(em) - 1, NULL, GFP_NOFS);
5589 5590
		goto insert;
	} else {
J
Julia Lawall 已提交
5591
		WARN(1, KERN_ERR "btrfs unknown found_type %d\n", found_type);
5592 5593 5594
	}
not_found:
	em->start = start;
5595
	em->orig_start = start;
5596
	em->len = len;
5597
not_found_em:
5598
	em->block_start = EXTENT_MAP_HOLE;
Y
Yan Zheng 已提交
5599
	set_bit(EXTENT_FLAG_VACANCY, &em->flags);
5600
insert:
5601
	btrfs_release_path(path);
5602
	if (em->start > start || extent_map_end(em) <= start) {
C
Chris Mason 已提交
5603 5604 5605 5606 5607
		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);
5608 5609 5610
		err = -EIO;
		goto out;
	}
5611 5612

	err = 0;
5613
	write_lock(&em_tree->lock);
5614
	ret = add_extent_mapping(em_tree, em);
5615 5616 5617 5618
	/* 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
	 */
5619
	if (ret == -EEXIST) {
5620
		struct extent_map *existing;
5621 5622 5623

		ret = 0;

5624
		existing = lookup_extent_mapping(em_tree, start, len);
5625 5626 5627 5628 5629
		if (existing && (existing->start > start ||
		    existing->start + existing->len <= start)) {
			free_extent_map(existing);
			existing = NULL;
		}
5630 5631 5632 5633 5634
		if (!existing) {
			existing = lookup_extent_mapping(em_tree, em->start,
							 em->len);
			if (existing) {
				err = merge_extent_mapping(em_tree, existing,
5635 5636
							   em, start,
							   root->sectorsize);
5637 5638 5639 5640 5641 5642 5643 5644 5645 5646 5647 5648 5649
				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;
5650
			err = 0;
5651 5652
		}
	}
5653
	write_unlock(&em_tree->lock);
5654
out:
5655

5656 5657
	if (em)
		trace_btrfs_get_extent(root, em);
5658

5659 5660
	if (path)
		btrfs_free_path(path);
5661 5662
	if (trans) {
		ret = btrfs_end_transaction(trans, root);
C
Chris Mason 已提交
5663
		if (!err)
5664 5665 5666 5667 5668 5669
			err = ret;
	}
	if (err) {
		free_extent_map(em);
		return ERR_PTR(err);
	}
5670
	BUG_ON(!em); /* Error is always set */
5671 5672 5673
	return em;
}

5674 5675 5676 5677 5678 5679 5680 5681 5682 5683 5684 5685 5686 5687 5688 5689 5690
struct extent_map *btrfs_get_extent_fiemap(struct inode *inode, struct page *page,
					   size_t pg_offset, u64 start, u64 len,
					   int create)
{
	struct extent_map *em;
	struct extent_map *hole_em = NULL;
	u64 range_start = start;
	u64 end;
	u64 found;
	u64 found_end;
	int err = 0;

	em = btrfs_get_extent(inode, page, pg_offset, start, len, create);
	if (IS_ERR(em))
		return em;
	if (em) {
		/*
5691 5692 5693 5694
		 * if our em maps to
		 * -  a hole or
		 * -  a pre-alloc extent,
		 * there might actually be delalloc bytes behind it.
5695
		 */
5696 5697
		if (em->block_start != EXTENT_MAP_HOLE &&
		    !test_bit(EXTENT_FLAG_PREALLOC, &em->flags))
5698 5699 5700 5701 5702 5703 5704 5705 5706 5707 5708 5709 5710 5711 5712 5713 5714 5715 5716 5717 5718 5719 5720 5721 5722 5723 5724 5725 5726 5727 5728 5729 5730 5731 5732 5733 5734 5735 5736 5737 5738
			return em;
		else
			hole_em = em;
	}

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

	em = NULL;

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

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

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

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

5739
		em = alloc_extent_map();
5740 5741 5742 5743 5744 5745 5746 5747 5748 5749 5750 5751 5752 5753 5754 5755 5756 5757 5758 5759 5760 5761 5762 5763 5764 5765 5766 5767 5768 5769 5770 5771 5772 5773 5774 5775 5776 5777 5778
		if (!em) {
			err = -ENOMEM;
			goto out;
		}
		/*
		 * when btrfs_get_extent can't find anything it
		 * returns one huge hole
		 *
		 * make sure what it found really fits our range, and
		 * adjust to make sure it is based on the start from
		 * the caller
		 */
		if (hole_em) {
			u64 calc_end = extent_map_end(hole_em);

			if (calc_end <= start || (hole_em->start > end)) {
				free_extent_map(hole_em);
				hole_em = NULL;
			} else {
				hole_start = max(hole_em->start, start);
				hole_len = calc_end - hole_start;
			}
		}
		em->bdev = NULL;
		if (hole_em && range_start > hole_start) {
			/* our hole starts before our delalloc, so we
			 * have to return just the parts of the hole
			 * that go until  the delalloc starts
			 */
			em->len = min(hole_len,
				      range_start - hole_start);
			em->start = hole_start;
			em->orig_start = hole_start;
			/*
			 * don't adjust block start at all,
			 * it is fixed at EXTENT_MAP_HOLE
			 */
			em->block_start = hole_em->block_start;
			em->block_len = hole_len;
5779 5780
			if (test_bit(EXTENT_FLAG_PREALLOC, &hole_em->flags))
				set_bit(EXTENT_FLAG_PREALLOC, &em->flags);
5781 5782 5783 5784 5785 5786 5787 5788 5789 5790 5791 5792 5793 5794 5795 5796 5797 5798 5799 5800
		} else {
			em->start = range_start;
			em->len = found;
			em->orig_start = range_start;
			em->block_start = EXTENT_MAP_DELALLOC;
			em->block_len = found;
		}
	} else if (hole_em) {
		return hole_em;
	}
out:

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

5801 5802 5803 5804 5805
static struct extent_map *btrfs_new_extent_direct(struct inode *inode,
						  u64 start, u64 len)
{
	struct btrfs_root *root = BTRFS_I(inode)->root;
	struct btrfs_trans_handle *trans;
5806
	struct extent_map *em;
5807 5808 5809 5810
	struct btrfs_key ins;
	u64 alloc_hint;
	int ret;

5811
	trans = btrfs_join_transaction(root);
5812 5813
	if (IS_ERR(trans))
		return ERR_CAST(trans);
5814 5815 5816 5817 5818

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

	alloc_hint = get_extent_allocation_hint(inode, start, len);
	ret = btrfs_reserve_extent(trans, root, len, root->sectorsize, 0,
5819
				   alloc_hint, &ins, 1);
5820 5821 5822 5823 5824
	if (ret) {
		em = ERR_PTR(ret);
		goto out;
	}

5825 5826 5827 5828
	em = create_pinned_em(inode, start, ins.offset, start, ins.objectid,
			      ins.offset, ins.offset, 0);
	if (IS_ERR(em))
		goto out;
5829 5830 5831 5832 5833 5834 5835 5836 5837 5838 5839 5840

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

5841 5842 5843 5844 5845 5846 5847 5848 5849 5850 5851 5852 5853 5854 5855 5856 5857 5858 5859 5860 5861 5862 5863 5864
/*
 * returns 1 when the nocow is safe, < 1 on error, 0 if the
 * block must be cow'd
 */
static noinline int can_nocow_odirect(struct btrfs_trans_handle *trans,
				      struct inode *inode, u64 offset, u64 len)
{
	struct btrfs_path *path;
	int ret;
	struct extent_buffer *leaf;
	struct btrfs_root *root = BTRFS_I(inode)->root;
	struct btrfs_file_extent_item *fi;
	struct btrfs_key key;
	u64 disk_bytenr;
	u64 backref_offset;
	u64 extent_end;
	u64 num_bytes;
	int slot;
	int found_type;

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

L
Li Zefan 已提交
5865
	ret = btrfs_lookup_file_extent(trans, root, path, btrfs_ino(inode),
5866 5867 5868 5869 5870 5871 5872 5873 5874 5875 5876 5877 5878 5879 5880 5881
				       offset, 0);
	if (ret < 0)
		goto out;

	slot = path->slots[0];
	if (ret == 1) {
		if (slot == 0) {
			/* can't find the item, must cow */
			ret = 0;
			goto out;
		}
		slot--;
	}
	ret = 0;
	leaf = path->nodes[0];
	btrfs_item_key_to_cpu(leaf, &key, slot);
L
Li Zefan 已提交
5882
	if (key.objectid != btrfs_ino(inode) ||
5883 5884 5885 5886 5887 5888 5889 5890 5891 5892 5893 5894 5895 5896 5897 5898 5899 5900 5901 5902 5903 5904 5905 5906 5907 5908 5909 5910 5911 5912 5913 5914 5915
	    key.type != BTRFS_EXTENT_DATA_KEY) {
		/* not our file or wrong item type, must cow */
		goto out;
	}

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

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

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

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

	/*
	 * look for other files referencing this extent, if we
	 * find any we must cow
	 */
L
Li Zefan 已提交
5916
	if (btrfs_cross_ref_exist(trans, root, btrfs_ino(inode),
5917 5918 5919 5920 5921 5922 5923 5924 5925 5926 5927 5928 5929 5930 5931 5932 5933 5934 5935 5936 5937 5938 5939 5940
				  key.offset - backref_offset, disk_bytenr))
		goto out;

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

5941 5942 5943 5944 5945 5946 5947 5948 5949 5950 5951 5952 5953 5954 5955 5956 5957 5958 5959 5960 5961 5962 5963 5964 5965 5966 5967 5968 5969 5970 5971 5972 5973 5974 5975 5976 5977 5978 5979 5980 5981 5982 5983 5984 5985 5986 5987 5988 5989 5990 5991 5992 5993 5994 5995 5996 5997 5998 5999 6000 6001
static int lock_extent_direct(struct inode *inode, u64 lockstart, u64 lockend,
			      struct extent_state **cached_state, int writing)
{
	struct btrfs_ordered_extent *ordered;
	int ret = 0;

	while (1) {
		lock_extent_bits(&BTRFS_I(inode)->io_tree, lockstart, lockend,
				 0, cached_state);
		/*
		 * We're concerned with the entire range that we're going to be
		 * doing DIO to, so we need to make sure theres no ordered
		 * extents in this range.
		 */
		ordered = btrfs_lookup_ordered_range(inode, lockstart,
						     lockend - lockstart + 1);

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

		unlock_extent_cached(&BTRFS_I(inode)->io_tree, lockstart, lockend,
				     cached_state, GFP_NOFS);

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

			/*
			 * If we found a page that couldn't be invalidated just
			 * fall back to buffered.
			 */
			ret = invalidate_inode_pages2_range(inode->i_mapping,
					lockstart >> PAGE_CACHE_SHIFT,
					lockend >> PAGE_CACHE_SHIFT);
			if (ret)
				break;
		}

		cond_resched();
	}

	return ret;
}

6002 6003 6004
static struct extent_map *create_pinned_em(struct inode *inode, u64 start,
					   u64 len, u64 orig_start,
					   u64 block_start, u64 block_len,
6005
					   u64 orig_block_len, int type)
6006 6007 6008 6009 6010 6011 6012 6013 6014 6015 6016 6017 6018
{
	struct extent_map_tree *em_tree;
	struct extent_map *em;
	struct btrfs_root *root = BTRFS_I(inode)->root;
	int ret;

	em_tree = &BTRFS_I(inode)->extent_tree;
	em = alloc_extent_map();
	if (!em)
		return ERR_PTR(-ENOMEM);

	em->start = start;
	em->orig_start = orig_start;
6019 6020
	em->mod_start = start;
	em->mod_len = len;
6021 6022 6023 6024
	em->len = len;
	em->block_len = block_len;
	em->block_start = block_start;
	em->bdev = root->fs_info->fs_devices->latest_bdev;
6025
	em->orig_block_len = orig_block_len;
6026
	em->generation = -1;
6027 6028
	set_bit(EXTENT_FLAG_PINNED, &em->flags);
	if (type == BTRFS_ORDERED_PREALLOC)
6029
		set_bit(EXTENT_FLAG_FILLING, &em->flags);
6030 6031 6032 6033 6034 6035

	do {
		btrfs_drop_extent_cache(inode, em->start,
				em->start + em->len - 1, 0);
		write_lock(&em_tree->lock);
		ret = add_extent_mapping(em_tree, em);
6036 6037 6038
		if (!ret)
			list_move(&em->list,
				  &em_tree->modified_extents);
6039 6040 6041 6042 6043 6044 6045 6046 6047 6048 6049 6050
		write_unlock(&em_tree->lock);
	} while (ret == -EEXIST);

	if (ret) {
		free_extent_map(em);
		return ERR_PTR(ret);
	}

	return em;
}


6051 6052 6053 6054 6055
static int btrfs_get_blocks_direct(struct inode *inode, sector_t iblock,
				   struct buffer_head *bh_result, int create)
{
	struct extent_map *em;
	struct btrfs_root *root = BTRFS_I(inode)->root;
6056
	struct extent_state *cached_state = NULL;
6057
	u64 start = iblock << inode->i_blkbits;
6058
	u64 lockstart, lockend;
6059
	u64 len = bh_result->b_size;
6060
	struct btrfs_trans_handle *trans;
6061 6062 6063 6064 6065 6066 6067 6068
	int unlock_bits = EXTENT_LOCKED;
	int ret;

	if (create) {
		ret = btrfs_delalloc_reserve_space(inode, len);
		if (ret)
			return ret;
		unlock_bits |= EXTENT_DELALLOC | EXTENT_DIRTY;
6069 6070
	} else {
		len = min_t(u64, len, root->sectorsize);
6071 6072
	}

6073 6074 6075
	lockstart = start;
	lockend = start + len - 1;

6076 6077 6078 6079 6080 6081 6082 6083 6084 6085 6086 6087 6088 6089
	/*
	 * If this errors out it's because we couldn't invalidate pagecache for
	 * this range and we need to fallback to buffered.
	 */
	if (lock_extent_direct(inode, lockstart, lockend, &cached_state, create))
		return -ENOTBLK;

	if (create) {
		ret = set_extent_bit(&BTRFS_I(inode)->io_tree, lockstart,
				     lockend, EXTENT_DELALLOC, NULL,
				     &cached_state, GFP_NOFS);
		if (ret)
			goto unlock_err;
	}
6090 6091

	em = btrfs_get_extent(inode, NULL, 0, start, len, 0);
6092 6093 6094 6095
	if (IS_ERR(em)) {
		ret = PTR_ERR(em);
		goto unlock_err;
	}
6096 6097 6098 6099 6100 6101 6102 6103 6104 6105 6106 6107 6108 6109 6110 6111 6112 6113

	/*
	 * Ok for INLINE and COMPRESSED extents we need to fallback on buffered
	 * io.  INLINE is special, and we could probably kludge it in here, but
	 * it's still buffered so for safety lets just fall back to the generic
	 * buffered path.
	 *
	 * For COMPRESSED we _have_ to read the entire extent in so we can
	 * decompress it, so there will be buffering required no matter what we
	 * do, so go ahead and fallback to buffered.
	 *
	 * We return -ENOTBLK because thats what makes DIO go ahead and go back
	 * to buffered IO.  Don't blame me, this is the price we pay for using
	 * the generic code.
	 */
	if (test_bit(EXTENT_FLAG_COMPRESSED, &em->flags) ||
	    em->block_start == EXTENT_MAP_INLINE) {
		free_extent_map(em);
6114 6115
		ret = -ENOTBLK;
		goto unlock_err;
6116 6117 6118 6119 6120 6121
	}

	/* Just a good old fashioned hole, return */
	if (!create && (em->block_start == EXTENT_MAP_HOLE ||
			test_bit(EXTENT_FLAG_PREALLOC, &em->flags))) {
		free_extent_map(em);
6122 6123
		ret = 0;
		goto unlock_err;
6124 6125 6126 6127 6128 6129 6130 6131 6132 6133 6134
	}

	/*
	 * We don't allocate a new extent in the following cases
	 *
	 * 1) The inode is marked as NODATACOW.  In this case we'll just use the
	 * existing extent.
	 * 2) The extent is marked as PREALLOC.  We're good to go here and can
	 * just use the extent.
	 *
	 */
6135
	if (!create) {
6136 6137 6138
		len = min(len, em->len - (start - em->start));
		lockstart = start + len;
		goto unlock;
6139
	}
6140 6141 6142 6143 6144 6145

	if (test_bit(EXTENT_FLAG_PREALLOC, &em->flags) ||
	    ((BTRFS_I(inode)->flags & BTRFS_INODE_NODATACOW) &&
	     em->block_start != EXTENT_MAP_HOLE)) {
		int type;
		int ret;
6146
		u64 block_start;
6147 6148 6149 6150 6151

		if (test_bit(EXTENT_FLAG_PREALLOC, &em->flags))
			type = BTRFS_ORDERED_PREALLOC;
		else
			type = BTRFS_ORDERED_NOCOW;
6152
		len = min(len, em->len - (start - em->start));
6153
		block_start = em->block_start + (start - em->start);
6154 6155 6156 6157 6158 6159

		/*
		 * we're not going to log anything, but we do need
		 * to make sure the current transaction stays open
		 * while we look for nocow cross refs
		 */
6160
		trans = btrfs_join_transaction(root);
6161
		if (IS_ERR(trans))
6162 6163 6164
			goto must_cow;

		if (can_nocow_odirect(trans, inode, start, len) == 1) {
6165
			u64 orig_start = em->orig_start;
6166
			u64 orig_block_len = em->orig_block_len;
6167 6168 6169 6170 6171

			if (type == BTRFS_ORDERED_PREALLOC) {
				free_extent_map(em);
				em = create_pinned_em(inode, start, len,
						       orig_start,
6172 6173
						       block_start, len,
						       orig_block_len, type);
6174 6175 6176 6177 6178 6179
				if (IS_ERR(em)) {
					btrfs_end_transaction(trans, root);
					goto unlock_err;
				}
			}

6180 6181 6182 6183 6184
			ret = btrfs_add_ordered_extent_dio(inode, start,
					   block_start, len, len, type);
			btrfs_end_transaction(trans, root);
			if (ret) {
				free_extent_map(em);
6185
				goto unlock_err;
6186 6187
			}
			goto unlock;
6188
		}
6189
		btrfs_end_transaction(trans, root);
6190
	}
6191 6192 6193 6194 6195 6196
must_cow:
	/*
	 * this will cow the extent, reset the len in case we changed
	 * it above
	 */
	len = bh_result->b_size;
6197 6198
	free_extent_map(em);
	em = btrfs_new_extent_direct(inode, start, len);
6199 6200 6201 6202
	if (IS_ERR(em)) {
		ret = PTR_ERR(em);
		goto unlock_err;
	}
6203 6204
	len = min(len, em->len - (start - em->start));
unlock:
6205 6206
	bh_result->b_blocknr = (em->block_start + (start - em->start)) >>
		inode->i_blkbits;
6207
	bh_result->b_size = len;
6208 6209
	bh_result->b_bdev = em->bdev;
	set_buffer_mapped(bh_result);
6210 6211 6212 6213 6214 6215 6216 6217 6218 6219 6220
	if (create) {
		if (!test_bit(EXTENT_FLAG_PREALLOC, &em->flags))
			set_buffer_new(bh_result);

		/*
		 * Need to update the i_size under the extent lock so buffered
		 * readers will get the updated i_size when we unlock.
		 */
		if (start + len > i_size_read(inode))
			i_size_write(inode, start + len);
	}
6221

6222 6223 6224 6225 6226
	/*
	 * In the case of write we need to clear and unlock the entire range,
	 * in the case of read we need to unlock only the end area that we
	 * aren't using if there is any left over space.
	 */
L
Liu Bo 已提交
6227 6228 6229
	if (lockstart < lockend) {
		if (create && len < lockend - lockstart) {
			clear_extent_bit(&BTRFS_I(inode)->io_tree, lockstart,
6230 6231
					 lockstart + len - 1,
					 unlock_bits | EXTENT_DEFRAG, 1, 0,
L
Liu Bo 已提交
6232 6233 6234 6235 6236 6237 6238
					 &cached_state, GFP_NOFS);
			/*
			 * Beside unlock, we also need to cleanup reserved space
			 * for the left range by attaching EXTENT_DO_ACCOUNTING.
			 */
			clear_extent_bit(&BTRFS_I(inode)->io_tree,
					 lockstart + len, lockend,
6239 6240
					 unlock_bits | EXTENT_DO_ACCOUNTING |
					 EXTENT_DEFRAG, 1, 0, NULL, GFP_NOFS);
L
Liu Bo 已提交
6241 6242 6243 6244 6245 6246
		} else {
			clear_extent_bit(&BTRFS_I(inode)->io_tree, lockstart,
					 lockend, unlock_bits, 1, 0,
					 &cached_state, GFP_NOFS);
		}
	} else {
6247
		free_extent_state(cached_state);
L
Liu Bo 已提交
6248
	}
6249

6250 6251 6252
	free_extent_map(em);

	return 0;
6253 6254 6255 6256 6257 6258 6259 6260

unlock_err:
	if (create)
		unlock_bits |= EXTENT_DO_ACCOUNTING;

	clear_extent_bit(&BTRFS_I(inode)->io_tree, lockstart, lockend,
			 unlock_bits, 1, 0, &cached_state, GFP_NOFS);
	return ret;
6261 6262 6263 6264 6265 6266 6267 6268
}

struct btrfs_dio_private {
	struct inode *inode;
	u64 logical_offset;
	u64 disk_bytenr;
	u64 bytes;
	void *private;
M
Miao Xie 已提交
6269 6270 6271 6272 6273 6274 6275 6276

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

	/* IO errors */
	int errors;

	struct bio *orig_bio;
6277 6278 6279 6280
};

static void btrfs_endio_direct_read(struct bio *bio, int err)
{
M
Miao Xie 已提交
6281
	struct btrfs_dio_private *dip = bio->bi_private;
6282 6283 6284 6285 6286 6287 6288 6289 6290 6291 6292 6293
	struct bio_vec *bvec_end = bio->bi_io_vec + bio->bi_vcnt - 1;
	struct bio_vec *bvec = bio->bi_io_vec;
	struct inode *inode = dip->inode;
	struct btrfs_root *root = BTRFS_I(inode)->root;
	u64 start;

	start = dip->logical_offset;
	do {
		if (!(BTRFS_I(inode)->flags & BTRFS_INODE_NODATASUM)) {
			struct page *page = bvec->bv_page;
			char *kaddr;
			u32 csum = ~(u32)0;
6294
			u64 private = ~(u32)0;
6295 6296
			unsigned long flags;

6297 6298 6299
			if (get_state_private(&BTRFS_I(inode)->io_tree,
					      start, &private))
				goto failed;
6300
			local_irq_save(flags);
6301
			kaddr = kmap_atomic(page);
6302 6303 6304
			csum = btrfs_csum_data(root, kaddr + bvec->bv_offset,
					       csum, bvec->bv_len);
			btrfs_csum_final(csum, (char *)&csum);
6305
			kunmap_atomic(kaddr);
6306 6307 6308
			local_irq_restore(flags);

			flush_dcache_page(bvec->bv_page);
6309 6310
			if (csum != private) {
failed:
L
Li Zefan 已提交
6311
				printk(KERN_ERR "btrfs csum failed ino %llu off"
6312
				      " %llu csum %u private %u\n",
L
Li Zefan 已提交
6313 6314
				      (unsigned long long)btrfs_ino(inode),
				      (unsigned long long)start,
6315
				      csum, (unsigned)private);
6316 6317 6318 6319 6320 6321 6322 6323 6324
				err = -EIO;
			}
		}

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

	unlock_extent(&BTRFS_I(inode)->io_tree, dip->logical_offset,
6325
		      dip->logical_offset + dip->bytes - 1);
6326 6327 6328
	bio->bi_private = dip->private;

	kfree(dip);
6329 6330 6331 6332

	/* If we had a csum failure make sure to clear the uptodate flag */
	if (err)
		clear_bit(BIO_UPTODATE, &bio->bi_flags);
6333 6334 6335 6336 6337 6338 6339 6340 6341
	dio_end_io(bio, err);
}

static void btrfs_endio_direct_write(struct bio *bio, int err)
{
	struct btrfs_dio_private *dip = bio->bi_private;
	struct inode *inode = dip->inode;
	struct btrfs_root *root = BTRFS_I(inode)->root;
	struct btrfs_ordered_extent *ordered = NULL;
6342 6343
	u64 ordered_offset = dip->logical_offset;
	u64 ordered_bytes = dip->bytes;
6344 6345 6346 6347
	int ret;

	if (err)
		goto out_done;
6348 6349 6350
again:
	ret = btrfs_dec_test_first_ordered_pending(inode, &ordered,
						   &ordered_offset,
6351
						   ordered_bytes, !err);
6352
	if (!ret)
6353
		goto out_test;
6354

6355 6356 6357 6358
	ordered->work.func = finish_ordered_fn;
	ordered->work.flags = 0;
	btrfs_queue_worker(&root->fs_info->endio_write_workers,
			   &ordered->work);
6359 6360 6361 6362 6363 6364 6365 6366
out_test:
	/*
	 * our bio might span multiple ordered extents.  If we haven't
	 * completed the accounting for the whole dio, go back and try again
	 */
	if (ordered_offset < dip->logical_offset + dip->bytes) {
		ordered_bytes = dip->logical_offset + dip->bytes -
			ordered_offset;
6367
		ordered = NULL;
6368 6369
		goto again;
	}
6370 6371 6372 6373
out_done:
	bio->bi_private = dip->private;

	kfree(dip);
6374 6375 6376 6377

	/* If we had an error make sure to clear the uptodate flag */
	if (err)
		clear_bit(BIO_UPTODATE, &bio->bi_flags);
6378 6379 6380
	dio_end_io(bio, err);
}

6381 6382 6383 6384 6385 6386 6387
static int __btrfs_submit_bio_start_direct_io(struct inode *inode, int rw,
				    struct bio *bio, int mirror_num,
				    unsigned long bio_flags, u64 offset)
{
	int ret;
	struct btrfs_root *root = BTRFS_I(inode)->root;
	ret = btrfs_csum_one_bio(root, inode, bio, offset, 1);
6388
	BUG_ON(ret); /* -ENOMEM */
6389 6390 6391
	return 0;
}

M
Miao Xie 已提交
6392 6393 6394 6395 6396
static void btrfs_end_dio_bio(struct bio *bio, int err)
{
	struct btrfs_dio_private *dip = bio->bi_private;

	if (err) {
L
Li Zefan 已提交
6397
		printk(KERN_ERR "btrfs direct IO failed ino %llu rw %lu "
J
Jan Beulich 已提交
6398
		      "sector %#Lx len %u err no %d\n",
L
Li Zefan 已提交
6399
		      (unsigned long long)btrfs_ino(dip->inode), bio->bi_rw,
J
Jan Beulich 已提交
6400
		      (unsigned long long)bio->bi_sector, bio->bi_size, err);
M
Miao Xie 已提交
6401 6402 6403 6404 6405 6406 6407 6408 6409 6410 6411 6412 6413 6414 6415 6416 6417 6418 6419 6420 6421 6422 6423 6424 6425 6426 6427 6428 6429 6430 6431 6432
		dip->errors = 1;

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

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

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

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

static inline int __btrfs_submit_dio_bio(struct bio *bio, struct inode *inode,
					 int rw, u64 file_offset, int skip_sum,
6433
					 int async_submit)
M
Miao Xie 已提交
6434 6435 6436 6437 6438
{
	int write = rw & REQ_WRITE;
	struct btrfs_root *root = BTRFS_I(inode)->root;
	int ret;

6439 6440 6441
	if (async_submit)
		async_submit = !atomic_read(&BTRFS_I(inode)->sync_writers);

M
Miao Xie 已提交
6442
	bio_get(bio);
6443 6444 6445 6446 6447 6448

	if (!write) {
		ret = btrfs_bio_wq_end_io(root->fs_info, bio, 0);
		if (ret)
			goto err;
	}
M
Miao Xie 已提交
6449

6450 6451 6452 6453
	if (skip_sum)
		goto map;

	if (write && async_submit) {
M
Miao Xie 已提交
6454 6455 6456 6457 6458 6459
		ret = btrfs_wq_submit_bio(root->fs_info,
				   inode, rw, bio, 0, 0,
				   file_offset,
				   __btrfs_submit_bio_start_direct_io,
				   __btrfs_submit_bio_done);
		goto err;
6460 6461 6462 6463 6464 6465 6466 6467
	} else if (write) {
		/*
		 * If we aren't doing async submit, calculate the csum of the
		 * bio now.
		 */
		ret = btrfs_csum_one_bio(root, inode, bio, file_offset, 1);
		if (ret)
			goto err;
6468
	} else if (!skip_sum) {
6469
		ret = btrfs_lookup_bio_sums_dio(root, inode, bio, file_offset);
6470 6471 6472
		if (ret)
			goto err;
	}
M
Miao Xie 已提交
6473

6474 6475
map:
	ret = btrfs_map_bio(root, rw, bio, 0, async_submit);
M
Miao Xie 已提交
6476 6477 6478 6479 6480 6481 6482 6483 6484 6485 6486 6487 6488 6489 6490 6491 6492 6493 6494
err:
	bio_put(bio);
	return ret;
}

static int btrfs_submit_direct_hook(int rw, struct btrfs_dio_private *dip,
				    int skip_sum)
{
	struct inode *inode = dip->inode;
	struct btrfs_root *root = BTRFS_I(inode)->root;
	struct bio *bio;
	struct bio *orig_bio = dip->orig_bio;
	struct bio_vec *bvec = orig_bio->bi_io_vec;
	u64 start_sector = orig_bio->bi_sector;
	u64 file_offset = dip->logical_offset;
	u64 submit_len = 0;
	u64 map_length;
	int nr_pages = 0;
	int ret = 0;
6495
	int async_submit = 0;
M
Miao Xie 已提交
6496 6497

	map_length = orig_bio->bi_size;
6498
	ret = btrfs_map_block(root->fs_info, READ, start_sector << 9,
M
Miao Xie 已提交
6499 6500
			      &map_length, NULL, 0);
	if (ret) {
6501
		bio_put(orig_bio);
M
Miao Xie 已提交
6502 6503 6504
		return -EIO;
	}

6505 6506 6507 6508 6509
	if (map_length >= orig_bio->bi_size) {
		bio = orig_bio;
		goto submit;
	}

6510
	async_submit = 1;
6511 6512 6513 6514 6515 6516 6517
	bio = btrfs_dio_bio_alloc(orig_bio->bi_bdev, start_sector, GFP_NOFS);
	if (!bio)
		return -ENOMEM;
	bio->bi_private = dip;
	bio->bi_end_io = btrfs_end_dio_bio;
	atomic_inc(&dip->pending_bios);

M
Miao Xie 已提交
6518 6519 6520 6521 6522 6523 6524 6525 6526 6527 6528 6529 6530
	while (bvec <= (orig_bio->bi_io_vec + orig_bio->bi_vcnt - 1)) {
		if (unlikely(map_length < submit_len + bvec->bv_len ||
		    bio_add_page(bio, bvec->bv_page, bvec->bv_len,
				 bvec->bv_offset) < bvec->bv_len)) {
			/*
			 * inc the count before we submit the bio so
			 * we know the end IO handler won't happen before
			 * we inc the count. Otherwise, the dip might get freed
			 * before we're done setting it up
			 */
			atomic_inc(&dip->pending_bios);
			ret = __btrfs_submit_dio_bio(bio, inode, rw,
						     file_offset, skip_sum,
6531
						     async_submit);
M
Miao Xie 已提交
6532 6533 6534 6535 6536 6537 6538 6539 6540 6541 6542 6543 6544 6545 6546 6547 6548 6549 6550 6551
			if (ret) {
				bio_put(bio);
				atomic_dec(&dip->pending_bios);
				goto out_err;
			}

			start_sector += submit_len >> 9;
			file_offset += submit_len;

			submit_len = 0;
			nr_pages = 0;

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

			map_length = orig_bio->bi_size;
6552 6553
			ret = btrfs_map_block(root->fs_info, READ,
					      start_sector << 9,
M
Miao Xie 已提交
6554 6555 6556 6557 6558 6559 6560 6561 6562 6563 6564 6565
					      &map_length, NULL, 0);
			if (ret) {
				bio_put(bio);
				goto out_err;
			}
		} else {
			submit_len += bvec->bv_len;
			nr_pages ++;
			bvec++;
		}
	}

6566
submit:
M
Miao Xie 已提交
6567
	ret = __btrfs_submit_dio_bio(bio, inode, rw, file_offset, skip_sum,
6568
				     async_submit);
M
Miao Xie 已提交
6569 6570 6571 6572 6573 6574 6575 6576 6577 6578 6579 6580 6581 6582 6583 6584 6585 6586
	if (!ret)
		return 0;

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

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

6587 6588 6589 6590 6591 6592 6593
static void btrfs_submit_direct(int rw, struct bio *bio, struct inode *inode,
				loff_t file_offset)
{
	struct btrfs_root *root = BTRFS_I(inode)->root;
	struct btrfs_dio_private *dip;
	struct bio_vec *bvec = bio->bi_io_vec;
	int skip_sum;
6594
	int write = rw & REQ_WRITE;
6595 6596 6597 6598 6599 6600 6601 6602 6603 6604 6605 6606 6607 6608 6609 6610 6611 6612 6613 6614
	int ret = 0;

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

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

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

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

6615
	dip->disk_bytenr = (u64)bio->bi_sector << 9;
6616
	bio->bi_private = dip;
M
Miao Xie 已提交
6617 6618 6619
	dip->errors = 0;
	dip->orig_bio = bio;
	atomic_set(&dip->pending_bios, 0);
6620 6621 6622 6623 6624 6625

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

M
Miao Xie 已提交
6626 6627
	ret = btrfs_submit_direct_hook(rw, dip, skip_sum);
	if (!ret)
6628
		return;
6629 6630 6631 6632 6633 6634 6635
free_ordered:
	/*
	 * If this is a write, we need to clean up the reserved space and kill
	 * the ordered extent.
	 */
	if (write) {
		struct btrfs_ordered_extent *ordered;
6636
		ordered = btrfs_lookup_ordered_extent(inode, file_offset);
6637 6638 6639 6640 6641 6642 6643 6644 6645 6646
		if (!test_bit(BTRFS_ORDERED_PREALLOC, &ordered->flags) &&
		    !test_bit(BTRFS_ORDERED_NOCOW, &ordered->flags))
			btrfs_free_reserved_extent(root, ordered->start,
						   ordered->disk_len);
		btrfs_put_ordered_extent(ordered);
		btrfs_put_ordered_extent(ordered);
	}
	bio_endio(bio, ret);
}

C
Chris Mason 已提交
6647 6648 6649 6650 6651
static ssize_t check_direct_IO(struct btrfs_root *root, int rw, struct kiocb *iocb,
			const struct iovec *iov, loff_t offset,
			unsigned long nr_segs)
{
	int seg;
6652
	int i;
C
Chris Mason 已提交
6653 6654 6655 6656 6657 6658 6659 6660 6661 6662 6663 6664 6665 6666
	size_t size;
	unsigned long addr;
	unsigned blocksize_mask = root->sectorsize - 1;
	ssize_t retval = -EINVAL;
	loff_t end = offset;

	if (offset & blocksize_mask)
		goto out;

	/* Check the memory alignment.  Blocks cannot straddle pages */
	for (seg = 0; seg < nr_segs; seg++) {
		addr = (unsigned long)iov[seg].iov_base;
		size = iov[seg].iov_len;
		end += size;
6667
		if ((addr & blocksize_mask) || (size & blocksize_mask))
C
Chris Mason 已提交
6668
			goto out;
6669 6670 6671 6672 6673 6674 6675 6676 6677 6678 6679 6680 6681 6682

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

		/*
		 * Check to make sure we don't have duplicate iov_base's in this
		 * iovec, if so return EINVAL, otherwise we'll get csum errors
		 * when reading back.
		 */
		for (i = seg + 1; i < nr_segs; i++) {
			if (iov[seg].iov_base == iov[i].iov_base)
				goto out;
		}
C
Chris Mason 已提交
6683 6684 6685 6686 6687
	}
	retval = 0;
out:
	return retval;
}
6688

6689 6690 6691 6692
static ssize_t btrfs_direct_IO(int rw, struct kiocb *iocb,
			const struct iovec *iov, loff_t offset,
			unsigned long nr_segs)
{
6693 6694 6695
	struct file *file = iocb->ki_filp;
	struct inode *inode = file->f_mapping->host;

C
Chris Mason 已提交
6696
	if (check_direct_IO(BTRFS_I(inode)->root, rw, iocb, iov,
6697
			    offset, nr_segs))
C
Chris Mason 已提交
6698
		return 0;
6699

6700
	return __blockdev_direct_IO(rw, iocb, inode,
C
Chris Mason 已提交
6701 6702 6703
		   BTRFS_I(inode)->root->fs_info->fs_devices->latest_bdev,
		   iov, offset, nr_segs, btrfs_get_blocks_direct, NULL,
		   btrfs_submit_direct, 0);
6704 6705
}

T
Tsutomu Itoh 已提交
6706 6707
#define BTRFS_FIEMAP_FLAGS	(FIEMAP_FLAG_SYNC)

Y
Yehuda Sadeh 已提交
6708 6709 6710
static int btrfs_fiemap(struct inode *inode, struct fiemap_extent_info *fieinfo,
		__u64 start, __u64 len)
{
T
Tsutomu Itoh 已提交
6711 6712 6713 6714 6715 6716
	int	ret;

	ret = fiemap_check_flags(fieinfo, BTRFS_FIEMAP_FLAGS);
	if (ret)
		return ret;

6717
	return extent_fiemap(inode, fieinfo, start, len, btrfs_get_extent_fiemap);
Y
Yehuda Sadeh 已提交
6718 6719
}

6720
int btrfs_readpage(struct file *file, struct page *page)
C
Chris Mason 已提交
6721
{
6722 6723
	struct extent_io_tree *tree;
	tree = &BTRFS_I(page->mapping->host)->io_tree;
6724
	return extent_read_full_page(tree, page, btrfs_get_extent, 0);
C
Chris Mason 已提交
6725
}
6726

6727
static int btrfs_writepage(struct page *page, struct writeback_control *wbc)
C
Chris Mason 已提交
6728
{
6729
	struct extent_io_tree *tree;
6730 6731 6732 6733 6734 6735 6736


	if (current->flags & PF_MEMALLOC) {
		redirty_page_for_writepage(wbc, page);
		unlock_page(page);
		return 0;
	}
6737
	tree = &BTRFS_I(page->mapping->host)->io_tree;
6738
	return extent_write_full_page(tree, page, btrfs_get_extent, wbc);
C
Chris Mason 已提交
6739 6740
}

6741 6742
int btrfs_writepages(struct address_space *mapping,
		     struct writeback_control *wbc)
C
Chris Mason 已提交
6743
{
6744
	struct extent_io_tree *tree;
6745

6746
	tree = &BTRFS_I(mapping->host)->io_tree;
C
Chris Mason 已提交
6747 6748 6749
	return extent_writepages(tree, mapping, btrfs_get_extent, wbc);
}

C
Chris Mason 已提交
6750 6751 6752 6753
static int
btrfs_readpages(struct file *file, struct address_space *mapping,
		struct list_head *pages, unsigned nr_pages)
{
6754 6755
	struct extent_io_tree *tree;
	tree = &BTRFS_I(mapping->host)->io_tree;
C
Chris Mason 已提交
6756 6757 6758
	return extent_readpages(tree, mapping, pages, nr_pages,
				btrfs_get_extent);
}
6759
static int __btrfs_releasepage(struct page *page, gfp_t gfp_flags)
C
Chris Mason 已提交
6760
{
6761 6762
	struct extent_io_tree *tree;
	struct extent_map_tree *map;
6763
	int ret;
6764

6765 6766
	tree = &BTRFS_I(page->mapping->host)->io_tree;
	map = &BTRFS_I(page->mapping->host)->extent_tree;
6767
	ret = try_release_extent_mapping(map, tree, page, gfp_flags);
6768 6769 6770 6771
	if (ret == 1) {
		ClearPagePrivate(page);
		set_page_private(page, 0);
		page_cache_release(page);
C
Chris Mason 已提交
6772
	}
6773
	return ret;
C
Chris Mason 已提交
6774 6775
}

6776 6777
static int btrfs_releasepage(struct page *page, gfp_t gfp_flags)
{
6778 6779
	if (PageWriteback(page) || PageDirty(page))
		return 0;
6780
	return __btrfs_releasepage(page, gfp_flags & GFP_NOFS);
6781 6782
}

6783
static void btrfs_invalidatepage(struct page *page, unsigned long offset)
C
Chris Mason 已提交
6784
{
6785
	struct inode *inode = page->mapping->host;
6786
	struct extent_io_tree *tree;
6787
	struct btrfs_ordered_extent *ordered;
6788
	struct extent_state *cached_state = NULL;
6789 6790
	u64 page_start = page_offset(page);
	u64 page_end = page_start + PAGE_CACHE_SIZE - 1;
C
Chris Mason 已提交
6791

6792 6793 6794 6795 6796 6797 6798
	/*
	 * 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
	 */
6799
	wait_on_page_writeback(page);
6800

6801
	tree = &BTRFS_I(inode)->io_tree;
6802 6803 6804 6805
	if (offset) {
		btrfs_releasepage(page, GFP_NOFS);
		return;
	}
6806
	lock_extent_bits(tree, page_start, page_end, 0, &cached_state);
M
Miao Xie 已提交
6807
	ordered = btrfs_lookup_ordered_extent(inode, page_offset(page));
6808
	if (ordered) {
6809 6810 6811 6812
		/*
		 * IO on this page will never be started, so we need
		 * to account for any ordered extents now
		 */
6813 6814
		clear_extent_bit(tree, page_start, page_end,
				 EXTENT_DIRTY | EXTENT_DELALLOC |
6815 6816
				 EXTENT_LOCKED | EXTENT_DO_ACCOUNTING |
				 EXTENT_DEFRAG, 1, 0, &cached_state, GFP_NOFS);
6817 6818 6819 6820
		/*
		 * whoever cleared the private bit is responsible
		 * for the finish_ordered_io
		 */
6821 6822 6823 6824
		if (TestClearPagePrivate2(page) &&
		    btrfs_dec_test_ordered_pending(inode, &ordered, page_start,
						   PAGE_CACHE_SIZE, 1)) {
			btrfs_finish_ordered_io(ordered);
6825
		}
6826
		btrfs_put_ordered_extent(ordered);
6827
		cached_state = NULL;
6828
		lock_extent_bits(tree, page_start, page_end, 0, &cached_state);
6829 6830
	}
	clear_extent_bit(tree, page_start, page_end,
6831
		 EXTENT_LOCKED | EXTENT_DIRTY | EXTENT_DELALLOC |
6832 6833
		 EXTENT_DO_ACCOUNTING | EXTENT_DEFRAG, 1, 1,
		 &cached_state, GFP_NOFS);
6834 6835
	__btrfs_releasepage(page, GFP_NOFS);

C
Chris Mason 已提交
6836
	ClearPageChecked(page);
6837 6838 6839 6840 6841
	if (PagePrivate(page)) {
		ClearPagePrivate(page);
		set_page_private(page, 0);
		page_cache_release(page);
	}
C
Chris Mason 已提交
6842 6843
}

C
Chris Mason 已提交
6844 6845 6846 6847 6848 6849 6850 6851 6852 6853 6854 6855 6856 6857 6858
/*
 * 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.
 */
6859
int btrfs_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf)
C
Chris Mason 已提交
6860
{
6861
	struct page *page = vmf->page;
6862
	struct inode *inode = fdentry(vma->vm_file)->d_inode;
6863
	struct btrfs_root *root = BTRFS_I(inode)->root;
6864 6865
	struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
	struct btrfs_ordered_extent *ordered;
6866
	struct extent_state *cached_state = NULL;
6867 6868
	char *kaddr;
	unsigned long zero_start;
C
Chris Mason 已提交
6869
	loff_t size;
6870
	int ret;
6871
	int reserved = 0;
6872
	u64 page_start;
6873
	u64 page_end;
C
Chris Mason 已提交
6874

6875
	sb_start_pagefault(inode->i_sb);
6876
	ret  = btrfs_delalloc_reserve_space(inode, PAGE_CACHE_SIZE);
6877
	if (!ret) {
6878
		ret = file_update_time(vma->vm_file);
6879 6880
		reserved = 1;
	}
6881 6882 6883 6884 6885
	if (ret) {
		if (ret == -ENOMEM)
			ret = VM_FAULT_OOM;
		else /* -ENOSPC, -EIO, etc */
			ret = VM_FAULT_SIGBUS;
6886 6887 6888
		if (reserved)
			goto out;
		goto out_noreserve;
6889
	}
6890

6891
	ret = VM_FAULT_NOPAGE; /* make the VM retry the fault */
6892
again:
C
Chris Mason 已提交
6893 6894
	lock_page(page);
	size = i_size_read(inode);
6895 6896
	page_start = page_offset(page);
	page_end = page_start + PAGE_CACHE_SIZE - 1;
6897

C
Chris Mason 已提交
6898
	if ((page->mapping != inode->i_mapping) ||
6899
	    (page_start >= size)) {
C
Chris Mason 已提交
6900 6901 6902
		/* page got truncated out from underneath us */
		goto out_unlock;
	}
6903 6904
	wait_on_page_writeback(page);

6905
	lock_extent_bits(io_tree, page_start, page_end, 0, &cached_state);
6906 6907
	set_page_extent_mapped(page);

6908 6909 6910 6911
	/*
	 * we can't set the delalloc bits if there are pending ordered
	 * extents.  Drop our locks and wait for them to finish
	 */
6912 6913
	ordered = btrfs_lookup_ordered_extent(inode, page_start);
	if (ordered) {
6914 6915
		unlock_extent_cached(io_tree, page_start, page_end,
				     &cached_state, GFP_NOFS);
6916
		unlock_page(page);
6917
		btrfs_start_ordered_extent(inode, ordered, 1);
6918 6919 6920 6921
		btrfs_put_ordered_extent(ordered);
		goto again;
	}

J
Josef Bacik 已提交
6922 6923 6924 6925 6926 6927 6928
	/*
	 * XXX - page_mkwrite gets called every time the page is dirtied, even
	 * if it was already dirty, so for space accounting reasons we need to
	 * clear any delalloc bits for the range we are fixing to save.  There
	 * is probably a better way to do this, but for now keep consistent with
	 * prepare_pages in the normal write path.
	 */
6929
	clear_extent_bit(&BTRFS_I(inode)->io_tree, page_start, page_end,
6930 6931
			  EXTENT_DIRTY | EXTENT_DELALLOC |
			  EXTENT_DO_ACCOUNTING | EXTENT_DEFRAG,
6932
			  0, 0, &cached_state, GFP_NOFS);
J
Josef Bacik 已提交
6933

6934 6935
	ret = btrfs_set_extent_delalloc(inode, page_start, page_end,
					&cached_state);
J
Josef Bacik 已提交
6936
	if (ret) {
6937 6938
		unlock_extent_cached(io_tree, page_start, page_end,
				     &cached_state, GFP_NOFS);
J
Josef Bacik 已提交
6939 6940 6941
		ret = VM_FAULT_SIGBUS;
		goto out_unlock;
	}
6942
	ret = 0;
C
Chris Mason 已提交
6943 6944

	/* page is wholly or partially inside EOF */
6945
	if (page_start + PAGE_CACHE_SIZE > size)
6946
		zero_start = size & ~PAGE_CACHE_MASK;
C
Chris Mason 已提交
6947
	else
6948
		zero_start = PAGE_CACHE_SIZE;
C
Chris Mason 已提交
6949

6950 6951 6952 6953 6954 6955
	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);
	}
6956
	ClearPageChecked(page);
6957
	set_page_dirty(page);
6958
	SetPageUptodate(page);
6959

6960 6961
	BTRFS_I(inode)->last_trans = root->fs_info->generation;
	BTRFS_I(inode)->last_sub_trans = BTRFS_I(inode)->root->log_transid;
6962
	BTRFS_I(inode)->last_log_commit = BTRFS_I(inode)->root->last_log_commit;
6963

6964
	unlock_extent_cached(io_tree, page_start, page_end, &cached_state, GFP_NOFS);
C
Chris Mason 已提交
6965 6966

out_unlock:
6967 6968
	if (!ret) {
		sb_end_pagefault(inode->i_sb);
6969
		return VM_FAULT_LOCKED;
6970
	}
C
Chris Mason 已提交
6971
	unlock_page(page);
6972
out:
6973
	btrfs_delalloc_release_space(inode, PAGE_CACHE_SIZE);
6974
out_noreserve:
6975
	sb_end_pagefault(inode->i_sb);
C
Chris Mason 已提交
6976 6977 6978
	return ret;
}

6979
static int btrfs_truncate(struct inode *inode)
C
Chris Mason 已提交
6980 6981
{
	struct btrfs_root *root = BTRFS_I(inode)->root;
6982
	struct btrfs_block_rsv *rsv;
C
Chris Mason 已提交
6983
	int ret;
6984
	int err = 0;
C
Chris Mason 已提交
6985
	struct btrfs_trans_handle *trans;
6986
	u64 mask = root->sectorsize - 1;
6987
	u64 min_size = btrfs_calc_trunc_metadata_size(root, 1);
C
Chris Mason 已提交
6988

J
Josef Bacik 已提交
6989
	ret = btrfs_truncate_page(inode, inode->i_size, 0, 0);
6990
	if (ret)
6991
		return ret;
6992

C
Chris Mason 已提交
6993
	btrfs_wait_ordered_range(inode, inode->i_size & (~mask), (u64)-1);
6994
	btrfs_ordered_update_i_size(inode, inode->i_size, NULL);
C
Chris Mason 已提交
6995

6996 6997 6998 6999 7000 7001 7002 7003 7004 7005 7006 7007 7008 7009 7010 7011 7012 7013 7014 7015 7016 7017 7018 7019 7020 7021 7022 7023 7024 7025 7026 7027 7028 7029 7030 7031
	/*
	 * Yes ladies and gentelment, this is indeed ugly.  The fact is we have
	 * 3 things going on here
	 *
	 * 1) We need to reserve space for our orphan item and the space to
	 * delete our orphan item.  Lord knows we don't want to have a dangling
	 * orphan item because we didn't reserve space to remove it.
	 *
	 * 2) We need to reserve space to update our inode.
	 *
	 * 3) We need to have something to cache all the space that is going to
	 * be free'd up by the truncate operation, but also have some slack
	 * space reserved in case it uses space during the truncate (thank you
	 * very much snapshotting).
	 *
	 * And we need these to all be seperate.  The fact is we can use alot of
	 * space doing the truncate, and we have no earthly idea how much space
	 * we will use, so we need the truncate reservation to be seperate so it
	 * doesn't end up using space reserved for updating the inode or
	 * removing the orphan item.  We also need to be able to stop the
	 * transaction and start a new one, which means we need to be able to
	 * update the inode several times, and we have no idea of knowing how
	 * many times that will be, so we can't just reserve 1 item for the
	 * entirety of the opration, so that has to be done seperately as well.
	 * Then there is the orphan item, which does indeed need to be held on
	 * to for the whole operation, and we need nobody to touch this reserved
	 * space except the orphan code.
	 *
	 * So that leaves us with
	 *
	 * 1) root->orphan_block_rsv - for the orphan deletion.
	 * 2) rsv - for the truncate reservation, which we will steal from the
	 * transaction reservation.
	 * 3) fs_info->trans_block_rsv - this will have 1 items worth left for
	 * updating the inode.
	 */
7032
	rsv = btrfs_alloc_block_rsv(root, BTRFS_BLOCK_RSV_TEMP);
7033 7034
	if (!rsv)
		return -ENOMEM;
J
Josef Bacik 已提交
7035
	rsv->size = min_size;
7036
	rsv->failfast = 1;
7037

7038
	/*
7039
	 * 1 for the truncate slack space
7040 7041
	 * 1 for updating the inode.
	 */
7042
	trans = btrfs_start_transaction(root, 2);
7043 7044 7045 7046
	if (IS_ERR(trans)) {
		err = PTR_ERR(trans);
		goto out;
	}
7047

7048 7049 7050
	/* Migrate the slack space for the truncate to our reserve */
	ret = btrfs_block_rsv_migrate(&root->fs_info->trans_block_rsv, rsv,
				      min_size);
7051
	BUG_ON(ret);
7052

7053 7054 7055 7056 7057 7058 7059 7060 7061 7062 7063 7064 7065 7066 7067 7068 7069
	/*
	 * 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.
	 */
7070 7071
	if (inode->i_size == 0 && test_bit(BTRFS_INODE_ORDERED_DATA_CLOSE,
					   &BTRFS_I(inode)->runtime_flags))
7072 7073
		btrfs_add_ordered_operation(trans, root, inode);

J
Josef Bacik 已提交
7074 7075 7076 7077 7078 7079 7080 7081
	/*
	 * So if we truncate and then write and fsync we normally would just
	 * write the extents that changed, which is a problem if we need to
	 * first truncate that entire inode.  So set this flag so we write out
	 * all of the extents in the inode to the sync log so we're completely
	 * safe.
	 */
	set_bit(BTRFS_INODE_NEEDS_FULL_SYNC, &BTRFS_I(inode)->runtime_flags);
7082
	trans->block_rsv = rsv;
7083

7084 7085 7086 7087
	while (1) {
		ret = btrfs_truncate_inode_items(trans, root, inode,
						 inode->i_size,
						 BTRFS_EXTENT_DATA_KEY);
7088
		if (ret != -ENOSPC) {
7089
			err = ret;
7090
			break;
7091
		}
C
Chris Mason 已提交
7092

7093
		trans->block_rsv = &root->fs_info->trans_block_rsv;
7094
		ret = btrfs_update_inode(trans, root, inode);
7095 7096 7097 7098
		if (ret) {
			err = ret;
			break;
		}
7099

7100
		btrfs_end_transaction(trans, root);
7101
		btrfs_btree_balance_dirty(root);
7102 7103 7104 7105 7106 7107 7108 7109 7110 7111 7112 7113

		trans = btrfs_start_transaction(root, 2);
		if (IS_ERR(trans)) {
			ret = err = PTR_ERR(trans);
			trans = NULL;
			break;
		}

		ret = btrfs_block_rsv_migrate(&root->fs_info->trans_block_rsv,
					      rsv, min_size);
		BUG_ON(ret);	/* shouldn't happen */
		trans->block_rsv = rsv;
7114 7115 7116
	}

	if (ret == 0 && inode->i_nlink > 0) {
7117
		trans->block_rsv = root->orphan_block_rsv;
7118
		ret = btrfs_orphan_del(trans, inode);
7119 7120
		if (ret)
			err = ret;
7121 7122
	}

7123 7124 7125 7126 7127
	if (trans) {
		trans->block_rsv = &root->fs_info->trans_block_rsv;
		ret = btrfs_update_inode(trans, root, inode);
		if (ret && !err)
			err = ret;
7128

7129
		ret = btrfs_end_transaction(trans, root);
7130
		btrfs_btree_balance_dirty(root);
7131
	}
7132 7133 7134 7135

out:
	btrfs_free_block_rsv(root, rsv);

7136 7137
	if (ret && !err)
		err = ret;
7138

7139
	return err;
C
Chris Mason 已提交
7140 7141
}

C
Chris Mason 已提交
7142 7143 7144
/*
 * create a new subvolume directory/inode (helper for the ioctl).
 */
7145
int btrfs_create_subvol_root(struct btrfs_trans_handle *trans,
7146
			     struct btrfs_root *new_root, u64 new_dirid)
C
Chris Mason 已提交
7147 7148
{
	struct inode *inode;
7149
	int err;
7150
	u64 index = 0;
C
Chris Mason 已提交
7151

7152 7153 7154 7155
	inode = btrfs_new_inode(trans, new_root, NULL, "..", 2,
				new_dirid, new_dirid,
				S_IFDIR | (~current_umask() & S_IRWXUGO),
				&index);
7156
	if (IS_ERR(inode))
C
Christoph Hellwig 已提交
7157
		return PTR_ERR(inode);
C
Chris Mason 已提交
7158 7159 7160
	inode->i_op = &btrfs_dir_inode_operations;
	inode->i_fop = &btrfs_dir_file_operations;

M
Miklos Szeredi 已提交
7161
	set_nlink(inode, 1);
7162
	btrfs_i_size_write(inode, 0);
7163

7164
	err = btrfs_update_inode(trans, new_root, inode);
7165

7166
	iput(inode);
7167
	return err;
C
Chris Mason 已提交
7168 7169 7170 7171 7172
}

struct inode *btrfs_alloc_inode(struct super_block *sb)
{
	struct btrfs_inode *ei;
Y
Yan, Zheng 已提交
7173
	struct inode *inode;
C
Chris Mason 已提交
7174 7175 7176 7177

	ei = kmem_cache_alloc(btrfs_inode_cachep, GFP_NOFS);
	if (!ei)
		return NULL;
Y
Yan, Zheng 已提交
7178 7179 7180

	ei->root = NULL;
	ei->generation = 0;
7181
	ei->last_trans = 0;
7182
	ei->last_sub_trans = 0;
7183
	ei->logged_trans = 0;
Y
Yan, Zheng 已提交
7184 7185 7186
	ei->delalloc_bytes = 0;
	ei->disk_i_size = 0;
	ei->flags = 0;
7187
	ei->csum_bytes = 0;
Y
Yan, Zheng 已提交
7188 7189
	ei->index_cnt = (u64)-1;
	ei->last_unlink_trans = 0;
7190
	ei->last_log_commit = 0;
Y
Yan, Zheng 已提交
7191

7192 7193 7194
	spin_lock_init(&ei->lock);
	ei->outstanding_extents = 0;
	ei->reserved_extents = 0;
Y
Yan, Zheng 已提交
7195

7196
	ei->runtime_flags = 0;
7197
	ei->force_compress = BTRFS_COMPRESS_NONE;
Y
Yan, Zheng 已提交
7198

7199 7200
	ei->delayed_node = NULL;

Y
Yan, Zheng 已提交
7201
	inode = &ei->vfs_inode;
7202
	extent_map_tree_init(&ei->extent_tree);
7203 7204
	extent_io_tree_init(&ei->io_tree, &inode->i_data);
	extent_io_tree_init(&ei->io_failure_tree, &inode->i_data);
7205 7206
	ei->io_tree.track_uptodate = 1;
	ei->io_failure_tree.track_uptodate = 1;
7207
	atomic_set(&ei->sync_writers, 0);
Y
Yan, Zheng 已提交
7208
	mutex_init(&ei->log_mutex);
7209
	mutex_init(&ei->delalloc_mutex);
7210
	btrfs_ordered_inode_tree_init(&ei->ordered_tree);
Y
Yan, Zheng 已提交
7211
	INIT_LIST_HEAD(&ei->delalloc_inodes);
7212
	INIT_LIST_HEAD(&ei->ordered_operations);
Y
Yan, Zheng 已提交
7213 7214 7215
	RB_CLEAR_NODE(&ei->rb_node);

	return inode;
C
Chris Mason 已提交
7216 7217
}

N
Nick Piggin 已提交
7218 7219 7220 7221 7222 7223
static void btrfs_i_callback(struct rcu_head *head)
{
	struct inode *inode = container_of(head, struct inode, i_rcu);
	kmem_cache_free(btrfs_inode_cachep, BTRFS_I(inode));
}

C
Chris Mason 已提交
7224 7225
void btrfs_destroy_inode(struct inode *inode)
{
7226
	struct btrfs_ordered_extent *ordered;
7227 7228
	struct btrfs_root *root = BTRFS_I(inode)->root;

A
Al Viro 已提交
7229
	WARN_ON(!hlist_empty(&inode->i_dentry));
C
Chris Mason 已提交
7230
	WARN_ON(inode->i_data.nrpages);
7231 7232
	WARN_ON(BTRFS_I(inode)->outstanding_extents);
	WARN_ON(BTRFS_I(inode)->reserved_extents);
7233 7234
	WARN_ON(BTRFS_I(inode)->delalloc_bytes);
	WARN_ON(BTRFS_I(inode)->csum_bytes);
C
Chris Mason 已提交
7235

7236 7237 7238 7239 7240 7241 7242 7243
	/*
	 * This can happen where we create an inode, but somebody else also
	 * created the same inode and we need to destroy the one we already
	 * created.
	 */
	if (!root)
		goto free;

7244 7245 7246 7247 7248 7249 7250 7251 7252 7253 7254
	/*
	 * 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);
	}

7255 7256
	if (test_bit(BTRFS_INODE_HAS_ORPHAN_ITEM,
		     &BTRFS_I(inode)->runtime_flags)) {
L
Li Zefan 已提交
7257 7258
		printk(KERN_INFO "BTRFS: inode %llu still on the orphan list\n",
		       (unsigned long long)btrfs_ino(inode));
7259
		atomic_dec(&root->orphan_inodes);
7260 7261
	}

C
Chris Mason 已提交
7262
	while (1) {
7263 7264 7265 7266
		ordered = btrfs_lookup_first_ordered_extent(inode, (u64)-1);
		if (!ordered)
			break;
		else {
C
Chris Mason 已提交
7267 7268 7269 7270
			printk(KERN_ERR "btrfs found ordered "
			       "extent %llu %llu on inode cleanup\n",
			       (unsigned long long)ordered->file_offset,
			       (unsigned long long)ordered->len);
7271 7272 7273 7274 7275
			btrfs_remove_ordered_extent(inode, ordered);
			btrfs_put_ordered_extent(ordered);
			btrfs_put_ordered_extent(ordered);
		}
	}
7276
	inode_tree_del(inode);
7277
	btrfs_drop_extent_cache(inode, 0, (u64)-1, 0);
7278
free:
7279
	btrfs_remove_delayed_node(inode);
N
Nick Piggin 已提交
7280
	call_rcu(&inode->i_rcu, btrfs_i_callback);
C
Chris Mason 已提交
7281 7282
}

7283
int btrfs_drop_inode(struct inode *inode)
7284 7285
{
	struct btrfs_root *root = BTRFS_I(inode)->root;
7286

7287
	if (btrfs_root_refs(&root->root_item) == 0 &&
7288
	    !btrfs_is_free_space_inode(inode))
7289
		return 1;
7290
	else
7291
		return generic_drop_inode(inode);
7292 7293
}

7294
static void init_once(void *foo)
C
Chris Mason 已提交
7295 7296 7297 7298 7299 7300 7301 7302
{
	struct btrfs_inode *ei = (struct btrfs_inode *) foo;

	inode_init_once(&ei->vfs_inode);
}

void btrfs_destroy_cachep(void)
{
7303 7304 7305 7306 7307
	/*
	 * Make sure all delayed rcu free inodes are flushed before we
	 * destroy cache.
	 */
	rcu_barrier();
C
Chris Mason 已提交
7308 7309 7310 7311 7312 7313 7314 7315
	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);
7316 7317
	if (btrfs_free_space_cachep)
		kmem_cache_destroy(btrfs_free_space_cachep);
7318 7319
	if (btrfs_delalloc_work_cachep)
		kmem_cache_destroy(btrfs_delalloc_work_cachep);
C
Chris Mason 已提交
7320 7321 7322 7323
}

int btrfs_init_cachep(void)
{
D
David Sterba 已提交
7324
	btrfs_inode_cachep = kmem_cache_create("btrfs_inode",
7325 7326
			sizeof(struct btrfs_inode), 0,
			SLAB_RECLAIM_ACCOUNT | SLAB_MEM_SPREAD, init_once);
C
Chris Mason 已提交
7327 7328
	if (!btrfs_inode_cachep)
		goto fail;
7329

D
David Sterba 已提交
7330
	btrfs_trans_handle_cachep = kmem_cache_create("btrfs_trans_handle",
7331 7332
			sizeof(struct btrfs_trans_handle), 0,
			SLAB_RECLAIM_ACCOUNT | SLAB_MEM_SPREAD, NULL);
C
Chris Mason 已提交
7333 7334
	if (!btrfs_trans_handle_cachep)
		goto fail;
7335

D
David Sterba 已提交
7336
	btrfs_transaction_cachep = kmem_cache_create("btrfs_transaction",
7337 7338
			sizeof(struct btrfs_transaction), 0,
			SLAB_RECLAIM_ACCOUNT | SLAB_MEM_SPREAD, NULL);
C
Chris Mason 已提交
7339 7340
	if (!btrfs_transaction_cachep)
		goto fail;
7341

D
David Sterba 已提交
7342
	btrfs_path_cachep = kmem_cache_create("btrfs_path",
7343 7344
			sizeof(struct btrfs_path), 0,
			SLAB_RECLAIM_ACCOUNT | SLAB_MEM_SPREAD, NULL);
C
Chris Mason 已提交
7345 7346
	if (!btrfs_path_cachep)
		goto fail;
7347

D
David Sterba 已提交
7348
	btrfs_free_space_cachep = kmem_cache_create("btrfs_free_space",
7349 7350 7351 7352 7353
			sizeof(struct btrfs_free_space), 0,
			SLAB_RECLAIM_ACCOUNT | SLAB_MEM_SPREAD, NULL);
	if (!btrfs_free_space_cachep)
		goto fail;

7354 7355 7356 7357 7358 7359 7360
	btrfs_delalloc_work_cachep = kmem_cache_create("btrfs_delalloc_work",
			sizeof(struct btrfs_delalloc_work), 0,
			SLAB_RECLAIM_ACCOUNT | SLAB_MEM_SPREAD,
			NULL);
	if (!btrfs_delalloc_work_cachep)
		goto fail;

C
Chris Mason 已提交
7361 7362 7363 7364 7365 7366 7367 7368 7369
	return 0;
fail:
	btrfs_destroy_cachep();
	return -ENOMEM;
}

static int btrfs_getattr(struct vfsmount *mnt,
			 struct dentry *dentry, struct kstat *stat)
{
7370
	u64 delalloc_bytes;
C
Chris Mason 已提交
7371
	struct inode *inode = dentry->d_inode;
D
David Sterba 已提交
7372 7373
	u32 blocksize = inode->i_sb->s_blocksize;

C
Chris Mason 已提交
7374
	generic_fillattr(inode, stat);
7375
	stat->dev = BTRFS_I(inode)->root->anon_dev;
C
Chris Mason 已提交
7376
	stat->blksize = PAGE_CACHE_SIZE;
7377 7378 7379 7380

	spin_lock(&BTRFS_I(inode)->lock);
	delalloc_bytes = BTRFS_I(inode)->delalloc_bytes;
	spin_unlock(&BTRFS_I(inode)->lock);
D
David Sterba 已提交
7381
	stat->blocks = (ALIGN(inode_get_bytes(inode), blocksize) +
7382
			ALIGN(delalloc_bytes, blocksize)) >> 9;
C
Chris Mason 已提交
7383 7384 7385
	return 0;
}

7386 7387 7388 7389 7390 7391 7392 7393 7394 7395 7396 7397 7398 7399
/*
 * If a file is moved, it will inherit the cow and compression flags of the new
 * directory.
 */
static void fixup_inode_flags(struct inode *dir, struct inode *inode)
{
	struct btrfs_inode *b_dir = BTRFS_I(dir);
	struct btrfs_inode *b_inode = BTRFS_I(inode);

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

7400
	if (b_dir->flags & BTRFS_INODE_COMPRESS) {
7401
		b_inode->flags |= BTRFS_INODE_COMPRESS;
7402 7403 7404 7405 7406
		b_inode->flags &= ~BTRFS_INODE_NOCOMPRESS;
	} else {
		b_inode->flags &= ~(BTRFS_INODE_COMPRESS |
				    BTRFS_INODE_NOCOMPRESS);
	}
7407 7408
}

C
Chris Mason 已提交
7409 7410
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
7411 7412 7413
{
	struct btrfs_trans_handle *trans;
	struct btrfs_root *root = BTRFS_I(old_dir)->root;
7414
	struct btrfs_root *dest = BTRFS_I(new_dir)->root;
C
Chris Mason 已提交
7415 7416 7417
	struct inode *new_inode = new_dentry->d_inode;
	struct inode *old_inode = old_dentry->d_inode;
	struct timespec ctime = CURRENT_TIME;
7418
	u64 index = 0;
7419
	u64 root_objectid;
C
Chris Mason 已提交
7420
	int ret;
L
Li Zefan 已提交
7421
	u64 old_ino = btrfs_ino(old_inode);
C
Chris Mason 已提交
7422

L
Li Zefan 已提交
7423
	if (btrfs_ino(new_dir) == BTRFS_EMPTY_SUBVOL_DIR_OBJECTID)
7424 7425
		return -EPERM;

7426
	/* we only allow rename subvolume link between subvolumes */
L
Li Zefan 已提交
7427
	if (old_ino != BTRFS_FIRST_FREE_OBJECTID && root != dest)
7428 7429
		return -EXDEV;

L
Li Zefan 已提交
7430 7431
	if (old_ino == BTRFS_EMPTY_SUBVOL_DIR_OBJECTID ||
	    (new_inode && btrfs_ino(new_inode) == BTRFS_FIRST_FREE_OBJECTID))
C
Chris Mason 已提交
7432
		return -ENOTEMPTY;
7433

7434 7435 7436
	if (S_ISDIR(old_inode->i_mode) && new_inode &&
	    new_inode->i_size > BTRFS_EMPTY_DIR_SIZE)
		return -ENOTEMPTY;
C
Chris Mason 已提交
7437 7438 7439 7440 7441 7442 7443 7444 7445 7446 7447 7448 7449 7450 7451 7452 7453 7454 7455 7456 7457 7458


	/* check for collisions, even if the  name isn't there */
	ret = btrfs_check_dir_item_collision(root, new_dir->i_ino,
			     new_dentry->d_name.name,
			     new_dentry->d_name.len);

	if (ret) {
		if (ret == -EEXIST) {
			/* we shouldn't get
			 * eexist without a new_inode */
			if (!new_inode) {
				WARN_ON(1);
				return ret;
			}
		} else {
			/* maybe -EOVERFLOW */
			return ret;
		}
	}
	ret = 0;

7459 7460 7461 7462 7463
	/*
	 * 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
	 */
7464
	if (new_inode && S_ISREG(old_inode->i_mode) && new_inode->i_size &&
7465 7466 7467
	    old_inode->i_size > BTRFS_ORDERED_OPERATIONS_FLUSH_LIMIT)
		filemap_flush(old_inode->i_mapping);

7468
	/* close the racy window with snapshot create/destroy ioctl */
L
Li Zefan 已提交
7469
	if (old_ino == BTRFS_FIRST_FREE_OBJECTID)
7470
		down_read(&root->fs_info->subvol_sem);
7471 7472 7473 7474 7475 7476 7477 7478 7479
	/*
	 * We want to reserve the absolute worst case amount of items.  So if
	 * both inodes are subvols and we need to unlink them then that would
	 * require 4 item modifications, but if they are both normal inodes it
	 * would require 5 item modifications, so we'll assume their normal
	 * inodes.  So 5 * 2 is 10, plus 1 for the new link, so 11 total items
	 * should cover the worst case number of items we'll modify.
	 */
	trans = btrfs_start_transaction(root, 20);
7480 7481 7482 7483
	if (IS_ERR(trans)) {
                ret = PTR_ERR(trans);
                goto out_notrans;
        }
7484

7485 7486
	if (dest != root)
		btrfs_record_root_in_trans(trans, dest);
7487

7488 7489 7490
	ret = btrfs_set_inode_index(new_dir, &index);
	if (ret)
		goto out_fail;
7491

L
Li Zefan 已提交
7492
	if (unlikely(old_ino == BTRFS_FIRST_FREE_OBJECTID)) {
7493 7494 7495
		/* force full log commit if subvolume involved. */
		root->fs_info->last_trans_log_full_commit = trans->transid;
	} else {
7496 7497 7498
		ret = btrfs_insert_inode_ref(trans, dest,
					     new_dentry->d_name.name,
					     new_dentry->d_name.len,
L
Li Zefan 已提交
7499 7500
					     old_ino,
					     btrfs_ino(new_dir), index);
7501 7502
		if (ret)
			goto out_fail;
7503 7504 7505 7506 7507 7508 7509 7510 7511
		/*
		 * 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);
	}
7512 7513 7514 7515
	/*
	 * make sure the inode gets flushed if it is replacing
	 * something.
	 */
L
Li Zefan 已提交
7516
	if (new_inode && new_inode->i_size && S_ISREG(old_inode->i_mode))
7517 7518
		btrfs_add_ordered_operation(trans, root, old_inode);

7519 7520 7521
	inode_inc_iversion(old_dir);
	inode_inc_iversion(new_dir);
	inode_inc_iversion(old_inode);
C
Chris Mason 已提交
7522 7523 7524
	old_dir->i_ctime = old_dir->i_mtime = ctime;
	new_dir->i_ctime = new_dir->i_mtime = ctime;
	old_inode->i_ctime = ctime;
7525

7526 7527 7528
	if (old_dentry->d_parent != new_dentry->d_parent)
		btrfs_record_unlink_dir(trans, old_dir, old_inode, 1);

L
Li Zefan 已提交
7529
	if (unlikely(old_ino == BTRFS_FIRST_FREE_OBJECTID)) {
7530 7531 7532 7533 7534
		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 {
7535 7536 7537 7538 7539 7540
		ret = __btrfs_unlink_inode(trans, root, old_dir,
					old_dentry->d_inode,
					old_dentry->d_name.name,
					old_dentry->d_name.len);
		if (!ret)
			ret = btrfs_update_inode(trans, root, old_inode);
7541
	}
7542 7543 7544 7545
	if (ret) {
		btrfs_abort_transaction(trans, root, ret);
		goto out_fail;
	}
C
Chris Mason 已提交
7546 7547

	if (new_inode) {
7548
		inode_inc_iversion(new_inode);
C
Chris Mason 已提交
7549
		new_inode->i_ctime = CURRENT_TIME;
L
Li Zefan 已提交
7550
		if (unlikely(btrfs_ino(new_inode) ==
7551 7552 7553 7554 7555 7556 7557 7558 7559 7560 7561 7562 7563
			     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);
		}
7564
		if (!ret && new_inode->i_nlink == 0) {
7565
			ret = btrfs_orphan_add(trans, new_dentry->d_inode);
7566
			BUG_ON(ret);
7567
		}
7568 7569 7570 7571
		if (ret) {
			btrfs_abort_transaction(trans, root, ret);
			goto out_fail;
		}
C
Chris Mason 已提交
7572
	}
7573

7574 7575
	fixup_inode_flags(new_dir, old_inode);

7576 7577
	ret = btrfs_add_link(trans, new_dir, old_inode,
			     new_dentry->d_name.name,
7578
			     new_dentry->d_name.len, 0, index);
7579 7580 7581 7582
	if (ret) {
		btrfs_abort_transaction(trans, root, ret);
		goto out_fail;
	}
C
Chris Mason 已提交
7583

L
Li Zefan 已提交
7584
	if (old_ino != BTRFS_FIRST_FREE_OBJECTID) {
7585
		struct dentry *parent = new_dentry->d_parent;
7586
		btrfs_log_new_name(trans, old_inode, old_dir, parent);
7587 7588
		btrfs_end_log_trans(root);
	}
C
Chris Mason 已提交
7589
out_fail:
7590
	btrfs_end_transaction(trans, root);
7591
out_notrans:
L
Li Zefan 已提交
7592
	if (old_ino == BTRFS_FIRST_FREE_OBJECTID)
7593
		up_read(&root->fs_info->subvol_sem);
J
Josef Bacik 已提交
7594

C
Chris Mason 已提交
7595 7596 7597
	return ret;
}

7598 7599 7600 7601 7602 7603 7604 7605 7606 7607 7608 7609 7610 7611 7612 7613 7614 7615 7616 7617 7618 7619 7620 7621 7622 7623 7624 7625 7626 7627 7628 7629 7630 7631 7632 7633 7634 7635 7636 7637 7638 7639 7640
static void btrfs_run_delalloc_work(struct btrfs_work *work)
{
	struct btrfs_delalloc_work *delalloc_work;

	delalloc_work = container_of(work, struct btrfs_delalloc_work,
				     work);
	if (delalloc_work->wait)
		btrfs_wait_ordered_range(delalloc_work->inode, 0, (u64)-1);
	else
		filemap_flush(delalloc_work->inode->i_mapping);

	if (delalloc_work->delay_iput)
		btrfs_add_delayed_iput(delalloc_work->inode);
	else
		iput(delalloc_work->inode);
	complete(&delalloc_work->completion);
}

struct btrfs_delalloc_work *btrfs_alloc_delalloc_work(struct inode *inode,
						    int wait, int delay_iput)
{
	struct btrfs_delalloc_work *work;

	work = kmem_cache_zalloc(btrfs_delalloc_work_cachep, GFP_NOFS);
	if (!work)
		return NULL;

	init_completion(&work->completion);
	INIT_LIST_HEAD(&work->list);
	work->inode = inode;
	work->wait = wait;
	work->delay_iput = delay_iput;
	work->work.func = btrfs_run_delalloc_work;

	return work;
}

void btrfs_wait_and_free_delalloc_work(struct btrfs_delalloc_work *work)
{
	wait_for_completion(&work->completion);
	kmem_cache_free(btrfs_delalloc_work_cachep, work);
}

C
Chris Mason 已提交
7641 7642 7643 7644
/*
 * some fairly slow code that needs optimization. This walks the list
 * of all the inodes with pending delalloc and forces them to disk.
 */
Y
Yan, Zheng 已提交
7645
int btrfs_start_delalloc_inodes(struct btrfs_root *root, int delay_iput)
7646 7647
{
	struct btrfs_inode *binode;
7648
	struct inode *inode;
7649 7650
	struct btrfs_delalloc_work *work, *next;
	struct list_head works;
7651
	struct list_head splice;
7652
	int ret = 0;
7653

Y
Yan Zheng 已提交
7654 7655 7656
	if (root->fs_info->sb->s_flags & MS_RDONLY)
		return -EROFS;

7657
	INIT_LIST_HEAD(&works);
7658
	INIT_LIST_HEAD(&splice);
7659

7660
	spin_lock(&root->fs_info->delalloc_lock);
7661 7662 7663
	list_splice_init(&root->fs_info->delalloc_inodes, &splice);
	while (!list_empty(&splice)) {
		binode = list_entry(splice.next, struct btrfs_inode,
7664
				    delalloc_inodes);
7665 7666 7667

		list_del_init(&binode->delalloc_inodes);

7668
		inode = igrab(&binode->vfs_inode);
7669 7670 7671
		if (!inode) {
			clear_bit(BTRFS_INODE_IN_DELALLOC_LIST,
				  &binode->runtime_flags);
7672
			continue;
7673
		}
7674 7675 7676

		list_add_tail(&binode->delalloc_inodes,
			      &root->fs_info->delalloc_inodes);
7677
		spin_unlock(&root->fs_info->delalloc_lock);
7678 7679 7680 7681 7682

		work = btrfs_alloc_delalloc_work(inode, 0, delay_iput);
		if (unlikely(!work)) {
			ret = -ENOMEM;
			goto out;
7683
		}
7684 7685 7686 7687
		list_add_tail(&work->list, &works);
		btrfs_queue_worker(&root->fs_info->flush_workers,
				   &work->work);

7688
		cond_resched();
7689
		spin_lock(&root->fs_info->delalloc_lock);
7690
	}
7691
	spin_unlock(&root->fs_info->delalloc_lock);
7692

7693 7694 7695 7696 7697
	list_for_each_entry_safe(work, next, &works, list) {
		list_del_init(&work->list);
		btrfs_wait_and_free_delalloc_work(work);
	}

7698 7699 7700 7701 7702
	/* 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
7703
	while (atomic_read(&root->fs_info->nr_async_submits) ||
7704
	      atomic_read(&root->fs_info->async_delalloc_pages)) {
7705
		wait_event(root->fs_info->async_submit_wait,
7706 7707
		   (atomic_read(&root->fs_info->nr_async_submits) == 0 &&
		    atomic_read(&root->fs_info->async_delalloc_pages) == 0));
7708 7709
	}
	atomic_dec(&root->fs_info->async_submit_draining);
7710
	return 0;
7711 7712 7713 7714 7715
out:
	list_for_each_entry_safe(work, next, &works, list) {
		list_del_init(&work->list);
		btrfs_wait_and_free_delalloc_work(work);
	}
7716 7717 7718 7719 7720 7721

	if (!list_empty_careful(&splice)) {
		spin_lock(&root->fs_info->delalloc_lock);
		list_splice_tail(&splice, &root->fs_info->delalloc_inodes);
		spin_unlock(&root->fs_info->delalloc_lock);
	}
7722
	return ret;
7723 7724
}

C
Chris Mason 已提交
7725 7726 7727 7728 7729 7730 7731
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;
7732
	struct inode *inode = NULL;
C
Chris Mason 已提交
7733 7734 7735
	int err;
	int drop_inode = 0;
	u64 objectid;
7736
	u64 index = 0 ;
C
Chris Mason 已提交
7737 7738
	int name_len;
	int datasize;
7739
	unsigned long ptr;
C
Chris Mason 已提交
7740
	struct btrfs_file_extent_item *ei;
7741
	struct extent_buffer *leaf;
C
Chris Mason 已提交
7742 7743 7744 7745

	name_len = strlen(symname) + 1;
	if (name_len > BTRFS_MAX_INLINE_DATA_SIZE(root))
		return -ENAMETOOLONG;
7746

J
Josef Bacik 已提交
7747 7748 7749 7750 7751
	/*
	 * 2 items for inode item and ref
	 * 2 items for dir items
	 * 1 item for xattr if selinux is on
	 */
7752 7753 7754
	trans = btrfs_start_transaction(root, 5);
	if (IS_ERR(trans))
		return PTR_ERR(trans);
7755

7756 7757 7758 7759
	err = btrfs_find_free_ino(root, &objectid);
	if (err)
		goto out_unlock;

7760
	inode = btrfs_new_inode(trans, root, dir, dentry->d_name.name,
L
Li Zefan 已提交
7761
				dentry->d_name.len, btrfs_ino(dir), objectid,
7762
				S_IFLNK|S_IRWXUGO, &index);
7763 7764
	if (IS_ERR(inode)) {
		err = PTR_ERR(inode);
C
Chris Mason 已提交
7765
		goto out_unlock;
7766
	}
C
Chris Mason 已提交
7767

7768
	err = btrfs_init_inode_security(trans, inode, dir, &dentry->d_name);
J
Josef Bacik 已提交
7769 7770 7771 7772 7773
	if (err) {
		drop_inode = 1;
		goto out_unlock;
	}

7774 7775 7776 7777 7778 7779 7780 7781 7782
	/*
	* If the active LSM wants to access the inode during
	* d_instantiate it needs these. Smack checks to see
	* if the filesystem supports xattrs by looking at the
	* ops vector.
	*/
	inode->i_fop = &btrfs_file_operations;
	inode->i_op = &btrfs_file_inode_operations;

7783
	err = btrfs_add_nondir(trans, dir, dentry, inode, 0, index);
C
Chris Mason 已提交
7784 7785 7786 7787
	if (err)
		drop_inode = 1;
	else {
		inode->i_mapping->a_ops = &btrfs_aops;
C
Chris Mason 已提交
7788
		inode->i_mapping->backing_dev_info = &root->fs_info->bdi;
7789
		BTRFS_I(inode)->io_tree.ops = &btrfs_extent_io_ops;
C
Chris Mason 已提交
7790 7791 7792 7793 7794
	}
	if (drop_inode)
		goto out_unlock;

	path = btrfs_alloc_path();
7795 7796 7797 7798 7799
	if (!path) {
		err = -ENOMEM;
		drop_inode = 1;
		goto out_unlock;
	}
L
Li Zefan 已提交
7800
	key.objectid = btrfs_ino(inode);
C
Chris Mason 已提交
7801 7802 7803 7804 7805
	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);
7806 7807
	if (err) {
		drop_inode = 1;
7808
		btrfs_free_path(path);
7809 7810
		goto out_unlock;
	}
7811 7812 7813 7814 7815
	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
7816
				   BTRFS_FILE_EXTENT_INLINE);
C
Chris Mason 已提交
7817 7818 7819 7820 7821
	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
7822
	ptr = btrfs_file_extent_inline_start(ei);
7823 7824
	write_extent_buffer(leaf, symname, ptr, name_len);
	btrfs_mark_buffer_dirty(leaf);
C
Chris Mason 已提交
7825
	btrfs_free_path(path);
7826

C
Chris Mason 已提交
7827 7828
	inode->i_op = &btrfs_symlink_inode_operations;
	inode->i_mapping->a_ops = &btrfs_symlink_aops;
C
Chris Mason 已提交
7829
	inode->i_mapping->backing_dev_info = &root->fs_info->bdi;
Y
Yan Zheng 已提交
7830
	inode_set_bytes(inode, name_len);
7831
	btrfs_i_size_write(inode, name_len - 1);
7832 7833 7834
	err = btrfs_update_inode(trans, root, inode);
	if (err)
		drop_inode = 1;
C
Chris Mason 已提交
7835 7836

out_unlock:
7837 7838
	if (!err)
		d_instantiate(dentry, inode);
7839
	btrfs_end_transaction(trans, root);
C
Chris Mason 已提交
7840 7841 7842 7843
	if (drop_inode) {
		inode_dec_link_count(inode);
		iput(inode);
	}
7844
	btrfs_btree_balance_dirty(root);
C
Chris Mason 已提交
7845 7846
	return err;
}
7847

7848 7849 7850 7851
static int __btrfs_prealloc_file_range(struct inode *inode, int mode,
				       u64 start, u64 num_bytes, u64 min_size,
				       loff_t actual_len, u64 *alloc_hint,
				       struct btrfs_trans_handle *trans)
Y
Yan Zheng 已提交
7852
{
J
Josef Bacik 已提交
7853 7854
	struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
	struct extent_map *em;
Y
Yan Zheng 已提交
7855 7856 7857
	struct btrfs_root *root = BTRFS_I(inode)->root;
	struct btrfs_key ins;
	u64 cur_offset = start;
7858
	u64 i_size;
Y
Yan Zheng 已提交
7859
	int ret = 0;
7860
	bool own_trans = true;
Y
Yan Zheng 已提交
7861

7862 7863
	if (trans)
		own_trans = false;
Y
Yan Zheng 已提交
7864
	while (num_bytes > 0) {
7865 7866 7867 7868 7869 7870
		if (own_trans) {
			trans = btrfs_start_transaction(root, 3);
			if (IS_ERR(trans)) {
				ret = PTR_ERR(trans);
				break;
			}
7871 7872
		}

7873
		ret = btrfs_reserve_extent(trans, root, num_bytes, min_size,
7874
					   0, *alloc_hint, &ins, 1);
7875
		if (ret) {
7876 7877
			if (own_trans)
				btrfs_end_transaction(trans, root);
7878
			break;
Y
Yan Zheng 已提交
7879
		}
7880

Y
Yan Zheng 已提交
7881 7882 7883
		ret = insert_reserved_file_extent(trans, inode,
						  cur_offset, ins.objectid,
						  ins.offset, ins.offset,
Y
Yan, Zheng 已提交
7884
						  ins.offset, 0, 0, 0,
Y
Yan Zheng 已提交
7885
						  BTRFS_FILE_EXTENT_PREALLOC);
7886 7887 7888 7889 7890 7891
		if (ret) {
			btrfs_abort_transaction(trans, root, ret);
			if (own_trans)
				btrfs_end_transaction(trans, root);
			break;
		}
C
Chris Mason 已提交
7892 7893
		btrfs_drop_extent_cache(inode, cur_offset,
					cur_offset + ins.offset -1, 0);
7894

J
Josef Bacik 已提交
7895 7896 7897 7898 7899 7900 7901 7902 7903 7904 7905 7906
		em = alloc_extent_map();
		if (!em) {
			set_bit(BTRFS_INODE_NEEDS_FULL_SYNC,
				&BTRFS_I(inode)->runtime_flags);
			goto next;
		}

		em->start = cur_offset;
		em->orig_start = cur_offset;
		em->len = ins.offset;
		em->block_start = ins.objectid;
		em->block_len = ins.offset;
7907
		em->orig_block_len = ins.offset;
J
Josef Bacik 已提交
7908 7909 7910 7911 7912 7913 7914 7915 7916 7917 7918 7919 7920 7921 7922 7923 7924 7925 7926
		em->bdev = root->fs_info->fs_devices->latest_bdev;
		set_bit(EXTENT_FLAG_PREALLOC, &em->flags);
		em->generation = trans->transid;

		while (1) {
			write_lock(&em_tree->lock);
			ret = add_extent_mapping(em_tree, em);
			if (!ret)
				list_move(&em->list,
					  &em_tree->modified_extents);
			write_unlock(&em_tree->lock);
			if (ret != -EEXIST)
				break;
			btrfs_drop_extent_cache(inode, cur_offset,
						cur_offset + ins.offset - 1,
						0);
		}
		free_extent_map(em);
next:
Y
Yan Zheng 已提交
7927 7928
		num_bytes -= ins.offset;
		cur_offset += ins.offset;
7929
		*alloc_hint = ins.objectid + ins.offset;
7930

7931
		inode_inc_iversion(inode);
Y
Yan Zheng 已提交
7932
		inode->i_ctime = CURRENT_TIME;
7933
		BTRFS_I(inode)->flags |= BTRFS_INODE_PREALLOC;
Y
Yan Zheng 已提交
7934
		if (!(mode & FALLOC_FL_KEEP_SIZE) &&
7935 7936
		    (actual_len > inode->i_size) &&
		    (cur_offset > inode->i_size)) {
7937
			if (cur_offset > actual_len)
7938
				i_size = actual_len;
7939
			else
7940 7941 7942
				i_size = cur_offset;
			i_size_write(inode, i_size);
			btrfs_ordered_update_i_size(inode, i_size, NULL);
7943 7944
		}

Y
Yan Zheng 已提交
7945
		ret = btrfs_update_inode(trans, root, inode);
7946 7947 7948 7949 7950 7951 7952

		if (ret) {
			btrfs_abort_transaction(trans, root, ret);
			if (own_trans)
				btrfs_end_transaction(trans, root);
			break;
		}
Y
Yan Zheng 已提交
7953

7954 7955
		if (own_trans)
			btrfs_end_transaction(trans, root);
7956
	}
Y
Yan Zheng 已提交
7957 7958 7959
	return ret;
}

7960 7961 7962 7963 7964 7965 7966 7967 7968 7969 7970 7971 7972 7973 7974 7975 7976 7977
int btrfs_prealloc_file_range(struct inode *inode, int mode,
			      u64 start, u64 num_bytes, u64 min_size,
			      loff_t actual_len, u64 *alloc_hint)
{
	return __btrfs_prealloc_file_range(inode, mode, start, num_bytes,
					   min_size, actual_len, alloc_hint,
					   NULL);
}

int btrfs_prealloc_file_range_trans(struct inode *inode,
				    struct btrfs_trans_handle *trans, int mode,
				    u64 start, u64 num_bytes, u64 min_size,
				    loff_t actual_len, u64 *alloc_hint)
{
	return __btrfs_prealloc_file_range(inode, mode, start, num_bytes,
					   min_size, actual_len, alloc_hint, trans);
}

7978 7979 7980 7981 7982
static int btrfs_set_page_dirty(struct page *page)
{
	return __set_page_dirty_nobuffers(page);
}

7983
static int btrfs_permission(struct inode *inode, int mask)
Y
Yan 已提交
7984
{
L
Li Zefan 已提交
7985
	struct btrfs_root *root = BTRFS_I(inode)->root;
7986
	umode_t mode = inode->i_mode;
L
Li Zefan 已提交
7987

7988 7989 7990 7991 7992 7993 7994
	if (mask & MAY_WRITE &&
	    (S_ISREG(mode) || S_ISDIR(mode) || S_ISLNK(mode))) {
		if (btrfs_root_readonly(root))
			return -EROFS;
		if (BTRFS_I(inode)->flags & BTRFS_INODE_READONLY)
			return -EACCES;
	}
7995
	return generic_permission(inode, mask);
Y
Yan 已提交
7996
}
C
Chris Mason 已提交
7997

7998
static const struct inode_operations btrfs_dir_inode_operations = {
7999
	.getattr	= btrfs_getattr,
C
Chris Mason 已提交
8000 8001 8002 8003 8004 8005 8006 8007 8008
	.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
8009
	.mknod		= btrfs_mknod,
8010 8011
	.setxattr	= btrfs_setxattr,
	.getxattr	= btrfs_getxattr,
J
Josef Bacik 已提交
8012
	.listxattr	= btrfs_listxattr,
8013
	.removexattr	= btrfs_removexattr,
Y
Yan 已提交
8014
	.permission	= btrfs_permission,
8015
	.get_acl	= btrfs_get_acl,
C
Chris Mason 已提交
8016
};
8017
static const struct inode_operations btrfs_dir_ro_inode_operations = {
C
Chris Mason 已提交
8018
	.lookup		= btrfs_lookup,
Y
Yan 已提交
8019
	.permission	= btrfs_permission,
8020
	.get_acl	= btrfs_get_acl,
C
Chris Mason 已提交
8021
};
8022

8023
static const struct file_operations btrfs_dir_file_operations = {
C
Chris Mason 已提交
8024 8025
	.llseek		= generic_file_llseek,
	.read		= generic_read_dir,
8026
	.readdir	= btrfs_real_readdir,
C
Christoph Hellwig 已提交
8027
	.unlocked_ioctl	= btrfs_ioctl,
C
Chris Mason 已提交
8028
#ifdef CONFIG_COMPAT
C
Christoph Hellwig 已提交
8029
	.compat_ioctl	= btrfs_ioctl,
C
Chris Mason 已提交
8030
#endif
S
Sage Weil 已提交
8031
	.release        = btrfs_release_file,
8032
	.fsync		= btrfs_sync_file,
C
Chris Mason 已提交
8033 8034
};

8035
static struct extent_io_ops btrfs_extent_io_ops = {
8036
	.fill_delalloc = run_delalloc_range,
8037
	.submit_bio_hook = btrfs_submit_bio_hook,
8038
	.merge_bio_hook = btrfs_merge_bio_hook,
8039
	.readpage_end_io_hook = btrfs_readpage_end_io_hook,
8040
	.writepage_end_io_hook = btrfs_writepage_end_io_hook,
8041
	.writepage_start_hook = btrfs_writepage_start_hook,
C
Chris Mason 已提交
8042 8043
	.set_bit_hook = btrfs_set_bit_hook,
	.clear_bit_hook = btrfs_clear_bit_hook,
J
Josef Bacik 已提交
8044 8045
	.merge_extent_hook = btrfs_merge_extent_hook,
	.split_extent_hook = btrfs_split_extent_hook,
8046 8047
};

8048 8049 8050 8051 8052 8053 8054 8055 8056 8057 8058 8059
/*
 * 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.
 */
8060
static const struct address_space_operations btrfs_aops = {
C
Chris Mason 已提交
8061 8062
	.readpage	= btrfs_readpage,
	.writepage	= btrfs_writepage,
C
Chris Mason 已提交
8063
	.writepages	= btrfs_writepages,
C
Chris Mason 已提交
8064
	.readpages	= btrfs_readpages,
8065
	.direct_IO	= btrfs_direct_IO,
8066 8067
	.invalidatepage = btrfs_invalidatepage,
	.releasepage	= btrfs_releasepage,
8068
	.set_page_dirty	= btrfs_set_page_dirty,
8069
	.error_remove_page = generic_error_remove_page,
C
Chris Mason 已提交
8070 8071
};

8072
static const struct address_space_operations btrfs_symlink_aops = {
C
Chris Mason 已提交
8073 8074
	.readpage	= btrfs_readpage,
	.writepage	= btrfs_writepage,
C
Chris Mason 已提交
8075 8076
	.invalidatepage = btrfs_invalidatepage,
	.releasepage	= btrfs_releasepage,
C
Chris Mason 已提交
8077 8078
};

8079
static const struct inode_operations btrfs_file_inode_operations = {
C
Chris Mason 已提交
8080 8081
	.getattr	= btrfs_getattr,
	.setattr	= btrfs_setattr,
8082 8083
	.setxattr	= btrfs_setxattr,
	.getxattr	= btrfs_getxattr,
J
Josef Bacik 已提交
8084
	.listxattr      = btrfs_listxattr,
8085
	.removexattr	= btrfs_removexattr,
Y
Yan 已提交
8086
	.permission	= btrfs_permission,
Y
Yehuda Sadeh 已提交
8087
	.fiemap		= btrfs_fiemap,
8088
	.get_acl	= btrfs_get_acl,
8089
	.update_time	= btrfs_update_time,
C
Chris Mason 已提交
8090
};
8091
static const struct inode_operations btrfs_special_inode_operations = {
J
Josef Bacik 已提交
8092 8093
	.getattr	= btrfs_getattr,
	.setattr	= btrfs_setattr,
Y
Yan 已提交
8094
	.permission	= btrfs_permission,
8095 8096
	.setxattr	= btrfs_setxattr,
	.getxattr	= btrfs_getxattr,
J
Josef Bacik 已提交
8097
	.listxattr	= btrfs_listxattr,
8098
	.removexattr	= btrfs_removexattr,
8099
	.get_acl	= btrfs_get_acl,
8100
	.update_time	= btrfs_update_time,
J
Josef Bacik 已提交
8101
};
8102
static const struct inode_operations btrfs_symlink_inode_operations = {
C
Chris Mason 已提交
8103 8104 8105
	.readlink	= generic_readlink,
	.follow_link	= page_follow_link_light,
	.put_link	= page_put_link,
8106
	.getattr	= btrfs_getattr,
8107
	.setattr	= btrfs_setattr,
Y
Yan 已提交
8108
	.permission	= btrfs_permission,
J
Jim Owens 已提交
8109 8110 8111 8112
	.setxattr	= btrfs_setxattr,
	.getxattr	= btrfs_getxattr,
	.listxattr	= btrfs_listxattr,
	.removexattr	= btrfs_removexattr,
8113
	.get_acl	= btrfs_get_acl,
8114
	.update_time	= btrfs_update_time,
C
Chris Mason 已提交
8115
};
8116

8117
const struct dentry_operations btrfs_dentry_operations = {
8118
	.d_delete	= btrfs_dentry_delete,
8119
	.d_release	= btrfs_dentry_release,
8120
};