inode.c 196.8 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>
C
Chris Mason 已提交
40
#include "compat.h"
C
Chris Mason 已提交
41 42 43 44 45 46
#include "ctree.h"
#include "disk-io.h"
#include "transaction.h"
#include "btrfs_inode.h"
#include "ioctl.h"
#include "print-tree.h"
47
#include "volumes.h"
48
#include "ordered-data.h"
49
#include "xattr.h"
50
#include "tree-log.h"
C
Chris Mason 已提交
51
#include "compression.h"
52
#include "locking.h"
53
#include "free-space-cache.h"
C
Chris Mason 已提交
54 55 56 57 58 59

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

60 61 62 63 64
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;
65 66
static const struct address_space_operations btrfs_aops;
static const struct address_space_operations btrfs_symlink_aops;
67
static const struct file_operations btrfs_dir_file_operations;
68
static struct extent_io_ops btrfs_extent_io_ops;
C
Chris Mason 已提交
69 70 71 72 73

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

#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,
};

87 88
static int btrfs_setsize(struct inode *inode, loff_t newsize);
static int btrfs_truncate(struct inode *inode);
C
Chris Mason 已提交
89
static int btrfs_finish_ordered_io(struct inode *inode, u64 start, u64 end);
90 91 92 93
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);
94

95 96
static int btrfs_init_inode_security(struct btrfs_trans_handle *trans,
				     struct inode *inode,  struct inode *dir)
J
Jim Owens 已提交
97 98 99
{
	int err;

100
	err = btrfs_init_acl(trans, inode, dir);
J
Jim Owens 已提交
101
	if (!err)
102
		err = btrfs_xattr_security_init(trans, inode, dir);
J
Jim Owens 已提交
103 104 105
	return err;
}

C
Chris Mason 已提交
106 107 108 109 110
/*
 * 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
111
static noinline int insert_inline_extent(struct btrfs_trans_handle *trans,
C
Chris Mason 已提交
112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127
				struct btrfs_root *root, struct inode *inode,
				u64 start, size_t size, size_t compressed_size,
				struct page **compressed_pages)
{
	struct btrfs_key key;
	struct btrfs_path *path;
	struct extent_buffer *leaf;
	struct page *page = NULL;
	char *kaddr;
	unsigned long ptr;
	struct btrfs_file_extent_item *ei;
	int err = 0;
	int ret;
	size_t cur_size = size;
	size_t datasize;
	unsigned long offset;
128
	int compress_type = BTRFS_COMPRESS_NONE;
C
Chris Mason 已提交
129 130

	if (compressed_size && compressed_pages) {
131
		compress_type = root->fs_info->compress_type;
C
Chris Mason 已提交
132 133 134
		cur_size = compressed_size;
	}

C
Chris Mason 已提交
135 136
	path = btrfs_alloc_path();
	if (!path)
C
Chris Mason 已提交
137 138
		return -ENOMEM;

139
	path->leave_spinning = 1;
C
Chris Mason 已提交
140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164
	btrfs_set_trans_block_group(trans, inode);

	key.objectid = inode->i_ino;
	key.offset = start;
	btrfs_set_key_type(&key, BTRFS_EXTENT_DATA_KEY);
	datasize = btrfs_file_extent_calc_inline_size(cur_size);

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

165
	if (compress_type != BTRFS_COMPRESS_NONE) {
C
Chris Mason 已提交
166 167
		struct page *cpage;
		int i = 0;
C
Chris Mason 已提交
168
		while (compressed_size > 0) {
C
Chris Mason 已提交
169
			cpage = compressed_pages[i];
170
			cur_size = min_t(unsigned long, compressed_size,
C
Chris Mason 已提交
171 172
				       PAGE_CACHE_SIZE);

173
			kaddr = kmap_atomic(cpage, KM_USER0);
C
Chris Mason 已提交
174
			write_extent_buffer(leaf, kaddr, ptr, cur_size);
175
			kunmap_atomic(kaddr, KM_USER0);
C
Chris Mason 已提交
176 177 178 179 180 181

			i++;
			ptr += cur_size;
			compressed_size -= cur_size;
		}
		btrfs_set_file_extent_compression(leaf, ei,
182
						  compress_type);
C
Chris Mason 已提交
183 184 185 186 187 188 189 190 191 192 193 194 195
	} else {
		page = find_get_page(inode->i_mapping,
				     start >> PAGE_CACHE_SHIFT);
		btrfs_set_file_extent_compression(leaf, ei, 0);
		kaddr = kmap_atomic(page, KM_USER0);
		offset = start & (PAGE_CACHE_SIZE - 1);
		write_extent_buffer(leaf, kaddr + offset, ptr, size);
		kunmap_atomic(kaddr, KM_USER0);
		page_cache_release(page);
	}
	btrfs_mark_buffer_dirty(leaf);
	btrfs_free_path(path);

196 197 198 199 200 201 202 203 204
	/*
	 * 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
205 206
	BTRFS_I(inode)->disk_i_size = inode->i_size;
	btrfs_update_inode(trans, root, inode);
207

C
Chris Mason 已提交
208 209 210 211 212 213 214 215 216 217 218 219
	return 0;
fail:
	btrfs_free_path(path);
	return err;
}


/*
 * conditionally insert an inline extent into the file.  This
 * does the checks required to make sure the data is small enough
 * to fit as an inline extent.
 */
220
static noinline int cow_file_range_inline(struct btrfs_trans_handle *trans,
C
Chris Mason 已提交
221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238
				 struct btrfs_root *root,
				 struct inode *inode, u64 start, u64 end,
				 size_t compressed_size,
				 struct page **compressed_pages)
{
	u64 isize = i_size_read(inode);
	u64 actual_end = min(end + 1, isize);
	u64 inline_len = actual_end - start;
	u64 aligned_end = (end + root->sectorsize - 1) &
			~((u64)root->sectorsize - 1);
	u64 hint_byte;
	u64 data_len = inline_len;
	int ret;

	if (compressed_size)
		data_len = compressed_size;

	if (start > 0 ||
C
Chris Mason 已提交
239
	    actual_end >= PAGE_CACHE_SIZE ||
C
Chris Mason 已提交
240 241 242 243 244 245 246 247
	    data_len >= BTRFS_MAX_INLINE_DATA_SIZE(root) ||
	    (!compressed_size &&
	    (actual_end & (root->sectorsize - 1)) == 0) ||
	    end + 1 < isize ||
	    data_len > root->fs_info->max_inline) {
		return 1;
	}

Y
Yan, Zheng 已提交
248
	ret = btrfs_drop_extents(trans, inode, start, aligned_end,
C
Chris Mason 已提交
249
				 &hint_byte, 1);
C
Chris Mason 已提交
250 251 252 253 254 255 256 257
	BUG_ON(ret);

	if (isize > actual_end)
		inline_len = min_t(u64, isize, actual_end);
	ret = insert_inline_extent(trans, root, inode, start,
				   inline_len, compressed_size,
				   compressed_pages);
	BUG_ON(ret);
258
	btrfs_delalloc_release_metadata(inode, end + 1 - start);
C
Chris Mason 已提交
259
	btrfs_drop_extent_cache(inode, start, aligned_end - 1, 0);
C
Chris Mason 已提交
260 261 262
	return 0;
}

263 264 265 266 267 268
struct async_extent {
	u64 start;
	u64 ram_size;
	u64 compressed_size;
	struct page **pages;
	unsigned long nr_pages;
269
	int compress_type;
270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286
	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,
287 288
				     unsigned long nr_pages,
				     int compress_type)
289 290 291 292 293 294 295 296 297
{
	struct async_extent *async_extent;

	async_extent = kmalloc(sizeof(*async_extent), GFP_NOFS);
	async_extent->start = start;
	async_extent->ram_size = ram_size;
	async_extent->compressed_size = compressed_size;
	async_extent->pages = pages;
	async_extent->nr_pages = nr_pages;
298
	async_extent->compress_type = compress_type;
299 300 301 302
	list_add_tail(&async_extent->list, &cow->extents);
	return 0;
}

C
Chris Mason 已提交
303
/*
304 305 306
 * 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
307
 *
308 309 310 311 312
 * 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
313
 *
314 315 316 317
 * If this code finds it can't get good compression, it puts an
 * entry onto the work queue to write the uncompressed bytes.  This
 * makes sure that both compressed inodes and uncompressed inodes
 * are written in the same order that pdflush sent them down.
C
Chris Mason 已提交
318
 */
319 320 321 322 323
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)
324 325 326
{
	struct btrfs_root *root = BTRFS_I(inode)->root;
	struct btrfs_trans_handle *trans;
327 328
	u64 num_bytes;
	u64 blocksize = root->sectorsize;
C
Chris Mason 已提交
329
	u64 actual_end;
330
	u64 isize = i_size_read(inode);
331
	int ret = 0;
C
Chris Mason 已提交
332 333 334 335 336 337
	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;
338
	unsigned long max_uncompressed = 128 * 1024;
C
Chris Mason 已提交
339 340
	int i;
	int will_compress;
341
	int compress_type = root->fs_info->compress_type;
342

343
	actual_end = min_t(u64, isize, end + 1);
C
Chris Mason 已提交
344 345 346 347
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);
348

349 350 351 352 353 354 355 356 357 358 359 360 361
	/*
	 * 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
362 363 364 365
	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
366 367 368 369 370 371 372
	 * 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
373 374
	 */
	total_compressed = min(total_compressed, max_uncompressed);
375
	num_bytes = (end - start + blocksize) & ~(blocksize - 1);
376
	num_bytes = max(blocksize,  num_bytes);
C
Chris Mason 已提交
377 378
	total_in = 0;
	ret = 0;
379

380 381 382 383
	/*
	 * 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
384
	 */
385
	if (!(BTRFS_I(inode)->flags & BTRFS_INODE_NOCOMPRESS) &&
C
Chris Mason 已提交
386 387
	    (btrfs_test_opt(root, COMPRESS) ||
	     (BTRFS_I(inode)->force_compress))) {
C
Chris Mason 已提交
388
		WARN_ON(pages);
389
		pages = kzalloc(sizeof(struct page *) * nr_pages, GFP_NOFS);
C
Chris Mason 已提交
390

391 392 393 394 395 396 397 398 399 400
		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
401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420

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

			/* zero the tail end of the last page, we might be
			 * sending it down to disk
			 */
			if (offset) {
				kaddr = kmap_atomic(page, KM_USER0);
				memset(kaddr + offset, 0,
				       PAGE_CACHE_SIZE - offset);
				kunmap_atomic(kaddr, KM_USER0);
			}
			will_compress = 1;
		}
	}
	if (start == 0) {
421
		trans = btrfs_join_transaction(root, 1);
422
		BUG_ON(IS_ERR(trans));
423
		btrfs_set_trans_block_group(trans, inode);
424
		trans->block_rsv = &root->fs_info->delalloc_block_rsv;
425

C
Chris Mason 已提交
426
		/* lets try to make an inline extent */
427
		if (ret || total_in < (actual_end - start)) {
C
Chris Mason 已提交
428
			/* we didn't compress the entire range, try
429
			 * to make an uncompressed inline extent.
C
Chris Mason 已提交
430 431 432 433
			 */
			ret = cow_file_range_inline(trans, root, inode,
						    start, end, 0, NULL);
		} else {
434
			/* try making a compressed inline extent */
C
Chris Mason 已提交
435 436 437 438 439
			ret = cow_file_range_inline(trans, root, inode,
						    start, end,
						    total_compressed, pages);
		}
		if (ret == 0) {
440 441 442 443 444
			/*
			 * inline extent creation worked, we don't need
			 * to create any more async work items.  Unlock
			 * and free up our temp pages.
			 */
C
Chris Mason 已提交
445
			extent_clear_unlock_delalloc(inode,
446 447 448
			     &BTRFS_I(inode)->io_tree,
			     start, end, NULL,
			     EXTENT_CLEAR_UNLOCK_PAGE | EXTENT_CLEAR_DIRTY |
449
			     EXTENT_CLEAR_DELALLOC |
450
			     EXTENT_SET_WRITEBACK | EXTENT_END_WRITEBACK);
451 452

			btrfs_end_transaction(trans, root);
C
Chris Mason 已提交
453 454
			goto free_pages_out;
		}
455
		btrfs_end_transaction(trans, root);
C
Chris Mason 已提交
456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484
	}

	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
485
			WARN_ON(pages[i]->mapping);
C
Chris Mason 已提交
486 487 488 489 490 491 492 493
			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
494 495
		if (!btrfs_test_opt(root, FORCE_COMPRESS) &&
		    !(BTRFS_I(inode)->force_compress)) {
C
Chris Mason 已提交
496
			BTRFS_I(inode)->flags |= BTRFS_INODE_NOCOMPRESS;
C
Chris Mason 已提交
497
		}
C
Chris Mason 已提交
498
	}
499 500
	if (will_compress) {
		*num_added += 1;
C
Chris Mason 已提交
501

502 503 504 505 506
		/* 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,
507 508
				 total_compressed, pages, nr_pages_ret,
				 compress_type);
509

510
		if (start + num_bytes < end) {
511 512 513 514 515 516
			start += num_bytes;
			pages = NULL;
			cond_resched();
			goto again;
		}
	} else {
517
cleanup_and_bail_uncompressed:
518 519 520 521 522 523 524 525 526 527 528 529
		/*
		 * 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 */
		}
530 531
		add_async_extent(async_cow, start, end - start + 1,
				 0, NULL, 0, BTRFS_COMPRESS_NONE);
532 533
		*num_added += 1;
	}
534

535 536 537 538 539 540 541 542
out:
	return 0;

free_pages_out:
	for (i = 0; i < nr_pages_ret; i++) {
		WARN_ON(pages[i]->mapping);
		page_cache_release(pages[i]);
	}
C
Chris Mason 已提交
543
	kfree(pages);
544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564

	goto out;
}

/*
 * phase two of compressed writeback.  This is the ordered portion
 * of the code, which only gets called in the order the work was
 * queued.  We walk all the async extents created by compress_file_range
 * and send them down to the disk.
 */
static noinline int submit_compressed_extents(struct inode *inode,
					      struct async_cow *async_cow)
{
	struct async_extent *async_extent;
	u64 alloc_hint = 0;
	struct btrfs_trans_handle *trans;
	struct btrfs_key ins;
	struct extent_map *em;
	struct btrfs_root *root = BTRFS_I(inode)->root;
	struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
	struct extent_io_tree *io_tree;
565
	int ret = 0;
566 567 568 569 570

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


C
Chris Mason 已提交
571
	while (!list_empty(&async_cow->extents)) {
572 573 574
		async_extent = list_entry(async_cow->extents.next,
					  struct async_extent, list);
		list_del(&async_extent->list);
C
Chris Mason 已提交
575

576 577
		io_tree = &BTRFS_I(inode)->io_tree;

578
retry:
579 580 581 582 583 584
		/* 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,
585 586
					 async_extent->start +
					 async_extent->ram_size - 1, GFP_NOFS);
587 588

			/* allocate blocks */
589 590 591 592 593
			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);
594 595 596 597 598 599 600

			/*
			 * 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.
			 */
601
			if (!page_started && !ret)
602 603
				extent_write_locked_range(io_tree,
						  inode, async_extent->start,
C
Chris Mason 已提交
604
						  async_extent->start +
605 606 607 608 609 610 611 612 613 614 615 616
						  async_extent->ram_size - 1,
						  btrfs_get_extent,
						  WB_SYNC_ALL);
			kfree(async_extent);
			cond_resched();
			continue;
		}

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

617
		trans = btrfs_join_transaction(root, 1);
618
		BUG_ON(IS_ERR(trans));
619 620 621 622 623
		ret = btrfs_reserve_extent(trans, root,
					   async_extent->compressed_size,
					   async_extent->compressed_size,
					   0, alloc_hint,
					   (u64)-1, &ins, 1);
624 625
		btrfs_end_transaction(trans, root);

626 627 628 629 630 631 632 633 634 635 636 637 638 639 640
		if (ret) {
			int i;
			for (i = 0; i < async_extent->nr_pages; i++) {
				WARN_ON(async_extent->pages[i]->mapping);
				page_cache_release(async_extent->pages[i]);
			}
			kfree(async_extent->pages);
			async_extent->nr_pages = 0;
			async_extent->pages = NULL;
			unlock_extent(io_tree, async_extent->start,
				      async_extent->start +
				      async_extent->ram_size - 1, GFP_NOFS);
			goto retry;
		}

641 642 643 644 645 646 647 648
		/*
		 * 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);

649
		em = alloc_extent_map(GFP_NOFS);
650
		BUG_ON(!em);
651 652
		em->start = async_extent->start;
		em->len = async_extent->ram_size;
653
		em->orig_start = em->start;
C
Chris Mason 已提交
654

655 656 657
		em->block_start = ins.objectid;
		em->block_len = ins.offset;
		em->bdev = root->fs_info->fs_devices->latest_bdev;
658
		em->compress_type = async_extent->compress_type;
659 660 661
		set_bit(EXTENT_FLAG_PINNED, &em->flags);
		set_bit(EXTENT_FLAG_COMPRESSED, &em->flags);

C
Chris Mason 已提交
662
		while (1) {
663
			write_lock(&em_tree->lock);
664
			ret = add_extent_mapping(em_tree, em);
665
			write_unlock(&em_tree->lock);
666 667 668 669 670 671 672 673 674
			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);
		}

675 676 677 678 679 680 681
		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);
682 683 684 685 686 687
		BUG_ON(ret);

		/*
		 * clear dirty, set writeback and unlock the pages.
		 */
		extent_clear_unlock_delalloc(inode,
688 689 690 691 692 693
				&BTRFS_I(inode)->io_tree,
				async_extent->start,
				async_extent->start +
				async_extent->ram_size - 1,
				NULL, EXTENT_CLEAR_UNLOCK_PAGE |
				EXTENT_CLEAR_UNLOCK |
694
				EXTENT_CLEAR_DELALLOC |
695
				EXTENT_CLEAR_DIRTY | EXTENT_SET_WRITEBACK);
696 697

		ret = btrfs_submit_compressed_write(inode,
C
Chris Mason 已提交
698 699 700 701 702
				    async_extent->start,
				    async_extent->ram_size,
				    ins.objectid,
				    ins.offset, async_extent->pages,
				    async_extent->nr_pages);
703 704 705 706 707 708 709 710 711 712

		BUG_ON(ret);
		alloc_hint = ins.objectid + ins.offset;
		kfree(async_extent);
		cond_resched();
	}

	return 0;
}

713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732 733 734 735 736 737 738 739 740 741 742 743 744
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;
}

745 746 747 748 749 750 751 752 753 754 755 756 757 758 759 760 761 762 763 764 765 766 767 768 769 770 771 772 773 774 775 776
/*
 * when extent_io.c finds a delayed allocation range in the file,
 * the call backs end up in this code.  The basic idea is to
 * allocate extents on disk for the range, and create ordered data structs
 * in ram to track those extents.
 *
 * locked_page is the page that writepage had locked already.  We use
 * it to make sure we don't do extra locks or unlocks.
 *
 * *page_started is set to one if we unlock locked_page and do everything
 * required to start IO on it.  It may be clean and already done with
 * IO when we return.
 */
static noinline int cow_file_range(struct inode *inode,
				   struct page *locked_page,
				   u64 start, u64 end, int *page_started,
				   unsigned long *nr_written,
				   int unlock)
{
	struct btrfs_root *root = BTRFS_I(inode)->root;
	struct btrfs_trans_handle *trans;
	u64 alloc_hint = 0;
	u64 num_bytes;
	unsigned long ram_size;
	u64 disk_num_bytes;
	u64 cur_alloc_size;
	u64 blocksize = root->sectorsize;
	struct btrfs_key ins;
	struct extent_map *em;
	struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
	int ret = 0;

J
Josef Bacik 已提交
777
	BUG_ON(root == root->fs_info->tree_root);
778
	trans = btrfs_join_transaction(root, 1);
779
	BUG_ON(IS_ERR(trans));
780
	btrfs_set_trans_block_group(trans, inode);
781
	trans->block_rsv = &root->fs_info->delalloc_block_rsv;
782 783 784 785 786 787 788 789 790 791 792 793

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

	if (start == 0) {
		/* lets try to make an inline extent */
		ret = cow_file_range_inline(trans, root, inode,
					    start, end, 0, NULL);
		if (ret == 0) {
			extent_clear_unlock_delalloc(inode,
794 795 796 797 798 799 800 801
				     &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);
802

803 804 805 806 807 808 809 810 811 812 813
			*nr_written = *nr_written +
			     (end - start + PAGE_CACHE_SIZE) / PAGE_CACHE_SIZE;
			*page_started = 1;
			ret = 0;
			goto out;
		}
	}

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

814
	alloc_hint = get_extent_allocation_hint(inode, start, num_bytes);
815 816
	btrfs_drop_extent_cache(inode, start, start + num_bytes - 1, 0);

C
Chris Mason 已提交
817
	while (disk_num_bytes > 0) {
818 819
		unsigned long op;

820
		cur_alloc_size = disk_num_bytes;
821
		ret = btrfs_reserve_extent(trans, root, cur_alloc_size,
822
					   root->sectorsize, 0, alloc_hint,
823
					   (u64)-1, &ins, 1);
C
Chris Mason 已提交
824 825
		BUG_ON(ret);

826
		em = alloc_extent_map(GFP_NOFS);
827
		BUG_ON(!em);
828
		em->start = start;
829
		em->orig_start = em->start;
830 831
		ram_size = ins.offset;
		em->len = ins.offset;
C
Chris Mason 已提交
832

833
		em->block_start = ins.objectid;
C
Chris Mason 已提交
834
		em->block_len = ins.offset;
835
		em->bdev = root->fs_info->fs_devices->latest_bdev;
836
		set_bit(EXTENT_FLAG_PINNED, &em->flags);
C
Chris Mason 已提交
837

C
Chris Mason 已提交
838
		while (1) {
839
			write_lock(&em_tree->lock);
840
			ret = add_extent_mapping(em_tree, em);
841
			write_unlock(&em_tree->lock);
842 843 844 845 846
			if (ret != -EEXIST) {
				free_extent_map(em);
				break;
			}
			btrfs_drop_extent_cache(inode, start,
C
Chris Mason 已提交
847
						start + ram_size - 1, 0);
848 849
		}

850
		cur_alloc_size = ins.offset;
851
		ret = btrfs_add_ordered_extent(inode, start, ins.objectid,
852
					       ram_size, cur_alloc_size, 0);
853
		BUG_ON(ret);
C
Chris Mason 已提交
854

855 856 857 858 859 860 861
		if (root->root_key.objectid ==
		    BTRFS_DATA_RELOC_TREE_OBJECTID) {
			ret = btrfs_reloc_clone_csums(inode, start,
						      cur_alloc_size);
			BUG_ON(ret);
		}

C
Chris Mason 已提交
862
		if (disk_num_bytes < cur_alloc_size)
863
			break;
C
Chris Mason 已提交
864

C
Chris Mason 已提交
865 866 867
		/* 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
868 869 870
		 *
		 * Do set the Private2 bit so we know this page was properly
		 * setup for writepage
C
Chris Mason 已提交
871
		 */
872 873 874 875
		op = unlock ? EXTENT_CLEAR_UNLOCK_PAGE : 0;
		op |= EXTENT_CLEAR_UNLOCK | EXTENT_CLEAR_DELALLOC |
			EXTENT_SET_PRIVATE2;

C
Chris Mason 已提交
876 877
		extent_clear_unlock_delalloc(inode, &BTRFS_I(inode)->io_tree,
					     start, start + ram_size - 1,
878
					     locked_page, op);
C
Chris Mason 已提交
879
		disk_num_bytes -= cur_alloc_size;
880 881 882
		num_bytes -= cur_alloc_size;
		alloc_hint = ins.objectid + ins.offset;
		start += cur_alloc_size;
883 884
	}
out:
885
	ret = 0;
886
	btrfs_end_transaction(trans, root);
C
Chris Mason 已提交
887

888
	return ret;
889
}
C
Chris Mason 已提交
890

891 892 893 894 895 896 897 898 899 900 901 902 903 904 905 906 907 908 909 910 911 912 913 914 915 916 917 918 919 920 921 922 923 924 925 926 927 928
/*
 * work queue call back to started compression on a file and pages
 */
static noinline void async_cow_start(struct btrfs_work *work)
{
	struct async_cow *async_cow;
	int num_added = 0;
	async_cow = container_of(work, struct async_cow, work);

	compress_file_range(async_cow->inode, async_cow->locked_page,
			    async_cow->start, async_cow->end, async_cow,
			    &num_added);
	if (num_added == 0)
		async_cow->inode = NULL;
}

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

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

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

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

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

C
Chris Mason 已提交
929
	if (async_cow->inode)
930 931
		submit_compressed_extents(async_cow->inode, async_cow);
}
C
Chris Mason 已提交
932

933 934 935 936 937 938 939 940 941 942 943 944 945 946 947 948 949
static noinline void async_cow_free(struct btrfs_work *work)
{
	struct async_cow *async_cow;
	async_cow = container_of(work, struct async_cow, work);
	kfree(async_cow);
}

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

950 951
	clear_extent_bit(&BTRFS_I(inode)->io_tree, start, end, EXTENT_LOCKED,
			 1, 0, NULL, GFP_NOFS);
C
Chris Mason 已提交
952
	while (start < end) {
953 954 955 956 957 958
		async_cow = kmalloc(sizeof(*async_cow), GFP_NOFS);
		async_cow->inode = inode;
		async_cow->root = root;
		async_cow->locked_page = locked_page;
		async_cow->start = start;

959
		if (BTRFS_I(inode)->flags & BTRFS_INODE_NOCOMPRESS)
960 961 962 963 964 965 966 967 968 969 970 971 972 973 974 975 976 977 978 979 980 981 982 983 984
			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
985
		while (atomic_read(&root->fs_info->async_submit_draining) &&
986 987 988 989 990 991 992 993 994 995 996
		      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;
997 998
}

C
Chris Mason 已提交
999
static noinline int csum_exist_in_range(struct btrfs_root *root,
1000 1001 1002 1003 1004 1005
					u64 bytenr, u64 num_bytes)
{
	int ret;
	struct btrfs_ordered_sum *sums;
	LIST_HEAD(list);

Y
Yan Zheng 已提交
1006 1007
	ret = btrfs_lookup_csums_range(root->fs_info->csum_root, bytenr,
				       bytenr + num_bytes - 1, &list);
1008 1009 1010 1011 1012 1013 1014 1015 1016 1017 1018
	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
1019 1020 1021 1022 1023 1024 1025
/*
 * 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
 */
1026 1027
static noinline int run_delalloc_nocow(struct inode *inode,
				       struct page *locked_page,
1028 1029
			      u64 start, u64 end, int *page_started, int force,
			      unsigned long *nr_written)
1030 1031
{
	struct btrfs_root *root = BTRFS_I(inode)->root;
1032
	struct btrfs_trans_handle *trans;
1033 1034
	struct extent_buffer *leaf;
	struct btrfs_path *path;
Y
Yan Zheng 已提交
1035
	struct btrfs_file_extent_item *fi;
1036
	struct btrfs_key found_key;
Y
Yan Zheng 已提交
1037 1038 1039
	u64 cow_start;
	u64 cur_offset;
	u64 extent_end;
1040
	u64 extent_offset;
Y
Yan Zheng 已提交
1041 1042 1043 1044
	u64 disk_bytenr;
	u64 num_bytes;
	int extent_type;
	int ret;
Y
Yan Zheng 已提交
1045
	int type;
Y
Yan Zheng 已提交
1046 1047
	int nocow;
	int check_prev = 1;
J
Josef Bacik 已提交
1048
	bool nolock = false;
1049 1050 1051

	path = btrfs_alloc_path();
	BUG_ON(!path);
J
Josef Bacik 已提交
1052 1053 1054 1055 1056 1057
	if (root == root->fs_info->tree_root) {
		nolock = true;
		trans = btrfs_join_transaction_nolock(root, 1);
	} else {
		trans = btrfs_join_transaction(root, 1);
	}
1058
	BUG_ON(IS_ERR(trans));
1059

Y
Yan Zheng 已提交
1060 1061 1062 1063 1064 1065 1066 1067 1068 1069 1070 1071 1072 1073 1074 1075 1076 1077 1078 1079 1080 1081 1082 1083 1084
	cow_start = (u64)-1;
	cur_offset = start;
	while (1) {
		ret = btrfs_lookup_file_extent(trans, root, path, inode->i_ino,
					       cur_offset, 0);
		BUG_ON(ret < 0);
		if (ret > 0 && path->slots[0] > 0 && check_prev) {
			leaf = path->nodes[0];
			btrfs_item_key_to_cpu(leaf, &found_key,
					      path->slots[0] - 1);
			if (found_key.objectid == inode->i_ino &&
			    found_key.type == BTRFS_EXTENT_DATA_KEY)
				path->slots[0]--;
		}
		check_prev = 0;
next_slot:
		leaf = path->nodes[0];
		if (path->slots[0] >= btrfs_header_nritems(leaf)) {
			ret = btrfs_next_leaf(root, path);
			if (ret < 0)
				BUG_ON(1);
			if (ret > 0)
				break;
			leaf = path->nodes[0];
		}
1085

Y
Yan Zheng 已提交
1086 1087
		nocow = 0;
		disk_bytenr = 0;
1088
		num_bytes = 0;
Y
Yan Zheng 已提交
1089 1090 1091 1092 1093 1094 1095 1096 1097
		btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);

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

		if (found_key.offset > cur_offset) {
			extent_end = found_key.offset;
1098
			extent_type = 0;
Y
Yan Zheng 已提交
1099 1100 1101 1102 1103 1104 1105
			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
1106 1107
		if (extent_type == BTRFS_FILE_EXTENT_REG ||
		    extent_type == BTRFS_FILE_EXTENT_PREALLOC) {
Y
Yan Zheng 已提交
1108
			disk_bytenr = btrfs_file_extent_disk_bytenr(leaf, fi);
1109
			extent_offset = btrfs_file_extent_offset(leaf, fi);
Y
Yan Zheng 已提交
1110 1111 1112 1113 1114 1115
			extent_end = found_key.offset +
				btrfs_file_extent_num_bytes(leaf, fi);
			if (extent_end <= start) {
				path->slots[0]++;
				goto next_slot;
			}
1116 1117
			if (disk_bytenr == 0)
				goto out_check;
Y
Yan Zheng 已提交
1118 1119 1120 1121
			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
1122 1123
			if (extent_type == BTRFS_FILE_EXTENT_REG && !force)
				goto out_check;
1124
			if (btrfs_extent_readonly(root, disk_bytenr))
Y
Yan Zheng 已提交
1125
				goto out_check;
1126
			if (btrfs_cross_ref_exist(trans, root, inode->i_ino,
1127 1128
						  found_key.offset -
						  extent_offset, disk_bytenr))
1129
				goto out_check;
1130
			disk_bytenr += extent_offset;
1131 1132 1133 1134 1135 1136 1137 1138 1139
			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
1140 1141 1142 1143 1144 1145 1146 1147 1148 1149 1150 1151 1152 1153 1154 1155 1156 1157 1158 1159 1160
			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;
1161 1162 1163
		}

		btrfs_release_path(root, path);
Y
Yan Zheng 已提交
1164 1165
		if (cow_start != (u64)-1) {
			ret = cow_file_range(inode, locked_page, cow_start,
1166 1167
					found_key.offset - 1, page_started,
					nr_written, 1);
Y
Yan Zheng 已提交
1168 1169
			BUG_ON(ret);
			cow_start = (u64)-1;
1170
		}
Y
Yan Zheng 已提交
1171

Y
Yan Zheng 已提交
1172 1173 1174 1175 1176
		if (extent_type == BTRFS_FILE_EXTENT_PREALLOC) {
			struct extent_map *em;
			struct extent_map_tree *em_tree;
			em_tree = &BTRFS_I(inode)->extent_tree;
			em = alloc_extent_map(GFP_NOFS);
1177
			BUG_ON(!em);
Y
Yan Zheng 已提交
1178
			em->start = cur_offset;
1179
			em->orig_start = em->start;
Y
Yan Zheng 已提交
1180 1181 1182 1183 1184 1185
			em->len = num_bytes;
			em->block_len = num_bytes;
			em->block_start = disk_bytenr;
			em->bdev = root->fs_info->fs_devices->latest_bdev;
			set_bit(EXTENT_FLAG_PINNED, &em->flags);
			while (1) {
1186
				write_lock(&em_tree->lock);
Y
Yan Zheng 已提交
1187
				ret = add_extent_mapping(em_tree, em);
1188
				write_unlock(&em_tree->lock);
Y
Yan Zheng 已提交
1189 1190 1191 1192 1193 1194 1195 1196 1197 1198 1199
				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
1200 1201

		ret = btrfs_add_ordered_extent(inode, cur_offset, disk_bytenr,
Y
Yan Zheng 已提交
1202 1203
					       num_bytes, num_bytes, type);
		BUG_ON(ret);
1204

1205 1206 1207 1208 1209 1210 1211
		if (root->root_key.objectid ==
		    BTRFS_DATA_RELOC_TREE_OBJECTID) {
			ret = btrfs_reloc_clone_csums(inode, cur_offset,
						      num_bytes);
			BUG_ON(ret);
		}

Y
Yan Zheng 已提交
1212
		extent_clear_unlock_delalloc(inode, &BTRFS_I(inode)->io_tree,
1213 1214 1215 1216
				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
1217 1218 1219
		cur_offset = extent_end;
		if (cur_offset > end)
			break;
1220
	}
Y
Yan Zheng 已提交
1221 1222 1223 1224 1225 1226
	btrfs_release_path(root, path);

	if (cur_offset <= end && cow_start == (u64)-1)
		cow_start = cur_offset;
	if (cow_start != (u64)-1) {
		ret = cow_file_range(inode, locked_page, cow_start, end,
1227
				     page_started, nr_written, 1);
Y
Yan Zheng 已提交
1228 1229 1230
		BUG_ON(ret);
	}

J
Josef Bacik 已提交
1231 1232 1233 1234 1235 1236 1237
	if (nolock) {
		ret = btrfs_end_transaction_nolock(trans, root);
		BUG_ON(ret);
	} else {
		ret = btrfs_end_transaction(trans, root);
		BUG_ON(ret);
	}
1238
	btrfs_free_path(path);
Y
Yan Zheng 已提交
1239
	return 0;
1240 1241
}

C
Chris Mason 已提交
1242 1243 1244
/*
 * extent_io.c call back to do delayed allocation processing
 */
C
Chris Mason 已提交
1245
static int run_delalloc_range(struct inode *inode, struct page *locked_page,
1246 1247
			      u64 start, u64 end, int *page_started,
			      unsigned long *nr_written)
1248 1249
{
	int ret;
1250
	struct btrfs_root *root = BTRFS_I(inode)->root;
1251

1252
	if (BTRFS_I(inode)->flags & BTRFS_INODE_NODATACOW)
C
Chris Mason 已提交
1253
		ret = run_delalloc_nocow(inode, locked_page, start, end,
C
Chris Mason 已提交
1254
					 page_started, 1, nr_written);
1255
	else if (BTRFS_I(inode)->flags & BTRFS_INODE_PREALLOC)
Y
Yan Zheng 已提交
1256
		ret = run_delalloc_nocow(inode, locked_page, start, end,
C
Chris Mason 已提交
1257
					 page_started, 0, nr_written);
C
Chris Mason 已提交
1258 1259
	else if (!btrfs_test_opt(root, COMPRESS) &&
		 !(BTRFS_I(inode)->force_compress))
1260 1261
		ret = cow_file_range(inode, locked_page, start, end,
				      page_started, nr_written, 1);
1262
	else
1263
		ret = cow_file_range_async(inode, locked_page, start, end,
C
Chris Mason 已提交
1264
					   page_started, nr_written);
1265 1266 1267
	return ret;
}

J
Josef Bacik 已提交
1268
static int btrfs_split_extent_hook(struct inode *inode,
1269
				   struct extent_state *orig, u64 split)
J
Josef Bacik 已提交
1270
{
1271
	/* not delalloc, ignore it */
J
Josef Bacik 已提交
1272 1273 1274
	if (!(orig->state & EXTENT_DELALLOC))
		return 0;

1275
	atomic_inc(&BTRFS_I(inode)->outstanding_extents);
J
Josef Bacik 已提交
1276 1277 1278 1279 1280 1281 1282 1283 1284 1285 1286 1287 1288 1289 1290 1291 1292
	return 0;
}

/*
 * 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.
 */
static int btrfs_merge_extent_hook(struct inode *inode,
				   struct extent_state *new,
				   struct extent_state *other)
{
	/* not delalloc, ignore it */
	if (!(other->state & EXTENT_DELALLOC))
		return 0;

1293
	atomic_dec(&BTRFS_I(inode)->outstanding_extents);
J
Josef Bacik 已提交
1294 1295 1296
	return 0;
}

C
Chris Mason 已提交
1297 1298 1299 1300 1301
/*
 * 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.
 */
1302 1303
static int btrfs_set_bit_hook(struct inode *inode,
			      struct extent_state *state, int *bits)
1304
{
J
Josef Bacik 已提交
1305

1306 1307 1308 1309 1310
	/*
	 * set_bit and clear bit hooks normally require _irqsave/restore
	 * but in this case, we are only testeing for the DELALLOC
	 * bit, which is only set or cleared with irqs on
	 */
1311
	if (!(state->state & EXTENT_DELALLOC) && (*bits & EXTENT_DELALLOC)) {
1312
		struct btrfs_root *root = BTRFS_I(inode)->root;
1313
		u64 len = state->end + 1 - state->start;
J
Josef Bacik 已提交
1314 1315
		int do_list = (root->root_key.objectid !=
			       BTRFS_ROOT_TREE_OBJECTID);
J
Josef Bacik 已提交
1316

1317 1318 1319 1320
		if (*bits & EXTENT_FIRST_DELALLOC)
			*bits &= ~EXTENT_FIRST_DELALLOC;
		else
			atomic_inc(&BTRFS_I(inode)->outstanding_extents);
1321

1322
		spin_lock(&root->fs_info->delalloc_lock);
1323 1324
		BTRFS_I(inode)->delalloc_bytes += len;
		root->fs_info->delalloc_bytes += len;
J
Josef Bacik 已提交
1325
		if (do_list && list_empty(&BTRFS_I(inode)->delalloc_inodes)) {
1326 1327 1328
			list_add_tail(&BTRFS_I(inode)->delalloc_inodes,
				      &root->fs_info->delalloc_inodes);
		}
1329
		spin_unlock(&root->fs_info->delalloc_lock);
1330 1331 1332 1333
	}
	return 0;
}

C
Chris Mason 已提交
1334 1335 1336
/*
 * extent_io.c clear_bit_hook, see set_bit_hook for why
 */
J
Josef Bacik 已提交
1337
static int btrfs_clear_bit_hook(struct inode *inode,
1338
				struct extent_state *state, int *bits)
1339
{
1340 1341 1342 1343 1344
	/*
	 * set_bit and clear bit hooks normally require _irqsave/restore
	 * but in this case, we are only testeing for the DELALLOC
	 * bit, which is only set or cleared with irqs on
	 */
1345
	if ((state->state & EXTENT_DELALLOC) && (*bits & EXTENT_DELALLOC)) {
1346
		struct btrfs_root *root = BTRFS_I(inode)->root;
1347
		u64 len = state->end + 1 - state->start;
J
Josef Bacik 已提交
1348 1349
		int do_list = (root->root_key.objectid !=
			       BTRFS_ROOT_TREE_OBJECTID);
1350

1351 1352 1353 1354 1355 1356 1357 1358
		if (*bits & EXTENT_FIRST_DELALLOC)
			*bits &= ~EXTENT_FIRST_DELALLOC;
		else if (!(*bits & EXTENT_DO_ACCOUNTING))
			atomic_dec(&BTRFS_I(inode)->outstanding_extents);

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

J
Josef Bacik 已提交
1359 1360
		if (root->root_key.objectid != BTRFS_DATA_RELOC_TREE_OBJECTID
		    && do_list)
1361
			btrfs_free_reserved_data_space(inode, len);
J
Josef Bacik 已提交
1362

1363
		spin_lock(&root->fs_info->delalloc_lock);
1364 1365 1366
		root->fs_info->delalloc_bytes -= len;
		BTRFS_I(inode)->delalloc_bytes -= len;

J
Josef Bacik 已提交
1367
		if (do_list && BTRFS_I(inode)->delalloc_bytes == 0 &&
1368 1369 1370
		    !list_empty(&BTRFS_I(inode)->delalloc_inodes)) {
			list_del_init(&BTRFS_I(inode)->delalloc_inodes);
		}
1371
		spin_unlock(&root->fs_info->delalloc_lock);
1372 1373 1374 1375
	}
	return 0;
}

C
Chris Mason 已提交
1376 1377 1378 1379
/*
 * 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
 */
1380
int btrfs_merge_bio_hook(struct page *page, unsigned long offset,
C
Chris Mason 已提交
1381 1382
			 size_t size, struct bio *bio,
			 unsigned long bio_flags)
1383 1384 1385
{
	struct btrfs_root *root = BTRFS_I(page->mapping->host)->root;
	struct btrfs_mapping_tree *map_tree;
1386
	u64 logical = (u64)bio->bi_sector << 9;
1387 1388 1389 1390
	u64 length = 0;
	u64 map_length;
	int ret;

1391 1392 1393
	if (bio_flags & EXTENT_BIO_COMPRESSED)
		return 0;

1394
	length = bio->bi_size;
1395 1396
	map_tree = &root->fs_info->mapping_tree;
	map_length = length;
1397
	ret = btrfs_map_block(map_tree, READ, logical,
1398
			      &map_length, NULL, 0);
1399

C
Chris Mason 已提交
1400
	if (map_length < length + size)
1401
		return 1;
1402
	return ret;
1403 1404
}

C
Chris Mason 已提交
1405 1406 1407 1408 1409 1410 1411 1412
/*
 * 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
1413 1414
static int __btrfs_submit_bio_start(struct inode *inode, int rw,
				    struct bio *bio, int mirror_num,
1415 1416
				    unsigned long bio_flags,
				    u64 bio_offset)
1417 1418 1419
{
	struct btrfs_root *root = BTRFS_I(inode)->root;
	int ret = 0;
1420

1421
	ret = btrfs_csum_one_bio(root, inode, bio, 0, 0);
1422
	BUG_ON(ret);
C
Chris Mason 已提交
1423 1424
	return 0;
}
1425

C
Chris Mason 已提交
1426 1427 1428 1429 1430 1431 1432 1433
/*
 * 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
 */
1434
static int __btrfs_submit_bio_done(struct inode *inode, int rw, struct bio *bio,
1435 1436
			  int mirror_num, unsigned long bio_flags,
			  u64 bio_offset)
C
Chris Mason 已提交
1437 1438
{
	struct btrfs_root *root = BTRFS_I(inode)->root;
1439
	return btrfs_map_bio(root, rw, bio, mirror_num, 1);
1440 1441
}

C
Chris Mason 已提交
1442
/*
1443 1444
 * 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
1445
 */
1446
static int btrfs_submit_bio_hook(struct inode *inode, int rw, struct bio *bio,
1447 1448
			  int mirror_num, unsigned long bio_flags,
			  u64 bio_offset)
1449 1450 1451
{
	struct btrfs_root *root = BTRFS_I(inode)->root;
	int ret = 0;
1452
	int skip_sum;
1453

1454
	skip_sum = BTRFS_I(inode)->flags & BTRFS_INODE_NODATASUM;
1455

J
Josef Bacik 已提交
1456 1457 1458 1459
	if (root == root->fs_info->tree_root)
		ret = btrfs_bio_wq_end_io(root->fs_info, bio, 2);
	else
		ret = btrfs_bio_wq_end_io(root->fs_info, bio, 0);
1460
	BUG_ON(ret);
1461

1462
	if (!(rw & REQ_WRITE)) {
1463
		if (bio_flags & EXTENT_BIO_COMPRESSED) {
C
Chris Mason 已提交
1464 1465
			return btrfs_submit_compressed_read(inode, bio,
						    mirror_num, bio_flags);
1466 1467
		} else if (!skip_sum)
			btrfs_lookup_bio_sums(root, inode, bio, NULL);
1468
		goto mapit;
1469
	} else if (!skip_sum) {
1470 1471 1472
		/* csum items have already been cloned */
		if (root->root_key.objectid == BTRFS_DATA_RELOC_TREE_OBJECTID)
			goto mapit;
1473 1474
		/* we're doing a write, do the async checksumming */
		return btrfs_wq_submit_bio(BTRFS_I(inode)->root->fs_info,
1475
				   inode, rw, bio, mirror_num,
1476 1477
				   bio_flags, bio_offset,
				   __btrfs_submit_bio_start,
C
Chris Mason 已提交
1478
				   __btrfs_submit_bio_done);
1479 1480
	}

1481
mapit:
1482
	return btrfs_map_bio(root, rw, bio, mirror_num, 0);
1483
}
C
Chris Mason 已提交
1484

C
Chris Mason 已提交
1485 1486 1487 1488
/*
 * 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.
 */
1489
static noinline int add_pending_csums(struct btrfs_trans_handle *trans,
1490 1491 1492 1493 1494 1495
			     struct inode *inode, u64 file_offset,
			     struct list_head *list)
{
	struct btrfs_ordered_sum *sum;

	btrfs_set_trans_block_group(trans, inode);
Q
Qinghuang Feng 已提交
1496 1497

	list_for_each_entry(sum, list, list) {
1498 1499
		btrfs_csum_file_blocks(trans,
		       BTRFS_I(inode)->root->fs_info->csum_root, sum);
1500 1501 1502 1503
	}
	return 0;
}

1504 1505
int btrfs_set_extent_delalloc(struct inode *inode, u64 start, u64 end,
			      struct extent_state **cached_state)
1506
{
C
Chris Mason 已提交
1507
	if ((end & (PAGE_CACHE_SIZE - 1)) == 0)
1508
		WARN_ON(1);
1509
	return set_extent_delalloc(&BTRFS_I(inode)->io_tree, start, end,
1510
				   cached_state, GFP_NOFS);
1511 1512
}

C
Chris Mason 已提交
1513
/* see btrfs_writepage_start_hook for details on why this is required */
1514 1515 1516 1517 1518
struct btrfs_writepage_fixup {
	struct page *page;
	struct btrfs_work work;
};

1519
static void btrfs_writepage_fixup_worker(struct btrfs_work *work)
1520 1521 1522
{
	struct btrfs_writepage_fixup *fixup;
	struct btrfs_ordered_extent *ordered;
1523
	struct extent_state *cached_state = NULL;
1524 1525 1526 1527 1528 1529 1530
	struct page *page;
	struct inode *inode;
	u64 page_start;
	u64 page_end;

	fixup = container_of(work, struct btrfs_writepage_fixup, work);
	page = fixup->page;
C
Chris Mason 已提交
1531
again:
1532 1533 1534 1535 1536 1537 1538 1539 1540 1541
	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;

1542 1543
	lock_extent_bits(&BTRFS_I(inode)->io_tree, page_start, page_end, 0,
			 &cached_state, GFP_NOFS);
C
Chris Mason 已提交
1544 1545

	/* already ordered? We're done */
1546
	if (PagePrivate2(page))
1547
		goto out;
C
Chris Mason 已提交
1548 1549 1550

	ordered = btrfs_lookup_ordered_extent(inode, page_start);
	if (ordered) {
1551 1552
		unlock_extent_cached(&BTRFS_I(inode)->io_tree, page_start,
				     page_end, &cached_state, GFP_NOFS);
C
Chris Mason 已提交
1553 1554 1555 1556
		unlock_page(page);
		btrfs_start_ordered_extent(inode, ordered, 1);
		goto again;
	}
1557

1558
	BUG();
1559
	btrfs_set_extent_delalloc(inode, page_start, page_end, &cached_state);
1560 1561
	ClearPageChecked(page);
out:
1562 1563
	unlock_extent_cached(&BTRFS_I(inode)->io_tree, page_start, page_end,
			     &cached_state, GFP_NOFS);
1564 1565 1566
out_page:
	unlock_page(page);
	page_cache_release(page);
1567
	kfree(fixup);
1568 1569 1570 1571 1572 1573 1574 1575
}

/*
 * 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
1576
 * In our case any range that doesn't have the ORDERED bit set
1577 1578 1579 1580
 * 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.
 */
1581
static int btrfs_writepage_start_hook(struct page *page, u64 start, u64 end)
1582 1583 1584 1585 1586
{
	struct inode *inode = page->mapping->host;
	struct btrfs_writepage_fixup *fixup;
	struct btrfs_root *root = BTRFS_I(inode)->root;

1587 1588
	/* this page is properly in the ordered list */
	if (TestClearPagePrivate2(page))
1589 1590 1591 1592 1593 1594 1595 1596
		return 0;

	if (PageChecked(page))
		return -EAGAIN;

	fixup = kzalloc(sizeof(*fixup), GFP_NOFS);
	if (!fixup)
		return -EAGAIN;
1597

1598 1599 1600 1601 1602 1603 1604 1605
	SetPageChecked(page);
	page_cache_get(page);
	fixup->work.func = btrfs_writepage_fixup_worker;
	fixup->page = page;
	btrfs_queue_worker(&root->fs_info->fixup_workers, &fixup->work);
	return -EAGAIN;
}

Y
Yan Zheng 已提交
1606 1607 1608 1609 1610 1611 1612 1613 1614 1615 1616 1617 1618 1619 1620 1621 1622 1623
static int insert_reserved_file_extent(struct btrfs_trans_handle *trans,
				       struct inode *inode, u64 file_pos,
				       u64 disk_bytenr, u64 disk_num_bytes,
				       u64 num_bytes, u64 ram_bytes,
				       u8 compression, u8 encryption,
				       u16 other_encoding, int extent_type)
{
	struct btrfs_root *root = BTRFS_I(inode)->root;
	struct btrfs_file_extent_item *fi;
	struct btrfs_path *path;
	struct extent_buffer *leaf;
	struct btrfs_key ins;
	u64 hint;
	int ret;

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

1624
	path->leave_spinning = 1;
C
Chris Mason 已提交
1625 1626 1627 1628 1629 1630 1631 1632 1633 1634

	/*
	 * we may be replacing one extent in the tree with another.
	 * The new extent is pinned in the extent map, and we don't want
	 * to drop it from the cache until it is completely in the btree.
	 *
	 * So, tell btrfs_drop_extents to leave this extent in the cache.
	 * the caller is expected to unpin it and allow it to be merged
	 * with the others.
	 */
Y
Yan, Zheng 已提交
1635 1636
	ret = btrfs_drop_extents(trans, inode, file_pos, file_pos + num_bytes,
				 &hint, 0);
Y
Yan Zheng 已提交
1637 1638 1639 1640 1641 1642 1643 1644 1645 1646 1647 1648 1649 1650 1651 1652 1653 1654 1655 1656
	BUG_ON(ret);

	ins.objectid = inode->i_ino;
	ins.offset = file_pos;
	ins.type = BTRFS_EXTENT_DATA_KEY;
	ret = btrfs_insert_empty_item(trans, root, path, &ins, sizeof(*fi));
	BUG_ON(ret);
	leaf = path->nodes[0];
	fi = btrfs_item_ptr(leaf, path->slots[0],
			    struct btrfs_file_extent_item);
	btrfs_set_file_extent_generation(leaf, fi, trans->transid);
	btrfs_set_file_extent_type(leaf, fi, extent_type);
	btrfs_set_file_extent_disk_bytenr(leaf, fi, disk_bytenr);
	btrfs_set_file_extent_disk_num_bytes(leaf, fi, disk_num_bytes);
	btrfs_set_file_extent_offset(leaf, fi, 0);
	btrfs_set_file_extent_num_bytes(leaf, fi, num_bytes);
	btrfs_set_file_extent_ram_bytes(leaf, fi, ram_bytes);
	btrfs_set_file_extent_compression(leaf, fi, compression);
	btrfs_set_file_extent_encryption(leaf, fi, encryption);
	btrfs_set_file_extent_other_encoding(leaf, fi, other_encoding);
1657 1658 1659 1660

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

Y
Yan Zheng 已提交
1661 1662 1663 1664 1665 1666 1667
	btrfs_mark_buffer_dirty(leaf);

	inode_add_bytes(inode, num_bytes);

	ins.objectid = disk_bytenr;
	ins.offset = disk_num_bytes;
	ins.type = BTRFS_EXTENT_ITEM_KEY;
1668 1669 1670
	ret = btrfs_alloc_reserved_file_extent(trans, root,
					root->root_key.objectid,
					inode->i_ino, file_pos, &ins);
Y
Yan Zheng 已提交
1671 1672
	BUG_ON(ret);
	btrfs_free_path(path);
1673

Y
Yan Zheng 已提交
1674 1675 1676
	return 0;
}

1677 1678 1679 1680 1681 1682
/*
 * 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
1683 1684 1685 1686
/* 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.
 */
1687
static int btrfs_finish_ordered_io(struct inode *inode, u64 start, u64 end)
1688 1689
{
	struct btrfs_root *root = BTRFS_I(inode)->root;
1690
	struct btrfs_trans_handle *trans = NULL;
1691
	struct btrfs_ordered_extent *ordered_extent = NULL;
1692
	struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
1693
	struct extent_state *cached_state = NULL;
1694
	int compress_type = 0;
1695
	int ret;
J
Josef Bacik 已提交
1696
	bool nolock = false;
1697

1698 1699
	ret = btrfs_dec_test_ordered_pending(inode, &ordered_extent, start,
					     end - start + 1);
1700
	if (!ret)
1701 1702
		return 0;
	BUG_ON(!ordered_extent);
1703

J
Josef Bacik 已提交
1704 1705
	nolock = (root == root->fs_info->tree_root);

1706 1707 1708 1709
	if (test_bit(BTRFS_ORDERED_NOCOW, &ordered_extent->flags)) {
		BUG_ON(!list_empty(&ordered_extent->list));
		ret = btrfs_ordered_update_i_size(inode, 0, ordered_extent);
		if (!ret) {
J
Josef Bacik 已提交
1710 1711 1712 1713
			if (nolock)
				trans = btrfs_join_transaction_nolock(root, 1);
			else
				trans = btrfs_join_transaction(root, 1);
1714
			BUG_ON(IS_ERR(trans));
1715 1716
			btrfs_set_trans_block_group(trans, inode);
			trans->block_rsv = &root->fs_info->delalloc_block_rsv;
1717 1718 1719 1720 1721
			ret = btrfs_update_inode(trans, root, inode);
			BUG_ON(ret);
		}
		goto out;
	}
1722

1723 1724 1725
	lock_extent_bits(io_tree, ordered_extent->file_offset,
			 ordered_extent->file_offset + ordered_extent->len - 1,
			 0, &cached_state, GFP_NOFS);
1726

J
Josef Bacik 已提交
1727 1728 1729 1730
	if (nolock)
		trans = btrfs_join_transaction_nolock(root, 1);
	else
		trans = btrfs_join_transaction(root, 1);
1731
	BUG_ON(IS_ERR(trans));
1732 1733
	btrfs_set_trans_block_group(trans, inode);
	trans->block_rsv = &root->fs_info->delalloc_block_rsv;
1734

C
Chris Mason 已提交
1735
	if (test_bit(BTRFS_ORDERED_COMPRESSED, &ordered_extent->flags))
1736
		compress_type = ordered_extent->compress_type;
Y
Yan Zheng 已提交
1737
	if (test_bit(BTRFS_ORDERED_PREALLOC, &ordered_extent->flags)) {
1738
		BUG_ON(compress_type);
Y
Yan, Zheng 已提交
1739
		ret = btrfs_mark_extent_written(trans, inode,
Y
Yan Zheng 已提交
1740 1741 1742 1743 1744
						ordered_extent->file_offset,
						ordered_extent->file_offset +
						ordered_extent->len);
		BUG_ON(ret);
	} else {
1745
		BUG_ON(root == root->fs_info->tree_root);
Y
Yan Zheng 已提交
1746 1747 1748 1749 1750 1751
		ret = insert_reserved_file_extent(trans, inode,
						ordered_extent->file_offset,
						ordered_extent->start,
						ordered_extent->disk_len,
						ordered_extent->len,
						ordered_extent->len,
1752
						compress_type, 0, 0,
Y
Yan Zheng 已提交
1753
						BTRFS_FILE_EXTENT_REG);
C
Chris Mason 已提交
1754 1755 1756
		unpin_extent_cache(&BTRFS_I(inode)->extent_tree,
				   ordered_extent->file_offset,
				   ordered_extent->len);
Y
Yan Zheng 已提交
1757 1758
		BUG_ON(ret);
	}
1759 1760 1761 1762
	unlock_extent_cached(io_tree, ordered_extent->file_offset,
			     ordered_extent->file_offset +
			     ordered_extent->len - 1, &cached_state, GFP_NOFS);

1763 1764 1765
	add_pending_csums(trans, inode, ordered_extent->file_offset,
			  &ordered_extent->list);

1766 1767 1768 1769
	btrfs_ordered_update_i_size(inode, 0, ordered_extent);
	ret = btrfs_update_inode(trans, root, inode);
	BUG_ON(ret);
out:
J
Josef Bacik 已提交
1770 1771 1772 1773 1774 1775 1776 1777 1778
	if (nolock) {
		if (trans)
			btrfs_end_transaction_nolock(trans, root);
	} else {
		btrfs_delalloc_release_metadata(inode, ordered_extent->len);
		if (trans)
			btrfs_end_transaction(trans, root);
	}

1779 1780 1781 1782 1783 1784 1785 1786
	/* once for us */
	btrfs_put_ordered_extent(ordered_extent);
	/* once for the tree */
	btrfs_put_ordered_extent(ordered_extent);

	return 0;
}

1787
static int btrfs_writepage_end_io_hook(struct page *page, u64 start, u64 end,
1788 1789
				struct extent_state *state, int uptodate)
{
1790
	ClearPagePrivate2(page);
1791 1792 1793
	return btrfs_finish_ordered_io(page->mapping->host, start, end);
}

C
Chris Mason 已提交
1794 1795 1796 1797 1798 1799 1800 1801
/*
 * When IO fails, either with EIO or csum verification fails, we
 * try other mirrors that might have a good copy of the data.  This
 * io_failure_record is used to record state as we go through all the
 * mirrors.  If another mirror has good data, the page is set up to date
 * and things continue.  If a good mirror can't be found, the original
 * bio end_io callback is called to indicate things have failed.
 */
1802 1803 1804 1805 1806
struct io_failure_record {
	struct page *page;
	u64 start;
	u64 len;
	u64 logical;
1807
	unsigned long bio_flags;
1808 1809 1810
	int last_mirror;
};

1811
static int btrfs_io_failed_hook(struct bio *failed_bio,
1812 1813
			 struct page *page, u64 start, u64 end,
			 struct extent_state *state)
1814 1815 1816 1817 1818 1819
{
	struct io_failure_record *failrec = NULL;
	u64 private;
	struct extent_map *em;
	struct inode *inode = page->mapping->host;
	struct extent_io_tree *failure_tree = &BTRFS_I(inode)->io_failure_tree;
1820
	struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
1821 1822 1823
	struct bio *bio;
	int num_copies;
	int ret;
1824
	int rw;
1825 1826 1827 1828 1829 1830 1831 1832 1833 1834
	u64 logical;

	ret = get_state_private(failure_tree, start, &private);
	if (ret) {
		failrec = kmalloc(sizeof(*failrec), GFP_NOFS);
		if (!failrec)
			return -ENOMEM;
		failrec->start = start;
		failrec->len = end - start + 1;
		failrec->last_mirror = 0;
1835
		failrec->bio_flags = 0;
1836

1837
		read_lock(&em_tree->lock);
1838 1839 1840 1841 1842
		em = lookup_extent_mapping(em_tree, start, failrec->len);
		if (em->start > start || em->start + em->len < start) {
			free_extent_map(em);
			em = NULL;
		}
1843
		read_unlock(&em_tree->lock);
1844 1845 1846 1847 1848 1849 1850

		if (!em || IS_ERR(em)) {
			kfree(failrec);
			return -EIO;
		}
		logical = start - em->start;
		logical = em->block_start + logical;
1851 1852 1853
		if (test_bit(EXTENT_FLAG_COMPRESSED, &em->flags)) {
			logical = em->block_start;
			failrec->bio_flags = EXTENT_BIO_COMPRESSED;
1854 1855
			extent_set_compress_type(&failrec->bio_flags,
						 em->compress_type);
1856
		}
1857 1858 1859 1860
		failrec->logical = logical;
		free_extent_map(em);
		set_extent_bits(failure_tree, start, end, EXTENT_LOCKED |
				EXTENT_DIRTY, GFP_NOFS);
1861 1862
		set_state_private(failure_tree, start,
				 (u64)(unsigned long)failrec);
1863
	} else {
1864
		failrec = (struct io_failure_record *)(unsigned long)private;
1865 1866 1867 1868 1869 1870
	}
	num_copies = btrfs_num_copies(
			      &BTRFS_I(inode)->root->fs_info->mapping_tree,
			      failrec->logical, failrec->len);
	failrec->last_mirror++;
	if (!state) {
1871
		spin_lock(&BTRFS_I(inode)->io_tree.lock);
1872 1873 1874 1875 1876
		state = find_first_extent_bit_state(&BTRFS_I(inode)->io_tree,
						    failrec->start,
						    EXTENT_LOCKED);
		if (state && state->start != failrec->start)
			state = NULL;
1877
		spin_unlock(&BTRFS_I(inode)->io_tree.lock);
1878 1879 1880 1881 1882 1883 1884 1885 1886 1887 1888 1889 1890 1891
	}
	if (!state || failrec->last_mirror > num_copies) {
		set_state_private(failure_tree, failrec->start, 0);
		clear_extent_bits(failure_tree, failrec->start,
				  failrec->start + failrec->len - 1,
				  EXTENT_LOCKED | EXTENT_DIRTY, GFP_NOFS);
		kfree(failrec);
		return -EIO;
	}
	bio = bio_alloc(GFP_NOFS, 1);
	bio->bi_private = state;
	bio->bi_end_io = failed_bio->bi_end_io;
	bio->bi_sector = failrec->logical >> 9;
	bio->bi_bdev = failed_bio->bi_bdev;
1892
	bio->bi_size = 0;
1893

1894
	bio_add_page(bio, page, failrec->len, start - page_offset(page));
1895
	if (failed_bio->bi_rw & REQ_WRITE)
1896 1897 1898 1899 1900
		rw = WRITE;
	else
		rw = READ;

	BTRFS_I(inode)->io_tree.ops->submit_bio_hook(inode, rw, bio,
C
Chris Mason 已提交
1901
						      failrec->last_mirror,
1902
						      failrec->bio_flags, 0);
1903 1904 1905
	return 0;
}

C
Chris Mason 已提交
1906 1907 1908 1909
/*
 * each time an IO finishes, we do a fast check in the IO failure tree
 * to see if we need to process or clean up an io_failure_record
 */
1910
static int btrfs_clean_io_failures(struct inode *inode, u64 start)
1911 1912 1913 1914 1915 1916 1917 1918
{
	u64 private;
	u64 private_failure;
	struct io_failure_record *failure;
	int ret;

	private = 0;
	if (count_range_bits(&BTRFS_I(inode)->io_failure_tree, &private,
1919
			     (u64)-1, 1, EXTENT_DIRTY, 0)) {
1920 1921 1922 1923 1924 1925 1926 1927 1928 1929 1930 1931 1932 1933 1934
		ret = get_state_private(&BTRFS_I(inode)->io_failure_tree,
					start, &private_failure);
		if (ret == 0) {
			failure = (struct io_failure_record *)(unsigned long)
				   private_failure;
			set_state_private(&BTRFS_I(inode)->io_failure_tree,
					  failure->start, 0);
			clear_extent_bits(&BTRFS_I(inode)->io_failure_tree,
					  failure->start,
					  failure->start + failure->len - 1,
					  EXTENT_DIRTY | EXTENT_LOCKED,
					  GFP_NOFS);
			kfree(failure);
		}
	}
1935 1936 1937
	return 0;
}

C
Chris Mason 已提交
1938 1939 1940 1941 1942
/*
 * when reads are done, we need to check csums to verify the data is correct
 * if there's a match, we allow the bio to finish.  If not, we go through
 * the io_failure_record routines to find good copies
 */
1943
static int btrfs_readpage_end_io_hook(struct page *page, u64 start, u64 end,
1944
			       struct extent_state *state)
1945
{
1946
	size_t offset = start - ((u64)page->index << PAGE_CACHE_SHIFT);
1947
	struct inode *inode = page->mapping->host;
1948
	struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
1949
	char *kaddr;
1950
	u64 private = ~(u32)0;
1951
	int ret;
1952 1953
	struct btrfs_root *root = BTRFS_I(inode)->root;
	u32 csum = ~(u32)0;
1954

1955 1956 1957 1958
	if (PageChecked(page)) {
		ClearPageChecked(page);
		goto good;
	}
1959 1960

	if (BTRFS_I(inode)->flags & BTRFS_INODE_NODATASUM)
1961 1962 1963
		return 0;

	if (root->root_key.objectid == BTRFS_DATA_RELOC_TREE_OBJECTID &&
1964
	    test_range_bit(io_tree, start, end, EXTENT_NODATASUM, 1, NULL)) {
1965 1966
		clear_extent_bits(io_tree, start, end, EXTENT_NODATASUM,
				  GFP_NOFS);
1967
		return 0;
1968
	}
1969

Y
Yan 已提交
1970
	if (state && state->start == start) {
1971 1972 1973 1974 1975
		private = state->private;
		ret = 0;
	} else {
		ret = get_state_private(io_tree, start, &private);
	}
1976
	kaddr = kmap_atomic(page, KM_USER0);
C
Chris Mason 已提交
1977
	if (ret)
1978
		goto zeroit;
C
Chris Mason 已提交
1979

1980 1981
	csum = btrfs_csum_data(root, kaddr + offset, csum,  end - start + 1);
	btrfs_csum_final(csum, (char *)&csum);
C
Chris Mason 已提交
1982
	if (csum != private)
1983
		goto zeroit;
C
Chris Mason 已提交
1984

1985
	kunmap_atomic(kaddr, KM_USER0);
1986
good:
1987 1988 1989
	/* if the io failure tree for this inode is non-empty,
	 * check to see if we've recovered from a failed IO
	 */
1990
	btrfs_clean_io_failures(inode, start);
1991 1992 1993
	return 0;

zeroit:
C
Chris Mason 已提交
1994 1995 1996 1997 1998 1999
	if (printk_ratelimit()) {
		printk(KERN_INFO "btrfs csum failed ino %lu off %llu csum %u "
		       "private %llu\n", page->mapping->host->i_ino,
		       (unsigned long long)start, csum,
		       (unsigned long long)private);
	}
2000 2001
	memset(kaddr + offset, 1, end - start + 1);
	flush_dcache_page(page);
2002
	kunmap_atomic(kaddr, KM_USER0);
2003 2004
	if (private == 0)
		return 0;
2005
	return -EIO;
2006
}
2007

Y
Yan, Zheng 已提交
2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 2020 2021 2022 2023 2024 2025 2026 2027 2028 2029 2030 2031 2032 2033 2034 2035 2036 2037 2038 2039 2040 2041 2042 2043 2044 2045 2046 2047 2048 2049 2050 2051 2052 2053 2054 2055
struct delayed_iput {
	struct list_head list;
	struct inode *inode;
};

void btrfs_add_delayed_iput(struct inode *inode)
{
	struct btrfs_fs_info *fs_info = BTRFS_I(inode)->root->fs_info;
	struct delayed_iput *delayed;

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

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

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

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

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

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

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

2056 2057 2058 2059 2060 2061 2062 2063 2064 2065 2066 2067 2068 2069 2070 2071 2072 2073 2074 2075 2076 2077 2078 2079 2080 2081 2082 2083 2084 2085 2086 2087 2088 2089 2090 2091 2092 2093 2094 2095 2096 2097 2098 2099 2100 2101 2102 2103 2104 2105 2106 2107 2108 2109 2110 2111 2112 2113 2114 2115 2116 2117 2118 2119 2120 2121 2122 2123 2124 2125 2126 2127 2128 2129 2130 2131 2132 2133 2134 2135 2136 2137 2138 2139 2140 2141 2142 2143 2144 2145 2146 2147 2148 2149 2150 2151 2152 2153 2154 2155 2156 2157 2158 2159 2160 2161 2162 2163 2164 2165 2166 2167 2168 2169 2170 2171 2172
/*
 * calculate extra metadata reservation when snapshotting a subvolume
 * contains orphan files.
 */
void btrfs_orphan_pre_snapshot(struct btrfs_trans_handle *trans,
				struct btrfs_pending_snapshot *pending,
				u64 *bytes_to_reserve)
{
	struct btrfs_root *root;
	struct btrfs_block_rsv *block_rsv;
	u64 num_bytes;
	int index;

	root = pending->root;
	if (!root->orphan_block_rsv || list_empty(&root->orphan_list))
		return;

	block_rsv = root->orphan_block_rsv;

	/* orphan block reservation for the snapshot */
	num_bytes = block_rsv->size;

	/*
	 * after the snapshot is created, COWing tree blocks may use more
	 * space than it frees. So we should make sure there is enough
	 * reserved space.
	 */
	index = trans->transid & 0x1;
	if (block_rsv->reserved + block_rsv->freed[index] < block_rsv->size) {
		num_bytes += block_rsv->size -
			     (block_rsv->reserved + block_rsv->freed[index]);
	}

	*bytes_to_reserve += num_bytes;
}

void btrfs_orphan_post_snapshot(struct btrfs_trans_handle *trans,
				struct btrfs_pending_snapshot *pending)
{
	struct btrfs_root *root = pending->root;
	struct btrfs_root *snap = pending->snap;
	struct btrfs_block_rsv *block_rsv;
	u64 num_bytes;
	int index;
	int ret;

	if (!root->orphan_block_rsv || list_empty(&root->orphan_list))
		return;

	/* refill source subvolume's orphan block reservation */
	block_rsv = root->orphan_block_rsv;
	index = trans->transid & 0x1;
	if (block_rsv->reserved + block_rsv->freed[index] < block_rsv->size) {
		num_bytes = block_rsv->size -
			    (block_rsv->reserved + block_rsv->freed[index]);
		ret = btrfs_block_rsv_migrate(&pending->block_rsv,
					      root->orphan_block_rsv,
					      num_bytes);
		BUG_ON(ret);
	}

	/* setup orphan block reservation for the snapshot */
	block_rsv = btrfs_alloc_block_rsv(snap);
	BUG_ON(!block_rsv);

	btrfs_add_durable_block_rsv(root->fs_info, block_rsv);
	snap->orphan_block_rsv = block_rsv;

	num_bytes = root->orphan_block_rsv->size;
	ret = btrfs_block_rsv_migrate(&pending->block_rsv,
				      block_rsv, num_bytes);
	BUG_ON(ret);

#if 0
	/* insert orphan item for the snapshot */
	WARN_ON(!root->orphan_item_inserted);
	ret = btrfs_insert_orphan_item(trans, root->fs_info->tree_root,
				       snap->root_key.objectid);
	BUG_ON(ret);
	snap->orphan_item_inserted = 1;
#endif
}

enum btrfs_orphan_cleanup_state {
	ORPHAN_CLEANUP_STARTED	= 1,
	ORPHAN_CLEANUP_DONE	= 2,
};

/*
 * This is called in transaction commmit time. If there are no orphan
 * 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)
{
	int ret;

	if (!list_empty(&root->orphan_list) ||
	    root->orphan_cleanup_state != ORPHAN_CLEANUP_DONE)
		return;

	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;
	}

	if (root->orphan_block_rsv) {
		WARN_ON(root->orphan_block_rsv->size > 0);
		btrfs_free_block_rsv(root, root->orphan_block_rsv);
		root->orphan_block_rsv = NULL;
	}
}

2173 2174 2175
/*
 * This creates an orphan entry for the given inode in case something goes
 * wrong in the middle of an unlink/truncate.
2176 2177 2178
 *
 * NOTE: caller of this function should reserve 5 units of metadata for
 *	 this function.
2179 2180 2181 2182
 */
int btrfs_orphan_add(struct btrfs_trans_handle *trans, struct inode *inode)
{
	struct btrfs_root *root = BTRFS_I(inode)->root;
2183 2184 2185 2186
	struct btrfs_block_rsv *block_rsv = NULL;
	int reserve = 0;
	int insert = 0;
	int ret;
2187

2188 2189 2190 2191
	if (!root->orphan_block_rsv) {
		block_rsv = btrfs_alloc_block_rsv(root);
		BUG_ON(!block_rsv);
	}
2192

2193 2194 2195 2196 2197 2198
	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;
2199 2200
	}

2201 2202 2203 2204 2205 2206 2207 2208 2209 2210 2211 2212 2213 2214 2215 2216
	if (list_empty(&BTRFS_I(inode)->i_orphan)) {
		list_add(&BTRFS_I(inode)->i_orphan, &root->orphan_list);
#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;
	} else {
		WARN_ON(!BTRFS_I(inode)->orphan_meta_reserved);
2217 2218
	}

2219 2220 2221 2222 2223
	if (!BTRFS_I(inode)->orphan_meta_reserved) {
		BTRFS_I(inode)->orphan_meta_reserved = 1;
		reserve = 1;
	}
	spin_unlock(&root->orphan_lock);
2224

2225 2226
	if (block_rsv)
		btrfs_add_durable_block_rsv(root->fs_info, block_rsv);
2227

2228 2229 2230 2231 2232
	/* grab metadata reservation from transaction handle */
	if (reserve) {
		ret = btrfs_orphan_reserve_metadata(trans, inode);
		BUG_ON(ret);
	}
2233

2234 2235 2236 2237 2238 2239 2240 2241 2242 2243 2244 2245 2246
	/* insert an orphan item to track this unlinked/truncated file */
	if (insert >= 1) {
		ret = btrfs_insert_orphan_item(trans, root, inode->i_ino);
		BUG_ON(ret);
	}

	/* 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);
		BUG_ON(ret);
	}
	return 0;
2247 2248 2249 2250 2251 2252 2253 2254 2255
}

/*
 * 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;
2256 2257
	int delete_item = 0;
	int release_rsv = 0;
2258 2259
	int ret = 0;

2260 2261 2262 2263
	spin_lock(&root->orphan_lock);
	if (!list_empty(&BTRFS_I(inode)->i_orphan)) {
		list_del_init(&BTRFS_I(inode)->i_orphan);
		delete_item = 1;
2264 2265
	}

2266 2267 2268
	if (BTRFS_I(inode)->orphan_meta_reserved) {
		BTRFS_I(inode)->orphan_meta_reserved = 0;
		release_rsv = 1;
2269
	}
2270
	spin_unlock(&root->orphan_lock);
2271

2272 2273 2274 2275
	if (trans && delete_item) {
		ret = btrfs_del_orphan_item(trans, root, inode->i_ino);
		BUG_ON(ret);
	}
2276

2277 2278
	if (release_rsv)
		btrfs_orphan_release_metadata(inode);
2279

2280
	return 0;
2281 2282 2283 2284 2285 2286
}

/*
 * this cleans up any orphans that may be left on the list from the last use
 * of this root.
 */
2287
int btrfs_orphan_cleanup(struct btrfs_root *root)
2288 2289 2290 2291 2292 2293 2294 2295
{
	struct btrfs_path *path;
	struct extent_buffer *leaf;
	struct btrfs_key key, found_key;
	struct btrfs_trans_handle *trans;
	struct inode *inode;
	int ret = 0, nr_unlink = 0, nr_truncate = 0;

2296
	if (cmpxchg(&root->orphan_cleanup_state, 0, ORPHAN_CLEANUP_STARTED))
2297
		return 0;
2298 2299

	path = btrfs_alloc_path();
2300 2301 2302 2303
	if (!path) {
		ret = -ENOMEM;
		goto out;
	}
2304 2305 2306 2307 2308 2309 2310 2311
	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);
2312 2313
		if (ret < 0)
			goto out;
2314 2315 2316 2317 2318 2319 2320

		/*
		 * if ret == 0 means we found what we were searching for, which
		 * is weird, but possible, so only screw with path if we didnt
		 * find the key and see if we have stuff that matches
		 */
		if (ret > 0) {
2321
			ret = 0;
2322 2323 2324 2325 2326 2327 2328 2329 2330 2331 2332 2333 2334 2335 2336 2337 2338 2339 2340 2341 2342 2343 2344
			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 */
		btrfs_release_path(root, path);

		/*
		 * this is where we are basically btrfs_lookup, without the
		 * crossing root thing.  we store the inode number in the
		 * offset of the orphan item.
		 */
2345 2346 2347
		found_key.objectid = found_key.offset;
		found_key.type = BTRFS_INODE_ITEM_KEY;
		found_key.offset = 0;
2348
		inode = btrfs_iget(root->fs_info->sb, &found_key, root, NULL);
2349 2350 2351 2352
		if (IS_ERR(inode)) {
			ret = PTR_ERR(inode);
			goto out;
		}
2353 2354 2355 2356 2357

		/*
		 * add this inode to the orphan list so btrfs_orphan_del does
		 * the proper thing when we hit it
		 */
2358
		spin_lock(&root->orphan_lock);
2359
		list_add(&BTRFS_I(inode)->i_orphan, &root->orphan_list);
2360
		spin_unlock(&root->orphan_lock);
2361 2362 2363 2364 2365 2366 2367 2368

		/*
		 * if this is a bad inode, means we actually succeeded in
		 * removing the inode, but not the orphan record, which means
		 * we need to manually delete the orphan since iput will just
		 * do a destroy_inode
		 */
		if (is_bad_inode(inode)) {
2369
			trans = btrfs_start_transaction(root, 0);
2370 2371 2372 2373
			if (IS_ERR(trans)) {
				ret = PTR_ERR(trans);
				goto out;
			}
2374
			btrfs_orphan_del(trans, inode);
2375
			btrfs_end_transaction(trans, root);
2376 2377 2378 2379 2380 2381
			iput(inode);
			continue;
		}

		/* if we have links, this was a truncate, lets do that */
		if (inode->i_nlink) {
2382 2383 2384 2385 2386
			if (!S_ISREG(inode->i_mode)) {
				WARN_ON(1);
				iput(inode);
				continue;
			}
2387
			nr_truncate++;
2388
			ret = btrfs_truncate(inode);
2389 2390 2391 2392 2393 2394
		} else {
			nr_unlink++;
		}

		/* this will do delete_inode and everything for us */
		iput(inode);
2395 2396
		if (ret)
			goto out;
2397
	}
2398 2399 2400 2401 2402 2403 2404 2405
	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) {
		trans = btrfs_join_transaction(root, 1);
2406 2407
		if (!IS_ERR(trans))
			btrfs_end_transaction(trans, root);
2408
	}
2409 2410 2411 2412 2413

	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);
2414 2415 2416 2417 2418 2419

out:
	if (ret)
		printk(KERN_CRIT "btrfs: could not do orphan cleanup %d\n", ret);
	btrfs_free_path(path);
	return ret;
2420 2421
}

2422 2423 2424 2425 2426 2427 2428 2429 2430 2431 2432 2433 2434 2435 2436 2437 2438 2439 2440 2441 2442 2443 2444 2445 2446 2447 2448 2449 2450 2451 2452 2453 2454 2455 2456 2457 2458 2459 2460 2461 2462 2463 2464 2465 2466 2467 2468 2469 2470 2471 2472
/*
 * 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
2473 2474 2475
/*
 * read an inode from the btree into the in-memory inode
 */
2476
static void btrfs_read_locked_inode(struct inode *inode)
C
Chris Mason 已提交
2477 2478
{
	struct btrfs_path *path;
2479
	struct extent_buffer *leaf;
C
Chris Mason 已提交
2480
	struct btrfs_inode_item *inode_item;
2481
	struct btrfs_timespec *tspec;
C
Chris Mason 已提交
2482 2483
	struct btrfs_root *root = BTRFS_I(inode)->root;
	struct btrfs_key location;
2484
	int maybe_acls;
C
Chris Mason 已提交
2485
	u64 alloc_group_block;
J
Josef Bacik 已提交
2486
	u32 rdev;
C
Chris Mason 已提交
2487 2488 2489 2490 2491
	int ret;

	path = btrfs_alloc_path();
	BUG_ON(!path);
	memcpy(&location, &BTRFS_I(inode)->location, sizeof(location));
C
Chris Mason 已提交
2492

C
Chris Mason 已提交
2493
	ret = btrfs_lookup_inode(NULL, root, path, &location, 0);
2494
	if (ret)
C
Chris Mason 已提交
2495 2496
		goto make_bad;

2497 2498 2499 2500 2501 2502 2503 2504
	leaf = path->nodes[0];
	inode_item = btrfs_item_ptr(leaf, path->slots[0],
				    struct btrfs_inode_item);

	inode->i_mode = btrfs_inode_mode(leaf, inode_item);
	inode->i_nlink = btrfs_inode_nlink(leaf, inode_item);
	inode->i_uid = btrfs_inode_uid(leaf, inode_item);
	inode->i_gid = btrfs_inode_gid(leaf, inode_item);
2505
	btrfs_i_size_write(inode, btrfs_inode_size(leaf, inode_item));
2506 2507 2508 2509 2510 2511 2512 2513 2514 2515 2516 2517 2518

	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);

2519
	inode_set_bytes(inode, btrfs_inode_nbytes(leaf, inode_item));
2520
	BTRFS_I(inode)->generation = btrfs_inode_generation(leaf, inode_item);
2521
	BTRFS_I(inode)->sequence = btrfs_inode_sequence(leaf, inode_item);
2522
	inode->i_generation = BTRFS_I(inode)->generation;
J
Josef Bacik 已提交
2523
	inode->i_rdev = 0;
2524 2525
	rdev = btrfs_inode_rdev(leaf, inode_item);

2526
	BTRFS_I(inode)->index_cnt = (u64)-1;
2527
	BTRFS_I(inode)->flags = btrfs_inode_flags(leaf, inode_item);
2528

2529
	alloc_group_block = btrfs_inode_block_group(leaf, inode_item);
2530

2531 2532 2533 2534 2535
	/*
	 * try to precache a NULL acl entry for files that don't have
	 * any xattrs or acls
	 */
	maybe_acls = acls_after_inode_item(leaf, path->slots[0], inode->i_ino);
2536 2537
	if (!maybe_acls)
		cache_no_acl(inode);
2538

2539 2540
	BTRFS_I(inode)->block_group = btrfs_find_block_group(root, 0,
						alloc_group_block, 0);
C
Chris Mason 已提交
2541 2542 2543 2544 2545 2546
	btrfs_free_path(path);
	inode_item = NULL;

	switch (inode->i_mode & S_IFMT) {
	case S_IFREG:
		inode->i_mapping->a_ops = &btrfs_aops;
C
Chris Mason 已提交
2547
		inode->i_mapping->backing_dev_info = &root->fs_info->bdi;
2548
		BTRFS_I(inode)->io_tree.ops = &btrfs_extent_io_ops;
C
Chris Mason 已提交
2549 2550 2551 2552 2553 2554 2555 2556 2557 2558 2559 2560 2561
		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
2562
		inode->i_mapping->backing_dev_info = &root->fs_info->bdi;
C
Chris Mason 已提交
2563
		break;
J
Josef Bacik 已提交
2564
	default:
J
Jim Owens 已提交
2565
		inode->i_op = &btrfs_special_inode_operations;
J
Josef Bacik 已提交
2566 2567
		init_special_inode(inode, inode->i_mode, rdev);
		break;
C
Chris Mason 已提交
2568
	}
2569 2570

	btrfs_update_iflags(inode);
C
Chris Mason 已提交
2571 2572 2573 2574 2575 2576 2577
	return;

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

C
Chris Mason 已提交
2578 2579 2580
/*
 * given a leaf and an inode, copy the inode fields into the leaf
 */
2581 2582
static void fill_inode_item(struct btrfs_trans_handle *trans,
			    struct extent_buffer *leaf,
2583
			    struct btrfs_inode_item *item,
C
Chris Mason 已提交
2584 2585
			    struct inode *inode)
{
2586 2587
	btrfs_set_inode_uid(leaf, item, inode->i_uid);
	btrfs_set_inode_gid(leaf, item, inode->i_gid);
2588
	btrfs_set_inode_size(leaf, item, BTRFS_I(inode)->disk_i_size);
2589 2590 2591 2592 2593 2594 2595 2596 2597 2598 2599 2600 2601 2602 2603 2604 2605 2606
	btrfs_set_inode_mode(leaf, item, inode->i_mode);
	btrfs_set_inode_nlink(leaf, item, inode->i_nlink);

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

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

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

2607
	btrfs_set_inode_nbytes(leaf, item, inode_get_bytes(inode));
2608
	btrfs_set_inode_generation(leaf, item, BTRFS_I(inode)->generation);
2609
	btrfs_set_inode_sequence(leaf, item, BTRFS_I(inode)->sequence);
2610
	btrfs_set_inode_transid(leaf, item, trans->transid);
2611
	btrfs_set_inode_rdev(leaf, item, inode->i_rdev);
Y
Yan 已提交
2612
	btrfs_set_inode_flags(leaf, item, BTRFS_I(inode)->flags);
2613
	btrfs_set_inode_block_group(leaf, item, BTRFS_I(inode)->block_group);
C
Chris Mason 已提交
2614 2615
}

C
Chris Mason 已提交
2616 2617 2618
/*
 * copy everything in the in-memory inode into the btree.
 */
C
Chris Mason 已提交
2619 2620
noinline int btrfs_update_inode(struct btrfs_trans_handle *trans,
				struct btrfs_root *root, struct inode *inode)
C
Chris Mason 已提交
2621 2622 2623
{
	struct btrfs_inode_item *inode_item;
	struct btrfs_path *path;
2624
	struct extent_buffer *leaf;
C
Chris Mason 已提交
2625 2626 2627 2628
	int ret;

	path = btrfs_alloc_path();
	BUG_ON(!path);
2629
	path->leave_spinning = 1;
C
Chris Mason 已提交
2630 2631 2632 2633 2634 2635 2636 2637
	ret = btrfs_lookup_inode(trans, root, path,
				 &BTRFS_I(inode)->location, 1);
	if (ret) {
		if (ret > 0)
			ret = -ENOENT;
		goto failed;
	}

2638
	btrfs_unlock_up_safe(path, 1);
2639 2640
	leaf = path->nodes[0];
	inode_item = btrfs_item_ptr(leaf, path->slots[0],
C
Chris Mason 已提交
2641 2642
				  struct btrfs_inode_item);

2643
	fill_inode_item(trans, leaf, inode_item, inode);
2644
	btrfs_mark_buffer_dirty(leaf);
2645
	btrfs_set_inode_last_trans(trans, inode);
C
Chris Mason 已提交
2646 2647 2648 2649 2650 2651 2652
	ret = 0;
failed:
	btrfs_free_path(path);
	return ret;
}


C
Chris Mason 已提交
2653 2654 2655 2656 2657
/*
 * 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
 */
2658 2659 2660 2661
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
2662 2663 2664
{
	struct btrfs_path *path;
	int ret = 0;
2665
	struct extent_buffer *leaf;
C
Chris Mason 已提交
2666
	struct btrfs_dir_item *di;
2667
	struct btrfs_key key;
2668
	u64 index;
C
Chris Mason 已提交
2669 2670

	path = btrfs_alloc_path();
2671 2672
	if (!path) {
		ret = -ENOMEM;
2673
		goto out;
2674 2675
	}

2676
	path->leave_spinning = 1;
C
Chris Mason 已提交
2677 2678 2679 2680 2681 2682 2683 2684 2685 2686
	di = btrfs_lookup_dir_item(trans, root, path, dir->i_ino,
				    name, name_len, -1);
	if (IS_ERR(di)) {
		ret = PTR_ERR(di);
		goto err;
	}
	if (!di) {
		ret = -ENOENT;
		goto err;
	}
2687 2688
	leaf = path->nodes[0];
	btrfs_dir_item_key_to_cpu(leaf, di, &key);
C
Chris Mason 已提交
2689
	ret = btrfs_delete_one_dir_name(trans, root, path, di);
2690 2691
	if (ret)
		goto err;
C
Chris Mason 已提交
2692 2693
	btrfs_release_path(root, path);

2694
	ret = btrfs_del_inode_ref(trans, root, name, name_len,
2695 2696
				  inode->i_ino,
				  dir->i_ino, &index);
2697
	if (ret) {
C
Chris Mason 已提交
2698
		printk(KERN_INFO "btrfs failed to delete reference to %.*s, "
2699
		       "inode %lu parent %lu\n", name_len, name,
2700
		       inode->i_ino, dir->i_ino);
2701 2702 2703
		goto err;
	}

C
Chris Mason 已提交
2704
	di = btrfs_lookup_dir_index_item(trans, root, path, dir->i_ino,
2705
					 index, name, name_len, -1);
C
Chris Mason 已提交
2706 2707 2708 2709 2710 2711 2712 2713 2714
	if (IS_ERR(di)) {
		ret = PTR_ERR(di);
		goto err;
	}
	if (!di) {
		ret = -ENOENT;
		goto err;
	}
	ret = btrfs_delete_one_dir_name(trans, root, path, di);
2715
	btrfs_release_path(root, path);
C
Chris Mason 已提交
2716

2717 2718
	ret = btrfs_del_inode_ref_in_log(trans, root, name, name_len,
					 inode, dir->i_ino);
C
Chris Mason 已提交
2719
	BUG_ON(ret != 0 && ret != -ENOENT);
2720 2721 2722

	ret = btrfs_del_dir_entries_in_log(trans, root, name, name_len,
					   dir, index);
2723 2724
	if (ret == -ENOENT)
		ret = 0;
C
Chris Mason 已提交
2725 2726
err:
	btrfs_free_path(path);
2727 2728 2729 2730 2731 2732 2733 2734 2735
	if (ret)
		goto out;

	btrfs_i_size_write(dir, dir->i_size - name_len * 2);
	inode->i_ctime = dir->i_mtime = dir->i_ctime = CURRENT_TIME;
	btrfs_update_inode(trans, root, dir);
	btrfs_drop_nlink(inode);
	ret = btrfs_update_inode(trans, root, inode);
out:
C
Chris Mason 已提交
2736 2737 2738
	return ret;
}

2739 2740 2741
/* 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
2742
{
2743 2744
	struct extent_buffer *eb;
	int level;
2745
	u64 refs = 1;
2746

2747
	for (level = 0; level < BTRFS_MAX_LEVEL; level++) {
2748 2749
		int ret;

2750 2751 2752 2753 2754 2755 2756 2757 2758
		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;
2759
	}
2760
	return 0;
C
Chris Mason 已提交
2761 2762
}

2763 2764 2765 2766 2767 2768 2769 2770 2771
/*
 * 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)
2772
{
C
Chris Mason 已提交
2773
	struct btrfs_trans_handle *trans;
2774
	struct btrfs_root *root = BTRFS_I(dir)->root;
2775
	struct btrfs_path *path;
2776
	struct btrfs_inode_ref *ref;
2777
	struct btrfs_dir_item *di;
2778
	struct inode *inode = dentry->d_inode;
2779
	u64 index;
2780 2781
	int check_link = 1;
	int err = -ENOSPC;
2782 2783
	int ret;

2784 2785 2786
	trans = btrfs_start_transaction(root, 10);
	if (!IS_ERR(trans) || PTR_ERR(trans) != -ENOSPC)
		return trans;
2787

2788 2789
	if (inode->i_ino == BTRFS_EMPTY_SUBVOL_DIR_OBJECTID)
		return ERR_PTR(-ENOSPC);
2790

2791 2792 2793
	/* check if there is someone else holds reference */
	if (S_ISDIR(inode->i_mode) && atomic_read(&inode->i_count) > 1)
		return ERR_PTR(-ENOSPC);
2794

2795 2796
	if (atomic_read(&inode->i_count) > 2)
		return ERR_PTR(-ENOSPC);
2797

2798 2799 2800 2801 2802 2803 2804
	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);
2805 2806
	}

2807
	trans = btrfs_start_transaction(root, 0);
2808
	if (IS_ERR(trans)) {
2809 2810 2811 2812
		btrfs_free_path(path);
		root->fs_info->enospc_unlink = 0;
		return trans;
	}
2813

2814 2815
	path->skip_locking = 1;
	path->search_commit_root = 1;
2816

2817 2818 2819 2820 2821 2822 2823 2824 2825 2826 2827
	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;
2828
	}
2829 2830 2831 2832 2833 2834 2835 2836 2837 2838 2839 2840 2841 2842 2843 2844 2845 2846 2847 2848 2849 2850 2851 2852 2853 2854 2855 2856 2857 2858 2859 2860 2861 2862 2863 2864 2865 2866 2867 2868 2869 2870 2871 2872 2873 2874 2875 2876 2877 2878 2879 2880 2881 2882 2883 2884 2885 2886 2887 2888 2889 2890 2891 2892 2893 2894 2895 2896 2897 2898 2899 2900 2901 2902 2903 2904 2905 2906 2907 2908 2909 2910 2911 2912 2913 2914 2915 2916 2917 2918 2919 2920 2921 2922 2923 2924 2925 2926 2927 2928 2929 2930 2931 2932 2933 2934
	btrfs_release_path(root, path);

	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;
	}
	btrfs_release_path(root, path);

	if (ret == 0 && S_ISREG(inode->i_mode)) {
		ret = btrfs_lookup_file_extent(trans, root, path,
					       inode->i_ino, (u64)-1, 0);
		if (ret < 0) {
			err = ret;
			goto out;
		}
		BUG_ON(ret == 0);
		if (check_path_shared(root, path))
			goto out;
		btrfs_release_path(root, path);
	}

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

	di = btrfs_lookup_dir_item(trans, root, path, dir->i_ino,
				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;
	}
	btrfs_release_path(root, path);

	ref = btrfs_lookup_inode_ref(trans, root, path,
				dentry->d_name.name, dentry->d_name.len,
				inode->i_ino, dir->i_ino, 0);
	if (IS_ERR(ref)) {
		err = PTR_ERR(ref);
		goto out;
	}
	BUG_ON(!ref);
	if (check_path_shared(root, path))
		goto out;
	index = btrfs_inode_ref_index(path->nodes[0], ref);
	btrfs_release_path(root, path);

	di = btrfs_lookup_dir_index_item(trans, root, path, dir->i_ino, index,
				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);
	if (err) {
		btrfs_end_transaction(trans, root);
		root->fs_info->enospc_unlink = 0;
		return ERR_PTR(err);
	}

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

static void __unlink_end_trans(struct btrfs_trans_handle *trans,
			       struct btrfs_root *root)
{
	if (trans->block_rsv == &root->fs_info->global_block_rsv) {
		BUG_ON(!root->fs_info->enospc_unlink);
		root->fs_info->enospc_unlink = 0;
	}
	btrfs_end_transaction_throttle(trans, root);
}

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

	trans = __unlink_start_trans(dir, dentry);
	if (IS_ERR(trans))
		return PTR_ERR(trans);
2935

C
Chris Mason 已提交
2936
	btrfs_set_trans_block_group(trans, dir);
2937 2938 2939

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

2940 2941
	ret = btrfs_unlink_inode(trans, root, dir, dentry->d_inode,
				 dentry->d_name.name, dentry->d_name.len);
2942
	BUG_ON(ret);
2943

2944
	if (inode->i_nlink == 0) {
2945
		ret = btrfs_orphan_add(trans, inode);
2946 2947
		BUG_ON(ret);
	}
2948

2949
	nr = trans->blocks_used;
2950
	__unlink_end_trans(trans, root);
2951
	btrfs_btree_balance_dirty(root, nr);
C
Chris Mason 已提交
2952 2953 2954
	return ret;
}

2955 2956 2957 2958 2959 2960 2961 2962 2963 2964 2965 2966 2967 2968 2969 2970 2971 2972 2973 2974 2975 2976 2977 2978 2979 2980 2981 2982 2983 2984 2985 2986 2987 2988 2989 2990 2991 2992 2993 2994 2995 2996 2997 2998 2999 3000 3001 3002 3003 3004 3005 3006 3007 3008 3009 3010 3011 3012 3013 3014 3015 3016
int btrfs_unlink_subvol(struct btrfs_trans_handle *trans,
			struct btrfs_root *root,
			struct inode *dir, u64 objectid,
			const char *name, int name_len)
{
	struct btrfs_path *path;
	struct extent_buffer *leaf;
	struct btrfs_dir_item *di;
	struct btrfs_key key;
	u64 index;
	int ret;

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

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

	leaf = path->nodes[0];
	btrfs_dir_item_key_to_cpu(leaf, di, &key);
	WARN_ON(key.type != BTRFS_ROOT_ITEM_KEY || key.objectid != objectid);
	ret = btrfs_delete_one_dir_name(trans, root, path, di);
	BUG_ON(ret);
	btrfs_release_path(root, path);

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

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

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

	leaf = path->nodes[0];
	btrfs_dir_item_key_to_cpu(leaf, di, &key);
	WARN_ON(key.type != BTRFS_ROOT_ITEM_KEY || key.objectid != objectid);
	ret = btrfs_delete_one_dir_name(trans, root, path, di);
	BUG_ON(ret);
	btrfs_release_path(root, path);

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

	btrfs_free_path(path);
	return 0;
}

C
Chris Mason 已提交
3017 3018 3019
static int btrfs_rmdir(struct inode *dir, struct dentry *dentry)
{
	struct inode *inode = dentry->d_inode;
3020
	int err = 0;
C
Chris Mason 已提交
3021 3022
	struct btrfs_root *root = BTRFS_I(dir)->root;
	struct btrfs_trans_handle *trans;
3023
	unsigned long nr = 0;
C
Chris Mason 已提交
3024

3025
	if (inode->i_size > BTRFS_EMPTY_DIR_SIZE ||
3026
	    inode->i_ino == BTRFS_FIRST_FREE_OBJECTID)
Y
Yan 已提交
3027 3028
		return -ENOTEMPTY;

3029 3030
	trans = __unlink_start_trans(dir, dentry);
	if (IS_ERR(trans))
3031 3032
		return PTR_ERR(trans);

C
Chris Mason 已提交
3033 3034
	btrfs_set_trans_block_group(trans, dir);

3035 3036 3037 3038 3039 3040 3041 3042
	if (unlikely(inode->i_ino == BTRFS_EMPTY_SUBVOL_DIR_OBJECTID)) {
		err = btrfs_unlink_subvol(trans, root, dir,
					  BTRFS_I(inode)->location.objectid,
					  dentry->d_name.name,
					  dentry->d_name.len);
		goto out;
	}

3043 3044
	err = btrfs_orphan_add(trans, inode);
	if (err)
3045
		goto out;
3046

C
Chris Mason 已提交
3047
	/* now the directory is empty */
3048 3049
	err = btrfs_unlink_inode(trans, root, dir, dentry->d_inode,
				 dentry->d_name.name, dentry->d_name.len);
C
Chris Mason 已提交
3050
	if (!err)
3051
		btrfs_i_size_write(inode, 0);
3052
out:
3053
	nr = trans->blocks_used;
3054
	__unlink_end_trans(trans, root);
3055
	btrfs_btree_balance_dirty(root, nr);
3056

C
Chris Mason 已提交
3057 3058 3059
	return err;
}

3060
#if 0
3061 3062 3063 3064 3065 3066 3067 3068 3069 3070 3071 3072 3073 3074 3075 3076 3077 3078 3079 3080 3081 3082 3083 3084 3085 3086 3087 3088
/*
 * when truncating bytes in a file, it is possible to avoid reading
 * the leaves that contain only checksum items.  This can be the
 * majority of the IO required to delete a large file, but it must
 * be done carefully.
 *
 * The keys in the level just above the leaves are checked to make sure
 * the lowest key in a given leaf is a csum key, and starts at an offset
 * after the new  size.
 *
 * Then the key for the next leaf is checked to make sure it also has
 * a checksum item for the same file.  If it does, we know our target leaf
 * contains only checksum items, and it can be safely freed without reading
 * it.
 *
 * This is just an optimization targeted at large files.  It may do
 * nothing.  It will return 0 unless things went badly.
 */
static noinline int drop_csum_leaves(struct btrfs_trans_handle *trans,
				     struct btrfs_root *root,
				     struct btrfs_path *path,
				     struct inode *inode, u64 new_size)
{
	struct btrfs_key key;
	int ret;
	int nritems;
	struct btrfs_key found_key;
	struct btrfs_key other_key;
Y
Yan Zheng 已提交
3089 3090 3091
	struct btrfs_leaf_ref *ref;
	u64 leaf_gen;
	u64 leaf_start;
3092 3093 3094 3095 3096 3097 3098 3099 3100 3101 3102 3103 3104 3105 3106 3107 3108 3109 3110 3111 3112 3113 3114 3115 3116 3117 3118 3119 3120 3121 3122 3123 3124 3125 3126 3127 3128 3129 3130 3131 3132 3133 3134 3135 3136 3137 3138 3139 3140 3141 3142 3143 3144 3145 3146 3147 3148 3149 3150 3151 3152 3153 3154 3155 3156 3157 3158 3159 3160 3161 3162 3163 3164 3165 3166 3167 3168 3169 3170 3171 3172 3173 3174 3175 3176 3177 3178 3179 3180 3181 3182 3183 3184

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

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

	if (!nritems)
		goto out;

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

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

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

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

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

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

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

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

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

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

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

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

Y
Yan Zheng 已提交
3185 3186
	leaf_start = btrfs_node_blockptr(path->nodes[1], path->slots[1]);
	leaf_gen = btrfs_node_ptr_generation(path->nodes[1], path->slots[1]);
3187 3188 3189 3190
	/*
	 * it is safe to delete this leaf, it contains only
	 * csum items from this inode at an offset >= new_size
	 */
Y
Yan Zheng 已提交
3191
	ret = btrfs_del_leaf(trans, root, path, leaf_start);
3192 3193
	BUG_ON(ret);

Y
Yan Zheng 已提交
3194 3195 3196 3197 3198 3199 3200 3201 3202
	if (root->ref_cows && leaf_gen < trans->transid) {
		ref = btrfs_alloc_leaf_ref(root, 0);
		if (ref) {
			ref->root_gen = root->root_key.offset;
			ref->bytenr = leaf_start;
			ref->owner = 0;
			ref->generation = leaf_gen;
			ref->nritems = 0;

3203 3204
			btrfs_sort_leaf_ref(ref);

Y
Yan Zheng 已提交
3205 3206 3207 3208 3209 3210 3211
			ret = btrfs_add_leaf_ref(root, ref, 0);
			WARN_ON(ret);
			btrfs_free_leaf_ref(root, ref);
		} else {
			WARN_ON(1);
		}
	}
3212 3213 3214 3215 3216 3217 3218 3219 3220 3221 3222 3223 3224 3225 3226 3227 3228 3229
next_key:
	btrfs_release_path(root, path);

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

3230 3231
#endif

C
Chris Mason 已提交
3232 3233 3234
/*
 * 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
3235
 * any higher than new_size
C
Chris Mason 已提交
3236 3237 3238
 *
 * csum items that cross the new i_size are truncated to the new size
 * as well.
3239 3240 3241
 *
 * 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
3242
 */
3243 3244 3245 3246
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
3247 3248
{
	struct btrfs_path *path;
3249
	struct extent_buffer *leaf;
C
Chris Mason 已提交
3250
	struct btrfs_file_extent_item *fi;
3251 3252
	struct btrfs_key key;
	struct btrfs_key found_key;
C
Chris Mason 已提交
3253
	u64 extent_start = 0;
3254
	u64 extent_num_bytes = 0;
3255
	u64 extent_offset = 0;
C
Chris Mason 已提交
3256
	u64 item_end = 0;
3257 3258
	u64 mask = root->sectorsize - 1;
	u32 found_type = (u8)-1;
C
Chris Mason 已提交
3259 3260
	int found_extent;
	int del_item;
3261 3262
	int pending_del_nr = 0;
	int pending_del_slot = 0;
3263
	int extent_type = -1;
3264
	int encoding;
3265 3266 3267 3268
	int ret;
	int err = 0;

	BUG_ON(new_size > 0 && min_type != BTRFS_EXTENT_DATA_KEY);
C
Chris Mason 已提交
3269

3270
	if (root->ref_cows || root == root->fs_info->tree_root)
3271
		btrfs_drop_extent_cache(inode, new_size & (~mask), (u64)-1, 0);
3272

C
Chris Mason 已提交
3273 3274
	path = btrfs_alloc_path();
	BUG_ON(!path);
J
Julia Lawall 已提交
3275
	path->reada = -1;
3276

C
Chris Mason 已提交
3277 3278
	key.objectid = inode->i_ino;
	key.offset = (u64)-1;
3279 3280
	key.type = (u8)-1;

3281
search_again:
3282
	path->leave_spinning = 1;
3283
	ret = btrfs_search_slot(trans, root, &key, path, -1, 1);
3284 3285 3286 3287
	if (ret < 0) {
		err = ret;
		goto out;
	}
C
Chris Mason 已提交
3288

3289
	if (ret > 0) {
3290 3291 3292
		/* there are no items in the tree for us to truncate, we're
		 * done
		 */
3293 3294
		if (path->slots[0] == 0)
			goto out;
3295 3296 3297
		path->slots[0]--;
	}

C
Chris Mason 已提交
3298
	while (1) {
C
Chris Mason 已提交
3299
		fi = NULL;
3300 3301 3302
		leaf = path->nodes[0];
		btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
		found_type = btrfs_key_type(&found_key);
3303
		encoding = 0;
C
Chris Mason 已提交
3304

3305
		if (found_key.objectid != inode->i_ino)
C
Chris Mason 已提交
3306
			break;
3307

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

3311
		item_end = found_key.offset;
C
Chris Mason 已提交
3312
		if (found_type == BTRFS_EXTENT_DATA_KEY) {
3313
			fi = btrfs_item_ptr(leaf, path->slots[0],
C
Chris Mason 已提交
3314
					    struct btrfs_file_extent_item);
3315
			extent_type = btrfs_file_extent_type(leaf, fi);
3316 3317 3318 3319
			encoding = btrfs_file_extent_compression(leaf, fi);
			encoding |= btrfs_file_extent_encryption(leaf, fi);
			encoding |= btrfs_file_extent_other_encoding(leaf, fi);

3320
			if (extent_type != BTRFS_FILE_EXTENT_INLINE) {
3321
				item_end +=
3322
				    btrfs_file_extent_num_bytes(leaf, fi);
3323 3324
			} else if (extent_type == BTRFS_FILE_EXTENT_INLINE) {
				item_end += btrfs_file_extent_inline_len(leaf,
C
Chris Mason 已提交
3325
									 fi);
C
Chris Mason 已提交
3326
			}
3327
			item_end--;
C
Chris Mason 已提交
3328
		}
3329 3330 3331 3332
		if (found_type > min_type) {
			del_item = 1;
		} else {
			if (item_end < new_size)
3333
				break;
3334 3335 3336 3337
			if (found_key.offset >= new_size)
				del_item = 1;
			else
				del_item = 0;
C
Chris Mason 已提交
3338 3339 3340
		}
		found_extent = 0;
		/* FIXME, shrink the extent if the ref count is only 1 */
3341 3342 3343 3344
		if (found_type != BTRFS_EXTENT_DATA_KEY)
			goto delete;

		if (extent_type != BTRFS_FILE_EXTENT_INLINE) {
C
Chris Mason 已提交
3345
			u64 num_dec;
3346
			extent_start = btrfs_file_extent_disk_bytenr(leaf, fi);
3347
			if (!del_item && !encoding) {
3348 3349
				u64 orig_num_bytes =
					btrfs_file_extent_num_bytes(leaf, fi);
3350
				extent_num_bytes = new_size -
3351
					found_key.offset + root->sectorsize - 1;
3352 3353
				extent_num_bytes = extent_num_bytes &
					~((u64)root->sectorsize - 1);
3354 3355 3356
				btrfs_set_file_extent_num_bytes(leaf, fi,
							 extent_num_bytes);
				num_dec = (orig_num_bytes -
C
Chris Mason 已提交
3357
					   extent_num_bytes);
3358
				if (root->ref_cows && extent_start != 0)
3359
					inode_sub_bytes(inode, num_dec);
3360
				btrfs_mark_buffer_dirty(leaf);
C
Chris Mason 已提交
3361
			} else {
3362 3363 3364
				extent_num_bytes =
					btrfs_file_extent_disk_num_bytes(leaf,
									 fi);
3365 3366 3367
				extent_offset = found_key.offset -
					btrfs_file_extent_offset(leaf, fi);

C
Chris Mason 已提交
3368
				/* FIXME blocksize != 4096 */
C
Chris Mason 已提交
3369
				num_dec = btrfs_file_extent_num_bytes(leaf, fi);
C
Chris Mason 已提交
3370 3371
				if (extent_start != 0) {
					found_extent = 1;
3372
					if (root->ref_cows)
3373
						inode_sub_bytes(inode, num_dec);
3374
				}
C
Chris Mason 已提交
3375
			}
C
Chris Mason 已提交
3376
		} else if (extent_type == BTRFS_FILE_EXTENT_INLINE) {
C
Chris Mason 已提交
3377 3378 3379 3380 3381 3382 3383 3384
			/*
			 * 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) {
3385 3386 3387
				u32 size = new_size - found_key.offset;

				if (root->ref_cows) {
3388 3389
					inode_sub_bytes(inode, item_end + 1 -
							new_size);
3390 3391 3392
				}
				size =
				    btrfs_file_extent_calc_inline_size(size);
C
Chris Mason 已提交
3393
				ret = btrfs_truncate_item(trans, root, path,
3394
							  size, 1);
C
Chris Mason 已提交
3395
				BUG_ON(ret);
3396
			} else if (root->ref_cows) {
3397 3398
				inode_sub_bytes(inode, item_end + 1 -
						found_key.offset);
C
Chris Mason 已提交
3399
			}
C
Chris Mason 已提交
3400
		}
3401
delete:
C
Chris Mason 已提交
3402
		if (del_item) {
3403 3404 3405 3406 3407 3408 3409 3410 3411 3412
			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
3413
				BUG();
3414
			}
C
Chris Mason 已提交
3415 3416 3417
		} else {
			break;
		}
3418 3419
		if (found_extent && (root->ref_cows ||
				     root == root->fs_info->tree_root)) {
3420
			btrfs_set_path_blocking(path);
C
Chris Mason 已提交
3421
			ret = btrfs_free_extent(trans, root, extent_start,
3422 3423 3424
						extent_num_bytes, 0,
						btrfs_header_owner(leaf),
						inode->i_ino, extent_offset);
C
Chris Mason 已提交
3425 3426
			BUG_ON(ret);
		}
3427

3428 3429 3430 3431 3432 3433 3434 3435 3436 3437 3438 3439 3440 3441 3442 3443
		if (found_type == BTRFS_INODE_ITEM_KEY)
			break;

		if (path->slots[0] == 0 ||
		    path->slots[0] != pending_del_slot) {
			if (root->ref_cows) {
				err = -EAGAIN;
				goto out;
			}
			if (pending_del_nr) {
				ret = btrfs_del_items(trans, root, path,
						pending_del_slot,
						pending_del_nr);
				BUG_ON(ret);
				pending_del_nr = 0;
			}
3444 3445
			btrfs_release_path(root, path);
			goto search_again;
3446 3447
		} else {
			path->slots[0]--;
3448
		}
C
Chris Mason 已提交
3449
	}
3450
out:
3451 3452 3453
	if (pending_del_nr) {
		ret = btrfs_del_items(trans, root, path, pending_del_slot,
				      pending_del_nr);
3454
		BUG_ON(ret);
3455
	}
C
Chris Mason 已提交
3456
	btrfs_free_path(path);
3457
	return err;
C
Chris Mason 已提交
3458 3459 3460 3461 3462 3463 3464 3465 3466
}

/*
 * taken from block_truncate_page, but does cow as it zeros out
 * any bytes left in the last page in the file.
 */
static int btrfs_truncate_page(struct address_space *mapping, loff_t from)
{
	struct inode *inode = mapping->host;
3467
	struct btrfs_root *root = BTRFS_I(inode)->root;
3468 3469
	struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
	struct btrfs_ordered_extent *ordered;
3470
	struct extent_state *cached_state = NULL;
3471
	char *kaddr;
3472
	u32 blocksize = root->sectorsize;
C
Chris Mason 已提交
3473 3474 3475 3476
	pgoff_t index = from >> PAGE_CACHE_SHIFT;
	unsigned offset = from & (PAGE_CACHE_SIZE-1);
	struct page *page;
	int ret = 0;
3477
	u64 page_start;
3478
	u64 page_end;
C
Chris Mason 已提交
3479 3480 3481

	if ((offset & (blocksize - 1)) == 0)
		goto out;
3482
	ret = btrfs_delalloc_reserve_space(inode, PAGE_CACHE_SIZE);
3483 3484
	if (ret)
		goto out;
C
Chris Mason 已提交
3485 3486

	ret = -ENOMEM;
3487
again:
C
Chris Mason 已提交
3488
	page = grab_cache_page(mapping, index);
3489
	if (!page) {
3490
		btrfs_delalloc_release_space(inode, PAGE_CACHE_SIZE);
C
Chris Mason 已提交
3491
		goto out;
3492
	}
3493 3494 3495 3496

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

C
Chris Mason 已提交
3497
	if (!PageUptodate(page)) {
C
Chris Mason 已提交
3498
		ret = btrfs_readpage(NULL, page);
C
Chris Mason 已提交
3499
		lock_page(page);
3500 3501 3502 3503 3504
		if (page->mapping != mapping) {
			unlock_page(page);
			page_cache_release(page);
			goto again;
		}
C
Chris Mason 已提交
3505 3506
		if (!PageUptodate(page)) {
			ret = -EIO;
3507
			goto out_unlock;
C
Chris Mason 已提交
3508 3509
		}
	}
3510
	wait_on_page_writeback(page);
3511

3512 3513
	lock_extent_bits(io_tree, page_start, page_end, 0, &cached_state,
			 GFP_NOFS);
3514 3515 3516 3517
	set_page_extent_mapped(page);

	ordered = btrfs_lookup_ordered_extent(inode, page_start);
	if (ordered) {
3518 3519
		unlock_extent_cached(io_tree, page_start, page_end,
				     &cached_state, GFP_NOFS);
3520 3521
		unlock_page(page);
		page_cache_release(page);
3522
		btrfs_start_ordered_extent(inode, ordered, 1);
3523 3524 3525 3526
		btrfs_put_ordered_extent(ordered);
		goto again;
	}

3527
	clear_extent_bit(&BTRFS_I(inode)->io_tree, page_start, page_end,
3528
			  EXTENT_DIRTY | EXTENT_DELALLOC | EXTENT_DO_ACCOUNTING,
3529
			  0, 0, &cached_state, GFP_NOFS);
3530

3531 3532
	ret = btrfs_set_extent_delalloc(inode, page_start, page_end,
					&cached_state);
J
Josef Bacik 已提交
3533
	if (ret) {
3534 3535
		unlock_extent_cached(io_tree, page_start, page_end,
				     &cached_state, GFP_NOFS);
J
Josef Bacik 已提交
3536 3537 3538
		goto out_unlock;
	}

3539 3540 3541 3542 3543 3544 3545
	ret = 0;
	if (offset != PAGE_CACHE_SIZE) {
		kaddr = kmap(page);
		memset(kaddr + offset, 0, PAGE_CACHE_SIZE - offset);
		flush_dcache_page(page);
		kunmap(page);
	}
3546
	ClearPageChecked(page);
3547
	set_page_dirty(page);
3548 3549
	unlock_extent_cached(io_tree, page_start, page_end, &cached_state,
			     GFP_NOFS);
C
Chris Mason 已提交
3550

3551
out_unlock:
3552
	if (ret)
3553
		btrfs_delalloc_release_space(inode, PAGE_CACHE_SIZE);
C
Chris Mason 已提交
3554 3555 3556 3557 3558 3559
	unlock_page(page);
	page_cache_release(page);
out:
	return ret;
}

3560 3561 3562 3563 3564 3565
/*
 * 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
 */
3566
int btrfs_cont_expand(struct inode *inode, loff_t oldsize, loff_t size)
C
Chris Mason 已提交
3567
{
Y
Yan Zheng 已提交
3568 3569 3570
	struct btrfs_trans_handle *trans;
	struct btrfs_root *root = BTRFS_I(inode)->root;
	struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
3571
	struct extent_map *em = NULL;
3572
	struct extent_state *cached_state = NULL;
Y
Yan Zheng 已提交
3573
	u64 mask = root->sectorsize - 1;
3574
	u64 hole_start = (oldsize + mask) & ~mask;
Y
Yan Zheng 已提交
3575 3576 3577 3578
	u64 block_end = (size + mask) & ~mask;
	u64 last_byte;
	u64 cur_offset;
	u64 hole_size;
J
Josef Bacik 已提交
3579
	int err = 0;
C
Chris Mason 已提交
3580

Y
Yan Zheng 已提交
3581 3582 3583 3584 3585 3586 3587
	if (size <= hole_start)
		return 0;

	while (1) {
		struct btrfs_ordered_extent *ordered;
		btrfs_wait_ordered_range(inode, hole_start,
					 block_end - hole_start);
3588 3589
		lock_extent_bits(io_tree, hole_start, block_end - 1, 0,
				 &cached_state, GFP_NOFS);
Y
Yan Zheng 已提交
3590 3591 3592
		ordered = btrfs_lookup_ordered_extent(inode, hole_start);
		if (!ordered)
			break;
3593 3594
		unlock_extent_cached(io_tree, hole_start, block_end - 1,
				     &cached_state, GFP_NOFS);
Y
Yan Zheng 已提交
3595 3596
		btrfs_put_ordered_extent(ordered);
	}
C
Chris Mason 已提交
3597

Y
Yan Zheng 已提交
3598 3599 3600 3601 3602 3603 3604
	cur_offset = hole_start;
	while (1) {
		em = btrfs_get_extent(inode, NULL, 0, cur_offset,
				block_end - cur_offset, 0);
		BUG_ON(IS_ERR(em) || !em);
		last_byte = min(extent_map_end(em), block_end);
		last_byte = (last_byte + mask) & ~mask;
3605
		if (!test_bit(EXTENT_FLAG_PREALLOC, &em->flags)) {
3606
			u64 hint_byte = 0;
Y
Yan Zheng 已提交
3607
			hole_size = last_byte - cur_offset;
J
Josef Bacik 已提交
3608

3609 3610 3611
			trans = btrfs_start_transaction(root, 2);
			if (IS_ERR(trans)) {
				err = PTR_ERR(trans);
J
Josef Bacik 已提交
3612
				break;
3613
			}
3614 3615 3616 3617 3618
			btrfs_set_trans_block_group(trans, inode);

			err = btrfs_drop_extents(trans, inode, cur_offset,
						 cur_offset + hole_size,
						 &hint_byte, 1);
3619 3620
			if (err)
				break;
3621

Y
Yan Zheng 已提交
3622 3623 3624 3625
			err = btrfs_insert_file_extent(trans, root,
					inode->i_ino, cur_offset, 0,
					0, hole_size, 0, hole_size,
					0, 0, 0);
3626 3627
			if (err)
				break;
3628

Y
Yan Zheng 已提交
3629 3630
			btrfs_drop_extent_cache(inode, hole_start,
					last_byte - 1, 0);
3631 3632

			btrfs_end_transaction(trans, root);
Y
Yan Zheng 已提交
3633 3634
		}
		free_extent_map(em);
3635
		em = NULL;
Y
Yan Zheng 已提交
3636
		cur_offset = last_byte;
3637
		if (cur_offset >= block_end)
Y
Yan Zheng 已提交
3638 3639
			break;
	}
3640

3641
	free_extent_map(em);
3642 3643
	unlock_extent_cached(io_tree, hole_start, block_end - 1, &cached_state,
			     GFP_NOFS);
Y
Yan Zheng 已提交
3644 3645
	return err;
}
C
Chris Mason 已提交
3646

3647
static int btrfs_setsize(struct inode *inode, loff_t newsize)
3648
{
3649
	loff_t oldsize = i_size_read(inode);
3650 3651
	int ret;

3652
	if (newsize == oldsize)
3653 3654
		return 0;

3655 3656 3657 3658 3659
	if (newsize > oldsize) {
		i_size_write(inode, newsize);
		btrfs_ordered_update_i_size(inode, i_size_read(inode), NULL);
		truncate_pagecache(inode, oldsize, newsize);
		ret = btrfs_cont_expand(inode, oldsize, newsize);
3660
		if (ret) {
3661
			btrfs_setsize(inode, oldsize);
3662 3663 3664
			return ret;
		}

3665
		mark_inode_dirty(inode);
3666
	} else {
3667

3668 3669 3670 3671 3672 3673 3674
		/*
		 * 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)
			BTRFS_I(inode)->ordered_data_close = 1;
3675

3676 3677 3678 3679
		/* we don't support swapfiles, so vmtruncate shouldn't fail */
		truncate_setsize(inode, newsize);
		ret = btrfs_truncate(inode);
	}
3680

3681
	return ret;
3682 3683
}

Y
Yan Zheng 已提交
3684 3685 3686
static int btrfs_setattr(struct dentry *dentry, struct iattr *attr)
{
	struct inode *inode = dentry->d_inode;
L
Li Zefan 已提交
3687
	struct btrfs_root *root = BTRFS_I(inode)->root;
Y
Yan Zheng 已提交
3688
	int err;
C
Chris Mason 已提交
3689

L
Li Zefan 已提交
3690 3691 3692
	if (btrfs_root_readonly(root))
		return -EROFS;

Y
Yan Zheng 已提交
3693 3694 3695
	err = inode_change_ok(inode, attr);
	if (err)
		return err;
C
Chris Mason 已提交
3696

3697
	if (S_ISREG(inode->i_mode) && (attr->ia_valid & ATTR_SIZE)) {
3698
		err = btrfs_setsize(inode, attr->ia_size);
3699 3700
		if (err)
			return err;
C
Chris Mason 已提交
3701
	}
Y
Yan Zheng 已提交
3702

C
Christoph Hellwig 已提交
3703 3704 3705 3706 3707 3708 3709
	if (attr->ia_valid) {
		setattr_copy(inode, attr);
		mark_inode_dirty(inode);

		if (attr->ia_valid & ATTR_MODE)
			err = btrfs_acl_chmod(inode);
	}
J
Josef Bacik 已提交
3710

C
Chris Mason 已提交
3711 3712
	return err;
}
3713

A
Al Viro 已提交
3714
void btrfs_evict_inode(struct inode *inode)
C
Chris Mason 已提交
3715 3716 3717
{
	struct btrfs_trans_handle *trans;
	struct btrfs_root *root = BTRFS_I(inode)->root;
3718
	unsigned long nr;
C
Chris Mason 已提交
3719 3720 3721
	int ret;

	truncate_inode_pages(&inode->i_data, 0);
3722 3723
	if (inode->i_nlink && (btrfs_root_refs(&root->root_item) != 0 ||
			       root == root->fs_info->tree_root))
A
Al Viro 已提交
3724 3725
		goto no_delete;

C
Chris Mason 已提交
3726
	if (is_bad_inode(inode)) {
3727
		btrfs_orphan_del(NULL, inode);
C
Chris Mason 已提交
3728 3729
		goto no_delete;
	}
A
Al Viro 已提交
3730
	/* do we really want it for ->i_nlink > 0 and zero btrfs_root_refs? */
C
Chris Mason 已提交
3731
	btrfs_wait_ordered_range(inode, 0, (u64)-1);
3732

3733 3734 3735 3736 3737
	if (root->fs_info->log_root_recovering) {
		BUG_ON(!list_empty(&BTRFS_I(inode)->i_orphan));
		goto no_delete;
	}

3738 3739 3740 3741 3742
	if (inode->i_nlink > 0) {
		BUG_ON(btrfs_root_refs(&root->root_item) != 0);
		goto no_delete;
	}

3743
	btrfs_i_size_write(inode, 0);
3744

3745
	while (1) {
3746 3747
		trans = btrfs_start_transaction(root, 0);
		BUG_ON(IS_ERR(trans));
3748
		btrfs_set_trans_block_group(trans, inode);
3749 3750 3751 3752 3753 3754 3755 3756 3757 3758
		trans->block_rsv = root->orphan_block_rsv;

		ret = btrfs_block_rsv_check(trans, root,
					    root->orphan_block_rsv, 0, 5);
		if (ret) {
			BUG_ON(ret != -EAGAIN);
			ret = btrfs_commit_transaction(trans, root);
			BUG_ON(ret);
			continue;
		}
3759

3760
		ret = btrfs_truncate_inode_items(trans, root, inode, 0, 0);
3761 3762
		if (ret != -EAGAIN)
			break;
3763

3764 3765 3766 3767
		nr = trans->blocks_used;
		btrfs_end_transaction(trans, root);
		trans = NULL;
		btrfs_btree_balance_dirty(root, nr);
3768

3769
	}
3770

3771 3772 3773 3774
	if (ret == 0) {
		ret = btrfs_orphan_del(trans, inode);
		BUG_ON(ret);
	}
3775

3776
	nr = trans->blocks_used;
3777
	btrfs_end_transaction(trans, root);
3778
	btrfs_btree_balance_dirty(root, nr);
C
Chris Mason 已提交
3779
no_delete:
A
Al Viro 已提交
3780
	end_writeback(inode);
3781
	return;
C
Chris Mason 已提交
3782 3783 3784 3785 3786 3787 3788 3789 3790 3791 3792 3793 3794 3795
}

/*
 * 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
3796
	int ret = 0;
C
Chris Mason 已提交
3797 3798 3799

	path = btrfs_alloc_path();
	BUG_ON(!path);
3800

C
Chris Mason 已提交
3801 3802
	di = btrfs_lookup_dir_item(NULL, root, path, dir->i_ino, name,
				    namelen, 0);
Y
Yan 已提交
3803 3804
	if (IS_ERR(di))
		ret = PTR_ERR(di);
C
Chris Mason 已提交
3805 3806

	if (!di || IS_ERR(di))
3807
		goto out_err;
C
Chris Mason 已提交
3808

3809
	btrfs_dir_item_key_to_cpu(path->nodes[0], di, location);
C
Chris Mason 已提交
3810 3811 3812
out:
	btrfs_free_path(path);
	return ret;
3813 3814 3815
out_err:
	location->objectid = 0;
	goto out;
C
Chris Mason 已提交
3816 3817 3818 3819 3820 3821 3822 3823
}

/*
 * 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,
3824 3825 3826 3827
				    struct inode *dir,
				    struct dentry *dentry,
				    struct btrfs_key *location,
				    struct btrfs_root **sub_root)
C
Chris Mason 已提交
3828
{
3829 3830 3831 3832 3833 3834
	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
3835

3836 3837 3838 3839 3840
	path = btrfs_alloc_path();
	if (!path) {
		err = -ENOMEM;
		goto out;
	}
C
Chris Mason 已提交
3841

3842 3843 3844 3845 3846 3847 3848 3849 3850
	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
3851

3852 3853 3854 3855 3856
	leaf = path->nodes[0];
	ref = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_root_ref);
	if (btrfs_root_ref_dirid(leaf, ref) != dir->i_ino ||
	    btrfs_root_ref_name_len(leaf, ref) != dentry->d_name.len)
		goto out;
C
Chris Mason 已提交
3857

3858 3859 3860 3861 3862 3863 3864 3865 3866 3867 3868 3869 3870 3871 3872 3873 3874 3875 3876 3877 3878 3879 3880 3881 3882 3883 3884
	ret = memcmp_extent_buffer(leaf, dentry->d_name.name,
				   (unsigned long)(ref + 1),
				   dentry->d_name.len);
	if (ret)
		goto out;

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

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

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

	*sub_root = new_root;
	location->objectid = btrfs_root_dirid(&new_root->root_item);
	location->type = BTRFS_INODE_ITEM_KEY;
	location->offset = 0;
	err = 0;
out:
	btrfs_free_path(path);
	return err;
C
Chris Mason 已提交
3885 3886
}

3887 3888 3889 3890
static void inode_tree_add(struct inode *inode)
{
	struct btrfs_root *root = BTRFS_I(inode)->root;
	struct btrfs_inode *entry;
3891 3892 3893 3894 3895
	struct rb_node **p;
	struct rb_node *parent;
again:
	p = &root->inode_tree.rb_node;
	parent = NULL;
3896

A
Al Viro 已提交
3897
	if (inode_unhashed(inode))
3898 3899
		return;

3900 3901 3902 3903 3904 3905
	spin_lock(&root->inode_lock);
	while (*p) {
		parent = *p;
		entry = rb_entry(parent, struct btrfs_inode, rb_node);

		if (inode->i_ino < entry->vfs_inode.i_ino)
3906
			p = &parent->rb_left;
3907
		else if (inode->i_ino > entry->vfs_inode.i_ino)
3908
			p = &parent->rb_right;
3909 3910
		else {
			WARN_ON(!(entry->vfs_inode.i_state &
A
Al Viro 已提交
3911
				  (I_WILL_FREE | I_FREEING)));
3912 3913 3914 3915
			rb_erase(parent, &root->inode_tree);
			RB_CLEAR_NODE(parent);
			spin_unlock(&root->inode_lock);
			goto again;
3916 3917 3918 3919 3920 3921 3922 3923 3924 3925
		}
	}
	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;
3926
	int empty = 0;
3927

3928
	spin_lock(&root->inode_lock);
3929 3930 3931
	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);
3932
		empty = RB_EMPTY_ROOT(&root->inode_tree);
3933
	}
3934
	spin_unlock(&root->inode_lock);
3935

3936 3937 3938 3939 3940 3941 3942 3943
	/*
	 * 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) {
3944 3945 3946 3947 3948 3949 3950 3951 3952 3953 3954 3955 3956 3957 3958 3959 3960 3961 3962 3963 3964 3965 3966 3967 3968 3969 3970 3971 3972 3973 3974 3975 3976 3977 3978 3979 3980 3981 3982 3983 3984 3985 3986 3987 3988 3989 3990 3991 3992 3993 3994 3995 3996
		synchronize_srcu(&root->fs_info->subvol_srcu);
		spin_lock(&root->inode_lock);
		empty = RB_EMPTY_ROOT(&root->inode_tree);
		spin_unlock(&root->inode_lock);
		if (empty)
			btrfs_add_dead_root(root);
	}
}

int btrfs_invalidate_inodes(struct btrfs_root *root)
{
	struct rb_node *node;
	struct rb_node *prev;
	struct btrfs_inode *entry;
	struct inode *inode;
	u64 objectid = 0;

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

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

		if (objectid < entry->vfs_inode.i_ino)
			node = node->rb_left;
		else if (objectid > entry->vfs_inode.i_ino)
			node = node->rb_right;
		else
			break;
	}
	if (!node) {
		while (prev) {
			entry = rb_entry(prev, struct btrfs_inode, rb_node);
			if (objectid <= entry->vfs_inode.i_ino) {
				node = prev;
				break;
			}
			prev = rb_next(prev);
		}
	}
	while (node) {
		entry = rb_entry(node, struct btrfs_inode, rb_node);
		objectid = entry->vfs_inode.i_ino + 1;
		inode = igrab(&entry->vfs_inode);
		if (inode) {
			spin_unlock(&root->inode_lock);
			if (atomic_read(&inode->i_count) > 1)
				d_prune_aliases(inode);
			/*
3997
			 * btrfs_drop_inode will have it removed from
3998 3999 4000 4001 4002 4003 4004 4005 4006 4007 4008 4009 4010 4011 4012 4013
			 * the inode cache when its usage count
			 * hits zero.
			 */
			iput(inode);
			cond_resched();
			spin_lock(&root->inode_lock);
			goto again;
		}

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

		node = rb_next(node);
	}
	spin_unlock(&root->inode_lock);
	return 0;
4014 4015
}

4016 4017 4018 4019 4020
static int btrfs_init_locked_inode(struct inode *inode, void *p)
{
	struct btrfs_iget_args *args = p;
	inode->i_ino = args->ino;
	BTRFS_I(inode)->root = args->root;
J
Josef Bacik 已提交
4021
	btrfs_set_inode_space_info(args->root, inode);
C
Chris Mason 已提交
4022 4023 4024 4025 4026 4027
	return 0;
}

static int btrfs_find_actor(struct inode *inode, void *opaque)
{
	struct btrfs_iget_args *args = opaque;
C
Chris Mason 已提交
4028 4029
	return args->ino == inode->i_ino &&
		args->root == BTRFS_I(inode)->root;
C
Chris Mason 已提交
4030 4031
}

4032 4033 4034
static struct inode *btrfs_iget_locked(struct super_block *s,
				       u64 objectid,
				       struct btrfs_root *root)
C
Chris Mason 已提交
4035 4036 4037 4038 4039 4040 4041 4042 4043 4044 4045 4046
{
	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
4047 4048 4049 4050
/* 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,
4051
			 struct btrfs_root *root, int *new)
B
Balaji Rao 已提交
4052 4053 4054 4055 4056
{
	struct inode *inode;

	inode = btrfs_iget_locked(s, location->objectid, root);
	if (!inode)
4057
		return ERR_PTR(-ENOMEM);
B
Balaji Rao 已提交
4058 4059 4060 4061 4062

	if (inode->i_state & I_NEW) {
		BTRFS_I(inode)->root = root;
		memcpy(&BTRFS_I(inode)->location, location, sizeof(*location));
		btrfs_read_locked_inode(inode);
4063 4064

		inode_tree_add(inode);
B
Balaji Rao 已提交
4065
		unlock_new_inode(inode);
4066 4067
		if (new)
			*new = 1;
B
Balaji Rao 已提交
4068 4069 4070 4071 4072
	}

	return inode;
}

4073 4074 4075 4076 4077 4078 4079 4080 4081 4082 4083 4084 4085 4086 4087 4088 4089 4090 4091 4092 4093 4094
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));
	BTRFS_I(inode)->dummy_inode = 1;

	inode->i_ino = BTRFS_EMPTY_SUBVOL_DIR_OBJECTID;
	inode->i_op = &simple_dir_inode_operations;
	inode->i_fop = &simple_dir_operations;
	inode->i_mode = S_IFDIR | S_IRUGO | S_IWUSR | S_IXUGO;
	inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;

	return inode;
}

4095
struct inode *btrfs_lookup_dentry(struct inode *dir, struct dentry *dentry)
C
Chris Mason 已提交
4096
{
C
Chris Mason 已提交
4097
	struct inode *inode;
4098
	struct btrfs_root *root = BTRFS_I(dir)->root;
C
Chris Mason 已提交
4099 4100
	struct btrfs_root *sub_root = root;
	struct btrfs_key location;
4101
	int index;
4102
	int ret;
C
Chris Mason 已提交
4103 4104 4105

	if (dentry->d_name.len > BTRFS_NAME_LEN)
		return ERR_PTR(-ENAMETOOLONG);
4106

C
Chris Mason 已提交
4107
	ret = btrfs_inode_by_name(dir, dentry, &location);
4108

C
Chris Mason 已提交
4109 4110
	if (ret < 0)
		return ERR_PTR(ret);
4111

4112 4113 4114 4115
	if (location.objectid == 0)
		return NULL;

	if (location.type == BTRFS_INODE_ITEM_KEY) {
4116
		inode = btrfs_iget(dir->i_sb, &location, root, NULL);
4117 4118 4119 4120 4121
		return inode;
	}

	BUG_ON(location.type != BTRFS_ROOT_ITEM_KEY);

4122
	index = srcu_read_lock(&root->fs_info->subvol_srcu);
4123 4124 4125 4126 4127 4128 4129 4130
	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 {
4131
		inode = btrfs_iget(dir->i_sb, &location, sub_root, NULL);
C
Chris Mason 已提交
4132
	}
4133 4134
	srcu_read_unlock(&root->fs_info->subvol_srcu, index);

4135
	if (!IS_ERR(inode) && root != sub_root) {
4136 4137
		down_read(&root->fs_info->cleanup_work_sem);
		if (!(inode->i_sb->s_flags & MS_RDONLY))
4138
			ret = btrfs_orphan_cleanup(sub_root);
4139
		up_read(&root->fs_info->cleanup_work_sem);
4140 4141
		if (ret)
			inode = ERR_PTR(ret);
4142 4143
	}

4144 4145 4146
	return inode;
}

N
Nick Piggin 已提交
4147
static int btrfs_dentry_delete(const struct dentry *dentry)
4148 4149 4150
{
	struct btrfs_root *root;

4151 4152
	if (!dentry->d_inode && !IS_ROOT(dentry))
		dentry = dentry->d_parent;
4153

4154 4155 4156 4157 4158
	if (dentry->d_inode) {
		root = BTRFS_I(dentry->d_inode)->root;
		if (btrfs_root_refs(&root->root_item) == 0)
			return 1;
	}
4159 4160 4161
	return 0;
}

4162 4163 4164 4165 4166 4167 4168 4169
static struct dentry *btrfs_lookup(struct inode *dir, struct dentry *dentry,
				   struct nameidata *nd)
{
	struct inode *inode;

	inode = btrfs_lookup_dentry(dir, dentry);
	if (IS_ERR(inode))
		return ERR_CAST(inode);
4170

C
Chris Mason 已提交
4171 4172 4173 4174 4175 4176 4177
	return d_splice_alias(inode, dentry);
}

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

4178 4179
static int btrfs_real_readdir(struct file *filp, void *dirent,
			      filldir_t filldir)
C
Chris Mason 已提交
4180
{
4181
	struct inode *inode = filp->f_dentry->d_inode;
C
Chris Mason 已提交
4182 4183 4184 4185
	struct btrfs_root *root = BTRFS_I(inode)->root;
	struct btrfs_item *item;
	struct btrfs_dir_item *di;
	struct btrfs_key key;
4186
	struct btrfs_key found_key;
C
Chris Mason 已提交
4187 4188 4189
	struct btrfs_path *path;
	int ret;
	u32 nritems;
4190
	struct extent_buffer *leaf;
C
Chris Mason 已提交
4191 4192 4193 4194 4195 4196 4197 4198
	int slot;
	int advance;
	unsigned char d_type;
	int over = 0;
	u32 di_cur;
	u32 di_total;
	u32 di_len;
	int key_type = BTRFS_DIR_INDEX_KEY;
4199 4200 4201
	char tmp_name[32];
	char *name_ptr;
	int name_len;
C
Chris Mason 已提交
4202 4203 4204 4205

	/* FIXME, use a real flag for deciding about the key type */
	if (root->fs_info->tree_root == root)
		key_type = BTRFS_DIR_ITEM_KEY;
4206

4207 4208 4209 4210 4211 4212 4213 4214 4215 4216 4217
	/* special case for "." */
	if (filp->f_pos == 0) {
		over = filldir(dirent, ".", 1,
			       1, inode->i_ino,
			       DT_DIR);
		if (over)
			return 0;
		filp->f_pos = 1;
	}
	/* special case for .., just use the back ref */
	if (filp->f_pos == 1) {
4218
		u64 pino = parent_ino(filp->f_path.dentry);
4219
		over = filldir(dirent, "..", 2,
4220
			       2, pino, DT_DIR);
4221
		if (over)
4222
			return 0;
4223 4224
		filp->f_pos = 2;
	}
4225 4226 4227
	path = btrfs_alloc_path();
	path->reada = 2;

C
Chris Mason 已提交
4228 4229
	btrfs_set_key_type(&key, key_type);
	key.offset = filp->f_pos;
4230
	key.objectid = inode->i_ino;
4231

C
Chris Mason 已提交
4232 4233 4234 4235
	ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
	if (ret < 0)
		goto err;
	advance = 0;
4236 4237

	while (1) {
4238 4239
		leaf = path->nodes[0];
		nritems = btrfs_header_nritems(leaf);
C
Chris Mason 已提交
4240 4241
		slot = path->slots[0];
		if (advance || slot >= nritems) {
4242
			if (slot >= nritems - 1) {
C
Chris Mason 已提交
4243 4244 4245
				ret = btrfs_next_leaf(root, path);
				if (ret)
					break;
4246 4247
				leaf = path->nodes[0];
				nritems = btrfs_header_nritems(leaf);
C
Chris Mason 已提交
4248 4249 4250 4251 4252 4253
				slot = path->slots[0];
			} else {
				slot++;
				path->slots[0]++;
			}
		}
4254

C
Chris Mason 已提交
4255
		advance = 1;
4256 4257 4258 4259
		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
4260
			break;
4261
		if (btrfs_key_type(&found_key) != key_type)
C
Chris Mason 已提交
4262
			break;
4263
		if (found_key.offset < filp->f_pos)
C
Chris Mason 已提交
4264
			continue;
4265 4266

		filp->f_pos = found_key.offset;
4267

C
Chris Mason 已提交
4268 4269
		di = btrfs_item_ptr(leaf, slot, struct btrfs_dir_item);
		di_cur = 0;
4270
		di_total = btrfs_item_size(leaf, item);
4271 4272

		while (di_cur < di_total) {
4273 4274 4275
			struct btrfs_key location;

			name_len = btrfs_dir_name_len(leaf, di);
4276
			if (name_len <= sizeof(tmp_name)) {
4277 4278 4279
				name_ptr = tmp_name;
			} else {
				name_ptr = kmalloc(name_len, GFP_NOFS);
4280 4281 4282 4283
				if (!name_ptr) {
					ret = -ENOMEM;
					goto err;
				}
4284 4285 4286 4287 4288 4289
			}
			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);
4290 4291 4292 4293 4294 4295 4296 4297 4298

			/* is this a reference to our own snapshot? If so
			 * skip it
			 */
			if (location.type == BTRFS_ROOT_ITEM_KEY &&
			    location.objectid == root->root_key.objectid) {
				over = 0;
				goto skip;
			}
4299
			over = filldir(dirent, name_ptr, name_len,
4300
				       found_key.offset, location.objectid,
C
Chris Mason 已提交
4301
				       d_type);
4302

4303
skip:
4304 4305 4306
			if (name_ptr != tmp_name)
				kfree(name_ptr);

C
Chris Mason 已提交
4307 4308
			if (over)
				goto nopos;
J
Josef Bacik 已提交
4309
			di_len = btrfs_dir_name_len(leaf, di) +
4310
				 btrfs_dir_data_len(leaf, di) + sizeof(*di);
C
Chris Mason 已提交
4311 4312 4313 4314
			di_cur += di_len;
			di = (struct btrfs_dir_item *)((char *)di + di_len);
		}
	}
4315 4316

	/* Reached end of directory/root. Bump pos past the last item. */
4317
	if (key_type == BTRFS_DIR_INDEX_KEY)
4318 4319 4320 4321 4322
		/*
		 * 32-bit glibc will use getdents64, but then strtol -
		 * so the last number we can serve is this.
		 */
		filp->f_pos = 0x7fffffff;
4323 4324
	else
		filp->f_pos++;
C
Chris Mason 已提交
4325 4326 4327 4328 4329 4330 4331
nopos:
	ret = 0;
err:
	btrfs_free_path(path);
	return ret;
}

4332
int btrfs_write_inode(struct inode *inode, struct writeback_control *wbc)
C
Chris Mason 已提交
4333 4334 4335 4336
{
	struct btrfs_root *root = BTRFS_I(inode)->root;
	struct btrfs_trans_handle *trans;
	int ret = 0;
4337
	bool nolock = false;
C
Chris Mason 已提交
4338

4339
	if (BTRFS_I(inode)->dummy_inode)
4340 4341
		return 0;

4342 4343 4344
	smp_mb();
	nolock = (root->fs_info->closing && root == root->fs_info->tree_root);

4345
	if (wbc->sync_mode == WB_SYNC_ALL) {
4346 4347 4348 4349
		if (nolock)
			trans = btrfs_join_transaction_nolock(root, 1);
		else
			trans = btrfs_join_transaction(root, 1);
4350 4351
		if (IS_ERR(trans))
			return PTR_ERR(trans);
C
Chris Mason 已提交
4352
		btrfs_set_trans_block_group(trans, inode);
4353 4354 4355 4356
		if (nolock)
			ret = btrfs_end_transaction_nolock(trans, root);
		else
			ret = btrfs_commit_transaction(trans, root);
C
Chris Mason 已提交
4357 4358 4359 4360 4361
	}
	return ret;
}

/*
4362
 * This is somewhat expensive, updating the tree every time the
C
Chris Mason 已提交
4363 4364 4365 4366 4367 4368 4369 4370
 * inode changes.  But, it is most likely to find the inode in cache.
 * FIXME, needs more benchmarking...there are no reasons other than performance
 * to keep or drop this code.
 */
void btrfs_dirty_inode(struct inode *inode)
{
	struct btrfs_root *root = BTRFS_I(inode)->root;
	struct btrfs_trans_handle *trans;
4371 4372 4373 4374
	int ret;

	if (BTRFS_I(inode)->dummy_inode)
		return;
C
Chris Mason 已提交
4375

4376
	trans = btrfs_join_transaction(root, 1);
4377
	BUG_ON(IS_ERR(trans));
C
Chris Mason 已提交
4378
	btrfs_set_trans_block_group(trans, inode);
4379 4380

	ret = btrfs_update_inode(trans, root, inode);
4381 4382 4383 4384
	if (ret && ret == -ENOSPC) {
		/* whoops, lets try again with the full transaction */
		btrfs_end_transaction(trans, root);
		trans = btrfs_start_transaction(root, 1);
4385 4386 4387 4388 4389 4390 4391 4392
		if (IS_ERR(trans)) {
			if (printk_ratelimit()) {
				printk(KERN_ERR "btrfs: fail to "
				       "dirty  inode %lu error %ld\n",
				       inode->i_ino, PTR_ERR(trans));
			}
			return;
		}
4393
		btrfs_set_trans_block_group(trans, inode);
4394

4395 4396
		ret = btrfs_update_inode(trans, root, inode);
		if (ret) {
4397 4398 4399 4400 4401
			if (printk_ratelimit()) {
				printk(KERN_ERR "btrfs: fail to "
				       "dirty  inode %lu error %d\n",
				       inode->i_ino, ret);
			}
4402 4403
		}
	}
C
Chris Mason 已提交
4404 4405 4406
	btrfs_end_transaction(trans, root);
}

C
Chris Mason 已提交
4407 4408 4409 4410 4411
/*
 * find the highest existing sequence number in a directory
 * and then set the in-memory index_cnt variable to reflect
 * free sequence numbers
 */
4412 4413 4414 4415 4416 4417 4418 4419 4420 4421 4422 4423 4424 4425 4426 4427 4428 4429 4430 4431 4432 4433 4434 4435 4436 4437 4438 4439 4440 4441 4442 4443 4444 4445 4446 4447 4448 4449 4450 4451 4452 4453 4454 4455 4456 4457 4458 4459 4460 4461 4462 4463
static int btrfs_set_inode_index_count(struct inode *inode)
{
	struct btrfs_root *root = BTRFS_I(inode)->root;
	struct btrfs_key key, found_key;
	struct btrfs_path *path;
	struct extent_buffer *leaf;
	int ret;

	key.objectid = inode->i_ino;
	btrfs_set_key_type(&key, BTRFS_DIR_INDEX_KEY);
	key.offset = (u64)-1;

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

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

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

	path->slots[0]--;

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

	if (found_key.objectid != inode->i_ino ||
	    btrfs_key_type(&found_key) != BTRFS_DIR_INDEX_KEY) {
		BTRFS_I(inode)->index_cnt = 2;
		goto out;
	}

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

C
Chris Mason 已提交
4464 4465 4466 4467
/*
 * 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
 */
4468
int btrfs_set_inode_index(struct inode *dir, u64 *index)
4469 4470 4471 4472 4473
{
	int ret = 0;

	if (BTRFS_I(dir)->index_cnt == (u64)-1) {
		ret = btrfs_set_inode_index_count(dir);
C
Chris Mason 已提交
4474
		if (ret)
4475 4476 4477
			return ret;
	}

4478
	*index = BTRFS_I(dir)->index_cnt;
4479 4480 4481 4482 4483
	BTRFS_I(dir)->index_cnt++;

	return ret;
}

C
Chris Mason 已提交
4484 4485
static struct inode *btrfs_new_inode(struct btrfs_trans_handle *trans,
				     struct btrfs_root *root,
4486
				     struct inode *dir,
4487
				     const char *name, int name_len,
4488 4489
				     u64 ref_objectid, u64 objectid,
				     u64 alloc_hint, int mode, u64 *index)
C
Chris Mason 已提交
4490 4491
{
	struct inode *inode;
4492
	struct btrfs_inode_item *inode_item;
C
Chris Mason 已提交
4493
	struct btrfs_key *location;
4494
	struct btrfs_path *path;
4495 4496 4497 4498
	struct btrfs_inode_ref *ref;
	struct btrfs_key key[2];
	u32 sizes[2];
	unsigned long ptr;
C
Chris Mason 已提交
4499 4500 4501
	int ret;
	int owner;

4502 4503 4504
	path = btrfs_alloc_path();
	BUG_ON(!path);

C
Chris Mason 已提交
4505 4506 4507 4508
	inode = new_inode(root->fs_info->sb);
	if (!inode)
		return ERR_PTR(-ENOMEM);

4509
	if (dir) {
4510
		ret = btrfs_set_inode_index(dir, index);
4511 4512
		if (ret) {
			iput(inode);
4513
			return ERR_PTR(ret);
4514
		}
4515 4516 4517 4518 4519 4520 4521
	}
	/*
	 * 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
4522
	BTRFS_I(inode)->root = root;
4523
	BTRFS_I(inode)->generation = trans->transid;
4524
	inode->i_generation = BTRFS_I(inode)->generation;
J
Josef Bacik 已提交
4525
	btrfs_set_inode_space_info(root, inode);
4526

C
Chris Mason 已提交
4527 4528 4529 4530
	if (mode & S_IFDIR)
		owner = 0;
	else
		owner = 1;
4531 4532
	BTRFS_I(inode)->block_group =
			btrfs_find_block_group(root, 0, alloc_hint, owner);
4533 4534 4535 4536 4537 4538 4539 4540 4541 4542 4543 4544

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

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

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

4545
	path->leave_spinning = 1;
4546 4547
	ret = btrfs_insert_empty_items(trans, root, path, key, sizes, 2);
	if (ret != 0)
4548 4549
		goto fail;

4550
	inode_init_owner(inode, dir, mode);
C
Chris Mason 已提交
4551
	inode->i_ino = objectid;
4552
	inode_set_bytes(inode, 0);
C
Chris Mason 已提交
4553
	inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
4554 4555
	inode_item = btrfs_item_ptr(path->nodes[0], path->slots[0],
				  struct btrfs_inode_item);
4556
	fill_inode_item(trans, path->nodes[0], inode_item, inode);
4557 4558 4559 4560

	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);
4561
	btrfs_set_inode_ref_index(path->nodes[0], ref, *index);
4562 4563 4564
	ptr = (unsigned long)(ref + 1);
	write_extent_buffer(path->nodes[0], name, ptr, name_len);

4565 4566 4567
	btrfs_mark_buffer_dirty(path->nodes[0]);
	btrfs_free_path(path);

C
Chris Mason 已提交
4568 4569 4570 4571 4572
	location = &BTRFS_I(inode)->location;
	location->objectid = objectid;
	location->offset = 0;
	btrfs_set_key_type(location, BTRFS_INODE_ITEM_KEY);

4573 4574
	btrfs_inherit_iflags(inode, dir);

4575 4576 4577 4578 4579 4580 4581
	if ((mode & S_IFREG)) {
		if (btrfs_test_opt(root, NODATASUM))
			BTRFS_I(inode)->flags |= BTRFS_INODE_NODATASUM;
		if (btrfs_test_opt(root, NODATACOW))
			BTRFS_I(inode)->flags |= BTRFS_INODE_NODATACOW;
	}

C
Chris Mason 已提交
4582
	insert_inode_hash(inode);
4583
	inode_tree_add(inode);
C
Chris Mason 已提交
4584
	return inode;
4585
fail:
4586 4587
	if (dir)
		BTRFS_I(dir)->index_cnt--;
4588
	btrfs_free_path(path);
4589
	iput(inode);
4590
	return ERR_PTR(ret);
C
Chris Mason 已提交
4591 4592 4593 4594 4595 4596 4597
}

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
4598 4599 4600 4601 4602 4603
/*
 * 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.
 */
4604 4605 4606
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
4607
{
4608
	int ret = 0;
C
Chris Mason 已提交
4609
	struct btrfs_key key;
4610
	struct btrfs_root *root = BTRFS_I(parent_inode)->root;
4611

4612 4613 4614 4615 4616 4617 4618 4619 4620 4621 4622 4623 4624 4625 4626 4627 4628 4629
	if (unlikely(inode->i_ino == BTRFS_FIRST_FREE_OBJECTID)) {
		memcpy(&key, &BTRFS_I(inode)->root->root_key, sizeof(key));
	} else {
		key.objectid = inode->i_ino;
		btrfs_set_key_type(&key, BTRFS_INODE_ITEM_KEY);
		key.offset = 0;
	}

	if (unlikely(inode->i_ino == BTRFS_FIRST_FREE_OBJECTID)) {
		ret = btrfs_add_root_ref(trans, root->fs_info->tree_root,
					 key.objectid, root->root_key.objectid,
					 parent_inode->i_ino,
					 index, name, name_len);
	} else if (add_backref) {
		ret = btrfs_insert_inode_ref(trans, root,
					     name, name_len, inode->i_ino,
					     parent_inode->i_ino, index);
	}
C
Chris Mason 已提交
4630 4631

	if (ret == 0) {
4632 4633 4634 4635 4636
		ret = btrfs_insert_dir_item(trans, root, name, name_len,
					    parent_inode->i_ino, &key,
					    btrfs_inode_type(inode), index);
		BUG_ON(ret);

4637
		btrfs_i_size_write(parent_inode, parent_inode->i_size +
4638
				   name_len * 2);
4639
		parent_inode->i_mtime = parent_inode->i_ctime = CURRENT_TIME;
4640
		ret = btrfs_update_inode(trans, root, parent_inode);
C
Chris Mason 已提交
4641 4642 4643 4644 4645
	}
	return ret;
}

static int btrfs_add_nondir(struct btrfs_trans_handle *trans,
4646 4647
			    struct inode *dir, struct dentry *dentry,
			    struct inode *inode, int backref, u64 index)
C
Chris Mason 已提交
4648
{
4649 4650 4651
	int err = btrfs_add_link(trans, dir, inode,
				 dentry->d_name.name, dentry->d_name.len,
				 backref, index);
C
Chris Mason 已提交
4652 4653 4654 4655 4656 4657 4658 4659 4660
	if (!err) {
		d_instantiate(dentry, inode);
		return 0;
	}
	if (err > 0)
		err = -EEXIST;
	return err;
}

J
Josef Bacik 已提交
4661 4662 4663 4664 4665
static int btrfs_mknod(struct inode *dir, struct dentry *dentry,
			int mode, dev_t rdev)
{
	struct btrfs_trans_handle *trans;
	struct btrfs_root *root = BTRFS_I(dir)->root;
4666
	struct inode *inode = NULL;
J
Josef Bacik 已提交
4667 4668 4669
	int err;
	int drop_inode = 0;
	u64 objectid;
4670
	unsigned long nr = 0;
4671
	u64 index = 0;
J
Josef Bacik 已提交
4672 4673 4674 4675

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

4676 4677 4678 4679
	err = btrfs_find_free_objectid(NULL, root, dir->i_ino, &objectid);
	if (err)
		return err;

J
Josef Bacik 已提交
4680 4681 4682 4683 4684
	/*
	 * 2 for inode item and ref
	 * 2 for dir items
	 * 1 for xattr if selinux is on
	 */
4685 4686 4687
	trans = btrfs_start_transaction(root, 5);
	if (IS_ERR(trans))
		return PTR_ERR(trans);
4688

J
Josef Bacik 已提交
4689 4690
	btrfs_set_trans_block_group(trans, dir);

4691
	inode = btrfs_new_inode(trans, root, dir, dentry->d_name.name,
4692
				dentry->d_name.len, dir->i_ino, objectid,
4693
				BTRFS_I(dir)->block_group, mode, &index);
J
Josef Bacik 已提交
4694 4695 4696 4697
	err = PTR_ERR(inode);
	if (IS_ERR(inode))
		goto out_unlock;

4698
	err = btrfs_init_inode_security(trans, inode, dir);
J
Josef Bacik 已提交
4699 4700 4701 4702 4703
	if (err) {
		drop_inode = 1;
		goto out_unlock;
	}

J
Josef Bacik 已提交
4704
	btrfs_set_trans_block_group(trans, inode);
4705
	err = btrfs_add_nondir(trans, dir, dentry, inode, 0, index);
J
Josef Bacik 已提交
4706 4707 4708 4709 4710
	if (err)
		drop_inode = 1;
	else {
		inode->i_op = &btrfs_special_inode_operations;
		init_special_inode(inode, inode->i_mode, rdev);
4711
		btrfs_update_inode(trans, root, inode);
J
Josef Bacik 已提交
4712 4713 4714 4715
	}
	btrfs_update_inode_block_group(trans, inode);
	btrfs_update_inode_block_group(trans, dir);
out_unlock:
4716
	nr = trans->blocks_used;
4717
	btrfs_end_transaction_throttle(trans, root);
4718
	btrfs_btree_balance_dirty(root, nr);
J
Josef Bacik 已提交
4719 4720 4721 4722 4723 4724 4725
	if (drop_inode) {
		inode_dec_link_count(inode);
		iput(inode);
	}
	return err;
}

C
Chris Mason 已提交
4726 4727 4728 4729 4730
static int btrfs_create(struct inode *dir, struct dentry *dentry,
			int mode, struct nameidata *nd)
{
	struct btrfs_trans_handle *trans;
	struct btrfs_root *root = BTRFS_I(dir)->root;
4731
	struct inode *inode = NULL;
C
Chris Mason 已提交
4732
	int drop_inode = 0;
4733
	int err;
4734
	unsigned long nr = 0;
C
Chris Mason 已提交
4735
	u64 objectid;
4736
	u64 index = 0;
C
Chris Mason 已提交
4737

4738 4739 4740
	err = btrfs_find_free_objectid(NULL, root, dir->i_ino, &objectid);
	if (err)
		return err;
J
Josef Bacik 已提交
4741 4742 4743 4744 4745
	/*
	 * 2 for inode item and ref
	 * 2 for dir items
	 * 1 for xattr if selinux is on
	 */
4746 4747 4748
	trans = btrfs_start_transaction(root, 5);
	if (IS_ERR(trans))
		return PTR_ERR(trans);
J
Josef Bacik 已提交
4749

C
Chris Mason 已提交
4750 4751
	btrfs_set_trans_block_group(trans, dir);

4752
	inode = btrfs_new_inode(trans, root, dir, dentry->d_name.name,
4753 4754
				dentry->d_name.len, dir->i_ino, objectid,
				BTRFS_I(dir)->block_group, mode, &index);
C
Chris Mason 已提交
4755 4756 4757 4758
	err = PTR_ERR(inode);
	if (IS_ERR(inode))
		goto out_unlock;

4759
	err = btrfs_init_inode_security(trans, inode, dir);
J
Josef Bacik 已提交
4760 4761 4762 4763 4764
	if (err) {
		drop_inode = 1;
		goto out_unlock;
	}

C
Chris Mason 已提交
4765
	btrfs_set_trans_block_group(trans, inode);
4766
	err = btrfs_add_nondir(trans, dir, dentry, inode, 0, index);
C
Chris Mason 已提交
4767 4768 4769 4770
	if (err)
		drop_inode = 1;
	else {
		inode->i_mapping->a_ops = &btrfs_aops;
C
Chris Mason 已提交
4771
		inode->i_mapping->backing_dev_info = &root->fs_info->bdi;
C
Chris Mason 已提交
4772 4773
		inode->i_fop = &btrfs_file_operations;
		inode->i_op = &btrfs_file_inode_operations;
4774
		BTRFS_I(inode)->io_tree.ops = &btrfs_extent_io_ops;
C
Chris Mason 已提交
4775 4776 4777 4778
	}
	btrfs_update_inode_block_group(trans, inode);
	btrfs_update_inode_block_group(trans, dir);
out_unlock:
4779
	nr = trans->blocks_used;
4780
	btrfs_end_transaction_throttle(trans, root);
C
Chris Mason 已提交
4781 4782 4783 4784
	if (drop_inode) {
		inode_dec_link_count(inode);
		iput(inode);
	}
4785
	btrfs_btree_balance_dirty(root, nr);
C
Chris Mason 已提交
4786 4787 4788 4789 4790 4791 4792 4793 4794
	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;
4795
	u64 index;
4796
	unsigned long nr = 0;
C
Chris Mason 已提交
4797 4798 4799 4800 4801 4802
	int err;
	int drop_inode = 0;

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

4803 4804 4805 4806
	/* do not allow sys_link's with other subvols of the same device */
	if (root->objectid != BTRFS_I(inode)->root->objectid)
		return -EPERM;

J
Josef Bacik 已提交
4807
	btrfs_inc_nlink(inode);
4808
	inode->i_ctime = CURRENT_TIME;
J
Josef Bacik 已提交
4809

4810
	err = btrfs_set_inode_index(dir, &index);
4811 4812 4813
	if (err)
		goto fail;

4814
	/*
M
Miao Xie 已提交
4815
	 * 2 items for inode and inode ref
4816
	 * 2 items for dir items
M
Miao Xie 已提交
4817
	 * 1 item for parent inode
4818
	 */
M
Miao Xie 已提交
4819
	trans = btrfs_start_transaction(root, 5);
4820 4821 4822 4823
	if (IS_ERR(trans)) {
		err = PTR_ERR(trans);
		goto fail;
	}
4824

C
Chris Mason 已提交
4825
	btrfs_set_trans_block_group(trans, dir);
A
Al Viro 已提交
4826
	ihold(inode);
4827

4828
	err = btrfs_add_nondir(trans, dir, dentry, inode, 1, index);
4829

4830
	if (err) {
4831
		drop_inode = 1;
4832
	} else {
4833
		struct dentry *parent = dget_parent(dentry);
4834 4835 4836
		btrfs_update_inode_block_group(trans, dir);
		err = btrfs_update_inode(trans, root, inode);
		BUG_ON(err);
4837 4838
		btrfs_log_new_name(trans, inode, NULL, parent);
		dput(parent);
4839
	}
C
Chris Mason 已提交
4840

4841
	nr = trans->blocks_used;
4842
	btrfs_end_transaction_throttle(trans, root);
4843
fail:
C
Chris Mason 已提交
4844 4845 4846 4847
	if (drop_inode) {
		inode_dec_link_count(inode);
		iput(inode);
	}
4848
	btrfs_btree_balance_dirty(root, nr);
C
Chris Mason 已提交
4849 4850 4851 4852 4853
	return err;
}

static int btrfs_mkdir(struct inode *dir, struct dentry *dentry, int mode)
{
4854
	struct inode *inode = NULL;
C
Chris Mason 已提交
4855 4856 4857 4858
	struct btrfs_trans_handle *trans;
	struct btrfs_root *root = BTRFS_I(dir)->root;
	int err = 0;
	int drop_on_err = 0;
4859
	u64 objectid = 0;
4860
	u64 index = 0;
4861
	unsigned long nr = 1;
C
Chris Mason 已提交
4862

4863 4864 4865 4866
	err = btrfs_find_free_objectid(NULL, root, dir->i_ino, &objectid);
	if (err)
		return err;

J
Josef Bacik 已提交
4867 4868 4869 4870 4871
	/*
	 * 2 items for inode and ref
	 * 2 items for dir items
	 * 1 for xattr if selinux is on
	 */
4872 4873 4874
	trans = btrfs_start_transaction(root, 5);
	if (IS_ERR(trans))
		return PTR_ERR(trans);
J
Josef Bacik 已提交
4875
	btrfs_set_trans_block_group(trans, dir);
C
Chris Mason 已提交
4876

4877
	inode = btrfs_new_inode(trans, root, dir, dentry->d_name.name,
4878
				dentry->d_name.len, dir->i_ino, objectid,
4879 4880
				BTRFS_I(dir)->block_group, S_IFDIR | mode,
				&index);
C
Chris Mason 已提交
4881 4882 4883 4884
	if (IS_ERR(inode)) {
		err = PTR_ERR(inode);
		goto out_fail;
	}
4885

C
Chris Mason 已提交
4886
	drop_on_err = 1;
J
Josef Bacik 已提交
4887

4888
	err = btrfs_init_inode_security(trans, inode, dir);
J
Josef Bacik 已提交
4889 4890 4891
	if (err)
		goto out_fail;

C
Chris Mason 已提交
4892 4893 4894 4895
	inode->i_op = &btrfs_dir_inode_operations;
	inode->i_fop = &btrfs_dir_file_operations;
	btrfs_set_trans_block_group(trans, inode);

4896
	btrfs_i_size_write(inode, 0);
C
Chris Mason 已提交
4897 4898 4899
	err = btrfs_update_inode(trans, root, inode);
	if (err)
		goto out_fail;
4900

4901 4902
	err = btrfs_add_link(trans, dir, inode, dentry->d_name.name,
			     dentry->d_name.len, 0, index);
C
Chris Mason 已提交
4903 4904
	if (err)
		goto out_fail;
4905

C
Chris Mason 已提交
4906 4907 4908 4909 4910 4911
	d_instantiate(dentry, inode);
	drop_on_err = 0;
	btrfs_update_inode_block_group(trans, inode);
	btrfs_update_inode_block_group(trans, dir);

out_fail:
4912
	nr = trans->blocks_used;
4913
	btrfs_end_transaction_throttle(trans, root);
C
Chris Mason 已提交
4914 4915
	if (drop_on_err)
		iput(inode);
4916
	btrfs_btree_balance_dirty(root, nr);
C
Chris Mason 已提交
4917 4918 4919
	return err;
}

C
Chris Mason 已提交
4920 4921 4922 4923
/* 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.
 */
4924 4925
static int merge_extent_mapping(struct extent_map_tree *em_tree,
				struct extent_map *existing,
4926 4927
				struct extent_map *em,
				u64 map_start, u64 map_len)
4928 4929 4930
{
	u64 start_diff;

4931 4932 4933 4934
	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
4935 4936
	if (em->block_start < EXTENT_MAP_LAST_BYTE &&
	    !test_bit(EXTENT_FLAG_COMPRESSED, &em->flags)) {
4937
		em->block_start += start_diff;
C
Chris Mason 已提交
4938 4939
		em->block_len -= start_diff;
	}
4940
	return add_extent_mapping(em_tree, em);
4941 4942
}

C
Chris Mason 已提交
4943 4944 4945 4946 4947 4948 4949 4950 4951 4952 4953
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;
4954
	int compress_type;
C
Chris Mason 已提交
4955 4956

	WARN_ON(pg_offset != 0);
4957
	compress_type = btrfs_file_extent_compression(leaf, item);
C
Chris Mason 已提交
4958 4959 4960 4961 4962 4963 4964 4965
	max_size = btrfs_file_extent_ram_bytes(leaf, item);
	inline_size = btrfs_file_extent_inline_item_len(leaf,
					btrfs_item_nr(leaf, path->slots[0]));
	tmp = kmalloc(inline_size, GFP_NOFS);
	ptr = btrfs_file_extent_inline_start(item);

	read_extent_buffer(leaf, tmp, ptr, inline_size);

4966
	max_size = min_t(unsigned long, PAGE_CACHE_SIZE, max_size);
4967 4968
	ret = btrfs_decompress(compress_type, tmp, page,
			       extent_offset, inline_size, max_size);
C
Chris Mason 已提交
4969 4970 4971 4972 4973 4974 4975 4976 4977 4978 4979 4980
	if (ret) {
		char *kaddr = kmap_atomic(page, KM_USER0);
		unsigned long copy_size = min_t(u64,
				  PAGE_CACHE_SIZE - pg_offset,
				  max_size - extent_offset);
		memset(kaddr + pg_offset, 0, copy_size);
		kunmap_atomic(kaddr, KM_USER0);
	}
	kfree(tmp);
	return 0;
}

C
Chris Mason 已提交
4981 4982
/*
 * a bit scary, this does extent mapping from logical file offset to the disk.
C
Chris Mason 已提交
4983 4984
 * 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
4985 4986 4987 4988
 * 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
4989

4990
struct extent_map *btrfs_get_extent(struct inode *inode, struct page *page,
4991
				    size_t pg_offset, u64 start, u64 len,
4992 4993 4994 4995
				    int create)
{
	int ret;
	int err = 0;
4996
	u64 bytenr;
4997 4998 4999 5000
	u64 extent_start = 0;
	u64 extent_end = 0;
	u64 objectid = inode->i_ino;
	u32 found_type;
5001
	struct btrfs_path *path = NULL;
5002 5003
	struct btrfs_root *root = BTRFS_I(inode)->root;
	struct btrfs_file_extent_item *item;
5004 5005
	struct extent_buffer *leaf;
	struct btrfs_key found_key;
5006 5007
	struct extent_map *em = NULL;
	struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
5008
	struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
5009
	struct btrfs_trans_handle *trans = NULL;
5010
	int compress_type;
5011 5012

again:
5013
	read_lock(&em_tree->lock);
5014
	em = lookup_extent_mapping(em_tree, start, len);
5015 5016
	if (em)
		em->bdev = root->fs_info->fs_devices->latest_bdev;
5017
	read_unlock(&em_tree->lock);
5018

5019
	if (em) {
5020 5021 5022
		if (em->start > start || em->start + em->len <= start)
			free_extent_map(em);
		else if (em->block_start == EXTENT_MAP_INLINE && page)
5023 5024 5025
			free_extent_map(em);
		else
			goto out;
5026
	}
5027
	em = alloc_extent_map(GFP_NOFS);
5028
	if (!em) {
5029 5030
		err = -ENOMEM;
		goto out;
5031
	}
5032
	em->bdev = root->fs_info->fs_devices->latest_bdev;
5033
	em->start = EXTENT_MAP_HOLE;
5034
	em->orig_start = EXTENT_MAP_HOLE;
5035
	em->len = (u64)-1;
C
Chris Mason 已提交
5036
	em->block_len = (u64)-1;
5037 5038 5039 5040 5041 5042

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

5043 5044
	ret = btrfs_lookup_file_extent(trans, root, path,
				       objectid, start, trans != NULL);
5045 5046 5047 5048 5049 5050 5051 5052 5053 5054 5055
	if (ret < 0) {
		err = ret;
		goto out;
	}

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

5056 5057
	leaf = path->nodes[0];
	item = btrfs_item_ptr(leaf, path->slots[0],
5058 5059
			      struct btrfs_file_extent_item);
	/* are we inside the extent that was found? */
5060 5061 5062
	btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
	found_type = btrfs_key_type(&found_key);
	if (found_key.objectid != objectid ||
5063 5064 5065 5066
	    found_type != BTRFS_EXTENT_DATA_KEY) {
		goto not_found;
	}

5067 5068
	found_type = btrfs_file_extent_type(leaf, item);
	extent_start = found_key.offset;
5069
	compress_type = btrfs_file_extent_compression(leaf, item);
Y
Yan Zheng 已提交
5070 5071
	if (found_type == BTRFS_FILE_EXTENT_REG ||
	    found_type == BTRFS_FILE_EXTENT_PREALLOC) {
5072
		extent_end = extent_start +
5073
		       btrfs_file_extent_num_bytes(leaf, item);
Y
Yan Zheng 已提交
5074 5075 5076 5077 5078 5079 5080 5081 5082 5083 5084 5085 5086 5087
	} 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;
5088
			}
Y
Yan Zheng 已提交
5089 5090 5091
			if (ret > 0)
				goto not_found;
			leaf = path->nodes[0];
5092
		}
Y
Yan Zheng 已提交
5093 5094 5095 5096 5097 5098 5099 5100 5101 5102 5103
		btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
		if (found_key.objectid != objectid ||
		    found_key.type != BTRFS_EXTENT_DATA_KEY)
			goto not_found;
		if (start + len <= found_key.offset)
			goto not_found;
		em->start = start;
		em->len = found_key.offset - start;
		goto not_found_em;
	}

Y
Yan Zheng 已提交
5104 5105
	if (found_type == BTRFS_FILE_EXTENT_REG ||
	    found_type == BTRFS_FILE_EXTENT_PREALLOC) {
Y
Yan Zheng 已提交
5106 5107
		em->start = extent_start;
		em->len = extent_end - extent_start;
5108 5109
		em->orig_start = extent_start -
				 btrfs_file_extent_offset(leaf, item);
5110 5111
		bytenr = btrfs_file_extent_disk_bytenr(leaf, item);
		if (bytenr == 0) {
5112
			em->block_start = EXTENT_MAP_HOLE;
5113 5114
			goto insert;
		}
5115
		if (compress_type != BTRFS_COMPRESS_NONE) {
C
Chris Mason 已提交
5116
			set_bit(EXTENT_FLAG_COMPRESSED, &em->flags);
5117
			em->compress_type = compress_type;
C
Chris Mason 已提交
5118 5119 5120 5121 5122 5123 5124
			em->block_start = bytenr;
			em->block_len = btrfs_file_extent_disk_num_bytes(leaf,
									 item);
		} else {
			bytenr += btrfs_file_extent_offset(leaf, item);
			em->block_start = bytenr;
			em->block_len = em->len;
Y
Yan Zheng 已提交
5125 5126
			if (found_type == BTRFS_FILE_EXTENT_PREALLOC)
				set_bit(EXTENT_FLAG_PREALLOC, &em->flags);
C
Chris Mason 已提交
5127
		}
5128 5129
		goto insert;
	} else if (found_type == BTRFS_FILE_EXTENT_INLINE) {
5130
		unsigned long ptr;
5131
		char *map;
5132 5133 5134
		size_t size;
		size_t extent_offset;
		size_t copy_size;
5135

5136
		em->block_start = EXTENT_MAP_INLINE;
C
Chris Mason 已提交
5137
		if (!page || create) {
5138
			em->start = extent_start;
Y
Yan Zheng 已提交
5139
			em->len = extent_end - extent_start;
5140 5141
			goto out;
		}
5142

Y
Yan Zheng 已提交
5143 5144
		size = btrfs_file_extent_inline_len(leaf, item);
		extent_offset = page_offset(page) + pg_offset - extent_start;
5145
		copy_size = min_t(u64, PAGE_CACHE_SIZE - pg_offset,
5146 5147
				size - extent_offset);
		em->start = extent_start + extent_offset;
5148 5149
		em->len = (copy_size + root->sectorsize - 1) &
			~((u64)root->sectorsize - 1);
5150
		em->orig_start = EXTENT_MAP_INLINE;
5151
		if (compress_type) {
C
Chris Mason 已提交
5152
			set_bit(EXTENT_FLAG_COMPRESSED, &em->flags);
5153 5154
			em->compress_type = compress_type;
		}
5155
		ptr = btrfs_file_extent_inline_start(item) + extent_offset;
5156
		if (create == 0 && !PageUptodate(page)) {
5157 5158
			if (btrfs_file_extent_compression(leaf, item) !=
			    BTRFS_COMPRESS_NONE) {
C
Chris Mason 已提交
5159 5160 5161 5162 5163 5164 5165 5166
				ret = uncompress_inline(path, inode, page,
							pg_offset,
							extent_offset, item);
				BUG_ON(ret);
			} else {
				map = kmap(page);
				read_extent_buffer(leaf, map + pg_offset, ptr,
						   copy_size);
5167 5168 5169 5170 5171
				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
5172 5173
				kunmap(page);
			}
5174 5175
			flush_dcache_page(page);
		} else if (create && PageUptodate(page)) {
5176
			WARN_ON(1);
5177 5178 5179 5180 5181
			if (!trans) {
				kunmap(page);
				free_extent_map(em);
				em = NULL;
				btrfs_release_path(root, path);
5182
				trans = btrfs_join_transaction(root, 1);
5183 5184
				if (IS_ERR(trans))
					return ERR_CAST(trans);
5185 5186
				goto again;
			}
C
Chris Mason 已提交
5187
			map = kmap(page);
5188
			write_extent_buffer(leaf, map + pg_offset, ptr,
5189
					    copy_size);
C
Chris Mason 已提交
5190
			kunmap(page);
5191
			btrfs_mark_buffer_dirty(leaf);
5192
		}
5193 5194
		set_extent_uptodate(io_tree, em->start,
				    extent_map_end(em) - 1, GFP_NOFS);
5195 5196
		goto insert;
	} else {
C
Chris Mason 已提交
5197
		printk(KERN_ERR "btrfs unknown found_type %d\n", found_type);
5198 5199 5200 5201
		WARN_ON(1);
	}
not_found:
	em->start = start;
5202
	em->len = len;
5203
not_found_em:
5204
	em->block_start = EXTENT_MAP_HOLE;
Y
Yan Zheng 已提交
5205
	set_bit(EXTENT_FLAG_VACANCY, &em->flags);
5206 5207
insert:
	btrfs_release_path(root, path);
5208
	if (em->start > start || extent_map_end(em) <= start) {
C
Chris Mason 已提交
5209 5210 5211 5212 5213
		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);
5214 5215 5216
		err = -EIO;
		goto out;
	}
5217 5218

	err = 0;
5219
	write_lock(&em_tree->lock);
5220
	ret = add_extent_mapping(em_tree, em);
5221 5222 5223 5224
	/* 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
	 */
5225
	if (ret == -EEXIST) {
5226
		struct extent_map *existing;
5227 5228 5229

		ret = 0;

5230
		existing = lookup_extent_mapping(em_tree, start, len);
5231 5232 5233 5234 5235
		if (existing && (existing->start > start ||
		    existing->start + existing->len <= start)) {
			free_extent_map(existing);
			existing = NULL;
		}
5236 5237 5238 5239 5240
		if (!existing) {
			existing = lookup_extent_mapping(em_tree, em->start,
							 em->len);
			if (existing) {
				err = merge_extent_mapping(em_tree, existing,
5241 5242
							   em, start,
							   root->sectorsize);
5243 5244 5245 5246 5247 5248 5249 5250 5251 5252 5253 5254 5255
				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;
5256
			err = 0;
5257 5258
		}
	}
5259
	write_unlock(&em_tree->lock);
5260
out:
5261 5262
	if (path)
		btrfs_free_path(path);
5263 5264
	if (trans) {
		ret = btrfs_end_transaction(trans, root);
C
Chris Mason 已提交
5265
		if (!err)
5266 5267 5268 5269 5270 5271 5272 5273 5274
			err = ret;
	}
	if (err) {
		free_extent_map(em);
		return ERR_PTR(err);
	}
	return em;
}

5275 5276 5277 5278 5279 5280 5281 5282 5283 5284 5285 5286 5287 5288 5289 5290 5291 5292 5293 5294 5295 5296 5297 5298 5299 5300 5301 5302 5303 5304 5305 5306 5307 5308 5309 5310 5311 5312 5313 5314 5315 5316 5317 5318 5319 5320 5321 5322 5323 5324 5325 5326 5327 5328 5329 5330 5331 5332 5333 5334 5335 5336 5337 5338 5339 5340 5341 5342 5343 5344 5345 5346 5347 5348 5349 5350 5351 5352 5353 5354 5355 5356 5357 5358 5359 5360 5361 5362 5363 5364 5365 5366 5367 5368 5369 5370 5371 5372 5373 5374 5375 5376 5377 5378 5379 5380 5381 5382 5383 5384 5385 5386 5387 5388 5389 5390 5391 5392 5393 5394 5395 5396
struct extent_map *btrfs_get_extent_fiemap(struct inode *inode, struct page *page,
					   size_t pg_offset, u64 start, u64 len,
					   int create)
{
	struct extent_map *em;
	struct extent_map *hole_em = NULL;
	u64 range_start = start;
	u64 end;
	u64 found;
	u64 found_end;
	int err = 0;

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

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

	em = NULL;

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

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

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

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

		em = alloc_extent_map(GFP_NOFS);
		if (!em) {
			err = -ENOMEM;
			goto out;
		}
		/*
		 * when btrfs_get_extent can't find anything it
		 * returns one huge hole
		 *
		 * make sure what it found really fits our range, and
		 * adjust to make sure it is based on the start from
		 * the caller
		 */
		if (hole_em) {
			u64 calc_end = extent_map_end(hole_em);

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

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

5397 5398 5399 5400 5401 5402 5403 5404 5405 5406 5407 5408 5409 5410
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;
	struct extent_map *em;
	struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
	struct btrfs_key ins;
	u64 alloc_hint;
	int ret;

	btrfs_drop_extent_cache(inode, start, start + len - 1, 0);

	trans = btrfs_join_transaction(root, 0);
5411 5412
	if (IS_ERR(trans))
		return ERR_CAST(trans);
5413 5414 5415 5416 5417 5418 5419 5420 5421 5422 5423 5424 5425 5426 5427 5428 5429 5430 5431 5432 5433 5434 5435 5436 5437 5438 5439 5440 5441 5442 5443 5444 5445 5446 5447 5448 5449 5450 5451 5452 5453 5454 5455 5456 5457 5458

	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,
				   alloc_hint, (u64)-1, &ins, 1);
	if (ret) {
		em = ERR_PTR(ret);
		goto out;
	}

	em = alloc_extent_map(GFP_NOFS);
	if (!em) {
		em = ERR_PTR(-ENOMEM);
		goto out;
	}

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

	em->block_start = ins.objectid;
	em->block_len = ins.offset;
	em->bdev = root->fs_info->fs_devices->latest_bdev;
	set_bit(EXTENT_FLAG_PINNED, &em->flags);

	while (1) {
		write_lock(&em_tree->lock);
		ret = add_extent_mapping(em_tree, em);
		write_unlock(&em_tree->lock);
		if (ret != -EEXIST)
			break;
		btrfs_drop_extent_cache(inode, start, start + em->len - 1, 0);
	}

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

5459 5460 5461 5462 5463 5464 5465 5466 5467 5468 5469 5470 5471 5472 5473 5474 5475 5476 5477 5478 5479 5480 5481 5482 5483 5484 5485 5486 5487 5488 5489 5490 5491 5492 5493 5494 5495 5496 5497 5498 5499 5500 5501 5502 5503 5504 5505 5506 5507 5508 5509 5510 5511 5512 5513 5514 5515 5516 5517 5518 5519 5520 5521 5522 5523 5524 5525 5526 5527 5528 5529 5530 5531 5532 5533 5534 5535 5536 5537 5538 5539 5540 5541 5542 5543 5544 5545 5546 5547 5548 5549 5550 5551 5552 5553 5554 5555 5556 5557 5558
/*
 * 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;

	ret = btrfs_lookup_file_extent(trans, root, path, inode->i_ino,
				       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);
	if (key.objectid != inode->i_ino ||
	    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
	 */
	if (btrfs_cross_ref_exist(trans, root, inode->i_ino,
				  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;
}

5559 5560 5561 5562 5563 5564 5565
static int btrfs_get_blocks_direct(struct inode *inode, sector_t iblock,
				   struct buffer_head *bh_result, int create)
{
	struct extent_map *em;
	struct btrfs_root *root = BTRFS_I(inode)->root;
	u64 start = iblock << inode->i_blkbits;
	u64 len = bh_result->b_size;
5566
	struct btrfs_trans_handle *trans;
5567 5568 5569 5570 5571 5572 5573 5574 5575 5576 5577 5578 5579 5580 5581 5582 5583 5584 5585 5586 5587 5588 5589 5590 5591 5592 5593 5594 5595 5596 5597 5598 5599 5600 5601 5602 5603 5604 5605 5606 5607 5608 5609 5610

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

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

	/* Just a good old fashioned hole, return */
	if (!create && (em->block_start == EXTENT_MAP_HOLE ||
			test_bit(EXTENT_FLAG_PREALLOC, &em->flags))) {
		free_extent_map(em);
		/* DIO will do one hole at a time, so just unlock a sector */
		unlock_extent(&BTRFS_I(inode)->io_tree, start,
			      start + root->sectorsize - 1, GFP_NOFS);
		return 0;
	}

	/*
	 * We don't allocate a new extent in the following cases
	 *
	 * 1) The inode is marked as NODATACOW.  In this case we'll just use the
	 * existing extent.
	 * 2) The extent is marked as PREALLOC.  We're good to go here and can
	 * just use the extent.
	 *
	 */
5611 5612
	if (!create) {
		len = em->len - (start - em->start);
5613
		goto map;
5614
	}
5615 5616 5617 5618 5619 5620

	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;
5621
		u64 block_start;
5622 5623 5624 5625 5626

		if (test_bit(EXTENT_FLAG_PREALLOC, &em->flags))
			type = BTRFS_ORDERED_PREALLOC;
		else
			type = BTRFS_ORDERED_NOCOW;
5627
		len = min(len, em->len - (start - em->start));
5628
		block_start = em->block_start + (start - em->start);
5629 5630 5631 5632 5633 5634 5635

		/*
		 * 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
		 */
		trans = btrfs_join_transaction(root, 0);
5636
		if (IS_ERR(trans))
5637 5638 5639 5640 5641 5642 5643 5644 5645 5646 5647
			goto must_cow;

		if (can_nocow_odirect(trans, inode, start, len) == 1) {
			ret = btrfs_add_ordered_extent_dio(inode, start,
					   block_start, len, len, type);
			btrfs_end_transaction(trans, root);
			if (ret) {
				free_extent_map(em);
				return ret;
			}
			goto unlock;
5648
		}
5649
		btrfs_end_transaction(trans, root);
5650
	}
5651 5652 5653 5654 5655 5656 5657 5658 5659 5660 5661 5662
must_cow:
	/*
	 * this will cow the extent, reset the len in case we changed
	 * it above
	 */
	len = bh_result->b_size;
	free_extent_map(em);
	em = btrfs_new_extent_direct(inode, start, len);
	if (IS_ERR(em))
		return PTR_ERR(em);
	len = min(len, em->len - (start - em->start));
unlock:
5663 5664 5665
	clear_extent_bit(&BTRFS_I(inode)->io_tree, start, start + len - 1,
			  EXTENT_LOCKED | EXTENT_DELALLOC | EXTENT_DIRTY, 1,
			  0, NULL, GFP_NOFS);
5666 5667 5668
map:
	bh_result->b_blocknr = (em->block_start + (start - em->start)) >>
		inode->i_blkbits;
5669
	bh_result->b_size = len;
5670 5671 5672 5673 5674 5675 5676 5677 5678 5679 5680 5681 5682 5683 5684 5685 5686
	bh_result->b_bdev = em->bdev;
	set_buffer_mapped(bh_result);
	if (create && !test_bit(EXTENT_FLAG_PREALLOC, &em->flags))
		set_buffer_new(bh_result);

	free_extent_map(em);

	return 0;
}

struct btrfs_dio_private {
	struct inode *inode;
	u64 logical_offset;
	u64 disk_bytenr;
	u64 bytes;
	u32 *csums;
	void *private;
M
Miao Xie 已提交
5687 5688 5689 5690 5691 5692 5693 5694

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

	/* IO errors */
	int errors;

	struct bio *orig_bio;
5695 5696 5697 5698
};

static void btrfs_endio_direct_read(struct bio *bio, int err)
{
M
Miao Xie 已提交
5699
	struct btrfs_dio_private *dip = bio->bi_private;
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 5739 5740 5741 5742 5743 5744 5745 5746 5747 5748 5749 5750 5751 5752 5753 5754
	struct bio_vec *bvec_end = bio->bi_io_vec + bio->bi_vcnt - 1;
	struct bio_vec *bvec = bio->bi_io_vec;
	struct inode *inode = dip->inode;
	struct btrfs_root *root = BTRFS_I(inode)->root;
	u64 start;
	u32 *private = dip->csums;

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

			local_irq_save(flags);
			kaddr = kmap_atomic(page, KM_IRQ0);
			csum = btrfs_csum_data(root, kaddr + bvec->bv_offset,
					       csum, bvec->bv_len);
			btrfs_csum_final(csum, (char *)&csum);
			kunmap_atomic(kaddr, KM_IRQ0);
			local_irq_restore(flags);

			flush_dcache_page(bvec->bv_page);
			if (csum != *private) {
				printk(KERN_ERR "btrfs csum failed ino %lu off"
				      " %llu csum %u private %u\n",
				      inode->i_ino, (unsigned long long)start,
				      csum, *private);
				err = -EIO;
			}
		}

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

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

	kfree(dip->csums);
	kfree(dip);
	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_trans_handle *trans;
	struct btrfs_ordered_extent *ordered = NULL;
	struct extent_state *cached_state = NULL;
5755 5756
	u64 ordered_offset = dip->logical_offset;
	u64 ordered_bytes = dip->bytes;
5757 5758 5759 5760
	int ret;

	if (err)
		goto out_done;
5761 5762 5763 5764
again:
	ret = btrfs_dec_test_first_ordered_pending(inode, &ordered,
						   &ordered_offset,
						   ordered_bytes);
5765
	if (!ret)
5766
		goto out_test;
5767 5768 5769 5770

	BUG_ON(!ordered);

	trans = btrfs_join_transaction(root, 1);
5771
	if (IS_ERR(trans)) {
5772 5773 5774 5775 5776 5777 5778 5779 5780 5781 5782 5783 5784 5785 5786 5787 5788 5789 5790 5791 5792 5793 5794 5795 5796 5797 5798 5799 5800 5801 5802 5803 5804 5805 5806 5807 5808 5809 5810 5811 5812 5813 5814 5815 5816 5817 5818 5819 5820 5821 5822 5823 5824 5825
		err = -ENOMEM;
		goto out;
	}
	trans->block_rsv = &root->fs_info->delalloc_block_rsv;

	if (test_bit(BTRFS_ORDERED_NOCOW, &ordered->flags)) {
		ret = btrfs_ordered_update_i_size(inode, 0, ordered);
		if (!ret)
			ret = btrfs_update_inode(trans, root, inode);
		err = ret;
		goto out;
	}

	lock_extent_bits(&BTRFS_I(inode)->io_tree, ordered->file_offset,
			 ordered->file_offset + ordered->len - 1, 0,
			 &cached_state, GFP_NOFS);

	if (test_bit(BTRFS_ORDERED_PREALLOC, &ordered->flags)) {
		ret = btrfs_mark_extent_written(trans, inode,
						ordered->file_offset,
						ordered->file_offset +
						ordered->len);
		if (ret) {
			err = ret;
			goto out_unlock;
		}
	} else {
		ret = insert_reserved_file_extent(trans, inode,
						  ordered->file_offset,
						  ordered->start,
						  ordered->disk_len,
						  ordered->len,
						  ordered->len,
						  0, 0, 0,
						  BTRFS_FILE_EXTENT_REG);
		unpin_extent_cache(&BTRFS_I(inode)->extent_tree,
				   ordered->file_offset, ordered->len);
		if (ret) {
			err = ret;
			WARN_ON(1);
			goto out_unlock;
		}
	}

	add_pending_csums(trans, inode, ordered->file_offset, &ordered->list);
	btrfs_ordered_update_i_size(inode, 0, ordered);
	btrfs_update_inode(trans, root, inode);
out_unlock:
	unlock_extent_cached(&BTRFS_I(inode)->io_tree, ordered->file_offset,
			     ordered->file_offset + ordered->len - 1,
			     &cached_state, GFP_NOFS);
out:
	btrfs_delalloc_release_metadata(inode, ordered->len);
	btrfs_end_transaction(trans, root);
5826
	ordered_offset = ordered->file_offset + ordered->len;
5827 5828
	btrfs_put_ordered_extent(ordered);
	btrfs_put_ordered_extent(ordered);
5829 5830 5831 5832 5833 5834 5835 5836 5837 5838 5839

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;
		goto again;
	}
5840 5841 5842 5843 5844 5845 5846 5847
out_done:
	bio->bi_private = dip->private;

	kfree(dip->csums);
	kfree(dip);
	dio_end_io(bio, err);
}

5848 5849 5850 5851 5852 5853 5854 5855 5856 5857 5858
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);
	BUG_ON(ret);
	return 0;
}

M
Miao Xie 已提交
5859 5860 5861 5862 5863 5864
static void btrfs_end_dio_bio(struct bio *bio, int err)
{
	struct btrfs_dio_private *dip = bio->bi_private;

	if (err) {
		printk(KERN_ERR "btrfs direct IO failed ino %lu rw %lu "
J
Jan Beulich 已提交
5865 5866 5867
		      "sector %#Lx len %u err no %d\n",
		      dip->inode->i_ino, bio->bi_rw,
		      (unsigned long long)bio->bi_sector, bio->bi_size, err);
M
Miao Xie 已提交
5868 5869 5870 5871 5872 5873 5874 5875 5876 5877 5878 5879 5880 5881 5882 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 5916 5917 5918 5919 5920 5921 5922 5923 5924 5925 5926 5927 5928 5929 5930 5931 5932 5933 5934 5935 5936 5937 5938 5939 5940 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 6002 6003 6004 6005 6006 6007 6008 6009 6010 6011 6012 6013 6014 6015 6016 6017 6018 6019 6020 6021 6022 6023 6024 6025 6026 6027 6028
		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,
					 u32 *csums)
{
	int write = rw & REQ_WRITE;
	struct btrfs_root *root = BTRFS_I(inode)->root;
	int ret;

	bio_get(bio);
	ret = btrfs_bio_wq_end_io(root->fs_info, bio, 0);
	if (ret)
		goto err;

	if (write && !skip_sum) {
		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;
	} else if (!skip_sum)
		btrfs_lookup_bio_sums_dio(root, inode, bio,
					  file_offset, csums);

	ret = btrfs_map_bio(root, rw, bio, 0, 1);
err:
	bio_put(bio);
	return ret;
}

static int btrfs_submit_direct_hook(int rw, struct btrfs_dio_private *dip,
				    int skip_sum)
{
	struct inode *inode = dip->inode;
	struct btrfs_root *root = BTRFS_I(inode)->root;
	struct btrfs_mapping_tree *map_tree = &root->fs_info->mapping_tree;
	struct bio *bio;
	struct bio *orig_bio = dip->orig_bio;
	struct bio_vec *bvec = orig_bio->bi_io_vec;
	u64 start_sector = orig_bio->bi_sector;
	u64 file_offset = dip->logical_offset;
	u64 submit_len = 0;
	u64 map_length;
	int nr_pages = 0;
	u32 *csums = dip->csums;
	int ret = 0;

	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);

	map_length = orig_bio->bi_size;
	ret = btrfs_map_block(map_tree, READ, start_sector << 9,
			      &map_length, NULL, 0);
	if (ret) {
		bio_put(bio);
		return -EIO;
	}

	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,
						     csums);
			if (ret) {
				bio_put(bio);
				atomic_dec(&dip->pending_bios);
				goto out_err;
			}

			if (!skip_sum)
				csums = csums + nr_pages;
			start_sector += submit_len >> 9;
			file_offset += submit_len;

			submit_len = 0;
			nr_pages = 0;

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

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

	ret = __btrfs_submit_dio_bio(bio, inode, rw, file_offset, skip_sum,
				     csums);
	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;
}

6029 6030 6031 6032 6033 6034 6035
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;
6036
	int write = rw & REQ_WRITE;
6037 6038 6039 6040 6041 6042 6043 6044 6045 6046 6047 6048 6049 6050
	int ret = 0;

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

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

	if (!skip_sum) {
		dip->csums = kmalloc(sizeof(u32) * bio->bi_vcnt, GFP_NOFS);
		if (!dip->csums) {
D
Daniel J Blueman 已提交
6051
			kfree(dip);
6052 6053 6054 6055 6056 6057 6058 6059 6060 6061 6062 6063 6064 6065 6066
			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));

6067
	dip->disk_bytenr = (u64)bio->bi_sector << 9;
6068
	bio->bi_private = dip;
M
Miao Xie 已提交
6069 6070 6071
	dip->errors = 0;
	dip->orig_bio = bio;
	atomic_set(&dip->pending_bios, 0);
6072 6073 6074 6075 6076 6077

	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
6078 6079
	ret = btrfs_submit_direct_hook(rw, dip, skip_sum);
	if (!ret)
6080
		return;
6081 6082 6083 6084 6085 6086 6087
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;
6088
		ordered = btrfs_lookup_ordered_extent(inode, file_offset);
6089 6090 6091 6092 6093 6094 6095 6096 6097 6098
		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
6099 6100 6101 6102 6103 6104 6105 6106 6107 6108 6109 6110 6111 6112 6113 6114 6115 6116 6117 6118 6119 6120 6121 6122 6123 6124
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;
	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;
		if ((addr & blocksize_mask) || (size & blocksize_mask)) 
			goto out;
	}
	retval = 0;
out:
	return retval;
}
6125 6126 6127 6128
static ssize_t btrfs_direct_IO(int rw, struct kiocb *iocb,
			const struct iovec *iov, loff_t offset,
			unsigned long nr_segs)
{
6129 6130 6131
	struct file *file = iocb->ki_filp;
	struct inode *inode = file->f_mapping->host;
	struct btrfs_ordered_extent *ordered;
6132
	struct extent_state *cached_state = NULL;
6133 6134
	u64 lockstart, lockend;
	ssize_t ret;
6135 6136
	int writing = rw & WRITE;
	int write_bits = 0;
6137
	size_t count = iov_length(iov, nr_segs);
6138

C
Chris Mason 已提交
6139 6140 6141 6142 6143
	if (check_direct_IO(BTRFS_I(inode)->root, rw, iocb, iov,
			    offset, nr_segs)) {
		return 0;
	}

6144
	lockstart = offset;
6145 6146 6147 6148 6149 6150 6151
	lockend = offset + count - 1;

	if (writing) {
		ret = btrfs_delalloc_reserve_space(inode, count);
		if (ret)
			goto out;
	}
6152

6153
	while (1) {
6154 6155
		lock_extent_bits(&BTRFS_I(inode)->io_tree, lockstart, lockend,
				 0, &cached_state, GFP_NOFS);
6156 6157 6158 6159 6160 6161 6162 6163 6164
		/*
		 * 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);
		if (!ordered)
			break;
6165 6166
		unlock_extent_cached(&BTRFS_I(inode)->io_tree, lockstart, lockend,
				     &cached_state, GFP_NOFS);
6167 6168 6169 6170 6171
		btrfs_start_ordered_extent(inode, ordered, 1);
		btrfs_put_ordered_extent(ordered);
		cond_resched();
	}

6172 6173 6174 6175 6176 6177 6178 6179 6180 6181 6182 6183 6184 6185 6186 6187 6188 6189 6190 6191
	/*
	 * we don't use btrfs_set_extent_delalloc because we don't want
	 * the dirty or uptodate bits
	 */
	if (writing) {
		write_bits = EXTENT_DELALLOC | EXTENT_DO_ACCOUNTING;
		ret = set_extent_bit(&BTRFS_I(inode)->io_tree, lockstart, lockend,
				     EXTENT_DELALLOC, 0, NULL, &cached_state,
				     GFP_NOFS);
		if (ret) {
			clear_extent_bit(&BTRFS_I(inode)->io_tree, lockstart,
					 lockend, EXTENT_LOCKED | write_bits,
					 1, 0, &cached_state, GFP_NOFS);
			goto out;
		}
	}

	free_extent_state(cached_state);
	cached_state = NULL;

C
Chris Mason 已提交
6192 6193 6194 6195
	ret = __blockdev_direct_IO(rw, iocb, inode,
		   BTRFS_I(inode)->root->fs_info->fs_devices->latest_bdev,
		   iov, offset, nr_segs, btrfs_get_blocks_direct, NULL,
		   btrfs_submit_direct, 0);
6196 6197

	if (ret < 0 && ret != -EIOCBQUEUED) {
6198 6199 6200 6201
		clear_extent_bit(&BTRFS_I(inode)->io_tree, offset,
			      offset + iov_length(iov, nr_segs) - 1,
			      EXTENT_LOCKED | write_bits, 1, 0,
			      &cached_state, GFP_NOFS);
6202 6203 6204 6205 6206
	} else if (ret >= 0 && ret < iov_length(iov, nr_segs)) {
		/*
		 * We're falling back to buffered, unlock the section we didn't
		 * do IO on.
		 */
6207 6208 6209 6210
		clear_extent_bit(&BTRFS_I(inode)->io_tree, offset + ret,
			      offset + iov_length(iov, nr_segs) - 1,
			      EXTENT_LOCKED | write_bits, 1, 0,
			      &cached_state, GFP_NOFS);
6211
	}
6212 6213
out:
	free_extent_state(cached_state);
6214
	return ret;
6215 6216
}

Y
Yehuda Sadeh 已提交
6217 6218 6219
static int btrfs_fiemap(struct inode *inode, struct fiemap_extent_info *fieinfo,
		__u64 start, __u64 len)
{
6220
	return extent_fiemap(inode, fieinfo, start, len, btrfs_get_extent_fiemap);
Y
Yehuda Sadeh 已提交
6221 6222
}

6223
int btrfs_readpage(struct file *file, struct page *page)
C
Chris Mason 已提交
6224
{
6225 6226
	struct extent_io_tree *tree;
	tree = &BTRFS_I(page->mapping->host)->io_tree;
6227
	return extent_read_full_page(tree, page, btrfs_get_extent);
C
Chris Mason 已提交
6228
}
6229

6230
static int btrfs_writepage(struct page *page, struct writeback_control *wbc)
C
Chris Mason 已提交
6231
{
6232
	struct extent_io_tree *tree;
6233 6234 6235 6236 6237 6238 6239


	if (current->flags & PF_MEMALLOC) {
		redirty_page_for_writepage(wbc, page);
		unlock_page(page);
		return 0;
	}
6240
	tree = &BTRFS_I(page->mapping->host)->io_tree;
6241
	return extent_write_full_page(tree, page, btrfs_get_extent, wbc);
C
Chris Mason 已提交
6242 6243
}

6244 6245
int btrfs_writepages(struct address_space *mapping,
		     struct writeback_control *wbc)
C
Chris Mason 已提交
6246
{
6247
	struct extent_io_tree *tree;
6248

6249
	tree = &BTRFS_I(mapping->host)->io_tree;
C
Chris Mason 已提交
6250 6251 6252
	return extent_writepages(tree, mapping, btrfs_get_extent, wbc);
}

C
Chris Mason 已提交
6253 6254 6255 6256
static int
btrfs_readpages(struct file *file, struct address_space *mapping,
		struct list_head *pages, unsigned nr_pages)
{
6257 6258
	struct extent_io_tree *tree;
	tree = &BTRFS_I(mapping->host)->io_tree;
C
Chris Mason 已提交
6259 6260 6261
	return extent_readpages(tree, mapping, pages, nr_pages,
				btrfs_get_extent);
}
6262
static int __btrfs_releasepage(struct page *page, gfp_t gfp_flags)
C
Chris Mason 已提交
6263
{
6264 6265
	struct extent_io_tree *tree;
	struct extent_map_tree *map;
6266
	int ret;
6267

6268 6269
	tree = &BTRFS_I(page->mapping->host)->io_tree;
	map = &BTRFS_I(page->mapping->host)->extent_tree;
6270
	ret = try_release_extent_mapping(map, tree, page, gfp_flags);
6271 6272 6273 6274
	if (ret == 1) {
		ClearPagePrivate(page);
		set_page_private(page, 0);
		page_cache_release(page);
C
Chris Mason 已提交
6275
	}
6276
	return ret;
C
Chris Mason 已提交
6277 6278
}

6279 6280
static int btrfs_releasepage(struct page *page, gfp_t gfp_flags)
{
6281 6282
	if (PageWriteback(page) || PageDirty(page))
		return 0;
6283
	return __btrfs_releasepage(page, gfp_flags & GFP_NOFS);
6284 6285
}

6286
static void btrfs_invalidatepage(struct page *page, unsigned long offset)
C
Chris Mason 已提交
6287
{
6288
	struct extent_io_tree *tree;
6289
	struct btrfs_ordered_extent *ordered;
6290
	struct extent_state *cached_state = NULL;
6291 6292
	u64 page_start = page_offset(page);
	u64 page_end = page_start + PAGE_CACHE_SIZE - 1;
C
Chris Mason 已提交
6293

6294 6295 6296 6297 6298 6299 6300 6301

	/*
	 * 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
	 */
6302
	wait_on_page_writeback(page);
6303

6304
	tree = &BTRFS_I(page->mapping->host)->io_tree;
6305 6306 6307 6308
	if (offset) {
		btrfs_releasepage(page, GFP_NOFS);
		return;
	}
6309 6310
	lock_extent_bits(tree, page_start, page_end, 0, &cached_state,
			 GFP_NOFS);
6311 6312 6313
	ordered = btrfs_lookup_ordered_extent(page->mapping->host,
					   page_offset(page));
	if (ordered) {
6314 6315 6316 6317
		/*
		 * IO on this page will never be started, so we need
		 * to account for any ordered extents now
		 */
6318 6319
		clear_extent_bit(tree, page_start, page_end,
				 EXTENT_DIRTY | EXTENT_DELALLOC |
6320
				 EXTENT_LOCKED | EXTENT_DO_ACCOUNTING, 1, 0,
6321
				 &cached_state, GFP_NOFS);
6322 6323 6324 6325 6326 6327 6328 6329
		/*
		 * whoever cleared the private bit is responsible
		 * for the finish_ordered_io
		 */
		if (TestClearPagePrivate2(page)) {
			btrfs_finish_ordered_io(page->mapping->host,
						page_start, page_end);
		}
6330
		btrfs_put_ordered_extent(ordered);
6331 6332 6333
		cached_state = NULL;
		lock_extent_bits(tree, page_start, page_end, 0, &cached_state,
				 GFP_NOFS);
6334 6335
	}
	clear_extent_bit(tree, page_start, page_end,
6336
		 EXTENT_LOCKED | EXTENT_DIRTY | EXTENT_DELALLOC |
6337
		 EXTENT_DO_ACCOUNTING, 1, 1, &cached_state, GFP_NOFS);
6338 6339
	__btrfs_releasepage(page, GFP_NOFS);

C
Chris Mason 已提交
6340
	ClearPageChecked(page);
6341 6342 6343 6344 6345
	if (PagePrivate(page)) {
		ClearPagePrivate(page);
		set_page_private(page, 0);
		page_cache_release(page);
	}
C
Chris Mason 已提交
6346 6347
}

C
Chris Mason 已提交
6348 6349 6350 6351 6352 6353 6354 6355 6356 6357 6358 6359 6360 6361 6362
/*
 * 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.
 */
6363
int btrfs_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf)
C
Chris Mason 已提交
6364
{
6365
	struct page *page = vmf->page;
6366
	struct inode *inode = fdentry(vma->vm_file)->d_inode;
6367
	struct btrfs_root *root = BTRFS_I(inode)->root;
6368 6369
	struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
	struct btrfs_ordered_extent *ordered;
6370
	struct extent_state *cached_state = NULL;
6371 6372
	char *kaddr;
	unsigned long zero_start;
C
Chris Mason 已提交
6373
	loff_t size;
6374
	int ret;
6375
	u64 page_start;
6376
	u64 page_end;
C
Chris Mason 已提交
6377

6378
	ret  = btrfs_delalloc_reserve_space(inode, PAGE_CACHE_SIZE);
6379 6380 6381 6382 6383
	if (ret) {
		if (ret == -ENOMEM)
			ret = VM_FAULT_OOM;
		else /* -ENOSPC, -EIO, etc */
			ret = VM_FAULT_SIGBUS;
6384
		goto out;
6385
	}
6386

6387
	ret = VM_FAULT_NOPAGE; /* make the VM retry the fault */
6388
again:
C
Chris Mason 已提交
6389 6390
	lock_page(page);
	size = i_size_read(inode);
6391 6392
	page_start = page_offset(page);
	page_end = page_start + PAGE_CACHE_SIZE - 1;
6393

C
Chris Mason 已提交
6394
	if ((page->mapping != inode->i_mapping) ||
6395
	    (page_start >= size)) {
C
Chris Mason 已提交
6396 6397 6398
		/* page got truncated out from underneath us */
		goto out_unlock;
	}
6399 6400
	wait_on_page_writeback(page);

6401 6402
	lock_extent_bits(io_tree, page_start, page_end, 0, &cached_state,
			 GFP_NOFS);
6403 6404
	set_page_extent_mapped(page);

6405 6406 6407 6408
	/*
	 * we can't set the delalloc bits if there are pending ordered
	 * extents.  Drop our locks and wait for them to finish
	 */
6409 6410
	ordered = btrfs_lookup_ordered_extent(inode, page_start);
	if (ordered) {
6411 6412
		unlock_extent_cached(io_tree, page_start, page_end,
				     &cached_state, GFP_NOFS);
6413
		unlock_page(page);
6414
		btrfs_start_ordered_extent(inode, ordered, 1);
6415 6416 6417 6418
		btrfs_put_ordered_extent(ordered);
		goto again;
	}

J
Josef Bacik 已提交
6419 6420 6421 6422 6423 6424 6425
	/*
	 * 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.
	 */
6426
	clear_extent_bit(&BTRFS_I(inode)->io_tree, page_start, page_end,
6427
			  EXTENT_DIRTY | EXTENT_DELALLOC | EXTENT_DO_ACCOUNTING,
6428
			  0, 0, &cached_state, GFP_NOFS);
J
Josef Bacik 已提交
6429

6430 6431
	ret = btrfs_set_extent_delalloc(inode, page_start, page_end,
					&cached_state);
J
Josef Bacik 已提交
6432
	if (ret) {
6433 6434
		unlock_extent_cached(io_tree, page_start, page_end,
				     &cached_state, GFP_NOFS);
J
Josef Bacik 已提交
6435 6436 6437
		ret = VM_FAULT_SIGBUS;
		goto out_unlock;
	}
6438
	ret = 0;
C
Chris Mason 已提交
6439 6440

	/* page is wholly or partially inside EOF */
6441
	if (page_start + PAGE_CACHE_SIZE > size)
6442
		zero_start = size & ~PAGE_CACHE_MASK;
C
Chris Mason 已提交
6443
	else
6444
		zero_start = PAGE_CACHE_SIZE;
C
Chris Mason 已提交
6445

6446 6447 6448 6449 6450 6451
	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);
	}
6452
	ClearPageChecked(page);
6453
	set_page_dirty(page);
6454
	SetPageUptodate(page);
6455

6456 6457 6458
	BTRFS_I(inode)->last_trans = root->fs_info->generation;
	BTRFS_I(inode)->last_sub_trans = BTRFS_I(inode)->root->log_transid;

6459
	unlock_extent_cached(io_tree, page_start, page_end, &cached_state, GFP_NOFS);
C
Chris Mason 已提交
6460 6461

out_unlock:
6462 6463
	if (!ret)
		return VM_FAULT_LOCKED;
C
Chris Mason 已提交
6464
	unlock_page(page);
6465
	btrfs_delalloc_release_space(inode, PAGE_CACHE_SIZE);
6466
out:
C
Chris Mason 已提交
6467 6468 6469
	return ret;
}

6470
static int btrfs_truncate(struct inode *inode)
C
Chris Mason 已提交
6471 6472 6473
{
	struct btrfs_root *root = BTRFS_I(inode)->root;
	int ret;
6474
	int err = 0;
C
Chris Mason 已提交
6475
	struct btrfs_trans_handle *trans;
6476
	unsigned long nr;
6477
	u64 mask = root->sectorsize - 1;
C
Chris Mason 已提交
6478

6479 6480
	ret = btrfs_truncate_page(inode->i_mapping, inode->i_size);
	if (ret)
6481
		return ret;
6482

C
Chris Mason 已提交
6483
	btrfs_wait_ordered_range(inode, inode->i_size & (~mask), (u64)-1);
6484
	btrfs_ordered_update_i_size(inode, inode->i_size, NULL);
C
Chris Mason 已提交
6485

6486 6487 6488 6489 6490 6491 6492 6493 6494 6495 6496 6497 6498 6499 6500 6501 6502
	trans = btrfs_start_transaction(root, 5);
	if (IS_ERR(trans))
		return PTR_ERR(trans);

	btrfs_set_trans_block_group(trans, inode);

	ret = btrfs_orphan_add(trans, inode);
	if (ret) {
		btrfs_end_transaction(trans, root);
		return ret;
	}

	nr = trans->blocks_used;
	btrfs_end_transaction(trans, root);
	btrfs_btree_balance_dirty(root, nr);

	/* Now start a transaction for the truncate */
6503
	trans = btrfs_start_transaction(root, 0);
6504 6505
	if (IS_ERR(trans))
		return PTR_ERR(trans);
6506
	btrfs_set_trans_block_group(trans, inode);
6507
	trans->block_rsv = root->orphan_block_rsv;
6508 6509 6510 6511 6512 6513 6514 6515 6516 6517 6518 6519 6520 6521 6522 6523 6524 6525 6526 6527 6528

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

6529
	while (1) {
6530 6531
		if (!trans) {
			trans = btrfs_start_transaction(root, 0);
6532 6533
			if (IS_ERR(trans))
				return PTR_ERR(trans);
6534 6535 6536 6537 6538 6539
			btrfs_set_trans_block_group(trans, inode);
			trans->block_rsv = root->orphan_block_rsv;
		}

		ret = btrfs_block_rsv_check(trans, root,
					    root->orphan_block_rsv, 0, 5);
6540
		if (ret == -EAGAIN) {
6541
			ret = btrfs_commit_transaction(trans, root);
6542 6543
			if (ret)
				return ret;
6544 6545
			trans = NULL;
			continue;
6546 6547 6548
		} else if (ret) {
			err = ret;
			break;
6549 6550
		}

6551 6552 6553
		ret = btrfs_truncate_inode_items(trans, root, inode,
						 inode->i_size,
						 BTRFS_EXTENT_DATA_KEY);
6554 6555
		if (ret != -EAGAIN) {
			err = ret;
6556
			break;
6557
		}
C
Chris Mason 已提交
6558

6559
		ret = btrfs_update_inode(trans, root, inode);
6560 6561 6562 6563
		if (ret) {
			err = ret;
			break;
		}
6564

6565 6566
		nr = trans->blocks_used;
		btrfs_end_transaction(trans, root);
6567
		trans = NULL;
6568 6569 6570 6571 6572
		btrfs_btree_balance_dirty(root, nr);
	}

	if (ret == 0 && inode->i_nlink > 0) {
		ret = btrfs_orphan_del(trans, inode);
6573 6574
		if (ret)
			err = ret;
6575 6576 6577 6578 6579 6580
	} else if (ret && inode->i_nlink > 0) {
		/*
		 * Failed to do the truncate, remove us from the in memory
		 * orphan list.
		 */
		ret = btrfs_orphan_del(NULL, inode);
6581 6582 6583
	}

	ret = btrfs_update_inode(trans, root, inode);
6584 6585
	if (ret && !err)
		err = ret;
6586 6587

	nr = trans->blocks_used;
6588
	ret = btrfs_end_transaction_throttle(trans, root);
6589 6590
	if (ret && !err)
		err = ret;
6591
	btrfs_btree_balance_dirty(root, nr);
6592

6593
	return err;
C
Chris Mason 已提交
6594 6595
}

C
Chris Mason 已提交
6596 6597 6598
/*
 * create a new subvolume directory/inode (helper for the ioctl).
 */
6599
int btrfs_create_subvol_root(struct btrfs_trans_handle *trans,
6600
			     struct btrfs_root *new_root,
6601
			     u64 new_dirid, u64 alloc_hint)
C
Chris Mason 已提交
6602 6603
{
	struct inode *inode;
6604
	int err;
6605
	u64 index = 0;
C
Chris Mason 已提交
6606

6607
	inode = btrfs_new_inode(trans, new_root, NULL, "..", 2, new_dirid,
6608
				new_dirid, alloc_hint, S_IFDIR | 0700, &index);
6609
	if (IS_ERR(inode))
C
Christoph Hellwig 已提交
6610
		return PTR_ERR(inode);
C
Chris Mason 已提交
6611 6612 6613 6614
	inode->i_op = &btrfs_dir_inode_operations;
	inode->i_fop = &btrfs_dir_file_operations;

	inode->i_nlink = 1;
6615
	btrfs_i_size_write(inode, 0);
6616

6617 6618
	err = btrfs_update_inode(trans, new_root, inode);
	BUG_ON(err);
6619

6620
	iput(inode);
6621
	return 0;
C
Chris Mason 已提交
6622 6623
}

C
Chris Mason 已提交
6624 6625 6626
/* helper function for file defrag and space balancing.  This
 * forces readahead on a given range of bytes in an inode
 */
6627
unsigned long btrfs_force_ra(struct address_space *mapping,
6628 6629 6630
			      struct file_ra_state *ra, struct file *file,
			      pgoff_t offset, pgoff_t last_index)
{
6631
	pgoff_t req_size = last_index - offset + 1;
6632 6633 6634 6635 6636

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

C
Chris Mason 已提交
6637 6638 6639
struct inode *btrfs_alloc_inode(struct super_block *sb)
{
	struct btrfs_inode *ei;
Y
Yan, Zheng 已提交
6640
	struct inode *inode;
C
Chris Mason 已提交
6641 6642 6643 6644

	ei = kmem_cache_alloc(btrfs_inode_cachep, GFP_NOFS);
	if (!ei)
		return NULL;
Y
Yan, Zheng 已提交
6645 6646 6647 6648 6649

	ei->root = NULL;
	ei->space_info = NULL;
	ei->generation = 0;
	ei->sequence = 0;
6650
	ei->last_trans = 0;
6651
	ei->last_sub_trans = 0;
6652
	ei->logged_trans = 0;
Y
Yan, Zheng 已提交
6653 6654 6655 6656 6657 6658 6659
	ei->delalloc_bytes = 0;
	ei->reserved_bytes = 0;
	ei->disk_i_size = 0;
	ei->flags = 0;
	ei->index_cnt = (u64)-1;
	ei->last_unlink_trans = 0;

6660
	atomic_set(&ei->outstanding_extents, 0);
6661
	atomic_set(&ei->reserved_extents, 0);
Y
Yan, Zheng 已提交
6662 6663

	ei->ordered_data_close = 0;
6664
	ei->orphan_meta_reserved = 0;
Y
Yan, Zheng 已提交
6665
	ei->dummy_inode = 0;
6666
	ei->force_compress = BTRFS_COMPRESS_NONE;
Y
Yan, Zheng 已提交
6667 6668 6669 6670 6671 6672

	inode = &ei->vfs_inode;
	extent_map_tree_init(&ei->extent_tree, GFP_NOFS);
	extent_io_tree_init(&ei->io_tree, &inode->i_data, GFP_NOFS);
	extent_io_tree_init(&ei->io_failure_tree, &inode->i_data, GFP_NOFS);
	mutex_init(&ei->log_mutex);
6673
	btrfs_ordered_inode_tree_init(&ei->ordered_tree);
6674
	INIT_LIST_HEAD(&ei->i_orphan);
Y
Yan, Zheng 已提交
6675
	INIT_LIST_HEAD(&ei->delalloc_inodes);
6676
	INIT_LIST_HEAD(&ei->ordered_operations);
Y
Yan, Zheng 已提交
6677 6678 6679
	RB_CLEAR_NODE(&ei->rb_node);

	return inode;
C
Chris Mason 已提交
6680 6681
}

N
Nick Piggin 已提交
6682 6683 6684 6685 6686 6687 6688
static void btrfs_i_callback(struct rcu_head *head)
{
	struct inode *inode = container_of(head, struct inode, i_rcu);
	INIT_LIST_HEAD(&inode->i_dentry);
	kmem_cache_free(btrfs_inode_cachep, BTRFS_I(inode));
}

C
Chris Mason 已提交
6689 6690
void btrfs_destroy_inode(struct inode *inode)
{
6691
	struct btrfs_ordered_extent *ordered;
6692 6693
	struct btrfs_root *root = BTRFS_I(inode)->root;

C
Chris Mason 已提交
6694 6695
	WARN_ON(!list_empty(&inode->i_dentry));
	WARN_ON(inode->i_data.nrpages);
6696
	WARN_ON(atomic_read(&BTRFS_I(inode)->outstanding_extents));
6697
	WARN_ON(atomic_read(&BTRFS_I(inode)->reserved_extents));
C
Chris Mason 已提交
6698

6699 6700 6701 6702 6703 6704 6705 6706
	/*
	 * 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;

6707 6708 6709 6710 6711 6712 6713 6714 6715 6716 6717
	/*
	 * 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);
	}

6718 6719 6720 6721 6722 6723 6724 6725 6726 6727 6728 6729 6730 6731 6732
	if (root == root->fs_info->tree_root) {
		struct btrfs_block_group_cache *block_group;

		block_group = btrfs_lookup_block_group(root->fs_info,
						BTRFS_I(inode)->block_group);
		if (block_group && block_group->inode == inode) {
			spin_lock(&block_group->lock);
			block_group->inode = NULL;
			spin_unlock(&block_group->lock);
			btrfs_put_block_group(block_group);
		} else if (block_group) {
			btrfs_put_block_group(block_group);
		}
	}

6733
	spin_lock(&root->orphan_lock);
6734
	if (!list_empty(&BTRFS_I(inode)->i_orphan)) {
6735 6736 6737
		printk(KERN_INFO "BTRFS: inode %lu still on the orphan list\n",
		       inode->i_ino);
		list_del_init(&BTRFS_I(inode)->i_orphan);
6738
	}
6739
	spin_unlock(&root->orphan_lock);
6740

C
Chris Mason 已提交
6741
	while (1) {
6742 6743 6744 6745
		ordered = btrfs_lookup_first_ordered_extent(inode, (u64)-1);
		if (!ordered)
			break;
		else {
C
Chris Mason 已提交
6746 6747 6748 6749
			printk(KERN_ERR "btrfs found ordered "
			       "extent %llu %llu on inode cleanup\n",
			       (unsigned long long)ordered->file_offset,
			       (unsigned long long)ordered->len);
6750 6751 6752 6753 6754
			btrfs_remove_ordered_extent(inode, ordered);
			btrfs_put_ordered_extent(ordered);
			btrfs_put_ordered_extent(ordered);
		}
	}
6755
	inode_tree_del(inode);
6756
	btrfs_drop_extent_cache(inode, 0, (u64)-1, 0);
6757
free:
N
Nick Piggin 已提交
6758
	call_rcu(&inode->i_rcu, btrfs_i_callback);
C
Chris Mason 已提交
6759 6760
}

6761
int btrfs_drop_inode(struct inode *inode)
6762 6763
{
	struct btrfs_root *root = BTRFS_I(inode)->root;
6764

6765 6766
	if (btrfs_root_refs(&root->root_item) == 0 &&
	    root != root->fs_info->tree_root)
6767
		return 1;
6768
	else
6769
		return generic_drop_inode(inode);
6770 6771
}

6772
static void init_once(void *foo)
C
Chris Mason 已提交
6773 6774 6775 6776 6777 6778 6779 6780 6781 6782 6783 6784 6785 6786 6787 6788
{
	struct btrfs_inode *ei = (struct btrfs_inode *) foo;

	inode_init_once(&ei->vfs_inode);
}

void btrfs_destroy_cachep(void)
{
	if (btrfs_inode_cachep)
		kmem_cache_destroy(btrfs_inode_cachep);
	if (btrfs_trans_handle_cachep)
		kmem_cache_destroy(btrfs_trans_handle_cachep);
	if (btrfs_transaction_cachep)
		kmem_cache_destroy(btrfs_transaction_cachep);
	if (btrfs_path_cachep)
		kmem_cache_destroy(btrfs_path_cachep);
6789 6790
	if (btrfs_free_space_cachep)
		kmem_cache_destroy(btrfs_free_space_cachep);
C
Chris Mason 已提交
6791 6792 6793 6794
}

int btrfs_init_cachep(void)
{
6795 6796 6797
	btrfs_inode_cachep = kmem_cache_create("btrfs_inode_cache",
			sizeof(struct btrfs_inode), 0,
			SLAB_RECLAIM_ACCOUNT | SLAB_MEM_SPREAD, init_once);
C
Chris Mason 已提交
6798 6799
	if (!btrfs_inode_cachep)
		goto fail;
6800 6801 6802 6803

	btrfs_trans_handle_cachep = kmem_cache_create("btrfs_trans_handle_cache",
			sizeof(struct btrfs_trans_handle), 0,
			SLAB_RECLAIM_ACCOUNT | SLAB_MEM_SPREAD, NULL);
C
Chris Mason 已提交
6804 6805
	if (!btrfs_trans_handle_cachep)
		goto fail;
6806 6807 6808 6809

	btrfs_transaction_cachep = kmem_cache_create("btrfs_transaction_cache",
			sizeof(struct btrfs_transaction), 0,
			SLAB_RECLAIM_ACCOUNT | SLAB_MEM_SPREAD, NULL);
C
Chris Mason 已提交
6810 6811
	if (!btrfs_transaction_cachep)
		goto fail;
6812 6813 6814 6815

	btrfs_path_cachep = kmem_cache_create("btrfs_path_cache",
			sizeof(struct btrfs_path), 0,
			SLAB_RECLAIM_ACCOUNT | SLAB_MEM_SPREAD, NULL);
C
Chris Mason 已提交
6816 6817
	if (!btrfs_path_cachep)
		goto fail;
6818

6819 6820 6821 6822 6823 6824
	btrfs_free_space_cachep = kmem_cache_create("btrfs_free_space_cache",
			sizeof(struct btrfs_free_space), 0,
			SLAB_RECLAIM_ACCOUNT | SLAB_MEM_SPREAD, NULL);
	if (!btrfs_free_space_cachep)
		goto fail;

C
Chris Mason 已提交
6825 6826 6827 6828 6829 6830 6831 6832 6833 6834 6835
	return 0;
fail:
	btrfs_destroy_cachep();
	return -ENOMEM;
}

static int btrfs_getattr(struct vfsmount *mnt,
			 struct dentry *dentry, struct kstat *stat)
{
	struct inode *inode = dentry->d_inode;
	generic_fillattr(inode, stat);
6836
	stat->dev = BTRFS_I(inode)->root->anon_super.s_dev;
C
Chris Mason 已提交
6837
	stat->blksize = PAGE_CACHE_SIZE;
6838 6839
	stat->blocks = (inode_get_bytes(inode) +
			BTRFS_I(inode)->delalloc_bytes) >> 9;
C
Chris Mason 已提交
6840 6841 6842
	return 0;
}

C
Chris Mason 已提交
6843 6844
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
6845 6846 6847
{
	struct btrfs_trans_handle *trans;
	struct btrfs_root *root = BTRFS_I(old_dir)->root;
6848
	struct btrfs_root *dest = BTRFS_I(new_dir)->root;
C
Chris Mason 已提交
6849 6850 6851
	struct inode *new_inode = new_dentry->d_inode;
	struct inode *old_inode = old_dentry->d_inode;
	struct timespec ctime = CURRENT_TIME;
6852
	u64 index = 0;
6853
	u64 root_objectid;
C
Chris Mason 已提交
6854 6855
	int ret;

6856 6857 6858
	if (new_dir->i_ino == BTRFS_EMPTY_SUBVOL_DIR_OBJECTID)
		return -EPERM;

6859 6860
	/* we only allow rename subvolume link between subvolumes */
	if (old_inode->i_ino != BTRFS_FIRST_FREE_OBJECTID && root != dest)
6861 6862
		return -EXDEV;

6863 6864
	if (old_inode->i_ino == BTRFS_EMPTY_SUBVOL_DIR_OBJECTID ||
	    (new_inode && new_inode->i_ino == BTRFS_FIRST_FREE_OBJECTID))
C
Chris Mason 已提交
6865
		return -ENOTEMPTY;
6866

6867 6868 6869
	if (S_ISDIR(old_inode->i_mode) && new_inode &&
	    new_inode->i_size > BTRFS_EMPTY_DIR_SIZE)
		return -ENOTEMPTY;
6870 6871 6872 6873 6874
	/*
	 * 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
	 */
6875
	if (new_inode && S_ISREG(old_inode->i_mode) && new_inode->i_size &&
6876 6877 6878
	    old_inode->i_size > BTRFS_ORDERED_OPERATIONS_FLUSH_LIMIT)
		filemap_flush(old_inode->i_mapping);

6879 6880 6881
	/* close the racy window with snapshot create/destroy ioctl */
	if (old_inode->i_ino == BTRFS_FIRST_FREE_OBJECTID)
		down_read(&root->fs_info->subvol_sem);
6882 6883 6884 6885 6886 6887 6888 6889 6890 6891 6892
	/*
	 * 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);
	if (IS_ERR(trans))
		return PTR_ERR(trans);
6893

6894
	btrfs_set_trans_block_group(trans, new_dir);
6895

6896 6897
	if (dest != root)
		btrfs_record_root_in_trans(trans, dest);
6898

6899 6900 6901
	ret = btrfs_set_inode_index(new_dir, &index);
	if (ret)
		goto out_fail;
6902

6903
	if (unlikely(old_inode->i_ino == BTRFS_FIRST_FREE_OBJECTID)) {
6904 6905 6906
		/* force full log commit if subvolume involved. */
		root->fs_info->last_trans_log_full_commit = trans->transid;
	} else {
6907 6908 6909 6910 6911 6912 6913
		ret = btrfs_insert_inode_ref(trans, dest,
					     new_dentry->d_name.name,
					     new_dentry->d_name.len,
					     old_inode->i_ino,
					     new_dir->i_ino, index);
		if (ret)
			goto out_fail;
6914 6915 6916 6917 6918 6919 6920 6921 6922
		/*
		 * 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);
	}
6923 6924 6925 6926 6927 6928 6929 6930 6931
	/*
	 * make sure the inode gets flushed if it is replacing
	 * something.
	 */
	if (new_inode && new_inode->i_size &&
	    old_inode && S_ISREG(old_inode->i_mode)) {
		btrfs_add_ordered_operation(trans, root, old_inode);
	}

C
Chris Mason 已提交
6932 6933 6934
	old_dir->i_ctime = old_dir->i_mtime = ctime;
	new_dir->i_ctime = new_dir->i_mtime = ctime;
	old_inode->i_ctime = ctime;
6935

6936 6937 6938
	if (old_dentry->d_parent != new_dentry->d_parent)
		btrfs_record_unlink_dir(trans, old_dir, old_inode, 1);

6939 6940 6941 6942 6943 6944 6945 6946 6947 6948 6949 6950 6951
	if (unlikely(old_inode->i_ino == BTRFS_FIRST_FREE_OBJECTID)) {
		root_objectid = BTRFS_I(old_inode)->root->root_key.objectid;
		ret = btrfs_unlink_subvol(trans, root, old_dir, root_objectid,
					old_dentry->d_name.name,
					old_dentry->d_name.len);
	} else {
		btrfs_inc_nlink(old_dentry->d_inode);
		ret = btrfs_unlink_inode(trans, root, old_dir,
					 old_dentry->d_inode,
					 old_dentry->d_name.name,
					 old_dentry->d_name.len);
	}
	BUG_ON(ret);
C
Chris Mason 已提交
6952 6953 6954

	if (new_inode) {
		new_inode->i_ctime = CURRENT_TIME;
6955 6956 6957 6958 6959 6960 6961 6962 6963 6964 6965 6966 6967 6968 6969
		if (unlikely(new_inode->i_ino ==
			     BTRFS_EMPTY_SUBVOL_DIR_OBJECTID)) {
			root_objectid = BTRFS_I(new_inode)->location.objectid;
			ret = btrfs_unlink_subvol(trans, dest, new_dir,
						root_objectid,
						new_dentry->d_name.name,
						new_dentry->d_name.len);
			BUG_ON(new_inode->i_nlink == 0);
		} else {
			ret = btrfs_unlink_inode(trans, dest, new_dir,
						 new_dentry->d_inode,
						 new_dentry->d_name.name,
						 new_dentry->d_name.len);
		}
		BUG_ON(ret);
6970
		if (new_inode->i_nlink == 0) {
6971
			ret = btrfs_orphan_add(trans, new_dentry->d_inode);
6972
			BUG_ON(ret);
6973
		}
C
Chris Mason 已提交
6974
	}
6975

6976 6977
	ret = btrfs_add_link(trans, new_dir, old_inode,
			     new_dentry->d_name.name,
6978
			     new_dentry->d_name.len, 0, index);
6979
	BUG_ON(ret);
C
Chris Mason 已提交
6980

6981
	if (old_inode->i_ino != BTRFS_FIRST_FREE_OBJECTID) {
6982 6983 6984
		struct dentry *parent = dget_parent(new_dentry);
		btrfs_log_new_name(trans, old_inode, old_dir, parent);
		dput(parent);
6985 6986
		btrfs_end_log_trans(root);
	}
C
Chris Mason 已提交
6987
out_fail:
6988
	btrfs_end_transaction_throttle(trans, root);
6989

6990 6991
	if (old_inode->i_ino == BTRFS_FIRST_FREE_OBJECTID)
		up_read(&root->fs_info->subvol_sem);
J
Josef Bacik 已提交
6992

C
Chris Mason 已提交
6993 6994 6995
	return ret;
}

C
Chris Mason 已提交
6996 6997 6998 6999
/*
 * 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
7000
int btrfs_start_delalloc_inodes(struct btrfs_root *root, int delay_iput)
7001 7002 7003
{
	struct list_head *head = &root->fs_info->delalloc_inodes;
	struct btrfs_inode *binode;
7004
	struct inode *inode;
7005

Y
Yan Zheng 已提交
7006 7007 7008
	if (root->fs_info->sb->s_flags & MS_RDONLY)
		return -EROFS;

7009
	spin_lock(&root->fs_info->delalloc_lock);
C
Chris Mason 已提交
7010
	while (!list_empty(head)) {
7011 7012
		binode = list_entry(head->next, struct btrfs_inode,
				    delalloc_inodes);
7013 7014 7015
		inode = igrab(&binode->vfs_inode);
		if (!inode)
			list_del_init(&binode->delalloc_inodes);
7016
		spin_unlock(&root->fs_info->delalloc_lock);
7017
		if (inode) {
7018
			filemap_flush(inode->i_mapping);
Y
Yan, Zheng 已提交
7019 7020 7021 7022
			if (delay_iput)
				btrfs_add_delayed_iput(inode);
			else
				iput(inode);
7023 7024
		}
		cond_resched();
7025
		spin_lock(&root->fs_info->delalloc_lock);
7026
	}
7027
	spin_unlock(&root->fs_info->delalloc_lock);
7028 7029 7030 7031 7032 7033

	/* 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
7034
	while (atomic_read(&root->fs_info->nr_async_submits) ||
7035
	      atomic_read(&root->fs_info->async_delalloc_pages)) {
7036
		wait_event(root->fs_info->async_submit_wait,
7037 7038
		   (atomic_read(&root->fs_info->nr_async_submits) == 0 &&
		    atomic_read(&root->fs_info->async_delalloc_pages) == 0));
7039 7040
	}
	atomic_dec(&root->fs_info->async_submit_draining);
7041 7042 7043
	return 0;
}

J
Josef Bacik 已提交
7044 7045
int btrfs_start_one_delalloc_inode(struct btrfs_root *root, int delay_iput,
				   int sync)
7046 7047 7048 7049 7050 7051 7052 7053 7054 7055 7056 7057 7058 7059 7060 7061 7062 7063 7064 7065 7066
{
	struct btrfs_inode *binode;
	struct inode *inode = NULL;

	spin_lock(&root->fs_info->delalloc_lock);
	while (!list_empty(&root->fs_info->delalloc_inodes)) {
		binode = list_entry(root->fs_info->delalloc_inodes.next,
				    struct btrfs_inode, delalloc_inodes);
		inode = igrab(&binode->vfs_inode);
		if (inode) {
			list_move_tail(&binode->delalloc_inodes,
				       &root->fs_info->delalloc_inodes);
			break;
		}

		list_del_init(&binode->delalloc_inodes);
		cond_resched_lock(&root->fs_info->delalloc_lock);
	}
	spin_unlock(&root->fs_info->delalloc_lock);

	if (inode) {
J
Josef Bacik 已提交
7067 7068 7069 7070 7071 7072 7073 7074 7075 7076 7077 7078 7079 7080 7081 7082 7083 7084 7085 7086
		if (sync) {
			filemap_write_and_wait(inode->i_mapping);
			/*
			 * We have to do this because compression doesn't
			 * actually set PG_writeback until it submits the pages
			 * for IO, which happens in an async thread, so we could
			 * race and not actually wait for any writeback pages
			 * because they've not been submitted yet.  Technically
			 * this could still be the case for the ordered stuff
			 * since the async thread may not have started to do its
			 * work yet.  If this becomes the case then we need to
			 * figure out a way to make sure that in writepage we
			 * wait for any async pages to be submitted before
			 * returning so that fdatawait does what its supposed to
			 * do.
			 */
			btrfs_wait_ordered_range(inode, 0, (u64)-1);
		} else {
			filemap_flush(inode->i_mapping);
		}
7087 7088 7089 7090 7091 7092 7093 7094 7095
		if (delay_iput)
			btrfs_add_delayed_iput(inode);
		else
			iput(inode);
		return 1;
	}
	return 0;
}

C
Chris Mason 已提交
7096 7097 7098 7099 7100 7101 7102
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;
7103
	struct inode *inode = NULL;
C
Chris Mason 已提交
7104 7105 7106
	int err;
	int drop_inode = 0;
	u64 objectid;
7107
	u64 index = 0 ;
C
Chris Mason 已提交
7108 7109
	int name_len;
	int datasize;
7110
	unsigned long ptr;
C
Chris Mason 已提交
7111
	struct btrfs_file_extent_item *ei;
7112
	struct extent_buffer *leaf;
7113
	unsigned long nr = 0;
C
Chris Mason 已提交
7114 7115 7116 7117

	name_len = strlen(symname) + 1;
	if (name_len > BTRFS_MAX_INLINE_DATA_SIZE(root))
		return -ENAMETOOLONG;
7118

7119 7120 7121
	err = btrfs_find_free_objectid(NULL, root, dir->i_ino, &objectid);
	if (err)
		return err;
J
Josef Bacik 已提交
7122 7123 7124 7125 7126
	/*
	 * 2 items for inode item and ref
	 * 2 items for dir items
	 * 1 item for xattr if selinux is on
	 */
7127 7128 7129
	trans = btrfs_start_transaction(root, 5);
	if (IS_ERR(trans))
		return PTR_ERR(trans);
7130

C
Chris Mason 已提交
7131 7132
	btrfs_set_trans_block_group(trans, dir);

7133
	inode = btrfs_new_inode(trans, root, dir, dentry->d_name.name,
7134
				dentry->d_name.len, dir->i_ino, objectid,
7135 7136
				BTRFS_I(dir)->block_group, S_IFLNK|S_IRWXUGO,
				&index);
C
Chris Mason 已提交
7137 7138 7139 7140
	err = PTR_ERR(inode);
	if (IS_ERR(inode))
		goto out_unlock;

7141
	err = btrfs_init_inode_security(trans, inode, dir);
J
Josef Bacik 已提交
7142 7143 7144 7145 7146
	if (err) {
		drop_inode = 1;
		goto out_unlock;
	}

C
Chris Mason 已提交
7147
	btrfs_set_trans_block_group(trans, inode);
7148
	err = btrfs_add_nondir(trans, dir, dentry, inode, 0, index);
C
Chris Mason 已提交
7149 7150 7151 7152
	if (err)
		drop_inode = 1;
	else {
		inode->i_mapping->a_ops = &btrfs_aops;
C
Chris Mason 已提交
7153
		inode->i_mapping->backing_dev_info = &root->fs_info->bdi;
C
Chris Mason 已提交
7154 7155
		inode->i_fop = &btrfs_file_operations;
		inode->i_op = &btrfs_file_inode_operations;
7156
		BTRFS_I(inode)->io_tree.ops = &btrfs_extent_io_ops;
C
Chris Mason 已提交
7157 7158 7159 7160 7161 7162 7163 7164 7165 7166 7167 7168 7169 7170
	}
	btrfs_update_inode_block_group(trans, inode);
	btrfs_update_inode_block_group(trans, dir);
	if (drop_inode)
		goto out_unlock;

	path = btrfs_alloc_path();
	BUG_ON(!path);
	key.objectid = inode->i_ino;
	key.offset = 0;
	btrfs_set_key_type(&key, BTRFS_EXTENT_DATA_KEY);
	datasize = btrfs_file_extent_calc_inline_size(name_len);
	err = btrfs_insert_empty_item(trans, root, path, &key,
				      datasize);
7171 7172 7173 7174
	if (err) {
		drop_inode = 1;
		goto out_unlock;
	}
7175 7176 7177 7178 7179
	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
7180
				   BTRFS_FILE_EXTENT_INLINE);
C
Chris Mason 已提交
7181 7182 7183 7184 7185
	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
7186
	ptr = btrfs_file_extent_inline_start(ei);
7187 7188
	write_extent_buffer(leaf, symname, ptr, name_len);
	btrfs_mark_buffer_dirty(leaf);
C
Chris Mason 已提交
7189
	btrfs_free_path(path);
7190

C
Chris Mason 已提交
7191 7192
	inode->i_op = &btrfs_symlink_inode_operations;
	inode->i_mapping->a_ops = &btrfs_symlink_aops;
C
Chris Mason 已提交
7193
	inode->i_mapping->backing_dev_info = &root->fs_info->bdi;
Y
Yan Zheng 已提交
7194
	inode_set_bytes(inode, name_len);
7195
	btrfs_i_size_write(inode, name_len - 1);
7196 7197 7198
	err = btrfs_update_inode(trans, root, inode);
	if (err)
		drop_inode = 1;
C
Chris Mason 已提交
7199 7200

out_unlock:
7201
	nr = trans->blocks_used;
7202
	btrfs_end_transaction_throttle(trans, root);
C
Chris Mason 已提交
7203 7204 7205 7206
	if (drop_inode) {
		inode_dec_link_count(inode);
		iput(inode);
	}
7207
	btrfs_btree_balance_dirty(root, nr);
C
Chris Mason 已提交
7208 7209
	return err;
}
7210

7211 7212 7213 7214
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
7215 7216 7217 7218
{
	struct btrfs_root *root = BTRFS_I(inode)->root;
	struct btrfs_key ins;
	u64 cur_offset = start;
7219
	u64 i_size;
Y
Yan Zheng 已提交
7220
	int ret = 0;
7221
	bool own_trans = true;
Y
Yan Zheng 已提交
7222

7223 7224
	if (trans)
		own_trans = false;
Y
Yan Zheng 已提交
7225
	while (num_bytes > 0) {
7226 7227 7228 7229 7230 7231
		if (own_trans) {
			trans = btrfs_start_transaction(root, 3);
			if (IS_ERR(trans)) {
				ret = PTR_ERR(trans);
				break;
			}
7232 7233
		}

7234 7235
		ret = btrfs_reserve_extent(trans, root, num_bytes, min_size,
					   0, *alloc_hint, (u64)-1, &ins, 1);
7236
		if (ret) {
7237 7238
			if (own_trans)
				btrfs_end_transaction(trans, root);
7239
			break;
Y
Yan Zheng 已提交
7240
		}
7241

Y
Yan Zheng 已提交
7242 7243 7244
		ret = insert_reserved_file_extent(trans, inode,
						  cur_offset, ins.objectid,
						  ins.offset, ins.offset,
Y
Yan, Zheng 已提交
7245
						  ins.offset, 0, 0, 0,
Y
Yan Zheng 已提交
7246 7247
						  BTRFS_FILE_EXTENT_PREALLOC);
		BUG_ON(ret);
C
Chris Mason 已提交
7248 7249
		btrfs_drop_extent_cache(inode, cur_offset,
					cur_offset + ins.offset -1, 0);
7250

Y
Yan Zheng 已提交
7251 7252
		num_bytes -= ins.offset;
		cur_offset += ins.offset;
7253
		*alloc_hint = ins.objectid + ins.offset;
7254

Y
Yan Zheng 已提交
7255
		inode->i_ctime = CURRENT_TIME;
7256
		BTRFS_I(inode)->flags |= BTRFS_INODE_PREALLOC;
Y
Yan Zheng 已提交
7257
		if (!(mode & FALLOC_FL_KEEP_SIZE) &&
7258 7259
		    (actual_len > inode->i_size) &&
		    (cur_offset > inode->i_size)) {
7260
			if (cur_offset > actual_len)
7261
				i_size = actual_len;
7262
			else
7263 7264 7265
				i_size = cur_offset;
			i_size_write(inode, i_size);
			btrfs_ordered_update_i_size(inode, i_size, NULL);
7266 7267
		}

Y
Yan Zheng 已提交
7268 7269 7270
		ret = btrfs_update_inode(trans, root, inode);
		BUG_ON(ret);

7271 7272
		if (own_trans)
			btrfs_end_transaction(trans, root);
7273
	}
Y
Yan Zheng 已提交
7274 7275 7276
	return ret;
}

7277 7278 7279 7280 7281 7282 7283 7284 7285 7286 7287 7288 7289 7290 7291 7292 7293 7294
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);
}

7295 7296 7297 7298 7299
static int btrfs_set_page_dirty(struct page *page)
{
	return __set_page_dirty_nobuffers(page);
}

7300
static int btrfs_permission(struct inode *inode, int mask, unsigned int flags)
Y
Yan 已提交
7301
{
L
Li Zefan 已提交
7302 7303 7304 7305
	struct btrfs_root *root = BTRFS_I(inode)->root;

	if (btrfs_root_readonly(root) && (mask & MAY_WRITE))
		return -EROFS;
7306
	if ((BTRFS_I(inode)->flags & BTRFS_INODE_READONLY) && (mask & MAY_WRITE))
Y
Yan 已提交
7307
		return -EACCES;
7308
	return generic_permission(inode, mask, flags, btrfs_check_acl);
Y
Yan 已提交
7309
}
C
Chris Mason 已提交
7310

7311
static const struct inode_operations btrfs_dir_inode_operations = {
7312
	.getattr	= btrfs_getattr,
C
Chris Mason 已提交
7313 7314 7315 7316 7317 7318 7319 7320 7321
	.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
7322
	.mknod		= btrfs_mknod,
7323 7324
	.setxattr	= btrfs_setxattr,
	.getxattr	= btrfs_getxattr,
J
Josef Bacik 已提交
7325
	.listxattr	= btrfs_listxattr,
7326
	.removexattr	= btrfs_removexattr,
Y
Yan 已提交
7327
	.permission	= btrfs_permission,
C
Chris Mason 已提交
7328
};
7329
static const struct inode_operations btrfs_dir_ro_inode_operations = {
C
Chris Mason 已提交
7330
	.lookup		= btrfs_lookup,
Y
Yan 已提交
7331
	.permission	= btrfs_permission,
C
Chris Mason 已提交
7332
};
7333

7334
static const struct file_operations btrfs_dir_file_operations = {
C
Chris Mason 已提交
7335 7336
	.llseek		= generic_file_llseek,
	.read		= generic_read_dir,
7337
	.readdir	= btrfs_real_readdir,
C
Christoph Hellwig 已提交
7338
	.unlocked_ioctl	= btrfs_ioctl,
C
Chris Mason 已提交
7339
#ifdef CONFIG_COMPAT
C
Christoph Hellwig 已提交
7340
	.compat_ioctl	= btrfs_ioctl,
C
Chris Mason 已提交
7341
#endif
S
Sage Weil 已提交
7342
	.release        = btrfs_release_file,
7343
	.fsync		= btrfs_sync_file,
C
Chris Mason 已提交
7344 7345
};

7346
static struct extent_io_ops btrfs_extent_io_ops = {
7347
	.fill_delalloc = run_delalloc_range,
7348
	.submit_bio_hook = btrfs_submit_bio_hook,
7349
	.merge_bio_hook = btrfs_merge_bio_hook,
7350
	.readpage_end_io_hook = btrfs_readpage_end_io_hook,
7351
	.writepage_end_io_hook = btrfs_writepage_end_io_hook,
7352
	.writepage_start_hook = btrfs_writepage_start_hook,
7353
	.readpage_io_failed_hook = btrfs_io_failed_hook,
C
Chris Mason 已提交
7354 7355
	.set_bit_hook = btrfs_set_bit_hook,
	.clear_bit_hook = btrfs_clear_bit_hook,
J
Josef Bacik 已提交
7356 7357
	.merge_extent_hook = btrfs_merge_extent_hook,
	.split_extent_hook = btrfs_split_extent_hook,
7358 7359
};

7360 7361 7362 7363 7364 7365 7366 7367 7368 7369 7370 7371
/*
 * 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.
 */
7372
static const struct address_space_operations btrfs_aops = {
C
Chris Mason 已提交
7373 7374
	.readpage	= btrfs_readpage,
	.writepage	= btrfs_writepage,
C
Chris Mason 已提交
7375
	.writepages	= btrfs_writepages,
C
Chris Mason 已提交
7376
	.readpages	= btrfs_readpages,
C
Chris Mason 已提交
7377
	.sync_page	= block_sync_page,
7378
	.direct_IO	= btrfs_direct_IO,
7379 7380
	.invalidatepage = btrfs_invalidatepage,
	.releasepage	= btrfs_releasepage,
7381
	.set_page_dirty	= btrfs_set_page_dirty,
7382
	.error_remove_page = generic_error_remove_page,
C
Chris Mason 已提交
7383 7384
};

7385
static const struct address_space_operations btrfs_symlink_aops = {
C
Chris Mason 已提交
7386 7387
	.readpage	= btrfs_readpage,
	.writepage	= btrfs_writepage,
C
Chris Mason 已提交
7388 7389
	.invalidatepage = btrfs_invalidatepage,
	.releasepage	= btrfs_releasepage,
C
Chris Mason 已提交
7390 7391
};

7392
static const struct inode_operations btrfs_file_inode_operations = {
C
Chris Mason 已提交
7393 7394
	.getattr	= btrfs_getattr,
	.setattr	= btrfs_setattr,
7395 7396
	.setxattr	= btrfs_setxattr,
	.getxattr	= btrfs_getxattr,
J
Josef Bacik 已提交
7397
	.listxattr      = btrfs_listxattr,
7398
	.removexattr	= btrfs_removexattr,
Y
Yan 已提交
7399
	.permission	= btrfs_permission,
Y
Yehuda Sadeh 已提交
7400
	.fiemap		= btrfs_fiemap,
C
Chris Mason 已提交
7401
};
7402
static const struct inode_operations btrfs_special_inode_operations = {
J
Josef Bacik 已提交
7403 7404
	.getattr	= btrfs_getattr,
	.setattr	= btrfs_setattr,
Y
Yan 已提交
7405
	.permission	= btrfs_permission,
7406 7407
	.setxattr	= btrfs_setxattr,
	.getxattr	= btrfs_getxattr,
J
Josef Bacik 已提交
7408
	.listxattr	= btrfs_listxattr,
7409
	.removexattr	= btrfs_removexattr,
J
Josef Bacik 已提交
7410
};
7411
static const struct inode_operations btrfs_symlink_inode_operations = {
C
Chris Mason 已提交
7412 7413 7414
	.readlink	= generic_readlink,
	.follow_link	= page_follow_link_light,
	.put_link	= page_put_link,
7415
	.getattr	= btrfs_getattr,
Y
Yan 已提交
7416
	.permission	= btrfs_permission,
J
Jim Owens 已提交
7417 7418 7419 7420
	.setxattr	= btrfs_setxattr,
	.getxattr	= btrfs_getxattr,
	.listxattr	= btrfs_listxattr,
	.removexattr	= btrfs_removexattr,
C
Chris Mason 已提交
7421
};
7422

7423
const struct dentry_operations btrfs_dentry_operations = {
7424 7425
	.d_delete	= btrfs_dentry_delete,
};