inode.c 194.6 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"
C
Chris Mason 已提交
53 54 55 56 57 58

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

59 60 61 62 63
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;
64 65
static const struct address_space_operations btrfs_aops;
static const struct address_space_operations btrfs_symlink_aops;
66
static const struct file_operations btrfs_dir_file_operations;
67
static struct extent_io_ops btrfs_extent_io_ops;
C
Chris Mason 已提交
68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84

static struct kmem_cache *btrfs_inode_cachep;
struct kmem_cache *btrfs_trans_handle_cachep;
struct kmem_cache *btrfs_transaction_cachep;
struct kmem_cache *btrfs_path_cachep;

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

85
static void btrfs_truncate(struct inode *inode);
C
Chris Mason 已提交
86
static int btrfs_finish_ordered_io(struct inode *inode, u64 start, u64 end);
87 88 89 90
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);
91

92 93
static int btrfs_init_inode_security(struct btrfs_trans_handle *trans,
				     struct inode *inode,  struct inode *dir)
J
Jim Owens 已提交
94 95 96
{
	int err;

97
	err = btrfs_init_acl(trans, inode, dir);
J
Jim Owens 已提交
98
	if (!err)
99
		err = btrfs_xattr_security_init(trans, inode, dir);
J
Jim Owens 已提交
100 101 102
	return err;
}

C
Chris Mason 已提交
103 104 105 106 107
/*
 * 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
108
static noinline int insert_inline_extent(struct btrfs_trans_handle *trans,
C
Chris Mason 已提交
109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131
				struct btrfs_root *root, struct inode *inode,
				u64 start, size_t size, size_t compressed_size,
				struct page **compressed_pages)
{
	struct btrfs_key key;
	struct btrfs_path *path;
	struct extent_buffer *leaf;
	struct page *page = NULL;
	char *kaddr;
	unsigned long ptr;
	struct btrfs_file_extent_item *ei;
	int err = 0;
	int ret;
	size_t cur_size = size;
	size_t datasize;
	unsigned long offset;
	int use_compress = 0;

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

C
Chris Mason 已提交
132 133
	path = btrfs_alloc_path();
	if (!path)
C
Chris Mason 已提交
134 135
		return -ENOMEM;

136
	path->leave_spinning = 1;
C
Chris Mason 已提交
137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164
	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);

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

170
			kaddr = kmap_atomic(cpage, KM_USER0);
C
Chris Mason 已提交
171
			write_extent_buffer(leaf, kaddr, ptr, cur_size);
172
			kunmap_atomic(kaddr, KM_USER0);
C
Chris Mason 已提交
173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192

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

193 194 195 196 197 198 199 200 201
	/*
	 * 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
202 203
	BTRFS_I(inode)->disk_i_size = inode->i_size;
	btrfs_update_inode(trans, root, inode);
204

C
Chris Mason 已提交
205 206 207 208 209 210 211 212 213 214 215 216
	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.
 */
217
static noinline int cow_file_range_inline(struct btrfs_trans_handle *trans,
C
Chris Mason 已提交
218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235
				 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
236
	    actual_end >= PAGE_CACHE_SIZE ||
C
Chris Mason 已提交
237 238 239 240 241 242 243 244
	    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
245
	ret = btrfs_drop_extents(trans, inode, start, aligned_end,
C
Chris Mason 已提交
246
				 &hint_byte, 1);
C
Chris Mason 已提交
247 248 249 250 251 252 253 254
	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);
255
	btrfs_delalloc_release_metadata(inode, end + 1 - start);
C
Chris Mason 已提交
256
	btrfs_drop_extent_cache(inode, start, aligned_end - 1, 0);
C
Chris Mason 已提交
257 258 259
	return 0;
}

260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296
struct async_extent {
	u64 start;
	u64 ram_size;
	u64 compressed_size;
	struct page **pages;
	unsigned long nr_pages;
	struct list_head list;
};

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

static noinline int add_async_extent(struct async_cow *cow,
				     u64 start, u64 ram_size,
				     u64 compressed_size,
				     struct page **pages,
				     unsigned long nr_pages)
{
	struct async_extent *async_extent;

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

C
Chris Mason 已提交
297
/*
298 299 300
 * 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
301
 *
302 303 304 305 306
 * 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
307
 *
308 309 310 311
 * 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
312
 */
313 314 315 316 317
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)
318 319 320
{
	struct btrfs_root *root = BTRFS_I(inode)->root;
	struct btrfs_trans_handle *trans;
321 322
	u64 num_bytes;
	u64 blocksize = root->sectorsize;
C
Chris Mason 已提交
323
	u64 actual_end;
324
	u64 isize = i_size_read(inode);
325
	int ret = 0;
C
Chris Mason 已提交
326 327 328 329 330 331
	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;
332
	unsigned long max_uncompressed = 128 * 1024;
C
Chris Mason 已提交
333 334
	int i;
	int will_compress;
335

336
	actual_end = min_t(u64, isize, end + 1);
C
Chris Mason 已提交
337 338 339 340
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);
341

342 343 344 345 346 347 348 349 350 351 352 353 354
	/*
	 * 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
355 356 357 358
	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
359 360 361 362 363 364 365
	 * 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
366 367
	 */
	total_compressed = min(total_compressed, max_uncompressed);
368
	num_bytes = (end - start + blocksize) & ~(blocksize - 1);
369
	num_bytes = max(blocksize,  num_bytes);
C
Chris Mason 已提交
370 371
	total_in = 0;
	ret = 0;
372

373 374 375 376
	/*
	 * 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
377
	 */
378
	if (!(BTRFS_I(inode)->flags & BTRFS_INODE_NOCOMPRESS) &&
C
Chris Mason 已提交
379 380
	    (btrfs_test_opt(root, COMPRESS) ||
	     (BTRFS_I(inode)->force_compress))) {
C
Chris Mason 已提交
381
		WARN_ON(pages);
382
		pages = kzalloc(sizeof(struct page *) * nr_pages, GFP_NOFS);
C
Chris Mason 已提交
383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409

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

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

			/* zero the tail end of the last page, we might be
			 * sending it down to disk
			 */
			if (offset) {
				kaddr = kmap_atomic(page, KM_USER0);
				memset(kaddr + offset, 0,
				       PAGE_CACHE_SIZE - offset);
				kunmap_atomic(kaddr, KM_USER0);
			}
			will_compress = 1;
		}
	}
	if (start == 0) {
410 411 412
		trans = btrfs_join_transaction(root, 1);
		BUG_ON(!trans);
		btrfs_set_trans_block_group(trans, inode);
413
		trans->block_rsv = &root->fs_info->delalloc_block_rsv;
414

C
Chris Mason 已提交
415
		/* lets try to make an inline extent */
416
		if (ret || total_in < (actual_end - start)) {
C
Chris Mason 已提交
417
			/* we didn't compress the entire range, try
418
			 * to make an uncompressed inline extent.
C
Chris Mason 已提交
419 420 421 422
			 */
			ret = cow_file_range_inline(trans, root, inode,
						    start, end, 0, NULL);
		} else {
423
			/* try making a compressed inline extent */
C
Chris Mason 已提交
424 425 426 427 428
			ret = cow_file_range_inline(trans, root, inode,
						    start, end,
						    total_compressed, pages);
		}
		if (ret == 0) {
429 430 431 432 433
			/*
			 * 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
434
			extent_clear_unlock_delalloc(inode,
435 436 437
			     &BTRFS_I(inode)->io_tree,
			     start, end, NULL,
			     EXTENT_CLEAR_UNLOCK_PAGE | EXTENT_CLEAR_DIRTY |
438
			     EXTENT_CLEAR_DELALLOC |
439
			     EXTENT_SET_WRITEBACK | EXTENT_END_WRITEBACK);
440 441

			btrfs_end_transaction(trans, root);
C
Chris Mason 已提交
442 443
			goto free_pages_out;
		}
444
		btrfs_end_transaction(trans, root);
C
Chris Mason 已提交
445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473
	}

	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
474
			WARN_ON(pages[i]->mapping);
C
Chris Mason 已提交
475 476 477 478 479 480 481 482
			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
483 484
		if (!btrfs_test_opt(root, FORCE_COMPRESS) &&
		    !(BTRFS_I(inode)->force_compress)) {
C
Chris Mason 已提交
485
			BTRFS_I(inode)->flags |= BTRFS_INODE_NOCOMPRESS;
C
Chris Mason 已提交
486
		}
C
Chris Mason 已提交
487
	}
488 489
	if (will_compress) {
		*num_added += 1;
C
Chris Mason 已提交
490

491 492 493 494 495 496
		/* the async work queues will take care of doing actual
		 * allocation on disk for these compressed pages,
		 * and will submit them to the elevator.
		 */
		add_async_extent(async_cow, start, num_bytes,
				 total_compressed, pages, nr_pages_ret);
497

498
		if (start + num_bytes < end && start + num_bytes < actual_end) {
499 500 501 502 503 504
			start += num_bytes;
			pages = NULL;
			cond_resched();
			goto again;
		}
	} else {
505
cleanup_and_bail_uncompressed:
506 507 508 509 510 511 512 513 514 515 516 517 518 519 520
		/*
		 * No compression, but we still need to write the pages in
		 * the file we've been given so far.  redirty the locked
		 * page if it corresponds to our extent and set things up
		 * for the async work queue to run cow_file_range to do
		 * the normal delalloc dance
		 */
		if (page_offset(locked_page) >= start &&
		    page_offset(locked_page) <= end) {
			__set_page_dirty_nobuffers(locked_page);
			/* unlocked later on in the async handlers */
		}
		add_async_extent(async_cow, start, end - start + 1, 0, NULL, 0);
		*num_added += 1;
	}
521

522 523 524 525 526 527 528 529
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
530
	kfree(pages);
531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551

	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;
552
	int ret = 0;
553 554 555 556 557

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


C
Chris Mason 已提交
558
	while (!list_empty(&async_cow->extents)) {
559 560 561
		async_extent = list_entry(async_cow->extents.next,
					  struct async_extent, list);
		list_del(&async_extent->list);
C
Chris Mason 已提交
562

563 564
		io_tree = &BTRFS_I(inode)->io_tree;

565
retry:
566 567 568 569 570 571
		/* 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,
572 573
					 async_extent->start +
					 async_extent->ram_size - 1, GFP_NOFS);
574 575

			/* allocate blocks */
576 577 578 579 580
			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);
581 582 583 584 585 586 587

			/*
			 * 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.
			 */
588
			if (!page_started && !ret)
589 590
				extent_write_locked_range(io_tree,
						  inode, async_extent->start,
C
Chris Mason 已提交
591
						  async_extent->start +
592 593 594 595 596 597 598 599 600 601 602 603
						  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);

604
		trans = btrfs_join_transaction(root, 1);
605 606 607 608 609
		ret = btrfs_reserve_extent(trans, root,
					   async_extent->compressed_size,
					   async_extent->compressed_size,
					   0, alloc_hint,
					   (u64)-1, &ins, 1);
610 611
		btrfs_end_transaction(trans, root);

612 613 614 615 616 617 618 619 620 621 622 623 624 625 626
		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;
		}

627 628 629 630 631 632 633 634
		/*
		 * 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);

635 636 637
		em = alloc_extent_map(GFP_NOFS);
		em->start = async_extent->start;
		em->len = async_extent->ram_size;
638
		em->orig_start = em->start;
C
Chris Mason 已提交
639

640 641 642 643 644 645
		em->block_start = ins.objectid;
		em->block_len = ins.offset;
		em->bdev = root->fs_info->fs_devices->latest_bdev;
		set_bit(EXTENT_FLAG_PINNED, &em->flags);
		set_bit(EXTENT_FLAG_COMPRESSED, &em->flags);

C
Chris Mason 已提交
646
		while (1) {
647
			write_lock(&em_tree->lock);
648
			ret = add_extent_mapping(em_tree, em);
649
			write_unlock(&em_tree->lock);
650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669
			if (ret != -EEXIST) {
				free_extent_map(em);
				break;
			}
			btrfs_drop_extent_cache(inode, async_extent->start,
						async_extent->start +
						async_extent->ram_size - 1, 0);
		}

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

		/*
		 * clear dirty, set writeback and unlock the pages.
		 */
		extent_clear_unlock_delalloc(inode,
670 671 672 673 674 675
				&BTRFS_I(inode)->io_tree,
				async_extent->start,
				async_extent->start +
				async_extent->ram_size - 1,
				NULL, EXTENT_CLEAR_UNLOCK_PAGE |
				EXTENT_CLEAR_UNLOCK |
676
				EXTENT_CLEAR_DELALLOC |
677
				EXTENT_CLEAR_DIRTY | EXTENT_SET_WRITEBACK);
678 679

		ret = btrfs_submit_compressed_write(inode,
C
Chris Mason 已提交
680 681 682 683 684
				    async_extent->start,
				    async_extent->ram_size,
				    ins.objectid,
				    ins.offset, async_extent->pages,
				    async_extent->nr_pages);
685 686 687 688 689 690 691 692 693 694

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

	return 0;
}

695 696 697 698 699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726
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;
}

727 728 729 730 731 732 733 734 735 736 737 738 739 740 741 742 743 744 745 746 747 748 749 750 751 752 753 754 755 756 757 758
/*
 * 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
759
	BUG_ON(root == root->fs_info->tree_root);
760 761 762
	trans = btrfs_join_transaction(root, 1);
	BUG_ON(!trans);
	btrfs_set_trans_block_group(trans, inode);
763
	trans->block_rsv = &root->fs_info->delalloc_block_rsv;
764 765 766 767 768 769 770 771 772 773 774 775

	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,
776 777 778 779 780 781 782 783
				     &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);
784

785 786 787 788 789 790 791 792 793 794 795
			*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));

796
	alloc_hint = get_extent_allocation_hint(inode, start, num_bytes);
797 798
	btrfs_drop_extent_cache(inode, start, start + num_bytes - 1, 0);

C
Chris Mason 已提交
799
	while (disk_num_bytes > 0) {
800 801
		unsigned long op;

802
		cur_alloc_size = disk_num_bytes;
803
		ret = btrfs_reserve_extent(trans, root, cur_alloc_size,
804
					   root->sectorsize, 0, alloc_hint,
805
					   (u64)-1, &ins, 1);
C
Chris Mason 已提交
806 807
		BUG_ON(ret);

808 809
		em = alloc_extent_map(GFP_NOFS);
		em->start = start;
810
		em->orig_start = em->start;
811 812
		ram_size = ins.offset;
		em->len = ins.offset;
C
Chris Mason 已提交
813

814
		em->block_start = ins.objectid;
C
Chris Mason 已提交
815
		em->block_len = ins.offset;
816
		em->bdev = root->fs_info->fs_devices->latest_bdev;
817
		set_bit(EXTENT_FLAG_PINNED, &em->flags);
C
Chris Mason 已提交
818

C
Chris Mason 已提交
819
		while (1) {
820
			write_lock(&em_tree->lock);
821
			ret = add_extent_mapping(em_tree, em);
822
			write_unlock(&em_tree->lock);
823 824 825 826 827
			if (ret != -EEXIST) {
				free_extent_map(em);
				break;
			}
			btrfs_drop_extent_cache(inode, start,
C
Chris Mason 已提交
828
						start + ram_size - 1, 0);
829 830
		}

831
		cur_alloc_size = ins.offset;
832
		ret = btrfs_add_ordered_extent(inode, start, ins.objectid,
833
					       ram_size, cur_alloc_size, 0);
834
		BUG_ON(ret);
C
Chris Mason 已提交
835

836 837 838 839 840 841 842
		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
843
		if (disk_num_bytes < cur_alloc_size)
844
			break;
C
Chris Mason 已提交
845

C
Chris Mason 已提交
846 847 848
		/* 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
849 850 851
		 *
		 * Do set the Private2 bit so we know this page was properly
		 * setup for writepage
C
Chris Mason 已提交
852
		 */
853 854 855 856
		op = unlock ? EXTENT_CLEAR_UNLOCK_PAGE : 0;
		op |= EXTENT_CLEAR_UNLOCK | EXTENT_CLEAR_DELALLOC |
			EXTENT_SET_PRIVATE2;

C
Chris Mason 已提交
857 858
		extent_clear_unlock_delalloc(inode, &BTRFS_I(inode)->io_tree,
					     start, start + ram_size - 1,
859
					     locked_page, op);
C
Chris Mason 已提交
860
		disk_num_bytes -= cur_alloc_size;
861 862 863
		num_bytes -= cur_alloc_size;
		alloc_hint = ins.objectid + ins.offset;
		start += cur_alloc_size;
864 865
	}
out:
866
	ret = 0;
867
	btrfs_end_transaction(trans, root);
C
Chris Mason 已提交
868

869
	return ret;
870
}
C
Chris Mason 已提交
871

872 873 874 875 876 877 878 879 880 881 882 883 884 885 886 887 888 889 890 891 892 893 894 895 896 897 898 899 900 901 902 903 904 905 906 907 908 909
/*
 * 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
910
	if (async_cow->inode)
911 912
		submit_compressed_extents(async_cow->inode, async_cow);
}
C
Chris Mason 已提交
913

914 915 916 917 918 919 920 921 922 923 924 925 926 927 928 929 930
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;

931 932
	clear_extent_bit(&BTRFS_I(inode)->io_tree, start, end, EXTENT_LOCKED,
			 1, 0, NULL, GFP_NOFS);
C
Chris Mason 已提交
933
	while (start < end) {
934 935 936 937 938 939
		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;

940
		if (BTRFS_I(inode)->flags & BTRFS_INODE_NOCOMPRESS)
941 942 943 944 945 946 947 948 949 950 951 952 953 954 955 956 957 958 959 960 961 962 963 964 965
			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
966
		while (atomic_read(&root->fs_info->async_submit_draining) &&
967 968 969 970 971 972 973 974 975 976 977
		      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;
978 979
}

C
Chris Mason 已提交
980
static noinline int csum_exist_in_range(struct btrfs_root *root,
981 982 983 984 985 986
					u64 bytenr, u64 num_bytes)
{
	int ret;
	struct btrfs_ordered_sum *sums;
	LIST_HEAD(list);

Y
Yan Zheng 已提交
987 988
	ret = btrfs_lookup_csums_range(root->fs_info->csum_root, bytenr,
				       bytenr + num_bytes - 1, &list);
989 990 991 992 993 994 995 996 997 998 999
	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
1000 1001 1002 1003 1004 1005 1006
/*
 * 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
 */
1007 1008
static noinline int run_delalloc_nocow(struct inode *inode,
				       struct page *locked_page,
1009 1010
			      u64 start, u64 end, int *page_started, int force,
			      unsigned long *nr_written)
1011 1012
{
	struct btrfs_root *root = BTRFS_I(inode)->root;
1013
	struct btrfs_trans_handle *trans;
1014 1015
	struct extent_buffer *leaf;
	struct btrfs_path *path;
Y
Yan Zheng 已提交
1016
	struct btrfs_file_extent_item *fi;
1017
	struct btrfs_key found_key;
Y
Yan Zheng 已提交
1018 1019 1020
	u64 cow_start;
	u64 cur_offset;
	u64 extent_end;
1021
	u64 extent_offset;
Y
Yan Zheng 已提交
1022 1023 1024 1025
	u64 disk_bytenr;
	u64 num_bytes;
	int extent_type;
	int ret;
Y
Yan Zheng 已提交
1026
	int type;
Y
Yan Zheng 已提交
1027 1028
	int nocow;
	int check_prev = 1;
J
Josef Bacik 已提交
1029
	bool nolock = false;
1030 1031 1032

	path = btrfs_alloc_path();
	BUG_ON(!path);
J
Josef Bacik 已提交
1033 1034 1035 1036 1037 1038
	if (root == root->fs_info->tree_root) {
		nolock = true;
		trans = btrfs_join_transaction_nolock(root, 1);
	} else {
		trans = btrfs_join_transaction(root, 1);
	}
1039
	BUG_ON(!trans);
1040

Y
Yan Zheng 已提交
1041 1042 1043 1044 1045 1046 1047 1048 1049 1050 1051 1052 1053 1054 1055 1056 1057 1058 1059 1060 1061 1062 1063 1064 1065
	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];
		}
1066

Y
Yan Zheng 已提交
1067 1068
		nocow = 0;
		disk_bytenr = 0;
1069
		num_bytes = 0;
Y
Yan Zheng 已提交
1070 1071 1072 1073 1074 1075 1076 1077 1078
		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;
1079
			extent_type = 0;
Y
Yan Zheng 已提交
1080 1081 1082 1083 1084 1085 1086
			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
1087 1088
		if (extent_type == BTRFS_FILE_EXTENT_REG ||
		    extent_type == BTRFS_FILE_EXTENT_PREALLOC) {
Y
Yan Zheng 已提交
1089
			disk_bytenr = btrfs_file_extent_disk_bytenr(leaf, fi);
1090
			extent_offset = btrfs_file_extent_offset(leaf, fi);
Y
Yan Zheng 已提交
1091 1092 1093 1094 1095 1096
			extent_end = found_key.offset +
				btrfs_file_extent_num_bytes(leaf, fi);
			if (extent_end <= start) {
				path->slots[0]++;
				goto next_slot;
			}
1097 1098
			if (disk_bytenr == 0)
				goto out_check;
Y
Yan Zheng 已提交
1099 1100 1101 1102
			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
1103 1104
			if (extent_type == BTRFS_FILE_EXTENT_REG && !force)
				goto out_check;
1105
			if (btrfs_extent_readonly(root, disk_bytenr))
Y
Yan Zheng 已提交
1106
				goto out_check;
1107
			if (btrfs_cross_ref_exist(trans, root, inode->i_ino,
1108 1109
						  found_key.offset -
						  extent_offset, disk_bytenr))
1110
				goto out_check;
1111
			disk_bytenr += extent_offset;
1112 1113 1114 1115 1116 1117 1118 1119 1120
			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
1121 1122 1123 1124 1125 1126 1127 1128 1129 1130 1131 1132 1133 1134 1135 1136 1137 1138 1139 1140 1141
			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;
1142 1143 1144
		}

		btrfs_release_path(root, path);
Y
Yan Zheng 已提交
1145 1146
		if (cow_start != (u64)-1) {
			ret = cow_file_range(inode, locked_page, cow_start,
1147 1148
					found_key.offset - 1, page_started,
					nr_written, 1);
Y
Yan Zheng 已提交
1149 1150
			BUG_ON(ret);
			cow_start = (u64)-1;
1151
		}
Y
Yan Zheng 已提交
1152

Y
Yan Zheng 已提交
1153 1154 1155 1156 1157 1158
		if (extent_type == BTRFS_FILE_EXTENT_PREALLOC) {
			struct extent_map *em;
			struct extent_map_tree *em_tree;
			em_tree = &BTRFS_I(inode)->extent_tree;
			em = alloc_extent_map(GFP_NOFS);
			em->start = cur_offset;
1159
			em->orig_start = em->start;
Y
Yan Zheng 已提交
1160 1161 1162 1163 1164 1165
			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) {
1166
				write_lock(&em_tree->lock);
Y
Yan Zheng 已提交
1167
				ret = add_extent_mapping(em_tree, em);
1168
				write_unlock(&em_tree->lock);
Y
Yan Zheng 已提交
1169 1170 1171 1172 1173 1174 1175 1176 1177 1178 1179
				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
1180 1181

		ret = btrfs_add_ordered_extent(inode, cur_offset, disk_bytenr,
Y
Yan Zheng 已提交
1182 1183
					       num_bytes, num_bytes, type);
		BUG_ON(ret);
1184

1185 1186 1187 1188 1189 1190 1191
		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
1192
		extent_clear_unlock_delalloc(inode, &BTRFS_I(inode)->io_tree,
1193 1194 1195 1196
				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
1197 1198 1199
		cur_offset = extent_end;
		if (cur_offset > end)
			break;
1200
	}
Y
Yan Zheng 已提交
1201 1202 1203 1204 1205 1206
	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,
1207
				     page_started, nr_written, 1);
Y
Yan Zheng 已提交
1208 1209 1210
		BUG_ON(ret);
	}

J
Josef Bacik 已提交
1211 1212 1213 1214 1215 1216 1217
	if (nolock) {
		ret = btrfs_end_transaction_nolock(trans, root);
		BUG_ON(ret);
	} else {
		ret = btrfs_end_transaction(trans, root);
		BUG_ON(ret);
	}
1218
	btrfs_free_path(path);
Y
Yan Zheng 已提交
1219
	return 0;
1220 1221
}

C
Chris Mason 已提交
1222 1223 1224
/*
 * extent_io.c call back to do delayed allocation processing
 */
C
Chris Mason 已提交
1225
static int run_delalloc_range(struct inode *inode, struct page *locked_page,
1226 1227
			      u64 start, u64 end, int *page_started,
			      unsigned long *nr_written)
1228 1229
{
	int ret;
1230
	struct btrfs_root *root = BTRFS_I(inode)->root;
1231

1232
	if (BTRFS_I(inode)->flags & BTRFS_INODE_NODATACOW)
C
Chris Mason 已提交
1233
		ret = run_delalloc_nocow(inode, locked_page, start, end,
C
Chris Mason 已提交
1234
					 page_started, 1, nr_written);
1235
	else if (BTRFS_I(inode)->flags & BTRFS_INODE_PREALLOC)
Y
Yan Zheng 已提交
1236
		ret = run_delalloc_nocow(inode, locked_page, start, end,
C
Chris Mason 已提交
1237
					 page_started, 0, nr_written);
C
Chris Mason 已提交
1238 1239
	else if (!btrfs_test_opt(root, COMPRESS) &&
		 !(BTRFS_I(inode)->force_compress))
1240 1241
		ret = cow_file_range(inode, locked_page, start, end,
				      page_started, nr_written, 1);
1242
	else
1243
		ret = cow_file_range_async(inode, locked_page, start, end,
C
Chris Mason 已提交
1244
					   page_started, nr_written);
1245 1246 1247
	return ret;
}

J
Josef Bacik 已提交
1248
static int btrfs_split_extent_hook(struct inode *inode,
1249
				   struct extent_state *orig, u64 split)
J
Josef Bacik 已提交
1250
{
1251
	/* not delalloc, ignore it */
J
Josef Bacik 已提交
1252 1253 1254
	if (!(orig->state & EXTENT_DELALLOC))
		return 0;

1255
	atomic_inc(&BTRFS_I(inode)->outstanding_extents);
J
Josef Bacik 已提交
1256 1257 1258 1259 1260 1261 1262 1263 1264 1265 1266 1267 1268 1269 1270 1271 1272
	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;

1273
	atomic_dec(&BTRFS_I(inode)->outstanding_extents);
J
Josef Bacik 已提交
1274 1275 1276
	return 0;
}

C
Chris Mason 已提交
1277 1278 1279 1280 1281
/*
 * 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.
 */
1282 1283
static int btrfs_set_bit_hook(struct inode *inode,
			      struct extent_state *state, int *bits)
1284
{
J
Josef Bacik 已提交
1285

1286 1287 1288 1289 1290
	/*
	 * 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
	 */
1291
	if (!(state->state & EXTENT_DELALLOC) && (*bits & EXTENT_DELALLOC)) {
1292
		struct btrfs_root *root = BTRFS_I(inode)->root;
1293
		u64 len = state->end + 1 - state->start;
J
Josef Bacik 已提交
1294 1295
		int do_list = (root->root_key.objectid !=
			       BTRFS_ROOT_TREE_OBJECTID);
J
Josef Bacik 已提交
1296

1297 1298 1299 1300
		if (*bits & EXTENT_FIRST_DELALLOC)
			*bits &= ~EXTENT_FIRST_DELALLOC;
		else
			atomic_inc(&BTRFS_I(inode)->outstanding_extents);
1301

1302
		spin_lock(&root->fs_info->delalloc_lock);
1303 1304
		BTRFS_I(inode)->delalloc_bytes += len;
		root->fs_info->delalloc_bytes += len;
J
Josef Bacik 已提交
1305
		if (do_list && list_empty(&BTRFS_I(inode)->delalloc_inodes)) {
1306 1307 1308
			list_add_tail(&BTRFS_I(inode)->delalloc_inodes,
				      &root->fs_info->delalloc_inodes);
		}
1309
		spin_unlock(&root->fs_info->delalloc_lock);
1310 1311 1312 1313
	}
	return 0;
}

C
Chris Mason 已提交
1314 1315 1316
/*
 * extent_io.c clear_bit_hook, see set_bit_hook for why
 */
J
Josef Bacik 已提交
1317
static int btrfs_clear_bit_hook(struct inode *inode,
1318
				struct extent_state *state, int *bits)
1319
{
1320 1321 1322 1323 1324
	/*
	 * 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
	 */
1325
	if ((state->state & EXTENT_DELALLOC) && (*bits & EXTENT_DELALLOC)) {
1326
		struct btrfs_root *root = BTRFS_I(inode)->root;
1327
		u64 len = state->end + 1 - state->start;
J
Josef Bacik 已提交
1328 1329
		int do_list = (root->root_key.objectid !=
			       BTRFS_ROOT_TREE_OBJECTID);
1330

1331 1332 1333 1334 1335 1336 1337 1338
		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
1339 1340
		if (root->root_key.objectid != BTRFS_DATA_RELOC_TREE_OBJECTID
		    && do_list)
1341
			btrfs_free_reserved_data_space(inode, len);
J
Josef Bacik 已提交
1342

1343
		spin_lock(&root->fs_info->delalloc_lock);
1344 1345 1346
		root->fs_info->delalloc_bytes -= len;
		BTRFS_I(inode)->delalloc_bytes -= len;

J
Josef Bacik 已提交
1347
		if (do_list && BTRFS_I(inode)->delalloc_bytes == 0 &&
1348 1349 1350
		    !list_empty(&BTRFS_I(inode)->delalloc_inodes)) {
			list_del_init(&BTRFS_I(inode)->delalloc_inodes);
		}
1351
		spin_unlock(&root->fs_info->delalloc_lock);
1352 1353 1354 1355
	}
	return 0;
}

C
Chris Mason 已提交
1356 1357 1358 1359
/*
 * 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
 */
1360
int btrfs_merge_bio_hook(struct page *page, unsigned long offset,
C
Chris Mason 已提交
1361 1362
			 size_t size, struct bio *bio,
			 unsigned long bio_flags)
1363 1364 1365
{
	struct btrfs_root *root = BTRFS_I(page->mapping->host)->root;
	struct btrfs_mapping_tree *map_tree;
1366
	u64 logical = (u64)bio->bi_sector << 9;
1367 1368 1369 1370
	u64 length = 0;
	u64 map_length;
	int ret;

1371 1372 1373
	if (bio_flags & EXTENT_BIO_COMPRESSED)
		return 0;

1374
	length = bio->bi_size;
1375 1376
	map_tree = &root->fs_info->mapping_tree;
	map_length = length;
1377
	ret = btrfs_map_block(map_tree, READ, logical,
1378
			      &map_length, NULL, 0);
1379

C
Chris Mason 已提交
1380
	if (map_length < length + size)
1381
		return 1;
1382
	return ret;
1383 1384
}

C
Chris Mason 已提交
1385 1386 1387 1388 1389 1390 1391 1392
/*
 * 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
1393 1394
static int __btrfs_submit_bio_start(struct inode *inode, int rw,
				    struct bio *bio, int mirror_num,
1395 1396
				    unsigned long bio_flags,
				    u64 bio_offset)
1397 1398 1399
{
	struct btrfs_root *root = BTRFS_I(inode)->root;
	int ret = 0;
1400

1401
	ret = btrfs_csum_one_bio(root, inode, bio, 0, 0);
1402
	BUG_ON(ret);
C
Chris Mason 已提交
1403 1404
	return 0;
}
1405

C
Chris Mason 已提交
1406 1407 1408 1409 1410 1411 1412 1413
/*
 * 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
 */
1414
static int __btrfs_submit_bio_done(struct inode *inode, int rw, struct bio *bio,
1415 1416
			  int mirror_num, unsigned long bio_flags,
			  u64 bio_offset)
C
Chris Mason 已提交
1417 1418
{
	struct btrfs_root *root = BTRFS_I(inode)->root;
1419
	return btrfs_map_bio(root, rw, bio, mirror_num, 1);
1420 1421
}

C
Chris Mason 已提交
1422
/*
1423 1424
 * 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
1425
 */
1426
static int btrfs_submit_bio_hook(struct inode *inode, int rw, struct bio *bio,
1427 1428
			  int mirror_num, unsigned long bio_flags,
			  u64 bio_offset)
1429 1430 1431
{
	struct btrfs_root *root = BTRFS_I(inode)->root;
	int ret = 0;
1432
	int skip_sum;
1433

1434
	skip_sum = BTRFS_I(inode)->flags & BTRFS_INODE_NODATASUM;
1435

J
Josef Bacik 已提交
1436 1437 1438 1439
	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);
1440
	BUG_ON(ret);
1441

1442
	if (!(rw & REQ_WRITE)) {
1443
		if (bio_flags & EXTENT_BIO_COMPRESSED) {
C
Chris Mason 已提交
1444 1445
			return btrfs_submit_compressed_read(inode, bio,
						    mirror_num, bio_flags);
1446 1447
		} else if (!skip_sum)
			btrfs_lookup_bio_sums(root, inode, bio, NULL);
1448
		goto mapit;
1449
	} else if (!skip_sum) {
1450 1451 1452
		/* csum items have already been cloned */
		if (root->root_key.objectid == BTRFS_DATA_RELOC_TREE_OBJECTID)
			goto mapit;
1453 1454
		/* we're doing a write, do the async checksumming */
		return btrfs_wq_submit_bio(BTRFS_I(inode)->root->fs_info,
1455
				   inode, rw, bio, mirror_num,
1456 1457
				   bio_flags, bio_offset,
				   __btrfs_submit_bio_start,
C
Chris Mason 已提交
1458
				   __btrfs_submit_bio_done);
1459 1460
	}

1461
mapit:
1462
	return btrfs_map_bio(root, rw, bio, mirror_num, 0);
1463
}
C
Chris Mason 已提交
1464

C
Chris Mason 已提交
1465 1466 1467 1468
/*
 * 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.
 */
1469
static noinline int add_pending_csums(struct btrfs_trans_handle *trans,
1470 1471 1472 1473 1474 1475
			     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
1476 1477

	list_for_each_entry(sum, list, list) {
1478 1479
		btrfs_csum_file_blocks(trans,
		       BTRFS_I(inode)->root->fs_info->csum_root, sum);
1480 1481 1482 1483
	}
	return 0;
}

1484 1485
int btrfs_set_extent_delalloc(struct inode *inode, u64 start, u64 end,
			      struct extent_state **cached_state)
1486
{
C
Chris Mason 已提交
1487
	if ((end & (PAGE_CACHE_SIZE - 1)) == 0)
1488
		WARN_ON(1);
1489
	return set_extent_delalloc(&BTRFS_I(inode)->io_tree, start, end,
1490
				   cached_state, GFP_NOFS);
1491 1492
}

C
Chris Mason 已提交
1493
/* see btrfs_writepage_start_hook for details on why this is required */
1494 1495 1496 1497 1498
struct btrfs_writepage_fixup {
	struct page *page;
	struct btrfs_work work;
};

1499
static void btrfs_writepage_fixup_worker(struct btrfs_work *work)
1500 1501 1502
{
	struct btrfs_writepage_fixup *fixup;
	struct btrfs_ordered_extent *ordered;
1503
	struct extent_state *cached_state = NULL;
1504 1505 1506 1507 1508 1509 1510
	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
1511
again:
1512 1513 1514 1515 1516 1517 1518 1519 1520 1521
	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;

1522 1523
	lock_extent_bits(&BTRFS_I(inode)->io_tree, page_start, page_end, 0,
			 &cached_state, GFP_NOFS);
C
Chris Mason 已提交
1524 1525

	/* already ordered? We're done */
1526
	if (PagePrivate2(page))
1527
		goto out;
C
Chris Mason 已提交
1528 1529 1530

	ordered = btrfs_lookup_ordered_extent(inode, page_start);
	if (ordered) {
1531 1532
		unlock_extent_cached(&BTRFS_I(inode)->io_tree, page_start,
				     page_end, &cached_state, GFP_NOFS);
C
Chris Mason 已提交
1533 1534 1535 1536
		unlock_page(page);
		btrfs_start_ordered_extent(inode, ordered, 1);
		goto again;
	}
1537

1538
	BUG();
1539
	btrfs_set_extent_delalloc(inode, page_start, page_end, &cached_state);
1540 1541
	ClearPageChecked(page);
out:
1542 1543
	unlock_extent_cached(&BTRFS_I(inode)->io_tree, page_start, page_end,
			     &cached_state, GFP_NOFS);
1544 1545 1546 1547 1548 1549 1550 1551 1552 1553 1554
out_page:
	unlock_page(page);
	page_cache_release(page);
}

/*
 * There are a few paths in the higher layers of the kernel that directly
 * set the page dirty bit without asking the filesystem if it is a
 * good idea.  This causes problems because we want to make sure COW
 * properly happens and the data=ordered rules are followed.
 *
C
Chris Mason 已提交
1555
 * In our case any range that doesn't have the ORDERED bit set
1556 1557 1558 1559
 * 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.
 */
1560
static int btrfs_writepage_start_hook(struct page *page, u64 start, u64 end)
1561 1562 1563 1564 1565
{
	struct inode *inode = page->mapping->host;
	struct btrfs_writepage_fixup *fixup;
	struct btrfs_root *root = BTRFS_I(inode)->root;

1566 1567
	/* this page is properly in the ordered list */
	if (TestClearPagePrivate2(page))
1568 1569 1570 1571 1572 1573 1574 1575
		return 0;

	if (PageChecked(page))
		return -EAGAIN;

	fixup = kzalloc(sizeof(*fixup), GFP_NOFS);
	if (!fixup)
		return -EAGAIN;
1576

1577 1578 1579 1580 1581 1582 1583 1584
	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
1585 1586 1587 1588 1589 1590 1591 1592 1593 1594 1595 1596 1597 1598 1599 1600 1601 1602
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);

1603
	path->leave_spinning = 1;
C
Chris Mason 已提交
1604 1605 1606 1607 1608 1609 1610 1611 1612 1613

	/*
	 * 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
1614 1615
	ret = btrfs_drop_extents(trans, inode, file_pos, file_pos + num_bytes,
				 &hint, 0);
Y
Yan Zheng 已提交
1616 1617 1618 1619 1620 1621 1622 1623 1624 1625 1626 1627 1628 1629 1630 1631 1632 1633 1634 1635
	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);
1636 1637 1638 1639

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

Y
Yan Zheng 已提交
1640 1641 1642 1643 1644 1645 1646
	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;
1647 1648 1649
	ret = btrfs_alloc_reserved_file_extent(trans, root,
					root->root_key.objectid,
					inode->i_ino, file_pos, &ins);
Y
Yan Zheng 已提交
1650 1651
	BUG_ON(ret);
	btrfs_free_path(path);
1652

Y
Yan Zheng 已提交
1653 1654 1655
	return 0;
}

1656 1657 1658 1659 1660 1661
/*
 * 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
1662 1663 1664 1665
/* 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.
 */
1666
static int btrfs_finish_ordered_io(struct inode *inode, u64 start, u64 end)
1667 1668
{
	struct btrfs_root *root = BTRFS_I(inode)->root;
1669
	struct btrfs_trans_handle *trans = NULL;
1670
	struct btrfs_ordered_extent *ordered_extent = NULL;
1671
	struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
1672
	struct extent_state *cached_state = NULL;
Y
Yan Zheng 已提交
1673
	int compressed = 0;
1674
	int ret;
J
Josef Bacik 已提交
1675
	bool nolock = false;
1676

1677 1678
	ret = btrfs_dec_test_ordered_pending(inode, &ordered_extent, start,
					     end - start + 1);
1679
	if (!ret)
1680 1681
		return 0;
	BUG_ON(!ordered_extent);
1682

J
Josef Bacik 已提交
1683 1684
	nolock = (root == root->fs_info->tree_root);

1685 1686 1687 1688
	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
1689 1690 1691 1692 1693
			if (nolock)
				trans = btrfs_join_transaction_nolock(root, 1);
			else
				trans = btrfs_join_transaction(root, 1);
			BUG_ON(!trans);
1694 1695
			btrfs_set_trans_block_group(trans, inode);
			trans->block_rsv = &root->fs_info->delalloc_block_rsv;
1696 1697 1698 1699 1700
			ret = btrfs_update_inode(trans, root, inode);
			BUG_ON(ret);
		}
		goto out;
	}
1701

1702 1703 1704
	lock_extent_bits(io_tree, ordered_extent->file_offset,
			 ordered_extent->file_offset + ordered_extent->len - 1,
			 0, &cached_state, GFP_NOFS);
1705

J
Josef Bacik 已提交
1706 1707 1708 1709
	if (nolock)
		trans = btrfs_join_transaction_nolock(root, 1);
	else
		trans = btrfs_join_transaction(root, 1);
1710 1711
	btrfs_set_trans_block_group(trans, inode);
	trans->block_rsv = &root->fs_info->delalloc_block_rsv;
1712

C
Chris Mason 已提交
1713
	if (test_bit(BTRFS_ORDERED_COMPRESSED, &ordered_extent->flags))
Y
Yan Zheng 已提交
1714 1715 1716
		compressed = 1;
	if (test_bit(BTRFS_ORDERED_PREALLOC, &ordered_extent->flags)) {
		BUG_ON(compressed);
Y
Yan, Zheng 已提交
1717
		ret = btrfs_mark_extent_written(trans, inode,
Y
Yan Zheng 已提交
1718 1719 1720 1721 1722
						ordered_extent->file_offset,
						ordered_extent->file_offset +
						ordered_extent->len);
		BUG_ON(ret);
	} else {
1723
		BUG_ON(root == root->fs_info->tree_root);
Y
Yan Zheng 已提交
1724 1725 1726 1727 1728 1729 1730 1731
		ret = insert_reserved_file_extent(trans, inode,
						ordered_extent->file_offset,
						ordered_extent->start,
						ordered_extent->disk_len,
						ordered_extent->len,
						ordered_extent->len,
						compressed, 0, 0,
						BTRFS_FILE_EXTENT_REG);
C
Chris Mason 已提交
1732 1733 1734
		unpin_extent_cache(&BTRFS_I(inode)->extent_tree,
				   ordered_extent->file_offset,
				   ordered_extent->len);
Y
Yan Zheng 已提交
1735 1736
		BUG_ON(ret);
	}
1737 1738 1739 1740
	unlock_extent_cached(io_tree, ordered_extent->file_offset,
			     ordered_extent->file_offset +
			     ordered_extent->len - 1, &cached_state, GFP_NOFS);

1741 1742 1743
	add_pending_csums(trans, inode, ordered_extent->file_offset,
			  &ordered_extent->list);

1744 1745 1746 1747
	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
1748 1749 1750 1751 1752 1753 1754 1755 1756
	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);
	}

1757 1758 1759 1760 1761 1762 1763 1764
	/* once for us */
	btrfs_put_ordered_extent(ordered_extent);
	/* once for the tree */
	btrfs_put_ordered_extent(ordered_extent);

	return 0;
}

1765
static int btrfs_writepage_end_io_hook(struct page *page, u64 start, u64 end,
1766 1767
				struct extent_state *state, int uptodate)
{
1768
	ClearPagePrivate2(page);
1769 1770 1771
	return btrfs_finish_ordered_io(page->mapping->host, start, end);
}

C
Chris Mason 已提交
1772 1773 1774 1775 1776 1777 1778 1779
/*
 * 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.
 */
1780 1781 1782 1783 1784
struct io_failure_record {
	struct page *page;
	u64 start;
	u64 len;
	u64 logical;
1785
	unsigned long bio_flags;
1786 1787 1788
	int last_mirror;
};

1789
static int btrfs_io_failed_hook(struct bio *failed_bio,
1790 1791
			 struct page *page, u64 start, u64 end,
			 struct extent_state *state)
1792 1793 1794 1795 1796 1797
{
	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;
1798
	struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
1799 1800 1801
	struct bio *bio;
	int num_copies;
	int ret;
1802
	int rw;
1803 1804 1805 1806 1807 1808 1809 1810 1811 1812
	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;
1813
		failrec->bio_flags = 0;
1814

1815
		read_lock(&em_tree->lock);
1816 1817 1818 1819 1820
		em = lookup_extent_mapping(em_tree, start, failrec->len);
		if (em->start > start || em->start + em->len < start) {
			free_extent_map(em);
			em = NULL;
		}
1821
		read_unlock(&em_tree->lock);
1822 1823 1824 1825 1826 1827 1828

		if (!em || IS_ERR(em)) {
			kfree(failrec);
			return -EIO;
		}
		logical = start - em->start;
		logical = em->block_start + logical;
1829 1830 1831 1832
		if (test_bit(EXTENT_FLAG_COMPRESSED, &em->flags)) {
			logical = em->block_start;
			failrec->bio_flags = EXTENT_BIO_COMPRESSED;
		}
1833 1834 1835 1836
		failrec->logical = logical;
		free_extent_map(em);
		set_extent_bits(failure_tree, start, end, EXTENT_LOCKED |
				EXTENT_DIRTY, GFP_NOFS);
1837 1838
		set_state_private(failure_tree, start,
				 (u64)(unsigned long)failrec);
1839
	} else {
1840
		failrec = (struct io_failure_record *)(unsigned long)private;
1841 1842 1843 1844 1845 1846
	}
	num_copies = btrfs_num_copies(
			      &BTRFS_I(inode)->root->fs_info->mapping_tree,
			      failrec->logical, failrec->len);
	failrec->last_mirror++;
	if (!state) {
1847
		spin_lock(&BTRFS_I(inode)->io_tree.lock);
1848 1849 1850 1851 1852
		state = find_first_extent_bit_state(&BTRFS_I(inode)->io_tree,
						    failrec->start,
						    EXTENT_LOCKED);
		if (state && state->start != failrec->start)
			state = NULL;
1853
		spin_unlock(&BTRFS_I(inode)->io_tree.lock);
1854 1855 1856 1857 1858 1859 1860 1861 1862 1863 1864 1865 1866 1867
	}
	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;
1868
	bio->bi_size = 0;
1869

1870
	bio_add_page(bio, page, failrec->len, start - page_offset(page));
1871
	if (failed_bio->bi_rw & REQ_WRITE)
1872 1873 1874 1875 1876
		rw = WRITE;
	else
		rw = READ;

	BTRFS_I(inode)->io_tree.ops->submit_bio_hook(inode, rw, bio,
C
Chris Mason 已提交
1877
						      failrec->last_mirror,
1878
						      failrec->bio_flags, 0);
1879 1880 1881
	return 0;
}

C
Chris Mason 已提交
1882 1883 1884 1885
/*
 * 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
 */
1886
static int btrfs_clean_io_failures(struct inode *inode, u64 start)
1887 1888 1889 1890 1891 1892 1893 1894 1895 1896 1897 1898 1899 1900 1901 1902 1903 1904 1905 1906 1907 1908 1909 1910
{
	u64 private;
	u64 private_failure;
	struct io_failure_record *failure;
	int ret;

	private = 0;
	if (count_range_bits(&BTRFS_I(inode)->io_failure_tree, &private,
			     (u64)-1, 1, EXTENT_DIRTY)) {
		ret = get_state_private(&BTRFS_I(inode)->io_failure_tree,
					start, &private_failure);
		if (ret == 0) {
			failure = (struct io_failure_record *)(unsigned long)
				   private_failure;
			set_state_private(&BTRFS_I(inode)->io_failure_tree,
					  failure->start, 0);
			clear_extent_bits(&BTRFS_I(inode)->io_failure_tree,
					  failure->start,
					  failure->start + failure->len - 1,
					  EXTENT_DIRTY | EXTENT_LOCKED,
					  GFP_NOFS);
			kfree(failure);
		}
	}
1911 1912 1913
	return 0;
}

C
Chris Mason 已提交
1914 1915 1916 1917 1918
/*
 * 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
 */
1919
static int btrfs_readpage_end_io_hook(struct page *page, u64 start, u64 end,
1920
			       struct extent_state *state)
1921
{
1922
	size_t offset = start - ((u64)page->index << PAGE_CACHE_SHIFT);
1923
	struct inode *inode = page->mapping->host;
1924
	struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
1925
	char *kaddr;
1926
	u64 private = ~(u32)0;
1927
	int ret;
1928 1929
	struct btrfs_root *root = BTRFS_I(inode)->root;
	u32 csum = ~(u32)0;
1930

1931 1932 1933 1934
	if (PageChecked(page)) {
		ClearPageChecked(page);
		goto good;
	}
1935 1936

	if (BTRFS_I(inode)->flags & BTRFS_INODE_NODATASUM)
1937 1938 1939
		return 0;

	if (root->root_key.objectid == BTRFS_DATA_RELOC_TREE_OBJECTID &&
1940
	    test_range_bit(io_tree, start, end, EXTENT_NODATASUM, 1, NULL)) {
1941 1942
		clear_extent_bits(io_tree, start, end, EXTENT_NODATASUM,
				  GFP_NOFS);
1943
		return 0;
1944
	}
1945

Y
Yan 已提交
1946
	if (state && state->start == start) {
1947 1948 1949 1950 1951
		private = state->private;
		ret = 0;
	} else {
		ret = get_state_private(io_tree, start, &private);
	}
1952
	kaddr = kmap_atomic(page, KM_USER0);
C
Chris Mason 已提交
1953
	if (ret)
1954
		goto zeroit;
C
Chris Mason 已提交
1955

1956 1957
	csum = btrfs_csum_data(root, kaddr + offset, csum,  end - start + 1);
	btrfs_csum_final(csum, (char *)&csum);
C
Chris Mason 已提交
1958
	if (csum != private)
1959
		goto zeroit;
C
Chris Mason 已提交
1960

1961
	kunmap_atomic(kaddr, KM_USER0);
1962
good:
1963 1964 1965
	/* if the io failure tree for this inode is non-empty,
	 * check to see if we've recovered from a failed IO
	 */
1966
	btrfs_clean_io_failures(inode, start);
1967 1968 1969
	return 0;

zeroit:
C
Chris Mason 已提交
1970 1971 1972 1973 1974 1975
	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);
	}
1976 1977
	memset(kaddr + offset, 1, end - start + 1);
	flush_dcache_page(page);
1978
	kunmap_atomic(kaddr, KM_USER0);
1979 1980
	if (private == 0)
		return 0;
1981
	return -EIO;
1982
}
1983

Y
Yan, Zheng 已提交
1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 2020 2021 2022 2023 2024 2025 2026 2027 2028 2029 2030 2031
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);
}

2032 2033 2034 2035 2036 2037 2038 2039 2040 2041 2042 2043 2044 2045 2046 2047 2048 2049 2050 2051 2052 2053 2054 2055 2056 2057 2058 2059 2060 2061 2062 2063 2064 2065 2066 2067 2068 2069 2070 2071 2072 2073 2074 2075 2076 2077 2078 2079 2080 2081 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
/*
 * 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;
	}
}

2149 2150 2151
/*
 * This creates an orphan entry for the given inode in case something goes
 * wrong in the middle of an unlink/truncate.
2152 2153 2154
 *
 * NOTE: caller of this function should reserve 5 units of metadata for
 *	 this function.
2155 2156 2157 2158
 */
int btrfs_orphan_add(struct btrfs_trans_handle *trans, struct inode *inode)
{
	struct btrfs_root *root = BTRFS_I(inode)->root;
2159 2160 2161 2162
	struct btrfs_block_rsv *block_rsv = NULL;
	int reserve = 0;
	int insert = 0;
	int ret;
2163

2164 2165 2166 2167
	if (!root->orphan_block_rsv) {
		block_rsv = btrfs_alloc_block_rsv(root);
		BUG_ON(!block_rsv);
	}
2168

2169 2170 2171 2172 2173 2174
	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;
2175 2176
	}

2177 2178 2179 2180 2181 2182 2183 2184 2185 2186 2187 2188 2189 2190 2191 2192
	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);
2193 2194
	}

2195 2196 2197 2198 2199
	if (!BTRFS_I(inode)->orphan_meta_reserved) {
		BTRFS_I(inode)->orphan_meta_reserved = 1;
		reserve = 1;
	}
	spin_unlock(&root->orphan_lock);
2200

2201 2202
	if (block_rsv)
		btrfs_add_durable_block_rsv(root->fs_info, block_rsv);
2203

2204 2205 2206 2207 2208
	/* grab metadata reservation from transaction handle */
	if (reserve) {
		ret = btrfs_orphan_reserve_metadata(trans, inode);
		BUG_ON(ret);
	}
2209

2210 2211 2212 2213 2214 2215 2216 2217 2218 2219 2220 2221 2222
	/* 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;
2223 2224 2225 2226 2227 2228 2229 2230 2231
}

/*
 * 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;
2232 2233
	int delete_item = 0;
	int release_rsv = 0;
2234 2235
	int ret = 0;

2236 2237 2238 2239
	spin_lock(&root->orphan_lock);
	if (!list_empty(&BTRFS_I(inode)->i_orphan)) {
		list_del_init(&BTRFS_I(inode)->i_orphan);
		delete_item = 1;
2240 2241
	}

2242 2243 2244
	if (BTRFS_I(inode)->orphan_meta_reserved) {
		BTRFS_I(inode)->orphan_meta_reserved = 0;
		release_rsv = 1;
2245
	}
2246
	spin_unlock(&root->orphan_lock);
2247

2248 2249 2250 2251
	if (trans && delete_item) {
		ret = btrfs_del_orphan_item(trans, root, inode->i_ino);
		BUG_ON(ret);
	}
2252

2253 2254
	if (release_rsv)
		btrfs_orphan_release_metadata(inode);
2255

2256
	return 0;
2257 2258 2259 2260 2261 2262 2263 2264 2265 2266 2267 2268 2269 2270 2271
}

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

2272
	if (cmpxchg(&root->orphan_cleanup_state, 0, ORPHAN_CLEANUP_STARTED))
2273
		return;
2274 2275 2276

	path = btrfs_alloc_path();
	BUG_ON(!path);
2277 2278 2279 2280 2281 2282 2283 2284 2285 2286 2287 2288 2289 2290 2291 2292 2293 2294 2295 2296 2297 2298 2299 2300 2301 2302 2303 2304 2305 2306 2307 2308 2309 2310 2311 2312 2313 2314 2315 2316 2317 2318 2319
	path->reada = -1;

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

	while (1) {
		ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
		if (ret < 0) {
			printk(KERN_ERR "Error searching slot for orphan: %d"
			       "\n", ret);
			break;
		}

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

		/* pull out the item */
		leaf = path->nodes[0];
		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.
		 */
2320 2321 2322
		found_key.objectid = found_key.offset;
		found_key.type = BTRFS_INODE_ITEM_KEY;
		found_key.offset = 0;
2323
		inode = btrfs_iget(root->fs_info->sb, &found_key, root, NULL);
2324
		BUG_ON(IS_ERR(inode));
2325 2326 2327 2328 2329

		/*
		 * add this inode to the orphan list so btrfs_orphan_del does
		 * the proper thing when we hit it
		 */
2330
		spin_lock(&root->orphan_lock);
2331
		list_add(&BTRFS_I(inode)->i_orphan, &root->orphan_list);
2332
		spin_unlock(&root->orphan_lock);
2333 2334 2335 2336 2337 2338 2339 2340

		/*
		 * 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)) {
2341
			trans = btrfs_start_transaction(root, 0);
2342
			btrfs_orphan_del(trans, inode);
2343
			btrfs_end_transaction(trans, root);
2344 2345 2346 2347 2348 2349 2350 2351 2352 2353 2354 2355 2356 2357 2358
			iput(inode);
			continue;
		}

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

		/* this will do delete_inode and everything for us */
		iput(inode);
	}
2359 2360 2361 2362 2363 2364 2365 2366 2367 2368 2369 2370
	btrfs_free_path(path);

	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);
		btrfs_end_transaction(trans, root);
	}
2371 2372 2373 2374 2375 2376 2377

	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);
}

2378 2379 2380 2381 2382 2383 2384 2385 2386 2387 2388 2389 2390 2391 2392 2393 2394 2395 2396 2397 2398 2399 2400 2401 2402 2403 2404 2405 2406 2407 2408 2409 2410 2411 2412 2413 2414 2415 2416 2417 2418 2419 2420 2421 2422 2423 2424 2425 2426 2427 2428
/*
 * 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
2429 2430 2431
/*
 * read an inode from the btree into the in-memory inode
 */
2432
static void btrfs_read_locked_inode(struct inode *inode)
C
Chris Mason 已提交
2433 2434
{
	struct btrfs_path *path;
2435
	struct extent_buffer *leaf;
C
Chris Mason 已提交
2436
	struct btrfs_inode_item *inode_item;
2437
	struct btrfs_timespec *tspec;
C
Chris Mason 已提交
2438 2439
	struct btrfs_root *root = BTRFS_I(inode)->root;
	struct btrfs_key location;
2440
	int maybe_acls;
C
Chris Mason 已提交
2441
	u64 alloc_group_block;
J
Josef Bacik 已提交
2442
	u32 rdev;
C
Chris Mason 已提交
2443 2444 2445 2446 2447
	int ret;

	path = btrfs_alloc_path();
	BUG_ON(!path);
	memcpy(&location, &BTRFS_I(inode)->location, sizeof(location));
C
Chris Mason 已提交
2448

C
Chris Mason 已提交
2449
	ret = btrfs_lookup_inode(NULL, root, path, &location, 0);
2450
	if (ret)
C
Chris Mason 已提交
2451 2452
		goto make_bad;

2453 2454 2455 2456 2457 2458 2459 2460
	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);
2461
	btrfs_i_size_write(inode, btrfs_inode_size(leaf, inode_item));
2462 2463 2464 2465 2466 2467 2468 2469 2470 2471 2472 2473 2474

	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);

2475
	inode_set_bytes(inode, btrfs_inode_nbytes(leaf, inode_item));
2476
	BTRFS_I(inode)->generation = btrfs_inode_generation(leaf, inode_item);
2477
	BTRFS_I(inode)->sequence = btrfs_inode_sequence(leaf, inode_item);
2478
	inode->i_generation = BTRFS_I(inode)->generation;
J
Josef Bacik 已提交
2479
	inode->i_rdev = 0;
2480 2481
	rdev = btrfs_inode_rdev(leaf, inode_item);

2482
	BTRFS_I(inode)->index_cnt = (u64)-1;
2483
	BTRFS_I(inode)->flags = btrfs_inode_flags(leaf, inode_item);
2484

2485
	alloc_group_block = btrfs_inode_block_group(leaf, inode_item);
2486

2487 2488 2489 2490 2491
	/*
	 * 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);
2492 2493
	if (!maybe_acls)
		cache_no_acl(inode);
2494

2495 2496
	BTRFS_I(inode)->block_group = btrfs_find_block_group(root, 0,
						alloc_group_block, 0);
C
Chris Mason 已提交
2497 2498 2499 2500 2501 2502
	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
2503
		inode->i_mapping->backing_dev_info = &root->fs_info->bdi;
2504
		BTRFS_I(inode)->io_tree.ops = &btrfs_extent_io_ops;
C
Chris Mason 已提交
2505 2506 2507 2508 2509 2510 2511 2512 2513 2514 2515 2516 2517
		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
2518
		inode->i_mapping->backing_dev_info = &root->fs_info->bdi;
C
Chris Mason 已提交
2519
		break;
J
Josef Bacik 已提交
2520
	default:
J
Jim Owens 已提交
2521
		inode->i_op = &btrfs_special_inode_operations;
J
Josef Bacik 已提交
2522 2523
		init_special_inode(inode, inode->i_mode, rdev);
		break;
C
Chris Mason 已提交
2524
	}
2525 2526

	btrfs_update_iflags(inode);
C
Chris Mason 已提交
2527 2528 2529 2530 2531 2532 2533
	return;

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

C
Chris Mason 已提交
2534 2535 2536
/*
 * given a leaf and an inode, copy the inode fields into the leaf
 */
2537 2538
static void fill_inode_item(struct btrfs_trans_handle *trans,
			    struct extent_buffer *leaf,
2539
			    struct btrfs_inode_item *item,
C
Chris Mason 已提交
2540 2541
			    struct inode *inode)
{
2542 2543
	btrfs_set_inode_uid(leaf, item, inode->i_uid);
	btrfs_set_inode_gid(leaf, item, inode->i_gid);
2544
	btrfs_set_inode_size(leaf, item, BTRFS_I(inode)->disk_i_size);
2545 2546 2547 2548 2549 2550 2551 2552 2553 2554 2555 2556 2557 2558 2559 2560 2561 2562
	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);

2563
	btrfs_set_inode_nbytes(leaf, item, inode_get_bytes(inode));
2564
	btrfs_set_inode_generation(leaf, item, BTRFS_I(inode)->generation);
2565
	btrfs_set_inode_sequence(leaf, item, BTRFS_I(inode)->sequence);
2566
	btrfs_set_inode_transid(leaf, item, trans->transid);
2567
	btrfs_set_inode_rdev(leaf, item, inode->i_rdev);
Y
Yan 已提交
2568
	btrfs_set_inode_flags(leaf, item, BTRFS_I(inode)->flags);
2569
	btrfs_set_inode_block_group(leaf, item, BTRFS_I(inode)->block_group);
C
Chris Mason 已提交
2570 2571
}

C
Chris Mason 已提交
2572 2573 2574
/*
 * copy everything in the in-memory inode into the btree.
 */
C
Chris Mason 已提交
2575 2576
noinline int btrfs_update_inode(struct btrfs_trans_handle *trans,
				struct btrfs_root *root, struct inode *inode)
C
Chris Mason 已提交
2577 2578 2579
{
	struct btrfs_inode_item *inode_item;
	struct btrfs_path *path;
2580
	struct extent_buffer *leaf;
C
Chris Mason 已提交
2581 2582 2583 2584
	int ret;

	path = btrfs_alloc_path();
	BUG_ON(!path);
2585
	path->leave_spinning = 1;
C
Chris Mason 已提交
2586 2587 2588 2589 2590 2591 2592 2593
	ret = btrfs_lookup_inode(trans, root, path,
				 &BTRFS_I(inode)->location, 1);
	if (ret) {
		if (ret > 0)
			ret = -ENOENT;
		goto failed;
	}

2594
	btrfs_unlock_up_safe(path, 1);
2595 2596
	leaf = path->nodes[0];
	inode_item = btrfs_item_ptr(leaf, path->slots[0],
C
Chris Mason 已提交
2597 2598
				  struct btrfs_inode_item);

2599
	fill_inode_item(trans, leaf, inode_item, inode);
2600
	btrfs_mark_buffer_dirty(leaf);
2601
	btrfs_set_inode_last_trans(trans, inode);
C
Chris Mason 已提交
2602 2603 2604 2605 2606 2607 2608
	ret = 0;
failed:
	btrfs_free_path(path);
	return ret;
}


C
Chris Mason 已提交
2609 2610 2611 2612 2613
/*
 * 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
 */
2614 2615 2616 2617
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
2618 2619 2620
{
	struct btrfs_path *path;
	int ret = 0;
2621
	struct extent_buffer *leaf;
C
Chris Mason 已提交
2622
	struct btrfs_dir_item *di;
2623
	struct btrfs_key key;
2624
	u64 index;
C
Chris Mason 已提交
2625 2626

	path = btrfs_alloc_path();
2627 2628 2629 2630 2631
	if (!path) {
		ret = -ENOMEM;
		goto err;
	}

2632
	path->leave_spinning = 1;
C
Chris Mason 已提交
2633 2634 2635 2636 2637 2638 2639 2640 2641 2642
	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;
	}
2643 2644
	leaf = path->nodes[0];
	btrfs_dir_item_key_to_cpu(leaf, di, &key);
C
Chris Mason 已提交
2645
	ret = btrfs_delete_one_dir_name(trans, root, path, di);
2646 2647
	if (ret)
		goto err;
C
Chris Mason 已提交
2648 2649
	btrfs_release_path(root, path);

2650
	ret = btrfs_del_inode_ref(trans, root, name, name_len,
2651 2652
				  inode->i_ino,
				  dir->i_ino, &index);
2653
	if (ret) {
C
Chris Mason 已提交
2654
		printk(KERN_INFO "btrfs failed to delete reference to %.*s, "
2655
		       "inode %lu parent %lu\n", name_len, name,
2656
		       inode->i_ino, dir->i_ino);
2657 2658 2659
		goto err;
	}

C
Chris Mason 已提交
2660
	di = btrfs_lookup_dir_index_item(trans, root, path, dir->i_ino,
2661
					 index, name, name_len, -1);
C
Chris Mason 已提交
2662 2663 2664 2665 2666 2667 2668 2669 2670
	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);
2671
	btrfs_release_path(root, path);
C
Chris Mason 已提交
2672

2673 2674
	ret = btrfs_del_inode_ref_in_log(trans, root, name, name_len,
					 inode, dir->i_ino);
C
Chris Mason 已提交
2675
	BUG_ON(ret != 0 && ret != -ENOENT);
2676 2677 2678

	ret = btrfs_del_dir_entries_in_log(trans, root, name, name_len,
					   dir, index);
2679 2680
	if (ret == -ENOENT)
		ret = 0;
C
Chris Mason 已提交
2681 2682
err:
	btrfs_free_path(path);
2683 2684 2685 2686 2687 2688 2689 2690 2691
	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
2692 2693 2694
	return ret;
}

2695 2696 2697
/* 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
2698
{
2699 2700
	struct extent_buffer *eb;
	int level;
2701
	u64 refs = 1;
2702
	int uninitialized_var(ret);
2703

2704 2705 2706 2707 2708 2709 2710 2711 2712 2713
	for (level = 0; level < BTRFS_MAX_LEVEL; level++) {
		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;
2714
	}
2715
	return ret; /* XXX callers? */
C
Chris Mason 已提交
2716 2717
}

2718 2719 2720 2721 2722 2723 2724 2725 2726
/*
 * 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)
2727
{
C
Chris Mason 已提交
2728
	struct btrfs_trans_handle *trans;
2729
	struct btrfs_root *root = BTRFS_I(dir)->root;
2730
	struct btrfs_path *path;
2731
	struct btrfs_inode_ref *ref;
2732
	struct btrfs_dir_item *di;
2733
	struct inode *inode = dentry->d_inode;
2734
	u64 index;
2735 2736
	int check_link = 1;
	int err = -ENOSPC;
2737 2738
	int ret;

2739 2740 2741
	trans = btrfs_start_transaction(root, 10);
	if (!IS_ERR(trans) || PTR_ERR(trans) != -ENOSPC)
		return trans;
2742

2743 2744
	if (inode->i_ino == BTRFS_EMPTY_SUBVOL_DIR_OBJECTID)
		return ERR_PTR(-ENOSPC);
2745

2746 2747 2748
	/* check if there is someone else holds reference */
	if (S_ISDIR(inode->i_mode) && atomic_read(&inode->i_count) > 1)
		return ERR_PTR(-ENOSPC);
2749

2750 2751
	if (atomic_read(&inode->i_count) > 2)
		return ERR_PTR(-ENOSPC);
2752

2753 2754 2755 2756 2757 2758 2759
	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);
2760 2761
	}

2762
	trans = btrfs_start_transaction(root, 0);
2763
	if (IS_ERR(trans)) {
2764 2765 2766 2767
		btrfs_free_path(path);
		root->fs_info->enospc_unlink = 0;
		return trans;
	}
2768

2769 2770
	path->skip_locking = 1;
	path->search_commit_root = 1;
2771

2772 2773 2774 2775 2776 2777 2778 2779 2780 2781 2782
	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;
2783
	}
2784 2785 2786 2787 2788 2789 2790 2791 2792 2793 2794 2795 2796 2797 2798 2799 2800 2801 2802 2803 2804 2805 2806 2807 2808 2809 2810 2811 2812 2813 2814 2815 2816 2817 2818 2819 2820 2821 2822 2823 2824 2825 2826 2827 2828 2829 2830 2831 2832 2833 2834 2835 2836 2837 2838 2839 2840 2841 2842 2843 2844 2845 2846 2847 2848 2849 2850 2851 2852 2853 2854 2855 2856 2857 2858 2859 2860 2861 2862 2863 2864 2865 2866 2867 2868 2869 2870 2871 2872 2873 2874 2875 2876 2877 2878 2879 2880 2881 2882 2883 2884 2885 2886 2887 2888 2889
	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);
2890

C
Chris Mason 已提交
2891
	btrfs_set_trans_block_group(trans, dir);
2892 2893 2894

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

2895 2896
	ret = btrfs_unlink_inode(trans, root, dir, dentry->d_inode,
				 dentry->d_name.name, dentry->d_name.len);
2897
	BUG_ON(ret);
2898

2899
	if (inode->i_nlink == 0) {
2900
		ret = btrfs_orphan_add(trans, inode);
2901 2902
		BUG_ON(ret);
	}
2903

2904
	nr = trans->blocks_used;
2905
	__unlink_end_trans(trans, root);
2906
	btrfs_btree_balance_dirty(root, nr);
C
Chris Mason 已提交
2907 2908 2909
	return ret;
}

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 2935 2936 2937 2938 2939 2940 2941 2942 2943 2944 2945 2946 2947 2948 2949 2950 2951 2952 2953 2954 2955 2956 2957 2958 2959 2960 2961 2962 2963 2964 2965 2966 2967 2968 2969 2970 2971
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
2972 2973 2974
static int btrfs_rmdir(struct inode *dir, struct dentry *dentry)
{
	struct inode *inode = dentry->d_inode;
2975
	int err = 0;
C
Chris Mason 已提交
2976 2977
	struct btrfs_root *root = BTRFS_I(dir)->root;
	struct btrfs_trans_handle *trans;
2978
	unsigned long nr = 0;
C
Chris Mason 已提交
2979

2980
	if (inode->i_size > BTRFS_EMPTY_DIR_SIZE ||
2981
	    inode->i_ino == BTRFS_FIRST_FREE_OBJECTID)
Y
Yan 已提交
2982 2983
		return -ENOTEMPTY;

2984 2985
	trans = __unlink_start_trans(dir, dentry);
	if (IS_ERR(trans))
2986 2987
		return PTR_ERR(trans);

C
Chris Mason 已提交
2988 2989
	btrfs_set_trans_block_group(trans, dir);

2990 2991 2992 2993 2994 2995 2996 2997
	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;
	}

2998 2999
	err = btrfs_orphan_add(trans, inode);
	if (err)
3000
		goto out;
3001

C
Chris Mason 已提交
3002
	/* now the directory is empty */
3003 3004
	err = btrfs_unlink_inode(trans, root, dir, dentry->d_inode,
				 dentry->d_name.name, dentry->d_name.len);
C
Chris Mason 已提交
3005
	if (!err)
3006
		btrfs_i_size_write(inode, 0);
3007
out:
3008
	nr = trans->blocks_used;
3009
	__unlink_end_trans(trans, root);
3010
	btrfs_btree_balance_dirty(root, nr);
3011

C
Chris Mason 已提交
3012 3013 3014
	return err;
}

3015
#if 0
3016 3017 3018 3019 3020 3021 3022 3023 3024 3025 3026 3027 3028 3029 3030 3031 3032 3033 3034 3035 3036 3037 3038 3039 3040 3041 3042 3043
/*
 * 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
3044 3045 3046
	struct btrfs_leaf_ref *ref;
	u64 leaf_gen;
	u64 leaf_start;
3047 3048 3049 3050 3051 3052 3053 3054 3055 3056 3057 3058 3059 3060 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 3089 3090 3091 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

	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
3140 3141
	leaf_start = btrfs_node_blockptr(path->nodes[1], path->slots[1]);
	leaf_gen = btrfs_node_ptr_generation(path->nodes[1], path->slots[1]);
3142 3143 3144 3145
	/*
	 * 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
3146
	ret = btrfs_del_leaf(trans, root, path, leaf_start);
3147 3148
	BUG_ON(ret);

Y
Yan Zheng 已提交
3149 3150 3151 3152 3153 3154 3155 3156 3157
	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;

3158 3159
			btrfs_sort_leaf_ref(ref);

Y
Yan Zheng 已提交
3160 3161 3162 3163 3164 3165 3166
			ret = btrfs_add_leaf_ref(root, ref, 0);
			WARN_ON(ret);
			btrfs_free_leaf_ref(root, ref);
		} else {
			WARN_ON(1);
		}
	}
3167 3168 3169 3170 3171 3172 3173 3174 3175 3176 3177 3178 3179 3180 3181 3182 3183 3184
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;
}

3185 3186
#endif

C
Chris Mason 已提交
3187 3188 3189
/*
 * 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
3190
 * any higher than new_size
C
Chris Mason 已提交
3191 3192 3193
 *
 * csum items that cross the new i_size are truncated to the new size
 * as well.
3194 3195 3196
 *
 * 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
3197
 */
3198 3199 3200 3201
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
3202 3203
{
	struct btrfs_path *path;
3204
	struct extent_buffer *leaf;
C
Chris Mason 已提交
3205
	struct btrfs_file_extent_item *fi;
3206 3207
	struct btrfs_key key;
	struct btrfs_key found_key;
C
Chris Mason 已提交
3208
	u64 extent_start = 0;
3209
	u64 extent_num_bytes = 0;
3210
	u64 extent_offset = 0;
C
Chris Mason 已提交
3211
	u64 item_end = 0;
3212 3213
	u64 mask = root->sectorsize - 1;
	u32 found_type = (u8)-1;
C
Chris Mason 已提交
3214 3215
	int found_extent;
	int del_item;
3216 3217
	int pending_del_nr = 0;
	int pending_del_slot = 0;
3218
	int extent_type = -1;
3219
	int encoding;
3220 3221 3222 3223
	int ret;
	int err = 0;

	BUG_ON(new_size > 0 && min_type != BTRFS_EXTENT_DATA_KEY);
C
Chris Mason 已提交
3224

3225
	if (root->ref_cows || root == root->fs_info->tree_root)
3226
		btrfs_drop_extent_cache(inode, new_size & (~mask), (u64)-1, 0);
3227

C
Chris Mason 已提交
3228 3229
	path = btrfs_alloc_path();
	BUG_ON(!path);
J
Julia Lawall 已提交
3230
	path->reada = -1;
3231

C
Chris Mason 已提交
3232 3233
	key.objectid = inode->i_ino;
	key.offset = (u64)-1;
3234 3235
	key.type = (u8)-1;

3236
search_again:
3237
	path->leave_spinning = 1;
3238
	ret = btrfs_search_slot(trans, root, &key, path, -1, 1);
3239 3240 3241 3242
	if (ret < 0) {
		err = ret;
		goto out;
	}
C
Chris Mason 已提交
3243

3244
	if (ret > 0) {
3245 3246 3247
		/* there are no items in the tree for us to truncate, we're
		 * done
		 */
3248 3249
		if (path->slots[0] == 0)
			goto out;
3250 3251 3252
		path->slots[0]--;
	}

C
Chris Mason 已提交
3253
	while (1) {
C
Chris Mason 已提交
3254
		fi = NULL;
3255 3256 3257
		leaf = path->nodes[0];
		btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
		found_type = btrfs_key_type(&found_key);
3258
		encoding = 0;
C
Chris Mason 已提交
3259

3260
		if (found_key.objectid != inode->i_ino)
C
Chris Mason 已提交
3261
			break;
3262

3263
		if (found_type < min_type)
C
Chris Mason 已提交
3264 3265
			break;

3266
		item_end = found_key.offset;
C
Chris Mason 已提交
3267
		if (found_type == BTRFS_EXTENT_DATA_KEY) {
3268
			fi = btrfs_item_ptr(leaf, path->slots[0],
C
Chris Mason 已提交
3269
					    struct btrfs_file_extent_item);
3270
			extent_type = btrfs_file_extent_type(leaf, fi);
3271 3272 3273 3274
			encoding = btrfs_file_extent_compression(leaf, fi);
			encoding |= btrfs_file_extent_encryption(leaf, fi);
			encoding |= btrfs_file_extent_other_encoding(leaf, fi);

3275
			if (extent_type != BTRFS_FILE_EXTENT_INLINE) {
3276
				item_end +=
3277
				    btrfs_file_extent_num_bytes(leaf, fi);
3278 3279
			} else if (extent_type == BTRFS_FILE_EXTENT_INLINE) {
				item_end += btrfs_file_extent_inline_len(leaf,
C
Chris Mason 已提交
3280
									 fi);
C
Chris Mason 已提交
3281
			}
3282
			item_end--;
C
Chris Mason 已提交
3283
		}
3284 3285 3286 3287
		if (found_type > min_type) {
			del_item = 1;
		} else {
			if (item_end < new_size)
3288
				break;
3289 3290 3291 3292
			if (found_key.offset >= new_size)
				del_item = 1;
			else
				del_item = 0;
C
Chris Mason 已提交
3293 3294 3295
		}
		found_extent = 0;
		/* FIXME, shrink the extent if the ref count is only 1 */
3296 3297 3298 3299
		if (found_type != BTRFS_EXTENT_DATA_KEY)
			goto delete;

		if (extent_type != BTRFS_FILE_EXTENT_INLINE) {
C
Chris Mason 已提交
3300
			u64 num_dec;
3301
			extent_start = btrfs_file_extent_disk_bytenr(leaf, fi);
3302
			if (!del_item && !encoding) {
3303 3304
				u64 orig_num_bytes =
					btrfs_file_extent_num_bytes(leaf, fi);
3305
				extent_num_bytes = new_size -
3306
					found_key.offset + root->sectorsize - 1;
3307 3308
				extent_num_bytes = extent_num_bytes &
					~((u64)root->sectorsize - 1);
3309 3310 3311
				btrfs_set_file_extent_num_bytes(leaf, fi,
							 extent_num_bytes);
				num_dec = (orig_num_bytes -
C
Chris Mason 已提交
3312
					   extent_num_bytes);
3313
				if (root->ref_cows && extent_start != 0)
3314
					inode_sub_bytes(inode, num_dec);
3315
				btrfs_mark_buffer_dirty(leaf);
C
Chris Mason 已提交
3316
			} else {
3317 3318 3319
				extent_num_bytes =
					btrfs_file_extent_disk_num_bytes(leaf,
									 fi);
3320 3321 3322
				extent_offset = found_key.offset -
					btrfs_file_extent_offset(leaf, fi);

C
Chris Mason 已提交
3323
				/* FIXME blocksize != 4096 */
C
Chris Mason 已提交
3324
				num_dec = btrfs_file_extent_num_bytes(leaf, fi);
C
Chris Mason 已提交
3325 3326
				if (extent_start != 0) {
					found_extent = 1;
3327
					if (root->ref_cows)
3328
						inode_sub_bytes(inode, num_dec);
3329
				}
C
Chris Mason 已提交
3330
			}
C
Chris Mason 已提交
3331
		} else if (extent_type == BTRFS_FILE_EXTENT_INLINE) {
C
Chris Mason 已提交
3332 3333 3334 3335 3336 3337 3338 3339
			/*
			 * 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) {
3340 3341 3342
				u32 size = new_size - found_key.offset;

				if (root->ref_cows) {
3343 3344
					inode_sub_bytes(inode, item_end + 1 -
							new_size);
3345 3346 3347
				}
				size =
				    btrfs_file_extent_calc_inline_size(size);
C
Chris Mason 已提交
3348
				ret = btrfs_truncate_item(trans, root, path,
3349
							  size, 1);
C
Chris Mason 已提交
3350
				BUG_ON(ret);
3351
			} else if (root->ref_cows) {
3352 3353
				inode_sub_bytes(inode, item_end + 1 -
						found_key.offset);
C
Chris Mason 已提交
3354
			}
C
Chris Mason 已提交
3355
		}
3356
delete:
C
Chris Mason 已提交
3357
		if (del_item) {
3358 3359 3360 3361 3362 3363 3364 3365 3366 3367
			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
3368
				BUG();
3369
			}
C
Chris Mason 已提交
3370 3371 3372
		} else {
			break;
		}
3373 3374
		if (found_extent && (root->ref_cows ||
				     root == root->fs_info->tree_root)) {
3375
			btrfs_set_path_blocking(path);
C
Chris Mason 已提交
3376
			ret = btrfs_free_extent(trans, root, extent_start,
3377 3378 3379
						extent_num_bytes, 0,
						btrfs_header_owner(leaf),
						inode->i_ino, extent_offset);
C
Chris Mason 已提交
3380 3381
			BUG_ON(ret);
		}
3382

3383 3384 3385 3386 3387 3388 3389 3390 3391 3392 3393 3394 3395 3396 3397 3398
		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;
			}
3399 3400
			btrfs_release_path(root, path);
			goto search_again;
3401 3402
		} else {
			path->slots[0]--;
3403
		}
C
Chris Mason 已提交
3404
	}
3405
out:
3406 3407 3408
	if (pending_del_nr) {
		ret = btrfs_del_items(trans, root, path, pending_del_slot,
				      pending_del_nr);
3409
		BUG_ON(ret);
3410
	}
C
Chris Mason 已提交
3411
	btrfs_free_path(path);
3412
	return err;
C
Chris Mason 已提交
3413 3414 3415 3416 3417 3418 3419 3420 3421
}

/*
 * 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;
3422
	struct btrfs_root *root = BTRFS_I(inode)->root;
3423 3424
	struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
	struct btrfs_ordered_extent *ordered;
3425
	struct extent_state *cached_state = NULL;
3426
	char *kaddr;
3427
	u32 blocksize = root->sectorsize;
C
Chris Mason 已提交
3428 3429 3430 3431
	pgoff_t index = from >> PAGE_CACHE_SHIFT;
	unsigned offset = from & (PAGE_CACHE_SIZE-1);
	struct page *page;
	int ret = 0;
3432
	u64 page_start;
3433
	u64 page_end;
C
Chris Mason 已提交
3434 3435 3436

	if ((offset & (blocksize - 1)) == 0)
		goto out;
3437
	ret = btrfs_delalloc_reserve_space(inode, PAGE_CACHE_SIZE);
3438 3439
	if (ret)
		goto out;
C
Chris Mason 已提交
3440 3441

	ret = -ENOMEM;
3442
again:
C
Chris Mason 已提交
3443
	page = grab_cache_page(mapping, index);
3444
	if (!page) {
3445
		btrfs_delalloc_release_space(inode, PAGE_CACHE_SIZE);
C
Chris Mason 已提交
3446
		goto out;
3447
	}
3448 3449 3450 3451

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

C
Chris Mason 已提交
3452
	if (!PageUptodate(page)) {
C
Chris Mason 已提交
3453
		ret = btrfs_readpage(NULL, page);
C
Chris Mason 已提交
3454
		lock_page(page);
3455 3456 3457 3458 3459
		if (page->mapping != mapping) {
			unlock_page(page);
			page_cache_release(page);
			goto again;
		}
C
Chris Mason 已提交
3460 3461
		if (!PageUptodate(page)) {
			ret = -EIO;
3462
			goto out_unlock;
C
Chris Mason 已提交
3463 3464
		}
	}
3465
	wait_on_page_writeback(page);
3466

3467 3468
	lock_extent_bits(io_tree, page_start, page_end, 0, &cached_state,
			 GFP_NOFS);
3469 3470 3471 3472
	set_page_extent_mapped(page);

	ordered = btrfs_lookup_ordered_extent(inode, page_start);
	if (ordered) {
3473 3474
		unlock_extent_cached(io_tree, page_start, page_end,
				     &cached_state, GFP_NOFS);
3475 3476
		unlock_page(page);
		page_cache_release(page);
3477
		btrfs_start_ordered_extent(inode, ordered, 1);
3478 3479 3480 3481
		btrfs_put_ordered_extent(ordered);
		goto again;
	}

3482
	clear_extent_bit(&BTRFS_I(inode)->io_tree, page_start, page_end,
3483
			  EXTENT_DIRTY | EXTENT_DELALLOC | EXTENT_DO_ACCOUNTING,
3484
			  0, 0, &cached_state, GFP_NOFS);
3485

3486 3487
	ret = btrfs_set_extent_delalloc(inode, page_start, page_end,
					&cached_state);
J
Josef Bacik 已提交
3488
	if (ret) {
3489 3490
		unlock_extent_cached(io_tree, page_start, page_end,
				     &cached_state, GFP_NOFS);
J
Josef Bacik 已提交
3491 3492 3493
		goto out_unlock;
	}

3494 3495 3496 3497 3498 3499 3500
	ret = 0;
	if (offset != PAGE_CACHE_SIZE) {
		kaddr = kmap(page);
		memset(kaddr + offset, 0, PAGE_CACHE_SIZE - offset);
		flush_dcache_page(page);
		kunmap(page);
	}
3501
	ClearPageChecked(page);
3502
	set_page_dirty(page);
3503 3504
	unlock_extent_cached(io_tree, page_start, page_end, &cached_state,
			     GFP_NOFS);
C
Chris Mason 已提交
3505

3506
out_unlock:
3507
	if (ret)
3508
		btrfs_delalloc_release_space(inode, PAGE_CACHE_SIZE);
C
Chris Mason 已提交
3509 3510 3511 3512 3513 3514
	unlock_page(page);
	page_cache_release(page);
out:
	return ret;
}

Y
Yan Zheng 已提交
3515
int btrfs_cont_expand(struct inode *inode, loff_t size)
C
Chris Mason 已提交
3516
{
Y
Yan Zheng 已提交
3517 3518 3519
	struct btrfs_trans_handle *trans;
	struct btrfs_root *root = BTRFS_I(inode)->root;
	struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
3520
	struct extent_map *em = NULL;
3521
	struct extent_state *cached_state = NULL;
Y
Yan Zheng 已提交
3522 3523 3524 3525 3526 3527
	u64 mask = root->sectorsize - 1;
	u64 hole_start = (inode->i_size + mask) & ~mask;
	u64 block_end = (size + mask) & ~mask;
	u64 last_byte;
	u64 cur_offset;
	u64 hole_size;
J
Josef Bacik 已提交
3528
	int err = 0;
C
Chris Mason 已提交
3529

Y
Yan Zheng 已提交
3530 3531 3532 3533 3534 3535 3536
	if (size <= hole_start)
		return 0;

	while (1) {
		struct btrfs_ordered_extent *ordered;
		btrfs_wait_ordered_range(inode, hole_start,
					 block_end - hole_start);
3537 3538
		lock_extent_bits(io_tree, hole_start, block_end - 1, 0,
				 &cached_state, GFP_NOFS);
Y
Yan Zheng 已提交
3539 3540 3541
		ordered = btrfs_lookup_ordered_extent(inode, hole_start);
		if (!ordered)
			break;
3542 3543
		unlock_extent_cached(io_tree, hole_start, block_end - 1,
				     &cached_state, GFP_NOFS);
Y
Yan Zheng 已提交
3544 3545
		btrfs_put_ordered_extent(ordered);
	}
C
Chris Mason 已提交
3546

Y
Yan Zheng 已提交
3547 3548 3549 3550 3551 3552 3553
	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;
3554
		if (!test_bit(EXTENT_FLAG_PREALLOC, &em->flags)) {
3555
			u64 hint_byte = 0;
Y
Yan Zheng 已提交
3556
			hole_size = last_byte - cur_offset;
J
Josef Bacik 已提交
3557

3558 3559 3560
			trans = btrfs_start_transaction(root, 2);
			if (IS_ERR(trans)) {
				err = PTR_ERR(trans);
J
Josef Bacik 已提交
3561
				break;
3562
			}
3563 3564 3565 3566 3567 3568 3569
			btrfs_set_trans_block_group(trans, inode);

			err = btrfs_drop_extents(trans, inode, cur_offset,
						 cur_offset + hole_size,
						 &hint_byte, 1);
			BUG_ON(err);

Y
Yan Zheng 已提交
3570 3571 3572 3573
			err = btrfs_insert_file_extent(trans, root,
					inode->i_ino, cur_offset, 0,
					0, hole_size, 0, hole_size,
					0, 0, 0);
3574 3575
			BUG_ON(err);

Y
Yan Zheng 已提交
3576 3577
			btrfs_drop_extent_cache(inode, hole_start,
					last_byte - 1, 0);
3578 3579

			btrfs_end_transaction(trans, root);
Y
Yan Zheng 已提交
3580 3581
		}
		free_extent_map(em);
3582
		em = NULL;
Y
Yan Zheng 已提交
3583
		cur_offset = last_byte;
3584
		if (cur_offset >= block_end)
Y
Yan Zheng 已提交
3585 3586
			break;
	}
3587

3588
	free_extent_map(em);
3589 3590
	unlock_extent_cached(io_tree, hole_start, block_end - 1, &cached_state,
			     GFP_NOFS);
Y
Yan Zheng 已提交
3591 3592
	return err;
}
C
Chris Mason 已提交
3593

3594 3595 3596 3597 3598 3599 3600 3601 3602 3603 3604 3605 3606 3607 3608 3609 3610 3611 3612 3613 3614
static int btrfs_setattr_size(struct inode *inode, struct iattr *attr)
{
	struct btrfs_root *root = BTRFS_I(inode)->root;
	struct btrfs_trans_handle *trans;
	unsigned long nr;
	int ret;

	if (attr->ia_size == inode->i_size)
		return 0;

	if (attr->ia_size > inode->i_size) {
		unsigned long limit;
		limit = current->signal->rlim[RLIMIT_FSIZE].rlim_cur;
		if (attr->ia_size > inode->i_sb->s_maxbytes)
			return -EFBIG;
		if (limit != RLIM_INFINITY && attr->ia_size > limit) {
			send_sig(SIGXFSZ, current, 0);
			return -EFBIG;
		}
	}

3615 3616 3617
	trans = btrfs_start_transaction(root, 5);
	if (IS_ERR(trans))
		return PTR_ERR(trans);
3618 3619 3620 3621 3622 3623 3624 3625 3626 3627 3628 3629 3630 3631 3632 3633 3634 3635 3636 3637

	btrfs_set_trans_block_group(trans, inode);

	ret = btrfs_orphan_add(trans, inode);
	BUG_ON(ret);

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

	if (attr->ia_size > inode->i_size) {
		ret = btrfs_cont_expand(inode, attr->ia_size);
		if (ret) {
			btrfs_truncate(inode);
			return ret;
		}

		i_size_write(inode, attr->ia_size);
		btrfs_ordered_update_i_size(inode, inode->i_size, NULL);

3638 3639
		trans = btrfs_start_transaction(root, 0);
		BUG_ON(IS_ERR(trans));
3640
		btrfs_set_trans_block_group(trans, inode);
3641 3642
		trans->block_rsv = root->orphan_block_rsv;
		BUG_ON(!trans->block_rsv);
3643 3644 3645 3646 3647 3648 3649 3650 3651 3652 3653 3654 3655 3656 3657 3658 3659 3660 3661 3662 3663 3664 3665 3666 3667 3668 3669 3670

		ret = btrfs_update_inode(trans, root, inode);
		BUG_ON(ret);
		if (inode->i_nlink > 0) {
			ret = btrfs_orphan_del(trans, inode);
			BUG_ON(ret);
		}
		nr = trans->blocks_used;
		btrfs_end_transaction(trans, root);
		btrfs_btree_balance_dirty(root, nr);
		return 0;
	}

	/*
	 * We're truncating a file that used to have good data down to
	 * zero. Make sure it gets into the ordered flush list so that
	 * any new writes get down to disk quickly.
	 */
	if (attr->ia_size == 0)
		BTRFS_I(inode)->ordered_data_close = 1;

	/* we don't support swapfiles, so vmtruncate shouldn't fail */
	ret = vmtruncate(inode, attr->ia_size);
	BUG_ON(ret);

	return 0;
}

Y
Yan Zheng 已提交
3671 3672 3673 3674
static int btrfs_setattr(struct dentry *dentry, struct iattr *attr)
{
	struct inode *inode = dentry->d_inode;
	int err;
C
Chris Mason 已提交
3675

Y
Yan Zheng 已提交
3676 3677 3678
	err = inode_change_ok(inode, attr);
	if (err)
		return err;
C
Chris Mason 已提交
3679

3680
	if (S_ISREG(inode->i_mode) && (attr->ia_valid & ATTR_SIZE)) {
3681 3682 3683
		err = btrfs_setattr_size(inode, attr);
		if (err)
			return err;
C
Chris Mason 已提交
3684
	}
Y
Yan Zheng 已提交
3685

C
Christoph Hellwig 已提交
3686 3687 3688 3689 3690 3691 3692
	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
3693

C
Chris Mason 已提交
3694 3695
	return err;
}
3696

A
Al Viro 已提交
3697
void btrfs_evict_inode(struct inode *inode)
C
Chris Mason 已提交
3698 3699 3700
{
	struct btrfs_trans_handle *trans;
	struct btrfs_root *root = BTRFS_I(inode)->root;
3701
	unsigned long nr;
C
Chris Mason 已提交
3702 3703 3704
	int ret;

	truncate_inode_pages(&inode->i_data, 0);
3705 3706
	if (inode->i_nlink && (btrfs_root_refs(&root->root_item) != 0 ||
			       root == root->fs_info->tree_root))
A
Al Viro 已提交
3707 3708
		goto no_delete;

C
Chris Mason 已提交
3709
	if (is_bad_inode(inode)) {
3710
		btrfs_orphan_del(NULL, inode);
C
Chris Mason 已提交
3711 3712
		goto no_delete;
	}
A
Al Viro 已提交
3713
	/* do we really want it for ->i_nlink > 0 and zero btrfs_root_refs? */
C
Chris Mason 已提交
3714
	btrfs_wait_ordered_range(inode, 0, (u64)-1);
3715

3716 3717 3718 3719 3720
	if (root->fs_info->log_root_recovering) {
		BUG_ON(!list_empty(&BTRFS_I(inode)->i_orphan));
		goto no_delete;
	}

3721 3722 3723 3724 3725
	if (inode->i_nlink > 0) {
		BUG_ON(btrfs_root_refs(&root->root_item) != 0);
		goto no_delete;
	}

3726
	btrfs_i_size_write(inode, 0);
3727

3728
	while (1) {
3729 3730
		trans = btrfs_start_transaction(root, 0);
		BUG_ON(IS_ERR(trans));
3731
		btrfs_set_trans_block_group(trans, inode);
3732 3733 3734 3735 3736 3737 3738 3739 3740 3741
		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;
		}
3742

3743
		ret = btrfs_truncate_inode_items(trans, root, inode, 0, 0);
3744 3745
		if (ret != -EAGAIN)
			break;
3746

3747 3748 3749 3750
		nr = trans->blocks_used;
		btrfs_end_transaction(trans, root);
		trans = NULL;
		btrfs_btree_balance_dirty(root, nr);
3751

3752
	}
3753

3754 3755 3756 3757
	if (ret == 0) {
		ret = btrfs_orphan_del(trans, inode);
		BUG_ON(ret);
	}
3758

3759
	nr = trans->blocks_used;
3760
	btrfs_end_transaction(trans, root);
3761
	btrfs_btree_balance_dirty(root, nr);
C
Chris Mason 已提交
3762
no_delete:
A
Al Viro 已提交
3763
	end_writeback(inode);
3764
	return;
C
Chris Mason 已提交
3765 3766 3767 3768 3769 3770 3771 3772 3773 3774 3775 3776 3777 3778
}

/*
 * 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
3779
	int ret = 0;
C
Chris Mason 已提交
3780 3781 3782

	path = btrfs_alloc_path();
	BUG_ON(!path);
3783

C
Chris Mason 已提交
3784 3785
	di = btrfs_lookup_dir_item(NULL, root, path, dir->i_ino, name,
				    namelen, 0);
Y
Yan 已提交
3786 3787
	if (IS_ERR(di))
		ret = PTR_ERR(di);
C
Chris Mason 已提交
3788 3789

	if (!di || IS_ERR(di))
3790
		goto out_err;
C
Chris Mason 已提交
3791

3792
	btrfs_dir_item_key_to_cpu(path->nodes[0], di, location);
C
Chris Mason 已提交
3793 3794 3795
out:
	btrfs_free_path(path);
	return ret;
3796 3797 3798
out_err:
	location->objectid = 0;
	goto out;
C
Chris Mason 已提交
3799 3800 3801 3802 3803 3804 3805 3806
}

/*
 * 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,
3807 3808 3809 3810
				    struct inode *dir,
				    struct dentry *dentry,
				    struct btrfs_key *location,
				    struct btrfs_root **sub_root)
C
Chris Mason 已提交
3811
{
3812 3813 3814 3815 3816 3817
	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
3818

3819 3820 3821 3822 3823
	path = btrfs_alloc_path();
	if (!path) {
		err = -ENOMEM;
		goto out;
	}
C
Chris Mason 已提交
3824

3825 3826 3827 3828 3829 3830 3831 3832 3833
	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
3834

3835 3836 3837 3838 3839
	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
3840

3841 3842 3843 3844 3845 3846 3847 3848 3849 3850 3851 3852 3853 3854 3855 3856 3857 3858 3859 3860 3861 3862 3863 3864 3865 3866 3867
	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
3868 3869
}

3870 3871 3872 3873
static void inode_tree_add(struct inode *inode)
{
	struct btrfs_root *root = BTRFS_I(inode)->root;
	struct btrfs_inode *entry;
3874 3875 3876 3877 3878
	struct rb_node **p;
	struct rb_node *parent;
again:
	p = &root->inode_tree.rb_node;
	parent = NULL;
3879

3880 3881 3882
	if (hlist_unhashed(&inode->i_hash))
		return;

3883 3884 3885 3886 3887 3888
	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)
3889
			p = &parent->rb_left;
3890
		else if (inode->i_ino > entry->vfs_inode.i_ino)
3891
			p = &parent->rb_right;
3892 3893
		else {
			WARN_ON(!(entry->vfs_inode.i_state &
A
Al Viro 已提交
3894
				  (I_WILL_FREE | I_FREEING)));
3895 3896 3897 3898
			rb_erase(parent, &root->inode_tree);
			RB_CLEAR_NODE(parent);
			spin_unlock(&root->inode_lock);
			goto again;
3899 3900 3901 3902 3903 3904 3905 3906 3907 3908
		}
	}
	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;
3909
	int empty = 0;
3910

3911
	spin_lock(&root->inode_lock);
3912 3913 3914
	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);
3915
		empty = RB_EMPTY_ROOT(&root->inode_tree);
3916
	}
3917
	spin_unlock(&root->inode_lock);
3918

3919 3920 3921 3922 3923 3924 3925 3926
	/*
	 * 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) {
3927 3928 3929 3930 3931 3932 3933 3934 3935 3936 3937 3938 3939 3940 3941 3942 3943 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
		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);
			/*
3980
			 * btrfs_drop_inode will have it removed from
3981 3982 3983 3984 3985 3986 3987 3988 3989 3990 3991 3992 3993 3994 3995 3996
			 * 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;
3997 3998
}

3999 4000 4001 4002 4003
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
4004
	btrfs_set_inode_space_info(args->root, inode);
C
Chris Mason 已提交
4005 4006 4007 4008 4009 4010
	return 0;
}

static int btrfs_find_actor(struct inode *inode, void *opaque)
{
	struct btrfs_iget_args *args = opaque;
C
Chris Mason 已提交
4011 4012
	return args->ino == inode->i_ino &&
		args->root == BTRFS_I(inode)->root;
C
Chris Mason 已提交
4013 4014
}

4015 4016 4017
static struct inode *btrfs_iget_locked(struct super_block *s,
				       u64 objectid,
				       struct btrfs_root *root)
C
Chris Mason 已提交
4018 4019 4020 4021 4022 4023 4024 4025 4026 4027 4028 4029
{
	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
4030 4031 4032 4033
/* 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,
4034
			 struct btrfs_root *root, int *new)
B
Balaji Rao 已提交
4035 4036 4037 4038 4039
{
	struct inode *inode;

	inode = btrfs_iget_locked(s, location->objectid, root);
	if (!inode)
4040
		return ERR_PTR(-ENOMEM);
B
Balaji Rao 已提交
4041 4042 4043 4044 4045

	if (inode->i_state & I_NEW) {
		BTRFS_I(inode)->root = root;
		memcpy(&BTRFS_I(inode)->location, location, sizeof(*location));
		btrfs_read_locked_inode(inode);
4046 4047

		inode_tree_add(inode);
B
Balaji Rao 已提交
4048
		unlock_new_inode(inode);
4049 4050
		if (new)
			*new = 1;
B
Balaji Rao 已提交
4051 4052 4053 4054 4055
	}

	return inode;
}

4056 4057 4058 4059 4060 4061 4062 4063 4064 4065 4066 4067 4068 4069 4070 4071 4072 4073 4074 4075 4076 4077
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;
}

4078
struct inode *btrfs_lookup_dentry(struct inode *dir, struct dentry *dentry)
C
Chris Mason 已提交
4079
{
C
Chris Mason 已提交
4080
	struct inode *inode;
4081
	struct btrfs_root *root = BTRFS_I(dir)->root;
C
Chris Mason 已提交
4082 4083
	struct btrfs_root *sub_root = root;
	struct btrfs_key location;
4084
	int index;
4085
	int ret;
C
Chris Mason 已提交
4086

4087 4088
	dentry->d_op = &btrfs_dentry_operations;

C
Chris Mason 已提交
4089 4090
	if (dentry->d_name.len > BTRFS_NAME_LEN)
		return ERR_PTR(-ENAMETOOLONG);
4091

C
Chris Mason 已提交
4092
	ret = btrfs_inode_by_name(dir, dentry, &location);
4093

C
Chris Mason 已提交
4094 4095
	if (ret < 0)
		return ERR_PTR(ret);
4096

4097 4098 4099 4100
	if (location.objectid == 0)
		return NULL;

	if (location.type == BTRFS_INODE_ITEM_KEY) {
4101
		inode = btrfs_iget(dir->i_sb, &location, root, NULL);
4102 4103 4104 4105 4106
		return inode;
	}

	BUG_ON(location.type != BTRFS_ROOT_ITEM_KEY);

4107
	index = srcu_read_lock(&root->fs_info->subvol_srcu);
4108 4109 4110 4111 4112 4113 4114 4115
	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 {
4116
		inode = btrfs_iget(dir->i_sb, &location, sub_root, NULL);
C
Chris Mason 已提交
4117
	}
4118 4119
	srcu_read_unlock(&root->fs_info->subvol_srcu, index);

4120 4121 4122 4123 4124 4125 4126
	if (root != sub_root) {
		down_read(&root->fs_info->cleanup_work_sem);
		if (!(inode->i_sb->s_flags & MS_RDONLY))
			btrfs_orphan_cleanup(sub_root);
		up_read(&root->fs_info->cleanup_work_sem);
	}

4127 4128 4129
	return inode;
}

4130 4131 4132 4133
static int btrfs_dentry_delete(struct dentry *dentry)
{
	struct btrfs_root *root;

4134 4135
	if (!dentry->d_inode && !IS_ROOT(dentry))
		dentry = dentry->d_parent;
4136

4137 4138 4139 4140 4141
	if (dentry->d_inode) {
		root = BTRFS_I(dentry->d_inode)->root;
		if (btrfs_root_refs(&root->root_item) == 0)
			return 1;
	}
4142 4143 4144
	return 0;
}

4145 4146 4147 4148 4149 4150 4151 4152
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);
4153

C
Chris Mason 已提交
4154 4155 4156 4157 4158 4159 4160
	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
};

4161 4162
static int btrfs_real_readdir(struct file *filp, void *dirent,
			      filldir_t filldir)
C
Chris Mason 已提交
4163
{
4164
	struct inode *inode = filp->f_dentry->d_inode;
C
Chris Mason 已提交
4165 4166 4167 4168
	struct btrfs_root *root = BTRFS_I(inode)->root;
	struct btrfs_item *item;
	struct btrfs_dir_item *di;
	struct btrfs_key key;
4169
	struct btrfs_key found_key;
C
Chris Mason 已提交
4170 4171 4172
	struct btrfs_path *path;
	int ret;
	u32 nritems;
4173
	struct extent_buffer *leaf;
C
Chris Mason 已提交
4174 4175 4176 4177 4178 4179 4180 4181
	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;
4182 4183 4184
	char tmp_name[32];
	char *name_ptr;
	int name_len;
C
Chris Mason 已提交
4185 4186 4187 4188

	/* FIXME, use a real flag for deciding about the key type */
	if (root->fs_info->tree_root == root)
		key_type = BTRFS_DIR_ITEM_KEY;
4189

4190 4191 4192 4193 4194 4195 4196 4197 4198 4199 4200
	/* 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) {
4201
		u64 pino = parent_ino(filp->f_path.dentry);
4202
		over = filldir(dirent, "..", 2,
4203
			       2, pino, DT_DIR);
4204
		if (over)
4205
			return 0;
4206 4207
		filp->f_pos = 2;
	}
4208 4209 4210
	path = btrfs_alloc_path();
	path->reada = 2;

C
Chris Mason 已提交
4211 4212
	btrfs_set_key_type(&key, key_type);
	key.offset = filp->f_pos;
4213
	key.objectid = inode->i_ino;
4214

C
Chris Mason 已提交
4215 4216 4217 4218
	ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
	if (ret < 0)
		goto err;
	advance = 0;
4219 4220

	while (1) {
4221 4222
		leaf = path->nodes[0];
		nritems = btrfs_header_nritems(leaf);
C
Chris Mason 已提交
4223 4224
		slot = path->slots[0];
		if (advance || slot >= nritems) {
4225
			if (slot >= nritems - 1) {
C
Chris Mason 已提交
4226 4227 4228
				ret = btrfs_next_leaf(root, path);
				if (ret)
					break;
4229 4230
				leaf = path->nodes[0];
				nritems = btrfs_header_nritems(leaf);
C
Chris Mason 已提交
4231 4232 4233 4234 4235 4236
				slot = path->slots[0];
			} else {
				slot++;
				path->slots[0]++;
			}
		}
4237

C
Chris Mason 已提交
4238
		advance = 1;
4239 4240 4241 4242
		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
4243
			break;
4244
		if (btrfs_key_type(&found_key) != key_type)
C
Chris Mason 已提交
4245
			break;
4246
		if (found_key.offset < filp->f_pos)
C
Chris Mason 已提交
4247
			continue;
4248 4249

		filp->f_pos = found_key.offset;
4250

C
Chris Mason 已提交
4251 4252
		di = btrfs_item_ptr(leaf, slot, struct btrfs_dir_item);
		di_cur = 0;
4253
		di_total = btrfs_item_size(leaf, item);
4254 4255

		while (di_cur < di_total) {
4256 4257 4258
			struct btrfs_key location;

			name_len = btrfs_dir_name_len(leaf, di);
4259
			if (name_len <= sizeof(tmp_name)) {
4260 4261 4262
				name_ptr = tmp_name;
			} else {
				name_ptr = kmalloc(name_len, GFP_NOFS);
4263 4264 4265 4266
				if (!name_ptr) {
					ret = -ENOMEM;
					goto err;
				}
4267 4268 4269 4270 4271 4272
			}
			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);
4273 4274 4275 4276 4277 4278 4279 4280 4281

			/* 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;
			}
4282
			over = filldir(dirent, name_ptr, name_len,
4283
				       found_key.offset, location.objectid,
C
Chris Mason 已提交
4284
				       d_type);
4285

4286
skip:
4287 4288 4289
			if (name_ptr != tmp_name)
				kfree(name_ptr);

C
Chris Mason 已提交
4290 4291
			if (over)
				goto nopos;
J
Josef Bacik 已提交
4292
			di_len = btrfs_dir_name_len(leaf, di) +
4293
				 btrfs_dir_data_len(leaf, di) + sizeof(*di);
C
Chris Mason 已提交
4294 4295 4296 4297
			di_cur += di_len;
			di = (struct btrfs_dir_item *)((char *)di + di_len);
		}
	}
4298 4299

	/* Reached end of directory/root. Bump pos past the last item. */
4300
	if (key_type == BTRFS_DIR_INDEX_KEY)
4301 4302 4303 4304 4305
		/*
		 * 32-bit glibc will use getdents64, but then strtol -
		 * so the last number we can serve is this.
		 */
		filp->f_pos = 0x7fffffff;
4306 4307
	else
		filp->f_pos++;
C
Chris Mason 已提交
4308 4309 4310 4311 4312 4313 4314
nopos:
	ret = 0;
err:
	btrfs_free_path(path);
	return ret;
}

4315
int btrfs_write_inode(struct inode *inode, struct writeback_control *wbc)
C
Chris Mason 已提交
4316 4317 4318 4319
{
	struct btrfs_root *root = BTRFS_I(inode)->root;
	struct btrfs_trans_handle *trans;
	int ret = 0;
4320
	bool nolock = false;
C
Chris Mason 已提交
4321

4322
	if (BTRFS_I(inode)->dummy_inode)
4323 4324
		return 0;

4325 4326 4327
	smp_mb();
	nolock = (root->fs_info->closing && root == root->fs_info->tree_root);

4328
	if (wbc->sync_mode == WB_SYNC_ALL) {
4329 4330 4331 4332
		if (nolock)
			trans = btrfs_join_transaction_nolock(root, 1);
		else
			trans = btrfs_join_transaction(root, 1);
C
Chris Mason 已提交
4333
		btrfs_set_trans_block_group(trans, inode);
4334 4335 4336 4337
		if (nolock)
			ret = btrfs_end_transaction_nolock(trans, root);
		else
			ret = btrfs_commit_transaction(trans, root);
C
Chris Mason 已提交
4338 4339 4340 4341 4342
	}
	return ret;
}

/*
4343
 * This is somewhat expensive, updating the tree every time the
C
Chris Mason 已提交
4344 4345 4346 4347 4348 4349 4350 4351
 * 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;
4352 4353 4354 4355
	int ret;

	if (BTRFS_I(inode)->dummy_inode)
		return;
C
Chris Mason 已提交
4356

4357
	trans = btrfs_join_transaction(root, 1);
C
Chris Mason 已提交
4358
	btrfs_set_trans_block_group(trans, inode);
4359 4360

	ret = btrfs_update_inode(trans, root, inode);
4361 4362 4363 4364
	if (ret && ret == -ENOSPC) {
		/* whoops, lets try again with the full transaction */
		btrfs_end_transaction(trans, root);
		trans = btrfs_start_transaction(root, 1);
4365 4366 4367 4368 4369 4370 4371 4372
		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;
		}
4373
		btrfs_set_trans_block_group(trans, inode);
4374

4375 4376
		ret = btrfs_update_inode(trans, root, inode);
		if (ret) {
4377 4378 4379 4380 4381
			if (printk_ratelimit()) {
				printk(KERN_ERR "btrfs: fail to "
				       "dirty  inode %lu error %d\n",
				       inode->i_ino, ret);
			}
4382 4383
		}
	}
C
Chris Mason 已提交
4384 4385 4386
	btrfs_end_transaction(trans, root);
}

C
Chris Mason 已提交
4387 4388 4389 4390 4391
/*
 * find the highest existing sequence number in a directory
 * and then set the in-memory index_cnt variable to reflect
 * free sequence numbers
 */
4392 4393 4394 4395 4396 4397 4398 4399 4400 4401 4402 4403 4404 4405 4406 4407 4408 4409 4410 4411 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
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
4444 4445 4446 4447
/*
 * 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
 */
4448
int btrfs_set_inode_index(struct inode *dir, u64 *index)
4449 4450 4451 4452 4453
{
	int ret = 0;

	if (BTRFS_I(dir)->index_cnt == (u64)-1) {
		ret = btrfs_set_inode_index_count(dir);
C
Chris Mason 已提交
4454
		if (ret)
4455 4456 4457
			return ret;
	}

4458
	*index = BTRFS_I(dir)->index_cnt;
4459 4460 4461 4462 4463
	BTRFS_I(dir)->index_cnt++;

	return ret;
}

C
Chris Mason 已提交
4464 4465
static struct inode *btrfs_new_inode(struct btrfs_trans_handle *trans,
				     struct btrfs_root *root,
4466
				     struct inode *dir,
4467
				     const char *name, int name_len,
4468 4469
				     u64 ref_objectid, u64 objectid,
				     u64 alloc_hint, int mode, u64 *index)
C
Chris Mason 已提交
4470 4471
{
	struct inode *inode;
4472
	struct btrfs_inode_item *inode_item;
C
Chris Mason 已提交
4473
	struct btrfs_key *location;
4474
	struct btrfs_path *path;
4475 4476 4477 4478
	struct btrfs_inode_ref *ref;
	struct btrfs_key key[2];
	u32 sizes[2];
	unsigned long ptr;
C
Chris Mason 已提交
4479 4480 4481
	int ret;
	int owner;

4482 4483 4484
	path = btrfs_alloc_path();
	BUG_ON(!path);

C
Chris Mason 已提交
4485 4486 4487 4488
	inode = new_inode(root->fs_info->sb);
	if (!inode)
		return ERR_PTR(-ENOMEM);

4489
	if (dir) {
4490
		ret = btrfs_set_inode_index(dir, index);
4491 4492
		if (ret) {
			iput(inode);
4493
			return ERR_PTR(ret);
4494
		}
4495 4496 4497 4498 4499 4500 4501
	}
	/*
	 * 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
4502
	BTRFS_I(inode)->root = root;
4503
	BTRFS_I(inode)->generation = trans->transid;
4504
	inode->i_generation = BTRFS_I(inode)->generation;
J
Josef Bacik 已提交
4505
	btrfs_set_inode_space_info(root, inode);
4506

C
Chris Mason 已提交
4507 4508 4509 4510
	if (mode & S_IFDIR)
		owner = 0;
	else
		owner = 1;
4511 4512
	BTRFS_I(inode)->block_group =
			btrfs_find_block_group(root, 0, alloc_hint, owner);
4513 4514 4515 4516 4517 4518 4519 4520 4521 4522 4523 4524

	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);

4525
	path->leave_spinning = 1;
4526 4527
	ret = btrfs_insert_empty_items(trans, root, path, key, sizes, 2);
	if (ret != 0)
4528 4529
		goto fail;

4530
	inode_init_owner(inode, dir, mode);
C
Chris Mason 已提交
4531
	inode->i_ino = objectid;
4532
	inode_set_bytes(inode, 0);
C
Chris Mason 已提交
4533
	inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
4534 4535
	inode_item = btrfs_item_ptr(path->nodes[0], path->slots[0],
				  struct btrfs_inode_item);
4536
	fill_inode_item(trans, path->nodes[0], inode_item, inode);
4537 4538 4539 4540

	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);
4541
	btrfs_set_inode_ref_index(path->nodes[0], ref, *index);
4542 4543 4544
	ptr = (unsigned long)(ref + 1);
	write_extent_buffer(path->nodes[0], name, ptr, name_len);

4545 4546 4547
	btrfs_mark_buffer_dirty(path->nodes[0]);
	btrfs_free_path(path);

C
Chris Mason 已提交
4548 4549 4550 4551 4552
	location = &BTRFS_I(inode)->location;
	location->objectid = objectid;
	location->offset = 0;
	btrfs_set_key_type(location, BTRFS_INODE_ITEM_KEY);

4553 4554
	btrfs_inherit_iflags(inode, dir);

4555 4556 4557 4558 4559 4560 4561
	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
4562
	insert_inode_hash(inode);
4563
	inode_tree_add(inode);
C
Chris Mason 已提交
4564
	return inode;
4565
fail:
4566 4567
	if (dir)
		BTRFS_I(dir)->index_cnt--;
4568
	btrfs_free_path(path);
4569
	iput(inode);
4570
	return ERR_PTR(ret);
C
Chris Mason 已提交
4571 4572 4573 4574 4575 4576 4577
}

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
4578 4579 4580 4581 4582 4583
/*
 * 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.
 */
4584 4585 4586
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
4587
{
4588
	int ret = 0;
C
Chris Mason 已提交
4589
	struct btrfs_key key;
4590
	struct btrfs_root *root = BTRFS_I(parent_inode)->root;
4591

4592 4593 4594 4595 4596 4597 4598 4599 4600 4601 4602 4603 4604 4605 4606 4607 4608 4609
	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
4610 4611

	if (ret == 0) {
4612 4613 4614 4615 4616
		ret = btrfs_insert_dir_item(trans, root, name, name_len,
					    parent_inode->i_ino, &key,
					    btrfs_inode_type(inode), index);
		BUG_ON(ret);

4617
		btrfs_i_size_write(parent_inode, parent_inode->i_size +
4618
				   name_len * 2);
4619
		parent_inode->i_mtime = parent_inode->i_ctime = CURRENT_TIME;
4620
		ret = btrfs_update_inode(trans, root, parent_inode);
C
Chris Mason 已提交
4621 4622 4623 4624 4625
	}
	return ret;
}

static int btrfs_add_nondir(struct btrfs_trans_handle *trans,
4626 4627
			    struct inode *dir, struct dentry *dentry,
			    struct inode *inode, int backref, u64 index)
C
Chris Mason 已提交
4628
{
4629 4630 4631
	int err = btrfs_add_link(trans, dir, inode,
				 dentry->d_name.name, dentry->d_name.len,
				 backref, index);
C
Chris Mason 已提交
4632 4633 4634 4635 4636 4637 4638 4639 4640
	if (!err) {
		d_instantiate(dentry, inode);
		return 0;
	}
	if (err > 0)
		err = -EEXIST;
	return err;
}

J
Josef Bacik 已提交
4641 4642 4643 4644 4645
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;
4646
	struct inode *inode = NULL;
J
Josef Bacik 已提交
4647 4648 4649
	int err;
	int drop_inode = 0;
	u64 objectid;
4650
	unsigned long nr = 0;
4651
	u64 index = 0;
J
Josef Bacik 已提交
4652 4653 4654 4655

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

4656 4657 4658 4659
	err = btrfs_find_free_objectid(NULL, root, dir->i_ino, &objectid);
	if (err)
		return err;

J
Josef Bacik 已提交
4660 4661 4662 4663 4664
	/*
	 * 2 for inode item and ref
	 * 2 for dir items
	 * 1 for xattr if selinux is on
	 */
4665 4666 4667
	trans = btrfs_start_transaction(root, 5);
	if (IS_ERR(trans))
		return PTR_ERR(trans);
4668

J
Josef Bacik 已提交
4669 4670
	btrfs_set_trans_block_group(trans, dir);

4671
	inode = btrfs_new_inode(trans, root, dir, dentry->d_name.name,
4672
				dentry->d_name.len, dir->i_ino, objectid,
4673
				BTRFS_I(dir)->block_group, mode, &index);
J
Josef Bacik 已提交
4674 4675 4676 4677
	err = PTR_ERR(inode);
	if (IS_ERR(inode))
		goto out_unlock;

4678
	err = btrfs_init_inode_security(trans, inode, dir);
J
Josef Bacik 已提交
4679 4680 4681 4682 4683
	if (err) {
		drop_inode = 1;
		goto out_unlock;
	}

J
Josef Bacik 已提交
4684
	btrfs_set_trans_block_group(trans, inode);
4685
	err = btrfs_add_nondir(trans, dir, dentry, inode, 0, index);
J
Josef Bacik 已提交
4686 4687 4688 4689 4690
	if (err)
		drop_inode = 1;
	else {
		inode->i_op = &btrfs_special_inode_operations;
		init_special_inode(inode, inode->i_mode, rdev);
4691
		btrfs_update_inode(trans, root, inode);
J
Josef Bacik 已提交
4692 4693 4694 4695
	}
	btrfs_update_inode_block_group(trans, inode);
	btrfs_update_inode_block_group(trans, dir);
out_unlock:
4696
	nr = trans->blocks_used;
4697
	btrfs_end_transaction_throttle(trans, root);
4698
	btrfs_btree_balance_dirty(root, nr);
J
Josef Bacik 已提交
4699 4700 4701 4702 4703 4704 4705
	if (drop_inode) {
		inode_dec_link_count(inode);
		iput(inode);
	}
	return err;
}

C
Chris Mason 已提交
4706 4707 4708 4709 4710
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;
4711
	struct inode *inode = NULL;
C
Chris Mason 已提交
4712
	int drop_inode = 0;
4713
	int err;
4714
	unsigned long nr = 0;
C
Chris Mason 已提交
4715
	u64 objectid;
4716
	u64 index = 0;
C
Chris Mason 已提交
4717

4718 4719 4720
	err = btrfs_find_free_objectid(NULL, root, dir->i_ino, &objectid);
	if (err)
		return err;
J
Josef Bacik 已提交
4721 4722 4723 4724 4725
	/*
	 * 2 for inode item and ref
	 * 2 for dir items
	 * 1 for xattr if selinux is on
	 */
4726 4727 4728
	trans = btrfs_start_transaction(root, 5);
	if (IS_ERR(trans))
		return PTR_ERR(trans);
J
Josef Bacik 已提交
4729

C
Chris Mason 已提交
4730 4731
	btrfs_set_trans_block_group(trans, dir);

4732
	inode = btrfs_new_inode(trans, root, dir, dentry->d_name.name,
4733 4734
				dentry->d_name.len, dir->i_ino, objectid,
				BTRFS_I(dir)->block_group, mode, &index);
C
Chris Mason 已提交
4735 4736 4737 4738
	err = PTR_ERR(inode);
	if (IS_ERR(inode))
		goto out_unlock;

4739
	err = btrfs_init_inode_security(trans, inode, dir);
J
Josef Bacik 已提交
4740 4741 4742 4743 4744
	if (err) {
		drop_inode = 1;
		goto out_unlock;
	}

C
Chris Mason 已提交
4745
	btrfs_set_trans_block_group(trans, inode);
4746
	err = btrfs_add_nondir(trans, dir, dentry, inode, 0, index);
C
Chris Mason 已提交
4747 4748 4749 4750
	if (err)
		drop_inode = 1;
	else {
		inode->i_mapping->a_ops = &btrfs_aops;
C
Chris Mason 已提交
4751
		inode->i_mapping->backing_dev_info = &root->fs_info->bdi;
C
Chris Mason 已提交
4752 4753
		inode->i_fop = &btrfs_file_operations;
		inode->i_op = &btrfs_file_inode_operations;
4754
		BTRFS_I(inode)->io_tree.ops = &btrfs_extent_io_ops;
C
Chris Mason 已提交
4755 4756 4757 4758
	}
	btrfs_update_inode_block_group(trans, inode);
	btrfs_update_inode_block_group(trans, dir);
out_unlock:
4759
	nr = trans->blocks_used;
4760
	btrfs_end_transaction_throttle(trans, root);
C
Chris Mason 已提交
4761 4762 4763 4764
	if (drop_inode) {
		inode_dec_link_count(inode);
		iput(inode);
	}
4765
	btrfs_btree_balance_dirty(root, nr);
C
Chris Mason 已提交
4766 4767 4768 4769 4770 4771 4772 4773 4774
	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;
4775
	u64 index;
4776
	unsigned long nr = 0;
C
Chris Mason 已提交
4777 4778 4779 4780 4781 4782
	int err;
	int drop_inode = 0;

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

4783 4784 4785 4786
	/* 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
4787
	btrfs_inc_nlink(inode);
4788
	inode->i_ctime = CURRENT_TIME;
J
Josef Bacik 已提交
4789

4790
	err = btrfs_set_inode_index(dir, &index);
4791 4792 4793
	if (err)
		goto fail;

4794 4795 4796 4797 4798 4799 4800 4801 4802
	/*
	 * 1 item for inode ref
	 * 2 items for dir items
	 */
	trans = btrfs_start_transaction(root, 3);
	if (IS_ERR(trans)) {
		err = PTR_ERR(trans);
		goto fail;
	}
4803

C
Chris Mason 已提交
4804 4805
	btrfs_set_trans_block_group(trans, dir);
	atomic_inc(&inode->i_count);
4806

4807
	err = btrfs_add_nondir(trans, dir, dentry, inode, 1, index);
4808

4809
	if (err) {
4810
		drop_inode = 1;
4811
	} else {
4812
		struct dentry *parent = dget_parent(dentry);
4813 4814 4815
		btrfs_update_inode_block_group(trans, dir);
		err = btrfs_update_inode(trans, root, inode);
		BUG_ON(err);
4816 4817
		btrfs_log_new_name(trans, inode, NULL, parent);
		dput(parent);
4818
	}
C
Chris Mason 已提交
4819

4820
	nr = trans->blocks_used;
4821
	btrfs_end_transaction_throttle(trans, root);
4822
fail:
C
Chris Mason 已提交
4823 4824 4825 4826
	if (drop_inode) {
		inode_dec_link_count(inode);
		iput(inode);
	}
4827
	btrfs_btree_balance_dirty(root, nr);
C
Chris Mason 已提交
4828 4829 4830 4831 4832
	return err;
}

static int btrfs_mkdir(struct inode *dir, struct dentry *dentry, int mode)
{
4833
	struct inode *inode = NULL;
C
Chris Mason 已提交
4834 4835 4836 4837
	struct btrfs_trans_handle *trans;
	struct btrfs_root *root = BTRFS_I(dir)->root;
	int err = 0;
	int drop_on_err = 0;
4838
	u64 objectid = 0;
4839
	u64 index = 0;
4840
	unsigned long nr = 1;
C
Chris Mason 已提交
4841

4842 4843 4844 4845
	err = btrfs_find_free_objectid(NULL, root, dir->i_ino, &objectid);
	if (err)
		return err;

J
Josef Bacik 已提交
4846 4847 4848 4849 4850
	/*
	 * 2 items for inode and ref
	 * 2 items for dir items
	 * 1 for xattr if selinux is on
	 */
4851 4852 4853
	trans = btrfs_start_transaction(root, 5);
	if (IS_ERR(trans))
		return PTR_ERR(trans);
J
Josef Bacik 已提交
4854
	btrfs_set_trans_block_group(trans, dir);
C
Chris Mason 已提交
4855

4856
	inode = btrfs_new_inode(trans, root, dir, dentry->d_name.name,
4857
				dentry->d_name.len, dir->i_ino, objectid,
4858 4859
				BTRFS_I(dir)->block_group, S_IFDIR | mode,
				&index);
C
Chris Mason 已提交
4860 4861 4862 4863
	if (IS_ERR(inode)) {
		err = PTR_ERR(inode);
		goto out_fail;
	}
4864

C
Chris Mason 已提交
4865
	drop_on_err = 1;
J
Josef Bacik 已提交
4866

4867
	err = btrfs_init_inode_security(trans, inode, dir);
J
Josef Bacik 已提交
4868 4869 4870
	if (err)
		goto out_fail;

C
Chris Mason 已提交
4871 4872 4873 4874
	inode->i_op = &btrfs_dir_inode_operations;
	inode->i_fop = &btrfs_dir_file_operations;
	btrfs_set_trans_block_group(trans, inode);

4875
	btrfs_i_size_write(inode, 0);
C
Chris Mason 已提交
4876 4877 4878
	err = btrfs_update_inode(trans, root, inode);
	if (err)
		goto out_fail;
4879

4880 4881
	err = btrfs_add_link(trans, dir, inode, dentry->d_name.name,
			     dentry->d_name.len, 0, index);
C
Chris Mason 已提交
4882 4883
	if (err)
		goto out_fail;
4884

C
Chris Mason 已提交
4885 4886 4887 4888 4889 4890
	d_instantiate(dentry, inode);
	drop_on_err = 0;
	btrfs_update_inode_block_group(trans, inode);
	btrfs_update_inode_block_group(trans, dir);

out_fail:
4891
	nr = trans->blocks_used;
4892
	btrfs_end_transaction_throttle(trans, root);
C
Chris Mason 已提交
4893 4894
	if (drop_on_err)
		iput(inode);
4895
	btrfs_btree_balance_dirty(root, nr);
C
Chris Mason 已提交
4896 4897 4898
	return err;
}

C
Chris Mason 已提交
4899 4900 4901 4902
/* 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.
 */
4903 4904
static int merge_extent_mapping(struct extent_map_tree *em_tree,
				struct extent_map *existing,
4905 4906
				struct extent_map *em,
				u64 map_start, u64 map_len)
4907 4908 4909
{
	u64 start_diff;

4910 4911 4912 4913
	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
4914 4915
	if (em->block_start < EXTENT_MAP_LAST_BYTE &&
	    !test_bit(EXTENT_FLAG_COMPRESSED, &em->flags)) {
4916
		em->block_start += start_diff;
C
Chris Mason 已提交
4917 4918
		em->block_len -= start_diff;
	}
4919
	return add_extent_mapping(em_tree, em);
4920 4921
}

C
Chris Mason 已提交
4922 4923 4924 4925 4926 4927 4928 4929 4930 4931 4932 4933 4934 4935 4936 4937 4938 4939 4940 4941 4942
static noinline int uncompress_inline(struct btrfs_path *path,
				      struct inode *inode, struct page *page,
				      size_t pg_offset, u64 extent_offset,
				      struct btrfs_file_extent_item *item)
{
	int ret;
	struct extent_buffer *leaf = path->nodes[0];
	char *tmp;
	size_t max_size;
	unsigned long inline_size;
	unsigned long ptr;

	WARN_ON(pg_offset != 0);
	max_size = btrfs_file_extent_ram_bytes(leaf, item);
	inline_size = btrfs_file_extent_inline_item_len(leaf,
					btrfs_item_nr(leaf, path->slots[0]));
	tmp = kmalloc(inline_size, GFP_NOFS);
	ptr = btrfs_file_extent_inline_start(item);

	read_extent_buffer(leaf, tmp, ptr, inline_size);

4943
	max_size = min_t(unsigned long, PAGE_CACHE_SIZE, max_size);
C
Chris Mason 已提交
4944 4945 4946 4947 4948 4949 4950 4951 4952 4953 4954 4955 4956 4957
	ret = btrfs_zlib_decompress(tmp, page, extent_offset,
				    inline_size, max_size);
	if (ret) {
		char *kaddr = kmap_atomic(page, KM_USER0);
		unsigned long copy_size = min_t(u64,
				  PAGE_CACHE_SIZE - pg_offset,
				  max_size - extent_offset);
		memset(kaddr + pg_offset, 0, copy_size);
		kunmap_atomic(kaddr, KM_USER0);
	}
	kfree(tmp);
	return 0;
}

C
Chris Mason 已提交
4958 4959
/*
 * a bit scary, this does extent mapping from logical file offset to the disk.
C
Chris Mason 已提交
4960 4961
 * 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
4962 4963 4964 4965
 * 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
4966

4967
struct extent_map *btrfs_get_extent(struct inode *inode, struct page *page,
4968
				    size_t pg_offset, u64 start, u64 len,
4969 4970 4971 4972
				    int create)
{
	int ret;
	int err = 0;
4973
	u64 bytenr;
4974 4975 4976 4977
	u64 extent_start = 0;
	u64 extent_end = 0;
	u64 objectid = inode->i_ino;
	u32 found_type;
4978
	struct btrfs_path *path = NULL;
4979 4980
	struct btrfs_root *root = BTRFS_I(inode)->root;
	struct btrfs_file_extent_item *item;
4981 4982
	struct extent_buffer *leaf;
	struct btrfs_key found_key;
4983 4984
	struct extent_map *em = NULL;
	struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
4985
	struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
4986
	struct btrfs_trans_handle *trans = NULL;
C
Chris Mason 已提交
4987
	int compressed;
4988 4989

again:
4990
	read_lock(&em_tree->lock);
4991
	em = lookup_extent_mapping(em_tree, start, len);
4992 4993
	if (em)
		em->bdev = root->fs_info->fs_devices->latest_bdev;
4994
	read_unlock(&em_tree->lock);
4995

4996
	if (em) {
4997 4998 4999
		if (em->start > start || em->start + em->len <= start)
			free_extent_map(em);
		else if (em->block_start == EXTENT_MAP_INLINE && page)
5000 5001 5002
			free_extent_map(em);
		else
			goto out;
5003
	}
5004
	em = alloc_extent_map(GFP_NOFS);
5005
	if (!em) {
5006 5007
		err = -ENOMEM;
		goto out;
5008
	}
5009
	em->bdev = root->fs_info->fs_devices->latest_bdev;
5010
	em->start = EXTENT_MAP_HOLE;
5011
	em->orig_start = EXTENT_MAP_HOLE;
5012
	em->len = (u64)-1;
C
Chris Mason 已提交
5013
	em->block_len = (u64)-1;
5014 5015 5016 5017 5018 5019

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

5020 5021
	ret = btrfs_lookup_file_extent(trans, root, path,
				       objectid, start, trans != NULL);
5022 5023 5024 5025 5026 5027 5028 5029 5030 5031 5032
	if (ret < 0) {
		err = ret;
		goto out;
	}

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

5033 5034
	leaf = path->nodes[0];
	item = btrfs_item_ptr(leaf, path->slots[0],
5035 5036
			      struct btrfs_file_extent_item);
	/* are we inside the extent that was found? */
5037 5038 5039
	btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
	found_type = btrfs_key_type(&found_key);
	if (found_key.objectid != objectid ||
5040 5041 5042 5043
	    found_type != BTRFS_EXTENT_DATA_KEY) {
		goto not_found;
	}

5044 5045
	found_type = btrfs_file_extent_type(leaf, item);
	extent_start = found_key.offset;
C
Chris Mason 已提交
5046
	compressed = btrfs_file_extent_compression(leaf, item);
Y
Yan Zheng 已提交
5047 5048
	if (found_type == BTRFS_FILE_EXTENT_REG ||
	    found_type == BTRFS_FILE_EXTENT_PREALLOC) {
5049
		extent_end = extent_start +
5050
		       btrfs_file_extent_num_bytes(leaf, item);
Y
Yan Zheng 已提交
5051 5052 5053 5054 5055 5056 5057 5058 5059 5060 5061 5062 5063 5064
	} 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;
5065
			}
Y
Yan Zheng 已提交
5066 5067 5068
			if (ret > 0)
				goto not_found;
			leaf = path->nodes[0];
5069
		}
Y
Yan Zheng 已提交
5070 5071 5072 5073 5074 5075 5076 5077 5078 5079 5080
		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
5081 5082
	if (found_type == BTRFS_FILE_EXTENT_REG ||
	    found_type == BTRFS_FILE_EXTENT_PREALLOC) {
Y
Yan Zheng 已提交
5083 5084
		em->start = extent_start;
		em->len = extent_end - extent_start;
5085 5086
		em->orig_start = extent_start -
				 btrfs_file_extent_offset(leaf, item);
5087 5088
		bytenr = btrfs_file_extent_disk_bytenr(leaf, item);
		if (bytenr == 0) {
5089
			em->block_start = EXTENT_MAP_HOLE;
5090 5091
			goto insert;
		}
C
Chris Mason 已提交
5092 5093 5094 5095 5096 5097 5098 5099 5100
		if (compressed) {
			set_bit(EXTENT_FLAG_COMPRESSED, &em->flags);
			em->block_start = bytenr;
			em->block_len = btrfs_file_extent_disk_num_bytes(leaf,
									 item);
		} else {
			bytenr += btrfs_file_extent_offset(leaf, item);
			em->block_start = bytenr;
			em->block_len = em->len;
Y
Yan Zheng 已提交
5101 5102
			if (found_type == BTRFS_FILE_EXTENT_PREALLOC)
				set_bit(EXTENT_FLAG_PREALLOC, &em->flags);
C
Chris Mason 已提交
5103
		}
5104 5105
		goto insert;
	} else if (found_type == BTRFS_FILE_EXTENT_INLINE) {
5106
		unsigned long ptr;
5107
		char *map;
5108 5109 5110
		size_t size;
		size_t extent_offset;
		size_t copy_size;
5111

5112
		em->block_start = EXTENT_MAP_INLINE;
C
Chris Mason 已提交
5113
		if (!page || create) {
5114
			em->start = extent_start;
Y
Yan Zheng 已提交
5115
			em->len = extent_end - extent_start;
5116 5117
			goto out;
		}
5118

Y
Yan Zheng 已提交
5119 5120
		size = btrfs_file_extent_inline_len(leaf, item);
		extent_offset = page_offset(page) + pg_offset - extent_start;
5121
		copy_size = min_t(u64, PAGE_CACHE_SIZE - pg_offset,
5122 5123
				size - extent_offset);
		em->start = extent_start + extent_offset;
5124 5125
		em->len = (copy_size + root->sectorsize - 1) &
			~((u64)root->sectorsize - 1);
5126
		em->orig_start = EXTENT_MAP_INLINE;
C
Chris Mason 已提交
5127 5128
		if (compressed)
			set_bit(EXTENT_FLAG_COMPRESSED, &em->flags);
5129
		ptr = btrfs_file_extent_inline_start(item) + extent_offset;
5130
		if (create == 0 && !PageUptodate(page)) {
C
Chris Mason 已提交
5131 5132 5133 5134 5135 5136 5137 5138 5139 5140
			if (btrfs_file_extent_compression(leaf, item) ==
			    BTRFS_COMPRESS_ZLIB) {
				ret = uncompress_inline(path, inode, page,
							pg_offset,
							extent_offset, item);
				BUG_ON(ret);
			} else {
				map = kmap(page);
				read_extent_buffer(leaf, map + pg_offset, ptr,
						   copy_size);
5141 5142 5143 5144 5145
				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
5146 5147
				kunmap(page);
			}
5148 5149
			flush_dcache_page(page);
		} else if (create && PageUptodate(page)) {
5150
			WARN_ON(1);
5151 5152 5153 5154 5155
			if (!trans) {
				kunmap(page);
				free_extent_map(em);
				em = NULL;
				btrfs_release_path(root, path);
5156
				trans = btrfs_join_transaction(root, 1);
5157 5158
				goto again;
			}
C
Chris Mason 已提交
5159
			map = kmap(page);
5160
			write_extent_buffer(leaf, map + pg_offset, ptr,
5161
					    copy_size);
C
Chris Mason 已提交
5162
			kunmap(page);
5163
			btrfs_mark_buffer_dirty(leaf);
5164
		}
5165 5166
		set_extent_uptodate(io_tree, em->start,
				    extent_map_end(em) - 1, GFP_NOFS);
5167 5168
		goto insert;
	} else {
C
Chris Mason 已提交
5169
		printk(KERN_ERR "btrfs unknown found_type %d\n", found_type);
5170 5171 5172 5173
		WARN_ON(1);
	}
not_found:
	em->start = start;
5174
	em->len = len;
5175
not_found_em:
5176
	em->block_start = EXTENT_MAP_HOLE;
Y
Yan Zheng 已提交
5177
	set_bit(EXTENT_FLAG_VACANCY, &em->flags);
5178 5179
insert:
	btrfs_release_path(root, path);
5180
	if (em->start > start || extent_map_end(em) <= start) {
C
Chris Mason 已提交
5181 5182 5183 5184 5185
		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);
5186 5187 5188
		err = -EIO;
		goto out;
	}
5189 5190

	err = 0;
5191
	write_lock(&em_tree->lock);
5192
	ret = add_extent_mapping(em_tree, em);
5193 5194 5195 5196
	/* 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
	 */
5197
	if (ret == -EEXIST) {
5198
		struct extent_map *existing;
5199 5200 5201

		ret = 0;

5202
		existing = lookup_extent_mapping(em_tree, start, len);
5203 5204 5205 5206 5207
		if (existing && (existing->start > start ||
		    existing->start + existing->len <= start)) {
			free_extent_map(existing);
			existing = NULL;
		}
5208 5209 5210 5211 5212
		if (!existing) {
			existing = lookup_extent_mapping(em_tree, em->start,
							 em->len);
			if (existing) {
				err = merge_extent_mapping(em_tree, existing,
5213 5214
							   em, start,
							   root->sectorsize);
5215 5216 5217 5218 5219 5220 5221 5222 5223 5224 5225 5226 5227
				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;
5228
			err = 0;
5229 5230
		}
	}
5231
	write_unlock(&em_tree->lock);
5232
out:
5233 5234
	if (path)
		btrfs_free_path(path);
5235 5236
	if (trans) {
		ret = btrfs_end_transaction(trans, root);
C
Chris Mason 已提交
5237
		if (!err)
5238 5239 5240 5241 5242 5243 5244 5245 5246
			err = ret;
	}
	if (err) {
		free_extent_map(em);
		return ERR_PTR(err);
	}
	return em;
}

5247 5248 5249 5250 5251 5252 5253 5254 5255 5256 5257 5258 5259 5260 5261 5262 5263 5264 5265 5266 5267 5268 5269 5270 5271 5272 5273 5274 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
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);
	if (!trans)
		return ERR_PTR(-ENOMEM);

	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;
}

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 5397 5398 5399 5400 5401 5402 5403 5404 5405 5406 5407 5408
/*
 * 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;
}

5409 5410 5411 5412 5413 5414 5415
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;
5416
	struct btrfs_trans_handle *trans;
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 5459 5460

	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.
	 *
	 */
5461 5462
	if (!create) {
		len = em->len - (start - em->start);
5463
		goto map;
5464
	}
5465 5466 5467 5468 5469 5470

	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;
5471
		u64 block_start;
5472 5473 5474 5475 5476

		if (test_bit(EXTENT_FLAG_PREALLOC, &em->flags))
			type = BTRFS_ORDERED_PREALLOC;
		else
			type = BTRFS_ORDERED_NOCOW;
5477
		len = min(len, em->len - (start - em->start));
5478
		block_start = em->block_start + (start - em->start);
5479 5480 5481 5482 5483 5484 5485 5486 5487 5488 5489 5490 5491 5492 5493 5494 5495 5496 5497

		/*
		 * 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);
		if (!trans)
			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;
5498
		}
5499
		btrfs_end_transaction(trans, root);
5500
	}
5501 5502 5503 5504 5505 5506 5507 5508 5509 5510 5511 5512
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:
5513 5514 5515
	clear_extent_bit(&BTRFS_I(inode)->io_tree, start, start + len - 1,
			  EXTENT_LOCKED | EXTENT_DELALLOC | EXTENT_DIRTY, 1,
			  0, NULL, GFP_NOFS);
5516 5517 5518
map:
	bh_result->b_blocknr = (em->block_start + (start - em->start)) >>
		inode->i_blkbits;
5519
	bh_result->b_size = len;
5520 5521 5522 5523 5524 5525 5526 5527 5528 5529 5530 5531 5532 5533 5534 5535 5536
	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
5537 5538 5539 5540 5541 5542 5543 5544

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

	/* IO errors */
	int errors;

	struct bio *orig_bio;
5545 5546 5547 5548
};

static void btrfs_endio_direct_read(struct bio *bio, int err)
{
M
Miao Xie 已提交
5549
	struct btrfs_dio_private *dip = bio->bi_private;
5550 5551 5552 5553 5554 5555 5556 5557 5558 5559 5560 5561 5562 5563 5564 5565 5566 5567 5568 5569 5570 5571 5572 5573 5574 5575 5576 5577 5578 5579 5580 5581 5582 5583 5584 5585 5586 5587 5588 5589 5590 5591 5592 5593 5594 5595 5596 5597 5598 5599 5600 5601 5602 5603 5604
	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;
5605 5606
	u64 ordered_offset = dip->logical_offset;
	u64 ordered_bytes = dip->bytes;
5607 5608 5609 5610
	int ret;

	if (err)
		goto out_done;
5611 5612 5613 5614
again:
	ret = btrfs_dec_test_first_ordered_pending(inode, &ordered,
						   &ordered_offset,
						   ordered_bytes);
5615
	if (!ret)
5616
		goto out_test;
5617 5618 5619 5620 5621 5622 5623 5624 5625 5626 5627 5628 5629 5630 5631 5632 5633 5634 5635 5636 5637 5638 5639 5640 5641 5642 5643 5644 5645 5646 5647 5648 5649 5650 5651 5652 5653 5654 5655 5656 5657 5658 5659 5660 5661 5662 5663 5664 5665 5666 5667 5668 5669 5670 5671 5672 5673 5674 5675

	BUG_ON(!ordered);

	trans = btrfs_join_transaction(root, 1);
	if (!trans) {
		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);
5676
	ordered_offset = ordered->file_offset + ordered->len;
5677 5678
	btrfs_put_ordered_extent(ordered);
	btrfs_put_ordered_extent(ordered);
5679 5680 5681 5682 5683 5684 5685 5686 5687 5688 5689

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;
	}
5690 5691 5692 5693 5694 5695 5696 5697
out_done:
	bio->bi_private = dip->private;

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

5698 5699 5700 5701 5702 5703 5704 5705 5706 5707 5708
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
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 5755 5756 5757 5758 5759 5760 5761 5762 5763 5764 5765 5766 5767 5768 5769 5770 5771 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 5826 5827 5828 5829 5830 5831 5832 5833 5834 5835 5836 5837 5838 5839 5840 5841 5842 5843 5844 5845 5846 5847 5848 5849 5850 5851 5852 5853 5854 5855 5856 5857 5858 5859 5860 5861 5862 5863 5864 5865 5866 5867 5868 5869 5870 5871 5872 5873 5874 5875 5876 5877 5878
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 "
		      "disk_bytenr %lu len %u err no %d\n",
		      dip->inode->i_ino, bio->bi_rw, bio->bi_sector,
		      bio->bi_size, 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 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;
}

5879 5880 5881 5882 5883 5884 5885
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;
5886
	int write = rw & REQ_WRITE;
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
	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) {
			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));

5916
	dip->disk_bytenr = (u64)bio->bi_sector << 9;
5917
	bio->bi_private = dip;
M
Miao Xie 已提交
5918 5919 5920
	dip->errors = 0;
	dip->orig_bio = bio;
	atomic_set(&dip->pending_bios, 0);
5921 5922 5923 5924 5925 5926

	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
5927 5928
	ret = btrfs_submit_direct_hook(rw, dip, skip_sum);
	if (!ret)
5929
		return;
5930 5931 5932 5933 5934 5935 5936
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;
5937
		ordered = btrfs_lookup_ordered_extent(inode, file_offset);
5938 5939 5940 5941 5942 5943 5944 5945 5946 5947
		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
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
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;
}
5974 5975 5976 5977
static ssize_t btrfs_direct_IO(int rw, struct kiocb *iocb,
			const struct iovec *iov, loff_t offset,
			unsigned long nr_segs)
{
5978 5979 5980
	struct file *file = iocb->ki_filp;
	struct inode *inode = file->f_mapping->host;
	struct btrfs_ordered_extent *ordered;
5981
	struct extent_state *cached_state = NULL;
5982 5983
	u64 lockstart, lockend;
	ssize_t ret;
5984 5985
	int writing = rw & WRITE;
	int write_bits = 0;
5986
	size_t count = iov_length(iov, nr_segs);
5987

C
Chris Mason 已提交
5988 5989 5990 5991 5992
	if (check_direct_IO(BTRFS_I(inode)->root, rw, iocb, iov,
			    offset, nr_segs)) {
		return 0;
	}

5993
	lockstart = offset;
5994 5995 5996 5997 5998 5999 6000
	lockend = offset + count - 1;

	if (writing) {
		ret = btrfs_delalloc_reserve_space(inode, count);
		if (ret)
			goto out;
	}
6001

6002
	while (1) {
6003 6004
		lock_extent_bits(&BTRFS_I(inode)->io_tree, lockstart, lockend,
				 0, &cached_state, GFP_NOFS);
6005 6006 6007 6008 6009 6010 6011 6012 6013
		/*
		 * 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;
6014 6015
		unlock_extent_cached(&BTRFS_I(inode)->io_tree, lockstart, lockend,
				     &cached_state, GFP_NOFS);
6016 6017 6018 6019 6020
		btrfs_start_ordered_extent(inode, ordered, 1);
		btrfs_put_ordered_extent(ordered);
		cond_resched();
	}

6021 6022 6023 6024 6025 6026 6027 6028 6029 6030 6031 6032 6033 6034 6035 6036 6037 6038 6039 6040
	/*
	 * 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
6041 6042 6043 6044
	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);
6045 6046

	if (ret < 0 && ret != -EIOCBQUEUED) {
6047 6048 6049 6050
		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);
6051 6052 6053 6054 6055
	} 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.
		 */
6056 6057 6058 6059
		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);
6060
	}
6061 6062
out:
	free_extent_state(cached_state);
6063
	return ret;
6064 6065
}

Y
Yehuda Sadeh 已提交
6066 6067 6068 6069 6070 6071
static int btrfs_fiemap(struct inode *inode, struct fiemap_extent_info *fieinfo,
		__u64 start, __u64 len)
{
	return extent_fiemap(inode, fieinfo, start, len, btrfs_get_extent);
}

6072
int btrfs_readpage(struct file *file, struct page *page)
C
Chris Mason 已提交
6073
{
6074 6075
	struct extent_io_tree *tree;
	tree = &BTRFS_I(page->mapping->host)->io_tree;
6076
	return extent_read_full_page(tree, page, btrfs_get_extent);
C
Chris Mason 已提交
6077
}
6078

6079
static int btrfs_writepage(struct page *page, struct writeback_control *wbc)
C
Chris Mason 已提交
6080
{
6081
	struct extent_io_tree *tree;
6082 6083 6084 6085 6086 6087 6088


	if (current->flags & PF_MEMALLOC) {
		redirty_page_for_writepage(wbc, page);
		unlock_page(page);
		return 0;
	}
6089
	tree = &BTRFS_I(page->mapping->host)->io_tree;
6090
	return extent_write_full_page(tree, page, btrfs_get_extent, wbc);
C
Chris Mason 已提交
6091 6092
}

6093 6094
int btrfs_writepages(struct address_space *mapping,
		     struct writeback_control *wbc)
C
Chris Mason 已提交
6095
{
6096
	struct extent_io_tree *tree;
6097

6098
	tree = &BTRFS_I(mapping->host)->io_tree;
C
Chris Mason 已提交
6099 6100 6101
	return extent_writepages(tree, mapping, btrfs_get_extent, wbc);
}

C
Chris Mason 已提交
6102 6103 6104 6105
static int
btrfs_readpages(struct file *file, struct address_space *mapping,
		struct list_head *pages, unsigned nr_pages)
{
6106 6107
	struct extent_io_tree *tree;
	tree = &BTRFS_I(mapping->host)->io_tree;
C
Chris Mason 已提交
6108 6109 6110
	return extent_readpages(tree, mapping, pages, nr_pages,
				btrfs_get_extent);
}
6111
static int __btrfs_releasepage(struct page *page, gfp_t gfp_flags)
C
Chris Mason 已提交
6112
{
6113 6114
	struct extent_io_tree *tree;
	struct extent_map_tree *map;
6115
	int ret;
6116

6117 6118
	tree = &BTRFS_I(page->mapping->host)->io_tree;
	map = &BTRFS_I(page->mapping->host)->extent_tree;
6119
	ret = try_release_extent_mapping(map, tree, page, gfp_flags);
6120 6121 6122 6123
	if (ret == 1) {
		ClearPagePrivate(page);
		set_page_private(page, 0);
		page_cache_release(page);
C
Chris Mason 已提交
6124
	}
6125
	return ret;
C
Chris Mason 已提交
6126 6127
}

6128 6129
static int btrfs_releasepage(struct page *page, gfp_t gfp_flags)
{
6130 6131
	if (PageWriteback(page) || PageDirty(page))
		return 0;
6132
	return __btrfs_releasepage(page, gfp_flags & GFP_NOFS);
6133 6134
}

6135
static void btrfs_invalidatepage(struct page *page, unsigned long offset)
C
Chris Mason 已提交
6136
{
6137
	struct extent_io_tree *tree;
6138
	struct btrfs_ordered_extent *ordered;
6139
	struct extent_state *cached_state = NULL;
6140 6141
	u64 page_start = page_offset(page);
	u64 page_end = page_start + PAGE_CACHE_SIZE - 1;
C
Chris Mason 已提交
6142

6143 6144 6145 6146 6147 6148 6149 6150

	/*
	 * 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
	 */
6151
	wait_on_page_writeback(page);
6152

6153
	tree = &BTRFS_I(page->mapping->host)->io_tree;
6154 6155 6156 6157
	if (offset) {
		btrfs_releasepage(page, GFP_NOFS);
		return;
	}
6158 6159
	lock_extent_bits(tree, page_start, page_end, 0, &cached_state,
			 GFP_NOFS);
6160 6161 6162
	ordered = btrfs_lookup_ordered_extent(page->mapping->host,
					   page_offset(page));
	if (ordered) {
6163 6164 6165 6166
		/*
		 * IO on this page will never be started, so we need
		 * to account for any ordered extents now
		 */
6167 6168
		clear_extent_bit(tree, page_start, page_end,
				 EXTENT_DIRTY | EXTENT_DELALLOC |
6169
				 EXTENT_LOCKED | EXTENT_DO_ACCOUNTING, 1, 0,
6170
				 &cached_state, GFP_NOFS);
6171 6172 6173 6174 6175 6176 6177 6178
		/*
		 * 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);
		}
6179
		btrfs_put_ordered_extent(ordered);
6180 6181 6182
		cached_state = NULL;
		lock_extent_bits(tree, page_start, page_end, 0, &cached_state,
				 GFP_NOFS);
6183 6184
	}
	clear_extent_bit(tree, page_start, page_end,
6185
		 EXTENT_LOCKED | EXTENT_DIRTY | EXTENT_DELALLOC |
6186
		 EXTENT_DO_ACCOUNTING, 1, 1, &cached_state, GFP_NOFS);
6187 6188
	__btrfs_releasepage(page, GFP_NOFS);

C
Chris Mason 已提交
6189
	ClearPageChecked(page);
6190 6191 6192 6193 6194
	if (PagePrivate(page)) {
		ClearPagePrivate(page);
		set_page_private(page, 0);
		page_cache_release(page);
	}
C
Chris Mason 已提交
6195 6196
}

C
Chris Mason 已提交
6197 6198 6199 6200 6201 6202 6203 6204 6205 6206 6207 6208 6209 6210 6211
/*
 * 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.
 */
6212
int btrfs_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf)
C
Chris Mason 已提交
6213
{
6214
	struct page *page = vmf->page;
6215
	struct inode *inode = fdentry(vma->vm_file)->d_inode;
6216
	struct btrfs_root *root = BTRFS_I(inode)->root;
6217 6218
	struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
	struct btrfs_ordered_extent *ordered;
6219
	struct extent_state *cached_state = NULL;
6220 6221
	char *kaddr;
	unsigned long zero_start;
C
Chris Mason 已提交
6222
	loff_t size;
6223
	int ret;
6224
	u64 page_start;
6225
	u64 page_end;
C
Chris Mason 已提交
6226

6227
	ret  = btrfs_delalloc_reserve_space(inode, PAGE_CACHE_SIZE);
6228 6229 6230 6231 6232
	if (ret) {
		if (ret == -ENOMEM)
			ret = VM_FAULT_OOM;
		else /* -ENOSPC, -EIO, etc */
			ret = VM_FAULT_SIGBUS;
6233
		goto out;
6234
	}
6235

6236
	ret = VM_FAULT_NOPAGE; /* make the VM retry the fault */
6237
again:
C
Chris Mason 已提交
6238 6239
	lock_page(page);
	size = i_size_read(inode);
6240 6241
	page_start = page_offset(page);
	page_end = page_start + PAGE_CACHE_SIZE - 1;
6242

C
Chris Mason 已提交
6243
	if ((page->mapping != inode->i_mapping) ||
6244
	    (page_start >= size)) {
C
Chris Mason 已提交
6245 6246 6247
		/* page got truncated out from underneath us */
		goto out_unlock;
	}
6248 6249
	wait_on_page_writeback(page);

6250 6251
	lock_extent_bits(io_tree, page_start, page_end, 0, &cached_state,
			 GFP_NOFS);
6252 6253
	set_page_extent_mapped(page);

6254 6255 6256 6257
	/*
	 * we can't set the delalloc bits if there are pending ordered
	 * extents.  Drop our locks and wait for them to finish
	 */
6258 6259
	ordered = btrfs_lookup_ordered_extent(inode, page_start);
	if (ordered) {
6260 6261
		unlock_extent_cached(io_tree, page_start, page_end,
				     &cached_state, GFP_NOFS);
6262
		unlock_page(page);
6263
		btrfs_start_ordered_extent(inode, ordered, 1);
6264 6265 6266 6267
		btrfs_put_ordered_extent(ordered);
		goto again;
	}

J
Josef Bacik 已提交
6268 6269 6270 6271 6272 6273 6274
	/*
	 * 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.
	 */
6275
	clear_extent_bit(&BTRFS_I(inode)->io_tree, page_start, page_end,
6276
			  EXTENT_DIRTY | EXTENT_DELALLOC | EXTENT_DO_ACCOUNTING,
6277
			  0, 0, &cached_state, GFP_NOFS);
J
Josef Bacik 已提交
6278

6279 6280
	ret = btrfs_set_extent_delalloc(inode, page_start, page_end,
					&cached_state);
J
Josef Bacik 已提交
6281
	if (ret) {
6282 6283
		unlock_extent_cached(io_tree, page_start, page_end,
				     &cached_state, GFP_NOFS);
J
Josef Bacik 已提交
6284 6285 6286
		ret = VM_FAULT_SIGBUS;
		goto out_unlock;
	}
6287
	ret = 0;
C
Chris Mason 已提交
6288 6289

	/* page is wholly or partially inside EOF */
6290
	if (page_start + PAGE_CACHE_SIZE > size)
6291
		zero_start = size & ~PAGE_CACHE_MASK;
C
Chris Mason 已提交
6292
	else
6293
		zero_start = PAGE_CACHE_SIZE;
C
Chris Mason 已提交
6294

6295 6296 6297 6298 6299 6300
	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);
	}
6301
	ClearPageChecked(page);
6302
	set_page_dirty(page);
6303
	SetPageUptodate(page);
6304

6305 6306 6307
	BTRFS_I(inode)->last_trans = root->fs_info->generation;
	BTRFS_I(inode)->last_sub_trans = BTRFS_I(inode)->root->log_transid;

6308
	unlock_extent_cached(io_tree, page_start, page_end, &cached_state, GFP_NOFS);
C
Chris Mason 已提交
6309 6310

out_unlock:
6311 6312
	if (!ret)
		return VM_FAULT_LOCKED;
C
Chris Mason 已提交
6313
	unlock_page(page);
6314
	btrfs_delalloc_release_space(inode, PAGE_CACHE_SIZE);
6315
out:
C
Chris Mason 已提交
6316 6317 6318
	return ret;
}

C
Chris Mason 已提交
6319 6320 6321 6322 6323
static void btrfs_truncate(struct inode *inode)
{
	struct btrfs_root *root = BTRFS_I(inode)->root;
	int ret;
	struct btrfs_trans_handle *trans;
6324
	unsigned long nr;
6325
	u64 mask = root->sectorsize - 1;
C
Chris Mason 已提交
6326

6327 6328
	if (!S_ISREG(inode->i_mode)) {
		WARN_ON(1);
C
Chris Mason 已提交
6329
		return;
6330
	}
C
Chris Mason 已提交
6331

6332 6333 6334
	ret = btrfs_truncate_page(inode->i_mapping, inode->i_size);
	if (ret)
		return;
6335

C
Chris Mason 已提交
6336
	btrfs_wait_ordered_range(inode, inode->i_size & (~mask), (u64)-1);
6337
	btrfs_ordered_update_i_size(inode, inode->i_size, NULL);
C
Chris Mason 已提交
6338

6339 6340
	trans = btrfs_start_transaction(root, 0);
	BUG_ON(IS_ERR(trans));
6341
	btrfs_set_trans_block_group(trans, inode);
6342
	trans->block_rsv = root->orphan_block_rsv;
6343 6344 6345 6346 6347 6348 6349 6350 6351 6352 6353 6354 6355 6356 6357 6358 6359 6360 6361 6362 6363

	/*
	 * 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);

6364
	while (1) {
6365 6366 6367 6368 6369 6370 6371 6372 6373 6374 6375 6376 6377 6378 6379 6380 6381
		if (!trans) {
			trans = btrfs_start_transaction(root, 0);
			BUG_ON(IS_ERR(trans));
			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);
		if (ret) {
			BUG_ON(ret != -EAGAIN);
			ret = btrfs_commit_transaction(trans, root);
			BUG_ON(ret);
			trans = NULL;
			continue;
		}

6382 6383 6384 6385 6386
		ret = btrfs_truncate_inode_items(trans, root, inode,
						 inode->i_size,
						 BTRFS_EXTENT_DATA_KEY);
		if (ret != -EAGAIN)
			break;
C
Chris Mason 已提交
6387

6388 6389
		ret = btrfs_update_inode(trans, root, inode);
		BUG_ON(ret);
6390

6391 6392
		nr = trans->blocks_used;
		btrfs_end_transaction(trans, root);
6393
		trans = NULL;
6394 6395 6396 6397 6398 6399 6400 6401 6402
		btrfs_btree_balance_dirty(root, nr);
	}

	if (ret == 0 && inode->i_nlink > 0) {
		ret = btrfs_orphan_del(trans, inode);
		BUG_ON(ret);
	}

	ret = btrfs_update_inode(trans, root, inode);
6403 6404 6405
	BUG_ON(ret);

	nr = trans->blocks_used;
6406
	ret = btrfs_end_transaction_throttle(trans, root);
C
Chris Mason 已提交
6407
	BUG_ON(ret);
6408
	btrfs_btree_balance_dirty(root, nr);
C
Chris Mason 已提交
6409 6410
}

C
Chris Mason 已提交
6411 6412 6413
/*
 * create a new subvolume directory/inode (helper for the ioctl).
 */
6414
int btrfs_create_subvol_root(struct btrfs_trans_handle *trans,
6415
			     struct btrfs_root *new_root,
6416
			     u64 new_dirid, u64 alloc_hint)
C
Chris Mason 已提交
6417 6418
{
	struct inode *inode;
6419
	int err;
6420
	u64 index = 0;
C
Chris Mason 已提交
6421

6422
	inode = btrfs_new_inode(trans, new_root, NULL, "..", 2, new_dirid,
6423
				new_dirid, alloc_hint, S_IFDIR | 0700, &index);
6424
	if (IS_ERR(inode))
C
Christoph Hellwig 已提交
6425
		return PTR_ERR(inode);
C
Chris Mason 已提交
6426 6427 6428 6429
	inode->i_op = &btrfs_dir_inode_operations;
	inode->i_fop = &btrfs_dir_file_operations;

	inode->i_nlink = 1;
6430
	btrfs_i_size_write(inode, 0);
6431

6432 6433
	err = btrfs_update_inode(trans, new_root, inode);
	BUG_ON(err);
6434

6435
	iput(inode);
6436
	return 0;
C
Chris Mason 已提交
6437 6438
}

C
Chris Mason 已提交
6439 6440 6441
/* helper function for file defrag and space balancing.  This
 * forces readahead on a given range of bytes in an inode
 */
6442
unsigned long btrfs_force_ra(struct address_space *mapping,
6443 6444 6445
			      struct file_ra_state *ra, struct file *file,
			      pgoff_t offset, pgoff_t last_index)
{
6446
	pgoff_t req_size = last_index - offset + 1;
6447 6448 6449 6450 6451

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

C
Chris Mason 已提交
6452 6453 6454
struct inode *btrfs_alloc_inode(struct super_block *sb)
{
	struct btrfs_inode *ei;
Y
Yan, Zheng 已提交
6455
	struct inode *inode;
C
Chris Mason 已提交
6456 6457 6458 6459

	ei = kmem_cache_alloc(btrfs_inode_cachep, GFP_NOFS);
	if (!ei)
		return NULL;
Y
Yan, Zheng 已提交
6460 6461 6462 6463 6464

	ei->root = NULL;
	ei->space_info = NULL;
	ei->generation = 0;
	ei->sequence = 0;
6465
	ei->last_trans = 0;
6466
	ei->last_sub_trans = 0;
6467
	ei->logged_trans = 0;
Y
Yan, Zheng 已提交
6468 6469 6470 6471 6472 6473 6474
	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;

6475
	spin_lock_init(&ei->accounting_lock);
6476
	atomic_set(&ei->outstanding_extents, 0);
6477
	ei->reserved_extents = 0;
Y
Yan, Zheng 已提交
6478 6479

	ei->ordered_data_close = 0;
6480
	ei->orphan_meta_reserved = 0;
Y
Yan, Zheng 已提交
6481 6482 6483 6484 6485 6486 6487 6488
	ei->dummy_inode = 0;
	ei->force_compress = 0;

	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);
6489
	btrfs_ordered_inode_tree_init(&ei->ordered_tree);
6490
	INIT_LIST_HEAD(&ei->i_orphan);
Y
Yan, Zheng 已提交
6491
	INIT_LIST_HEAD(&ei->delalloc_inodes);
6492
	INIT_LIST_HEAD(&ei->ordered_operations);
Y
Yan, Zheng 已提交
6493 6494 6495
	RB_CLEAR_NODE(&ei->rb_node);

	return inode;
C
Chris Mason 已提交
6496 6497 6498 6499
}

void btrfs_destroy_inode(struct inode *inode)
{
6500
	struct btrfs_ordered_extent *ordered;
6501 6502
	struct btrfs_root *root = BTRFS_I(inode)->root;

C
Chris Mason 已提交
6503 6504
	WARN_ON(!list_empty(&inode->i_dentry));
	WARN_ON(inode->i_data.nrpages);
6505 6506
	WARN_ON(atomic_read(&BTRFS_I(inode)->outstanding_extents));
	WARN_ON(BTRFS_I(inode)->reserved_extents);
C
Chris Mason 已提交
6507

6508 6509 6510 6511 6512 6513 6514 6515
	/*
	 * 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;

6516 6517 6518 6519 6520 6521 6522 6523 6524 6525 6526
	/*
	 * 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);
	}

6527 6528 6529 6530 6531 6532 6533 6534 6535 6536 6537 6538 6539 6540 6541
	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);
		}
	}

6542
	spin_lock(&root->orphan_lock);
6543
	if (!list_empty(&BTRFS_I(inode)->i_orphan)) {
6544 6545 6546
		printk(KERN_INFO "BTRFS: inode %lu still on the orphan list\n",
		       inode->i_ino);
		list_del_init(&BTRFS_I(inode)->i_orphan);
6547
	}
6548
	spin_unlock(&root->orphan_lock);
6549

C
Chris Mason 已提交
6550
	while (1) {
6551 6552 6553 6554
		ordered = btrfs_lookup_first_ordered_extent(inode, (u64)-1);
		if (!ordered)
			break;
		else {
C
Chris Mason 已提交
6555 6556 6557 6558
			printk(KERN_ERR "btrfs found ordered "
			       "extent %llu %llu on inode cleanup\n",
			       (unsigned long long)ordered->file_offset,
			       (unsigned long long)ordered->len);
6559 6560 6561 6562 6563
			btrfs_remove_ordered_extent(inode, ordered);
			btrfs_put_ordered_extent(ordered);
			btrfs_put_ordered_extent(ordered);
		}
	}
6564
	inode_tree_del(inode);
6565
	btrfs_drop_extent_cache(inode, 0, (u64)-1, 0);
6566
free:
C
Chris Mason 已提交
6567 6568 6569
	kmem_cache_free(btrfs_inode_cachep, BTRFS_I(inode));
}

6570
int btrfs_drop_inode(struct inode *inode)
6571 6572
{
	struct btrfs_root *root = BTRFS_I(inode)->root;
6573

6574 6575
	if (btrfs_root_refs(&root->root_item) == 0 &&
	    root != root->fs_info->tree_root)
6576
		return 1;
6577
	else
6578
		return generic_drop_inode(inode);
6579 6580
}

6581
static void init_once(void *foo)
C
Chris Mason 已提交
6582 6583 6584 6585 6586 6587 6588 6589 6590 6591 6592 6593 6594 6595 6596 6597 6598 6599 6600 6601
{
	struct btrfs_inode *ei = (struct btrfs_inode *) foo;

	inode_init_once(&ei->vfs_inode);
}

void btrfs_destroy_cachep(void)
{
	if (btrfs_inode_cachep)
		kmem_cache_destroy(btrfs_inode_cachep);
	if (btrfs_trans_handle_cachep)
		kmem_cache_destroy(btrfs_trans_handle_cachep);
	if (btrfs_transaction_cachep)
		kmem_cache_destroy(btrfs_transaction_cachep);
	if (btrfs_path_cachep)
		kmem_cache_destroy(btrfs_path_cachep);
}

int btrfs_init_cachep(void)
{
6602 6603 6604
	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
6605 6606
	if (!btrfs_inode_cachep)
		goto fail;
6607 6608 6609 6610

	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
6611 6612
	if (!btrfs_trans_handle_cachep)
		goto fail;
6613 6614 6615 6616

	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
6617 6618
	if (!btrfs_transaction_cachep)
		goto fail;
6619 6620 6621 6622

	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
6623 6624
	if (!btrfs_path_cachep)
		goto fail;
6625

C
Chris Mason 已提交
6626 6627 6628 6629 6630 6631 6632 6633 6634 6635 6636
	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);
6637
	stat->dev = BTRFS_I(inode)->root->anon_super.s_dev;
C
Chris Mason 已提交
6638
	stat->blksize = PAGE_CACHE_SIZE;
6639 6640
	stat->blocks = (inode_get_bytes(inode) +
			BTRFS_I(inode)->delalloc_bytes) >> 9;
C
Chris Mason 已提交
6641 6642 6643
	return 0;
}

C
Chris Mason 已提交
6644 6645
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
6646 6647 6648
{
	struct btrfs_trans_handle *trans;
	struct btrfs_root *root = BTRFS_I(old_dir)->root;
6649
	struct btrfs_root *dest = BTRFS_I(new_dir)->root;
C
Chris Mason 已提交
6650 6651 6652
	struct inode *new_inode = new_dentry->d_inode;
	struct inode *old_inode = old_dentry->d_inode;
	struct timespec ctime = CURRENT_TIME;
6653
	u64 index = 0;
6654
	u64 root_objectid;
C
Chris Mason 已提交
6655 6656
	int ret;

6657 6658 6659
	if (new_dir->i_ino == BTRFS_EMPTY_SUBVOL_DIR_OBJECTID)
		return -EPERM;

6660 6661
	/* we only allow rename subvolume link between subvolumes */
	if (old_inode->i_ino != BTRFS_FIRST_FREE_OBJECTID && root != dest)
6662 6663
		return -EXDEV;

6664 6665
	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
6666
		return -ENOTEMPTY;
6667

6668 6669 6670
	if (S_ISDIR(old_inode->i_mode) && new_inode &&
	    new_inode->i_size > BTRFS_EMPTY_DIR_SIZE)
		return -ENOTEMPTY;
6671 6672 6673 6674 6675
	/*
	 * 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
	 */
6676
	if (new_inode && S_ISREG(old_inode->i_mode) && new_inode->i_size &&
6677 6678 6679
	    old_inode->i_size > BTRFS_ORDERED_OPERATIONS_FLUSH_LIMIT)
		filemap_flush(old_inode->i_mapping);

6680 6681 6682
	/* 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);
6683 6684 6685 6686 6687 6688 6689 6690 6691 6692 6693
	/*
	 * 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);
6694

6695
	btrfs_set_trans_block_group(trans, new_dir);
6696

6697 6698
	if (dest != root)
		btrfs_record_root_in_trans(trans, dest);
6699

6700 6701 6702
	ret = btrfs_set_inode_index(new_dir, &index);
	if (ret)
		goto out_fail;
6703

6704
	if (unlikely(old_inode->i_ino == BTRFS_FIRST_FREE_OBJECTID)) {
6705 6706 6707
		/* force full log commit if subvolume involved. */
		root->fs_info->last_trans_log_full_commit = trans->transid;
	} else {
6708 6709 6710 6711 6712 6713 6714
		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;
6715 6716 6717 6718 6719 6720 6721 6722 6723
		/*
		 * 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);
	}
6724 6725 6726 6727 6728 6729 6730 6731 6732
	/*
	 * 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
6733 6734 6735
	old_dir->i_ctime = old_dir->i_mtime = ctime;
	new_dir->i_ctime = new_dir->i_mtime = ctime;
	old_inode->i_ctime = ctime;
6736

6737 6738 6739
	if (old_dentry->d_parent != new_dentry->d_parent)
		btrfs_record_unlink_dir(trans, old_dir, old_inode, 1);

6740 6741 6742 6743 6744 6745 6746 6747 6748 6749 6750 6751 6752
	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
6753 6754 6755

	if (new_inode) {
		new_inode->i_ctime = CURRENT_TIME;
6756 6757 6758 6759 6760 6761 6762 6763 6764 6765 6766 6767 6768 6769 6770
		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);
6771
		if (new_inode->i_nlink == 0) {
6772
			ret = btrfs_orphan_add(trans, new_dentry->d_inode);
6773
			BUG_ON(ret);
6774
		}
C
Chris Mason 已提交
6775
	}
6776

6777 6778
	ret = btrfs_add_link(trans, new_dir, old_inode,
			     new_dentry->d_name.name,
6779
			     new_dentry->d_name.len, 0, index);
6780
	BUG_ON(ret);
C
Chris Mason 已提交
6781

6782
	if (old_inode->i_ino != BTRFS_FIRST_FREE_OBJECTID) {
6783 6784 6785
		struct dentry *parent = dget_parent(new_dentry);
		btrfs_log_new_name(trans, old_inode, old_dir, parent);
		dput(parent);
6786 6787
		btrfs_end_log_trans(root);
	}
C
Chris Mason 已提交
6788
out_fail:
6789
	btrfs_end_transaction_throttle(trans, root);
6790

6791 6792
	if (old_inode->i_ino == BTRFS_FIRST_FREE_OBJECTID)
		up_read(&root->fs_info->subvol_sem);
J
Josef Bacik 已提交
6793

C
Chris Mason 已提交
6794 6795 6796
	return ret;
}

C
Chris Mason 已提交
6797 6798 6799 6800
/*
 * 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
6801
int btrfs_start_delalloc_inodes(struct btrfs_root *root, int delay_iput)
6802 6803 6804
{
	struct list_head *head = &root->fs_info->delalloc_inodes;
	struct btrfs_inode *binode;
6805
	struct inode *inode;
6806

Y
Yan Zheng 已提交
6807 6808 6809
	if (root->fs_info->sb->s_flags & MS_RDONLY)
		return -EROFS;

6810
	spin_lock(&root->fs_info->delalloc_lock);
C
Chris Mason 已提交
6811
	while (!list_empty(head)) {
6812 6813
		binode = list_entry(head->next, struct btrfs_inode,
				    delalloc_inodes);
6814 6815 6816
		inode = igrab(&binode->vfs_inode);
		if (!inode)
			list_del_init(&binode->delalloc_inodes);
6817
		spin_unlock(&root->fs_info->delalloc_lock);
6818
		if (inode) {
6819
			filemap_flush(inode->i_mapping);
Y
Yan, Zheng 已提交
6820 6821 6822 6823
			if (delay_iput)
				btrfs_add_delayed_iput(inode);
			else
				iput(inode);
6824 6825
		}
		cond_resched();
6826
		spin_lock(&root->fs_info->delalloc_lock);
6827
	}
6828
	spin_unlock(&root->fs_info->delalloc_lock);
6829 6830 6831 6832 6833 6834

	/* 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
6835
	while (atomic_read(&root->fs_info->nr_async_submits) ||
6836
	      atomic_read(&root->fs_info->async_delalloc_pages)) {
6837
		wait_event(root->fs_info->async_submit_wait,
6838 6839
		   (atomic_read(&root->fs_info->nr_async_submits) == 0 &&
		    atomic_read(&root->fs_info->async_delalloc_pages) == 0));
6840 6841
	}
	atomic_dec(&root->fs_info->async_submit_draining);
6842 6843 6844
	return 0;
}

J
Josef Bacik 已提交
6845 6846
int btrfs_start_one_delalloc_inode(struct btrfs_root *root, int delay_iput,
				   int sync)
6847 6848 6849 6850 6851 6852 6853 6854 6855 6856 6857 6858 6859 6860 6861 6862 6863 6864 6865 6866 6867
{
	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
6868 6869 6870 6871 6872 6873 6874 6875 6876 6877 6878 6879 6880 6881 6882 6883 6884 6885 6886 6887
		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);
		}
6888 6889 6890 6891 6892 6893 6894 6895 6896
		if (delay_iput)
			btrfs_add_delayed_iput(inode);
		else
			iput(inode);
		return 1;
	}
	return 0;
}

C
Chris Mason 已提交
6897 6898 6899 6900 6901 6902 6903
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;
6904
	struct inode *inode = NULL;
C
Chris Mason 已提交
6905 6906 6907
	int err;
	int drop_inode = 0;
	u64 objectid;
6908
	u64 index = 0 ;
C
Chris Mason 已提交
6909 6910
	int name_len;
	int datasize;
6911
	unsigned long ptr;
C
Chris Mason 已提交
6912
	struct btrfs_file_extent_item *ei;
6913
	struct extent_buffer *leaf;
6914
	unsigned long nr = 0;
C
Chris Mason 已提交
6915 6916 6917 6918

	name_len = strlen(symname) + 1;
	if (name_len > BTRFS_MAX_INLINE_DATA_SIZE(root))
		return -ENAMETOOLONG;
6919

6920 6921 6922
	err = btrfs_find_free_objectid(NULL, root, dir->i_ino, &objectid);
	if (err)
		return err;
J
Josef Bacik 已提交
6923 6924 6925 6926 6927
	/*
	 * 2 items for inode item and ref
	 * 2 items for dir items
	 * 1 item for xattr if selinux is on
	 */
6928 6929 6930
	trans = btrfs_start_transaction(root, 5);
	if (IS_ERR(trans))
		return PTR_ERR(trans);
6931

C
Chris Mason 已提交
6932 6933
	btrfs_set_trans_block_group(trans, dir);

6934
	inode = btrfs_new_inode(trans, root, dir, dentry->d_name.name,
6935
				dentry->d_name.len, dir->i_ino, objectid,
6936 6937
				BTRFS_I(dir)->block_group, S_IFLNK|S_IRWXUGO,
				&index);
C
Chris Mason 已提交
6938 6939 6940 6941
	err = PTR_ERR(inode);
	if (IS_ERR(inode))
		goto out_unlock;

6942
	err = btrfs_init_inode_security(trans, inode, dir);
J
Josef Bacik 已提交
6943 6944 6945 6946 6947
	if (err) {
		drop_inode = 1;
		goto out_unlock;
	}

C
Chris Mason 已提交
6948
	btrfs_set_trans_block_group(trans, inode);
6949
	err = btrfs_add_nondir(trans, dir, dentry, inode, 0, index);
C
Chris Mason 已提交
6950 6951 6952 6953
	if (err)
		drop_inode = 1;
	else {
		inode->i_mapping->a_ops = &btrfs_aops;
C
Chris Mason 已提交
6954
		inode->i_mapping->backing_dev_info = &root->fs_info->bdi;
C
Chris Mason 已提交
6955 6956
		inode->i_fop = &btrfs_file_operations;
		inode->i_op = &btrfs_file_inode_operations;
6957
		BTRFS_I(inode)->io_tree.ops = &btrfs_extent_io_ops;
C
Chris Mason 已提交
6958 6959 6960 6961 6962 6963 6964 6965 6966 6967 6968 6969 6970 6971
	}
	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);
6972 6973 6974 6975
	if (err) {
		drop_inode = 1;
		goto out_unlock;
	}
6976 6977 6978 6979 6980
	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
6981
				   BTRFS_FILE_EXTENT_INLINE);
C
Chris Mason 已提交
6982 6983 6984 6985 6986
	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
6987
	ptr = btrfs_file_extent_inline_start(ei);
6988 6989
	write_extent_buffer(leaf, symname, ptr, name_len);
	btrfs_mark_buffer_dirty(leaf);
C
Chris Mason 已提交
6990
	btrfs_free_path(path);
6991

C
Chris Mason 已提交
6992 6993
	inode->i_op = &btrfs_symlink_inode_operations;
	inode->i_mapping->a_ops = &btrfs_symlink_aops;
C
Chris Mason 已提交
6994
	inode->i_mapping->backing_dev_info = &root->fs_info->bdi;
Y
Yan Zheng 已提交
6995
	inode_set_bytes(inode, name_len);
6996
	btrfs_i_size_write(inode, name_len - 1);
6997 6998 6999
	err = btrfs_update_inode(trans, root, inode);
	if (err)
		drop_inode = 1;
C
Chris Mason 已提交
7000 7001

out_unlock:
7002
	nr = trans->blocks_used;
7003
	btrfs_end_transaction_throttle(trans, root);
C
Chris Mason 已提交
7004 7005 7006 7007
	if (drop_inode) {
		inode_dec_link_count(inode);
		iput(inode);
	}
7008
	btrfs_btree_balance_dirty(root, nr);
C
Chris Mason 已提交
7009 7010
	return err;
}
7011

7012 7013 7014 7015
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
7016 7017 7018 7019
{
	struct btrfs_root *root = BTRFS_I(inode)->root;
	struct btrfs_key ins;
	u64 cur_offset = start;
7020
	u64 i_size;
Y
Yan Zheng 已提交
7021
	int ret = 0;
7022
	bool own_trans = true;
Y
Yan Zheng 已提交
7023

7024 7025
	if (trans)
		own_trans = false;
Y
Yan Zheng 已提交
7026
	while (num_bytes > 0) {
7027 7028 7029 7030 7031 7032
		if (own_trans) {
			trans = btrfs_start_transaction(root, 3);
			if (IS_ERR(trans)) {
				ret = PTR_ERR(trans);
				break;
			}
7033 7034
		}

7035 7036
		ret = btrfs_reserve_extent(trans, root, num_bytes, min_size,
					   0, *alloc_hint, (u64)-1, &ins, 1);
7037
		if (ret) {
7038 7039
			if (own_trans)
				btrfs_end_transaction(trans, root);
7040
			break;
Y
Yan Zheng 已提交
7041
		}
7042

Y
Yan Zheng 已提交
7043 7044 7045
		ret = insert_reserved_file_extent(trans, inode,
						  cur_offset, ins.objectid,
						  ins.offset, ins.offset,
Y
Yan, Zheng 已提交
7046
						  ins.offset, 0, 0, 0,
Y
Yan Zheng 已提交
7047 7048
						  BTRFS_FILE_EXTENT_PREALLOC);
		BUG_ON(ret);
C
Chris Mason 已提交
7049 7050
		btrfs_drop_extent_cache(inode, cur_offset,
					cur_offset + ins.offset -1, 0);
7051

Y
Yan Zheng 已提交
7052 7053
		num_bytes -= ins.offset;
		cur_offset += ins.offset;
7054
		*alloc_hint = ins.objectid + ins.offset;
7055

Y
Yan Zheng 已提交
7056
		inode->i_ctime = CURRENT_TIME;
7057
		BTRFS_I(inode)->flags |= BTRFS_INODE_PREALLOC;
Y
Yan Zheng 已提交
7058
		if (!(mode & FALLOC_FL_KEEP_SIZE) &&
7059 7060
		    (actual_len > inode->i_size) &&
		    (cur_offset > inode->i_size)) {
7061
			if (cur_offset > actual_len)
7062
				i_size = actual_len;
7063
			else
7064 7065 7066
				i_size = cur_offset;
			i_size_write(inode, i_size);
			btrfs_ordered_update_i_size(inode, i_size, NULL);
7067 7068
		}

Y
Yan Zheng 已提交
7069 7070 7071
		ret = btrfs_update_inode(trans, root, inode);
		BUG_ON(ret);

7072 7073
		if (own_trans)
			btrfs_end_transaction(trans, root);
7074
	}
Y
Yan Zheng 已提交
7075 7076 7077
	return ret;
}

7078 7079 7080 7081 7082 7083 7084 7085 7086 7087 7088 7089 7090 7091 7092 7093 7094 7095
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);
}

Y
Yan Zheng 已提交
7096 7097 7098
static long btrfs_fallocate(struct inode *inode, int mode,
			    loff_t offset, loff_t len)
{
7099
	struct extent_state *cached_state = NULL;
Y
Yan Zheng 已提交
7100 7101 7102 7103 7104
	u64 cur_offset;
	u64 last_byte;
	u64 alloc_start;
	u64 alloc_end;
	u64 alloc_hint = 0;
7105
	u64 locked_end;
Y
Yan Zheng 已提交
7106 7107 7108 7109 7110 7111 7112
	u64 mask = BTRFS_I(inode)->root->sectorsize - 1;
	struct extent_map *em;
	int ret;

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

7113 7114 7115 7116 7117 7118
	/*
	 * wait for ordered IO before we have any locks.  We'll loop again
	 * below with the locks held.
	 */
	btrfs_wait_ordered_range(inode, alloc_start, alloc_end - alloc_start);

Y
Yan Zheng 已提交
7119
	mutex_lock(&inode->i_mutex);
7120 7121 7122 7123
	ret = inode_newsize_ok(inode, alloc_end);
	if (ret)
		goto out;

Y
Yan Zheng 已提交
7124 7125 7126 7127 7128 7129
	if (alloc_start > inode->i_size) {
		ret = btrfs_cont_expand(inode, alloc_start);
		if (ret)
			goto out;
	}

7130
	ret = btrfs_check_data_free_space(inode, alloc_end - alloc_start);
7131 7132 7133
	if (ret)
		goto out;

7134
	locked_end = alloc_end - 1;
Y
Yan Zheng 已提交
7135 7136
	while (1) {
		struct btrfs_ordered_extent *ordered;
7137 7138 7139 7140

		/* the extent lock is ordered inside the running
		 * transaction
		 */
7141 7142
		lock_extent_bits(&BTRFS_I(inode)->io_tree, alloc_start,
				 locked_end, 0, &cached_state, GFP_NOFS);
Y
Yan Zheng 已提交
7143 7144 7145 7146 7147 7148
		ordered = btrfs_lookup_first_ordered_extent(inode,
							    alloc_end - 1);
		if (ordered &&
		    ordered->file_offset + ordered->len > alloc_start &&
		    ordered->file_offset < alloc_end) {
			btrfs_put_ordered_extent(ordered);
7149 7150 7151
			unlock_extent_cached(&BTRFS_I(inode)->io_tree,
					     alloc_start, locked_end,
					     &cached_state, GFP_NOFS);
7152 7153 7154 7155
			/*
			 * we can't wait on the range with the transaction
			 * running or with the extent lock held
			 */
Y
Yan Zheng 已提交
7156 7157 7158 7159 7160 7161 7162 7163 7164 7165 7166 7167 7168 7169 7170 7171
			btrfs_wait_ordered_range(inode, alloc_start,
						 alloc_end - alloc_start);
		} else {
			if (ordered)
				btrfs_put_ordered_extent(ordered);
			break;
		}
	}

	cur_offset = alloc_start;
	while (1) {
		em = btrfs_get_extent(inode, NULL, 0, cur_offset,
				      alloc_end - cur_offset, 0);
		BUG_ON(IS_ERR(em) || !em);
		last_byte = min(extent_map_end(em), alloc_end);
		last_byte = (last_byte + mask) & ~mask;
7172 7173 7174
		if (em->block_start == EXTENT_MAP_HOLE ||
		    (cur_offset >= inode->i_size &&
		     !test_bit(EXTENT_FLAG_PREALLOC, &em->flags))) {
J
Josef Bacik 已提交
7175
			ret = btrfs_prealloc_file_range(inode, mode, cur_offset,
7176 7177 7178 7179
							last_byte - cur_offset,
							1 << inode->i_blkbits,
							offset + len,
							&alloc_hint);
Y
Yan Zheng 已提交
7180 7181 7182 7183 7184 7185 7186 7187 7188 7189 7190 7191 7192
			if (ret < 0) {
				free_extent_map(em);
				break;
			}
		}
		free_extent_map(em);

		cur_offset = last_byte;
		if (cur_offset >= alloc_end) {
			ret = 0;
			break;
		}
	}
7193 7194
	unlock_extent_cached(&BTRFS_I(inode)->io_tree, alloc_start, locked_end,
			     &cached_state, GFP_NOFS);
7195

7196
	btrfs_free_reserved_data_space(inode, alloc_end - alloc_start);
Y
Yan Zheng 已提交
7197 7198 7199 7200 7201
out:
	mutex_unlock(&inode->i_mutex);
	return ret;
}

7202 7203 7204 7205 7206
static int btrfs_set_page_dirty(struct page *page)
{
	return __set_page_dirty_nobuffers(page);
}

7207
static int btrfs_permission(struct inode *inode, int mask)
Y
Yan 已提交
7208
{
7209
	if ((BTRFS_I(inode)->flags & BTRFS_INODE_READONLY) && (mask & MAY_WRITE))
Y
Yan 已提交
7210
		return -EACCES;
J
Josef Bacik 已提交
7211
	return generic_permission(inode, mask, btrfs_check_acl);
Y
Yan 已提交
7212
}
C
Chris Mason 已提交
7213

7214
static const struct inode_operations btrfs_dir_inode_operations = {
7215
	.getattr	= btrfs_getattr,
C
Chris Mason 已提交
7216 7217 7218 7219 7220 7221 7222 7223 7224
	.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
7225
	.mknod		= btrfs_mknod,
7226 7227
	.setxattr	= btrfs_setxattr,
	.getxattr	= btrfs_getxattr,
J
Josef Bacik 已提交
7228
	.listxattr	= btrfs_listxattr,
7229
	.removexattr	= btrfs_removexattr,
Y
Yan 已提交
7230
	.permission	= btrfs_permission,
C
Chris Mason 已提交
7231
};
7232
static const struct inode_operations btrfs_dir_ro_inode_operations = {
C
Chris Mason 已提交
7233
	.lookup		= btrfs_lookup,
Y
Yan 已提交
7234
	.permission	= btrfs_permission,
C
Chris Mason 已提交
7235
};
7236

7237
static const struct file_operations btrfs_dir_file_operations = {
C
Chris Mason 已提交
7238 7239
	.llseek		= generic_file_llseek,
	.read		= generic_read_dir,
7240
	.readdir	= btrfs_real_readdir,
C
Christoph Hellwig 已提交
7241
	.unlocked_ioctl	= btrfs_ioctl,
C
Chris Mason 已提交
7242
#ifdef CONFIG_COMPAT
C
Christoph Hellwig 已提交
7243
	.compat_ioctl	= btrfs_ioctl,
C
Chris Mason 已提交
7244
#endif
S
Sage Weil 已提交
7245
	.release        = btrfs_release_file,
7246
	.fsync		= btrfs_sync_file,
C
Chris Mason 已提交
7247 7248
};

7249
static struct extent_io_ops btrfs_extent_io_ops = {
7250
	.fill_delalloc = run_delalloc_range,
7251
	.submit_bio_hook = btrfs_submit_bio_hook,
7252
	.merge_bio_hook = btrfs_merge_bio_hook,
7253
	.readpage_end_io_hook = btrfs_readpage_end_io_hook,
7254
	.writepage_end_io_hook = btrfs_writepage_end_io_hook,
7255
	.writepage_start_hook = btrfs_writepage_start_hook,
7256
	.readpage_io_failed_hook = btrfs_io_failed_hook,
C
Chris Mason 已提交
7257 7258
	.set_bit_hook = btrfs_set_bit_hook,
	.clear_bit_hook = btrfs_clear_bit_hook,
J
Josef Bacik 已提交
7259 7260
	.merge_extent_hook = btrfs_merge_extent_hook,
	.split_extent_hook = btrfs_split_extent_hook,
7261 7262
};

7263 7264 7265 7266 7267 7268 7269 7270 7271 7272 7273 7274
/*
 * 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.
 */
7275
static const struct address_space_operations btrfs_aops = {
C
Chris Mason 已提交
7276 7277
	.readpage	= btrfs_readpage,
	.writepage	= btrfs_writepage,
C
Chris Mason 已提交
7278
	.writepages	= btrfs_writepages,
C
Chris Mason 已提交
7279
	.readpages	= btrfs_readpages,
C
Chris Mason 已提交
7280
	.sync_page	= block_sync_page,
7281
	.direct_IO	= btrfs_direct_IO,
7282 7283
	.invalidatepage = btrfs_invalidatepage,
	.releasepage	= btrfs_releasepage,
7284
	.set_page_dirty	= btrfs_set_page_dirty,
7285
	.error_remove_page = generic_error_remove_page,
C
Chris Mason 已提交
7286 7287
};

7288
static const struct address_space_operations btrfs_symlink_aops = {
C
Chris Mason 已提交
7289 7290
	.readpage	= btrfs_readpage,
	.writepage	= btrfs_writepage,
C
Chris Mason 已提交
7291 7292
	.invalidatepage = btrfs_invalidatepage,
	.releasepage	= btrfs_releasepage,
C
Chris Mason 已提交
7293 7294
};

7295
static const struct inode_operations btrfs_file_inode_operations = {
C
Chris Mason 已提交
7296 7297 7298
	.truncate	= btrfs_truncate,
	.getattr	= btrfs_getattr,
	.setattr	= btrfs_setattr,
7299 7300
	.setxattr	= btrfs_setxattr,
	.getxattr	= btrfs_getxattr,
J
Josef Bacik 已提交
7301
	.listxattr      = btrfs_listxattr,
7302
	.removexattr	= btrfs_removexattr,
Y
Yan 已提交
7303
	.permission	= btrfs_permission,
Y
Yan Zheng 已提交
7304
	.fallocate	= btrfs_fallocate,
Y
Yehuda Sadeh 已提交
7305
	.fiemap		= btrfs_fiemap,
C
Chris Mason 已提交
7306
};
7307
static const struct inode_operations btrfs_special_inode_operations = {
J
Josef Bacik 已提交
7308 7309
	.getattr	= btrfs_getattr,
	.setattr	= btrfs_setattr,
Y
Yan 已提交
7310
	.permission	= btrfs_permission,
7311 7312
	.setxattr	= btrfs_setxattr,
	.getxattr	= btrfs_getxattr,
J
Josef Bacik 已提交
7313
	.listxattr	= btrfs_listxattr,
7314
	.removexattr	= btrfs_removexattr,
J
Josef Bacik 已提交
7315
};
7316
static const struct inode_operations btrfs_symlink_inode_operations = {
C
Chris Mason 已提交
7317 7318 7319
	.readlink	= generic_readlink,
	.follow_link	= page_follow_link_light,
	.put_link	= page_put_link,
7320
	.getattr	= btrfs_getattr,
Y
Yan 已提交
7321
	.permission	= btrfs_permission,
J
Jim Owens 已提交
7322 7323 7324 7325
	.setxattr	= btrfs_setxattr,
	.getxattr	= btrfs_getxattr,
	.listxattr	= btrfs_listxattr,
	.removexattr	= btrfs_removexattr,
C
Chris Mason 已提交
7326
};
7327

7328
const struct dentry_operations btrfs_dentry_operations = {
7329 7330
	.d_delete	= btrfs_dentry_delete,
};