1// SPDX-License-Identifier: GPL-2.0
   2/*
   3 * Copyright (C) 2008 Oracle.  All rights reserved.
   4 */
   5
   6#include <linux/sched.h>
   7#include <linux/slab.h>
   8#include <linux/blkdev.h>
   9#include <linux/list_sort.h>
  10#include <linux/iversion.h>
  11#include "misc.h"
  12#include "ctree.h"
  13#include "tree-log.h"
  14#include "disk-io.h"
  15#include "locking.h"
  16#include "print-tree.h"
  17#include "backref.h"
  18#include "compression.h"
  19#include "qgroup.h"
  20#include "block-group.h"
  21#include "space-info.h"
  22#include "zoned.h"
  23#include "inode-item.h"
  24#include "fs.h"
  25#include "accessors.h"
  26#include "extent-tree.h"
  27#include "root-tree.h"
  28#include "dir-item.h"
  29#include "file-item.h"
  30#include "file.h"
  31#include "orphan.h"
  32#include "tree-checker.h"
  33
  34#define MAX_CONFLICT_INODES 10
  35
  36/* magic values for the inode_only field in btrfs_log_inode:
  37 *
  38 * LOG_INODE_ALL means to log everything
  39 * LOG_INODE_EXISTS means to log just enough to recreate the inode
  40 * during log replay
  41 */
  42enum {
  43	LOG_INODE_ALL,
  44	LOG_INODE_EXISTS,
  45};
  46
  47/*
  48 * directory trouble cases
  49 *
  50 * 1) on rename or unlink, if the inode being unlinked isn't in the fsync
  51 * log, we must force a full commit before doing an fsync of the directory
  52 * where the unlink was done.
  53 * ---> record transid of last unlink/rename per directory
  54 *
  55 * mkdir foo/some_dir
  56 * normal commit
  57 * rename foo/some_dir foo2/some_dir
  58 * mkdir foo/some_dir
  59 * fsync foo/some_dir/some_file
  60 *
  61 * The fsync above will unlink the original some_dir without recording
  62 * it in its new location (foo2).  After a crash, some_dir will be gone
  63 * unless the fsync of some_file forces a full commit
  64 *
  65 * 2) we must log any new names for any file or dir that is in the fsync
  66 * log. ---> check inode while renaming/linking.
  67 *
  68 * 2a) we must log any new names for any file or dir during rename
  69 * when the directory they are being removed from was logged.
  70 * ---> check inode and old parent dir during rename
  71 *
  72 *  2a is actually the more important variant.  With the extra logging
  73 *  a crash might unlink the old name without recreating the new one
  74 *
  75 * 3) after a crash, we must go through any directories with a link count
  76 * of zero and redo the rm -rf
  77 *
  78 * mkdir f1/foo
  79 * normal commit
  80 * rm -rf f1/foo
  81 * fsync(f1)
  82 *
  83 * The directory f1 was fully removed from the FS, but fsync was never
  84 * called on f1, only its parent dir.  After a crash the rm -rf must
  85 * be replayed.  This must be able to recurse down the entire
  86 * directory tree.  The inode link count fixup code takes care of the
  87 * ugly details.
  88 */
  89
  90/*
  91 * stages for the tree walking.  The first
  92 * stage (0) is to only pin down the blocks we find
  93 * the second stage (1) is to make sure that all the inodes
  94 * we find in the log are created in the subvolume.
  95 *
  96 * The last stage is to deal with directories and links and extents
  97 * and all the other fun semantics
  98 */
  99enum {
 100	LOG_WALK_PIN_ONLY,
 101	LOG_WALK_REPLAY_INODES,
 102	LOG_WALK_REPLAY_DIR_INDEX,
 103	LOG_WALK_REPLAY_ALL,
 104};
 105
 106static int btrfs_log_inode(struct btrfs_trans_handle *trans,
 107			   struct btrfs_inode *inode,
 108			   int inode_only,
 109			   struct btrfs_log_ctx *ctx);
 110static int link_to_fixup_dir(struct btrfs_trans_handle *trans,
 111			     struct btrfs_root *root,
 112			     struct btrfs_path *path, u64 objectid);
 113static noinline int replay_dir_deletes(struct btrfs_trans_handle *trans,
 114				       struct btrfs_root *root,
 115				       struct btrfs_root *log,
 116				       struct btrfs_path *path,
 117				       u64 dirid, int del_all);
 118static void wait_log_commit(struct btrfs_root *root, int transid);
 119
 120/*
 121 * tree logging is a special write ahead log used to make sure that
 122 * fsyncs and O_SYNCs can happen without doing full tree commits.
 123 *
 124 * Full tree commits are expensive because they require commonly
 125 * modified blocks to be recowed, creating many dirty pages in the
 126 * extent tree an 4x-6x higher write load than ext3.
 127 *
 128 * Instead of doing a tree commit on every fsync, we use the
 129 * key ranges and transaction ids to find items for a given file or directory
 130 * that have changed in this transaction.  Those items are copied into
 131 * a special tree (one per subvolume root), that tree is written to disk
 132 * and then the fsync is considered complete.
 133 *
 134 * After a crash, items are copied out of the log-tree back into the
 135 * subvolume tree.  Any file data extents found are recorded in the extent
 136 * allocation tree, and the log-tree freed.
 137 *
 138 * The log tree is read three times, once to pin down all the extents it is
 139 * using in ram and once, once to create all the inodes logged in the tree
 140 * and once to do all the other items.
 141 */
 142
 143/*
 144 * start a sub transaction and setup the log tree
 145 * this increments the log tree writer count to make the people
 146 * syncing the tree wait for us to finish
 147 */
 148static int start_log_trans(struct btrfs_trans_handle *trans,
 149			   struct btrfs_root *root,
 150			   struct btrfs_log_ctx *ctx)
 151{
 152	struct btrfs_fs_info *fs_info = root->fs_info;
 153	struct btrfs_root *tree_root = fs_info->tree_root;
 154	const bool zoned = btrfs_is_zoned(fs_info);
 155	int ret = 0;
 156	bool created = false;
 157
 158	/*
 159	 * First check if the log root tree was already created. If not, create
 160	 * it before locking the root's log_mutex, just to keep lockdep happy.
 161	 */
 162	if (!test_bit(BTRFS_ROOT_HAS_LOG_TREE, &tree_root->state)) {
 163		mutex_lock(&tree_root->log_mutex);
 164		if (!fs_info->log_root_tree) {
 165			ret = btrfs_init_log_root_tree(trans, fs_info);
 166			if (!ret) {
 167				set_bit(BTRFS_ROOT_HAS_LOG_TREE, &tree_root->state);
 168				created = true;
 169			}
 170		}
 171		mutex_unlock(&tree_root->log_mutex);
 172		if (ret)
 173			return ret;
 174	}
 175
 176	mutex_lock(&root->log_mutex);
 177
 178again:
 179	if (root->log_root) {
 180		int index = (root->log_transid + 1) % 2;
 181
 182		if (btrfs_need_log_full_commit(trans)) {
 183			ret = BTRFS_LOG_FORCE_COMMIT;
 184			goto out;
 185		}
 186
 187		if (zoned && atomic_read(&root->log_commit[index])) {
 188			wait_log_commit(root, root->log_transid - 1);
 189			goto again;
 190		}
 191
 192		if (!root->log_start_pid) {
 193			clear_bit(BTRFS_ROOT_MULTI_LOG_TASKS, &root->state);
 194			root->log_start_pid = current->pid;
 195		} else if (root->log_start_pid != current->pid) {
 196			set_bit(BTRFS_ROOT_MULTI_LOG_TASKS, &root->state);
 197		}
 198	} else {
 199		/*
 200		 * This means fs_info->log_root_tree was already created
 201		 * for some other FS trees. Do the full commit not to mix
 202		 * nodes from multiple log transactions to do sequential
 203		 * writing.
 204		 */
 205		if (zoned && !created) {
 206			ret = BTRFS_LOG_FORCE_COMMIT;
 207			goto out;
 208		}
 209
 210		ret = btrfs_add_log_tree(trans, root);
 211		if (ret)
 212			goto out;
 213
 214		set_bit(BTRFS_ROOT_HAS_LOG_TREE, &root->state);
 215		clear_bit(BTRFS_ROOT_MULTI_LOG_TASKS, &root->state);
 216		root->log_start_pid = current->pid;
 217	}
 218
 219	atomic_inc(&root->log_writers);
 220	if (!ctx->logging_new_name) {
 221		int index = root->log_transid % 2;
 222		list_add_tail(&ctx->list, &root->log_ctxs[index]);
 223		ctx->log_transid = root->log_transid;
 224	}
 225
 226out:
 227	mutex_unlock(&root->log_mutex);
 228	return ret;
 229}
 230
 231/*
 232 * returns 0 if there was a log transaction running and we were able
 233 * to join, or returns -ENOENT if there were not transactions
 234 * in progress
 235 */
 236static int join_running_log_trans(struct btrfs_root *root)
 237{
 238	const bool zoned = btrfs_is_zoned(root->fs_info);
 239	int ret = -ENOENT;
 240
 241	if (!test_bit(BTRFS_ROOT_HAS_LOG_TREE, &root->state))
 242		return ret;
 243
 244	mutex_lock(&root->log_mutex);
 245again:
 246	if (root->log_root) {
 247		int index = (root->log_transid + 1) % 2;
 248
 249		ret = 0;
 250		if (zoned && atomic_read(&root->log_commit[index])) {
 251			wait_log_commit(root, root->log_transid - 1);
 252			goto again;
 253		}
 254		atomic_inc(&root->log_writers);
 255	}
 256	mutex_unlock(&root->log_mutex);
 257	return ret;
 258}
 259
 260/*
 261 * This either makes the current running log transaction wait
 262 * until you call btrfs_end_log_trans() or it makes any future
 263 * log transactions wait until you call btrfs_end_log_trans()
 264 */
 265void btrfs_pin_log_trans(struct btrfs_root *root)
 266{
 267	atomic_inc(&root->log_writers);
 268}
 269
 270/*
 271 * indicate we're done making changes to the log tree
 272 * and wake up anyone waiting to do a sync
 273 */
 274void btrfs_end_log_trans(struct btrfs_root *root)
 275{
 276	if (atomic_dec_and_test(&root->log_writers)) {
 277		/* atomic_dec_and_test implies a barrier */
 278		cond_wake_up_nomb(&root->log_writer_wait);
 279	}
 280}
 281
 282static void btrfs_wait_tree_block_writeback(struct extent_buffer *buf)
 283{
 284	filemap_fdatawait_range(buf->pages[0]->mapping,
 285			        buf->start, buf->start + buf->len - 1);
 286}
 287
 288/*
 289 * the walk control struct is used to pass state down the chain when
 290 * processing the log tree.  The stage field tells us which part
 291 * of the log tree processing we are currently doing.  The others
 292 * are state fields used for that specific part
 293 */
 294struct walk_control {
 295	/* should we free the extent on disk when done?  This is used
 296	 * at transaction commit time while freeing a log tree
 297	 */
 298	int free;
 299
 300	/* pin only walk, we record which extents on disk belong to the
 301	 * log trees
 302	 */
 303	int pin;
 304
 305	/* what stage of the replay code we're currently in */
 306	int stage;
 307
 308	/*
 309	 * Ignore any items from the inode currently being processed. Needs
 310	 * to be set every time we find a BTRFS_INODE_ITEM_KEY and we are in
 311	 * the LOG_WALK_REPLAY_INODES stage.
 312	 */
 313	bool ignore_cur_inode;
 314
 315	/* the root we are currently replaying */
 316	struct btrfs_root *replay_dest;
 317
 318	/* the trans handle for the current replay */
 319	struct btrfs_trans_handle *trans;
 320
 321	/* the function that gets used to process blocks we find in the
 322	 * tree.  Note the extent_buffer might not be up to date when it is
 323	 * passed in, and it must be checked or read if you need the data
 324	 * inside it
 325	 */
 326	int (*process_func)(struct btrfs_root *log, struct extent_buffer *eb,
 327			    struct walk_control *wc, u64 gen, int level);
 328};
 329
 330/*
 331 * process_func used to pin down extents, write them or wait on them
 332 */
 333static int process_one_buffer(struct btrfs_root *log,
 334			      struct extent_buffer *eb,
 335			      struct walk_control *wc, u64 gen, int level)
 336{
 337	struct btrfs_fs_info *fs_info = log->fs_info;
 338	int ret = 0;
 339
 340	/*
 341	 * If this fs is mixed then we need to be able to process the leaves to
 342	 * pin down any logged extents, so we have to read the block.
 343	 */
 344	if (btrfs_fs_incompat(fs_info, MIXED_GROUPS)) {
 345		struct btrfs_tree_parent_check check = {
 346			.level = level,
 347			.transid = gen
 348		};
 349
 350		ret = btrfs_read_extent_buffer(eb, &check);
 351		if (ret)
 352			return ret;
 353	}
 354
 355	if (wc->pin) {
 356		ret = btrfs_pin_extent_for_log_replay(wc->trans, eb->start,
 357						      eb->len);
 358		if (ret)
 359			return ret;
 360
 361		if (btrfs_buffer_uptodate(eb, gen, 0) &&
 362		    btrfs_header_level(eb) == 0)
 363			ret = btrfs_exclude_logged_extents(eb);
 364	}
 365	return ret;
 366}
 367
 368/*
 369 * Item overwrite used by replay and tree logging.  eb, slot and key all refer
 370 * to the src data we are copying out.
 371 *
 372 * root is the tree we are copying into, and path is a scratch
 373 * path for use in this function (it should be released on entry and
 374 * will be released on exit).
 375 *
 376 * If the key is already in the destination tree the existing item is
 377 * overwritten.  If the existing item isn't big enough, it is extended.
 378 * If it is too large, it is truncated.
 379 *
 380 * If the key isn't in the destination yet, a new item is inserted.
 381 */
 382static int overwrite_item(struct btrfs_trans_handle *trans,
 383			  struct btrfs_root *root,
 384			  struct btrfs_path *path,
 385			  struct extent_buffer *eb, int slot,
 386			  struct btrfs_key *key)
 387{
 388	int ret;
 389	u32 item_size;
 390	u64 saved_i_size = 0;
 391	int save_old_i_size = 0;
 392	unsigned long src_ptr;
 393	unsigned long dst_ptr;
 394	bool inode_item = key->type == BTRFS_INODE_ITEM_KEY;
 395
 396	/*
 397	 * This is only used during log replay, so the root is always from a
 398	 * fs/subvolume tree. In case we ever need to support a log root, then
 399	 * we'll have to clone the leaf in the path, release the path and use
 400	 * the leaf before writing into the log tree. See the comments at
 401	 * copy_items() for more details.
 402	 */
 403	ASSERT(root->root_key.objectid != BTRFS_TREE_LOG_OBJECTID);
 404
 405	item_size = btrfs_item_size(eb, slot);
 406	src_ptr = btrfs_item_ptr_offset(eb, slot);
 407
 408	/* Look for the key in the destination tree. */
 409	ret = btrfs_search_slot(NULL, root, key, path, 0, 0);
 410	if (ret < 0)
 411		return ret;
 412
 413	if (ret == 0) {
 414		char *src_copy;
 415		char *dst_copy;
 416		u32 dst_size = btrfs_item_size(path->nodes[0],
 417						  path->slots[0]);
 418		if (dst_size != item_size)
 419			goto insert;
 420
 421		if (item_size == 0) {
 422			btrfs_release_path(path);
 423			return 0;
 424		}
 425		dst_copy = kmalloc(item_size, GFP_NOFS);
 426		src_copy = kmalloc(item_size, GFP_NOFS);
 427		if (!dst_copy || !src_copy) {
 428			btrfs_release_path(path);
 429			kfree(dst_copy);
 430			kfree(src_copy);
 431			return -ENOMEM;
 432		}
 433
 434		read_extent_buffer(eb, src_copy, src_ptr, item_size);
 435
 436		dst_ptr = btrfs_item_ptr_offset(path->nodes[0], path->slots[0]);
 437		read_extent_buffer(path->nodes[0], dst_copy, dst_ptr,
 438				   item_size);
 439		ret = memcmp(dst_copy, src_copy, item_size);
 440
 441		kfree(dst_copy);
 442		kfree(src_copy);
 443		/*
 444		 * they have the same contents, just return, this saves
 445		 * us from cowing blocks in the destination tree and doing
 446		 * extra writes that may not have been done by a previous
 447		 * sync
 448		 */
 449		if (ret == 0) {
 450			btrfs_release_path(path);
 451			return 0;
 452		}
 453
 454		/*
 455		 * We need to load the old nbytes into the inode so when we
 456		 * replay the extents we've logged we get the right nbytes.
 457		 */
 458		if (inode_item) {
 459			struct btrfs_inode_item *item;
 460			u64 nbytes;
 461			u32 mode;
 462
 463			item = btrfs_item_ptr(path->nodes[0], path->slots[0],
 464					      struct btrfs_inode_item);
 465			nbytes = btrfs_inode_nbytes(path->nodes[0], item);
 466			item = btrfs_item_ptr(eb, slot,
 467					      struct btrfs_inode_item);
 468			btrfs_set_inode_nbytes(eb, item, nbytes);
 469
 470			/*
 471			 * If this is a directory we need to reset the i_size to
 472			 * 0 so that we can set it up properly when replaying
 473			 * the rest of the items in this log.
 474			 */
 475			mode = btrfs_inode_mode(eb, item);
 476			if (S_ISDIR(mode))
 477				btrfs_set_inode_size(eb, item, 0);
 478		}
 479	} else if (inode_item) {
 480		struct btrfs_inode_item *item;
 481		u32 mode;
 482
 483		/*
 484		 * New inode, set nbytes to 0 so that the nbytes comes out
 485		 * properly when we replay the extents.
 486		 */
 487		item = btrfs_item_ptr(eb, slot, struct btrfs_inode_item);
 488		btrfs_set_inode_nbytes(eb, item, 0);
 489
 490		/*
 491		 * If this is a directory we need to reset the i_size to 0 so
 492		 * that we can set it up properly when replaying the rest of
 493		 * the items in this log.
 494		 */
 495		mode = btrfs_inode_mode(eb, item);
 496		if (S_ISDIR(mode))
 497			btrfs_set_inode_size(eb, item, 0);
 498	}
 499insert:
 500	btrfs_release_path(path);
 501	/* try to insert the key into the destination tree */
 502	path->skip_release_on_error = 1;
 503	ret = btrfs_insert_empty_item(trans, root, path,
 504				      key, item_size);
 505	path->skip_release_on_error = 0;
 506
 507	/* make sure any existing item is the correct size */
 508	if (ret == -EEXIST || ret == -EOVERFLOW) {
 509		u32 found_size;
 510		found_size = btrfs_item_size(path->nodes[0],
 511						path->slots[0]);
 512		if (found_size > item_size)
 513			btrfs_truncate_item(path, item_size, 1);
 514		else if (found_size < item_size)
 515			btrfs_extend_item(path, item_size - found_size);
 516	} else if (ret) {
 517		return ret;
 518	}
 519	dst_ptr = btrfs_item_ptr_offset(path->nodes[0],
 520					path->slots[0]);
 521
 522	/* don't overwrite an existing inode if the generation number
 523	 * was logged as zero.  This is done when the tree logging code
 524	 * is just logging an inode to make sure it exists after recovery.
 525	 *
 526	 * Also, don't overwrite i_size on directories during replay.
 527	 * log replay inserts and removes directory items based on the
 528	 * state of the tree found in the subvolume, and i_size is modified
 529	 * as it goes
 530	 */
 531	if (key->type == BTRFS_INODE_ITEM_KEY && ret == -EEXIST) {
 532		struct btrfs_inode_item *src_item;
 533		struct btrfs_inode_item *dst_item;
 534
 535		src_item = (struct btrfs_inode_item *)src_ptr;
 536		dst_item = (struct btrfs_inode_item *)dst_ptr;
 537
 538		if (btrfs_inode_generation(eb, src_item) == 0) {
 539			struct extent_buffer *dst_eb = path->nodes[0];
 540			const u64 ino_size = btrfs_inode_size(eb, src_item);
 541
 542			/*
 543			 * For regular files an ino_size == 0 is used only when
 544			 * logging that an inode exists, as part of a directory
 545			 * fsync, and the inode wasn't fsynced before. In this
 546			 * case don't set the size of the inode in the fs/subvol
 547			 * tree, otherwise we would be throwing valid data away.
 548			 */
 549			if (S_ISREG(btrfs_inode_mode(eb, src_item)) &&
 550			    S_ISREG(btrfs_inode_mode(dst_eb, dst_item)) &&
 551			    ino_size != 0)
 552				btrfs_set_inode_size(dst_eb, dst_item, ino_size);
 553			goto no_copy;
 554		}
 555
 556		if (S_ISDIR(btrfs_inode_mode(eb, src_item)) &&
 557		    S_ISDIR(btrfs_inode_mode(path->nodes[0], dst_item))) {
 558			save_old_i_size = 1;
 559			saved_i_size = btrfs_inode_size(path->nodes[0],
 560							dst_item);
 561		}
 562	}
 563
 564	copy_extent_buffer(path->nodes[0], eb, dst_ptr,
 565			   src_ptr, item_size);
 566
 567	if (save_old_i_size) {
 568		struct btrfs_inode_item *dst_item;
 569		dst_item = (struct btrfs_inode_item *)dst_ptr;
 570		btrfs_set_inode_size(path->nodes[0], dst_item, saved_i_size);
 571	}
 572
 573	/* make sure the generation is filled in */
 574	if (key->type == BTRFS_INODE_ITEM_KEY) {
 575		struct btrfs_inode_item *dst_item;
 576		dst_item = (struct btrfs_inode_item *)dst_ptr;
 577		if (btrfs_inode_generation(path->nodes[0], dst_item) == 0) {
 578			btrfs_set_inode_generation(path->nodes[0], dst_item,
 579						   trans->transid);
 580		}
 581	}
 582no_copy:
 583	btrfs_mark_buffer_dirty(path->nodes[0]);
 584	btrfs_release_path(path);
 585	return 0;
 586}
 587
 588static int read_alloc_one_name(struct extent_buffer *eb, void *start, int len,
 589			       struct fscrypt_str *name)
 590{
 591	char *buf;
 592
 593	buf = kmalloc(len, GFP_NOFS);
 594	if (!buf)
 595		return -ENOMEM;
 596
 597	read_extent_buffer(eb, buf, (unsigned long)start, len);
 598	name->name = buf;
 599	name->len = len;
 600	return 0;
 601}
 602
 603/*
 604 * simple helper to read an inode off the disk from a given root
 605 * This can only be called for subvolume roots and not for the log
 606 */
 607static noinline struct inode *read_one_inode(struct btrfs_root *root,
 608					     u64 objectid)
 609{
 610	struct inode *inode;
 611
 612	inode = btrfs_iget(root->fs_info->sb, objectid, root);
 613	if (IS_ERR(inode))
 614		inode = NULL;
 615	return inode;
 616}
 617
 618/* replays a single extent in 'eb' at 'slot' with 'key' into the
 619 * subvolume 'root'.  path is released on entry and should be released
 620 * on exit.
 621 *
 622 * extents in the log tree have not been allocated out of the extent
 623 * tree yet.  So, this completes the allocation, taking a reference
 624 * as required if the extent already exists or creating a new extent
 625 * if it isn't in the extent allocation tree yet.
 626 *
 627 * The extent is inserted into the file, dropping any existing extents
 628 * from the file that overlap the new one.
 629 */
 630static noinline int replay_one_extent(struct btrfs_trans_handle *trans,
 631				      struct btrfs_root *root,
 632				      struct btrfs_path *path,
 633				      struct extent_buffer *eb, int slot,
 634				      struct btrfs_key *key)
 635{
 636	struct btrfs_drop_extents_args drop_args = { 0 };
 637	struct btrfs_fs_info *fs_info = root->fs_info;
 638	int found_type;
 639	u64 extent_end;
 640	u64 start = key->offset;
 641	u64 nbytes = 0;
 642	struct btrfs_file_extent_item *item;
 643	struct inode *inode = NULL;
 644	unsigned long size;
 645	int ret = 0;
 646
 647	item = btrfs_item_ptr(eb, slot, struct btrfs_file_extent_item);
 648	found_type = btrfs_file_extent_type(eb, item);
 649
 650	if (found_type == BTRFS_FILE_EXTENT_REG ||
 651	    found_type == BTRFS_FILE_EXTENT_PREALLOC) {
 652		nbytes = btrfs_file_extent_num_bytes(eb, item);
 653		extent_end = start + nbytes;
 654
 655		/*
 656		 * We don't add to the inodes nbytes if we are prealloc or a
 657		 * hole.
 658		 */
 659		if (btrfs_file_extent_disk_bytenr(eb, item) == 0)
 660			nbytes = 0;
 661	} else if (found_type == BTRFS_FILE_EXTENT_INLINE) {
 662		size = btrfs_file_extent_ram_bytes(eb, item);
 663		nbytes = btrfs_file_extent_ram_bytes(eb, item);
 664		extent_end = ALIGN(start + size,
 665				   fs_info->sectorsize);
 666	} else {
 667		ret = 0;
 668		goto out;
 669	}
 670
 671	inode = read_one_inode(root, key->objectid);
 672	if (!inode) {
 673		ret = -EIO;
 674		goto out;
 675	}
 676
 677	/*
 678	 * first check to see if we already have this extent in the
 679	 * file.  This must be done before the btrfs_drop_extents run
 680	 * so we don't try to drop this extent.
 681	 */
 682	ret = btrfs_lookup_file_extent(trans, root, path,
 683			btrfs_ino(BTRFS_I(inode)), start, 0);
 684
 685	if (ret == 0 &&
 686	    (found_type == BTRFS_FILE_EXTENT_REG ||
 687	     found_type == BTRFS_FILE_EXTENT_PREALLOC)) {
 688		struct btrfs_file_extent_item cmp1;
 689		struct btrfs_file_extent_item cmp2;
 690		struct btrfs_file_extent_item *existing;
 691		struct extent_buffer *leaf;
 692
 693		leaf = path->nodes[0];
 694		existing = btrfs_item_ptr(leaf, path->slots[0],
 695					  struct btrfs_file_extent_item);
 696
 697		read_extent_buffer(eb, &cmp1, (unsigned long)item,
 698				   sizeof(cmp1));
 699		read_extent_buffer(leaf, &cmp2, (unsigned long)existing,
 700				   sizeof(cmp2));
 701
 702		/*
 703		 * we already have a pointer to this exact extent,
 704		 * we don't have to do anything
 705		 */
 706		if (memcmp(&cmp1, &cmp2, sizeof(cmp1)) == 0) {
 707			btrfs_release_path(path);
 708			goto out;
 709		}
 710	}
 711	btrfs_release_path(path);
 712
 713	/* drop any overlapping extents */
 714	drop_args.start = start;
 715	drop_args.end = extent_end;
 716	drop_args.drop_cache = true;
 717	ret = btrfs_drop_extents(trans, root, BTRFS_I(inode), &drop_args);
 718	if (ret)
 719		goto out;
 720
 721	if (found_type == BTRFS_FILE_EXTENT_REG ||
 722	    found_type == BTRFS_FILE_EXTENT_PREALLOC) {
 723		u64 offset;
 724		unsigned long dest_offset;
 725		struct btrfs_key ins;
 726
 727		if (btrfs_file_extent_disk_bytenr(eb, item) == 0 &&
 728		    btrfs_fs_incompat(fs_info, NO_HOLES))
 729			goto update_inode;
 730
 731		ret = btrfs_insert_empty_item(trans, root, path, key,
 732					      sizeof(*item));
 733		if (ret)
 734			goto out;
 735		dest_offset = btrfs_item_ptr_offset(path->nodes[0],
 736						    path->slots[0]);
 737		copy_extent_buffer(path->nodes[0], eb, dest_offset,
 738				(unsigned long)item,  sizeof(*item));
 739
 740		ins.objectid = btrfs_file_extent_disk_bytenr(eb, item);
 741		ins.offset = btrfs_file_extent_disk_num_bytes(eb, item);
 742		ins.type = BTRFS_EXTENT_ITEM_KEY;
 743		offset = key->offset - btrfs_file_extent_offset(eb, item);
 744
 745		/*
 746		 * Manually record dirty extent, as here we did a shallow
 747		 * file extent item copy and skip normal backref update,
 748		 * but modifying extent tree all by ourselves.
 749		 * So need to manually record dirty extent for qgroup,
 750		 * as the owner of the file extent changed from log tree
 751		 * (doesn't affect qgroup) to fs/file tree(affects qgroup)
 752		 */
 753		ret = btrfs_qgroup_trace_extent(trans,
 754				btrfs_file_extent_disk_bytenr(eb, item),
 755				btrfs_file_extent_disk_num_bytes(eb, item));
 756		if (ret < 0)
 757			goto out;
 758
 759		if (ins.objectid > 0) {
 760			struct btrfs_ref ref = { 0 };
 761			u64 csum_start;
 762			u64 csum_end;
 763			LIST_HEAD(ordered_sums);
 764
 765			/*
 766			 * is this extent already allocated in the extent
 767			 * allocation tree?  If so, just add a reference
 768			 */
 769			ret = btrfs_lookup_data_extent(fs_info, ins.objectid,
 770						ins.offset);
 771			if (ret < 0) {
 772				goto out;
 773			} else if (ret == 0) {
 774				btrfs_init_generic_ref(&ref,
 775						BTRFS_ADD_DELAYED_REF,
 776						ins.objectid, ins.offset, 0);
 
 777				btrfs_init_data_ref(&ref,
 778						root->root_key.objectid,
 779						key->objectid, offset, 0, false);
 780				ret = btrfs_inc_extent_ref(trans, &ref);
 781				if (ret)
 782					goto out;
 783			} else {
 784				/*
 785				 * insert the extent pointer in the extent
 786				 * allocation tree
 787				 */
 788				ret = btrfs_alloc_logged_file_extent(trans,
 789						root->root_key.objectid,
 790						key->objectid, offset, &ins);
 791				if (ret)
 792					goto out;
 793			}
 794			btrfs_release_path(path);
 795
 796			if (btrfs_file_extent_compression(eb, item)) {
 797				csum_start = ins.objectid;
 798				csum_end = csum_start + ins.offset;
 799			} else {
 800				csum_start = ins.objectid +
 801					btrfs_file_extent_offset(eb, item);
 802				csum_end = csum_start +
 803					btrfs_file_extent_num_bytes(eb, item);
 804			}
 805
 806			ret = btrfs_lookup_csums_list(root->log_root,
 807						csum_start, csum_end - 1,
 808						&ordered_sums, 0, false);
 809			if (ret)
 810				goto out;
 811			/*
 812			 * Now delete all existing cums in the csum root that
 813			 * cover our range. We do this because we can have an
 814			 * extent that is completely referenced by one file
 815			 * extent item and partially referenced by another
 816			 * file extent item (like after using the clone or
 817			 * extent_same ioctls). In this case if we end up doing
 818			 * the replay of the one that partially references the
 819			 * extent first, and we do not do the csum deletion
 820			 * below, we can get 2 csum items in the csum tree that
 821			 * overlap each other. For example, imagine our log has
 822			 * the two following file extent items:
 823			 *
 824			 * key (257 EXTENT_DATA 409600)
 825			 *     extent data disk byte 12845056 nr 102400
 826			 *     extent data offset 20480 nr 20480 ram 102400
 827			 *
 828			 * key (257 EXTENT_DATA 819200)
 829			 *     extent data disk byte 12845056 nr 102400
 830			 *     extent data offset 0 nr 102400 ram 102400
 831			 *
 832			 * Where the second one fully references the 100K extent
 833			 * that starts at disk byte 12845056, and the log tree
 834			 * has a single csum item that covers the entire range
 835			 * of the extent:
 836			 *
 837			 * key (EXTENT_CSUM EXTENT_CSUM 12845056) itemsize 100
 838			 *
 839			 * After the first file extent item is replayed, the
 840			 * csum tree gets the following csum item:
 841			 *
 842			 * key (EXTENT_CSUM EXTENT_CSUM 12865536) itemsize 20
 843			 *
 844			 * Which covers the 20K sub-range starting at offset 20K
 845			 * of our extent. Now when we replay the second file
 846			 * extent item, if we do not delete existing csum items
 847			 * that cover any of its blocks, we end up getting two
 848			 * csum items in our csum tree that overlap each other:
 849			 *
 850			 * key (EXTENT_CSUM EXTENT_CSUM 12845056) itemsize 100
 851			 * key (EXTENT_CSUM EXTENT_CSUM 12865536) itemsize 20
 852			 *
 853			 * Which is a problem, because after this anyone trying
 854			 * to lookup up for the checksum of any block of our
 855			 * extent starting at an offset of 40K or higher, will
 856			 * end up looking at the second csum item only, which
 857			 * does not contain the checksum for any block starting
 858			 * at offset 40K or higher of our extent.
 859			 */
 860			while (!list_empty(&ordered_sums)) {
 861				struct btrfs_ordered_sum *sums;
 862				struct btrfs_root *csum_root;
 863
 864				sums = list_entry(ordered_sums.next,
 865						struct btrfs_ordered_sum,
 866						list);
 867				csum_root = btrfs_csum_root(fs_info,
 868							    sums->bytenr);
 869				if (!ret)
 870					ret = btrfs_del_csums(trans, csum_root,
 871							      sums->bytenr,
 872							      sums->len);
 873				if (!ret)
 874					ret = btrfs_csum_file_blocks(trans,
 875								     csum_root,
 876								     sums);
 877				list_del(&sums->list);
 878				kfree(sums);
 879			}
 880			if (ret)
 881				goto out;
 882		} else {
 883			btrfs_release_path(path);
 884		}
 885	} else if (found_type == BTRFS_FILE_EXTENT_INLINE) {
 886		/* inline extents are easy, we just overwrite them */
 887		ret = overwrite_item(trans, root, path, eb, slot, key);
 888		if (ret)
 889			goto out;
 890	}
 891
 892	ret = btrfs_inode_set_file_extent_range(BTRFS_I(inode), start,
 893						extent_end - start);
 894	if (ret)
 895		goto out;
 896
 897update_inode:
 898	btrfs_update_inode_bytes(BTRFS_I(inode), nbytes, drop_args.bytes_found);
 899	ret = btrfs_update_inode(trans, root, BTRFS_I(inode));
 900out:
 901	iput(inode);
 902	return ret;
 903}
 904
 905static int unlink_inode_for_log_replay(struct btrfs_trans_handle *trans,
 906				       struct btrfs_inode *dir,
 907				       struct btrfs_inode *inode,
 908				       const struct fscrypt_str *name)
 909{
 910	int ret;
 911
 912	ret = btrfs_unlink_inode(trans, dir, inode, name);
 913	if (ret)
 914		return ret;
 915	/*
 916	 * Whenever we need to check if a name exists or not, we check the
 917	 * fs/subvolume tree. So after an unlink we must run delayed items, so
 918	 * that future checks for a name during log replay see that the name
 919	 * does not exists anymore.
 920	 */
 921	return btrfs_run_delayed_items(trans);
 922}
 923
 924/*
 925 * when cleaning up conflicts between the directory names in the
 926 * subvolume, directory names in the log and directory names in the
 927 * inode back references, we may have to unlink inodes from directories.
 928 *
 929 * This is a helper function to do the unlink of a specific directory
 930 * item
 931 */
 932static noinline int drop_one_dir_item(struct btrfs_trans_handle *trans,
 933				      struct btrfs_path *path,
 934				      struct btrfs_inode *dir,
 935				      struct btrfs_dir_item *di)
 936{
 937	struct btrfs_root *root = dir->root;
 938	struct inode *inode;
 939	struct fscrypt_str name;
 940	struct extent_buffer *leaf;
 941	struct btrfs_key location;
 942	int ret;
 943
 944	leaf = path->nodes[0];
 945
 946	btrfs_dir_item_key_to_cpu(leaf, di, &location);
 947	ret = read_alloc_one_name(leaf, di + 1, btrfs_dir_name_len(leaf, di), &name);
 948	if (ret)
 949		return -ENOMEM;
 950
 951	btrfs_release_path(path);
 952
 953	inode = read_one_inode(root, location.objectid);
 954	if (!inode) {
 955		ret = -EIO;
 956		goto out;
 957	}
 958
 959	ret = link_to_fixup_dir(trans, root, path, location.objectid);
 960	if (ret)
 961		goto out;
 962
 963	ret = unlink_inode_for_log_replay(trans, dir, BTRFS_I(inode), &name);
 964out:
 965	kfree(name.name);
 966	iput(inode);
 967	return ret;
 968}
 969
 970/*
 971 * See if a given name and sequence number found in an inode back reference are
 972 * already in a directory and correctly point to this inode.
 973 *
 974 * Returns: < 0 on error, 0 if the directory entry does not exists and 1 if it
 975 * exists.
 976 */
 977static noinline int inode_in_dir(struct btrfs_root *root,
 978				 struct btrfs_path *path,
 979				 u64 dirid, u64 objectid, u64 index,
 980				 struct fscrypt_str *name)
 981{
 982	struct btrfs_dir_item *di;
 983	struct btrfs_key location;
 984	int ret = 0;
 985
 986	di = btrfs_lookup_dir_index_item(NULL, root, path, dirid,
 987					 index, name, 0);
 988	if (IS_ERR(di)) {
 989		ret = PTR_ERR(di);
 990		goto out;
 991	} else if (di) {
 992		btrfs_dir_item_key_to_cpu(path->nodes[0], di, &location);
 993		if (location.objectid != objectid)
 994			goto out;
 995	} else {
 996		goto out;
 997	}
 998
 999	btrfs_release_path(path);
1000	di = btrfs_lookup_dir_item(NULL, root, path, dirid, name, 0);
1001	if (IS_ERR(di)) {
1002		ret = PTR_ERR(di);
1003		goto out;
1004	} else if (di) {
1005		btrfs_dir_item_key_to_cpu(path->nodes[0], di, &location);
1006		if (location.objectid == objectid)
1007			ret = 1;
1008	}
1009out:
1010	btrfs_release_path(path);
1011	return ret;
1012}
1013
1014/*
1015 * helper function to check a log tree for a named back reference in
1016 * an inode.  This is used to decide if a back reference that is
1017 * found in the subvolume conflicts with what we find in the log.
1018 *
1019 * inode backreferences may have multiple refs in a single item,
1020 * during replay we process one reference at a time, and we don't
1021 * want to delete valid links to a file from the subvolume if that
1022 * link is also in the log.
1023 */
1024static noinline int backref_in_log(struct btrfs_root *log,
1025				   struct btrfs_key *key,
1026				   u64 ref_objectid,
1027				   const struct fscrypt_str *name)
1028{
1029	struct btrfs_path *path;
1030	int ret;
1031
1032	path = btrfs_alloc_path();
1033	if (!path)
1034		return -ENOMEM;
1035
1036	ret = btrfs_search_slot(NULL, log, key, path, 0, 0);
1037	if (ret < 0) {
1038		goto out;
1039	} else if (ret == 1) {
1040		ret = 0;
1041		goto out;
1042	}
1043
1044	if (key->type == BTRFS_INODE_EXTREF_KEY)
1045		ret = !!btrfs_find_name_in_ext_backref(path->nodes[0],
1046						       path->slots[0],
1047						       ref_objectid, name);
1048	else
1049		ret = !!btrfs_find_name_in_backref(path->nodes[0],
1050						   path->slots[0], name);
1051out:
1052	btrfs_free_path(path);
1053	return ret;
1054}
1055
1056static inline int __add_inode_ref(struct btrfs_trans_handle *trans,
1057				  struct btrfs_root *root,
1058				  struct btrfs_path *path,
1059				  struct btrfs_root *log_root,
1060				  struct btrfs_inode *dir,
1061				  struct btrfs_inode *inode,
1062				  u64 inode_objectid, u64 parent_objectid,
1063				  u64 ref_index, struct fscrypt_str *name)
1064{
1065	int ret;
1066	struct extent_buffer *leaf;
1067	struct btrfs_dir_item *di;
1068	struct btrfs_key search_key;
1069	struct btrfs_inode_extref *extref;
1070
1071again:
1072	/* Search old style refs */
1073	search_key.objectid = inode_objectid;
1074	search_key.type = BTRFS_INODE_REF_KEY;
1075	search_key.offset = parent_objectid;
1076	ret = btrfs_search_slot(NULL, root, &search_key, path, 0, 0);
1077	if (ret == 0) {
1078		struct btrfs_inode_ref *victim_ref;
1079		unsigned long ptr;
1080		unsigned long ptr_end;
1081
1082		leaf = path->nodes[0];
1083
1084		/* are we trying to overwrite a back ref for the root directory
1085		 * if so, just jump out, we're done
1086		 */
1087		if (search_key.objectid == search_key.offset)
1088			return 1;
1089
1090		/* check all the names in this back reference to see
1091		 * if they are in the log.  if so, we allow them to stay
1092		 * otherwise they must be unlinked as a conflict
1093		 */
1094		ptr = btrfs_item_ptr_offset(leaf, path->slots[0]);
1095		ptr_end = ptr + btrfs_item_size(leaf, path->slots[0]);
1096		while (ptr < ptr_end) {
1097			struct fscrypt_str victim_name;
1098
1099			victim_ref = (struct btrfs_inode_ref *)ptr;
1100			ret = read_alloc_one_name(leaf, (victim_ref + 1),
1101				 btrfs_inode_ref_name_len(leaf, victim_ref),
1102				 &victim_name);
1103			if (ret)
1104				return ret;
1105
1106			ret = backref_in_log(log_root, &search_key,
1107					     parent_objectid, &victim_name);
1108			if (ret < 0) {
1109				kfree(victim_name.name);
1110				return ret;
1111			} else if (!ret) {
1112				inc_nlink(&inode->vfs_inode);
1113				btrfs_release_path(path);
1114
1115				ret = unlink_inode_for_log_replay(trans, dir, inode,
1116						&victim_name);
1117				kfree(victim_name.name);
1118				if (ret)
1119					return ret;
1120				goto again;
1121			}
1122			kfree(victim_name.name);
1123
1124			ptr = (unsigned long)(victim_ref + 1) + victim_name.len;
1125		}
1126	}
1127	btrfs_release_path(path);
1128
1129	/* Same search but for extended refs */
1130	extref = btrfs_lookup_inode_extref(NULL, root, path, name,
1131					   inode_objectid, parent_objectid, 0,
1132					   0);
1133	if (IS_ERR(extref)) {
1134		return PTR_ERR(extref);
1135	} else if (extref) {
1136		u32 item_size;
1137		u32 cur_offset = 0;
1138		unsigned long base;
1139		struct inode *victim_parent;
1140
1141		leaf = path->nodes[0];
1142
1143		item_size = btrfs_item_size(leaf, path->slots[0]);
1144		base = btrfs_item_ptr_offset(leaf, path->slots[0]);
1145
1146		while (cur_offset < item_size) {
1147			struct fscrypt_str victim_name;
1148
1149			extref = (struct btrfs_inode_extref *)(base + cur_offset);
1150
1151			if (btrfs_inode_extref_parent(leaf, extref) != parent_objectid)
1152				goto next;
1153
1154			ret = read_alloc_one_name(leaf, &extref->name,
1155				 btrfs_inode_extref_name_len(leaf, extref),
1156				 &victim_name);
1157			if (ret)
1158				return ret;
1159
1160			search_key.objectid = inode_objectid;
1161			search_key.type = BTRFS_INODE_EXTREF_KEY;
1162			search_key.offset = btrfs_extref_hash(parent_objectid,
1163							      victim_name.name,
1164							      victim_name.len);
1165			ret = backref_in_log(log_root, &search_key,
1166					     parent_objectid, &victim_name);
1167			if (ret < 0) {
1168				kfree(victim_name.name);
1169				return ret;
1170			} else if (!ret) {
1171				ret = -ENOENT;
1172				victim_parent = read_one_inode(root,
1173						parent_objectid);
1174				if (victim_parent) {
1175					inc_nlink(&inode->vfs_inode);
1176					btrfs_release_path(path);
1177
1178					ret = unlink_inode_for_log_replay(trans,
1179							BTRFS_I(victim_parent),
1180							inode, &victim_name);
1181				}
1182				iput(victim_parent);
1183				kfree(victim_name.name);
1184				if (ret)
1185					return ret;
1186				goto again;
1187			}
1188			kfree(victim_name.name);
1189next:
1190			cur_offset += victim_name.len + sizeof(*extref);
1191		}
1192	}
1193	btrfs_release_path(path);
1194
1195	/* look for a conflicting sequence number */
1196	di = btrfs_lookup_dir_index_item(trans, root, path, btrfs_ino(dir),
1197					 ref_index, name, 0);
1198	if (IS_ERR(di)) {
1199		return PTR_ERR(di);
1200	} else if (di) {
1201		ret = drop_one_dir_item(trans, path, dir, di);
1202		if (ret)
1203			return ret;
1204	}
1205	btrfs_release_path(path);
1206
1207	/* look for a conflicting name */
1208	di = btrfs_lookup_dir_item(trans, root, path, btrfs_ino(dir), name, 0);
1209	if (IS_ERR(di)) {
1210		return PTR_ERR(di);
1211	} else if (di) {
1212		ret = drop_one_dir_item(trans, path, dir, di);
1213		if (ret)
1214			return ret;
1215	}
1216	btrfs_release_path(path);
1217
1218	return 0;
1219}
1220
1221static int extref_get_fields(struct extent_buffer *eb, unsigned long ref_ptr,
1222			     struct fscrypt_str *name, u64 *index,
1223			     u64 *parent_objectid)
1224{
1225	struct btrfs_inode_extref *extref;
1226	int ret;
1227
1228	extref = (struct btrfs_inode_extref *)ref_ptr;
1229
1230	ret = read_alloc_one_name(eb, &extref->name,
1231				  btrfs_inode_extref_name_len(eb, extref), name);
1232	if (ret)
1233		return ret;
1234
1235	if (index)
1236		*index = btrfs_inode_extref_index(eb, extref);
1237	if (parent_objectid)
1238		*parent_objectid = btrfs_inode_extref_parent(eb, extref);
1239
1240	return 0;
1241}
1242
1243static int ref_get_fields(struct extent_buffer *eb, unsigned long ref_ptr,
1244			  struct fscrypt_str *name, u64 *index)
1245{
1246	struct btrfs_inode_ref *ref;
1247	int ret;
1248
1249	ref = (struct btrfs_inode_ref *)ref_ptr;
1250
1251	ret = read_alloc_one_name(eb, ref + 1, btrfs_inode_ref_name_len(eb, ref),
1252				  name);
1253	if (ret)
1254		return ret;
1255
1256	if (index)
1257		*index = btrfs_inode_ref_index(eb, ref);
1258
1259	return 0;
1260}
1261
1262/*
1263 * Take an inode reference item from the log tree and iterate all names from the
1264 * inode reference item in the subvolume tree with the same key (if it exists).
1265 * For any name that is not in the inode reference item from the log tree, do a
1266 * proper unlink of that name (that is, remove its entry from the inode
1267 * reference item and both dir index keys).
1268 */
1269static int unlink_old_inode_refs(struct btrfs_trans_handle *trans,
1270				 struct btrfs_root *root,
1271				 struct btrfs_path *path,
1272				 struct btrfs_inode *inode,
1273				 struct extent_buffer *log_eb,
1274				 int log_slot,
1275				 struct btrfs_key *key)
1276{
1277	int ret;
1278	unsigned long ref_ptr;
1279	unsigned long ref_end;
1280	struct extent_buffer *eb;
1281
1282again:
1283	btrfs_release_path(path);
1284	ret = btrfs_search_slot(NULL, root, key, path, 0, 0);
1285	if (ret > 0) {
1286		ret = 0;
1287		goto out;
1288	}
1289	if (ret < 0)
1290		goto out;
1291
1292	eb = path->nodes[0];
1293	ref_ptr = btrfs_item_ptr_offset(eb, path->slots[0]);
1294	ref_end = ref_ptr + btrfs_item_size(eb, path->slots[0]);
1295	while (ref_ptr < ref_end) {
1296		struct fscrypt_str name;
1297		u64 parent_id;
1298
1299		if (key->type == BTRFS_INODE_EXTREF_KEY) {
1300			ret = extref_get_fields(eb, ref_ptr, &name,
1301						NULL, &parent_id);
1302		} else {
1303			parent_id = key->offset;
1304			ret = ref_get_fields(eb, ref_ptr, &name, NULL);
1305		}
1306		if (ret)
1307			goto out;
1308
1309		if (key->type == BTRFS_INODE_EXTREF_KEY)
1310			ret = !!btrfs_find_name_in_ext_backref(log_eb, log_slot,
1311							       parent_id, &name);
1312		else
1313			ret = !!btrfs_find_name_in_backref(log_eb, log_slot, &name);
1314
1315		if (!ret) {
1316			struct inode *dir;
1317
1318			btrfs_release_path(path);
1319			dir = read_one_inode(root, parent_id);
1320			if (!dir) {
1321				ret = -ENOENT;
1322				kfree(name.name);
1323				goto out;
1324			}
1325			ret = unlink_inode_for_log_replay(trans, BTRFS_I(dir),
1326						 inode, &name);
1327			kfree(name.name);
1328			iput(dir);
1329			if (ret)
1330				goto out;
1331			goto again;
1332		}
1333
1334		kfree(name.name);
1335		ref_ptr += name.len;
1336		if (key->type == BTRFS_INODE_EXTREF_KEY)
1337			ref_ptr += sizeof(struct btrfs_inode_extref);
1338		else
1339			ref_ptr += sizeof(struct btrfs_inode_ref);
1340	}
1341	ret = 0;
1342 out:
1343	btrfs_release_path(path);
1344	return ret;
1345}
1346
1347/*
1348 * replay one inode back reference item found in the log tree.
1349 * eb, slot and key refer to the buffer and key found in the log tree.
1350 * root is the destination we are replaying into, and path is for temp
1351 * use by this function.  (it should be released on return).
1352 */
1353static noinline int add_inode_ref(struct btrfs_trans_handle *trans,
1354				  struct btrfs_root *root,
1355				  struct btrfs_root *log,
1356				  struct btrfs_path *path,
1357				  struct extent_buffer *eb, int slot,
1358				  struct btrfs_key *key)
1359{
1360	struct inode *dir = NULL;
1361	struct inode *inode = NULL;
1362	unsigned long ref_ptr;
1363	unsigned long ref_end;
1364	struct fscrypt_str name;
1365	int ret;
1366	int log_ref_ver = 0;
1367	u64 parent_objectid;
1368	u64 inode_objectid;
1369	u64 ref_index = 0;
1370	int ref_struct_size;
1371
1372	ref_ptr = btrfs_item_ptr_offset(eb, slot);
1373	ref_end = ref_ptr + btrfs_item_size(eb, slot);
1374
1375	if (key->type == BTRFS_INODE_EXTREF_KEY) {
1376		struct btrfs_inode_extref *r;
1377
1378		ref_struct_size = sizeof(struct btrfs_inode_extref);
1379		log_ref_ver = 1;
1380		r = (struct btrfs_inode_extref *)ref_ptr;
1381		parent_objectid = btrfs_inode_extref_parent(eb, r);
1382	} else {
1383		ref_struct_size = sizeof(struct btrfs_inode_ref);
1384		parent_objectid = key->offset;
1385	}
1386	inode_objectid = key->objectid;
1387
1388	/*
1389	 * it is possible that we didn't log all the parent directories
1390	 * for a given inode.  If we don't find the dir, just don't
1391	 * copy the back ref in.  The link count fixup code will take
1392	 * care of the rest
1393	 */
1394	dir = read_one_inode(root, parent_objectid);
1395	if (!dir) {
1396		ret = -ENOENT;
1397		goto out;
1398	}
1399
1400	inode = read_one_inode(root, inode_objectid);
1401	if (!inode) {
1402		ret = -EIO;
1403		goto out;
1404	}
1405
1406	while (ref_ptr < ref_end) {
1407		if (log_ref_ver) {
1408			ret = extref_get_fields(eb, ref_ptr, &name,
1409						&ref_index, &parent_objectid);
1410			/*
1411			 * parent object can change from one array
1412			 * item to another.
1413			 */
1414			if (!dir)
1415				dir = read_one_inode(root, parent_objectid);
1416			if (!dir) {
1417				ret = -ENOENT;
1418				goto out;
1419			}
1420		} else {
1421			ret = ref_get_fields(eb, ref_ptr, &name, &ref_index);
1422		}
1423		if (ret)
1424			goto out;
1425
1426		ret = inode_in_dir(root, path, btrfs_ino(BTRFS_I(dir)),
1427				   btrfs_ino(BTRFS_I(inode)), ref_index, &name);
1428		if (ret < 0) {
1429			goto out;
1430		} else if (ret == 0) {
1431			/*
1432			 * look for a conflicting back reference in the
1433			 * metadata. if we find one we have to unlink that name
1434			 * of the file before we add our new link.  Later on, we
1435			 * overwrite any existing back reference, and we don't
1436			 * want to create dangling pointers in the directory.
1437			 */
1438			ret = __add_inode_ref(trans, root, path, log,
1439					      BTRFS_I(dir), BTRFS_I(inode),
1440					      inode_objectid, parent_objectid,
1441					      ref_index, &name);
1442			if (ret) {
1443				if (ret == 1)
1444					ret = 0;
1445				goto out;
1446			}
1447
1448			/* insert our name */
1449			ret = btrfs_add_link(trans, BTRFS_I(dir), BTRFS_I(inode),
1450					     &name, 0, ref_index);
1451			if (ret)
1452				goto out;
1453
1454			ret = btrfs_update_inode(trans, root, BTRFS_I(inode));
1455			if (ret)
1456				goto out;
1457		}
1458		/* Else, ret == 1, we already have a perfect match, we're done. */
1459
1460		ref_ptr = (unsigned long)(ref_ptr + ref_struct_size) + name.len;
1461		kfree(name.name);
1462		name.name = NULL;
1463		if (log_ref_ver) {
1464			iput(dir);
1465			dir = NULL;
1466		}
1467	}
1468
1469	/*
1470	 * Before we overwrite the inode reference item in the subvolume tree
1471	 * with the item from the log tree, we must unlink all names from the
1472	 * parent directory that are in the subvolume's tree inode reference
1473	 * item, otherwise we end up with an inconsistent subvolume tree where
1474	 * dir index entries exist for a name but there is no inode reference
1475	 * item with the same name.
1476	 */
1477	ret = unlink_old_inode_refs(trans, root, path, BTRFS_I(inode), eb, slot,
1478				    key);
1479	if (ret)
1480		goto out;
1481
1482	/* finally write the back reference in the inode */
1483	ret = overwrite_item(trans, root, path, eb, slot, key);
1484out:
1485	btrfs_release_path(path);
1486	kfree(name.name);
1487	iput(dir);
1488	iput(inode);
1489	return ret;
1490}
1491
1492static int count_inode_extrefs(struct btrfs_root *root,
1493		struct btrfs_inode *inode, struct btrfs_path *path)
1494{
1495	int ret = 0;
1496	int name_len;
1497	unsigned int nlink = 0;
1498	u32 item_size;
1499	u32 cur_offset = 0;
1500	u64 inode_objectid = btrfs_ino(inode);
1501	u64 offset = 0;
1502	unsigned long ptr;
1503	struct btrfs_inode_extref *extref;
1504	struct extent_buffer *leaf;
1505
1506	while (1) {
1507		ret = btrfs_find_one_extref(root, inode_objectid, offset, path,
1508					    &extref, &offset);
1509		if (ret)
1510			break;
1511
1512		leaf = path->nodes[0];
1513		item_size = btrfs_item_size(leaf, path->slots[0]);
1514		ptr = btrfs_item_ptr_offset(leaf, path->slots[0]);
1515		cur_offset = 0;
1516
1517		while (cur_offset < item_size) {
1518			extref = (struct btrfs_inode_extref *) (ptr + cur_offset);
1519			name_len = btrfs_inode_extref_name_len(leaf, extref);
1520
1521			nlink++;
1522
1523			cur_offset += name_len + sizeof(*extref);
1524		}
1525
1526		offset++;
1527		btrfs_release_path(path);
1528	}
1529	btrfs_release_path(path);
1530
1531	if (ret < 0 && ret != -ENOENT)
1532		return ret;
1533	return nlink;
1534}
1535
1536static int count_inode_refs(struct btrfs_root *root,
1537			struct btrfs_inode *inode, struct btrfs_path *path)
1538{
1539	int ret;
1540	struct btrfs_key key;
1541	unsigned int nlink = 0;
1542	unsigned long ptr;
1543	unsigned long ptr_end;
1544	int name_len;
1545	u64 ino = btrfs_ino(inode);
1546
1547	key.objectid = ino;
1548	key.type = BTRFS_INODE_REF_KEY;
1549	key.offset = (u64)-1;
1550
1551	while (1) {
1552		ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
1553		if (ret < 0)
1554			break;
1555		if (ret > 0) {
1556			if (path->slots[0] == 0)
1557				break;
1558			path->slots[0]--;
1559		}
1560process_slot:
1561		btrfs_item_key_to_cpu(path->nodes[0], &key,
1562				      path->slots[0]);
1563		if (key.objectid != ino ||
1564		    key.type != BTRFS_INODE_REF_KEY)
1565			break;
1566		ptr = btrfs_item_ptr_offset(path->nodes[0], path->slots[0]);
1567		ptr_end = ptr + btrfs_item_size(path->nodes[0],
1568						   path->slots[0]);
1569		while (ptr < ptr_end) {
1570			struct btrfs_inode_ref *ref;
1571
1572			ref = (struct btrfs_inode_ref *)ptr;
1573			name_len = btrfs_inode_ref_name_len(path->nodes[0],
1574							    ref);
1575			ptr = (unsigned long)(ref + 1) + name_len;
1576			nlink++;
1577		}
1578
1579		if (key.offset == 0)
1580			break;
1581		if (path->slots[0] > 0) {
1582			path->slots[0]--;
1583			goto process_slot;
1584		}
1585		key.offset--;
1586		btrfs_release_path(path);
1587	}
1588	btrfs_release_path(path);
1589
1590	return nlink;
1591}
1592
1593/*
1594 * There are a few corners where the link count of the file can't
1595 * be properly maintained during replay.  So, instead of adding
1596 * lots of complexity to the log code, we just scan the backrefs
1597 * for any file that has been through replay.
1598 *
1599 * The scan will update the link count on the inode to reflect the
1600 * number of back refs found.  If it goes down to zero, the iput
1601 * will free the inode.
1602 */
1603static noinline int fixup_inode_link_count(struct btrfs_trans_handle *trans,
1604					   struct btrfs_root *root,
1605					   struct inode *inode)
1606{
 
1607	struct btrfs_path *path;
1608	int ret;
1609	u64 nlink = 0;
1610	u64 ino = btrfs_ino(BTRFS_I(inode));
1611
1612	path = btrfs_alloc_path();
1613	if (!path)
1614		return -ENOMEM;
1615
1616	ret = count_inode_refs(root, BTRFS_I(inode), path);
1617	if (ret < 0)
1618		goto out;
1619
1620	nlink = ret;
1621
1622	ret = count_inode_extrefs(root, BTRFS_I(inode), path);
1623	if (ret < 0)
1624		goto out;
1625
1626	nlink += ret;
1627
1628	ret = 0;
1629
1630	if (nlink != inode->i_nlink) {
1631		set_nlink(inode, nlink);
1632		ret = btrfs_update_inode(trans, root, BTRFS_I(inode));
1633		if (ret)
1634			goto out;
1635	}
1636	BTRFS_I(inode)->index_cnt = (u64)-1;
1637
1638	if (inode->i_nlink == 0) {
1639		if (S_ISDIR(inode->i_mode)) {
1640			ret = replay_dir_deletes(trans, root, NULL, path,
1641						 ino, 1);
1642			if (ret)
1643				goto out;
1644		}
1645		ret = btrfs_insert_orphan_item(trans, root, ino);
1646		if (ret == -EEXIST)
1647			ret = 0;
1648	}
1649
1650out:
1651	btrfs_free_path(path);
1652	return ret;
1653}
1654
1655static noinline int fixup_inode_link_counts(struct btrfs_trans_handle *trans,
1656					    struct btrfs_root *root,
1657					    struct btrfs_path *path)
1658{
1659	int ret;
1660	struct btrfs_key key;
1661	struct inode *inode;
1662
1663	key.objectid = BTRFS_TREE_LOG_FIXUP_OBJECTID;
1664	key.type = BTRFS_ORPHAN_ITEM_KEY;
1665	key.offset = (u64)-1;
1666	while (1) {
1667		ret = btrfs_search_slot(trans, root, &key, path, -1, 1);
1668		if (ret < 0)
1669			break;
1670
1671		if (ret == 1) {
1672			ret = 0;
1673			if (path->slots[0] == 0)
1674				break;
1675			path->slots[0]--;
1676		}
1677
1678		btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
1679		if (key.objectid != BTRFS_TREE_LOG_FIXUP_OBJECTID ||
1680		    key.type != BTRFS_ORPHAN_ITEM_KEY)
1681			break;
1682
1683		ret = btrfs_del_item(trans, root, path);
1684		if (ret)
1685			break;
1686
1687		btrfs_release_path(path);
1688		inode = read_one_inode(root, key.offset);
1689		if (!inode) {
1690			ret = -EIO;
1691			break;
1692		}
1693
1694		ret = fixup_inode_link_count(trans, root, inode);
1695		iput(inode);
1696		if (ret)
1697			break;
1698
1699		/*
1700		 * fixup on a directory may create new entries,
1701		 * make sure we always look for the highset possible
1702		 * offset
1703		 */
1704		key.offset = (u64)-1;
1705	}
1706	btrfs_release_path(path);
1707	return ret;
1708}
1709
1710
1711/*
1712 * record a given inode in the fixup dir so we can check its link
1713 * count when replay is done.  The link count is incremented here
1714 * so the inode won't go away until we check it
1715 */
1716static noinline int link_to_fixup_dir(struct btrfs_trans_handle *trans,
1717				      struct btrfs_root *root,
1718				      struct btrfs_path *path,
1719				      u64 objectid)
1720{
1721	struct btrfs_key key;
1722	int ret = 0;
1723	struct inode *inode;
1724
1725	inode = read_one_inode(root, objectid);
1726	if (!inode)
1727		return -EIO;
1728
1729	key.objectid = BTRFS_TREE_LOG_FIXUP_OBJECTID;
1730	key.type = BTRFS_ORPHAN_ITEM_KEY;
1731	key.offset = objectid;
1732
1733	ret = btrfs_insert_empty_item(trans, root, path, &key, 0);
1734
1735	btrfs_release_path(path);
1736	if (ret == 0) {
1737		if (!inode->i_nlink)
1738			set_nlink(inode, 1);
1739		else
1740			inc_nlink(inode);
1741		ret = btrfs_update_inode(trans, root, BTRFS_I(inode));
1742	} else if (ret == -EEXIST) {
1743		ret = 0;
1744	}
1745	iput(inode);
1746
1747	return ret;
1748}
1749
1750/*
1751 * when replaying the log for a directory, we only insert names
1752 * for inodes that actually exist.  This means an fsync on a directory
1753 * does not implicitly fsync all the new files in it
1754 */
1755static noinline int insert_one_name(struct btrfs_trans_handle *trans,
1756				    struct btrfs_root *root,
1757				    u64 dirid, u64 index,
1758				    const struct fscrypt_str *name,
1759				    struct btrfs_key *location)
1760{
1761	struct inode *inode;
1762	struct inode *dir;
1763	int ret;
1764
1765	inode = read_one_inode(root, location->objectid);
1766	if (!inode)
1767		return -ENOENT;
1768
1769	dir = read_one_inode(root, dirid);
1770	if (!dir) {
1771		iput(inode);
1772		return -EIO;
1773	}
1774
1775	ret = btrfs_add_link(trans, BTRFS_I(dir), BTRFS_I(inode), name,
1776			     1, index);
1777
1778	/* FIXME, put inode into FIXUP list */
1779
1780	iput(inode);
1781	iput(dir);
1782	return ret;
1783}
1784
1785static int delete_conflicting_dir_entry(struct btrfs_trans_handle *trans,
1786					struct btrfs_inode *dir,
1787					struct btrfs_path *path,
1788					struct btrfs_dir_item *dst_di,
1789					const struct btrfs_key *log_key,
1790					u8 log_flags,
1791					bool exists)
1792{
1793	struct btrfs_key found_key;
1794
1795	btrfs_dir_item_key_to_cpu(path->nodes[0], dst_di, &found_key);
1796	/* The existing dentry points to the same inode, don't delete it. */
1797	if (found_key.objectid == log_key->objectid &&
1798	    found_key.type == log_key->type &&
1799	    found_key.offset == log_key->offset &&
1800	    btrfs_dir_flags(path->nodes[0], dst_di) == log_flags)
1801		return 1;
1802
1803	/*
1804	 * Don't drop the conflicting directory entry if the inode for the new
1805	 * entry doesn't exist.
1806	 */
1807	if (!exists)
1808		return 0;
1809
1810	return drop_one_dir_item(trans, path, dir, dst_di);
1811}
1812
1813/*
1814 * take a single entry in a log directory item and replay it into
1815 * the subvolume.
1816 *
1817 * if a conflicting item exists in the subdirectory already,
1818 * the inode it points to is unlinked and put into the link count
1819 * fix up tree.
1820 *
1821 * If a name from the log points to a file or directory that does
1822 * not exist in the FS, it is skipped.  fsyncs on directories
1823 * do not force down inodes inside that directory, just changes to the
1824 * names or unlinks in a directory.
1825 *
1826 * Returns < 0 on error, 0 if the name wasn't replayed (dentry points to a
1827 * non-existing inode) and 1 if the name was replayed.
1828 */
1829static noinline int replay_one_name(struct btrfs_trans_handle *trans,
1830				    struct btrfs_root *root,
1831				    struct btrfs_path *path,
1832				    struct extent_buffer *eb,
1833				    struct btrfs_dir_item *di,
1834				    struct btrfs_key *key)
1835{
1836	struct fscrypt_str name;
1837	struct btrfs_dir_item *dir_dst_di;
1838	struct btrfs_dir_item *index_dst_di;
1839	bool dir_dst_matches = false;
1840	bool index_dst_matches = false;
1841	struct btrfs_key log_key;
1842	struct btrfs_key search_key;
1843	struct inode *dir;
1844	u8 log_flags;
1845	bool exists;
1846	int ret;
1847	bool update_size = true;
1848	bool name_added = false;
1849
1850	dir = read_one_inode(root, key->objectid);
1851	if (!dir)
1852		return -EIO;
1853
1854	ret = read_alloc_one_name(eb, di + 1, btrfs_dir_name_len(eb, di), &name);
1855	if (ret)
1856		goto out;
1857
1858	log_flags = btrfs_dir_flags(eb, di);
1859	btrfs_dir_item_key_to_cpu(eb, di, &log_key);
1860	ret = btrfs_lookup_inode(trans, root, path, &log_key, 0);
1861	btrfs_release_path(path);
1862	if (ret < 0)
1863		goto out;
1864	exists = (ret == 0);
1865	ret = 0;
1866
1867	dir_dst_di = btrfs_lookup_dir_item(trans, root, path, key->objectid,
1868					   &name, 1);
1869	if (IS_ERR(dir_dst_di)) {
1870		ret = PTR_ERR(dir_dst_di);
1871		goto out;
1872	} else if (dir_dst_di) {
1873		ret = delete_conflicting_dir_entry(trans, BTRFS_I(dir), path,
1874						   dir_dst_di, &log_key,
1875						   log_flags, exists);
1876		if (ret < 0)
1877			goto out;
1878		dir_dst_matches = (ret == 1);
1879	}
1880
1881	btrfs_release_path(path);
1882
1883	index_dst_di = btrfs_lookup_dir_index_item(trans, root, path,
1884						   key->objectid, key->offset,
1885						   &name, 1);
1886	if (IS_ERR(index_dst_di)) {
1887		ret = PTR_ERR(index_dst_di);
1888		goto out;
1889	} else if (index_dst_di) {
1890		ret = delete_conflicting_dir_entry(trans, BTRFS_I(dir), path,
1891						   index_dst_di, &log_key,
1892						   log_flags, exists);
1893		if (ret < 0)
1894			goto out;
1895		index_dst_matches = (ret == 1);
1896	}
1897
1898	btrfs_release_path(path);
1899
1900	if (dir_dst_matches && index_dst_matches) {
1901		ret = 0;
1902		update_size = false;
1903		goto out;
1904	}
1905
1906	/*
1907	 * Check if the inode reference exists in the log for the given name,
1908	 * inode and parent inode
1909	 */
1910	search_key.objectid = log_key.objectid;
1911	search_key.type = BTRFS_INODE_REF_KEY;
1912	search_key.offset = key->objectid;
1913	ret = backref_in_log(root->log_root, &search_key, 0, &name);
1914	if (ret < 0) {
1915	        goto out;
1916	} else if (ret) {
1917	        /* The dentry will be added later. */
1918	        ret = 0;
1919	        update_size = false;
1920	        goto out;
1921	}
1922
1923	search_key.objectid = log_key.objectid;
1924	search_key.type = BTRFS_INODE_EXTREF_KEY;
1925	search_key.offset = key->objectid;
1926	ret = backref_in_log(root->log_root, &search_key, key->objectid, &name);
1927	if (ret < 0) {
1928		goto out;
1929	} else if (ret) {
1930		/* The dentry will be added later. */
1931		ret = 0;
1932		update_size = false;
1933		goto out;
1934	}
1935	btrfs_release_path(path);
1936	ret = insert_one_name(trans, root, key->objectid, key->offset,
1937			      &name, &log_key);
1938	if (ret && ret != -ENOENT && ret != -EEXIST)
1939		goto out;
1940	if (!ret)
1941		name_added = true;
1942	update_size = false;
1943	ret = 0;
1944
1945out:
1946	if (!ret && update_size) {
1947		btrfs_i_size_write(BTRFS_I(dir), dir->i_size + name.len * 2);
1948		ret = btrfs_update_inode(trans, root, BTRFS_I(dir));
1949	}
1950	kfree(name.name);
1951	iput(dir);
1952	if (!ret && name_added)
1953		ret = 1;
1954	return ret;
1955}
1956
1957/* Replay one dir item from a BTRFS_DIR_INDEX_KEY key. */
1958static noinline int replay_one_dir_item(struct btrfs_trans_handle *trans,
1959					struct btrfs_root *root,
1960					struct btrfs_path *path,
1961					struct extent_buffer *eb, int slot,
1962					struct btrfs_key *key)
1963{
1964	int ret;
1965	struct btrfs_dir_item *di;
1966
1967	/* We only log dir index keys, which only contain a single dir item. */
1968	ASSERT(key->type == BTRFS_DIR_INDEX_KEY);
1969
1970	di = btrfs_item_ptr(eb, slot, struct btrfs_dir_item);
1971	ret = replay_one_name(trans, root, path, eb, di, key);
1972	if (ret < 0)
1973		return ret;
1974
1975	/*
1976	 * If this entry refers to a non-directory (directories can not have a
1977	 * link count > 1) and it was added in the transaction that was not
1978	 * committed, make sure we fixup the link count of the inode the entry
1979	 * points to. Otherwise something like the following would result in a
1980	 * directory pointing to an inode with a wrong link that does not account
1981	 * for this dir entry:
1982	 *
1983	 * mkdir testdir
1984	 * touch testdir/foo
1985	 * touch testdir/bar
1986	 * sync
1987	 *
1988	 * ln testdir/bar testdir/bar_link
1989	 * ln testdir/foo testdir/foo_link
1990	 * xfs_io -c "fsync" testdir/bar
1991	 *
1992	 * <power failure>
1993	 *
1994	 * mount fs, log replay happens
1995	 *
1996	 * File foo would remain with a link count of 1 when it has two entries
1997	 * pointing to it in the directory testdir. This would make it impossible
1998	 * to ever delete the parent directory has it would result in stale
1999	 * dentries that can never be deleted.
2000	 */
2001	if (ret == 1 && btrfs_dir_ftype(eb, di) != BTRFS_FT_DIR) {
2002		struct btrfs_path *fixup_path;
2003		struct btrfs_key di_key;
2004
2005		fixup_path = btrfs_alloc_path();
2006		if (!fixup_path)
2007			return -ENOMEM;
2008
2009		btrfs_dir_item_key_to_cpu(eb, di, &di_key);
2010		ret = link_to_fixup_dir(trans, root, fixup_path, di_key.objectid);
2011		btrfs_free_path(fixup_path);
2012	}
2013
2014	return ret;
2015}
2016
2017/*
2018 * directory replay has two parts.  There are the standard directory
2019 * items in the log copied from the subvolume, and range items
2020 * created in the log while the subvolume was logged.
2021 *
2022 * The range items tell us which parts of the key space the log
2023 * is authoritative for.  During replay, if a key in the subvolume
2024 * directory is in a logged range item, but not actually in the log
2025 * that means it was deleted from the directory before the fsync
2026 * and should be removed.
2027 */
2028static noinline int find_dir_range(struct btrfs_root *root,
2029				   struct btrfs_path *path,
2030				   u64 dirid,
2031				   u64 *start_ret, u64 *end_ret)
2032{
2033	struct btrfs_key key;
2034	u64 found_end;
2035	struct btrfs_dir_log_item *item;
2036	int ret;
2037	int nritems;
2038
2039	if (*start_ret == (u64)-1)
2040		return 1;
2041
2042	key.objectid = dirid;
2043	key.type = BTRFS_DIR_LOG_INDEX_KEY;
2044	key.offset = *start_ret;
2045
2046	ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
2047	if (ret < 0)
2048		goto out;
2049	if (ret > 0) {
2050		if (path->slots[0] == 0)
2051			goto out;
2052		path->slots[0]--;
2053	}
2054	if (ret != 0)
2055		btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
2056
2057	if (key.type != BTRFS_DIR_LOG_INDEX_KEY || key.objectid != dirid) {
2058		ret = 1;
2059		goto next;
2060	}
2061	item = btrfs_item_ptr(path->nodes[0], path->slots[0],
2062			      struct btrfs_dir_log_item);
2063	found_end = btrfs_dir_log_end(path->nodes[0], item);
2064
2065	if (*start_ret >= key.offset && *start_ret <= found_end) {
2066		ret = 0;
2067		*start_ret = key.offset;
2068		*end_ret = found_end;
2069		goto out;
2070	}
2071	ret = 1;
2072next:
2073	/* check the next slot in the tree to see if it is a valid item */
2074	nritems = btrfs_header_nritems(path->nodes[0]);
2075	path->slots[0]++;
2076	if (path->slots[0] >= nritems) {
2077		ret = btrfs_next_leaf(root, path);
2078		if (ret)
2079			goto out;
2080	}
2081
2082	btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
2083
2084	if (key.type != BTRFS_DIR_LOG_INDEX_KEY || key.objectid != dirid) {
2085		ret = 1;
2086		goto out;
2087	}
2088	item = btrfs_item_ptr(path->nodes[0], path->slots[0],
2089			      struct btrfs_dir_log_item);
2090	found_end = btrfs_dir_log_end(path->nodes[0], item);
2091	*start_ret = key.offset;
2092	*end_ret = found_end;
2093	ret = 0;
2094out:
2095	btrfs_release_path(path);
2096	return ret;
2097}
2098
2099/*
2100 * this looks for a given directory item in the log.  If the directory
2101 * item is not in the log, the item is removed and the inode it points
2102 * to is unlinked
2103 */
2104static noinline int check_item_in_log(struct btrfs_trans_handle *trans,
2105				      struct btrfs_root *log,
2106				      struct btrfs_path *path,
2107				      struct btrfs_path *log_path,
2108				      struct inode *dir,
2109				      struct btrfs_key *dir_key)
2110{
2111	struct btrfs_root *root = BTRFS_I(dir)->root;
2112	int ret;
2113	struct extent_buffer *eb;
2114	int slot;
2115	struct btrfs_dir_item *di;
2116	struct fscrypt_str name;
2117	struct inode *inode = NULL;
2118	struct btrfs_key location;
2119
2120	/*
2121	 * Currently we only log dir index keys. Even if we replay a log created
2122	 * by an older kernel that logged both dir index and dir item keys, all
2123	 * we need to do is process the dir index keys, we (and our caller) can
2124	 * safely ignore dir item keys (key type BTRFS_DIR_ITEM_KEY).
2125	 */
2126	ASSERT(dir_key->type == BTRFS_DIR_INDEX_KEY);
2127
2128	eb = path->nodes[0];
2129	slot = path->slots[0];
2130	di = btrfs_item_ptr(eb, slot, struct btrfs_dir_item);
2131	ret = read_alloc_one_name(eb, di + 1, btrfs_dir_name_len(eb, di), &name);
2132	if (ret)
2133		goto out;
2134
2135	if (log) {
2136		struct btrfs_dir_item *log_di;
2137
2138		log_di = btrfs_lookup_dir_index_item(trans, log, log_path,
2139						     dir_key->objectid,
2140						     dir_key->offset, &name, 0);
2141		if (IS_ERR(log_di)) {
2142			ret = PTR_ERR(log_di);
2143			goto out;
2144		} else if (log_di) {
2145			/* The dentry exists in the log, we have nothing to do. */
2146			ret = 0;
2147			goto out;
2148		}
2149	}
2150
2151	btrfs_dir_item_key_to_cpu(eb, di, &location);
2152	btrfs_release_path(path);
2153	btrfs_release_path(log_path);
2154	inode = read_one_inode(root, location.objectid);
2155	if (!inode) {
2156		ret = -EIO;
2157		goto out;
2158	}
2159
2160	ret = link_to_fixup_dir(trans, root, path, location.objectid);
2161	if (ret)
2162		goto out;
2163
2164	inc_nlink(inode);
2165	ret = unlink_inode_for_log_replay(trans, BTRFS_I(dir), BTRFS_I(inode),
2166					  &name);
2167	/*
2168	 * Unlike dir item keys, dir index keys can only have one name (entry) in
2169	 * them, as there are no key collisions since each key has a unique offset
2170	 * (an index number), so we're done.
2171	 */
2172out:
2173	btrfs_release_path(path);
2174	btrfs_release_path(log_path);
2175	kfree(name.name);
2176	iput(inode);
2177	return ret;
2178}
2179
2180static int replay_xattr_deletes(struct btrfs_trans_handle *trans,
2181			      struct btrfs_root *root,
2182			      struct btrfs_root *log,
2183			      struct btrfs_path *path,
2184			      const u64 ino)
2185{
2186	struct btrfs_key search_key;
2187	struct btrfs_path *log_path;
2188	int i;
2189	int nritems;
2190	int ret;
2191
2192	log_path = btrfs_alloc_path();
2193	if (!log_path)
2194		return -ENOMEM;
2195
2196	search_key.objectid = ino;
2197	search_key.type = BTRFS_XATTR_ITEM_KEY;
2198	search_key.offset = 0;
2199again:
2200	ret = btrfs_search_slot(NULL, root, &search_key, path, 0, 0);
2201	if (ret < 0)
2202		goto out;
2203process_leaf:
2204	nritems = btrfs_header_nritems(path->nodes[0]);
2205	for (i = path->slots[0]; i < nritems; i++) {
2206		struct btrfs_key key;
2207		struct btrfs_dir_item *di;
2208		struct btrfs_dir_item *log_di;
2209		u32 total_size;
2210		u32 cur;
2211
2212		btrfs_item_key_to_cpu(path->nodes[0], &key, i);
2213		if (key.objectid != ino || key.type != BTRFS_XATTR_ITEM_KEY) {
2214			ret = 0;
2215			goto out;
2216		}
2217
2218		di = btrfs_item_ptr(path->nodes[0], i, struct btrfs_dir_item);
2219		total_size = btrfs_item_size(path->nodes[0], i);
2220		cur = 0;
2221		while (cur < total_size) {
2222			u16 name_len = btrfs_dir_name_len(path->nodes[0], di);
2223			u16 data_len = btrfs_dir_data_len(path->nodes[0], di);
2224			u32 this_len = sizeof(*di) + name_len + data_len;
2225			char *name;
2226
2227			name = kmalloc(name_len, GFP_NOFS);
2228			if (!name) {
2229				ret = -ENOMEM;
2230				goto out;
2231			}
2232			read_extent_buffer(path->nodes[0], name,
2233					   (unsigned long)(di + 1), name_len);
2234
2235			log_di = btrfs_lookup_xattr(NULL, log, log_path, ino,
2236						    name, name_len, 0);
2237			btrfs_release_path(log_path);
2238			if (!log_di) {
2239				/* Doesn't exist in log tree, so delete it. */
2240				btrfs_release_path(path);
2241				di = btrfs_lookup_xattr(trans, root, path, ino,
2242							name, name_len, -1);
2243				kfree(name);
2244				if (IS_ERR(di)) {
2245					ret = PTR_ERR(di);
2246					goto out;
2247				}
2248				ASSERT(di);
2249				ret = btrfs_delete_one_dir_name(trans, root,
2250								path, di);
2251				if (ret)
2252					goto out;
2253				btrfs_release_path(path);
2254				search_key = key;
2255				goto again;
2256			}
2257			kfree(name);
2258			if (IS_ERR(log_di)) {
2259				ret = PTR_ERR(log_di);
2260				goto out;
2261			}
2262			cur += this_len;
2263			di = (struct btrfs_dir_item *)((char *)di + this_len);
2264		}
2265	}
2266	ret = btrfs_next_leaf(root, path);
2267	if (ret > 0)
2268		ret = 0;
2269	else if (ret == 0)
2270		goto process_leaf;
2271out:
2272	btrfs_free_path(log_path);
2273	btrfs_release_path(path);
2274	return ret;
2275}
2276
2277
2278/*
2279 * deletion replay happens before we copy any new directory items
2280 * out of the log or out of backreferences from inodes.  It
2281 * scans the log to find ranges of keys that log is authoritative for,
2282 * and then scans the directory to find items in those ranges that are
2283 * not present in the log.
2284 *
2285 * Anything we don't find in the log is unlinked and removed from the
2286 * directory.
2287 */
2288static noinline int replay_dir_deletes(struct btrfs_trans_handle *trans,
2289				       struct btrfs_root *root,
2290				       struct btrfs_root *log,
2291				       struct btrfs_path *path,
2292				       u64 dirid, int del_all)
2293{
2294	u64 range_start;
2295	u64 range_end;
2296	int ret = 0;
2297	struct btrfs_key dir_key;
2298	struct btrfs_key found_key;
2299	struct btrfs_path *log_path;
2300	struct inode *dir;
2301
2302	dir_key.objectid = dirid;
2303	dir_key.type = BTRFS_DIR_INDEX_KEY;
2304	log_path = btrfs_alloc_path();
2305	if (!log_path)
2306		return -ENOMEM;
2307
2308	dir = read_one_inode(root, dirid);
2309	/* it isn't an error if the inode isn't there, that can happen
2310	 * because we replay the deletes before we copy in the inode item
2311	 * from the log
2312	 */
2313	if (!dir) {
2314		btrfs_free_path(log_path);
2315		return 0;
2316	}
2317
2318	range_start = 0;
2319	range_end = 0;
2320	while (1) {
2321		if (del_all)
2322			range_end = (u64)-1;
2323		else {
2324			ret = find_dir_range(log, path, dirid,
2325					     &range_start, &range_end);
2326			if (ret < 0)
2327				goto out;
2328			else if (ret > 0)
2329				break;
2330		}
2331
2332		dir_key.offset = range_start;
2333		while (1) {
2334			int nritems;
2335			ret = btrfs_search_slot(NULL, root, &dir_key, path,
2336						0, 0);
2337			if (ret < 0)
2338				goto out;
2339
2340			nritems = btrfs_header_nritems(path->nodes[0]);
2341			if (path->slots[0] >= nritems) {
2342				ret = btrfs_next_leaf(root, path);
2343				if (ret == 1)
2344					break;
2345				else if (ret < 0)
2346					goto out;
2347			}
2348			btrfs_item_key_to_cpu(path->nodes[0], &found_key,
2349					      path->slots[0]);
2350			if (found_key.objectid != dirid ||
2351			    found_key.type != dir_key.type) {
2352				ret = 0;
2353				goto out;
2354			}
2355
2356			if (found_key.offset > range_end)
2357				break;
2358
2359			ret = check_item_in_log(trans, log, path,
2360						log_path, dir,
2361						&found_key);
2362			if (ret)
2363				goto out;
2364			if (found_key.offset == (u64)-1)
2365				break;
2366			dir_key.offset = found_key.offset + 1;
2367		}
2368		btrfs_release_path(path);
2369		if (range_end == (u64)-1)
2370			break;
2371		range_start = range_end + 1;
2372	}
2373	ret = 0;
2374out:
2375	btrfs_release_path(path);
2376	btrfs_free_path(log_path);
2377	iput(dir);
2378	return ret;
2379}
2380
2381/*
2382 * the process_func used to replay items from the log tree.  This
2383 * gets called in two different stages.  The first stage just looks
2384 * for inodes and makes sure they are all copied into the subvolume.
2385 *
2386 * The second stage copies all the other item types from the log into
2387 * the subvolume.  The two stage approach is slower, but gets rid of
2388 * lots of complexity around inodes referencing other inodes that exist
2389 * only in the log (references come from either directory items or inode
2390 * back refs).
2391 */
2392static int replay_one_buffer(struct btrfs_root *log, struct extent_buffer *eb,
2393			     struct walk_control *wc, u64 gen, int level)
2394{
2395	int nritems;
2396	struct btrfs_tree_parent_check check = {
2397		.transid = gen,
2398		.level = level
2399	};
2400	struct btrfs_path *path;
2401	struct btrfs_root *root = wc->replay_dest;
2402	struct btrfs_key key;
2403	int i;
2404	int ret;
2405
2406	ret = btrfs_read_extent_buffer(eb, &check);
2407	if (ret)
2408		return ret;
2409
2410	level = btrfs_header_level(eb);
2411
2412	if (level != 0)
2413		return 0;
2414
2415	path = btrfs_alloc_path();
2416	if (!path)
2417		return -ENOMEM;
2418
2419	nritems = btrfs_header_nritems(eb);
2420	for (i = 0; i < nritems; i++) {
2421		btrfs_item_key_to_cpu(eb, &key, i);
2422
2423		/* inode keys are done during the first stage */
2424		if (key.type == BTRFS_INODE_ITEM_KEY &&
2425		    wc->stage == LOG_WALK_REPLAY_INODES) {
2426			struct btrfs_inode_item *inode_item;
2427			u32 mode;
2428
2429			inode_item = btrfs_item_ptr(eb, i,
2430					    struct btrfs_inode_item);
2431			/*
2432			 * If we have a tmpfile (O_TMPFILE) that got fsync'ed
2433			 * and never got linked before the fsync, skip it, as
2434			 * replaying it is pointless since it would be deleted
2435			 * later. We skip logging tmpfiles, but it's always
2436			 * possible we are replaying a log created with a kernel
2437			 * that used to log tmpfiles.
2438			 */
2439			if (btrfs_inode_nlink(eb, inode_item) == 0) {
2440				wc->ignore_cur_inode = true;
2441				continue;
2442			} else {
2443				wc->ignore_cur_inode = false;
2444			}
2445			ret = replay_xattr_deletes(wc->trans, root, log,
2446						   path, key.objectid);
2447			if (ret)
2448				break;
2449			mode = btrfs_inode_mode(eb, inode_item);
2450			if (S_ISDIR(mode)) {
2451				ret = replay_dir_deletes(wc->trans,
2452					 root, log, path, key.objectid, 0);
2453				if (ret)
2454					break;
2455			}
2456			ret = overwrite_item(wc->trans, root, path,
2457					     eb, i, &key);
2458			if (ret)
2459				break;
2460
2461			/*
2462			 * Before replaying extents, truncate the inode to its
2463			 * size. We need to do it now and not after log replay
2464			 * because before an fsync we can have prealloc extents
2465			 * added beyond the inode's i_size. If we did it after,
2466			 * through orphan cleanup for example, we would drop
2467			 * those prealloc extents just after replaying them.
2468			 */
2469			if (S_ISREG(mode)) {
2470				struct btrfs_drop_extents_args drop_args = { 0 };
2471				struct inode *inode;
2472				u64 from;
2473
2474				inode = read_one_inode(root, key.objectid);
2475				if (!inode) {
2476					ret = -EIO;
2477					break;
2478				}
2479				from = ALIGN(i_size_read(inode),
2480					     root->fs_info->sectorsize);
2481				drop_args.start = from;
2482				drop_args.end = (u64)-1;
2483				drop_args.drop_cache = true;
2484				ret = btrfs_drop_extents(wc->trans, root,
2485							 BTRFS_I(inode),
2486							 &drop_args);
2487				if (!ret) {
2488					inode_sub_bytes(inode,
2489							drop_args.bytes_found);
2490					/* Update the inode's nbytes. */
2491					ret = btrfs_update_inode(wc->trans,
2492							root, BTRFS_I(inode));
2493				}
2494				iput(inode);
2495				if (ret)
2496					break;
2497			}
2498
2499			ret = link_to_fixup_dir(wc->trans, root,
2500						path, key.objectid);
2501			if (ret)
2502				break;
2503		}
2504
2505		if (wc->ignore_cur_inode)
2506			continue;
2507
2508		if (key.type == BTRFS_DIR_INDEX_KEY &&
2509		    wc->stage == LOG_WALK_REPLAY_DIR_INDEX) {
2510			ret = replay_one_dir_item(wc->trans, root, path,
2511						  eb, i, &key);
2512			if (ret)
2513				break;
2514		}
2515
2516		if (wc->stage < LOG_WALK_REPLAY_ALL)
2517			continue;
2518
2519		/* these keys are simply copied */
2520		if (key.type == BTRFS_XATTR_ITEM_KEY) {
2521			ret = overwrite_item(wc->trans, root, path,
2522					     eb, i, &key);
2523			if (ret)
2524				break;
2525		} else if (key.type == BTRFS_INODE_REF_KEY ||
2526			   key.type == BTRFS_INODE_EXTREF_KEY) {
2527			ret = add_inode_ref(wc->trans, root, log, path,
2528					    eb, i, &key);
2529			if (ret && ret != -ENOENT)
2530				break;
2531			ret = 0;
2532		} else if (key.type == BTRFS_EXTENT_DATA_KEY) {
2533			ret = replay_one_extent(wc->trans, root, path,
2534						eb, i, &key);
2535			if (ret)
2536				break;
2537		}
2538		/*
2539		 * We don't log BTRFS_DIR_ITEM_KEY keys anymore, only the
2540		 * BTRFS_DIR_INDEX_KEY items which we use to derive the
2541		 * BTRFS_DIR_ITEM_KEY items. If we are replaying a log from an
2542		 * older kernel with such keys, ignore them.
2543		 */
2544	}
2545	btrfs_free_path(path);
2546	return ret;
2547}
2548
2549/*
2550 * Correctly adjust the reserved bytes occupied by a log tree extent buffer
2551 */
2552static void unaccount_log_buffer(struct btrfs_fs_info *fs_info, u64 start)
2553{
2554	struct btrfs_block_group *cache;
2555
2556	cache = btrfs_lookup_block_group(fs_info, start);
2557	if (!cache) {
2558		btrfs_err(fs_info, "unable to find block group for %llu", start);
2559		return;
2560	}
2561
2562	spin_lock(&cache->space_info->lock);
2563	spin_lock(&cache->lock);
2564	cache->reserved -= fs_info->nodesize;
2565	cache->space_info->bytes_reserved -= fs_info->nodesize;
2566	spin_unlock(&cache->lock);
2567	spin_unlock(&cache->space_info->lock);
2568
2569	btrfs_put_block_group(cache);
2570}
2571
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
2572static noinline int walk_down_log_tree(struct btrfs_trans_handle *trans,
2573				   struct btrfs_root *root,
2574				   struct btrfs_path *path, int *level,
2575				   struct walk_control *wc)
2576{
2577	struct btrfs_fs_info *fs_info = root->fs_info;
2578	u64 bytenr;
2579	u64 ptr_gen;
2580	struct extent_buffer *next;
2581	struct extent_buffer *cur;
2582	u32 blocksize;
2583	int ret = 0;
2584
2585	while (*level > 0) {
2586		struct btrfs_tree_parent_check check = { 0 };
2587
2588		cur = path->nodes[*level];
2589
2590		WARN_ON(btrfs_header_level(cur) != *level);
2591
2592		if (path->slots[*level] >=
2593		    btrfs_header_nritems(cur))
2594			break;
2595
2596		bytenr = btrfs_node_blockptr(cur, path->slots[*level]);
2597		ptr_gen = btrfs_node_ptr_generation(cur, path->slots[*level]);
2598		check.transid = ptr_gen;
2599		check.level = *level - 1;
2600		check.has_first_key = true;
2601		btrfs_node_key_to_cpu(cur, &check.first_key, path->slots[*level]);
2602		blocksize = fs_info->nodesize;
2603
2604		next = btrfs_find_create_tree_block(fs_info, bytenr,
2605						    btrfs_header_owner(cur),
2606						    *level - 1);
2607		if (IS_ERR(next))
2608			return PTR_ERR(next);
2609
2610		if (*level == 1) {
2611			ret = wc->process_func(root, next, wc, ptr_gen,
2612					       *level - 1);
2613			if (ret) {
2614				free_extent_buffer(next);
2615				return ret;
2616			}
2617
2618			path->slots[*level]++;
2619			if (wc->free) {
2620				ret = btrfs_read_extent_buffer(next, &check);
2621				if (ret) {
2622					free_extent_buffer(next);
2623					return ret;
2624				}
2625
2626				if (trans) {
2627					btrfs_tree_lock(next);
2628					btrfs_clean_tree_block(next);
2629					btrfs_wait_tree_block_writeback(next);
2630					btrfs_tree_unlock(next);
2631					ret = btrfs_pin_reserved_extent(trans,
2632							bytenr, blocksize);
2633					if (ret) {
2634						free_extent_buffer(next);
2635						return ret;
2636					}
2637					btrfs_redirty_list_add(
2638						trans->transaction, next);
2639				} else {
2640					if (test_and_clear_bit(EXTENT_BUFFER_DIRTY, &next->bflags))
2641						clear_extent_buffer_dirty(next);
2642					unaccount_log_buffer(fs_info, bytenr);
2643				}
2644			}
2645			free_extent_buffer(next);
2646			continue;
2647		}
2648		ret = btrfs_read_extent_buffer(next, &check);
2649		if (ret) {
2650			free_extent_buffer(next);
2651			return ret;
2652		}
2653
2654		if (path->nodes[*level-1])
2655			free_extent_buffer(path->nodes[*level-1]);
2656		path->nodes[*level-1] = next;
2657		*level = btrfs_header_level(next);
2658		path->slots[*level] = 0;
2659		cond_resched();
2660	}
2661	path->slots[*level] = btrfs_header_nritems(path->nodes[*level]);
2662
2663	cond_resched();
2664	return 0;
2665}
2666
2667static noinline int walk_up_log_tree(struct btrfs_trans_handle *trans,
2668				 struct btrfs_root *root,
2669				 struct btrfs_path *path, int *level,
2670				 struct walk_control *wc)
2671{
2672	struct btrfs_fs_info *fs_info = root->fs_info;
2673	int i;
2674	int slot;
2675	int ret;
2676
2677	for (i = *level; i < BTRFS_MAX_LEVEL - 1 && path->nodes[i]; i++) {
2678		slot = path->slots[i];
2679		if (slot + 1 < btrfs_header_nritems(path->nodes[i])) {
2680			path->slots[i]++;
2681			*level = i;
2682			WARN_ON(*level == 0);
2683			return 0;
2684		} else {
2685			ret = wc->process_func(root, path->nodes[*level], wc,
2686				 btrfs_header_generation(path->nodes[*level]),
2687				 *level);
2688			if (ret)
2689				return ret;
2690
2691			if (wc->free) {
2692				struct extent_buffer *next;
2693
2694				next = path->nodes[*level];
2695
2696				if (trans) {
2697					btrfs_tree_lock(next);
2698					btrfs_clean_tree_block(next);
2699					btrfs_wait_tree_block_writeback(next);
2700					btrfs_tree_unlock(next);
2701					ret = btrfs_pin_reserved_extent(trans,
2702						     path->nodes[*level]->start,
2703						     path->nodes[*level]->len);
2704					if (ret)
2705						return ret;
2706					btrfs_redirty_list_add(trans->transaction,
2707							       next);
2708				} else {
2709					if (test_and_clear_bit(EXTENT_BUFFER_DIRTY, &next->bflags))
2710						clear_extent_buffer_dirty(next);
2711
2712					unaccount_log_buffer(fs_info,
2713						path->nodes[*level]->start);
2714				}
2715			}
2716			free_extent_buffer(path->nodes[*level]);
2717			path->nodes[*level] = NULL;
2718			*level = i + 1;
2719		}
2720	}
2721	return 1;
2722}
2723
2724/*
2725 * drop the reference count on the tree rooted at 'snap'.  This traverses
2726 * the tree freeing any blocks that have a ref count of zero after being
2727 * decremented.
2728 */
2729static int walk_log_tree(struct btrfs_trans_handle *trans,
2730			 struct btrfs_root *log, struct walk_control *wc)
2731{
2732	struct btrfs_fs_info *fs_info = log->fs_info;
2733	int ret = 0;
2734	int wret;
2735	int level;
2736	struct btrfs_path *path;
2737	int orig_level;
2738
2739	path = btrfs_alloc_path();
2740	if (!path)
2741		return -ENOMEM;
2742
2743	level = btrfs_header_level(log->node);
2744	orig_level = level;
2745	path->nodes[level] = log->node;
2746	atomic_inc(&log->node->refs);
2747	path->slots[level] = 0;
2748
2749	while (1) {
2750		wret = walk_down_log_tree(trans, log, path, &level, wc);
2751		if (wret > 0)
2752			break;
2753		if (wret < 0) {
2754			ret = wret;
2755			goto out;
2756		}
2757
2758		wret = walk_up_log_tree(trans, log, path, &level, wc);
2759		if (wret > 0)
2760			break;
2761		if (wret < 0) {
2762			ret = wret;
2763			goto out;
2764		}
2765	}
2766
2767	/* was the root node processed? if not, catch it here */
2768	if (path->nodes[orig_level]) {
2769		ret = wc->process_func(log, path->nodes[orig_level], wc,
2770			 btrfs_header_generation(path->nodes[orig_level]),
2771			 orig_level);
2772		if (ret)
2773			goto out;
2774		if (wc->free) {
2775			struct extent_buffer *next;
2776
2777			next = path->nodes[orig_level];
2778
2779			if (trans) {
2780				btrfs_tree_lock(next);
2781				btrfs_clean_tree_block(next);
2782				btrfs_wait_tree_block_writeback(next);
2783				btrfs_tree_unlock(next);
2784				ret = btrfs_pin_reserved_extent(trans,
2785						next->start, next->len);
2786				if (ret)
2787					goto out;
2788				btrfs_redirty_list_add(trans->transaction, next);
2789			} else {
2790				if (test_and_clear_bit(EXTENT_BUFFER_DIRTY, &next->bflags))
2791					clear_extent_buffer_dirty(next);
2792				unaccount_log_buffer(fs_info, next->start);
2793			}
2794		}
2795	}
2796
2797out:
2798	btrfs_free_path(path);
2799	return ret;
2800}
2801
2802/*
2803 * helper function to update the item for a given subvolumes log root
2804 * in the tree of log roots
2805 */
2806static int update_log_root(struct btrfs_trans_handle *trans,
2807			   struct btrfs_root *log,
2808			   struct btrfs_root_item *root_item)
2809{
2810	struct btrfs_fs_info *fs_info = log->fs_info;
2811	int ret;
2812
2813	if (log->log_transid == 1) {
2814		/* insert root item on the first sync */
2815		ret = btrfs_insert_root(trans, fs_info->log_root_tree,
2816				&log->root_key, root_item);
2817	} else {
2818		ret = btrfs_update_root(trans, fs_info->log_root_tree,
2819				&log->root_key, root_item);
2820	}
2821	return ret;
2822}
2823
2824static void wait_log_commit(struct btrfs_root *root, int transid)
2825{
2826	DEFINE_WAIT(wait);
2827	int index = transid % 2;
2828
2829	/*
2830	 * we only allow two pending log transactions at a time,
2831	 * so we know that if ours is more than 2 older than the
2832	 * current transaction, we're done
2833	 */
2834	for (;;) {
2835		prepare_to_wait(&root->log_commit_wait[index],
2836				&wait, TASK_UNINTERRUPTIBLE);
2837
2838		if (!(root->log_transid_committed < transid &&
2839		      atomic_read(&root->log_commit[index])))
2840			break;
2841
2842		mutex_unlock(&root->log_mutex);
2843		schedule();
2844		mutex_lock(&root->log_mutex);
2845	}
2846	finish_wait(&root->log_commit_wait[index], &wait);
2847}
2848
2849static void wait_for_writer(struct btrfs_root *root)
2850{
2851	DEFINE_WAIT(wait);
2852
2853	for (;;) {
2854		prepare_to_wait(&root->log_writer_wait, &wait,
2855				TASK_UNINTERRUPTIBLE);
2856		if (!atomic_read(&root->log_writers))
2857			break;
2858
2859		mutex_unlock(&root->log_mutex);
2860		schedule();
2861		mutex_lock(&root->log_mutex);
2862	}
2863	finish_wait(&root->log_writer_wait, &wait);
2864}
2865
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
2866static inline void btrfs_remove_log_ctx(struct btrfs_root *root,
2867					struct btrfs_log_ctx *ctx)
2868{
2869	mutex_lock(&root->log_mutex);
2870	list_del_init(&ctx->list);
2871	mutex_unlock(&root->log_mutex);
2872}
2873
2874/* 
2875 * Invoked in log mutex context, or be sure there is no other task which
2876 * can access the list.
2877 */
2878static inline void btrfs_remove_all_log_ctxs(struct btrfs_root *root,
2879					     int index, int error)
2880{
2881	struct btrfs_log_ctx *ctx;
2882	struct btrfs_log_ctx *safe;
2883
2884	list_for_each_entry_safe(ctx, safe, &root->log_ctxs[index], list) {
2885		list_del_init(&ctx->list);
2886		ctx->log_ret = error;
2887	}
2888}
2889
2890/*
2891 * btrfs_sync_log does sends a given tree log down to the disk and
2892 * updates the super blocks to record it.  When this call is done,
2893 * you know that any inodes previously logged are safely on disk only
2894 * if it returns 0.
2895 *
2896 * Any other return value means you need to call btrfs_commit_transaction.
2897 * Some of the edge cases for fsyncing directories that have had unlinks
2898 * or renames done in the past mean that sometimes the only safe
2899 * fsync is to commit the whole FS.  When btrfs_sync_log returns -EAGAIN,
2900 * that has happened.
2901 */
2902int btrfs_sync_log(struct btrfs_trans_handle *trans,
2903		   struct btrfs_root *root, struct btrfs_log_ctx *ctx)
2904{
2905	int index1;
2906	int index2;
2907	int mark;
2908	int ret;
2909	struct btrfs_fs_info *fs_info = root->fs_info;
2910	struct btrfs_root *log = root->log_root;
2911	struct btrfs_root *log_root_tree = fs_info->log_root_tree;
2912	struct btrfs_root_item new_root_item;
2913	int log_transid = 0;
2914	struct btrfs_log_ctx root_log_ctx;
2915	struct blk_plug plug;
2916	u64 log_root_start;
2917	u64 log_root_level;
2918
2919	mutex_lock(&root->log_mutex);
2920	log_transid = ctx->log_transid;
2921	if (root->log_transid_committed >= log_transid) {
2922		mutex_unlock(&root->log_mutex);
2923		return ctx->log_ret;
2924	}
2925
2926	index1 = log_transid % 2;
2927	if (atomic_read(&root->log_commit[index1])) {
2928		wait_log_commit(root, log_transid);
2929		mutex_unlock(&root->log_mutex);
2930		return ctx->log_ret;
2931	}
2932	ASSERT(log_transid == root->log_transid);
2933	atomic_set(&root->log_commit[index1], 1);
2934
2935	/* wait for previous tree log sync to complete */
2936	if (atomic_read(&root->log_commit[(index1 + 1) % 2]))
2937		wait_log_commit(root, log_transid - 1);
2938
2939	while (1) {
2940		int batch = atomic_read(&root->log_batch);
2941		/* when we're on an ssd, just kick the log commit out */
2942		if (!btrfs_test_opt(fs_info, SSD) &&
2943		    test_bit(BTRFS_ROOT_MULTI_LOG_TASKS, &root->state)) {
2944			mutex_unlock(&root->log_mutex);
2945			schedule_timeout_uninterruptible(1);
2946			mutex_lock(&root->log_mutex);
2947		}
2948		wait_for_writer(root);
2949		if (batch == atomic_read(&root->log_batch))
2950			break;
2951	}
2952
2953	/* bail out if we need to do a full commit */
2954	if (btrfs_need_log_full_commit(trans)) {
2955		ret = BTRFS_LOG_FORCE_COMMIT;
2956		mutex_unlock(&root->log_mutex);
2957		goto out;
2958	}
2959
2960	if (log_transid % 2 == 0)
2961		mark = EXTENT_DIRTY;
2962	else
2963		mark = EXTENT_NEW;
2964
2965	/* we start IO on  all the marked extents here, but we don't actually
2966	 * wait for them until later.
2967	 */
2968	blk_start_plug(&plug);
2969	ret = btrfs_write_marked_extents(fs_info, &log->dirty_log_pages, mark);
2970	/*
2971	 * -EAGAIN happens when someone, e.g., a concurrent transaction
2972	 *  commit, writes a dirty extent in this tree-log commit. This
2973	 *  concurrent write will create a hole writing out the extents,
2974	 *  and we cannot proceed on a zoned filesystem, requiring
2975	 *  sequential writing. While we can bail out to a full commit
2976	 *  here, but we can continue hoping the concurrent writing fills
2977	 *  the hole.
2978	 */
2979	if (ret == -EAGAIN && btrfs_is_zoned(fs_info))
2980		ret = 0;
2981	if (ret) {
2982		blk_finish_plug(&plug);
2983		btrfs_set_log_full_commit(trans);
2984		mutex_unlock(&root->log_mutex);
2985		goto out;
2986	}
2987
2988	/*
2989	 * We _must_ update under the root->log_mutex in order to make sure we
2990	 * have a consistent view of the log root we are trying to commit at
2991	 * this moment.
2992	 *
2993	 * We _must_ copy this into a local copy, because we are not holding the
2994	 * log_root_tree->log_mutex yet.  This is important because when we
2995	 * commit the log_root_tree we must have a consistent view of the
2996	 * log_root_tree when we update the super block to point at the
2997	 * log_root_tree bytenr.  If we update the log_root_tree here we'll race
2998	 * with the commit and possibly point at the new block which we may not
2999	 * have written out.
3000	 */
3001	btrfs_set_root_node(&log->root_item, log->node);
3002	memcpy(&new_root_item, &log->root_item, sizeof(new_root_item));
3003
3004	root->log_transid++;
3005	log->log_transid = root->log_transid;
3006	root->log_start_pid = 0;
3007	/*
3008	 * IO has been started, blocks of the log tree have WRITTEN flag set
3009	 * in their headers. new modifications of the log will be written to
3010	 * new positions. so it's safe to allow log writers to go in.
3011	 */
3012	mutex_unlock(&root->log_mutex);
3013
3014	if (btrfs_is_zoned(fs_info)) {
3015		mutex_lock(&fs_info->tree_root->log_mutex);
3016		if (!log_root_tree->node) {
3017			ret = btrfs_alloc_log_tree_node(trans, log_root_tree);
3018			if (ret) {
3019				mutex_unlock(&fs_info->tree_root->log_mutex);
3020				blk_finish_plug(&plug);
3021				goto out;
3022			}
3023		}
3024		mutex_unlock(&fs_info->tree_root->log_mutex);
3025	}
3026
3027	btrfs_init_log_ctx(&root_log_ctx, NULL);
3028
3029	mutex_lock(&log_root_tree->log_mutex);
3030
3031	index2 = log_root_tree->log_transid % 2;
3032	list_add_tail(&root_log_ctx.list, &log_root_tree->log_ctxs[index2]);
3033	root_log_ctx.log_transid = log_root_tree->log_transid;
3034
3035	/*
3036	 * Now we are safe to update the log_root_tree because we're under the
3037	 * log_mutex, and we're a current writer so we're holding the commit
3038	 * open until we drop the log_mutex.
3039	 */
3040	ret = update_log_root(trans, log, &new_root_item);
3041	if (ret) {
3042		if (!list_empty(&root_log_ctx.list))
3043			list_del_init(&root_log_ctx.list);
3044
3045		blk_finish_plug(&plug);
3046		btrfs_set_log_full_commit(trans);
3047		if (ret != -ENOSPC)
3048			btrfs_err(fs_info,
3049				  "failed to update log for root %llu ret %d",
3050				  root->root_key.objectid, ret);
3051		btrfs_wait_tree_log_extents(log, mark);
3052		mutex_unlock(&log_root_tree->log_mutex);
3053		goto out;
3054	}
3055
3056	if (log_root_tree->log_transid_committed >= root_log_ctx.log_transid) {
3057		blk_finish_plug(&plug);
3058		list_del_init(&root_log_ctx.list);
3059		mutex_unlock(&log_root_tree->log_mutex);
3060		ret = root_log_ctx.log_ret;
3061		goto out;
3062	}
3063
3064	index2 = root_log_ctx.log_transid % 2;
3065	if (atomic_read(&log_root_tree->log_commit[index2])) {
3066		blk_finish_plug(&plug);
3067		ret = btrfs_wait_tree_log_extents(log, mark);
3068		wait_log_commit(log_root_tree,
3069				root_log_ctx.log_transid);
3070		mutex_unlock(&log_root_tree->log_mutex);
3071		if (!ret)
3072			ret = root_log_ctx.log_ret;
3073		goto out;
3074	}
3075	ASSERT(root_log_ctx.log_transid == log_root_tree->log_transid);
3076	atomic_set(&log_root_tree->log_commit[index2], 1);
3077
3078	if (atomic_read(&log_root_tree->log_commit[(index2 + 1) % 2])) {
3079		wait_log_commit(log_root_tree,
3080				root_log_ctx.log_transid - 1);
3081	}
3082
3083	/*
3084	 * now that we've moved on to the tree of log tree roots,
3085	 * check the full commit flag again
3086	 */
3087	if (btrfs_need_log_full_commit(trans)) {
3088		blk_finish_plug(&plug);
3089		btrfs_wait_tree_log_extents(log, mark);
3090		mutex_unlock(&log_root_tree->log_mutex);
3091		ret = BTRFS_LOG_FORCE_COMMIT;
3092		goto out_wake_log_root;
3093	}
3094
3095	ret = btrfs_write_marked_extents(fs_info,
3096					 &log_root_tree->dirty_log_pages,
3097					 EXTENT_DIRTY | EXTENT_NEW);
3098	blk_finish_plug(&plug);
3099	/*
3100	 * As described above, -EAGAIN indicates a hole in the extents. We
3101	 * cannot wait for these write outs since the waiting cause a
3102	 * deadlock. Bail out to the full commit instead.
3103	 */
3104	if (ret == -EAGAIN && btrfs_is_zoned(fs_info)) {
3105		btrfs_set_log_full_commit(trans);
3106		btrfs_wait_tree_log_extents(log, mark);
3107		mutex_unlock(&log_root_tree->log_mutex);
3108		goto out_wake_log_root;
3109	} else if (ret) {
3110		btrfs_set_log_full_commit(trans);
3111		mutex_unlock(&log_root_tree->log_mutex);
3112		goto out_wake_log_root;
3113	}
3114	ret = btrfs_wait_tree_log_extents(log, mark);
3115	if (!ret)
3116		ret = btrfs_wait_tree_log_extents(log_root_tree,
3117						  EXTENT_NEW | EXTENT_DIRTY);
3118	if (ret) {
3119		btrfs_set_log_full_commit(trans);
3120		mutex_unlock(&log_root_tree->log_mutex);
3121		goto out_wake_log_root;
3122	}
3123
3124	log_root_start = log_root_tree->node->start;
3125	log_root_level = btrfs_header_level(log_root_tree->node);
3126	log_root_tree->log_transid++;
3127	mutex_unlock(&log_root_tree->log_mutex);
3128
3129	/*
3130	 * Here we are guaranteed that nobody is going to write the superblock
3131	 * for the current transaction before us and that neither we do write
3132	 * our superblock before the previous transaction finishes its commit
3133	 * and writes its superblock, because:
3134	 *
3135	 * 1) We are holding a handle on the current transaction, so no body
3136	 *    can commit it until we release the handle;
3137	 *
3138	 * 2) Before writing our superblock we acquire the tree_log_mutex, so
3139	 *    if the previous transaction is still committing, and hasn't yet
3140	 *    written its superblock, we wait for it to do it, because a
3141	 *    transaction commit acquires the tree_log_mutex when the commit
3142	 *    begins and releases it only after writing its superblock.
3143	 */
3144	mutex_lock(&fs_info->tree_log_mutex);
3145
3146	/*
3147	 * The previous transaction writeout phase could have failed, and thus
3148	 * marked the fs in an error state.  We must not commit here, as we
3149	 * could have updated our generation in the super_for_commit and
3150	 * writing the super here would result in transid mismatches.  If there
3151	 * is an error here just bail.
3152	 */
3153	if (BTRFS_FS_ERROR(fs_info)) {
3154		ret = -EIO;
3155		btrfs_set_log_full_commit(trans);
3156		btrfs_abort_transaction(trans, ret);
3157		mutex_unlock(&fs_info->tree_log_mutex);
3158		goto out_wake_log_root;
3159	}
3160
3161	btrfs_set_super_log_root(fs_info->super_for_commit, log_root_start);
3162	btrfs_set_super_log_root_level(fs_info->super_for_commit, log_root_level);
3163	ret = write_all_supers(fs_info, 1);
3164	mutex_unlock(&fs_info->tree_log_mutex);
3165	if (ret) {
3166		btrfs_set_log_full_commit(trans);
3167		btrfs_abort_transaction(trans, ret);
3168		goto out_wake_log_root;
3169	}
3170
3171	/*
3172	 * We know there can only be one task here, since we have not yet set
3173	 * root->log_commit[index1] to 0 and any task attempting to sync the
3174	 * log must wait for the previous log transaction to commit if it's
3175	 * still in progress or wait for the current log transaction commit if
3176	 * someone else already started it. We use <= and not < because the
3177	 * first log transaction has an ID of 0.
3178	 */
3179	ASSERT(root->last_log_commit <= log_transid);
3180	root->last_log_commit = log_transid;
3181
3182out_wake_log_root:
3183	mutex_lock(&log_root_tree->log_mutex);
3184	btrfs_remove_all_log_ctxs(log_root_tree, index2, ret);
3185
3186	log_root_tree->log_transid_committed++;
3187	atomic_set(&log_root_tree->log_commit[index2], 0);
3188	mutex_unlock(&log_root_tree->log_mutex);
3189
3190	/*
3191	 * The barrier before waitqueue_active (in cond_wake_up) is needed so
3192	 * all the updates above are seen by the woken threads. It might not be
3193	 * necessary, but proving that seems to be hard.
3194	 */
3195	cond_wake_up(&log_root_tree->log_commit_wait[index2]);
3196out:
3197	mutex_lock(&root->log_mutex);
3198	btrfs_remove_all_log_ctxs(root, index1, ret);
3199	root->log_transid_committed++;
3200	atomic_set(&root->log_commit[index1], 0);
3201	mutex_unlock(&root->log_mutex);
3202
3203	/*
3204	 * The barrier before waitqueue_active (in cond_wake_up) is needed so
3205	 * all the updates above are seen by the woken threads. It might not be
3206	 * necessary, but proving that seems to be hard.
3207	 */
3208	cond_wake_up(&root->log_commit_wait[index1]);
3209	return ret;
3210}
3211
3212static void free_log_tree(struct btrfs_trans_handle *trans,
3213			  struct btrfs_root *log)
3214{
3215	int ret;
3216	struct walk_control wc = {
3217		.free = 1,
3218		.process_func = process_one_buffer
3219	};
3220
3221	if (log->node) {
3222		ret = walk_log_tree(trans, log, &wc);
3223		if (ret) {
3224			/*
3225			 * We weren't able to traverse the entire log tree, the
3226			 * typical scenario is getting an -EIO when reading an
3227			 * extent buffer of the tree, due to a previous writeback
3228			 * failure of it.
3229			 */
3230			set_bit(BTRFS_FS_STATE_LOG_CLEANUP_ERROR,
3231				&log->fs_info->fs_state);
3232
3233			/*
3234			 * Some extent buffers of the log tree may still be dirty
3235			 * and not yet written back to storage, because we may
3236			 * have updates to a log tree without syncing a log tree,
3237			 * such as during rename and link operations. So flush
3238			 * them out and wait for their writeback to complete, so
3239			 * that we properly cleanup their state and pages.
3240			 */
3241			btrfs_write_marked_extents(log->fs_info,
3242						   &log->dirty_log_pages,
3243						   EXTENT_DIRTY | EXTENT_NEW);
3244			btrfs_wait_tree_log_extents(log,
3245						    EXTENT_DIRTY | EXTENT_NEW);
3246
3247			if (trans)
3248				btrfs_abort_transaction(trans, ret);
3249			else
3250				btrfs_handle_fs_error(log->fs_info, ret, NULL);
3251		}
3252	}
3253
3254	clear_extent_bits(&log->dirty_log_pages, 0, (u64)-1,
3255			  EXTENT_DIRTY | EXTENT_NEW | EXTENT_NEED_WAIT);
3256	extent_io_tree_release(&log->log_csum_range);
3257
3258	btrfs_put_root(log);
3259}
3260
3261/*
3262 * free all the extents used by the tree log.  This should be called
3263 * at commit time of the full transaction
3264 */
3265int btrfs_free_log(struct btrfs_trans_handle *trans, struct btrfs_root *root)
3266{
3267	if (root->log_root) {
3268		free_log_tree(trans, root->log_root);
3269		root->log_root = NULL;
3270		clear_bit(BTRFS_ROOT_HAS_LOG_TREE, &root->state);
3271	}
3272	return 0;
3273}
3274
3275int btrfs_free_log_root_tree(struct btrfs_trans_handle *trans,
3276			     struct btrfs_fs_info *fs_info)
3277{
3278	if (fs_info->log_root_tree) {
3279		free_log_tree(trans, fs_info->log_root_tree);
3280		fs_info->log_root_tree = NULL;
3281		clear_bit(BTRFS_ROOT_HAS_LOG_TREE, &fs_info->tree_root->state);
3282	}
3283	return 0;
3284}
3285
3286/*
3287 * Check if an inode was logged in the current transaction. This correctly deals
3288 * with the case where the inode was logged but has a logged_trans of 0, which
3289 * happens if the inode is evicted and loaded again, as logged_trans is an in
3290 * memory only field (not persisted).
3291 *
3292 * Returns 1 if the inode was logged before in the transaction, 0 if it was not,
3293 * and < 0 on error.
3294 */
3295static int inode_logged(struct btrfs_trans_handle *trans,
3296			struct btrfs_inode *inode,
3297			struct btrfs_path *path_in)
3298{
3299	struct btrfs_path *path = path_in;
3300	struct btrfs_key key;
3301	int ret;
3302
3303	if (inode->logged_trans == trans->transid)
3304		return 1;
3305
3306	/*
3307	 * If logged_trans is not 0, then we know the inode logged was not logged
3308	 * in this transaction, so we can return false right away.
3309	 */
3310	if (inode->logged_trans > 0)
3311		return 0;
3312
3313	/*
3314	 * If no log tree was created for this root in this transaction, then
3315	 * the inode can not have been logged in this transaction. In that case
3316	 * set logged_trans to anything greater than 0 and less than the current
3317	 * transaction's ID, to avoid the search below in a future call in case
3318	 * a log tree gets created after this.
3319	 */
3320	if (!test_bit(BTRFS_ROOT_HAS_LOG_TREE, &inode->root->state)) {
3321		inode->logged_trans = trans->transid - 1;
3322		return 0;
3323	}
3324
3325	/*
3326	 * We have a log tree and the inode's logged_trans is 0. We can't tell
3327	 * for sure if the inode was logged before in this transaction by looking
3328	 * only at logged_trans. We could be pessimistic and assume it was, but
3329	 * that can lead to unnecessarily logging an inode during rename and link
3330	 * operations, and then further updating the log in followup rename and
3331	 * link operations, specially if it's a directory, which adds latency
3332	 * visible to applications doing a series of rename or link operations.
3333	 *
3334	 * A logged_trans of 0 here can mean several things:
3335	 *
3336	 * 1) The inode was never logged since the filesystem was mounted, and may
3337	 *    or may have not been evicted and loaded again;
3338	 *
3339	 * 2) The inode was logged in a previous transaction, then evicted and
3340	 *    then loaded again;
3341	 *
3342	 * 3) The inode was logged in the current transaction, then evicted and
3343	 *    then loaded again.
3344	 *
3345	 * For cases 1) and 2) we don't want to return true, but we need to detect
3346	 * case 3) and return true. So we do a search in the log root for the inode
3347	 * item.
3348	 */
3349	key.objectid = btrfs_ino(inode);
3350	key.type = BTRFS_INODE_ITEM_KEY;
3351	key.offset = 0;
3352
3353	if (!path) {
3354		path = btrfs_alloc_path();
3355		if (!path)
3356			return -ENOMEM;
3357	}
3358
3359	ret = btrfs_search_slot(NULL, inode->root->log_root, &key, path, 0, 0);
3360
3361	if (path_in)
3362		btrfs_release_path(path);
3363	else
3364		btrfs_free_path(path);
3365
3366	/*
3367	 * Logging an inode always results in logging its inode item. So if we
3368	 * did not find the item we know the inode was not logged for sure.
3369	 */
3370	if (ret < 0) {
3371		return ret;
3372	} else if (ret > 0) {
3373		/*
3374		 * Set logged_trans to a value greater than 0 and less then the
3375		 * current transaction to avoid doing the search in future calls.
3376		 */
3377		inode->logged_trans = trans->transid - 1;
3378		return 0;
3379	}
3380
3381	/*
3382	 * The inode was previously logged and then evicted, set logged_trans to
3383	 * the current transacion's ID, to avoid future tree searches as long as
3384	 * the inode is not evicted again.
3385	 */
3386	inode->logged_trans = trans->transid;
3387
3388	/*
3389	 * If it's a directory, then we must set last_dir_index_offset to the
3390	 * maximum possible value, so that the next attempt to log the inode does
3391	 * not skip checking if dir index keys found in modified subvolume tree
3392	 * leaves have been logged before, otherwise it would result in attempts
3393	 * to insert duplicate dir index keys in the log tree. This must be done
3394	 * because last_dir_index_offset is an in-memory only field, not persisted
3395	 * in the inode item or any other on-disk structure, so its value is lost
3396	 * once the inode is evicted.
3397	 */
3398	if (S_ISDIR(inode->vfs_inode.i_mode))
3399		inode->last_dir_index_offset = (u64)-1;
3400
3401	return 1;
3402}
3403
3404/*
3405 * Delete a directory entry from the log if it exists.
3406 *
3407 * Returns < 0 on error
3408 *           1 if the entry does not exists
3409 *           0 if the entry existed and was successfully deleted
3410 */
3411static int del_logged_dentry(struct btrfs_trans_handle *trans,
3412			     struct btrfs_root *log,
3413			     struct btrfs_path *path,
3414			     u64 dir_ino,
3415			     const struct fscrypt_str *name,
3416			     u64 index)
3417{
3418	struct btrfs_dir_item *di;
3419
3420	/*
3421	 * We only log dir index items of a directory, so we don't need to look
3422	 * for dir item keys.
3423	 */
3424	di = btrfs_lookup_dir_index_item(trans, log, path, dir_ino,
3425					 index, name, -1);
3426	if (IS_ERR(di))
3427		return PTR_ERR(di);
3428	else if (!di)
3429		return 1;
3430
3431	/*
3432	 * We do not need to update the size field of the directory's
3433	 * inode item because on log replay we update the field to reflect
3434	 * all existing entries in the directory (see overwrite_item()).
3435	 */
3436	return btrfs_delete_one_dir_name(trans, log, path, di);
3437}
3438
3439/*
3440 * If both a file and directory are logged, and unlinks or renames are
3441 * mixed in, we have a few interesting corners:
3442 *
3443 * create file X in dir Y
3444 * link file X to X.link in dir Y
3445 * fsync file X
3446 * unlink file X but leave X.link
3447 * fsync dir Y
3448 *
3449 * After a crash we would expect only X.link to exist.  But file X
3450 * didn't get fsync'd again so the log has back refs for X and X.link.
3451 *
3452 * We solve this by removing directory entries and inode backrefs from the
3453 * log when a file that was logged in the current transaction is
3454 * unlinked.  Any later fsync will include the updated log entries, and
3455 * we'll be able to reconstruct the proper directory items from backrefs.
3456 *
3457 * This optimizations allows us to avoid relogging the entire inode
3458 * or the entire directory.
3459 */
3460void btrfs_del_dir_entries_in_log(struct btrfs_trans_handle *trans,
3461				  struct btrfs_root *root,
3462				  const struct fscrypt_str *name,
3463				  struct btrfs_inode *dir, u64 index)
3464{
3465	struct btrfs_path *path;
3466	int ret;
3467
3468	ret = inode_logged(trans, dir, NULL);
3469	if (ret == 0)
3470		return;
3471	else if (ret < 0) {
3472		btrfs_set_log_full_commit(trans);
3473		return;
3474	}
3475
3476	ret = join_running_log_trans(root);
3477	if (ret)
3478		return;
3479
3480	mutex_lock(&dir->log_mutex);
3481
3482	path = btrfs_alloc_path();
3483	if (!path) {
3484		ret = -ENOMEM;
3485		goto out_unlock;
3486	}
3487
3488	ret = del_logged_dentry(trans, root->log_root, path, btrfs_ino(dir),
3489				name, index);
3490	btrfs_free_path(path);
3491out_unlock:
3492	mutex_unlock(&dir->log_mutex);
3493	if (ret < 0)
3494		btrfs_set_log_full_commit(trans);
3495	btrfs_end_log_trans(root);
3496}
3497
3498/* see comments for btrfs_del_dir_entries_in_log */
3499void btrfs_del_inode_ref_in_log(struct btrfs_trans_handle *trans,
3500				struct btrfs_root *root,
3501				const struct fscrypt_str *name,
3502				struct btrfs_inode *inode, u64 dirid)
3503{
3504	struct btrfs_root *log;
3505	u64 index;
3506	int ret;
3507
3508	ret = inode_logged(trans, inode, NULL);
3509	if (ret == 0)
3510		return;
3511	else if (ret < 0) {
3512		btrfs_set_log_full_commit(trans);
3513		return;
3514	}
3515
3516	ret = join_running_log_trans(root);
3517	if (ret)
3518		return;
3519	log = root->log_root;
3520	mutex_lock(&inode->log_mutex);
3521
3522	ret = btrfs_del_inode_ref(trans, log, name, btrfs_ino(inode),
3523				  dirid, &index);
3524	mutex_unlock(&inode->log_mutex);
3525	if (ret < 0 && ret != -ENOENT)
3526		btrfs_set_log_full_commit(trans);
3527	btrfs_end_log_trans(root);
3528}
3529
3530/*
3531 * creates a range item in the log for 'dirid'.  first_offset and
3532 * last_offset tell us which parts of the key space the log should
3533 * be considered authoritative for.
3534 */
3535static noinline int insert_dir_log_key(struct btrfs_trans_handle *trans,
3536				       struct btrfs_root *log,
3537				       struct btrfs_path *path,
3538				       u64 dirid,
3539				       u64 first_offset, u64 last_offset)
3540{
3541	int ret;
3542	struct btrfs_key key;
3543	struct btrfs_dir_log_item *item;
3544
3545	key.objectid = dirid;
3546	key.offset = first_offset;
3547	key.type = BTRFS_DIR_LOG_INDEX_KEY;
3548	ret = btrfs_insert_empty_item(trans, log, path, &key, sizeof(*item));
3549	/*
3550	 * -EEXIST is fine and can happen sporadically when we are logging a
3551	 * directory and have concurrent insertions in the subvolume's tree for
3552	 * items from other inodes and that result in pushing off some dir items
3553	 * from one leaf to another in order to accommodate for the new items.
3554	 * This results in logging the same dir index range key.
3555	 */
3556	if (ret && ret != -EEXIST)
3557		return ret;
3558
3559	item = btrfs_item_ptr(path->nodes[0], path->slots[0],
3560			      struct btrfs_dir_log_item);
3561	if (ret == -EEXIST) {
3562		const u64 curr_end = btrfs_dir_log_end(path->nodes[0], item);
3563
3564		/*
3565		 * btrfs_del_dir_entries_in_log() might have been called during
3566		 * an unlink between the initial insertion of this key and the
3567		 * current update, or we might be logging a single entry deletion
3568		 * during a rename, so set the new last_offset to the max value.
3569		 */
3570		last_offset = max(last_offset, curr_end);
3571	}
3572	btrfs_set_dir_log_end(path->nodes[0], item, last_offset);
3573	btrfs_mark_buffer_dirty(path->nodes[0]);
3574	btrfs_release_path(path);
3575	return 0;
3576}
3577
3578static int flush_dir_items_batch(struct btrfs_trans_handle *trans,
3579				 struct btrfs_inode *inode,
3580				 struct extent_buffer *src,
3581				 struct btrfs_path *dst_path,
3582				 int start_slot,
3583				 int count)
3584{
3585	struct btrfs_root *log = inode->root->log_root;
3586	char *ins_data = NULL;
3587	struct btrfs_item_batch batch;
3588	struct extent_buffer *dst;
3589	unsigned long src_offset;
3590	unsigned long dst_offset;
3591	u64 last_index;
3592	struct btrfs_key key;
3593	u32 item_size;
3594	int ret;
3595	int i;
3596
3597	ASSERT(count > 0);
3598	batch.nr = count;
3599
3600	if (count == 1) {
3601		btrfs_item_key_to_cpu(src, &key, start_slot);
3602		item_size = btrfs_item_size(src, start_slot);
3603		batch.keys = &key;
3604		batch.data_sizes = &item_size;
3605		batch.total_data_size = item_size;
3606	} else {
3607		struct btrfs_key *ins_keys;
3608		u32 *ins_sizes;
3609
3610		ins_data = kmalloc(count * sizeof(u32) +
3611				   count * sizeof(struct btrfs_key), GFP_NOFS);
3612		if (!ins_data)
3613			return -ENOMEM;
3614
3615		ins_sizes = (u32 *)ins_data;
3616		ins_keys = (struct btrfs_key *)(ins_data + count * sizeof(u32));
3617		batch.keys = ins_keys;
3618		batch.data_sizes = ins_sizes;
3619		batch.total_data_size = 0;
3620
3621		for (i = 0; i < count; i++) {
3622			const int slot = start_slot + i;
3623
3624			btrfs_item_key_to_cpu(src, &ins_keys[i], slot);
3625			ins_sizes[i] = btrfs_item_size(src, slot);
3626			batch.total_data_size += ins_sizes[i];
3627		}
3628	}
3629
3630	ret = btrfs_insert_empty_items(trans, log, dst_path, &batch);
3631	if (ret)
3632		goto out;
3633
3634	dst = dst_path->nodes[0];
3635	/*
3636	 * Copy all the items in bulk, in a single copy operation. Item data is
3637	 * organized such that it's placed at the end of a leaf and from right
3638	 * to left. For example, the data for the second item ends at an offset
3639	 * that matches the offset where the data for the first item starts, the
3640	 * data for the third item ends at an offset that matches the offset
3641	 * where the data of the second items starts, and so on.
3642	 * Therefore our source and destination start offsets for copy match the
3643	 * offsets of the last items (highest slots).
3644	 */
3645	dst_offset = btrfs_item_ptr_offset(dst, dst_path->slots[0] + count - 1);
3646	src_offset = btrfs_item_ptr_offset(src, start_slot + count - 1);
3647	copy_extent_buffer(dst, src, dst_offset, src_offset, batch.total_data_size);
3648	btrfs_release_path(dst_path);
3649
3650	last_index = batch.keys[count - 1].offset;
3651	ASSERT(last_index > inode->last_dir_index_offset);
3652
3653	/*
3654	 * If for some unexpected reason the last item's index is not greater
3655	 * than the last index we logged, warn and return an error to fallback
3656	 * to a transaction commit.
3657	 */
3658	if (WARN_ON(last_index <= inode->last_dir_index_offset))
3659		ret = -EUCLEAN;
3660	else
3661		inode->last_dir_index_offset = last_index;
 
 
 
3662out:
3663	kfree(ins_data);
3664
3665	return ret;
3666}
3667
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
3668static int process_dir_items_leaf(struct btrfs_trans_handle *trans,
3669				  struct btrfs_inode *inode,
3670				  struct btrfs_path *path,
3671				  struct btrfs_path *dst_path,
3672				  struct btrfs_log_ctx *ctx,
3673				  u64 *last_old_dentry_offset)
3674{
3675	struct btrfs_root *log = inode->root->log_root;
3676	struct extent_buffer *src;
3677	const int nritems = btrfs_header_nritems(path->nodes[0]);
3678	const u64 ino = btrfs_ino(inode);
3679	bool last_found = false;
3680	int batch_start = 0;
3681	int batch_size = 0;
3682	int i;
3683
3684	/*
3685	 * We need to clone the leaf, release the read lock on it, and use the
3686	 * clone before modifying the log tree. See the comment at copy_items()
3687	 * about why we need to do this.
3688	 */
3689	src = btrfs_clone_extent_buffer(path->nodes[0]);
3690	if (!src)
3691		return -ENOMEM;
3692
3693	i = path->slots[0];
3694	btrfs_release_path(path);
3695	path->nodes[0] = src;
3696	path->slots[0] = i;
3697
3698	for (; i < nritems; i++) {
3699		struct btrfs_dir_item *di;
3700		struct btrfs_key key;
3701		int ret;
3702
3703		btrfs_item_key_to_cpu(src, &key, i);
3704
3705		if (key.objectid != ino || key.type != BTRFS_DIR_INDEX_KEY) {
3706			last_found = true;
3707			break;
3708		}
3709
3710		di = btrfs_item_ptr(src, i, struct btrfs_dir_item);
3711
3712		/*
3713		 * Skip ranges of items that consist only of dir item keys created
3714		 * in past transactions. However if we find a gap, we must log a
3715		 * dir index range item for that gap, so that index keys in that
3716		 * gap are deleted during log replay.
3717		 */
3718		if (btrfs_dir_transid(src, di) < trans->transid) {
3719			if (key.offset > *last_old_dentry_offset + 1) {
3720				ret = insert_dir_log_key(trans, log, dst_path,
3721						 ino, *last_old_dentry_offset + 1,
3722						 key.offset - 1);
3723				if (ret < 0)
3724					return ret;
3725			}
3726
3727			*last_old_dentry_offset = key.offset;
3728			continue;
3729		}
3730
3731		/* If we logged this dir index item before, we can skip it. */
3732		if (key.offset <= inode->last_dir_index_offset)
3733			continue;
3734
3735		/*
3736		 * We must make sure that when we log a directory entry, the
3737		 * corresponding inode, after log replay, has a matching link
3738		 * count. For example:
3739		 *
3740		 * touch foo
3741		 * mkdir mydir
3742		 * sync
3743		 * ln foo mydir/bar
3744		 * xfs_io -c "fsync" mydir
3745		 * <crash>
3746		 * <mount fs and log replay>
3747		 *
3748		 * Would result in a fsync log that when replayed, our file inode
3749		 * would have a link count of 1, but we get two directory entries
3750		 * pointing to the same inode. After removing one of the names,
3751		 * it would not be possible to remove the other name, which
3752		 * resulted always in stale file handle errors, and would not be
3753		 * possible to rmdir the parent directory, since its i_size could
3754		 * never be decremented to the value BTRFS_EMPTY_DIR_SIZE,
3755		 * resulting in -ENOTEMPTY errors.
3756		 */
3757		if (!ctx->log_new_dentries) {
3758			struct btrfs_key di_key;
3759
3760			btrfs_dir_item_key_to_cpu(src, di, &di_key);
3761			if (di_key.type != BTRFS_ROOT_ITEM_KEY)
3762				ctx->log_new_dentries = true;
3763		}
3764
3765		if (batch_size == 0)
3766			batch_start = i;
3767		batch_size++;
3768	}
3769
3770	if (batch_size > 0) {
3771		int ret;
3772
3773		ret = flush_dir_items_batch(trans, inode, src, dst_path,
3774					    batch_start, batch_size);
3775		if (ret < 0)
3776			return ret;
3777	}
3778
3779	return last_found ? 1 : 0;
3780}
3781
3782/*
3783 * log all the items included in the current transaction for a given
3784 * directory.  This also creates the range items in the log tree required
3785 * to replay anything deleted before the fsync
3786 */
3787static noinline int log_dir_items(struct btrfs_trans_handle *trans,
3788			  struct btrfs_inode *inode,
3789			  struct btrfs_path *path,
3790			  struct btrfs_path *dst_path,
3791			  struct btrfs_log_ctx *ctx,
3792			  u64 min_offset, u64 *last_offset_ret)
3793{
3794	struct btrfs_key min_key;
3795	struct btrfs_root *root = inode->root;
3796	struct btrfs_root *log = root->log_root;
3797	int err = 0;
3798	int ret;
3799	u64 last_old_dentry_offset = min_offset - 1;
3800	u64 last_offset = (u64)-1;
3801	u64 ino = btrfs_ino(inode);
3802
3803	min_key.objectid = ino;
3804	min_key.type = BTRFS_DIR_INDEX_KEY;
3805	min_key.offset = min_offset;
3806
3807	ret = btrfs_search_forward(root, &min_key, path, trans->transid);
3808
3809	/*
3810	 * we didn't find anything from this transaction, see if there
3811	 * is anything at all
3812	 */
3813	if (ret != 0 || min_key.objectid != ino ||
3814	    min_key.type != BTRFS_DIR_INDEX_KEY) {
3815		min_key.objectid = ino;
3816		min_key.type = BTRFS_DIR_INDEX_KEY;
3817		min_key.offset = (u64)-1;
3818		btrfs_release_path(path);
3819		ret = btrfs_search_slot(NULL, root, &min_key, path, 0, 0);
3820		if (ret < 0) {
3821			btrfs_release_path(path);
3822			return ret;
3823		}
3824		ret = btrfs_previous_item(root, path, ino, BTRFS_DIR_INDEX_KEY);
3825
3826		/* if ret == 0 there are items for this type,
3827		 * create a range to tell us the last key of this type.
3828		 * otherwise, there are no items in this directory after
3829		 * *min_offset, and we create a range to indicate that.
3830		 */
3831		if (ret == 0) {
3832			struct btrfs_key tmp;
3833
3834			btrfs_item_key_to_cpu(path->nodes[0], &tmp,
3835					      path->slots[0]);
3836			if (tmp.type == BTRFS_DIR_INDEX_KEY)
3837				last_old_dentry_offset = tmp.offset;
3838		} else if (ret < 0) {
3839			err = ret;
3840		}
3841
3842		goto done;
3843	}
3844
3845	/* go backward to find any previous key */
3846	ret = btrfs_previous_item(root, path, ino, BTRFS_DIR_INDEX_KEY);
3847	if (ret == 0) {
3848		struct btrfs_key tmp;
3849
3850		btrfs_item_key_to_cpu(path->nodes[0], &tmp, path->slots[0]);
3851		/*
3852		 * The dir index key before the first one we found that needs to
3853		 * be logged might be in a previous leaf, and there might be a
3854		 * gap between these keys, meaning that we had deletions that
3855		 * happened. So the key range item we log (key type
3856		 * BTRFS_DIR_LOG_INDEX_KEY) must cover a range that starts at the
3857		 * previous key's offset plus 1, so that those deletes are replayed.
3858		 */
3859		if (tmp.type == BTRFS_DIR_INDEX_KEY)
3860			last_old_dentry_offset = tmp.offset;
3861	} else if (ret < 0) {
3862		err = ret;
3863		goto done;
3864	}
3865
3866	btrfs_release_path(path);
3867
3868	/*
3869	 * Find the first key from this transaction again or the one we were at
3870	 * in the loop below in case we had to reschedule. We may be logging the
3871	 * directory without holding its VFS lock, which happen when logging new
3872	 * dentries (through log_new_dir_dentries()) or in some cases when we
3873	 * need to log the parent directory of an inode. This means a dir index
3874	 * key might be deleted from the inode's root, and therefore we may not
3875	 * find it anymore. If we can't find it, just move to the next key. We
3876	 * can not bail out and ignore, because if we do that we will simply
3877	 * not log dir index keys that come after the one that was just deleted
3878	 * and we can end up logging a dir index range that ends at (u64)-1
3879	 * (@last_offset is initialized to that), resulting in removing dir
3880	 * entries we should not remove at log replay time.
3881	 */
3882search:
3883	ret = btrfs_search_slot(NULL, root, &min_key, path, 0, 0);
3884	if (ret > 0)
3885		ret = btrfs_next_item(root, path);
 
 
 
 
 
 
3886	if (ret < 0)
3887		err = ret;
3888	/* If ret is 1, there are no more keys in the inode's root. */
3889	if (ret != 0)
3890		goto done;
3891
3892	/*
3893	 * we have a block from this transaction, log every item in it
3894	 * from our directory
3895	 */
3896	while (1) {
3897		ret = process_dir_items_leaf(trans, inode, path, dst_path, ctx,
3898					     &last_old_dentry_offset);
3899		if (ret != 0) {
3900			if (ret < 0)
3901				err = ret;
3902			goto done;
3903		}
3904		path->slots[0] = btrfs_header_nritems(path->nodes[0]);
3905
3906		/*
3907		 * look ahead to the next item and see if it is also
3908		 * from this directory and from this transaction
3909		 */
3910		ret = btrfs_next_leaf(root, path);
3911		if (ret) {
3912			if (ret == 1)
3913				last_offset = (u64)-1;
3914			else
3915				err = ret;
3916			goto done;
3917		}
3918		btrfs_item_key_to_cpu(path->nodes[0], &min_key, path->slots[0]);
3919		if (min_key.objectid != ino || min_key.type != BTRFS_DIR_INDEX_KEY) {
3920			last_offset = (u64)-1;
3921			goto done;
3922		}
3923		if (btrfs_header_generation(path->nodes[0]) != trans->transid) {
3924			/*
3925			 * The next leaf was not changed in the current transaction
3926			 * and has at least one dir index key.
3927			 * We check for the next key because there might have been
3928			 * one or more deletions between the last key we logged and
3929			 * that next key. So the key range item we log (key type
3930			 * BTRFS_DIR_LOG_INDEX_KEY) must end at the next key's
3931			 * offset minus 1, so that those deletes are replayed.
3932			 */
3933			last_offset = min_key.offset - 1;
3934			goto done;
3935		}
3936		if (need_resched()) {
3937			btrfs_release_path(path);
3938			cond_resched();
3939			goto search;
3940		}
3941	}
3942done:
3943	btrfs_release_path(path);
3944	btrfs_release_path(dst_path);
3945
3946	if (err == 0) {
3947		*last_offset_ret = last_offset;
3948		/*
3949		 * In case the leaf was changed in the current transaction but
3950		 * all its dir items are from a past transaction, the last item
3951		 * in the leaf is a dir item and there's no gap between that last
3952		 * dir item and the first one on the next leaf (which did not
3953		 * change in the current transaction), then we don't need to log
3954		 * a range, last_old_dentry_offset is == to last_offset.
3955		 */
3956		ASSERT(last_old_dentry_offset <= last_offset);
3957		if (last_old_dentry_offset < last_offset) {
3958			ret = insert_dir_log_key(trans, log, path, ino,
3959						 last_old_dentry_offset + 1,
3960						 last_offset);
3961			if (ret)
3962				err = ret;
3963		}
3964	}
3965	return err;
 
3966}
3967
3968/*
3969 * If the inode was logged before and it was evicted, then its
3970 * last_dir_index_offset is (u64)-1, so we don't the value of the last index
3971 * key offset. If that's the case, search for it and update the inode. This
3972 * is to avoid lookups in the log tree every time we try to insert a dir index
3973 * key from a leaf changed in the current transaction, and to allow us to always
3974 * do batch insertions of dir index keys.
3975 */
3976static int update_last_dir_index_offset(struct btrfs_inode *inode,
3977					struct btrfs_path *path,
3978					const struct btrfs_log_ctx *ctx)
3979{
3980	const u64 ino = btrfs_ino(inode);
3981	struct btrfs_key key;
3982	int ret;
3983
3984	lockdep_assert_held(&inode->log_mutex);
3985
3986	if (inode->last_dir_index_offset != (u64)-1)
3987		return 0;
3988
3989	if (!ctx->logged_before) {
3990		inode->last_dir_index_offset = BTRFS_DIR_START_INDEX - 1;
3991		return 0;
3992	}
3993
3994	key.objectid = ino;
3995	key.type = BTRFS_DIR_INDEX_KEY;
3996	key.offset = (u64)-1;
3997
3998	ret = btrfs_search_slot(NULL, inode->root->log_root, &key, path, 0, 0);
3999	/*
4000	 * An error happened or we actually have an index key with an offset
4001	 * value of (u64)-1. Bail out, we're done.
4002	 */
4003	if (ret <= 0)
4004		goto out;
4005
4006	ret = 0;
4007	inode->last_dir_index_offset = BTRFS_DIR_START_INDEX - 1;
4008
4009	/*
4010	 * No dir index items, bail out and leave last_dir_index_offset with
4011	 * the value right before the first valid index value.
4012	 */
4013	if (path->slots[0] == 0)
4014		goto out;
4015
4016	/*
4017	 * btrfs_search_slot() left us at one slot beyond the slot with the last
4018	 * index key, or beyond the last key of the directory that is not an
4019	 * index key. If we have an index key before, set last_dir_index_offset
4020	 * to its offset value, otherwise leave it with a value right before the
4021	 * first valid index value, as it means we have an empty directory.
4022	 */
4023	btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0] - 1);
4024	if (key.objectid == ino && key.type == BTRFS_DIR_INDEX_KEY)
4025		inode->last_dir_index_offset = key.offset;
4026
4027out:
4028	btrfs_release_path(path);
4029
4030	return ret;
4031}
4032
4033/*
4034 * logging directories is very similar to logging inodes, We find all the items
4035 * from the current transaction and write them to the log.
4036 *
4037 * The recovery code scans the directory in the subvolume, and if it finds a
4038 * key in the range logged that is not present in the log tree, then it means
4039 * that dir entry was unlinked during the transaction.
4040 *
4041 * In order for that scan to work, we must include one key smaller than
4042 * the smallest logged by this transaction and one key larger than the largest
4043 * key logged by this transaction.
4044 */
4045static noinline int log_directory_changes(struct btrfs_trans_handle *trans,
4046			  struct btrfs_inode *inode,
4047			  struct btrfs_path *path,
4048			  struct btrfs_path *dst_path,
4049			  struct btrfs_log_ctx *ctx)
4050{
4051	u64 min_key;
4052	u64 max_key;
4053	int ret;
4054
4055	ret = update_last_dir_index_offset(inode, path, ctx);
4056	if (ret)
4057		return ret;
4058
4059	min_key = BTRFS_DIR_START_INDEX;
4060	max_key = 0;
4061
4062	while (1) {
4063		ret = log_dir_items(trans, inode, path, dst_path,
4064				ctx, min_key, &max_key);
4065		if (ret)
4066			return ret;
4067		if (max_key == (u64)-1)
4068			break;
4069		min_key = max_key + 1;
4070	}
4071
4072	return 0;
4073}
4074
4075/*
4076 * a helper function to drop items from the log before we relog an
4077 * inode.  max_key_type indicates the highest item type to remove.
4078 * This cannot be run for file data extents because it does not
4079 * free the extents they point to.
4080 */
4081static int drop_inode_items(struct btrfs_trans_handle *trans,
4082				  struct btrfs_root *log,
4083				  struct btrfs_path *path,
4084				  struct btrfs_inode *inode,
4085				  int max_key_type)
4086{
4087	int ret;
4088	struct btrfs_key key;
4089	struct btrfs_key found_key;
4090	int start_slot;
4091
4092	key.objectid = btrfs_ino(inode);
4093	key.type = max_key_type;
4094	key.offset = (u64)-1;
4095
4096	while (1) {
4097		ret = btrfs_search_slot(trans, log, &key, path, -1, 1);
4098		BUG_ON(ret == 0); /* Logic error */
4099		if (ret < 0)
4100			break;
4101
4102		if (path->slots[0] == 0)
4103			break;
 
 
 
 
 
4104
4105		path->slots[0]--;
4106		btrfs_item_key_to_cpu(path->nodes[0], &found_key,
4107				      path->slots[0]);
4108
4109		if (found_key.objectid != key.objectid)
4110			break;
4111
4112		found_key.offset = 0;
4113		found_key.type = 0;
4114		ret = btrfs_bin_search(path->nodes[0], &found_key, &start_slot);
4115		if (ret < 0)
4116			break;
4117
4118		ret = btrfs_del_items(trans, log, path, start_slot,
4119				      path->slots[0] - start_slot + 1);
4120		/*
4121		 * If start slot isn't 0 then we don't need to re-search, we've
4122		 * found the last guy with the objectid in this tree.
4123		 */
4124		if (ret || start_slot != 0)
4125			break;
4126		btrfs_release_path(path);
4127	}
4128	btrfs_release_path(path);
4129	if (ret > 0)
4130		ret = 0;
4131	return ret;
4132}
4133
4134static int truncate_inode_items(struct btrfs_trans_handle *trans,
4135				struct btrfs_root *log_root,
4136				struct btrfs_inode *inode,
4137				u64 new_size, u32 min_type)
4138{
4139	struct btrfs_truncate_control control = {
4140		.new_size = new_size,
4141		.ino = btrfs_ino(inode),
4142		.min_type = min_type,
4143		.skip_ref_updates = true,
4144	};
4145
4146	return btrfs_truncate_inode_items(trans, log_root, &control);
4147}
4148
4149static void fill_inode_item(struct btrfs_trans_handle *trans,
4150			    struct extent_buffer *leaf,
4151			    struct btrfs_inode_item *item,
4152			    struct inode *inode, int log_inode_only,
4153			    u64 logged_isize)
4154{
4155	struct btrfs_map_token token;
4156	u64 flags;
4157
4158	btrfs_init_map_token(&token, leaf);
4159
4160	if (log_inode_only) {
4161		/* set the generation to zero so the recover code
4162		 * can tell the difference between an logging
4163		 * just to say 'this inode exists' and a logging
4164		 * to say 'update this inode with these values'
4165		 */
4166		btrfs_set_token_inode_generation(&token, item, 0);
4167		btrfs_set_token_inode_size(&token, item, logged_isize);
4168	} else {
4169		btrfs_set_token_inode_generation(&token, item,
4170						 BTRFS_I(inode)->generation);
4171		btrfs_set_token_inode_size(&token, item, inode->i_size);
4172	}
4173
4174	btrfs_set_token_inode_uid(&token, item, i_uid_read(inode));
4175	btrfs_set_token_inode_gid(&token, item, i_gid_read(inode));
4176	btrfs_set_token_inode_mode(&token, item, inode->i_mode);
4177	btrfs_set_token_inode_nlink(&token, item, inode->i_nlink);
4178
4179	btrfs_set_token_timespec_sec(&token, &item->atime,
4180				     inode->i_atime.tv_sec);
4181	btrfs_set_token_timespec_nsec(&token, &item->atime,
4182				      inode->i_atime.tv_nsec);
4183
4184	btrfs_set_token_timespec_sec(&token, &item->mtime,
4185				     inode->i_mtime.tv_sec);
4186	btrfs_set_token_timespec_nsec(&token, &item->mtime,
4187				      inode->i_mtime.tv_nsec);
4188
4189	btrfs_set_token_timespec_sec(&token, &item->ctime,
4190				     inode->i_ctime.tv_sec);
4191	btrfs_set_token_timespec_nsec(&token, &item->ctime,
4192				      inode->i_ctime.tv_nsec);
4193
4194	/*
4195	 * We do not need to set the nbytes field, in fact during a fast fsync
4196	 * its value may not even be correct, since a fast fsync does not wait
4197	 * for ordered extent completion, which is where we update nbytes, it
4198	 * only waits for writeback to complete. During log replay as we find
4199	 * file extent items and replay them, we adjust the nbytes field of the
4200	 * inode item in subvolume tree as needed (see overwrite_item()).
4201	 */
4202
4203	btrfs_set_token_inode_sequence(&token, item, inode_peek_iversion(inode));
4204	btrfs_set_token_inode_transid(&token, item, trans->transid);
4205	btrfs_set_token_inode_rdev(&token, item, inode->i_rdev);
4206	flags = btrfs_inode_combine_flags(BTRFS_I(inode)->flags,
4207					  BTRFS_I(inode)->ro_flags);
4208	btrfs_set_token_inode_flags(&token, item, flags);
4209	btrfs_set_token_inode_block_group(&token, item, 0);
4210}
4211
4212static int log_inode_item(struct btrfs_trans_handle *trans,
4213			  struct btrfs_root *log, struct btrfs_path *path,
4214			  struct btrfs_inode *inode, bool inode_item_dropped)
4215{
4216	struct btrfs_inode_item *inode_item;
4217	int ret;
4218
4219	/*
4220	 * If we are doing a fast fsync and the inode was logged before in the
4221	 * current transaction, then we know the inode was previously logged and
4222	 * it exists in the log tree. For performance reasons, in this case use
4223	 * btrfs_search_slot() directly with ins_len set to 0 so that we never
4224	 * attempt a write lock on the leaf's parent, which adds unnecessary lock
4225	 * contention in case there are concurrent fsyncs for other inodes of the
4226	 * same subvolume. Using btrfs_insert_empty_item() when the inode item
4227	 * already exists can also result in unnecessarily splitting a leaf.
4228	 */
4229	if (!inode_item_dropped && inode->logged_trans == trans->transid) {
4230		ret = btrfs_search_slot(trans, log, &inode->location, path, 0, 1);
4231		ASSERT(ret <= 0);
4232		if (ret > 0)
4233			ret = -ENOENT;
4234	} else {
4235		/*
4236		 * This means it is the first fsync in the current transaction,
4237		 * so the inode item is not in the log and we need to insert it.
4238		 * We can never get -EEXIST because we are only called for a fast
4239		 * fsync and in case an inode eviction happens after the inode was
4240		 * logged before in the current transaction, when we load again
4241		 * the inode, we set BTRFS_INODE_NEEDS_FULL_SYNC on its runtime
4242		 * flags and set ->logged_trans to 0.
4243		 */
4244		ret = btrfs_insert_empty_item(trans, log, path, &inode->location,
4245					      sizeof(*inode_item));
4246		ASSERT(ret != -EEXIST);
4247	}
4248	if (ret)
4249		return ret;
4250	inode_item = btrfs_item_ptr(path->nodes[0], path->slots[0],
4251				    struct btrfs_inode_item);
4252	fill_inode_item(trans, path->nodes[0], inode_item, &inode->vfs_inode,
4253			0, 0);
4254	btrfs_release_path(path);
4255	return 0;
4256}
4257
4258static int log_csums(struct btrfs_trans_handle *trans,
4259		     struct btrfs_inode *inode,
4260		     struct btrfs_root *log_root,
4261		     struct btrfs_ordered_sum *sums)
4262{
4263	const u64 lock_end = sums->bytenr + sums->len - 1;
4264	struct extent_state *cached_state = NULL;
4265	int ret;
4266
4267	/*
4268	 * If this inode was not used for reflink operations in the current
4269	 * transaction with new extents, then do the fast path, no need to
4270	 * worry about logging checksum items with overlapping ranges.
4271	 */
4272	if (inode->last_reflink_trans < trans->transid)
4273		return btrfs_csum_file_blocks(trans, log_root, sums);
4274
4275	/*
4276	 * Serialize logging for checksums. This is to avoid racing with the
4277	 * same checksum being logged by another task that is logging another
4278	 * file which happens to refer to the same extent as well. Such races
4279	 * can leave checksum items in the log with overlapping ranges.
4280	 */
4281	ret = lock_extent(&log_root->log_csum_range, sums->bytenr, lock_end,
4282			  &cached_state);
4283	if (ret)
4284		return ret;
4285	/*
4286	 * Due to extent cloning, we might have logged a csum item that covers a
4287	 * subrange of a cloned extent, and later we can end up logging a csum
4288	 * item for a larger subrange of the same extent or the entire range.
4289	 * This would leave csum items in the log tree that cover the same range
4290	 * and break the searches for checksums in the log tree, resulting in
4291	 * some checksums missing in the fs/subvolume tree. So just delete (or
4292	 * trim and adjust) any existing csum items in the log for this range.
4293	 */
4294	ret = btrfs_del_csums(trans, log_root, sums->bytenr, sums->len);
4295	if (!ret)
4296		ret = btrfs_csum_file_blocks(trans, log_root, sums);
4297
4298	unlock_extent(&log_root->log_csum_range, sums->bytenr, lock_end,
4299		      &cached_state);
4300
4301	return ret;
4302}
4303
4304static noinline int copy_items(struct btrfs_trans_handle *trans,
4305			       struct btrfs_inode *inode,
4306			       struct btrfs_path *dst_path,
4307			       struct btrfs_path *src_path,
4308			       int start_slot, int nr, int inode_only,
4309			       u64 logged_isize)
4310{
4311	struct btrfs_root *log = inode->root->log_root;
4312	struct btrfs_file_extent_item *extent;
4313	struct extent_buffer *src;
4314	int ret = 0;
4315	struct btrfs_key *ins_keys;
4316	u32 *ins_sizes;
4317	struct btrfs_item_batch batch;
4318	char *ins_data;
4319	int i;
4320	int dst_index;
4321	const bool skip_csum = (inode->flags & BTRFS_INODE_NODATASUM);
4322	const u64 i_size = i_size_read(&inode->vfs_inode);
4323
4324	/*
4325	 * To keep lockdep happy and avoid deadlocks, clone the source leaf and
4326	 * use the clone. This is because otherwise we would be changing the log
4327	 * tree, to insert items from the subvolume tree or insert csum items,
4328	 * while holding a read lock on a leaf from the subvolume tree, which
4329	 * creates a nasty lock dependency when COWing log tree nodes/leaves:
4330	 *
4331	 * 1) Modifying the log tree triggers an extent buffer allocation while
4332	 *    holding a write lock on a parent extent buffer from the log tree.
4333	 *    Allocating the pages for an extent buffer, or the extent buffer
4334	 *    struct, can trigger inode eviction and finally the inode eviction
4335	 *    will trigger a release/remove of a delayed node, which requires
4336	 *    taking the delayed node's mutex;
4337	 *
4338	 * 2) Allocating a metadata extent for a log tree can trigger the async
4339	 *    reclaim thread and make us wait for it to release enough space and
4340	 *    unblock our reservation ticket. The reclaim thread can start
4341	 *    flushing delayed items, and that in turn results in the need to
4342	 *    lock delayed node mutexes and in the need to write lock extent
4343	 *    buffers of a subvolume tree - all this while holding a write lock
4344	 *    on the parent extent buffer in the log tree.
4345	 *
4346	 * So one task in scenario 1) running in parallel with another task in
4347	 * scenario 2) could lead to a deadlock, one wanting to lock a delayed
4348	 * node mutex while having a read lock on a leaf from the subvolume,
4349	 * while the other is holding the delayed node's mutex and wants to
4350	 * write lock the same subvolume leaf for flushing delayed items.
4351	 */
4352	src = btrfs_clone_extent_buffer(src_path->nodes[0]);
4353	if (!src)
4354		return -ENOMEM;
4355
4356	i = src_path->slots[0];
4357	btrfs_release_path(src_path);
4358	src_path->nodes[0] = src;
4359	src_path->slots[0] = i;
4360
4361	ins_data = kmalloc(nr * sizeof(struct btrfs_key) +
4362			   nr * sizeof(u32), GFP_NOFS);
4363	if (!ins_data)
4364		return -ENOMEM;
4365
4366	ins_sizes = (u32 *)ins_data;
4367	ins_keys = (struct btrfs_key *)(ins_data + nr * sizeof(u32));
4368	batch.keys = ins_keys;
4369	batch.data_sizes = ins_sizes;
4370	batch.total_data_size = 0;
4371	batch.nr = 0;
4372
4373	dst_index = 0;
4374	for (i = 0; i < nr; i++) {
4375		const int src_slot = start_slot + i;
4376		struct btrfs_root *csum_root;
4377		struct btrfs_ordered_sum *sums;
4378		struct btrfs_ordered_sum *sums_next;
4379		LIST_HEAD(ordered_sums);
4380		u64 disk_bytenr;
4381		u64 disk_num_bytes;
4382		u64 extent_offset;
4383		u64 extent_num_bytes;
4384		bool is_old_extent;
4385
4386		btrfs_item_key_to_cpu(src, &ins_keys[dst_index], src_slot);
4387
4388		if (ins_keys[dst_index].type != BTRFS_EXTENT_DATA_KEY)
4389			goto add_to_batch;
4390
4391		extent = btrfs_item_ptr(src, src_slot,
4392					struct btrfs_file_extent_item);
4393
4394		is_old_extent = (btrfs_file_extent_generation(src, extent) <
4395				 trans->transid);
4396
4397		/*
4398		 * Don't copy extents from past generations. That would make us
4399		 * log a lot more metadata for common cases like doing only a
4400		 * few random writes into a file and then fsync it for the first
4401		 * time or after the full sync flag is set on the inode. We can
4402		 * get leaves full of extent items, most of which are from past
4403		 * generations, so we can skip them - as long as the inode has
4404		 * not been the target of a reflink operation in this transaction,
4405		 * as in that case it might have had file extent items with old
4406		 * generations copied into it. We also must always log prealloc
4407		 * extents that start at or beyond eof, otherwise we would lose
4408		 * them on log replay.
4409		 */
4410		if (is_old_extent &&
4411		    ins_keys[dst_index].offset < i_size &&
4412		    inode->last_reflink_trans < trans->transid)
4413			continue;
4414
4415		if (skip_csum)
4416			goto add_to_batch;
4417
4418		/* Only regular extents have checksums. */
4419		if (btrfs_file_extent_type(src, extent) != BTRFS_FILE_EXTENT_REG)
4420			goto add_to_batch;
4421
4422		/*
4423		 * If it's an extent created in a past transaction, then its
4424		 * checksums are already accessible from the committed csum tree,
4425		 * no need to log them.
4426		 */
4427		if (is_old_extent)
4428			goto add_to_batch;
4429
4430		disk_bytenr = btrfs_file_extent_disk_bytenr(src, extent);
4431		/* If it's an explicit hole, there are no checksums. */
4432		if (disk_bytenr == 0)
4433			goto add_to_batch;
4434
4435		disk_num_bytes = btrfs_file_extent_disk_num_bytes(src, extent);
4436
4437		if (btrfs_file_extent_compression(src, extent)) {
4438			extent_offset = 0;
4439			extent_num_bytes = disk_num_bytes;
4440		} else {
4441			extent_offset = btrfs_file_extent_offset(src, extent);
4442			extent_num_bytes = btrfs_file_extent_num_bytes(src, extent);
4443		}
4444
4445		csum_root = btrfs_csum_root(trans->fs_info, disk_bytenr);
4446		disk_bytenr += extent_offset;
4447		ret = btrfs_lookup_csums_list(csum_root, disk_bytenr,
4448					      disk_bytenr + extent_num_bytes - 1,
4449					      &ordered_sums, 0, false);
4450		if (ret)
4451			goto out;
4452
4453		list_for_each_entry_safe(sums, sums_next, &ordered_sums, list) {
4454			if (!ret)
4455				ret = log_csums(trans, inode, log, sums);
4456			list_del(&sums->list);
4457			kfree(sums);
4458		}
4459		if (ret)
4460			goto out;
4461
4462add_to_batch:
4463		ins_sizes[dst_index] = btrfs_item_size(src, src_slot);
4464		batch.total_data_size += ins_sizes[dst_index];
4465		batch.nr++;
4466		dst_index++;
4467	}
4468
4469	/*
4470	 * We have a leaf full of old extent items that don't need to be logged,
4471	 * so we don't need to do anything.
4472	 */
4473	if (batch.nr == 0)
4474		goto out;
4475
4476	ret = btrfs_insert_empty_items(trans, log, dst_path, &batch);
4477	if (ret)
4478		goto out;
4479
4480	dst_index = 0;
4481	for (i = 0; i < nr; i++) {
4482		const int src_slot = start_slot + i;
4483		const int dst_slot = dst_path->slots[0] + dst_index;
4484		struct btrfs_key key;
4485		unsigned long src_offset;
4486		unsigned long dst_offset;
4487
4488		/*
4489		 * We're done, all the remaining items in the source leaf
4490		 * correspond to old file extent items.
4491		 */
4492		if (dst_index >= batch.nr)
4493			break;
4494
4495		btrfs_item_key_to_cpu(src, &key, src_slot);
4496
4497		if (key.type != BTRFS_EXTENT_DATA_KEY)
4498			goto copy_item;
4499
4500		extent = btrfs_item_ptr(src, src_slot,
4501					struct btrfs_file_extent_item);
4502
4503		/* See the comment in the previous loop, same logic. */
4504		if (btrfs_file_extent_generation(src, extent) < trans->transid &&
4505		    key.offset < i_size &&
4506		    inode->last_reflink_trans < trans->transid)
4507			continue;
4508
4509copy_item:
4510		dst_offset = btrfs_item_ptr_offset(dst_path->nodes[0], dst_slot);
4511		src_offset = btrfs_item_ptr_offset(src, src_slot);
4512
4513		if (key.type == BTRFS_INODE_ITEM_KEY) {
4514			struct btrfs_inode_item *inode_item;
4515
4516			inode_item = btrfs_item_ptr(dst_path->nodes[0], dst_slot,
4517						    struct btrfs_inode_item);
4518			fill_inode_item(trans, dst_path->nodes[0], inode_item,
4519					&inode->vfs_inode,
4520					inode_only == LOG_INODE_EXISTS,
4521					logged_isize);
4522		} else {
4523			copy_extent_buffer(dst_path->nodes[0], src, dst_offset,
4524					   src_offset, ins_sizes[dst_index]);
4525		}
4526
4527		dst_index++;
4528	}
4529
4530	btrfs_mark_buffer_dirty(dst_path->nodes[0]);
4531	btrfs_release_path(dst_path);
4532out:
4533	kfree(ins_data);
4534
4535	return ret;
4536}
4537
4538static int extent_cmp(void *priv, const struct list_head *a,
4539		      const struct list_head *b)
4540{
4541	const struct extent_map *em1, *em2;
4542
4543	em1 = list_entry(a, struct extent_map, list);
4544	em2 = list_entry(b, struct extent_map, list);
4545
4546	if (em1->start < em2->start)
4547		return -1;
4548	else if (em1->start > em2->start)
4549		return 1;
4550	return 0;
4551}
4552
4553static int log_extent_csums(struct btrfs_trans_handle *trans,
4554			    struct btrfs_inode *inode,
4555			    struct btrfs_root *log_root,
4556			    const struct extent_map *em,
4557			    struct btrfs_log_ctx *ctx)
4558{
4559	struct btrfs_ordered_extent *ordered;
4560	struct btrfs_root *csum_root;
4561	u64 csum_offset;
4562	u64 csum_len;
4563	u64 mod_start = em->mod_start;
4564	u64 mod_len = em->mod_len;
4565	LIST_HEAD(ordered_sums);
4566	int ret = 0;
4567
4568	if (inode->flags & BTRFS_INODE_NODATASUM ||
4569	    test_bit(EXTENT_FLAG_PREALLOC, &em->flags) ||
4570	    em->block_start == EXTENT_MAP_HOLE)
4571		return 0;
4572
4573	list_for_each_entry(ordered, &ctx->ordered_extents, log_list) {
4574		const u64 ordered_end = ordered->file_offset + ordered->num_bytes;
4575		const u64 mod_end = mod_start + mod_len;
4576		struct btrfs_ordered_sum *sums;
4577
4578		if (mod_len == 0)
4579			break;
4580
4581		if (ordered_end <= mod_start)
4582			continue;
4583		if (mod_end <= ordered->file_offset)
4584			break;
4585
4586		/*
4587		 * We are going to copy all the csums on this ordered extent, so
4588		 * go ahead and adjust mod_start and mod_len in case this ordered
4589		 * extent has already been logged.
4590		 */
4591		if (ordered->file_offset > mod_start) {
4592			if (ordered_end >= mod_end)
4593				mod_len = ordered->file_offset - mod_start;
4594			/*
4595			 * If we have this case
4596			 *
4597			 * |--------- logged extent ---------|
4598			 *       |----- ordered extent ----|
4599			 *
4600			 * Just don't mess with mod_start and mod_len, we'll
4601			 * just end up logging more csums than we need and it
4602			 * will be ok.
4603			 */
4604		} else {
4605			if (ordered_end < mod_end) {
4606				mod_len = mod_end - ordered_end;
4607				mod_start = ordered_end;
4608			} else {
4609				mod_len = 0;
4610			}
4611		}
4612
4613		/*
4614		 * To keep us from looping for the above case of an ordered
4615		 * extent that falls inside of the logged extent.
4616		 */
4617		if (test_and_set_bit(BTRFS_ORDERED_LOGGED_CSUM, &ordered->flags))
4618			continue;
4619
4620		list_for_each_entry(sums, &ordered->list, list) {
4621			ret = log_csums(trans, inode, log_root, sums);
4622			if (ret)
4623				return ret;
4624		}
4625	}
4626
4627	/* We're done, found all csums in the ordered extents. */
4628	if (mod_len == 0)
4629		return 0;
4630
4631	/* If we're compressed we have to save the entire range of csums. */
4632	if (em->compress_type) {
4633		csum_offset = 0;
4634		csum_len = max(em->block_len, em->orig_block_len);
4635	} else {
4636		csum_offset = mod_start - em->start;
4637		csum_len = mod_len;
4638	}
4639
4640	/* block start is already adjusted for the file extent offset. */
4641	csum_root = btrfs_csum_root(trans->fs_info, em->block_start);
4642	ret = btrfs_lookup_csums_list(csum_root, em->block_start + csum_offset,
4643				      em->block_start + csum_offset +
4644				      csum_len - 1, &ordered_sums, 0, false);
4645	if (ret)
4646		return ret;
4647
4648	while (!list_empty(&ordered_sums)) {
4649		struct btrfs_ordered_sum *sums = list_entry(ordered_sums.next,
4650						   struct btrfs_ordered_sum,
4651						   list);
4652		if (!ret)
4653			ret = log_csums(trans, inode, log_root, sums);
4654		list_del(&sums->list);
4655		kfree(sums);
4656	}
4657
4658	return ret;
4659}
4660
4661static int log_one_extent(struct btrfs_trans_handle *trans,
4662			  struct btrfs_inode *inode,
4663			  const struct extent_map *em,
4664			  struct btrfs_path *path,
4665			  struct btrfs_log_ctx *ctx)
4666{
4667	struct btrfs_drop_extents_args drop_args = { 0 };
4668	struct btrfs_root *log = inode->root->log_root;
4669	struct btrfs_file_extent_item fi = { 0 };
4670	struct extent_buffer *leaf;
4671	struct btrfs_key key;
 
4672	u64 extent_offset = em->start - em->orig_start;
4673	u64 block_len;
4674	int ret;
4675
4676	btrfs_set_stack_file_extent_generation(&fi, trans->transid);
4677	if (test_bit(EXTENT_FLAG_PREALLOC, &em->flags))
4678		btrfs_set_stack_file_extent_type(&fi, BTRFS_FILE_EXTENT_PREALLOC);
4679	else
4680		btrfs_set_stack_file_extent_type(&fi, BTRFS_FILE_EXTENT_REG);
4681
4682	block_len = max(em->block_len, em->orig_block_len);
4683	if (em->compress_type != BTRFS_COMPRESS_NONE) {
 
4684		btrfs_set_stack_file_extent_disk_bytenr(&fi, em->block_start);
4685		btrfs_set_stack_file_extent_disk_num_bytes(&fi, block_len);
4686	} else if (em->block_start < EXTENT_MAP_LAST_BYTE) {
4687		btrfs_set_stack_file_extent_disk_bytenr(&fi, em->block_start -
4688							extent_offset);
4689		btrfs_set_stack_file_extent_disk_num_bytes(&fi, block_len);
4690	}
4691
4692	btrfs_set_stack_file_extent_offset(&fi, extent_offset);
4693	btrfs_set_stack_file_extent_num_bytes(&fi, em->len);
4694	btrfs_set_stack_file_extent_ram_bytes(&fi, em->ram_bytes);
4695	btrfs_set_stack_file_extent_compression(&fi, em->compress_type);
4696
4697	ret = log_extent_csums(trans, inode, log, em, ctx);
4698	if (ret)
4699		return ret;
4700
4701	/*
4702	 * If this is the first time we are logging the inode in the current
4703	 * transaction, we can avoid btrfs_drop_extents(), which is expensive
4704	 * because it does a deletion search, which always acquires write locks
4705	 * for extent buffers at levels 2, 1 and 0. This not only wastes time
4706	 * but also adds significant contention in a log tree, since log trees
4707	 * are small, with a root at level 2 or 3 at most, due to their short
4708	 * life span.
4709	 */
4710	if (ctx->logged_before) {
4711		drop_args.path = path;
4712		drop_args.start = em->start;
4713		drop_args.end = em->start + em->len;
4714		drop_args.replace_extent = true;
4715		drop_args.extent_item_size = sizeof(fi);
4716		ret = btrfs_drop_extents(trans, log, inode, &drop_args);
4717		if (ret)
4718			return ret;
4719	}
4720
4721	if (!drop_args.extent_inserted) {
4722		key.objectid = btrfs_ino(inode);
4723		key.type = BTRFS_EXTENT_DATA_KEY;
4724		key.offset = em->start;
4725
4726		ret = btrfs_insert_empty_item(trans, log, path, &key,
4727					      sizeof(fi));
4728		if (ret)
4729			return ret;
4730	}
4731	leaf = path->nodes[0];
4732	write_extent_buffer(leaf, &fi,
4733			    btrfs_item_ptr_offset(leaf, path->slots[0]),
4734			    sizeof(fi));
4735	btrfs_mark_buffer_dirty(leaf);
4736
4737	btrfs_release_path(path);
4738
4739	return ret;
4740}
4741
4742/*
4743 * Log all prealloc extents beyond the inode's i_size to make sure we do not
4744 * lose them after doing a full/fast fsync and replaying the log. We scan the
4745 * subvolume's root instead of iterating the inode's extent map tree because
4746 * otherwise we can log incorrect extent items based on extent map conversion.
4747 * That can happen due to the fact that extent maps are merged when they
4748 * are not in the extent map tree's list of modified extents.
4749 */
4750static int btrfs_log_prealloc_extents(struct btrfs_trans_handle *trans,
4751				      struct btrfs_inode *inode,
4752				      struct btrfs_path *path)
 
4753{
4754	struct btrfs_root *root = inode->root;
4755	struct btrfs_key key;
4756	const u64 i_size = i_size_read(&inode->vfs_inode);
4757	const u64 ino = btrfs_ino(inode);
4758	struct btrfs_path *dst_path = NULL;
4759	bool dropped_extents = false;
4760	u64 truncate_offset = i_size;
4761	struct extent_buffer *leaf;
4762	int slot;
4763	int ins_nr = 0;
4764	int start_slot;
4765	int ret;
4766
4767	if (!(inode->flags & BTRFS_INODE_PREALLOC))
4768		return 0;
4769
4770	key.objectid = ino;
4771	key.type = BTRFS_EXTENT_DATA_KEY;
4772	key.offset = i_size;
4773	ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
4774	if (ret < 0)
4775		goto out;
4776
4777	/*
4778	 * We must check if there is a prealloc extent that starts before the
4779	 * i_size and crosses the i_size boundary. This is to ensure later we
4780	 * truncate down to the end of that extent and not to the i_size, as
4781	 * otherwise we end up losing part of the prealloc extent after a log
4782	 * replay and with an implicit hole if there is another prealloc extent
4783	 * that starts at an offset beyond i_size.
4784	 */
4785	ret = btrfs_previous_item(root, path, ino, BTRFS_EXTENT_DATA_KEY);
4786	if (ret < 0)
4787		goto out;
4788
4789	if (ret == 0) {
4790		struct btrfs_file_extent_item *ei;
4791
4792		leaf = path->nodes[0];
4793		slot = path->slots[0];
4794		ei = btrfs_item_ptr(leaf, slot, struct btrfs_file_extent_item);
4795
4796		if (btrfs_file_extent_type(leaf, ei) ==
4797		    BTRFS_FILE_EXTENT_PREALLOC) {
4798			u64 extent_end;
4799
4800			btrfs_item_key_to_cpu(leaf, &key, slot);
4801			extent_end = key.offset +
4802				btrfs_file_extent_num_bytes(leaf, ei);
4803
4804			if (extent_end > i_size)
4805				truncate_offset = extent_end;
4806		}
4807	} else {
4808		ret = 0;
4809	}
4810
4811	while (true) {
4812		leaf = path->nodes[0];
4813		slot = path->slots[0];
4814
4815		if (slot >= btrfs_header_nritems(leaf)) {
4816			if (ins_nr > 0) {
4817				ret = copy_items(trans, inode, dst_path, path,
4818						 start_slot, ins_nr, 1, 0);
4819				if (ret < 0)
4820					goto out;
4821				ins_nr = 0;
4822			}
4823			ret = btrfs_next_leaf(root, path);
4824			if (ret < 0)
4825				goto out;
4826			if (ret > 0) {
4827				ret = 0;
4828				break;
4829			}
4830			continue;
4831		}
4832
4833		btrfs_item_key_to_cpu(leaf, &key, slot);
4834		if (key.objectid > ino)
4835			break;
4836		if (WARN_ON_ONCE(key.objectid < ino) ||
4837		    key.type < BTRFS_EXTENT_DATA_KEY ||
4838		    key.offset < i_size) {
4839			path->slots[0]++;
4840			continue;
4841		}
4842		if (!dropped_extents) {
4843			/*
4844			 * Avoid logging extent items logged in past fsync calls
4845			 * and leading to duplicate keys in the log tree.
4846			 */
4847			ret = truncate_inode_items(trans, root->log_root, inode,
4848						   truncate_offset,
4849						   BTRFS_EXTENT_DATA_KEY);
4850			if (ret)
4851				goto out;
4852			dropped_extents = true;
4853		}
4854		if (ins_nr == 0)
4855			start_slot = slot;
4856		ins_nr++;
4857		path->slots[0]++;
4858		if (!dst_path) {
4859			dst_path = btrfs_alloc_path();
4860			if (!dst_path) {
4861				ret = -ENOMEM;
4862				goto out;
4863			}
4864		}
4865	}
4866	if (ins_nr > 0)
4867		ret = copy_items(trans, inode, dst_path, path,
4868				 start_slot, ins_nr, 1, 0);
4869out:
4870	btrfs_release_path(path);
4871	btrfs_free_path(dst_path);
4872	return ret;
4873}
4874
4875static int btrfs_log_changed_extents(struct btrfs_trans_handle *trans,
4876				     struct btrfs_inode *inode,
4877				     struct btrfs_path *path,
4878				     struct btrfs_log_ctx *ctx)
4879{
4880	struct btrfs_ordered_extent *ordered;
4881	struct btrfs_ordered_extent *tmp;
4882	struct extent_map *em, *n;
4883	struct list_head extents;
4884	struct extent_map_tree *tree = &inode->extent_tree;
4885	int ret = 0;
4886	int num = 0;
4887
4888	INIT_LIST_HEAD(&extents);
4889
4890	write_lock(&tree->lock);
4891
4892	list_for_each_entry_safe(em, n, &tree->modified_extents, list) {
4893		list_del_init(&em->list);
4894		/*
4895		 * Just an arbitrary number, this can be really CPU intensive
4896		 * once we start getting a lot of extents, and really once we
4897		 * have a bunch of extents we just want to commit since it will
4898		 * be faster.
4899		 */
4900		if (++num > 32768) {
4901			list_del_init(&tree->modified_extents);
4902			ret = -EFBIG;
4903			goto process;
4904		}
4905
4906		if (em->generation < trans->transid)
4907			continue;
4908
4909		/* We log prealloc extents beyond eof later. */
4910		if (test_bit(EXTENT_FLAG_PREALLOC, &em->flags) &&
4911		    em->start >= i_size_read(&inode->vfs_inode))
4912			continue;
4913
4914		/* Need a ref to keep it from getting evicted from cache */
4915		refcount_inc(&em->refs);
4916		set_bit(EXTENT_FLAG_LOGGING, &em->flags);
4917		list_add_tail(&em->list, &extents);
4918		num++;
4919	}
4920
4921	list_sort(NULL, &extents, extent_cmp);
4922process:
4923	while (!list_empty(&extents)) {
4924		em = list_entry(extents.next, struct extent_map, list);
4925
4926		list_del_init(&em->list);
4927
4928		/*
4929		 * If we had an error we just need to delete everybody from our
4930		 * private list.
4931		 */
4932		if (ret) {
4933			clear_em_logging(tree, em);
4934			free_extent_map(em);
4935			continue;
4936		}
4937
4938		write_unlock(&tree->lock);
4939
4940		ret = log_one_extent(trans, inode, em, path, ctx);
4941		write_lock(&tree->lock);
4942		clear_em_logging(tree, em);
4943		free_extent_map(em);
4944	}
4945	WARN_ON(!list_empty(&extents));
4946	write_unlock(&tree->lock);
4947
4948	if (!ret)
4949		ret = btrfs_log_prealloc_extents(trans, inode, path);
4950	if (ret)
4951		return ret;
4952
4953	/*
4954	 * We have logged all extents successfully, now make sure the commit of
4955	 * the current transaction waits for the ordered extents to complete
4956	 * before it commits and wipes out the log trees, otherwise we would
4957	 * lose data if an ordered extents completes after the transaction
4958	 * commits and a power failure happens after the transaction commit.
4959	 */
4960	list_for_each_entry_safe(ordered, tmp, &ctx->ordered_extents, log_list) {
4961		list_del_init(&ordered->log_list);
4962		set_bit(BTRFS_ORDERED_LOGGED, &ordered->flags);
4963
4964		if (!test_bit(BTRFS_ORDERED_COMPLETE, &ordered->flags)) {
4965			spin_lock_irq(&inode->ordered_tree.lock);
4966			if (!test_bit(BTRFS_ORDERED_COMPLETE, &ordered->flags)) {
4967				set_bit(BTRFS_ORDERED_PENDING, &ordered->flags);
4968				atomic_inc(&trans->transaction->pending_ordered);
4969			}
4970			spin_unlock_irq(&inode->ordered_tree.lock);
4971		}
4972		btrfs_put_ordered_extent(ordered);
4973	}
4974
4975	return 0;
4976}
4977
4978static int logged_inode_size(struct btrfs_root *log, struct btrfs_inode *inode,
4979			     struct btrfs_path *path, u64 *size_ret)
4980{
4981	struct btrfs_key key;
4982	int ret;
4983
4984	key.objectid = btrfs_ino(inode);
4985	key.type = BTRFS_INODE_ITEM_KEY;
4986	key.offset = 0;
4987
4988	ret = btrfs_search_slot(NULL, log, &key, path, 0, 0);
4989	if (ret < 0) {
4990		return ret;
4991	} else if (ret > 0) {
4992		*size_ret = 0;
4993	} else {
4994		struct btrfs_inode_item *item;
4995
4996		item = btrfs_item_ptr(path->nodes[0], path->slots[0],
4997				      struct btrfs_inode_item);
4998		*size_ret = btrfs_inode_size(path->nodes[0], item);
4999		/*
5000		 * If the in-memory inode's i_size is smaller then the inode
5001		 * size stored in the btree, return the inode's i_size, so
5002		 * that we get a correct inode size after replaying the log
5003		 * when before a power failure we had a shrinking truncate
5004		 * followed by addition of a new name (rename / new hard link).
5005		 * Otherwise return the inode size from the btree, to avoid
5006		 * data loss when replaying a log due to previously doing a
5007		 * write that expands the inode's size and logging a new name
5008		 * immediately after.
5009		 */
5010		if (*size_ret > inode->vfs_inode.i_size)
5011			*size_ret = inode->vfs_inode.i_size;
5012	}
5013
5014	btrfs_release_path(path);
5015	return 0;
5016}
5017
5018/*
5019 * At the moment we always log all xattrs. This is to figure out at log replay
5020 * time which xattrs must have their deletion replayed. If a xattr is missing
5021 * in the log tree and exists in the fs/subvol tree, we delete it. This is
5022 * because if a xattr is deleted, the inode is fsynced and a power failure
5023 * happens, causing the log to be replayed the next time the fs is mounted,
5024 * we want the xattr to not exist anymore (same behaviour as other filesystems
5025 * with a journal, ext3/4, xfs, f2fs, etc).
5026 */
5027static int btrfs_log_all_xattrs(struct btrfs_trans_handle *trans,
5028				struct btrfs_inode *inode,
5029				struct btrfs_path *path,
5030				struct btrfs_path *dst_path)
 
5031{
5032	struct btrfs_root *root = inode->root;
5033	int ret;
5034	struct btrfs_key key;
5035	const u64 ino = btrfs_ino(inode);
5036	int ins_nr = 0;
5037	int start_slot = 0;
5038	bool found_xattrs = false;
5039
5040	if (test_bit(BTRFS_INODE_NO_XATTRS, &inode->runtime_flags))
5041		return 0;
5042
5043	key.objectid = ino;
5044	key.type = BTRFS_XATTR_ITEM_KEY;
5045	key.offset = 0;
5046
5047	ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
5048	if (ret < 0)
5049		return ret;
5050
5051	while (true) {
5052		int slot = path->slots[0];
5053		struct extent_buffer *leaf = path->nodes[0];
5054		int nritems = btrfs_header_nritems(leaf);
5055
5056		if (slot >= nritems) {
5057			if (ins_nr > 0) {
5058				ret = copy_items(trans, inode, dst_path, path,
5059						 start_slot, ins_nr, 1, 0);
5060				if (ret < 0)
5061					return ret;
5062				ins_nr = 0;
5063			}
5064			ret = btrfs_next_leaf(root, path);
5065			if (ret < 0)
5066				return ret;
5067			else if (ret > 0)
5068				break;
5069			continue;
5070		}
5071
5072		btrfs_item_key_to_cpu(leaf, &key, slot);
5073		if (key.objectid != ino || key.type != BTRFS_XATTR_ITEM_KEY)
5074			break;
5075
5076		if (ins_nr == 0)
5077			start_slot = slot;
5078		ins_nr++;
5079		path->slots[0]++;
5080		found_xattrs = true;
5081		cond_resched();
5082	}
5083	if (ins_nr > 0) {
5084		ret = copy_items(trans, inode, dst_path, path,
5085				 start_slot, ins_nr, 1, 0);
5086		if (ret < 0)
5087			return ret;
5088	}
5089
5090	if (!found_xattrs)
5091		set_bit(BTRFS_INODE_NO_XATTRS, &inode->runtime_flags);
5092
5093	return 0;
5094}
5095
5096/*
5097 * When using the NO_HOLES feature if we punched a hole that causes the
5098 * deletion of entire leafs or all the extent items of the first leaf (the one
5099 * that contains the inode item and references) we may end up not processing
5100 * any extents, because there are no leafs with a generation matching the
5101 * current transaction that have extent items for our inode. So we need to find
5102 * if any holes exist and then log them. We also need to log holes after any
5103 * truncate operation that changes the inode's size.
5104 */
5105static int btrfs_log_holes(struct btrfs_trans_handle *trans,
5106			   struct btrfs_inode *inode,
5107			   struct btrfs_path *path)
5108{
5109	struct btrfs_root *root = inode->root;
5110	struct btrfs_fs_info *fs_info = root->fs_info;
5111	struct btrfs_key key;
5112	const u64 ino = btrfs_ino(inode);
5113	const u64 i_size = i_size_read(&inode->vfs_inode);
5114	u64 prev_extent_end = 0;
5115	int ret;
5116
5117	if (!btrfs_fs_incompat(fs_info, NO_HOLES) || i_size == 0)
5118		return 0;
5119
5120	key.objectid = ino;
5121	key.type = BTRFS_EXTENT_DATA_KEY;
5122	key.offset = 0;
5123
5124	ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
5125	if (ret < 0)
5126		return ret;
5127
5128	while (true) {
5129		struct extent_buffer *leaf = path->nodes[0];
5130
5131		if (path->slots[0] >= btrfs_header_nritems(path->nodes[0])) {
5132			ret = btrfs_next_leaf(root, path);
5133			if (ret < 0)
5134				return ret;
5135			if (ret > 0) {
5136				ret = 0;
5137				break;
5138			}
5139			leaf = path->nodes[0];
5140		}
5141
5142		btrfs_item_key_to_cpu(leaf, &key, path->slots[0]);
5143		if (key.objectid != ino || key.type != BTRFS_EXTENT_DATA_KEY)
5144			break;
5145
5146		/* We have a hole, log it. */
5147		if (prev_extent_end < key.offset) {
5148			const u64 hole_len = key.offset - prev_extent_end;
5149
5150			/*
5151			 * Release the path to avoid deadlocks with other code
5152			 * paths that search the root while holding locks on
5153			 * leafs from the log root.
5154			 */
5155			btrfs_release_path(path);
5156			ret = btrfs_insert_hole_extent(trans, root->log_root,
5157						       ino, prev_extent_end,
5158						       hole_len);
5159			if (ret < 0)
5160				return ret;
5161
5162			/*
5163			 * Search for the same key again in the root. Since it's
5164			 * an extent item and we are holding the inode lock, the
5165			 * key must still exist. If it doesn't just emit warning
5166			 * and return an error to fall back to a transaction
5167			 * commit.
5168			 */
5169			ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
5170			if (ret < 0)
5171				return ret;
5172			if (WARN_ON(ret > 0))
5173				return -ENOENT;
5174			leaf = path->nodes[0];
5175		}
5176
5177		prev_extent_end = btrfs_file_extent_end(path);
5178		path->slots[0]++;
5179		cond_resched();
5180	}
5181
5182	if (prev_extent_end < i_size) {
5183		u64 hole_len;
5184
5185		btrfs_release_path(path);
5186		hole_len = ALIGN(i_size - prev_extent_end, fs_info->sectorsize);
5187		ret = btrfs_insert_hole_extent(trans, root->log_root, ino,
5188					       prev_extent_end, hole_len);
5189		if (ret < 0)
5190			return ret;
5191	}
5192
5193	return 0;
5194}
5195
5196/*
5197 * When we are logging a new inode X, check if it doesn't have a reference that
5198 * matches the reference from some other inode Y created in a past transaction
5199 * and that was renamed in the current transaction. If we don't do this, then at
5200 * log replay time we can lose inode Y (and all its files if it's a directory):
5201 *
5202 * mkdir /mnt/x
5203 * echo "hello world" > /mnt/x/foobar
5204 * sync
5205 * mv /mnt/x /mnt/y
5206 * mkdir /mnt/x                 # or touch /mnt/x
5207 * xfs_io -c fsync /mnt/x
5208 * <power fail>
5209 * mount fs, trigger log replay
5210 *
5211 * After the log replay procedure, we would lose the first directory and all its
5212 * files (file foobar).
5213 * For the case where inode Y is not a directory we simply end up losing it:
5214 *
5215 * echo "123" > /mnt/foo
5216 * sync
5217 * mv /mnt/foo /mnt/bar
5218 * echo "abc" > /mnt/foo
5219 * xfs_io -c fsync /mnt/foo
5220 * <power fail>
5221 *
5222 * We also need this for cases where a snapshot entry is replaced by some other
5223 * entry (file or directory) otherwise we end up with an unreplayable log due to
5224 * attempts to delete the snapshot entry (entry of type BTRFS_ROOT_ITEM_KEY) as
5225 * if it were a regular entry:
5226 *
5227 * mkdir /mnt/x
5228 * btrfs subvolume snapshot /mnt /mnt/x/snap
5229 * btrfs subvolume delete /mnt/x/snap
5230 * rmdir /mnt/x
5231 * mkdir /mnt/x
5232 * fsync /mnt/x or fsync some new file inside it
5233 * <power fail>
5234 *
5235 * The snapshot delete, rmdir of x, mkdir of a new x and the fsync all happen in
5236 * the same transaction.
5237 */
5238static int btrfs_check_ref_name_override(struct extent_buffer *eb,
5239					 const int slot,
5240					 const struct btrfs_key *key,
5241					 struct btrfs_inode *inode,
5242					 u64 *other_ino, u64 *other_parent)
5243{
5244	int ret;
5245	struct btrfs_path *search_path;
5246	char *name = NULL;
5247	u32 name_len = 0;
5248	u32 item_size = btrfs_item_size(eb, slot);
5249	u32 cur_offset = 0;
5250	unsigned long ptr = btrfs_item_ptr_offset(eb, slot);
5251
5252	search_path = btrfs_alloc_path();
5253	if (!search_path)
5254		return -ENOMEM;
5255	search_path->search_commit_root = 1;
5256	search_path->skip_locking = 1;
5257
5258	while (cur_offset < item_size) {
5259		u64 parent;
5260		u32 this_name_len;
5261		u32 this_len;
5262		unsigned long name_ptr;
5263		struct btrfs_dir_item *di;
5264		struct fscrypt_str name_str;
5265
5266		if (key->type == BTRFS_INODE_REF_KEY) {
5267			struct btrfs_inode_ref *iref;
5268
5269			iref = (struct btrfs_inode_ref *)(ptr + cur_offset);
5270			parent = key->offset;
5271			this_name_len = btrfs_inode_ref_name_len(eb, iref);
5272			name_ptr = (unsigned long)(iref + 1);
5273			this_len = sizeof(*iref) + this_name_len;
5274		} else {
5275			struct btrfs_inode_extref *extref;
5276
5277			extref = (struct btrfs_inode_extref *)(ptr +
5278							       cur_offset);
5279			parent = btrfs_inode_extref_parent(eb, extref);
5280			this_name_len = btrfs_inode_extref_name_len(eb, extref);
5281			name_ptr = (unsigned long)&extref->name;
5282			this_len = sizeof(*extref) + this_name_len;
5283		}
5284
5285		if (this_name_len > name_len) {
5286			char *new_name;
5287
5288			new_name = krealloc(name, this_name_len, GFP_NOFS);
5289			if (!new_name) {
5290				ret = -ENOMEM;
5291				goto out;
5292			}
5293			name_len = this_name_len;
5294			name = new_name;
5295		}
5296
5297		read_extent_buffer(eb, name, name_ptr, this_name_len);
5298
5299		name_str.name = name;
5300		name_str.len = this_name_len;
5301		di = btrfs_lookup_dir_item(NULL, inode->root, search_path,
5302				parent, &name_str, 0);
5303		if (di && !IS_ERR(di)) {
5304			struct btrfs_key di_key;
5305
5306			btrfs_dir_item_key_to_cpu(search_path->nodes[0],
5307						  di, &di_key);
5308			if (di_key.type == BTRFS_INODE_ITEM_KEY) {
5309				if (di_key.objectid != key->objectid) {
5310					ret = 1;
5311					*other_ino = di_key.objectid;
5312					*other_parent = parent;
5313				} else {
5314					ret = 0;
5315				}
5316			} else {
5317				ret = -EAGAIN;
5318			}
5319			goto out;
5320		} else if (IS_ERR(di)) {
5321			ret = PTR_ERR(di);
5322			goto out;
5323		}
5324		btrfs_release_path(search_path);
5325
5326		cur_offset += this_len;
5327	}
5328	ret = 0;
5329out:
5330	btrfs_free_path(search_path);
5331	kfree(name);
5332	return ret;
5333}
5334
5335/*
5336 * Check if we need to log an inode. This is used in contexts where while
5337 * logging an inode we need to log another inode (either that it exists or in
5338 * full mode). This is used instead of btrfs_inode_in_log() because the later
5339 * requires the inode to be in the log and have the log transaction committed,
5340 * while here we do not care if the log transaction was already committed - our
5341 * caller will commit the log later - and we want to avoid logging an inode
5342 * multiple times when multiple tasks have joined the same log transaction.
5343 */
5344static bool need_log_inode(const struct btrfs_trans_handle *trans,
5345			   const struct btrfs_inode *inode)
5346{
5347	/*
5348	 * If a directory was not modified, no dentries added or removed, we can
5349	 * and should avoid logging it.
5350	 */
5351	if (S_ISDIR(inode->vfs_inode.i_mode) && inode->last_trans < trans->transid)
5352		return false;
5353
5354	/*
5355	 * If this inode does not have new/updated/deleted xattrs since the last
5356	 * time it was logged and is flagged as logged in the current transaction,
5357	 * we can skip logging it. As for new/deleted names, those are updated in
5358	 * the log by link/unlink/rename operations.
5359	 * In case the inode was logged and then evicted and reloaded, its
5360	 * logged_trans will be 0, in which case we have to fully log it since
5361	 * logged_trans is a transient field, not persisted.
5362	 */
5363	if (inode->logged_trans == trans->transid &&
5364	    !test_bit(BTRFS_INODE_COPY_EVERYTHING, &inode->runtime_flags))
5365		return false;
5366
5367	return true;
5368}
5369
5370struct btrfs_dir_list {
5371	u64 ino;
5372	struct list_head list;
5373};
5374
5375/*
5376 * Log the inodes of the new dentries of a directory.
5377 * See process_dir_items_leaf() for details about why it is needed.
5378 * This is a recursive operation - if an existing dentry corresponds to a
5379 * directory, that directory's new entries are logged too (same behaviour as
5380 * ext3/4, xfs, f2fs, reiserfs, nilfs2). Note that when logging the inodes
5381 * the dentries point to we do not acquire their VFS lock, otherwise lockdep
5382 * complains about the following circular lock dependency / possible deadlock:
5383 *
5384 *        CPU0                                        CPU1
5385 *        ----                                        ----
5386 * lock(&type->i_mutex_dir_key#3/2);
5387 *                                            lock(sb_internal#2);
5388 *                                            lock(&type->i_mutex_dir_key#3/2);
5389 * lock(&sb->s_type->i_mutex_key#14);
5390 *
5391 * Where sb_internal is the lock (a counter that works as a lock) acquired by
5392 * sb_start_intwrite() in btrfs_start_transaction().
5393 * Not acquiring the VFS lock of the inodes is still safe because:
5394 *
5395 * 1) For regular files we log with a mode of LOG_INODE_EXISTS. It's possible
5396 *    that while logging the inode new references (names) are added or removed
5397 *    from the inode, leaving the logged inode item with a link count that does
5398 *    not match the number of logged inode reference items. This is fine because
5399 *    at log replay time we compute the real number of links and correct the
5400 *    link count in the inode item (see replay_one_buffer() and
5401 *    link_to_fixup_dir());
5402 *
5403 * 2) For directories we log with a mode of LOG_INODE_ALL. It's possible that
5404 *    while logging the inode's items new index items (key type
5405 *    BTRFS_DIR_INDEX_KEY) are added to fs/subvol tree and the logged inode item
5406 *    has a size that doesn't match the sum of the lengths of all the logged
5407 *    names - this is ok, not a problem, because at log replay time we set the
5408 *    directory's i_size to the correct value (see replay_one_name() and
5409 *    overwrite_item()).
5410 */
5411static int log_new_dir_dentries(struct btrfs_trans_handle *trans,
5412				struct btrfs_inode *start_inode,
5413				struct btrfs_log_ctx *ctx)
5414{
5415	struct btrfs_root *root = start_inode->root;
5416	struct btrfs_fs_info *fs_info = root->fs_info;
5417	struct btrfs_path *path;
5418	LIST_HEAD(dir_list);
5419	struct btrfs_dir_list *dir_elem;
5420	u64 ino = btrfs_ino(start_inode);
 
5421	int ret = 0;
5422
5423	/*
5424	 * If we are logging a new name, as part of a link or rename operation,
5425	 * don't bother logging new dentries, as we just want to log the names
5426	 * of an inode and that any new parents exist.
5427	 */
5428	if (ctx->logging_new_name)
5429		return 0;
5430
5431	path = btrfs_alloc_path();
5432	if (!path)
5433		return -ENOMEM;
5434
 
 
 
5435	while (true) {
5436		struct extent_buffer *leaf;
5437		struct btrfs_key min_key;
 
 
5438		bool continue_curr_inode = true;
5439		int nritems;
5440		int i;
5441
5442		min_key.objectid = ino;
5443		min_key.type = BTRFS_DIR_INDEX_KEY;
5444		min_key.offset = 0;
 
5445again:
5446		btrfs_release_path(path);
5447		ret = btrfs_search_forward(root, &min_key, path, trans->transid);
5448		if (ret < 0) {
5449			break;
5450		} else if (ret > 0) {
5451			ret = 0;
5452			goto next;
5453		}
5454
5455		leaf = path->nodes[0];
5456		nritems = btrfs_header_nritems(leaf);
5457		for (i = path->slots[0]; i < nritems; i++) {
5458			struct btrfs_dir_item *di;
5459			struct btrfs_key di_key;
5460			struct inode *di_inode;
5461			int log_mode = LOG_INODE_EXISTS;
5462			int type;
5463
5464			btrfs_item_key_to_cpu(leaf, &min_key, i);
5465			if (min_key.objectid != ino ||
5466			    min_key.type != BTRFS_DIR_INDEX_KEY) {
5467				continue_curr_inode = false;
5468				break;
5469			}
5470
5471			di = btrfs_item_ptr(leaf, i, struct btrfs_dir_item);
 
 
5472			type = btrfs_dir_ftype(leaf, di);
5473			if (btrfs_dir_transid(leaf, di) < trans->transid)
5474				continue;
5475			btrfs_dir_item_key_to_cpu(leaf, di, &di_key);
5476			if (di_key.type == BTRFS_ROOT_ITEM_KEY)
5477				continue;
5478
5479			btrfs_release_path(path);
5480			di_inode = btrfs_iget(fs_info->sb, di_key.objectid, root);
5481			if (IS_ERR(di_inode)) {
5482				ret = PTR_ERR(di_inode);
5483				goto out;
5484			}
5485
5486			if (!need_log_inode(trans, BTRFS_I(di_inode))) {
5487				btrfs_add_delayed_iput(BTRFS_I(di_inode));
5488				break;
5489			}
5490
5491			ctx->log_new_dentries = false;
5492			if (type == BTRFS_FT_DIR)
5493				log_mode = LOG_INODE_ALL;
5494			ret = btrfs_log_inode(trans, BTRFS_I(di_inode),
5495					      log_mode, ctx);
5496			btrfs_add_delayed_iput(BTRFS_I(di_inode));
5497			if (ret)
5498				goto out;
5499			if (ctx->log_new_dentries) {
5500				dir_elem = kmalloc(sizeof(*dir_elem), GFP_NOFS);
5501				if (!dir_elem) {
5502					ret = -ENOMEM;
5503					goto out;
5504				}
5505				dir_elem->ino = di_key.objectid;
5506				list_add_tail(&dir_elem->list, &dir_list);
5507			}
5508			break;
5509		}
5510
5511		if (continue_curr_inode && min_key.offset < (u64)-1) {
5512			min_key.offset++;
 
 
 
 
 
 
 
 
 
 
 
5513			goto again;
5514		}
5515
5516next:
 
5517		if (list_empty(&dir_list))
5518			break;
5519
5520		dir_elem = list_first_entry(&dir_list, struct btrfs_dir_list, list);
5521		ino = dir_elem->ino;
5522		list_del(&dir_elem->list);
5523		kfree(dir_elem);
 
 
 
 
 
 
 
 
 
 
5524	}
5525out:
5526	btrfs_free_path(path);
 
 
 
5527	if (ret) {
5528		struct btrfs_dir_list *next;
5529
5530		list_for_each_entry_safe(dir_elem, next, &dir_list, list)
5531			kfree(dir_elem);
5532	}
5533
5534	return ret;
5535}
5536
5537struct btrfs_ino_list {
5538	u64 ino;
5539	u64 parent;
5540	struct list_head list;
5541};
5542
5543static void free_conflicting_inodes(struct btrfs_log_ctx *ctx)
5544{
5545	struct btrfs_ino_list *curr;
5546	struct btrfs_ino_list *next;
5547
5548	list_for_each_entry_safe(curr, next, &ctx->conflict_inodes, list) {
5549		list_del(&curr->list);
5550		kfree(curr);
5551	}
5552}
5553
5554static int conflicting_inode_is_dir(struct btrfs_root *root, u64 ino,
5555				    struct btrfs_path *path)
5556{
5557	struct btrfs_key key;
5558	int ret;
5559
5560	key.objectid = ino;
5561	key.type = BTRFS_INODE_ITEM_KEY;
5562	key.offset = 0;
5563
5564	path->search_commit_root = 1;
5565	path->skip_locking = 1;
5566
5567	ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
5568	if (WARN_ON_ONCE(ret > 0)) {
5569		/*
5570		 * We have previously found the inode through the commit root
5571		 * so this should not happen. If it does, just error out and
5572		 * fallback to a transaction commit.
5573		 */
5574		ret = -ENOENT;
5575	} else if (ret == 0) {
5576		struct btrfs_inode_item *item;
5577
5578		item = btrfs_item_ptr(path->nodes[0], path->slots[0],
5579				      struct btrfs_inode_item);
5580		if (S_ISDIR(btrfs_inode_mode(path->nodes[0], item)))
5581			ret = 1;
5582	}
5583
5584	btrfs_release_path(path);
5585	path->search_commit_root = 0;
5586	path->skip_locking = 0;
5587
5588	return ret;
5589}
5590
5591static int add_conflicting_inode(struct btrfs_trans_handle *trans,
5592				 struct btrfs_root *root,
5593				 struct btrfs_path *path,
5594				 u64 ino, u64 parent,
5595				 struct btrfs_log_ctx *ctx)
5596{
5597	struct btrfs_ino_list *ino_elem;
5598	struct inode *inode;
5599
5600	/*
5601	 * It's rare to have a lot of conflicting inodes, in practice it is not
5602	 * common to have more than 1 or 2. We don't want to collect too many,
5603	 * as we could end up logging too many inodes (even if only in
5604	 * LOG_INODE_EXISTS mode) and slow down other fsyncs or transaction
5605	 * commits.
5606	 */
5607	if (ctx->num_conflict_inodes >= MAX_CONFLICT_INODES) {
5608		btrfs_set_log_full_commit(trans);
5609		return BTRFS_LOG_FORCE_COMMIT;
5610	}
5611
5612	inode = btrfs_iget(root->fs_info->sb, ino, root);
5613	/*
5614	 * If the other inode that had a conflicting dir entry was deleted in
5615	 * the current transaction then we either:
5616	 *
5617	 * 1) Log the parent directory (later after adding it to the list) if
5618	 *    the inode is a directory. This is because it may be a deleted
5619	 *    subvolume/snapshot or it may be a regular directory that had
5620	 *    deleted subvolumes/snapshots (or subdirectories that had them),
5621	 *    and at the moment we can't deal with dropping subvolumes/snapshots
5622	 *    during log replay. So we just log the parent, which will result in
5623	 *    a fallback to a transaction commit if we are dealing with those
5624	 *    cases (last_unlink_trans will match the current transaction);
5625	 *
5626	 * 2) Do nothing if it's not a directory. During log replay we simply
5627	 *    unlink the conflicting dentry from the parent directory and then
5628	 *    add the dentry for our inode. Like this we can avoid logging the
5629	 *    parent directory (and maybe fallback to a transaction commit in
5630	 *    case it has a last_unlink_trans == trans->transid, due to moving
5631	 *    some inode from it to some other directory).
5632	 */
5633	if (IS_ERR(inode)) {
5634		int ret = PTR_ERR(inode);
5635
5636		if (ret != -ENOENT)
5637			return ret;
5638
5639		ret = conflicting_inode_is_dir(root, ino, path);
5640		/* Not a directory or we got an error. */
5641		if (ret <= 0)
5642			return ret;
5643
5644		/* Conflicting inode is a directory, so we'll log its parent. */
5645		ino_elem = kmalloc(sizeof(*ino_elem), GFP_NOFS);
5646		if (!ino_elem)
5647			return -ENOMEM;
5648		ino_elem->ino = ino;
5649		ino_elem->parent = parent;
5650		list_add_tail(&ino_elem->list, &ctx->conflict_inodes);
5651		ctx->num_conflict_inodes++;
5652
5653		return 0;
5654	}
5655
5656	/*
5657	 * If the inode was already logged skip it - otherwise we can hit an
5658	 * infinite loop. Example:
5659	 *
5660	 * From the commit root (previous transaction) we have the following
5661	 * inodes:
5662	 *
5663	 * inode 257 a directory
5664	 * inode 258 with references "zz" and "zz_link" on inode 257
5665	 * inode 259 with reference "a" on inode 257
5666	 *
5667	 * And in the current (uncommitted) transaction we have:
5668	 *
5669	 * inode 257 a directory, unchanged
5670	 * inode 258 with references "a" and "a2" on inode 257
5671	 * inode 259 with reference "zz_link" on inode 257
5672	 * inode 261 with reference "zz" on inode 257
5673	 *
5674	 * When logging inode 261 the following infinite loop could
5675	 * happen if we don't skip already logged inodes:
5676	 *
5677	 * - we detect inode 258 as a conflicting inode, with inode 261
5678	 *   on reference "zz", and log it;
5679	 *
5680	 * - we detect inode 259 as a conflicting inode, with inode 258
5681	 *   on reference "a", and log it;
5682	 *
5683	 * - we detect inode 258 as a conflicting inode, with inode 259
5684	 *   on reference "zz_link", and log it - again! After this we
5685	 *   repeat the above steps forever.
5686	 *
5687	 * Here we can use need_log_inode() because we only need to log the
5688	 * inode in LOG_INODE_EXISTS mode and rename operations update the log,
5689	 * so that the log ends up with the new name and without the old name.
5690	 */
5691	if (!need_log_inode(trans, BTRFS_I(inode))) {
5692		btrfs_add_delayed_iput(BTRFS_I(inode));
5693		return 0;
5694	}
5695
5696	btrfs_add_delayed_iput(BTRFS_I(inode));
5697
5698	ino_elem = kmalloc(sizeof(*ino_elem), GFP_NOFS);
5699	if (!ino_elem)
5700		return -ENOMEM;
5701	ino_elem->ino = ino;
5702	ino_elem->parent = parent;
5703	list_add_tail(&ino_elem->list, &ctx->conflict_inodes);
5704	ctx->num_conflict_inodes++;
5705
5706	return 0;
5707}
5708
5709static int log_conflicting_inodes(struct btrfs_trans_handle *trans,
5710				  struct btrfs_root *root,
5711				  struct btrfs_log_ctx *ctx)
5712{
5713	struct btrfs_fs_info *fs_info = root->fs_info;
5714	int ret = 0;
5715
5716	/*
5717	 * Conflicting inodes are logged by the first call to btrfs_log_inode(),
5718	 * otherwise we could have unbounded recursion of btrfs_log_inode()
5719	 * calls. This check guarantees we can have only 1 level of recursion.
5720	 */
5721	if (ctx->logging_conflict_inodes)
5722		return 0;
5723
5724	ctx->logging_conflict_inodes = true;
5725
5726	/*
5727	 * New conflicting inodes may be found and added to the list while we
5728	 * are logging a conflicting inode, so keep iterating while the list is
5729	 * not empty.
5730	 */
5731	while (!list_empty(&ctx->conflict_inodes)) {
5732		struct btrfs_ino_list *curr;
5733		struct inode *inode;
5734		u64 ino;
5735		u64 parent;
5736
5737		curr = list_first_entry(&ctx->conflict_inodes,
5738					struct btrfs_ino_list, list);
5739		ino = curr->ino;
5740		parent = curr->parent;
5741		list_del(&curr->list);
5742		kfree(curr);
5743
5744		inode = btrfs_iget(fs_info->sb, ino, root);
5745		/*
5746		 * If the other inode that had a conflicting dir entry was
5747		 * deleted in the current transaction, we need to log its parent
5748		 * directory. See the comment at add_conflicting_inode().
5749		 */
5750		if (IS_ERR(inode)) {
5751			ret = PTR_ERR(inode);
5752			if (ret != -ENOENT)
5753				break;
5754
5755			inode = btrfs_iget(fs_info->sb, parent, root);
5756			if (IS_ERR(inode)) {
5757				ret = PTR_ERR(inode);
5758				break;
5759			}
5760
5761			/*
5762			 * Always log the directory, we cannot make this
5763			 * conditional on need_log_inode() because the directory
5764			 * might have been logged in LOG_INODE_EXISTS mode or
5765			 * the dir index of the conflicting inode is not in a
5766			 * dir index key range logged for the directory. So we
5767			 * must make sure the deletion is recorded.
5768			 */
5769			ret = btrfs_log_inode(trans, BTRFS_I(inode),
5770					      LOG_INODE_ALL, ctx);
5771			btrfs_add_delayed_iput(BTRFS_I(inode));
5772			if (ret)
5773				break;
5774			continue;
5775		}
5776
5777		/*
5778		 * Here we can use need_log_inode() because we only need to log
5779		 * the inode in LOG_INODE_EXISTS mode and rename operations
5780		 * update the log, so that the log ends up with the new name and
5781		 * without the old name.
5782		 *
5783		 * We did this check at add_conflicting_inode(), but here we do
5784		 * it again because if some other task logged the inode after
5785		 * that, we can avoid doing it again.
5786		 */
5787		if (!need_log_inode(trans, BTRFS_I(inode))) {
5788			btrfs_add_delayed_iput(BTRFS_I(inode));
5789			continue;
5790		}
5791
5792		/*
5793		 * We are safe logging the other inode without acquiring its
5794		 * lock as long as we log with the LOG_INODE_EXISTS mode. We
5795		 * are safe against concurrent renames of the other inode as
5796		 * well because during a rename we pin the log and update the
5797		 * log with the new name before we unpin it.
5798		 */
5799		ret = btrfs_log_inode(trans, BTRFS_I(inode), LOG_INODE_EXISTS, ctx);
5800		btrfs_add_delayed_iput(BTRFS_I(inode));
5801		if (ret)
5802			break;
5803	}
5804
5805	ctx->logging_conflict_inodes = false;
5806	if (ret)
5807		free_conflicting_inodes(ctx);
5808
5809	return ret;
5810}
5811
5812static int copy_inode_items_to_log(struct btrfs_trans_handle *trans,
5813				   struct btrfs_inode *inode,
5814				   struct btrfs_key *min_key,
5815				   const struct btrfs_key *max_key,
5816				   struct btrfs_path *path,
5817				   struct btrfs_path *dst_path,
5818				   const u64 logged_isize,
5819				   const int inode_only,
5820				   struct btrfs_log_ctx *ctx,
5821				   bool *need_log_inode_item)
5822{
5823	const u64 i_size = i_size_read(&inode->vfs_inode);
5824	struct btrfs_root *root = inode->root;
5825	int ins_start_slot = 0;
5826	int ins_nr = 0;
5827	int ret;
5828
5829	while (1) {
5830		ret = btrfs_search_forward(root, min_key, path, trans->transid);
5831		if (ret < 0)
5832			return ret;
5833		if (ret > 0) {
5834			ret = 0;
5835			break;
5836		}
5837again:
5838		/* Note, ins_nr might be > 0 here, cleanup outside the loop */
5839		if (min_key->objectid != max_key->objectid)
5840			break;
5841		if (min_key->type > max_key->type)
5842			break;
5843
5844		if (min_key->type == BTRFS_INODE_ITEM_KEY) {
5845			*need_log_inode_item = false;
5846		} else if (min_key->type == BTRFS_EXTENT_DATA_KEY &&
5847			   min_key->offset >= i_size) {
5848			/*
5849			 * Extents at and beyond eof are logged with
5850			 * btrfs_log_prealloc_extents().
5851			 * Only regular files have BTRFS_EXTENT_DATA_KEY keys,
5852			 * and no keys greater than that, so bail out.
5853			 */
5854			break;
5855		} else if ((min_key->type == BTRFS_INODE_REF_KEY ||
5856			    min_key->type == BTRFS_INODE_EXTREF_KEY) &&
5857			   (inode->generation == trans->transid ||
5858			    ctx->logging_conflict_inodes)) {
5859			u64 other_ino = 0;
5860			u64 other_parent = 0;
5861
5862			ret = btrfs_check_ref_name_override(path->nodes[0],
5863					path->slots[0], min_key, inode,
5864					&other_ino, &other_parent);
5865			if (ret < 0) {
5866				return ret;
5867			} else if (ret > 0 &&
5868				   other_ino != btrfs_ino(BTRFS_I(ctx->inode))) {
5869				if (ins_nr > 0) {
5870					ins_nr++;
5871				} else {
5872					ins_nr = 1;
5873					ins_start_slot = path->slots[0];
5874				}
5875				ret = copy_items(trans, inode, dst_path, path,
5876						 ins_start_slot, ins_nr,
5877						 inode_only, logged_isize);
5878				if (ret < 0)
5879					return ret;
5880				ins_nr = 0;
5881
5882				btrfs_release_path(path);
5883				ret = add_conflicting_inode(trans, root, path,
5884							    other_ino,
5885							    other_parent, ctx);
5886				if (ret)
5887					return ret;
5888				goto next_key;
5889			}
5890		} else if (min_key->type == BTRFS_XATTR_ITEM_KEY) {
5891			/* Skip xattrs, logged later with btrfs_log_all_xattrs() */
5892			if (ins_nr == 0)
5893				goto next_slot;
5894			ret = copy_items(trans, inode, dst_path, path,
5895					 ins_start_slot,
5896					 ins_nr, inode_only, logged_isize);
5897			if (ret < 0)
5898				return ret;
5899			ins_nr = 0;
5900			goto next_slot;
5901		}
5902
5903		if (ins_nr && ins_start_slot + ins_nr == path->slots[0]) {
5904			ins_nr++;
5905			goto next_slot;
5906		} else if (!ins_nr) {
5907			ins_start_slot = path->slots[0];
5908			ins_nr = 1;
5909			goto next_slot;
5910		}
5911
5912		ret = copy_items(trans, inode, dst_path, path, ins_start_slot,
5913				 ins_nr, inode_only, logged_isize);
5914		if (ret < 0)
5915			return ret;
5916		ins_nr = 1;
5917		ins_start_slot = path->slots[0];
5918next_slot:
5919		path->slots[0]++;
5920		if (path->slots[0] < btrfs_header_nritems(path->nodes[0])) {
5921			btrfs_item_key_to_cpu(path->nodes[0], min_key,
5922					      path->slots[0]);
5923			goto again;
5924		}
5925		if (ins_nr) {
5926			ret = copy_items(trans, inode, dst_path, path,
5927					 ins_start_slot, ins_nr, inode_only,
5928					 logged_isize);
5929			if (ret < 0)
5930				return ret;
5931			ins_nr = 0;
5932		}
5933		btrfs_release_path(path);
5934next_key:
5935		if (min_key->offset < (u64)-1) {
5936			min_key->offset++;
5937		} else if (min_key->type < max_key->type) {
5938			min_key->type++;
5939			min_key->offset = 0;
5940		} else {
5941			break;
5942		}
5943
5944		/*
5945		 * We may process many leaves full of items for our inode, so
5946		 * avoid monopolizing a cpu for too long by rescheduling while
5947		 * not holding locks on any tree.
5948		 */
5949		cond_resched();
5950	}
5951	if (ins_nr) {
5952		ret = copy_items(trans, inode, dst_path, path, ins_start_slot,
5953				 ins_nr, inode_only, logged_isize);
5954		if (ret)
5955			return ret;
5956	}
5957
5958	if (inode_only == LOG_INODE_ALL && S_ISREG(inode->vfs_inode.i_mode)) {
5959		/*
5960		 * Release the path because otherwise we might attempt to double
5961		 * lock the same leaf with btrfs_log_prealloc_extents() below.
5962		 */
5963		btrfs_release_path(path);
5964		ret = btrfs_log_prealloc_extents(trans, inode, dst_path);
5965	}
5966
5967	return ret;
5968}
5969
5970static int insert_delayed_items_batch(struct btrfs_trans_handle *trans,
5971				      struct btrfs_root *log,
5972				      struct btrfs_path *path,
5973				      const struct btrfs_item_batch *batch,
5974				      const struct btrfs_delayed_item *first_item)
5975{
5976	const struct btrfs_delayed_item *curr = first_item;
5977	int ret;
5978
5979	ret = btrfs_insert_empty_items(trans, log, path, batch);
5980	if (ret)
5981		return ret;
5982
5983	for (int i = 0; i < batch->nr; i++) {
5984		char *data_ptr;
5985
5986		data_ptr = btrfs_item_ptr(path->nodes[0], path->slots[0], char);
5987		write_extent_buffer(path->nodes[0], &curr->data,
5988				    (unsigned long)data_ptr, curr->data_len);
5989		curr = list_next_entry(curr, log_list);
5990		path->slots[0]++;
5991	}
5992
5993	btrfs_release_path(path);
5994
5995	return 0;
5996}
5997
5998static int log_delayed_insertion_items(struct btrfs_trans_handle *trans,
5999				       struct btrfs_inode *inode,
6000				       struct btrfs_path *path,
6001				       const struct list_head *delayed_ins_list,
6002				       struct btrfs_log_ctx *ctx)
6003{
6004	/* 195 (4095 bytes of keys and sizes) fits in a single 4K page. */
6005	const int max_batch_size = 195;
6006	const int leaf_data_size = BTRFS_LEAF_DATA_SIZE(trans->fs_info);
6007	const u64 ino = btrfs_ino(inode);
6008	struct btrfs_root *log = inode->root->log_root;
6009	struct btrfs_item_batch batch = {
6010		.nr = 0,
6011		.total_data_size = 0,
6012	};
6013	const struct btrfs_delayed_item *first = NULL;
6014	const struct btrfs_delayed_item *curr;
6015	char *ins_data;
6016	struct btrfs_key *ins_keys;
6017	u32 *ins_sizes;
6018	u64 curr_batch_size = 0;
6019	int batch_idx = 0;
6020	int ret;
6021
6022	/* We are adding dir index items to the log tree. */
6023	lockdep_assert_held(&inode->log_mutex);
6024
6025	/*
6026	 * We collect delayed items before copying index keys from the subvolume
6027	 * to the log tree. However just after we collected them, they may have
6028	 * been flushed (all of them or just some of them), and therefore we
6029	 * could have copied them from the subvolume tree to the log tree.
6030	 * So find the first delayed item that was not yet logged (they are
6031	 * sorted by index number).
6032	 */
6033	list_for_each_entry(curr, delayed_ins_list, log_list) {
6034		if (curr->index > inode->last_dir_index_offset) {
6035			first = curr;
6036			break;
6037		}
6038	}
6039
6040	/* Empty list or all delayed items were already logged. */
6041	if (!first)
6042		return 0;
6043
6044	ins_data = kmalloc(max_batch_size * sizeof(u32) +
6045			   max_batch_size * sizeof(struct btrfs_key), GFP_NOFS);
6046	if (!ins_data)
6047		return -ENOMEM;
6048	ins_sizes = (u32 *)ins_data;
6049	batch.data_sizes = ins_sizes;
6050	ins_keys = (struct btrfs_key *)(ins_data + max_batch_size * sizeof(u32));
6051	batch.keys = ins_keys;
6052
6053	curr = first;
6054	while (!list_entry_is_head(curr, delayed_ins_list, log_list)) {
6055		const u32 curr_size = curr->data_len + sizeof(struct btrfs_item);
6056
6057		if (curr_batch_size + curr_size > leaf_data_size ||
6058		    batch.nr == max_batch_size) {
6059			ret = insert_delayed_items_batch(trans, log, path,
6060							 &batch, first);
6061			if (ret)
6062				goto out;
6063			batch_idx = 0;
6064			batch.nr = 0;
6065			batch.total_data_size = 0;
6066			curr_batch_size = 0;
6067			first = curr;
6068		}
6069
6070		ins_sizes[batch_idx] = curr->data_len;
6071		ins_keys[batch_idx].objectid = ino;
6072		ins_keys[batch_idx].type = BTRFS_DIR_INDEX_KEY;
6073		ins_keys[batch_idx].offset = curr->index;
6074		curr_batch_size += curr_size;
6075		batch.total_data_size += curr->data_len;
6076		batch.nr++;
6077		batch_idx++;
6078		curr = list_next_entry(curr, log_list);
6079	}
6080
6081	ASSERT(batch.nr >= 1);
6082	ret = insert_delayed_items_batch(trans, log, path, &batch, first);
6083
6084	curr = list_last_entry(delayed_ins_list, struct btrfs_delayed_item,
6085			       log_list);
6086	inode->last_dir_index_offset = curr->index;
6087out:
6088	kfree(ins_data);
6089
6090	return ret;
6091}
6092
6093static int log_delayed_deletions_full(struct btrfs_trans_handle *trans,
6094				      struct btrfs_inode *inode,
6095				      struct btrfs_path *path,
6096				      const struct list_head *delayed_del_list,
6097				      struct btrfs_log_ctx *ctx)
6098{
6099	const u64 ino = btrfs_ino(inode);
6100	const struct btrfs_delayed_item *curr;
6101
6102	curr = list_first_entry(delayed_del_list, struct btrfs_delayed_item,
6103				log_list);
6104
6105	while (!list_entry_is_head(curr, delayed_del_list, log_list)) {
6106		u64 first_dir_index = curr->index;
6107		u64 last_dir_index;
6108		const struct btrfs_delayed_item *next;
6109		int ret;
6110
6111		/*
6112		 * Find a range of consecutive dir index items to delete. Like
6113		 * this we log a single dir range item spanning several contiguous
6114		 * dir items instead of logging one range item per dir index item.
6115		 */
6116		next = list_next_entry(curr, log_list);
6117		while (!list_entry_is_head(next, delayed_del_list, log_list)) {
6118			if (next->index != curr->index + 1)
6119				break;
6120			curr = next;
6121			next = list_next_entry(next, log_list);
6122		}
6123
6124		last_dir_index = curr->index;
6125		ASSERT(last_dir_index >= first_dir_index);
6126
6127		ret = insert_dir_log_key(trans, inode->root->log_root, path,
6128					 ino, first_dir_index, last_dir_index);
6129		if (ret)
6130			return ret;
6131		curr = list_next_entry(curr, log_list);
6132	}
6133
6134	return 0;
6135}
6136
6137static int batch_delete_dir_index_items(struct btrfs_trans_handle *trans,
6138					struct btrfs_inode *inode,
6139					struct btrfs_path *path,
6140					struct btrfs_log_ctx *ctx,
6141					const struct list_head *delayed_del_list,
6142					const struct btrfs_delayed_item *first,
6143					const struct btrfs_delayed_item **last_ret)
6144{
6145	const struct btrfs_delayed_item *next;
6146	struct extent_buffer *leaf = path->nodes[0];
6147	const int last_slot = btrfs_header_nritems(leaf) - 1;
6148	int slot = path->slots[0] + 1;
6149	const u64 ino = btrfs_ino(inode);
6150
6151	next = list_next_entry(first, log_list);
6152
6153	while (slot < last_slot &&
6154	       !list_entry_is_head(next, delayed_del_list, log_list)) {
6155		struct btrfs_key key;
6156
6157		btrfs_item_key_to_cpu(leaf, &key, slot);
6158		if (key.objectid != ino ||
6159		    key.type != BTRFS_DIR_INDEX_KEY ||
6160		    key.offset != next->index)
6161			break;
6162
6163		slot++;
6164		*last_ret = next;
6165		next = list_next_entry(next, log_list);
6166	}
6167
6168	return btrfs_del_items(trans, inode->root->log_root, path,
6169			       path->slots[0], slot - path->slots[0]);
6170}
6171
6172static int log_delayed_deletions_incremental(struct btrfs_trans_handle *trans,
6173					     struct btrfs_inode *inode,
6174					     struct btrfs_path *path,
6175					     const struct list_head *delayed_del_list,
6176					     struct btrfs_log_ctx *ctx)
6177{
6178	struct btrfs_root *log = inode->root->log_root;
6179	const struct btrfs_delayed_item *curr;
6180	u64 last_range_start;
6181	u64 last_range_end = 0;
6182	struct btrfs_key key;
6183
6184	key.objectid = btrfs_ino(inode);
6185	key.type = BTRFS_DIR_INDEX_KEY;
6186	curr = list_first_entry(delayed_del_list, struct btrfs_delayed_item,
6187				log_list);
6188
6189	while (!list_entry_is_head(curr, delayed_del_list, log_list)) {
6190		const struct btrfs_delayed_item *last = curr;
6191		u64 first_dir_index = curr->index;
6192		u64 last_dir_index;
6193		bool deleted_items = false;
6194		int ret;
6195
6196		key.offset = curr->index;
6197		ret = btrfs_search_slot(trans, log, &key, path, -1, 1);
6198		if (ret < 0) {
6199			return ret;
6200		} else if (ret == 0) {
6201			ret = batch_delete_dir_index_items(trans, inode, path, ctx,
6202							   delayed_del_list, curr,
6203							   &last);
6204			if (ret)
6205				return ret;
6206			deleted_items = true;
6207		}
6208
6209		btrfs_release_path(path);
6210
6211		/*
6212		 * If we deleted items from the leaf, it means we have a range
6213		 * item logging their range, so no need to add one or update an
6214		 * existing one. Otherwise we have to log a dir range item.
6215		 */
6216		if (deleted_items)
6217			goto next_batch;
6218
6219		last_dir_index = last->index;
6220		ASSERT(last_dir_index >= first_dir_index);
6221		/*
6222		 * If this range starts right after where the previous one ends,
6223		 * then we want to reuse the previous range item and change its
6224		 * end offset to the end of this range. This is just to minimize
6225		 * leaf space usage, by avoiding adding a new range item.
6226		 */
6227		if (last_range_end != 0 && first_dir_index == last_range_end + 1)
6228			first_dir_index = last_range_start;
6229
6230		ret = insert_dir_log_key(trans, log, path, key.objectid,
6231					 first_dir_index, last_dir_index);
6232		if (ret)
6233			return ret;
6234
6235		last_range_start = first_dir_index;
6236		last_range_end = last_dir_index;
6237next_batch:
6238		curr = list_next_entry(last, log_list);
6239	}
6240
6241	return 0;
6242}
6243
6244static int log_delayed_deletion_items(struct btrfs_trans_handle *trans,
6245				      struct btrfs_inode *inode,
6246				      struct btrfs_path *path,
6247				      const struct list_head *delayed_del_list,
6248				      struct btrfs_log_ctx *ctx)
6249{
6250	/*
6251	 * We are deleting dir index items from the log tree or adding range
6252	 * items to it.
6253	 */
6254	lockdep_assert_held(&inode->log_mutex);
6255
6256	if (list_empty(delayed_del_list))
6257		return 0;
6258
6259	if (ctx->logged_before)
6260		return log_delayed_deletions_incremental(trans, inode, path,
6261							 delayed_del_list, ctx);
6262
6263	return log_delayed_deletions_full(trans, inode, path, delayed_del_list,
6264					  ctx);
6265}
6266
6267/*
6268 * Similar logic as for log_new_dir_dentries(), but it iterates over the delayed
6269 * items instead of the subvolume tree.
6270 */
6271static int log_new_delayed_dentries(struct btrfs_trans_handle *trans,
6272				    struct btrfs_inode *inode,
6273				    const struct list_head *delayed_ins_list,
6274				    struct btrfs_log_ctx *ctx)
6275{
6276	const bool orig_log_new_dentries = ctx->log_new_dentries;
6277	struct btrfs_fs_info *fs_info = trans->fs_info;
6278	struct btrfs_delayed_item *item;
6279	int ret = 0;
6280
6281	/*
6282	 * No need for the log mutex, plus to avoid potential deadlocks or
6283	 * lockdep annotations due to nesting of delayed inode mutexes and log
6284	 * mutexes.
6285	 */
6286	lockdep_assert_not_held(&inode->log_mutex);
6287
6288	ASSERT(!ctx->logging_new_delayed_dentries);
6289	ctx->logging_new_delayed_dentries = true;
6290
6291	list_for_each_entry(item, delayed_ins_list, log_list) {
6292		struct btrfs_dir_item *dir_item;
6293		struct inode *di_inode;
6294		struct btrfs_key key;
6295		int log_mode = LOG_INODE_EXISTS;
6296
6297		dir_item = (struct btrfs_dir_item *)item->data;
6298		btrfs_disk_key_to_cpu(&key, &dir_item->location);
6299
6300		if (key.type == BTRFS_ROOT_ITEM_KEY)
6301			continue;
6302
6303		di_inode = btrfs_iget(fs_info->sb, key.objectid, inode->root);
6304		if (IS_ERR(di_inode)) {
6305			ret = PTR_ERR(di_inode);
6306			break;
6307		}
6308
6309		if (!need_log_inode(trans, BTRFS_I(di_inode))) {
6310			btrfs_add_delayed_iput(BTRFS_I(di_inode));
6311			continue;
6312		}
6313
6314		if (btrfs_stack_dir_ftype(dir_item) == BTRFS_FT_DIR)
6315			log_mode = LOG_INODE_ALL;
6316
6317		ctx->log_new_dentries = false;
6318		ret = btrfs_log_inode(trans, BTRFS_I(di_inode), log_mode, ctx);
6319
6320		if (!ret && ctx->log_new_dentries)
6321			ret = log_new_dir_dentries(trans, BTRFS_I(di_inode), ctx);
6322
6323		btrfs_add_delayed_iput(BTRFS_I(di_inode));
6324
6325		if (ret)
6326			break;
6327	}
6328
6329	ctx->log_new_dentries = orig_log_new_dentries;
6330	ctx->logging_new_delayed_dentries = false;
6331
6332	return ret;
6333}
6334
6335/* log a single inode in the tree log.
6336 * At least one parent directory for this inode must exist in the tree
6337 * or be logged already.
6338 *
6339 * Any items from this inode changed by the current transaction are copied
6340 * to the log tree.  An extra reference is taken on any extents in this
6341 * file, allowing us to avoid a whole pile of corner cases around logging
6342 * blocks that have been removed from the tree.
6343 *
6344 * See LOG_INODE_ALL and related defines for a description of what inode_only
6345 * does.
6346 *
6347 * This handles both files and directories.
6348 */
6349static int btrfs_log_inode(struct btrfs_trans_handle *trans,
6350			   struct btrfs_inode *inode,
6351			   int inode_only,
6352			   struct btrfs_log_ctx *ctx)
6353{
6354	struct btrfs_path *path;
6355	struct btrfs_path *dst_path;
6356	struct btrfs_key min_key;
6357	struct btrfs_key max_key;
6358	struct btrfs_root *log = inode->root->log_root;
6359	int ret;
6360	bool fast_search = false;
6361	u64 ino = btrfs_ino(inode);
6362	struct extent_map_tree *em_tree = &inode->extent_tree;
6363	u64 logged_isize = 0;
6364	bool need_log_inode_item = true;
6365	bool xattrs_logged = false;
6366	bool inode_item_dropped = true;
6367	bool full_dir_logging = false;
6368	LIST_HEAD(delayed_ins_list);
6369	LIST_HEAD(delayed_del_list);
6370
6371	path = btrfs_alloc_path();
6372	if (!path)
6373		return -ENOMEM;
6374	dst_path = btrfs_alloc_path();
6375	if (!dst_path) {
6376		btrfs_free_path(path);
6377		return -ENOMEM;
6378	}
6379
6380	min_key.objectid = ino;
6381	min_key.type = BTRFS_INODE_ITEM_KEY;
6382	min_key.offset = 0;
6383
6384	max_key.objectid = ino;
6385
6386
6387	/* today the code can only do partial logging of directories */
6388	if (S_ISDIR(inode->vfs_inode.i_mode) ||
6389	    (!test_bit(BTRFS_INODE_NEEDS_FULL_SYNC,
6390		       &inode->runtime_flags) &&
6391	     inode_only >= LOG_INODE_EXISTS))
6392		max_key.type = BTRFS_XATTR_ITEM_KEY;
6393	else
6394		max_key.type = (u8)-1;
6395	max_key.offset = (u64)-1;
6396
6397	if (S_ISDIR(inode->vfs_inode.i_mode) && inode_only == LOG_INODE_ALL)
6398		full_dir_logging = true;
6399
6400	/*
6401	 * If we are logging a directory while we are logging dentries of the
6402	 * delayed items of some other inode, then we need to flush the delayed
6403	 * items of this directory and not log the delayed items directly. This
6404	 * is to prevent more than one level of recursion into btrfs_log_inode()
6405	 * by having something like this:
6406	 *
6407	 *     $ mkdir -p a/b/c/d/e/f/g/h/...
6408	 *     $ xfs_io -c "fsync" a
6409	 *
6410	 * Where all directories in the path did not exist before and are
6411	 * created in the current transaction.
6412	 * So in such a case we directly log the delayed items of the main
6413	 * directory ("a") without flushing them first, while for each of its
6414	 * subdirectories we flush their delayed items before logging them.
6415	 * This prevents a potential unbounded recursion like this:
6416	 *
6417	 * btrfs_log_inode()
6418	 *   log_new_delayed_dentries()
6419	 *      btrfs_log_inode()
6420	 *        log_new_delayed_dentries()
6421	 *          btrfs_log_inode()
6422	 *            log_new_delayed_dentries()
6423	 *              (...)
6424	 *
6425	 * We have thresholds for the maximum number of delayed items to have in
6426	 * memory, and once they are hit, the items are flushed asynchronously.
6427	 * However the limit is quite high, so lets prevent deep levels of
6428	 * recursion to happen by limiting the maximum depth to be 1.
6429	 */
6430	if (full_dir_logging && ctx->logging_new_delayed_dentries) {
6431		ret = btrfs_commit_inode_delayed_items(trans, inode);
6432		if (ret)
6433			goto out;
6434	}
6435
6436	mutex_lock(&inode->log_mutex);
6437
6438	/*
6439	 * For symlinks, we must always log their content, which is stored in an
6440	 * inline extent, otherwise we could end up with an empty symlink after
6441	 * log replay, which is invalid on linux (symlink(2) returns -ENOENT if
6442	 * one attempts to create an empty symlink).
6443	 * We don't need to worry about flushing delalloc, because when we create
6444	 * the inline extent when the symlink is created (we never have delalloc
6445	 * for symlinks).
6446	 */
6447	if (S_ISLNK(inode->vfs_inode.i_mode))
6448		inode_only = LOG_INODE_ALL;
6449
6450	/*
6451	 * Before logging the inode item, cache the value returned by
6452	 * inode_logged(), because after that we have the need to figure out if
6453	 * the inode was previously logged in this transaction.
6454	 */
6455	ret = inode_logged(trans, inode, path);
6456	if (ret < 0)
6457		goto out_unlock;
6458	ctx->logged_before = (ret == 1);
6459	ret = 0;
6460
6461	/*
6462	 * This is for cases where logging a directory could result in losing a
6463	 * a file after replaying the log. For example, if we move a file from a
6464	 * directory A to a directory B, then fsync directory A, we have no way
6465	 * to known the file was moved from A to B, so logging just A would
6466	 * result in losing the file after a log replay.
6467	 */
6468	if (full_dir_logging && inode->last_unlink_trans >= trans->transid) {
6469		btrfs_set_log_full_commit(trans);
6470		ret = BTRFS_LOG_FORCE_COMMIT;
6471		goto out_unlock;
6472	}
6473
6474	/*
6475	 * a brute force approach to making sure we get the most uptodate
6476	 * copies of everything.
6477	 */
6478	if (S_ISDIR(inode->vfs_inode.i_mode)) {
6479		clear_bit(BTRFS_INODE_COPY_EVERYTHING, &inode->runtime_flags);
6480		if (ctx->logged_before)
6481			ret = drop_inode_items(trans, log, path, inode,
6482					       BTRFS_XATTR_ITEM_KEY);
6483	} else {
6484		if (inode_only == LOG_INODE_EXISTS && ctx->logged_before) {
6485			/*
6486			 * Make sure the new inode item we write to the log has
6487			 * the same isize as the current one (if it exists).
6488			 * This is necessary to prevent data loss after log
6489			 * replay, and also to prevent doing a wrong expanding
6490			 * truncate - for e.g. create file, write 4K into offset
6491			 * 0, fsync, write 4K into offset 4096, add hard link,
6492			 * fsync some other file (to sync log), power fail - if
6493			 * we use the inode's current i_size, after log replay
6494			 * we get a 8Kb file, with the last 4Kb extent as a hole
6495			 * (zeroes), as if an expanding truncate happened,
6496			 * instead of getting a file of 4Kb only.
6497			 */
6498			ret = logged_inode_size(log, inode, path, &logged_isize);
6499			if (ret)
6500				goto out_unlock;
6501		}
6502		if (test_bit(BTRFS_INODE_NEEDS_FULL_SYNC,
6503			     &inode->runtime_flags)) {
6504			if (inode_only == LOG_INODE_EXISTS) {
6505				max_key.type = BTRFS_XATTR_ITEM_KEY;
6506				if (ctx->logged_before)
6507					ret = drop_inode_items(trans, log, path,
6508							       inode, max_key.type);
6509			} else {
6510				clear_bit(BTRFS_INODE_NEEDS_FULL_SYNC,
6511					  &inode->runtime_flags);
6512				clear_bit(BTRFS_INODE_COPY_EVERYTHING,
6513					  &inode->runtime_flags);
6514				if (ctx->logged_before)
6515					ret = truncate_inode_items(trans, log,
6516								   inode, 0, 0);
6517			}
6518		} else if (test_and_clear_bit(BTRFS_INODE_COPY_EVERYTHING,
6519					      &inode->runtime_flags) ||
6520			   inode_only == LOG_INODE_EXISTS) {
6521			if (inode_only == LOG_INODE_ALL)
6522				fast_search = true;
6523			max_key.type = BTRFS_XATTR_ITEM_KEY;
6524			if (ctx->logged_before)
6525				ret = drop_inode_items(trans, log, path, inode,
6526						       max_key.type);
6527		} else {
6528			if (inode_only == LOG_INODE_ALL)
6529				fast_search = true;
6530			inode_item_dropped = false;
6531			goto log_extents;
6532		}
6533
6534	}
6535	if (ret)
6536		goto out_unlock;
6537
6538	/*
6539	 * If we are logging a directory in full mode, collect the delayed items
6540	 * before iterating the subvolume tree, so that we don't miss any new
6541	 * dir index items in case they get flushed while or right after we are
6542	 * iterating the subvolume tree.
6543	 */
6544	if (full_dir_logging && !ctx->logging_new_delayed_dentries)
6545		btrfs_log_get_delayed_items(inode, &delayed_ins_list,
6546					    &delayed_del_list);
6547
6548	ret = copy_inode_items_to_log(trans, inode, &min_key, &max_key,
6549				      path, dst_path, logged_isize,
6550				      inode_only, ctx,
6551				      &need_log_inode_item);
6552	if (ret)
6553		goto out_unlock;
6554
6555	btrfs_release_path(path);
6556	btrfs_release_path(dst_path);
6557	ret = btrfs_log_all_xattrs(trans, inode, path, dst_path);
6558	if (ret)
6559		goto out_unlock;
6560	xattrs_logged = true;
6561	if (max_key.type >= BTRFS_EXTENT_DATA_KEY && !fast_search) {
6562		btrfs_release_path(path);
6563		btrfs_release_path(dst_path);
6564		ret = btrfs_log_holes(trans, inode, path);
6565		if (ret)
6566			goto out_unlock;
6567	}
6568log_extents:
6569	btrfs_release_path(path);
6570	btrfs_release_path(dst_path);
6571	if (need_log_inode_item) {
6572		ret = log_inode_item(trans, log, dst_path, inode, inode_item_dropped);
6573		if (ret)
6574			goto out_unlock;
6575		/*
6576		 * If we are doing a fast fsync and the inode was logged before
6577		 * in this transaction, we don't need to log the xattrs because
6578		 * they were logged before. If xattrs were added, changed or
6579		 * deleted since the last time we logged the inode, then we have
6580		 * already logged them because the inode had the runtime flag
6581		 * BTRFS_INODE_COPY_EVERYTHING set.
6582		 */
6583		if (!xattrs_logged && inode->logged_trans < trans->transid) {
6584			ret = btrfs_log_all_xattrs(trans, inode, path, dst_path);
6585			if (ret)
6586				goto out_unlock;
6587			btrfs_release_path(path);
6588		}
6589	}
6590	if (fast_search) {
6591		ret = btrfs_log_changed_extents(trans, inode, dst_path, ctx);
6592		if (ret)
6593			goto out_unlock;
6594	} else if (inode_only == LOG_INODE_ALL) {
6595		struct extent_map *em, *n;
6596
6597		write_lock(&em_tree->lock);
6598		list_for_each_entry_safe(em, n, &em_tree->modified_extents, list)
6599			list_del_init(&em->list);
6600		write_unlock(&em_tree->lock);
6601	}
6602
6603	if (full_dir_logging) {
6604		ret = log_directory_changes(trans, inode, path, dst_path, ctx);
6605		if (ret)
6606			goto out_unlock;
6607		ret = log_delayed_insertion_items(trans, inode, path,
6608						  &delayed_ins_list, ctx);
6609		if (ret)
6610			goto out_unlock;
6611		ret = log_delayed_deletion_items(trans, inode, path,
6612						 &delayed_del_list, ctx);
6613		if (ret)
6614			goto out_unlock;
6615	}
6616
6617	spin_lock(&inode->lock);
6618	inode->logged_trans = trans->transid;
6619	/*
6620	 * Don't update last_log_commit if we logged that an inode exists.
6621	 * We do this for three reasons:
6622	 *
6623	 * 1) We might have had buffered writes to this inode that were
6624	 *    flushed and had their ordered extents completed in this
6625	 *    transaction, but we did not previously log the inode with
6626	 *    LOG_INODE_ALL. Later the inode was evicted and after that
6627	 *    it was loaded again and this LOG_INODE_EXISTS log operation
6628	 *    happened. We must make sure that if an explicit fsync against
6629	 *    the inode is performed later, it logs the new extents, an
6630	 *    updated inode item, etc, and syncs the log. The same logic
6631	 *    applies to direct IO writes instead of buffered writes.
6632	 *
6633	 * 2) When we log the inode with LOG_INODE_EXISTS, its inode item
6634	 *    is logged with an i_size of 0 or whatever value was logged
6635	 *    before. If later the i_size of the inode is increased by a
6636	 *    truncate operation, the log is synced through an fsync of
6637	 *    some other inode and then finally an explicit fsync against
6638	 *    this inode is made, we must make sure this fsync logs the
6639	 *    inode with the new i_size, the hole between old i_size and
6640	 *    the new i_size, and syncs the log.
6641	 *
6642	 * 3) If we are logging that an ancestor inode exists as part of
6643	 *    logging a new name from a link or rename operation, don't update
6644	 *    its last_log_commit - otherwise if an explicit fsync is made
6645	 *    against an ancestor, the fsync considers the inode in the log
6646	 *    and doesn't sync the log, resulting in the ancestor missing after
6647	 *    a power failure unless the log was synced as part of an fsync
6648	 *    against any other unrelated inode.
6649	 */
6650	if (inode_only != LOG_INODE_EXISTS)
6651		inode->last_log_commit = inode->last_sub_trans;
6652	spin_unlock(&inode->lock);
6653
6654	/*
6655	 * Reset the last_reflink_trans so that the next fsync does not need to
6656	 * go through the slower path when logging extents and their checksums.
6657	 */
6658	if (inode_only == LOG_INODE_ALL)
6659		inode->last_reflink_trans = 0;
6660
6661out_unlock:
6662	mutex_unlock(&inode->log_mutex);
6663out:
6664	btrfs_free_path(path);
6665	btrfs_free_path(dst_path);
6666
6667	if (ret)
6668		free_conflicting_inodes(ctx);
6669	else
6670		ret = log_conflicting_inodes(trans, inode->root, ctx);
6671
6672	if (full_dir_logging && !ctx->logging_new_delayed_dentries) {
6673		if (!ret)
6674			ret = log_new_delayed_dentries(trans, inode,
6675						       &delayed_ins_list, ctx);
6676
6677		btrfs_log_put_delayed_items(inode, &delayed_ins_list,
6678					    &delayed_del_list);
6679	}
6680
6681	return ret;
6682}
6683
6684static int btrfs_log_all_parents(struct btrfs_trans_handle *trans,
6685				 struct btrfs_inode *inode,
6686				 struct btrfs_log_ctx *ctx)
6687{
6688	struct btrfs_fs_info *fs_info = trans->fs_info;
6689	int ret;
6690	struct btrfs_path *path;
6691	struct btrfs_key key;
6692	struct btrfs_root *root = inode->root;
6693	const u64 ino = btrfs_ino(inode);
6694
6695	path = btrfs_alloc_path();
6696	if (!path)
6697		return -ENOMEM;
6698	path->skip_locking = 1;
6699	path->search_commit_root = 1;
6700
6701	key.objectid = ino;
6702	key.type = BTRFS_INODE_REF_KEY;
6703	key.offset = 0;
6704	ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
6705	if (ret < 0)
6706		goto out;
6707
6708	while (true) {
6709		struct extent_buffer *leaf = path->nodes[0];
6710		int slot = path->slots[0];
6711		u32 cur_offset = 0;
6712		u32 item_size;
6713		unsigned long ptr;
6714
6715		if (slot >= btrfs_header_nritems(leaf)) {
6716			ret = btrfs_next_leaf(root, path);
6717			if (ret < 0)
6718				goto out;
6719			else if (ret > 0)
6720				break;
6721			continue;
6722		}
6723
6724		btrfs_item_key_to_cpu(leaf, &key, slot);
6725		/* BTRFS_INODE_EXTREF_KEY is BTRFS_INODE_REF_KEY + 1 */
6726		if (key.objectid != ino || key.type > BTRFS_INODE_EXTREF_KEY)
6727			break;
6728
6729		item_size = btrfs_item_size(leaf, slot);
6730		ptr = btrfs_item_ptr_offset(leaf, slot);
6731		while (cur_offset < item_size) {
6732			struct btrfs_key inode_key;
6733			struct inode *dir_inode;
6734
6735			inode_key.type = BTRFS_INODE_ITEM_KEY;
6736			inode_key.offset = 0;
6737
6738			if (key.type == BTRFS_INODE_EXTREF_KEY) {
6739				struct btrfs_inode_extref *extref;
6740
6741				extref = (struct btrfs_inode_extref *)
6742					(ptr + cur_offset);
6743				inode_key.objectid = btrfs_inode_extref_parent(
6744					leaf, extref);
6745				cur_offset += sizeof(*extref);
6746				cur_offset += btrfs_inode_extref_name_len(leaf,
6747					extref);
6748			} else {
6749				inode_key.objectid = key.offset;
6750				cur_offset = item_size;
6751			}
6752
6753			dir_inode = btrfs_iget(fs_info->sb, inode_key.objectid,
6754					       root);
6755			/*
6756			 * If the parent inode was deleted, return an error to
6757			 * fallback to a transaction commit. This is to prevent
6758			 * getting an inode that was moved from one parent A to
6759			 * a parent B, got its former parent A deleted and then
6760			 * it got fsync'ed, from existing at both parents after
6761			 * a log replay (and the old parent still existing).
6762			 * Example:
6763			 *
6764			 * mkdir /mnt/A
6765			 * mkdir /mnt/B
6766			 * touch /mnt/B/bar
6767			 * sync
6768			 * mv /mnt/B/bar /mnt/A/bar
6769			 * mv -T /mnt/A /mnt/B
6770			 * fsync /mnt/B/bar
6771			 * <power fail>
6772			 *
6773			 * If we ignore the old parent B which got deleted,
6774			 * after a log replay we would have file bar linked
6775			 * at both parents and the old parent B would still
6776			 * exist.
6777			 */
6778			if (IS_ERR(dir_inode)) {
6779				ret = PTR_ERR(dir_inode);
6780				goto out;
6781			}
6782
6783			if (!need_log_inode(trans, BTRFS_I(dir_inode))) {
6784				btrfs_add_delayed_iput(BTRFS_I(dir_inode));
6785				continue;
6786			}
6787
6788			ctx->log_new_dentries = false;
6789			ret = btrfs_log_inode(trans, BTRFS_I(dir_inode),
6790					      LOG_INODE_ALL, ctx);
6791			if (!ret && ctx->log_new_dentries)
6792				ret = log_new_dir_dentries(trans,
6793						   BTRFS_I(dir_inode), ctx);
6794			btrfs_add_delayed_iput(BTRFS_I(dir_inode));
6795			if (ret)
6796				goto out;
6797		}
6798		path->slots[0]++;
6799	}
6800	ret = 0;
6801out:
6802	btrfs_free_path(path);
6803	return ret;
6804}
6805
6806static int log_new_ancestors(struct btrfs_trans_handle *trans,
6807			     struct btrfs_root *root,
6808			     struct btrfs_path *path,
6809			     struct btrfs_log_ctx *ctx)
6810{
6811	struct btrfs_key found_key;
6812
6813	btrfs_item_key_to_cpu(path->nodes[0], &found_key, path->slots[0]);
6814
6815	while (true) {
6816		struct btrfs_fs_info *fs_info = root->fs_info;
6817		struct extent_buffer *leaf = path->nodes[0];
6818		int slot = path->slots[0];
6819		struct btrfs_key search_key;
6820		struct inode *inode;
6821		u64 ino;
6822		int ret = 0;
6823
6824		btrfs_release_path(path);
6825
6826		ino = found_key.offset;
6827
6828		search_key.objectid = found_key.offset;
6829		search_key.type = BTRFS_INODE_ITEM_KEY;
6830		search_key.offset = 0;
6831		inode = btrfs_iget(fs_info->sb, ino, root);
6832		if (IS_ERR(inode))
6833			return PTR_ERR(inode);
6834
6835		if (BTRFS_I(inode)->generation >= trans->transid &&
6836		    need_log_inode(trans, BTRFS_I(inode)))
6837			ret = btrfs_log_inode(trans, BTRFS_I(inode),
6838					      LOG_INODE_EXISTS, ctx);
6839		btrfs_add_delayed_iput(BTRFS_I(inode));
6840		if (ret)
6841			return ret;
6842
6843		if (search_key.objectid == BTRFS_FIRST_FREE_OBJECTID)
6844			break;
6845
6846		search_key.type = BTRFS_INODE_REF_KEY;
6847		ret = btrfs_search_slot(NULL, root, &search_key, path, 0, 0);
6848		if (ret < 0)
6849			return ret;
6850
6851		leaf = path->nodes[0];
6852		slot = path->slots[0];
6853		if (slot >= btrfs_header_nritems(leaf)) {
6854			ret = btrfs_next_leaf(root, path);
6855			if (ret < 0)
6856				return ret;
6857			else if (ret > 0)
6858				return -ENOENT;
6859			leaf = path->nodes[0];
6860			slot = path->slots[0];
6861		}
6862
6863		btrfs_item_key_to_cpu(leaf, &found_key, slot);
6864		if (found_key.objectid != search_key.objectid ||
6865		    found_key.type != BTRFS_INODE_REF_KEY)
6866			return -ENOENT;
6867	}
6868	return 0;
6869}
6870
6871static int log_new_ancestors_fast(struct btrfs_trans_handle *trans,
6872				  struct btrfs_inode *inode,
6873				  struct dentry *parent,
6874				  struct btrfs_log_ctx *ctx)
6875{
6876	struct btrfs_root *root = inode->root;
6877	struct dentry *old_parent = NULL;
6878	struct super_block *sb = inode->vfs_inode.i_sb;
6879	int ret = 0;
6880
6881	while (true) {
6882		if (!parent || d_really_is_negative(parent) ||
6883		    sb != parent->d_sb)
6884			break;
6885
6886		inode = BTRFS_I(d_inode(parent));
6887		if (root != inode->root)
6888			break;
6889
6890		if (inode->generation >= trans->transid &&
6891		    need_log_inode(trans, inode)) {
6892			ret = btrfs_log_inode(trans, inode,
6893					      LOG_INODE_EXISTS, ctx);
6894			if (ret)
6895				break;
6896		}
6897		if (IS_ROOT(parent))
6898			break;
6899
6900		parent = dget_parent(parent);
6901		dput(old_parent);
6902		old_parent = parent;
6903	}
6904	dput(old_parent);
6905
6906	return ret;
6907}
6908
6909static int log_all_new_ancestors(struct btrfs_trans_handle *trans,
6910				 struct btrfs_inode *inode,
6911				 struct dentry *parent,
6912				 struct btrfs_log_ctx *ctx)
6913{
6914	struct btrfs_root *root = inode->root;
6915	const u64 ino = btrfs_ino(inode);
6916	struct btrfs_path *path;
6917	struct btrfs_key search_key;
6918	int ret;
6919
6920	/*
6921	 * For a single hard link case, go through a fast path that does not
6922	 * need to iterate the fs/subvolume tree.
6923	 */
6924	if (inode->vfs_inode.i_nlink < 2)
6925		return log_new_ancestors_fast(trans, inode, parent, ctx);
6926
6927	path = btrfs_alloc_path();
6928	if (!path)
6929		return -ENOMEM;
6930
6931	search_key.objectid = ino;
6932	search_key.type = BTRFS_INODE_REF_KEY;
6933	search_key.offset = 0;
6934again:
6935	ret = btrfs_search_slot(NULL, root, &search_key, path, 0, 0);
6936	if (ret < 0)
6937		goto out;
6938	if (ret == 0)
6939		path->slots[0]++;
6940
6941	while (true) {
6942		struct extent_buffer *leaf = path->nodes[0];
6943		int slot = path->slots[0];
6944		struct btrfs_key found_key;
6945
6946		if (slot >= btrfs_header_nritems(leaf)) {
6947			ret = btrfs_next_leaf(root, path);
6948			if (ret < 0)
6949				goto out;
6950			else if (ret > 0)
6951				break;
6952			continue;
6953		}
6954
6955		btrfs_item_key_to_cpu(leaf, &found_key, slot);
6956		if (found_key.objectid != ino ||
6957		    found_key.type > BTRFS_INODE_EXTREF_KEY)
6958			break;
6959
6960		/*
6961		 * Don't deal with extended references because they are rare
6962		 * cases and too complex to deal with (we would need to keep
6963		 * track of which subitem we are processing for each item in
6964		 * this loop, etc). So just return some error to fallback to
6965		 * a transaction commit.
6966		 */
6967		if (found_key.type == BTRFS_INODE_EXTREF_KEY) {
6968			ret = -EMLINK;
6969			goto out;
6970		}
6971
6972		/*
6973		 * Logging ancestors needs to do more searches on the fs/subvol
6974		 * tree, so it releases the path as needed to avoid deadlocks.
6975		 * Keep track of the last inode ref key and resume from that key
6976		 * after logging all new ancestors for the current hard link.
6977		 */
6978		memcpy(&search_key, &found_key, sizeof(search_key));
6979
6980		ret = log_new_ancestors(trans, root, path, ctx);
6981		if (ret)
6982			goto out;
6983		btrfs_release_path(path);
6984		goto again;
6985	}
6986	ret = 0;
6987out:
6988	btrfs_free_path(path);
6989	return ret;
6990}
6991
6992/*
6993 * helper function around btrfs_log_inode to make sure newly created
6994 * parent directories also end up in the log.  A minimal inode and backref
6995 * only logging is done of any parent directories that are older than
6996 * the last committed transaction
6997 */
6998static int btrfs_log_inode_parent(struct btrfs_trans_handle *trans,
6999				  struct btrfs_inode *inode,
7000				  struct dentry *parent,
7001				  int inode_only,
7002				  struct btrfs_log_ctx *ctx)
7003{
7004	struct btrfs_root *root = inode->root;
7005	struct btrfs_fs_info *fs_info = root->fs_info;
7006	int ret = 0;
7007	bool log_dentries = false;
7008
7009	if (btrfs_test_opt(fs_info, NOTREELOG)) {
7010		ret = BTRFS_LOG_FORCE_COMMIT;
7011		goto end_no_trans;
7012	}
7013
7014	if (btrfs_root_refs(&root->root_item) == 0) {
7015		ret = BTRFS_LOG_FORCE_COMMIT;
7016		goto end_no_trans;
7017	}
7018
7019	/*
7020	 * Skip already logged inodes or inodes corresponding to tmpfiles
7021	 * (since logging them is pointless, a link count of 0 means they
7022	 * will never be accessible).
7023	 */
7024	if ((btrfs_inode_in_log(inode, trans->transid) &&
7025	     list_empty(&ctx->ordered_extents)) ||
7026	    inode->vfs_inode.i_nlink == 0) {
7027		ret = BTRFS_NO_LOG_SYNC;
7028		goto end_no_trans;
7029	}
7030
7031	ret = start_log_trans(trans, root, ctx);
7032	if (ret)
7033		goto end_no_trans;
7034
7035	ret = btrfs_log_inode(trans, inode, inode_only, ctx);
7036	if (ret)
7037		goto end_trans;
7038
7039	/*
7040	 * for regular files, if its inode is already on disk, we don't
7041	 * have to worry about the parents at all.  This is because
7042	 * we can use the last_unlink_trans field to record renames
7043	 * and other fun in this file.
7044	 */
7045	if (S_ISREG(inode->vfs_inode.i_mode) &&
7046	    inode->generation < trans->transid &&
7047	    inode->last_unlink_trans < trans->transid) {
7048		ret = 0;
7049		goto end_trans;
7050	}
7051
7052	if (S_ISDIR(inode->vfs_inode.i_mode) && ctx->log_new_dentries)
7053		log_dentries = true;
7054
7055	/*
7056	 * On unlink we must make sure all our current and old parent directory
7057	 * inodes are fully logged. This is to prevent leaving dangling
7058	 * directory index entries in directories that were our parents but are
7059	 * not anymore. Not doing this results in old parent directory being
7060	 * impossible to delete after log replay (rmdir will always fail with
7061	 * error -ENOTEMPTY).
7062	 *
7063	 * Example 1:
7064	 *
7065	 * mkdir testdir
7066	 * touch testdir/foo
7067	 * ln testdir/foo testdir/bar
7068	 * sync
7069	 * unlink testdir/bar
7070	 * xfs_io -c fsync testdir/foo
7071	 * <power failure>
7072	 * mount fs, triggers log replay
7073	 *
7074	 * If we don't log the parent directory (testdir), after log replay the
7075	 * directory still has an entry pointing to the file inode using the bar
7076	 * name, but a matching BTRFS_INODE_[REF|EXTREF]_KEY does not exist and
7077	 * the file inode has a link count of 1.
7078	 *
7079	 * Example 2:
7080	 *
7081	 * mkdir testdir
7082	 * touch foo
7083	 * ln foo testdir/foo2
7084	 * ln foo testdir/foo3
7085	 * sync
7086	 * unlink testdir/foo3
7087	 * xfs_io -c fsync foo
7088	 * <power failure>
7089	 * mount fs, triggers log replay
7090	 *
7091	 * Similar as the first example, after log replay the parent directory
7092	 * testdir still has an entry pointing to the inode file with name foo3
7093	 * but the file inode does not have a matching BTRFS_INODE_REF_KEY item
7094	 * and has a link count of 2.
7095	 */
7096	if (inode->last_unlink_trans >= trans->transid) {
7097		ret = btrfs_log_all_parents(trans, inode, ctx);
7098		if (ret)
7099			goto end_trans;
7100	}
7101
7102	ret = log_all_new_ancestors(trans, inode, parent, ctx);
7103	if (ret)
7104		goto end_trans;
7105
7106	if (log_dentries)
7107		ret = log_new_dir_dentries(trans, inode, ctx);
7108	else
7109		ret = 0;
7110end_trans:
7111	if (ret < 0) {
7112		btrfs_set_log_full_commit(trans);
7113		ret = BTRFS_LOG_FORCE_COMMIT;
7114	}
7115
7116	if (ret)
7117		btrfs_remove_log_ctx(root, ctx);
7118	btrfs_end_log_trans(root);
7119end_no_trans:
7120	return ret;
7121}
7122
7123/*
7124 * it is not safe to log dentry if the chunk root has added new
7125 * chunks.  This returns 0 if the dentry was logged, and 1 otherwise.
7126 * If this returns 1, you must commit the transaction to safely get your
7127 * data on disk.
7128 */
7129int btrfs_log_dentry_safe(struct btrfs_trans_handle *trans,
7130			  struct dentry *dentry,
7131			  struct btrfs_log_ctx *ctx)
7132{
7133	struct dentry *parent = dget_parent(dentry);
7134	int ret;
7135
7136	ret = btrfs_log_inode_parent(trans, BTRFS_I(d_inode(dentry)), parent,
7137				     LOG_INODE_ALL, ctx);
7138	dput(parent);
7139
7140	return ret;
7141}
7142
7143/*
7144 * should be called during mount to recover any replay any log trees
7145 * from the FS
7146 */
7147int btrfs_recover_log_trees(struct btrfs_root *log_root_tree)
7148{
7149	int ret;
7150	struct btrfs_path *path;
7151	struct btrfs_trans_handle *trans;
7152	struct btrfs_key key;
7153	struct btrfs_key found_key;
7154	struct btrfs_root *log;
7155	struct btrfs_fs_info *fs_info = log_root_tree->fs_info;
7156	struct walk_control wc = {
7157		.process_func = process_one_buffer,
7158		.stage = LOG_WALK_PIN_ONLY,
7159	};
7160
7161	path = btrfs_alloc_path();
7162	if (!path)
7163		return -ENOMEM;
7164
7165	set_bit(BTRFS_FS_LOG_RECOVERING, &fs_info->flags);
7166
7167	trans = btrfs_start_transaction(fs_info->tree_root, 0);
7168	if (IS_ERR(trans)) {
7169		ret = PTR_ERR(trans);
7170		goto error;
7171	}
7172
7173	wc.trans = trans;
7174	wc.pin = 1;
7175
7176	ret = walk_log_tree(trans, log_root_tree, &wc);
7177	if (ret) {
7178		btrfs_abort_transaction(trans, ret);
7179		goto error;
7180	}
7181
7182again:
7183	key.objectid = BTRFS_TREE_LOG_OBJECTID;
7184	key.offset = (u64)-1;
7185	key.type = BTRFS_ROOT_ITEM_KEY;
7186
7187	while (1) {
7188		ret = btrfs_search_slot(NULL, log_root_tree, &key, path, 0, 0);
7189
7190		if (ret < 0) {
7191			btrfs_abort_transaction(trans, ret);
7192			goto error;
7193		}
7194		if (ret > 0) {
7195			if (path->slots[0] == 0)
7196				break;
7197			path->slots[0]--;
7198		}
7199		btrfs_item_key_to_cpu(path->nodes[0], &found_key,
7200				      path->slots[0]);
7201		btrfs_release_path(path);
7202		if (found_key.objectid != BTRFS_TREE_LOG_OBJECTID)
7203			break;
7204
7205		log = btrfs_read_tree_root(log_root_tree, &found_key);
7206		if (IS_ERR(log)) {
7207			ret = PTR_ERR(log);
7208			btrfs_abort_transaction(trans, ret);
7209			goto error;
7210		}
7211
7212		wc.replay_dest = btrfs_get_fs_root(fs_info, found_key.offset,
7213						   true);
7214		if (IS_ERR(wc.replay_dest)) {
7215			ret = PTR_ERR(wc.replay_dest);
7216
7217			/*
7218			 * We didn't find the subvol, likely because it was
7219			 * deleted.  This is ok, simply skip this log and go to
7220			 * the next one.
7221			 *
7222			 * We need to exclude the root because we can't have
7223			 * other log replays overwriting this log as we'll read
7224			 * it back in a few more times.  This will keep our
7225			 * block from being modified, and we'll just bail for
7226			 * each subsequent pass.
7227			 */
7228			if (ret == -ENOENT)
7229				ret = btrfs_pin_extent_for_log_replay(trans,
7230							log->node->start,
7231							log->node->len);
7232			btrfs_put_root(log);
7233
7234			if (!ret)
7235				goto next;
7236			btrfs_abort_transaction(trans, ret);
7237			goto error;
7238		}
7239
7240		wc.replay_dest->log_root = log;
7241		ret = btrfs_record_root_in_trans(trans, wc.replay_dest);
7242		if (ret)
7243			/* The loop needs to continue due to the root refs */
7244			btrfs_abort_transaction(trans, ret);
7245		else
7246			ret = walk_log_tree(trans, log, &wc);
7247
7248		if (!ret && wc.stage == LOG_WALK_REPLAY_ALL) {
7249			ret = fixup_inode_link_counts(trans, wc.replay_dest,
7250						      path);
7251			if (ret)
7252				btrfs_abort_transaction(trans, ret);
7253		}
7254
7255		if (!ret && wc.stage == LOG_WALK_REPLAY_ALL) {
7256			struct btrfs_root *root = wc.replay_dest;
7257
7258			btrfs_release_path(path);
7259
7260			/*
7261			 * We have just replayed everything, and the highest
7262			 * objectid of fs roots probably has changed in case
7263			 * some inode_item's got replayed.
7264			 *
7265			 * root->objectid_mutex is not acquired as log replay
7266			 * could only happen during mount.
7267			 */
7268			ret = btrfs_init_root_free_objectid(root);
7269			if (ret)
7270				btrfs_abort_transaction(trans, ret);
7271		}
7272
7273		wc.replay_dest->log_root = NULL;
7274		btrfs_put_root(wc.replay_dest);
7275		btrfs_put_root(log);
7276
7277		if (ret)
7278			goto error;
7279next:
7280		if (found_key.offset == 0)
7281			break;
7282		key.offset = found_key.offset - 1;
7283	}
7284	btrfs_release_path(path);
7285
7286	/* step one is to pin it all, step two is to replay just inodes */
7287	if (wc.pin) {
7288		wc.pin = 0;
7289		wc.process_func = replay_one_buffer;
7290		wc.stage = LOG_WALK_REPLAY_INODES;
7291		goto again;
7292	}
7293	/* step three is to replay everything */
7294	if (wc.stage < LOG_WALK_REPLAY_ALL) {
7295		wc.stage++;
7296		goto again;
7297	}
7298
7299	btrfs_free_path(path);
7300
7301	/* step 4: commit the transaction, which also unpins the blocks */
7302	ret = btrfs_commit_transaction(trans);
7303	if (ret)
7304		return ret;
7305
7306	log_root_tree->log_root = NULL;
7307	clear_bit(BTRFS_FS_LOG_RECOVERING, &fs_info->flags);
7308	btrfs_put_root(log_root_tree);
7309
7310	return 0;
7311error:
7312	if (wc.trans)
7313		btrfs_end_transaction(wc.trans);
7314	clear_bit(BTRFS_FS_LOG_RECOVERING, &fs_info->flags);
7315	btrfs_free_path(path);
7316	return ret;
7317}
7318
7319/*
7320 * there are some corner cases where we want to force a full
7321 * commit instead of allowing a directory to be logged.
7322 *
7323 * They revolve around files there were unlinked from the directory, and
7324 * this function updates the parent directory so that a full commit is
7325 * properly done if it is fsync'd later after the unlinks are done.
7326 *
7327 * Must be called before the unlink operations (updates to the subvolume tree,
7328 * inodes, etc) are done.
7329 */
7330void btrfs_record_unlink_dir(struct btrfs_trans_handle *trans,
7331			     struct btrfs_inode *dir, struct btrfs_inode *inode,
7332			     int for_rename)
7333{
7334	/*
7335	 * when we're logging a file, if it hasn't been renamed
7336	 * or unlinked, and its inode is fully committed on disk,
7337	 * we don't have to worry about walking up the directory chain
7338	 * to log its parents.
7339	 *
7340	 * So, we use the last_unlink_trans field to put this transid
7341	 * into the file.  When the file is logged we check it and
7342	 * don't log the parents if the file is fully on disk.
7343	 */
7344	mutex_lock(&inode->log_mutex);
7345	inode->last_unlink_trans = trans->transid;
7346	mutex_unlock(&inode->log_mutex);
7347
 
 
 
7348	/*
7349	 * if this directory was already logged any new
7350	 * names for this file/dir will get recorded
 
 
7351	 */
7352	if (dir->logged_trans == trans->transid)
7353		return;
7354
7355	/*
7356	 * if the inode we're about to unlink was logged,
7357	 * the log will be properly updated for any new names
 
 
7358	 */
7359	if (inode->logged_trans == trans->transid)
7360		return;
7361
7362	/*
7363	 * when renaming files across directories, if the directory
7364	 * there we're unlinking from gets fsync'd later on, there's
7365	 * no way to find the destination directory later and fsync it
7366	 * properly.  So, we have to be conservative and force commits
7367	 * so the new name gets discovered.
7368	 */
7369	if (for_rename)
7370		goto record;
7371
7372	/* we can safely do the unlink without any special recording */
7373	return;
7374
7375record:
7376	mutex_lock(&dir->log_mutex);
7377	dir->last_unlink_trans = trans->transid;
7378	mutex_unlock(&dir->log_mutex);
7379}
7380
7381/*
7382 * Make sure that if someone attempts to fsync the parent directory of a deleted
7383 * snapshot, it ends up triggering a transaction commit. This is to guarantee
7384 * that after replaying the log tree of the parent directory's root we will not
7385 * see the snapshot anymore and at log replay time we will not see any log tree
7386 * corresponding to the deleted snapshot's root, which could lead to replaying
7387 * it after replaying the log tree of the parent directory (which would replay
7388 * the snapshot delete operation).
7389 *
7390 * Must be called before the actual snapshot destroy operation (updates to the
7391 * parent root and tree of tree roots trees, etc) are done.
7392 */
7393void btrfs_record_snapshot_destroy(struct btrfs_trans_handle *trans,
7394				   struct btrfs_inode *dir)
7395{
7396	mutex_lock(&dir->log_mutex);
7397	dir->last_unlink_trans = trans->transid;
7398	mutex_unlock(&dir->log_mutex);
7399}
7400
7401/*
7402 * Update the log after adding a new name for an inode.
7403 *
7404 * @trans:              Transaction handle.
7405 * @old_dentry:         The dentry associated with the old name and the old
7406 *                      parent directory.
7407 * @old_dir:            The inode of the previous parent directory for the case
7408 *                      of a rename. For a link operation, it must be NULL.
7409 * @old_dir_index:      The index number associated with the old name, meaningful
7410 *                      only for rename operations (when @old_dir is not NULL).
7411 *                      Ignored for link operations.
7412 * @parent:             The dentry associated with the directory under which the
7413 *                      new name is located.
7414 *
7415 * Call this after adding a new name for an inode, as a result of a link or
7416 * rename operation, and it will properly update the log to reflect the new name.
7417 */
7418void btrfs_log_new_name(struct btrfs_trans_handle *trans,
7419			struct dentry *old_dentry, struct btrfs_inode *old_dir,
7420			u64 old_dir_index, struct dentry *parent)
7421{
7422	struct btrfs_inode *inode = BTRFS_I(d_inode(old_dentry));
7423	struct btrfs_root *root = inode->root;
7424	struct btrfs_log_ctx ctx;
7425	bool log_pinned = false;
7426	int ret;
7427
7428	/*
7429	 * this will force the logging code to walk the dentry chain
7430	 * up for the file
7431	 */
7432	if (!S_ISDIR(inode->vfs_inode.i_mode))
7433		inode->last_unlink_trans = trans->transid;
7434
7435	/*
7436	 * if this inode hasn't been logged and directory we're renaming it
7437	 * from hasn't been logged, we don't need to log it
7438	 */
7439	ret = inode_logged(trans, inode, NULL);
7440	if (ret < 0) {
7441		goto out;
7442	} else if (ret == 0) {
7443		if (!old_dir)
7444			return;
7445		/*
7446		 * If the inode was not logged and we are doing a rename (old_dir is not
7447		 * NULL), check if old_dir was logged - if it was not we can return and
7448		 * do nothing.
7449		 */
7450		ret = inode_logged(trans, old_dir, NULL);
7451		if (ret < 0)
7452			goto out;
7453		else if (ret == 0)
7454			return;
7455	}
7456	ret = 0;
7457
7458	/*
7459	 * If we are doing a rename (old_dir is not NULL) from a directory that
7460	 * was previously logged, make sure that on log replay we get the old
7461	 * dir entry deleted. This is needed because we will also log the new
7462	 * name of the renamed inode, so we need to make sure that after log
7463	 * replay we don't end up with both the new and old dir entries existing.
7464	 */
7465	if (old_dir && old_dir->logged_trans == trans->transid) {
7466		struct btrfs_root *log = old_dir->root->log_root;
7467		struct btrfs_path *path;
7468		struct fscrypt_name fname;
7469
7470		ASSERT(old_dir_index >= BTRFS_DIR_START_INDEX);
7471
7472		ret = fscrypt_setup_filename(&old_dir->vfs_inode,
7473					     &old_dentry->d_name, 0, &fname);
7474		if (ret)
7475			goto out;
7476		/*
7477		 * We have two inodes to update in the log, the old directory and
7478		 * the inode that got renamed, so we must pin the log to prevent
7479		 * anyone from syncing the log until we have updated both inodes
7480		 * in the log.
7481		 */
7482		ret = join_running_log_trans(root);
7483		/*
7484		 * At least one of the inodes was logged before, so this should
7485		 * not fail, but if it does, it's not serious, just bail out and
7486		 * mark the log for a full commit.
7487		 */
7488		if (WARN_ON_ONCE(ret < 0)) {
7489			fscrypt_free_filename(&fname);
7490			goto out;
7491		}
7492
7493		log_pinned = true;
7494
7495		path = btrfs_alloc_path();
7496		if (!path) {
7497			ret = -ENOMEM;
7498			fscrypt_free_filename(&fname);
7499			goto out;
7500		}
7501
7502		/*
7503		 * Other concurrent task might be logging the old directory,
7504		 * as it can be triggered when logging other inode that had or
7505		 * still has a dentry in the old directory. We lock the old
7506		 * directory's log_mutex to ensure the deletion of the old
7507		 * name is persisted, because during directory logging we
7508		 * delete all BTRFS_DIR_LOG_INDEX_KEY keys and the deletion of
7509		 * the old name's dir index item is in the delayed items, so
7510		 * it could be missed by an in progress directory logging.
7511		 */
7512		mutex_lock(&old_dir->log_mutex);
7513		ret = del_logged_dentry(trans, log, path, btrfs_ino(old_dir),
7514					&fname.disk_name, old_dir_index);
7515		if (ret > 0) {
7516			/*
7517			 * The dentry does not exist in the log, so record its
7518			 * deletion.
7519			 */
7520			btrfs_release_path(path);
7521			ret = insert_dir_log_key(trans, log, path,
7522						 btrfs_ino(old_dir),
7523						 old_dir_index, old_dir_index);
7524		}
7525		mutex_unlock(&old_dir->log_mutex);
7526
7527		btrfs_free_path(path);
7528		fscrypt_free_filename(&fname);
7529		if (ret < 0)
7530			goto out;
7531	}
7532
7533	btrfs_init_log_ctx(&ctx, &inode->vfs_inode);
7534	ctx.logging_new_name = true;
 
7535	/*
7536	 * We don't care about the return value. If we fail to log the new name
7537	 * then we know the next attempt to sync the log will fallback to a full
7538	 * transaction commit (due to a call to btrfs_set_log_full_commit()), so
7539	 * we don't need to worry about getting a log committed that has an
7540	 * inconsistent state after a rename operation.
7541	 */
7542	btrfs_log_inode_parent(trans, inode, parent, LOG_INODE_EXISTS, &ctx);
 
7543	ASSERT(list_empty(&ctx.conflict_inodes));
7544out:
7545	/*
7546	 * If an error happened mark the log for a full commit because it's not
7547	 * consistent and up to date or we couldn't find out if one of the
7548	 * inodes was logged before in this transaction. Do it before unpinning
7549	 * the log, to avoid any races with someone else trying to commit it.
7550	 */
7551	if (ret < 0)
7552		btrfs_set_log_full_commit(trans);
7553	if (log_pinned)
7554		btrfs_end_log_trans(root);
7555}
7556