inode.c

/*
 * Copyright (C) 2007 Oracle.  All rights reserved.
 *
 * This program is free software; you can redistribute it and/or
 * modify it under the terms of the GNU General Public
 * License v2 as published by the Free Software Foundation.
 *
 * This program is distributed in the hope that it will be useful,
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
 * General Public License for more details.
 *
 * You should have received a copy of the GNU General Public
 * License along with this program; if not, write to the
 * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
 * Boston, MA 021110-1307, USA.
 */

#include <linux/kernel.h>
#include <linux/bio.h>
#include <linux/buffer_head.h>
#include <linux/file.h>
#include <linux/fs.h>
#include <linux/pagemap.h>
#include <linux/highmem.h>
#include <linux/time.h>
#include <linux/init.h>
#include <linux/string.h>
#include <linux/backing-dev.h>
#include <linux/mpage.h>
#include <linux/swap.h>
#include <linux/writeback.h>
#include <linux/statfs.h>
#include <linux/compat.h>
#include <linux/bit_spinlock.h>
#include <linux/xattr.h>
#include <linux/posix_acl.h>
#include <linux/falloc.h>
#include "compat.h"
#include "ctree.h"
#include "disk-io.h"
#include "transaction.h"
#include "btrfs_inode.h"
#include "ioctl.h"
#include "print-tree.h"
#include "volumes.h"
#include "ordered-data.h"
#include "xattr.h"
#include "tree-log.h"
#include "compression.h"
#include "locking.h"

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

static const struct inode_operations btrfs_dir_inode_operations;
static const struct inode_operations btrfs_symlink_inode_operations;
static const struct inode_operations btrfs_dir_ro_inode_operations;
static const struct inode_operations btrfs_special_inode_operations;
static const struct inode_operations btrfs_file_inode_operations;
static const struct address_space_operations btrfs_aops;
static const struct address_space_operations btrfs_symlink_aops;
static const struct file_operations btrfs_dir_file_operations;
static struct extent_io_ops btrfs_extent_io_ops;

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

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

static void btrfs_truncate(struct inode *inode);
static int btrfs_finish_ordered_io(struct inode *inode, u64 start, u64 end);
static noinline int cow_file_range(struct inode *inode,
				   struct page *locked_page,
				   u64 start, u64 end, int *page_started,
				   unsigned long *nr_written, int unlock);

static int btrfs_init_inode_security(struct btrfs_trans_handle *trans,
				     struct inode *inode,  struct inode *dir)
{
	int err;

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

/*
 * this does all the hard work for inserting an inline extent into
 * the btree.  The caller should have done a btrfs_drop_extents so that
 * no overlapping inline items exist in the btree
 */
static noinline int insert_inline_extent(struct btrfs_trans_handle *trans,
				struct btrfs_root *root, struct inode *inode,
				u64 start, size_t size, size_t compressed_size,
				struct page **compressed_pages)
{
	struct btrfs_key key;
	struct btrfs_path *path;
	struct extent_buffer *leaf;
	struct page *page = NULL;
	char *kaddr;
	unsigned long ptr;
	struct btrfs_file_extent_item *ei;
	int err = 0;
	int ret;
	size_t cur_size = size;
	size_t datasize;
	unsigned long offset;
	int use_compress = 0;

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

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

	path->leave_spinning = 1;
	btrfs_set_trans_block_group(trans, inode);

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

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

	if (use_compress) {
		struct page *cpage;
		int i = 0;
		while (compressed_size > 0) {
			cpage = compressed_pages[i];
			cur_size = min_t(unsigned long, compressed_size,
				       PAGE_CACHE_SIZE);

			kaddr = kmap_atomic(cpage, KM_USER0);
			write_extent_buffer(leaf, kaddr, ptr, cur_size);
			kunmap_atomic(kaddr, KM_USER0);

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

	/*
	 * we're an inline extent, so nobody can
	 * extend the file past i_size without locking
	 * a page we already have locked.
	 *
	 * We must do any isize and inode updates
	 * before we unlock the pages.  Otherwise we
	 * could end up racing with unlink.
	 */
	BTRFS_I(inode)->disk_i_size = inode->i_size;
	btrfs_update_inode(trans, root, inode);

	return 0;
fail:
	btrfs_free_path(path);
	return err;
}


/*
 * conditionally insert an inline extent into the file.  This
 * does the checks required to make sure the data is small enough
 * to fit as an inline extent.
 */
static noinline int cow_file_range_inline(struct btrfs_trans_handle *trans,
				 struct btrfs_root *root,
				 struct inode *inode, u64 start, u64 end,
				 size_t compressed_size,
				 struct page **compressed_pages)
{
	u64 isize = i_size_read(inode);
	u64 actual_end = min(end + 1, isize);
	u64 inline_len = actual_end - start;
	u64 aligned_end = (end + root->sectorsize - 1) &
			~((u64)root->sectorsize - 1);
	u64 hint_byte;
	u64 data_len = inline_len;
	int ret;

	if (compressed_size)
		data_len = compressed_size;

	if (start > 0 ||
	    actual_end >= PAGE_CACHE_SIZE ||
	    data_len >= BTRFS_MAX_INLINE_DATA_SIZE(root) ||
	    (!compressed_size &&
	    (actual_end & (root->sectorsize - 1)) == 0) ||
	    end + 1 < isize ||
	    data_len > root->fs_info->max_inline) {
		return 1;
	}

	ret = btrfs_drop_extents(trans, inode, start, aligned_end,
				 &hint_byte, 1);
	BUG_ON(ret);

	if (isize > actual_end)
		inline_len = min_t(u64, isize, actual_end);
	ret = insert_inline_extent(trans, root, inode, start,
				   inline_len, compressed_size,
				   compressed_pages);
	BUG_ON(ret);
	btrfs_drop_extent_cache(inode, start, aligned_end - 1, 0);
	return 0;
}

struct async_extent {
	u64 start;
	u64 ram_size;
	u64 compressed_size;
	struct page **pages;
	unsigned long nr_pages;
	struct list_head list;
};

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

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

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

/*
 * we create compressed extents in two phases.  The first
 * phase compresses a range of pages that have already been
 * locked (both pages and state bits are locked).
 *
 * This is done inside an ordered work queue, and the compression
 * is spread across many cpus.  The actual IO submission is step
 * two, and the ordered work queue takes care of making sure that
 * happens in the same order things were put onto the queue by
 * writepages and friends.
 *
 * If this code finds it can't get good compression, it puts an
 * entry onto the work queue to write the uncompressed bytes.  This
 * makes sure that both compressed inodes and uncompressed inodes
 * are written in the same order that pdflush sent them down.
 */
static noinline int compress_file_range(struct inode *inode,
					struct page *locked_page,
					u64 start, u64 end,
					struct async_cow *async_cow,
					int *num_added)
{
	struct btrfs_root *root = BTRFS_I(inode)->root;
	struct btrfs_trans_handle *trans;
	u64 num_bytes;
	u64 orig_start;
	u64 disk_num_bytes;
	u64 blocksize = root->sectorsize;
	u64 actual_end;
	u64 isize = i_size_read(inode);
	int ret = 0;
	struct page **pages = NULL;
	unsigned long nr_pages;
	unsigned long nr_pages_ret = 0;
	unsigned long total_compressed = 0;
	unsigned long total_in = 0;
	unsigned long max_compressed = 128 * 1024;
	unsigned long max_uncompressed = 128 * 1024;
	int i;
	int will_compress;

	orig_start = start;

	actual_end = min_t(u64, isize, end + 1);
again:
	will_compress = 0;
	nr_pages = (end >> PAGE_CACHE_SHIFT) - (start >> PAGE_CACHE_SHIFT) + 1;
	nr_pages = min(nr_pages, (128 * 1024UL) / PAGE_CACHE_SIZE);

	/*
	 * we don't want to send crud past the end of i_size through
	 * compression, that's just a waste of CPU time.  So, if the
	 * end of the file is before the start of our current
	 * requested range of bytes, we bail out to the uncompressed
	 * cleanup code that can deal with all of this.
	 *
	 * It isn't really the fastest way to fix things, but this is a
	 * very uncommon corner.
	 */
	if (actual_end <= start)
		goto cleanup_and_bail_uncompressed;

	total_compressed = actual_end - start;

	/* we want to make sure that amount of ram required to uncompress
	 * an extent is reasonable, so we limit the total size in ram
	 * of a compressed extent to 128k.  This is a crucial number
	 * because it also controls how easily we can spread reads across
	 * cpus for decompression.
	 *
	 * We also want to make sure the amount of IO required to do
	 * a random read is reasonably small, so we limit the size of
	 * a compressed extent to 128k.
	 */
	total_compressed = min(total_compressed, max_uncompressed);
	num_bytes = (end - start + blocksize) & ~(blocksize - 1);
	num_bytes = max(blocksize,  num_bytes);
	disk_num_bytes = num_bytes;
	total_in = 0;
	ret = 0;

	/*
	 * we do compression for mount -o compress and when the
	 * inode has not been flagged as nocompress.  This flag can
	 * change at any time if we discover bad compression ratios.
	 */
	if (!(BTRFS_I(inode)->flags & BTRFS_INODE_NOCOMPRESS) &&
	    (btrfs_test_opt(root, COMPRESS) ||
	     (BTRFS_I(inode)->force_compress))) {
		WARN_ON(pages);
		pages = kzalloc(sizeof(struct page *) * nr_pages, GFP_NOFS);

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

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

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

		/* lets try to make an inline extent */
		if (ret || total_in < (actual_end - start)) {
			/* we didn't compress the entire range, try
			 * to make an uncompressed inline extent.
			 */
			ret = cow_file_range_inline(trans, root, inode,
						    start, end, 0, NULL);
		} else {
			/* try making a compressed inline extent */
			ret = cow_file_range_inline(trans, root, inode,
						    start, end,
						    total_compressed, pages);
		}
		if (ret == 0) {
			/*
			 * inline extent creation worked, we don't need
			 * to create any more async work items.  Unlock
			 * and free up our temp pages.
			 */
			extent_clear_unlock_delalloc(inode,
			     &BTRFS_I(inode)->io_tree,
			     start, end, NULL,
			     EXTENT_CLEAR_UNLOCK_PAGE | EXTENT_CLEAR_DIRTY |
			     EXTENT_CLEAR_DELALLOC |
			     EXTENT_CLEAR_ACCOUNTING |
			     EXTENT_SET_WRITEBACK | EXTENT_END_WRITEBACK);

			btrfs_end_transaction(trans, root);
			goto free_pages_out;
		}
		btrfs_end_transaction(trans, root);
	}

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

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

		/* flag the file so we don't compress in the future */
		if (!btrfs_test_opt(root, FORCE_COMPRESS) &&
		    !(BTRFS_I(inode)->force_compress)) {
			BTRFS_I(inode)->flags |= BTRFS_INODE_NOCOMPRESS;
		}
	}
	if (will_compress) {
		*num_added += 1;

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

		if (start + num_bytes < end && start + num_bytes < actual_end) {
			start += num_bytes;
			pages = NULL;
			cond_resched();
			goto again;
		}
	} else {
cleanup_and_bail_uncompressed:
		/*
		 * No compression, but we still need to write the pages in
		 * the file we've been given so far.  redirty the locked
		 * page if it corresponds to our extent and set things up
		 * for the async work queue to run cow_file_range to do
		 * the normal delalloc dance
		 */
		if (page_offset(locked_page) >= start &&
		    page_offset(locked_page) <= end) {
			__set_page_dirty_nobuffers(locked_page);
			/* unlocked later on in the async handlers */
		}
		add_async_extent(async_cow, start, end - start + 1, 0, NULL, 0);
		*num_added += 1;
	}

out:
	return 0;

free_pages_out:
	for (i = 0; i < nr_pages_ret; i++) {
		WARN_ON(pages[i]->mapping);
		page_cache_release(pages[i]);
	}
	kfree(pages);

	goto out;
}

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

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


	while (!list_empty(&async_cow->extents)) {
		async_extent = list_entry(async_cow->extents.next,
					  struct async_extent, list);
		list_del(&async_extent->list);

		io_tree = &BTRFS_I(inode)->io_tree;

retry:
		/* did the compression code fall back to uncompressed IO? */
		if (!async_extent->pages) {
			int page_started = 0;
			unsigned long nr_written = 0;

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

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

			/*
			 * if page_started, cow_file_range inserted an
			 * inline extent and took care of all the unlocking
			 * and IO for us.  Otherwise, we need to submit
			 * all those pages down to the drive.
			 */
			if (!page_started && !ret)
				extent_write_locked_range(io_tree,
						  inode, async_extent->start,
						  async_extent->start +
						  async_extent->ram_size - 1,
						  btrfs_get_extent,
						  WB_SYNC_ALL);
			kfree(async_extent);
			cond_resched();
			continue;
		}

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

		trans = btrfs_join_transaction(root, 1);
		ret = btrfs_reserve_extent(trans, root,
					   async_extent->compressed_size,
					   async_extent->compressed_size,
					   0, alloc_hint,
					   (u64)-1, &ins, 1);
		btrfs_end_transaction(trans, root);

		if (ret) {
			int i;
			for (i = 0; i < async_extent->nr_pages; i++) {
				WARN_ON(async_extent->pages[i]->mapping);
				page_cache_release(async_extent->pages[i]);
			}
			kfree(async_extent->pages);
			async_extent->nr_pages = 0;
			async_extent->pages = NULL;
			unlock_extent(io_tree, async_extent->start,
				      async_extent->start +
				      async_extent->ram_size - 1, GFP_NOFS);
			goto retry;
		}

		/*
		 * here we're doing allocation and writeback of the
		 * compressed pages
		 */
		btrfs_drop_extent_cache(inode, async_extent->start,
					async_extent->start +
					async_extent->ram_size - 1, 0);

		em = alloc_extent_map(GFP_NOFS);
		em->start = async_extent->start;
		em->len = async_extent->ram_size;
		em->orig_start = em->start;

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

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

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

		/*
		 * clear dirty, set writeback and unlock the pages.
		 */
		extent_clear_unlock_delalloc(inode,
				&BTRFS_I(inode)->io_tree,
				async_extent->start,
				async_extent->start +
				async_extent->ram_size - 1,
				NULL, EXTENT_CLEAR_UNLOCK_PAGE |
				EXTENT_CLEAR_UNLOCK |
				EXTENT_CLEAR_DELALLOC |
				EXTENT_CLEAR_DIRTY | EXTENT_SET_WRITEBACK);

		ret = btrfs_submit_compressed_write(inode,
				    async_extent->start,
				    async_extent->ram_size,
				    ins.objectid,
				    ins.offset, async_extent->pages,
				    async_extent->nr_pages);

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

	return 0;
}

/*
 * when extent_io.c finds a delayed allocation range in the file,
 * the call backs end up in this code.  The basic idea is to
 * allocate extents on disk for the range, and create ordered data structs
 * in ram to track those extents.
 *
 * locked_page is the page that writepage had locked already.  We use
 * it to make sure we don't do extra locks or unlocks.
 *
 * *page_started is set to one if we unlock locked_page and do everything
 * required to start IO on it.  It may be clean and already done with
 * IO when we return.
 */
static noinline int cow_file_range(struct inode *inode,
				   struct page *locked_page,
				   u64 start, u64 end, int *page_started,
				   unsigned long *nr_written,
				   int unlock)
{
	struct btrfs_root *root = BTRFS_I(inode)->root;
	struct btrfs_trans_handle *trans;
	u64 alloc_hint = 0;
	u64 num_bytes;
	unsigned long ram_size;
	u64 disk_num_bytes;
	u64 cur_alloc_size;
	u64 blocksize = root->sectorsize;
	u64 actual_end;
	u64 isize = i_size_read(inode);
	struct btrfs_key ins;
	struct extent_map *em;
	struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
	int ret = 0;

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

	actual_end = min_t(u64, isize, end + 1);

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

	if (start == 0) {
		/* lets try to make an inline extent */
		ret = cow_file_range_inline(trans, root, inode,
					    start, end, 0, NULL);
		if (ret == 0) {
			extent_clear_unlock_delalloc(inode,
				     &BTRFS_I(inode)->io_tree,
				     start, end, NULL,
				     EXTENT_CLEAR_UNLOCK_PAGE |
				     EXTENT_CLEAR_UNLOCK |
				     EXTENT_CLEAR_DELALLOC |
				     EXTENT_CLEAR_ACCOUNTING |
				     EXTENT_CLEAR_DIRTY |
				     EXTENT_SET_WRITEBACK |
				     EXTENT_END_WRITEBACK);

			*nr_written = *nr_written +
			     (end - start + PAGE_CACHE_SIZE) / PAGE_CACHE_SIZE;
			*page_started = 1;
			ret = 0;
			goto out;
		}
	}

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


	read_lock(&BTRFS_I(inode)->extent_tree.lock);
	em = search_extent_mapping(&BTRFS_I(inode)->extent_tree,
				   start, num_bytes);
	if (em) {
		/*
		 * if block start isn't an actual block number then find the
		 * first block in this inode and use that as a hint.  If that
		 * block is also bogus then just don't worry about it.
		 */
		if (em->block_start >= EXTENT_MAP_LAST_BYTE) {
			free_extent_map(em);
			em = search_extent_mapping(em_tree, 0, 0);
			if (em && em->block_start < EXTENT_MAP_LAST_BYTE)
				alloc_hint = em->block_start;
			if (em)
				free_extent_map(em);
		} else {
			alloc_hint = em->block_start;
			free_extent_map(em);
		}
	}
	read_unlock(&BTRFS_I(inode)->extent_tree.lock);
	btrfs_drop_extent_cache(inode, start, start + num_bytes - 1, 0);

	while (disk_num_bytes > 0) {
		unsigned long op;

		cur_alloc_size = min(disk_num_bytes, root->fs_info->max_extent);
		ret = btrfs_reserve_extent(trans, root, cur_alloc_size,
					   root->sectorsize, 0, alloc_hint,
					   (u64)-1, &ins, 1);
		BUG_ON(ret);

		em = alloc_extent_map(GFP_NOFS);
		em->start = start;
		em->orig_start = em->start;
		ram_size = ins.offset;
		em->len = ins.offset;

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

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

		cur_alloc_size = ins.offset;
		ret = btrfs_add_ordered_extent(inode, start, ins.objectid,
					       ram_size, cur_alloc_size, 0);
		BUG_ON(ret);

		if (root->root_key.objectid ==
		    BTRFS_DATA_RELOC_TREE_OBJECTID) {
			ret = btrfs_reloc_clone_csums(inode, start,
						      cur_alloc_size);
			BUG_ON(ret);
		}

		if (disk_num_bytes < cur_alloc_size)
			break;

		/* we're not doing compressed IO, don't unlock the first
		 * page (which the caller expects to stay locked), don't
		 * clear any dirty bits and don't set any writeback bits
		 *
		 * Do set the Private2 bit so we know this page was properly
		 * setup for writepage
		 */
		op = unlock ? EXTENT_CLEAR_UNLOCK_PAGE : 0;
		op |= EXTENT_CLEAR_UNLOCK | EXTENT_CLEAR_DELALLOC |
			EXTENT_SET_PRIVATE2;

		extent_clear_unlock_delalloc(inode, &BTRFS_I(inode)->io_tree,
					     start, start + ram_size - 1,
					     locked_page, op);
		disk_num_bytes -= cur_alloc_size;
		num_bytes -= cur_alloc_size;
		alloc_hint = ins.objectid + ins.offset;
		start += cur_alloc_size;
	}
out:
	ret = 0;
	btrfs_end_transaction(trans, root);

	return ret;
}

/*
 * work queue call back to started compression on a file and pages
 */
static noinline void async_cow_start(struct btrfs_work *work)
{
	struct async_cow *async_cow;
	int num_added = 0;
	async_cow = container_of(work, struct async_cow, work);

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

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

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

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

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

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

	if (async_cow->inode)
		submit_compressed_extents(async_cow->inode, async_cow);
}

static noinline void async_cow_free(struct btrfs_work *work)
{
	struct async_cow *async_cow;
	async_cow = container_of(work, struct async_cow, work);
	kfree(async_cow);
}

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

	clear_extent_bit(&BTRFS_I(inode)->io_tree, start, end, EXTENT_LOCKED,
			 1, 0, NULL, GFP_NOFS);
	while (start < end) {
		async_cow = kmalloc(sizeof(*async_cow), GFP_NOFS);
		async_cow->inode = inode;
		async_cow->root = root;
		async_cow->locked_page = locked_page;
		async_cow->start = start;

		if (BTRFS_I(inode)->flags & BTRFS_INODE_NOCOMPRESS)
			cur_end = end;
		else
			cur_end = min(end, start + 512 * 1024 - 1);

		async_cow->end = cur_end;
		INIT_LIST_HEAD(&async_cow->extents);

		async_cow->work.func = async_cow_start;
		async_cow->work.ordered_func = async_cow_submit;
		async_cow->work.ordered_free = async_cow_free;
		async_cow->work.flags = 0;

		nr_pages = (cur_end - start + PAGE_CACHE_SIZE) >>
			PAGE_CACHE_SHIFT;
		atomic_add(nr_pages, &root->fs_info->async_delalloc_pages);

		btrfs_queue_worker(&root->fs_info->delalloc_workers,
				   &async_cow->work);

		if (atomic_read(&root->fs_info->async_delalloc_pages) > limit) {
			wait_event(root->fs_info->async_submit_wait,
			   (atomic_read(&root->fs_info->async_delalloc_pages) <
			    limit));
		}

		while (atomic_read(&root->fs_info->async_submit_draining) &&
		      atomic_read(&root->fs_info->async_delalloc_pages)) {
			wait_event(root->fs_info->async_submit_wait,
			  (atomic_read(&root->fs_info->async_delalloc_pages) ==
			   0));
		}

		*nr_written += nr_pages;
		start = cur_end + 1;
	}
	*page_started = 1;
	return 0;
}

static noinline int csum_exist_in_range(struct btrfs_root *root,
					u64 bytenr, u64 num_bytes)
{
	int ret;
	struct btrfs_ordered_sum *sums;
	LIST_HEAD(list);

	ret = btrfs_lookup_csums_range(root->fs_info->csum_root, bytenr,
				       bytenr + num_bytes - 1, &list);
	if (ret == 0 && list_empty(&list))
		return 0;

	while (!list_empty(&list)) {
		sums = list_entry(list.next, struct btrfs_ordered_sum, list);
		list_del(&sums->list);
		kfree(sums);
	}
	return 1;
}

/*
 * when nowcow writeback call back.  This checks for snapshots or COW copies
 * of the extents that exist in the file, and COWs the file as required.
 *
 * If no cow copies or snapshots exist, we write directly to the existing
 * blocks on disk
 */
static noinline int run_delalloc_nocow(struct inode *inode,
				       struct page *locked_page,
			      u64 start, u64 end, int *page_started, int force,
			      unsigned long *nr_written)
{
	struct btrfs_root *root = BTRFS_I(inode)->root;
	struct btrfs_trans_handle *trans;
	struct extent_buffer *leaf;
	struct btrfs_path *path;
	struct btrfs_file_extent_item *fi;
	struct btrfs_key found_key;
	u64 cow_start;
	u64 cur_offset;
	u64 extent_end;
	u64 extent_offset;
	u64 disk_bytenr;
	u64 num_bytes;
	int extent_type;
	int ret;
	int type;
	int nocow;
	int check_prev = 1;

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

	cow_start = (u64)-1;
	cur_offset = start;
	while (1) {
		ret = btrfs_lookup_file_extent(trans, root, path, inode->i_ino,
					       cur_offset, 0);
		BUG_ON(ret < 0);
		if (ret > 0 && path->slots[0] > 0 && check_prev) {
			leaf = path->nodes[0];
			btrfs_item_key_to_cpu(leaf, &found_key,
					      path->slots[0] - 1);
			if (found_key.objectid == inode->i_ino &&
			    found_key.type == BTRFS_EXTENT_DATA_KEY)
				path->slots[0]--;
		}
		check_prev = 0;
next_slot:
		leaf = path->nodes[0];
		if (path->slots[0] >= btrfs_header_nritems(leaf)) {
			ret = btrfs_next_leaf(root, path);
			if (ret < 0)
				BUG_ON(1);
			if (ret > 0)
				break;
			leaf = path->nodes[0];
		}

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

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

		if (found_key.offset > cur_offset) {
			extent_end = found_key.offset;
			extent_type = 0;
			goto out_check;
		}

		fi = btrfs_item_ptr(leaf, path->slots[0],
				    struct btrfs_file_extent_item);
		extent_type = btrfs_file_extent_type(leaf, fi);

		if (extent_type == BTRFS_FILE_EXTENT_REG ||
		    extent_type == BTRFS_FILE_EXTENT_PREALLOC) {
			disk_bytenr = btrfs_file_extent_disk_bytenr(leaf, fi);
			extent_offset = btrfs_file_extent_offset(leaf, fi);
			extent_end = found_key.offset +
				btrfs_file_extent_num_bytes(leaf, fi);
			if (extent_end <= start) {
				path->slots[0]++;
				goto next_slot;
			}
			if (disk_bytenr == 0)
				goto out_check;
			if (btrfs_file_extent_compression(leaf, fi) ||
			    btrfs_file_extent_encryption(leaf, fi) ||
			    btrfs_file_extent_other_encoding(leaf, fi))
				goto out_check;
			if (extent_type == BTRFS_FILE_EXTENT_REG && !force)
				goto out_check;
			if (btrfs_extent_readonly(root, disk_bytenr))
				goto out_check;
			if (btrfs_cross_ref_exist(trans, root, inode->i_ino,
						  found_key.offset -
						  extent_offset, disk_bytenr))
				goto out_check;
			disk_bytenr += extent_offset;
			disk_bytenr += cur_offset - found_key.offset;
			num_bytes = min(end + 1, extent_end) - cur_offset;
			/*
			 * force cow if csum exists in the range.
			 * this ensure that csum for a given extent are
			 * either valid or do not exist.
			 */
			if (csum_exist_in_range(root, disk_bytenr, num_bytes))
				goto out_check;
			nocow = 1;
		} else if (extent_type == BTRFS_FILE_EXTENT_INLINE) {
			extent_end = found_key.offset +
				btrfs_file_extent_inline_len(leaf, fi);
			extent_end = ALIGN(extent_end, root->sectorsize);
		} else {
			BUG_ON(1);
		}
out_check:
		if (extent_end <= start) {
			path->slots[0]++;
			goto next_slot;
		}
		if (!nocow) {
			if (cow_start == (u64)-1)
				cow_start = cur_offset;
			cur_offset = extent_end;
			if (cur_offset > end)
				break;
			path->slots[0]++;
			goto next_slot;
		}

		btrfs_release_path(root, path);
		if (cow_start != (u64)-1) {
			ret = cow_file_range(inode, locked_page, cow_start,
					found_key.offset - 1, page_started,
					nr_written, 1);
			BUG_ON(ret);
			cow_start = (u64)-1;
		}

		if (extent_type == BTRFS_FILE_EXTENT_PREALLOC) {
			struct extent_map *em;
			struct extent_map_tree *em_tree;
			em_tree = &BTRFS_I(inode)->extent_tree;
			em = alloc_extent_map(GFP_NOFS);
			em->start = cur_offset;
			em->orig_start = em->start;
			em->len = num_bytes;
			em->block_len = num_bytes;
			em->block_start = disk_bytenr;
			em->bdev = root->fs_info->fs_devices->latest_bdev;
			set_bit(EXTENT_FLAG_PINNED, &em->flags);
			while (1) {
				write_lock(&em_tree->lock);
				ret = add_extent_mapping(em_tree, em);
				write_unlock(&em_tree->lock);
				if (ret != -EEXIST) {
					free_extent_map(em);
					break;
				}
				btrfs_drop_extent_cache(inode, em->start,
						em->start + em->len - 1, 0);
			}
			type = BTRFS_ORDERED_PREALLOC;
		} else {
			type = BTRFS_ORDERED_NOCOW;
		}

		ret = btrfs_add_ordered_extent(inode, cur_offset, disk_bytenr,
					       num_bytes, num_bytes, type);
		BUG_ON(ret);

		extent_clear_unlock_delalloc(inode, &BTRFS_I(inode)->io_tree,
				cur_offset, cur_offset + num_bytes - 1,
				locked_page, EXTENT_CLEAR_UNLOCK_PAGE |
				EXTENT_CLEAR_UNLOCK | EXTENT_CLEAR_DELALLOC |
				EXTENT_SET_PRIVATE2);
		cur_offset = extent_end;
		if (cur_offset > end)
			break;
	}
	btrfs_release_path(root, path);

	if (cur_offset <= end && cow_start == (u64)-1)
		cow_start = cur_offset;
	if (cow_start != (u64)-1) {
		ret = cow_file_range(inode, locked_page, cow_start, end,
				     page_started, nr_written, 1);
		BUG_ON(ret);
	}

	ret = btrfs_end_transaction(trans, root);
	BUG_ON(ret);
	btrfs_free_path(path);
	return 0;
}

/*
 * extent_io.c call back to do delayed allocation processing
 */
static int run_delalloc_range(struct inode *inode, struct page *locked_page,
			      u64 start, u64 end, int *page_started,
			      unsigned long *nr_written)
{
	int ret;
	struct btrfs_root *root = BTRFS_I(inode)->root;

	if (BTRFS_I(inode)->flags & BTRFS_INODE_NODATACOW)
		ret = run_delalloc_nocow(inode, locked_page, start, end,
					 page_started, 1, nr_written);
	else if (BTRFS_I(inode)->flags & BTRFS_INODE_PREALLOC)
		ret = run_delalloc_nocow(inode, locked_page, start, end,
					 page_started, 0, nr_written);
	else if (!btrfs_test_opt(root, COMPRESS) &&
		 !(BTRFS_I(inode)->force_compress))
		ret = cow_file_range(inode, locked_page, start, end,
				      page_started, nr_written, 1);
	else
		ret = cow_file_range_async(inode, locked_page, start, end,
					   page_started, nr_written);
	return ret;
}

static int btrfs_split_extent_hook(struct inode *inode,
				    struct extent_state *orig, u64 split)
{
	struct btrfs_root *root = BTRFS_I(inode)->root;
	u64 size;

	if (!(orig->state & EXTENT_DELALLOC))
		return 0;

	size = orig->end - orig->start + 1;
	if (size > root->fs_info->max_extent) {
		u64 num_extents;
		u64 new_size;

		new_size = orig->end - split + 1;
		num_extents = div64_u64(size + root->fs_info->max_extent - 1,
					root->fs_info->max_extent);

		/*
		 * if we break a large extent up then leave oustanding_extents
		 * be, since we've already accounted for the large extent.
		 */
		if (div64_u64(new_size + root->fs_info->max_extent - 1,
			      root->fs_info->max_extent) < num_extents)
			return 0;
	}

	spin_lock(&BTRFS_I(inode)->accounting_lock);
	BTRFS_I(inode)->outstanding_extents++;
	spin_unlock(&BTRFS_I(inode)->accounting_lock);

	return 0;
}

/*
 * extent_io.c merge_extent_hook, used to track merged delayed allocation
 * extents so we can keep track of new extents that are just merged onto old
 * extents, such as when we are doing sequential writes, so we can properly
 * account for the metadata space we'll need.
 */
static int btrfs_merge_extent_hook(struct inode *inode,
				   struct extent_state *new,
				   struct extent_state *other)
{
	struct btrfs_root *root = BTRFS_I(inode)->root;
	u64 new_size, old_size;
	u64 num_extents;

	/* not delalloc, ignore it */
	if (!(other->state & EXTENT_DELALLOC))
		return 0;

	old_size = other->end - other->start + 1;
	if (new->start < other->start)
		new_size = other->end - new->start + 1;
	else
		new_size = new->end - other->start + 1;

	/* we're not bigger than the max, unreserve the space and go */
	if (new_size <= root->fs_info->max_extent) {
		spin_lock(&BTRFS_I(inode)->accounting_lock);
		BTRFS_I(inode)->outstanding_extents--;
		spin_unlock(&BTRFS_I(inode)->accounting_lock);
		return 0;
	}

	/*
	 * If we grew by another max_extent, just return, we want to keep that
	 * reserved amount.
	 */
	num_extents = div64_u64(old_size + root->fs_info->max_extent - 1,
				root->fs_info->max_extent);
	if (div64_u64(new_size + root->fs_info->max_extent - 1,
		      root->fs_info->max_extent) > num_extents)
		return 0;

	spin_lock(&BTRFS_I(inode)->accounting_lock);
	BTRFS_I(inode)->outstanding_extents--;
	spin_unlock(&BTRFS_I(inode)->accounting_lock);

	return 0;
}

/*
 * extent_io.c set_bit_hook, used to track delayed allocation
 * bytes in this file, and to maintain the list of inodes that
 * have pending delalloc work to be done.
 */
static int btrfs_set_bit_hook(struct inode *inode, u64 start, u64 end,
		       unsigned long old, unsigned long bits)
{

	/*
	 * set_bit and clear bit hooks normally require _irqsave/restore
	 * but in this case, we are only testeing for the DELALLOC
	 * bit, which is only set or cleared with irqs on
	 */
	if (!(old & EXTENT_DELALLOC) && (bits & EXTENT_DELALLOC)) {
		struct btrfs_root *root = BTRFS_I(inode)->root;

		spin_lock(&BTRFS_I(inode)->accounting_lock);
		BTRFS_I(inode)->outstanding_extents++;
		spin_unlock(&BTRFS_I(inode)->accounting_lock);
		btrfs_delalloc_reserve_space(root, inode, end - start + 1);
		spin_lock(&root->fs_info->