Merge branch 'for-chris' of git://git.kernel.org/pub/scm/linux/kernel/git/josef/btrfs-work into for-linus

7 files changed, 180 insertions, 124 deletions

diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h
index 3458b572554..0d00a07b5b2 100644
--- a/fs/btrfs/ctree.h
+++ b/fs/btrfs/ctree.h
@@ -2576,6 +2576,11 @@ int btrfs_drop_extents(struct btrfs_trans_handle *trans, struct inode *inode,
 int btrfs_mark_extent_written(struct btrfs_trans_handle *trans,
 			      struct inode *inode, u64 start, u64 end);
 int btrfs_release_file(struct inode *inode, struct file *file);
+void btrfs_drop_pages(struct page **pages, size_t num_pages);
+int btrfs_dirty_pages(struct btrfs_root *root, struct inode *inode,
+		      struct page **pages, size_t num_pages,
+		      loff_t pos, size_t write_bytes,
+		      struct extent_state **cached);
 
 /* tree-defrag.c */
 int btrfs_defrag_leaves(struct btrfs_trans_handle *trans,
diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c
index a272bfd74ea..ef6865c17cd 100644
--- a/fs/btrfs/disk-io.c
+++ b/fs/btrfs/disk-io.c
@@ -3136,7 +3136,7 @@ static int btrfs_cleanup_transaction(struct btrfs_root *root)
 		btrfs_destroy_pinned_extent(root,
 					    root->fs_info->pinned_extents);
 
-		t->use_count = 0;
+		atomic_set(&t->use_count, 0);
 		list_del_init(&t->list);
 		memset(t, 0, sizeof(*t));
 		kmem_cache_free(btrfs_transaction_cachep, t);
diff --git a/fs/btrfs/file.c b/fs/btrfs/file.c
index e621ea54a3f..75899a01dde 100644
--- a/fs/btrfs/file.c
+++ b/fs/btrfs/file.c
@@ -104,7 +104,7 @@ static noinline int btrfs_copy_from_user(loff_t pos, int num_pages,
 /*
  * unlocks pages after btrfs_file_write is done with them
  */
-static noinline void btrfs_drop_pages(struct page **pages, size_t num_pages)
+void btrfs_drop_pages(struct page **pages, size_t num_pages)
 {
 	size_t i;
 	for (i = 0; i < num_pages; i++) {
@@ -127,16 +127,13 @@ static noinline void btrfs_drop_pages(struct page **pages, size_t num_pages)
  * this also makes the decision about creating an inline extent vs
  * doing real data extents, marking pages dirty and delalloc as required.
  */
-static noinline int dirty_and_release_pages(struct btrfs_root *root,
-					    struct file *file,
-					    struct page **pages,
-					    size_t num_pages,
-					    loff_t pos,
-					    size_t write_bytes)
+int btrfs_dirty_pages(struct btrfs_root *root, struct inode *inode,
+		      struct page **pages, size_t num_pages,
+		      loff_t pos, size_t write_bytes,
+		      struct extent_state **cached)
 {
 	int err = 0;
 	int i;
-	struct inode *inode = fdentry(file)->d_inode;
 	u64 num_bytes;
 	u64 start_pos;
 	u64 end_of_last_block;
@@ -149,7 +146,7 @@ static noinline int dirty_and_release_pages(struct btrfs_root *root,
 
 	end_of_last_block = start_pos + num_bytes - 1;
 	err = btrfs_set_extent_delalloc(inode, start_pos, end_of_last_block,
-					NULL);
+					cached);
 	if (err)
 		return err;
 
@@ -992,9 +989,9 @@ static noinline ssize_t __btrfs_buffered_write(struct file *file,
 		}
 
 		if (copied > 0) {
-			ret = dirty_and_release_pages(root, file, pages,
-						      dirty_pages, pos,
-						      copied);
+			ret = btrfs_dirty_pages(root, inode, pages,
+						dirty_pages, pos, copied,
+						NULL);
 			if (ret) {
 				btrfs_delalloc_release_space(inode,
 					dirty_pages << PAGE_CACHE_SHIFT);
diff --git a/fs/btrfs/free-space-cache.c b/fs/btrfs/free-space-cache.c
index f561c953205..a3f420def0e 100644
--- a/fs/btrfs/free-space-cache.c
+++ b/fs/btrfs/free-space-cache.c
@@ -508,6 +508,7 @@ int btrfs_write_out_cache(struct btrfs_root *root,
 	struct inode *inode;
 	struct rb_node *node;
 	struct list_head *pos, *n;
+	struct page **pages;
 	struct page *page;
 	struct extent_state *cached_state = NULL;
 	struct btrfs_free_cluster *cluster = NULL;
@@ -517,13 +518,13 @@ int btrfs_write_out_cache(struct btrfs_root *root,
 	u64 start, end, len;
 	u64 bytes = 0;
 	u32 *crc, *checksums;
-	pgoff_t index = 0, last_index = 0;
 	unsigned long first_page_offset;
-	int num_checksums;
+	int index = 0, num_pages = 0;
 	int entries = 0;
 	int bitmaps = 0;
 	int ret = 0;
 	bool next_page = false;
+	bool out_of_space = false;
 
 	root = root->fs_info->tree_root;
 
@@ -551,24 +552,31 @@ int btrfs_write_out_cache(struct btrfs_root *root,
 		return 0;
 	}
 
-	last_index = (i_size_read(inode) - 1) >> PAGE_CACHE_SHIFT;
+	num_pages = (i_size_read(inode) + PAGE_CACHE_SIZE - 1) >>
+		PAGE_CACHE_SHIFT;
 	filemap_write_and_wait(inode->i_mapping);
 	btrfs_wait_ordered_range(inode, inode->i_size &
 				 ~(root->sectorsize - 1), (u64)-1);
 
 	/* We need a checksum per page. */
-	num_checksums = i_size_read(inode) / PAGE_CACHE_SIZE;
-	crc = checksums  = kzalloc(sizeof(u32) * num_checksums, GFP_NOFS);
+	crc = checksums = kzalloc(sizeof(u32) * num_pages, GFP_NOFS);
 	if (!crc) {
 		iput(inode);
 		return 0;
 	}
 
+	pages = kzalloc(sizeof(struct page *) * num_pages, GFP_NOFS);
+	if (!pages) {
+		kfree(crc);
+		iput(inode);
+		return 0;
+	}
+
 	/* Since the first page has all of our checksums and our generation we
 	 * need to calculate the offset into the page that we can start writing
 	 * our entries.
 	 */
-	first_page_offset = (sizeof(u32) * num_checksums) + sizeof(u64);
+	first_page_offset = (sizeof(u32) * num_pages) + sizeof(u64);
 
 	/* Get the cluster for this block_group if it exists */
 	if (!list_empty(&block_group->cluster_list))
@@ -590,20 +598,18 @@ int btrfs_write_out_cache(struct btrfs_root *root,
 	 * after find_get_page at this point.  Just putting this here so people
 	 * know and don't freak out.
 	 */
-	while (index <= last_index) {
+	while (index < num_pages) {
 		page = grab_cache_page(inode->i_mapping, index);
 		if (!page) {
-			pgoff_t i = 0;
+			int i;
 
-			while (i < index) {
-				page = find_get_page(inode->i_mapping, i);
-				unlock_page(page);
-				page_cache_release(page);
-				page_cache_release(page);
-				i++;
+			for (i = 0; i < num_pages; i++) {
+				unlock_page(pages[i]);
+				page_cache_release(pages[i]);
 			}
 			goto out_free;
 		}
+		pages[index] = page;
 		index++;
 	}
 
@@ -631,7 +637,12 @@ int btrfs_write_out_cache(struct btrfs_root *root,
 			offset = start_offset;
 		}
 
-		page = find_get_page(inode->i_mapping, index);
+		if (index >= num_pages) {
+			out_of_space = true;
+			break;
+		}
+
+		page = pages[index];
 
 		addr = kmap(page);
 		entry = addr + start_offset;
@@ -708,23 +719,6 @@ int btrfs_write_out_cache(struct btrfs_root *root,
 
 		bytes += PAGE_CACHE_SIZE;
 
-		ClearPageChecked(page);
-		set_page_extent_mapped(page);
-		SetPageUptodate(page);
-		set_page_dirty(page);
-
-		/*
-		 * We need to release our reference we got for grab_cache_page,
-		 * except for the first page which will hold our checksums, we
-		 * do that below.
-		 */
-		if (index != 0) {
-			unlock_page(page);
-			page_cache_release(page);
-		}
-
-		page_cache_release(page);
-
 		index++;
 	} while (node || next_page);
 
@@ -734,6 +728,10 @@ int btrfs_write_out_cache(struct btrfs_root *root,
 		struct btrfs_free_space *entry =
 			list_entry(pos, struct btrfs_free_space, list);
 
+		if (index >= num_pages) {
+			out_of_space = true;
+			break;
+		}
 		page = find_get_page(inode->i_mapping, index);
 
 		addr = kmap(page);
@@ -745,64 +743,58 @@ int btrfs_write_out_cache(struct btrfs_root *root,
 		crc++;
 		bytes += PAGE_CACHE_SIZE;
 
-		ClearPageChecked(page);
-		set_page_extent_mapped(page);
-		SetPageUptodate(page);
-		set_page_dirty(page);
-		unlock_page(page);
-		page_cache_release(page);
-		page_cache_release(page);
 		list_del_init(&entry->list);
 		index++;
 	}
 
+	if (out_of_space) {
+		btrfs_drop_pages(pages, num_pages);
+		unlock_extent_cached(&BTRFS_I(inode)->io_tree, 0,
+				     i_size_read(inode) - 1, &cached_state,
+				     GFP_NOFS);
+		ret = 0;
+		goto out_free;
+	}
+
 	/* Zero out the rest of the pages just to make sure */
-	while (index <= last_index) {
+	while (index < num_pages) {
 		void *addr;
 
-		page = find_get_page(inode->i_mapping, index);
-
+		page = pages[index];
 		addr = kmap(page);
 		memset(addr, 0, PAGE_CACHE_SIZE);
 		kunmap(page);
-		ClearPageChecked(page);
-		set_page_extent_mapped(page);
-		SetPageUptodate(page);
-		set_page_dirty(page);
-		unlock_page(page);
-		page_cache_release(page);
-		page_cache_release(page);
 		bytes += PAGE_CACHE_SIZE;
 		index++;
 	}
 
-	btrfs_set_extent_delalloc(inode, 0, bytes - 1, &cached_state);
-
 	/* Write the checksums and trans id to the first page */
 	{
 		void *addr;
 		u64 *gen;
 
-		page = find_get_page(inode->i_mapping, 0);
+		page = pages[0];
 
 		addr = kmap(page);
-		memcpy(addr, checksums, sizeof(u32) * num_checksums);
-		gen = addr + (sizeof(u32) * num_checksums);
+		memcpy(addr, checksums, sizeof(u32) * num_pages);
+		gen = addr + (sizeof(u32) * num_pages);
 		*gen = trans->transid;
 		kunmap(page);
-		ClearPageChecked(page);
-		set_page_extent_mapped(page);
-		SetPageUptodate(page);
-		set_page_dirty(page);
-		unlock_page(page);
-		page_cache_release(page);
-		page_cache_release(page);
 	}
-	BTRFS_I(inode)->generation = trans->transid;
 
+	ret = btrfs_dirty_pages(root, inode, pages, num_pages, 0,
+					    bytes, &cached_state);
+	btrfs_drop_pages(pages, num_pages);
 	unlock_extent_cached(&BTRFS_I(inode)->io_tree, 0,
 			     i_size_read(inode) - 1, &cached_state, GFP_NOFS);
 
+	if (ret) {
+		ret = 0;
+		goto out_free;
+	}
+
+	BTRFS_I(inode)->generation = trans->transid;
+
 	filemap_write_and_wait(inode->i_mapping);
 
 	key.objectid = BTRFS_FREE_SPACE_OBJECTID;
@@ -853,6 +845,7 @@ out_free:
 		BTRFS_I(inode)->generation = 0;
 	}
 	kfree(checksums);
+	kfree(pages);
 	btrfs_update_inode(trans, root, inode);
 	iput(inode);
 	return ret;
diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c
index 5a993e0ec86..55a6a0b416d 100644
--- a/fs/btrfs/inode.c
+++ b/fs/btrfs/inode.c
@@ -1769,9 +1769,12 @@ static int btrfs_finish_ordered_io(struct inode *inode, u64 start, u64 end)
 	add_pending_csums(trans, inode, ordered_extent->file_offset,
 			  &ordered_extent->list);
 
-	btrfs_ordered_update_i_size(inode, 0, ordered_extent);
-	ret = btrfs_update_inode(trans, root, inode);
-	BUG_ON(ret);
+	ret = btrfs_ordered_update_i_size(inode, 0, ordered_extent);
+	if (!ret) {
+		ret = btrfs_update_inode(trans, root, inode);
+		BUG_ON(ret);
+	}
+	ret = 0;
 out:
 	if (nolock) {
 		if (trans)
@@ -2589,6 +2592,13 @@ static void fill_inode_item(struct btrfs_trans_handle *trans,
 			    struct btrfs_inode_item *item,
 			    struct inode *inode)
 {
+	if (!leaf->map_token)
+		map_private_extent_buffer(leaf, (unsigned long)item,
+					  sizeof(struct btrfs_inode_item),
+					  &leaf->map_token, &leaf->kaddr,
+					  &leaf->map_start, &leaf->map_len,
+					  KM_USER1);
+
 	btrfs_set_inode_uid(leaf, item, inode->i_uid);
 	btrfs_set_inode_gid(leaf, item, inode->i_gid);
 	btrfs_set_inode_size(leaf, item, BTRFS_I(inode)->disk_i_size);
@@ -2617,6 +2627,11 @@ static void fill_inode_item(struct btrfs_trans_handle *trans,
 	btrfs_set_inode_rdev(leaf, item, inode->i_rdev);
 	btrfs_set_inode_flags(leaf, item, BTRFS_I(inode)->flags);
 	btrfs_set_inode_block_group(leaf, item, BTRFS_I(inode)->block_group);
+
+	if (leaf->map_token) {
+		unmap_extent_buffer(leaf, leaf->map_token, KM_USER1);
+		leaf->map_token = NULL;
+	}
 }
 
 /*
@@ -5433,17 +5448,30 @@ out:
 }
 
 static struct extent_map *btrfs_new_extent_direct(struct inode *inode,
+						  struct extent_map *em,
 						  u64 start, u64 len)
 {
 	struct btrfs_root *root = BTRFS_I(inode)->root;
 	struct btrfs_trans_handle *trans;
-	struct extent_map *em;
 	struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
 	struct btrfs_key ins;
 	u64 alloc_hint;
 	int ret;
+	bool insert = false;
 
-	btrfs_drop_extent_cache(inode, start, start + len - 1, 0);
+	/*
+	 * Ok if the extent map we looked up is a hole and is for the exact
+	 * range we want, there is no reason to allocate a new one, however if
+	 * it is not right then we need to free this one and drop the cache for
+	 * our range.
+	 */
+	if (em->block_start != EXTENT_MAP_HOLE || em->start != start ||
+	    em->len != len) {
+		free_extent_map(em);
+		em = NULL;
+		insert = true;
+		btrfs_drop_extent_cache(inode, start, start + len - 1, 0);
+	}
 
 	trans = btrfs_join_transaction(root, 0);
 	if (IS_ERR(trans))
@@ -5459,10 +5487,12 @@ static struct extent_map *btrfs_new_extent_direct(struct inode *inode,
 		goto out;
 	}
 
-	em = alloc_extent_map(GFP_NOFS);
 	if (!em) {
-		em = ERR_PTR(-ENOMEM);
-		goto out;
+		em = alloc_extent_map(GFP_NOFS);
+		if (!em) {
+			em = ERR_PTR(-ENOMEM);
+			goto out;
+		}
 	}
 
 	em->start = start;
@@ -5472,9 +5502,15 @@ static struct extent_map *btrfs_new_extent_direct(struct inode *inode,
 	em->block_start = ins.objectid;
 	em->block_len = ins.offset;
 	em->bdev = root->fs_info->fs_devices->latest_bdev;
+
+	/*
+	 * We need to do this because if we're using the original em we searched
+	 * for, we could have EXTENT_FLAG_VACANCY set, and we don't want that.
+	 */
+	em->flags = 0;
 	set_bit(EXTENT_FLAG_PINNED, &em->flags);
 
-	while (1) {
+	while (insert) {
 		write_lock(&em_tree->lock);
 		ret = add_extent_mapping(em_tree, em);
 		write_unlock(&em_tree->lock);
@@ -5692,8 +5728,7 @@ must_cow:
 	 * it above
 	 */
 	len = bh_result->b_size;
-	free_extent_map(em);
-	em = btrfs_new_extent_direct(inode, start, len);
+	em = btrfs_new_extent_direct(inode, em, start, len);
 	if (IS_ERR(em))
 		return PTR_ERR(em);
 	len = min(len, em->len - (start - em->start));
@@ -5856,8 +5891,10 @@ again:
 	}
 
 	add_pending_csums(trans, inode, ordered->file_offset, &ordered->list);
-	btrfs_ordered_update_i_size(inode, 0, ordered);
-	btrfs_update_inode(trans, root, inode);
+	ret = btrfs_ordered_update_i_size(inode, 0, ordered);
+	if (!ret)
+		btrfs_update_inode(trans, root, inode);
+	ret = 0;
 out_unlock:
 	unlock_extent_cached(&BTRFS_I(inode)->io_tree, ordered->file_offset,
 			     ordered->file_offset + ordered->len - 1,
@@ -5943,7 +5980,7 @@ static struct bio *btrfs_dio_bio_alloc(struct block_device *bdev,
 
 static inline int __btrfs_submit_dio_bio(struct bio *bio, struct inode *inode,
 					 int rw, u64 file_offset, int skip_sum,
-					 u32 *csums)
+					 u32 *csums, int async_submit)
 {
 	int write = rw & REQ_WRITE;
 	struct btrfs_root *root = BTRFS_I(inode)->root;
@@ -5954,13 +5991,24 @@ static inline int __btrfs_submit_dio_bio(struct bio *bio, struct inode *inode,
 	if (ret)
 		goto err;
 
-	if (write && !skip_sum) {
+	if (skip_sum)
+		goto map;
+
+	if (write && async_submit) {
 		ret = btrfs_wq_submit_bio(root->fs_info,
 				   inode, rw, bio, 0, 0,
 				   file_offset,
 				   __btrfs_submit_bio_start_direct_io,
 				   __btrfs_submit_bio_done);
 		goto err;
+	} else if (write) {
+		/*
+		 * If we aren't doing async submit, calculate the csum of the
+		 * bio now.
+		 */
+		ret = btrfs_csum_one_bio(root, inode, bio, file_offset, 1);
+		if (ret)
+			goto err;
 	} else if (!skip_sum) {
 		ret = btrfs_lookup_bio_sums_dio(root, inode, bio,
 					  file_offset, csums);
@@ -5968,7 +6016,8 @@ static inline int __btrfs_submit_dio_bio(struct bio *bio, struct inode *inode,
 			goto err;
 	}
 
-	ret = btrfs_map_bio(root, rw, bio, 0, 1);
+map:
+	ret = btrfs_map_bio(root, rw, bio, 0, async_submit);
 err:
 	bio_put(bio);
 	return ret;
@@ -5990,15 +6039,9 @@ static int btrfs_submit_direct_hook(int rw, struct btrfs_dio_private *dip,
 	int nr_pages = 0;
 	u32 *csums = dip->csums;
 	int ret = 0;
+	int async_submit = 0;
 	int write = rw & REQ_WRITE;
 
-	bio = btrfs_dio_bio_alloc(orig_bio->bi_bdev, start_sector, GFP_NOFS);
-	if (!bio)
-		return -ENOMEM;
-	bio->bi_private = dip;
-	bio->bi_end_io = btrfs_end_dio_bio;
-	atomic_inc(&dip->pending_bios);
-
 	map_length = orig_bio->bi_size;
 	ret = btrfs_map_block(map_tree, READ, start_sector << 9,
 			      &map_length, NULL, 0);
@@ -6007,6 +6050,19 @@ static int btrfs_submit_direct_hook(int rw, struct btrfs_dio_private *dip,
 		return -EIO;
 	}
 
+	if (map_length >= orig_bio->bi_size) {
+		bio = orig_bio;
+		goto submit;
+	}
+
+	async_submit = 1;
+	bio = btrfs_dio_bio_alloc(orig_bio->bi_bdev, start_sector, GFP_NOFS);
+	if (!bio)
+		return -ENOMEM;
+	bio->bi_private = dip;
+	bio->bi_end_io = btrfs_end_dio_bio;
+	atomic_inc(&dip->pending_bios);
+
 	while (bvec <= (orig_bio->bi_io_vec + orig_bio->bi_vcnt - 1)) {
 		if (unlikely(map_length < submit_len + bvec->bv_len ||
 		    bio_add_page(bio, bvec->bv_page, bvec->bv_len,
@@ -6020,7 +6076,7 @@ static int btrfs_submit_direct_hook(int rw, struct btrfs_dio_private *dip,
 			atomic_inc(&dip->pending_bios);
 			ret = __btrfs_submit_dio_bio(bio, inode, rw,
 						     file_offset, skip_sum,
-						     csums);
+						     csums, async_submit);
 			if (ret) {
 				bio_put(bio);
 				atomic_dec(&dip->pending_bios);
@@ -6057,8 +6113,9 @@ static int btrfs_submit_direct_hook(int rw, struct btrfs_dio_private *dip,
 		}
 	}
 
+submit:
 	ret = __btrfs_submit_dio_bio(bio, inode, rw, file_offset, skip_sum,
-				     csums);
+				     csums, async_submit);
 	if (!ret)
 		return 0;
 
diff --git a/fs/btrfs/transaction.c b/fs/btrfs/transaction.c
index 5b158da7e0b..c571734d5e5 100644
--- a/fs/btrfs/transaction.c
+++ b/fs/btrfs/transaction.c
@@ -32,10 +32,8 @@
 
 static noinline void put_transaction(struct btrfs_transaction *transaction)
 {
-	WARN_ON(transaction->use_count == 0);
-	transaction->use_count--;
-	if (transaction->use_count == 0) {
-		list_del_init(&transaction->list);
+	WARN_ON(atomic_read(&transaction->use_count) == 0);
+	if (atomic_dec_and_test(&transaction->use_count)) {
 		memset(transaction, 0, sizeof(*transaction));
 		kmem_cache_free(btrfs_transaction_cachep, transaction);
 	}
@@ -60,14 +58,14 @@ static noinline int join_transaction(struct btrfs_root *root)
 		if (!cur_trans)
 			return -ENOMEM;
 		root->fs_info->generation++;
-		cur_trans->num_writers = 1;
+		atomic_set(&cur_trans->num_writers, 1);
 		cur_trans->num_joined = 0;
 		cur_trans->transid = root->fs_info->generation;
 		init_waitqueue_head(&cur_trans->writer_wait);
 		init_waitqueue_head(&cur_trans->commit_wait);
 		cur_trans->in_commit = 0;
 		cur_trans->blocked = 0;
-		cur_trans->use_count = 1;
+		atomic_set(&cur_trans->use_count, 1);
 		cur_trans->commit_done = 0;
 		cur_trans->start_time = get_seconds();
 
@@ -88,7 +86,7 @@ static noinline int join_transaction(struct btrfs_root *root)
 		root->fs_info->running_transaction = cur_trans;
 		spin_unlock(&root->fs_info->new_trans_lock);
 	} else {
-		cur_trans->num_writers++;
+		atomic_inc(&cur_trans->num_writers);
 		cur_trans->num_joined++;
 	}
 
@@ -145,7 +143,7 @@ static void wait_current_trans(struct btrfs_root *root)
 	cur_trans = root->fs_info->running_transaction;
 	if (cur_trans && cur_trans->blocked) {
 		DEFINE_WAIT(wait);
-		cur_trans->use_count++;
+		atomic_inc(&cur_trans->use_count);
 		while (1) {
 			prepare_to_wait(&root->fs_info->transaction_wait, &wait,
 					TASK_UNINTERRUPTIBLE);
@@ -181,6 +179,7 @@ static struct btrfs_trans_handle *start_transaction(struct btrfs_root *root,
 {
 	struct btrfs_trans_handle *h;
 	struct btrfs_transaction *cur_trans;
+	int retries = 0;
 	int ret;
 
 	if (root->fs_info->fs_state & BTRFS_SUPER_FLAG_ERROR)
@@ -204,7 +203,7 @@ again:
 	}
 
 	cur_trans = root->fs_info->running_transaction;
-	cur_trans->use_count++;
+	atomic_inc(&cur_trans->use_count);
 	if (type != TRANS_JOIN_NOLOCK)
 		mutex_unlock(&root->fs_info->trans_mutex);
 
@@ -224,10 +223,18 @@ again:
 
 	if (num_items > 0) {
 		ret = btrfs_trans_reserve_metadata(h, root, num_items);
-		if (ret == -EAGAIN) {
+		if (ret == -EAGAIN && !retries) {
+			retries++;
 			btrfs_commit_transaction(h, root);
 			goto again;
+		} else if (ret == -EAGAIN) {
+			/*
+			 * We have already retried and got EAGAIN, so really we
+			 * don't have space, so set ret to -ENOSPC.
+			 */
+			ret = -ENOSPC;
 		}
+
 		if (ret < 0) {
 			btrfs_end_transaction(h, root);
 			return ERR_PTR(ret);
@@ -327,7 +334,7 @@ int btrfs_wait_for_commit(struct btrfs_root *root, u64 transid)
 			goto out_unlock;  /* nothing committing|committed */
 	}
 
-	cur_trans->use_count++;
+	atomic_inc(&cur_trans->use_count);
 	mutex_unlock(&root->fs_info->trans_mutex);
 
 	wait_for_commit(root, cur_trans);
@@ -457,18 +464,14 @@ static int __btrfs_end_transaction(struct btrfs_trans_handle *trans,
 			wake_up_process(info->transaction_kthread);
 	}
 
-	if (lock)
-		mutex_lock(&info->trans_mutex);
 	WARN_ON(cur_trans != info->running_transaction);
-	WARN_ON(cur_trans->num_writers < 1);
-	cur_trans->num_writers--;
+	WARN_ON(atomic_read(&cur_trans->num_writers) < 1);
+	atomic_dec(&cur_trans->num_writers);
 
 	smp_mb();
 	if (waitqueue_active(&cur_trans->writer_wait))
 		wake_up(&cur_trans->writer_wait);
 	put_transaction(cur_trans);
-	if (lock)
-		mutex_unlock(&info->trans_mutex);
 
 	if (current->journal_info == trans)
 		current->journal_info = NULL;
@@ -1178,7 +1181,7 @@ int btrfs_commit_transaction_async(struct btrfs_trans_handle *trans,
 	/* take transaction reference */
 	mutex_lock(&root->fs_info->trans_mutex);
 	cur_trans = trans->transaction;
-	cur_trans->use_count++;
+	atomic_inc(&cur_trans->use_count);
 	mutex_unlock(&root->fs_info->trans_mutex);
 
 	btrfs_end_transaction(trans, root);
@@ -1237,7 +1240,7 @@ int btrfs_commit_transaction(struct btrfs_trans_handle *trans,
 
 	mutex_lock(&root->fs_info->trans_mutex);
 	if (cur_trans->in_commit) {
-		cur_trans->use_count++;
+		atomic_inc(&cur_trans->use_count);
 		mutex_unlock(&root->fs_info->trans_mutex);
 		btrfs_end_transaction(trans, root);
 
@@ -1259,7 +1262,7 @@ int btrfs_commit_transaction(struct btrfs_trans_handle *trans,
 		prev_trans = list_entry(cur_trans->list.prev,
 					struct btrfs_transaction, list);
 		if (!prev_trans->commit_done) {
-			prev_trans->use_count++;
+			atomic_inc(&prev_trans->use_count);
 			mutex_unlock(&root->fs_info->trans_mutex);
 
 			wait_for_commit(root, prev_trans);
@@ -1300,14 +1303,14 @@ int btrfs_commit_transaction(struct btrfs_trans_handle *trans,
 				TASK_UNINTERRUPTIBLE);
 
 		smp_mb();
-		if (cur_trans->num_writers > 1)
+		if (atomic_read(&cur_trans->num_writers) > 1)
 			schedule_timeout(MAX_SCHEDULE_TIMEOUT);
 		else if (should_grow)
 			schedule_timeout(1);
 
 		mutex_lock(&root->fs_info->trans_mutex);
 		finish_wait(&cur_trans->writer_wait, &wait);
-	} while (cur_trans->num_writers > 1 ||
+	} while (atomic_read(&cur_trans->num_writers) > 1 ||
 		 (should_grow && cur_trans->num_joined != joined));
 
 	ret = create_pending_snapshots(trans, root->fs_info);
@@ -1394,6 +1397,7 @@ int btrfs_commit_transaction(struct btrfs_trans_handle *trans,
 
 	wake_up(&cur_trans->commit_wait);
 
+	list_del_init(&cur_trans->list);
 	put_transaction(cur_trans);
 	put_transaction(cur_trans);
 
diff --git a/fs/btrfs/transaction.h b/fs/btrfs/transaction.h
index 229a594cacd..e441acc6c58 100644
--- a/fs/btrfs/transaction.h
+++ b/fs/btrfs/transaction.h
@@ -27,11 +27,11 @@ struct btrfs_transaction {
 	 * total writers in this transaction, it must be zero before the
 	 * transaction can end
 	 */
-	unsigned long num_writers;
+	atomic_t num_writers;
 
 	unsigned long num_joined;
 	int in_commit;
-	int use_count;
+	atomic_t use_count;
 	int commit_done;
 	int blocked;
 	struct list_head list;