Btrfs: Data ordered fixes

* In btrfs_delete_inode, wait for ordered extents after calling
truncate_inode_pages.  This is much faster, and more correct

* Properly clear our the PageChecked bit everywhere we redirty the page.

* Change the writepage fixup handler to lock the page range and check to
see if an ordered extent had been inserted since the improperly dirtied
page was discovered

* Wait for ordered extents outside the transaction.  This isn't required
for locking rules but does improve transaction latencies

* Reduce contention on the alloc_mutex by dropping it while incrementing
refs on a node/leaf and while dropping refs on a leaf.

Signed-off-by: Chris Mason <chris.mason@oracle.com>

4 files changed, 43 insertions, 10 deletions

diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c
index be2aef1..ccd4932 100644
--- a/fs/btrfs/extent-tree.c
+++ b/fs/btrfs/extent-tree.c
@@ -934,7 +934,6 @@ int btrfs_inc_ref(struct btrfs_trans_handle *trans, struct btrfs_root *root,
 	if (!root->ref_cows)
 		return 0;
 
-	mutex_lock(&root->fs_info->alloc_mutex);
 	level = btrfs_header_level(buf);
 	nritems = btrfs_header_nritems(buf);
 	for (i = 0; i < nritems; i++) {
@@ -951,29 +950,36 @@ int btrfs_inc_ref(struct btrfs_trans_handle *trans, struct btrfs_root *root,
 			disk_bytenr = btrfs_file_extent_disk_bytenr(buf, fi);
 			if (disk_bytenr == 0)
 				continue;
+
+			mutex_lock(&root->fs_info->alloc_mutex);
 			ret = __btrfs_inc_extent_ref(trans, root, disk_bytenr,
 				    btrfs_file_extent_disk_num_bytes(buf, fi),
 				    root->root_key.objectid, trans->transid,
 				    key.objectid, key.offset);
+			mutex_unlock(&root->fs_info->alloc_mutex);
 			if (ret) {
 				faili = i;
+				WARN_ON(1);
 				goto fail;
 			}
 		} else {
 			bytenr = btrfs_node_blockptr(buf, i);
 			btrfs_node_key_to_cpu(buf, &key, i);
+
+			mutex_lock(&root->fs_info->alloc_mutex);
 			ret = __btrfs_inc_extent_ref(trans, root, bytenr,
 					   btrfs_level_size(root, level - 1),
 					   root->root_key.objectid,
 					   trans->transid,
 					   level - 1, key.objectid);
+			mutex_unlock(&root->fs_info->alloc_mutex);
 			if (ret) {
 				faili = i;
+				WARN_ON(1);
 				goto fail;
 			}
 		}
 	}
-	mutex_unlock(&root->fs_info->alloc_mutex);
 	return 0;
 fail:
 	WARN_ON(1);
@@ -1004,7 +1010,6 @@ fail:
 		}
 	}
 #endif
-	mutex_unlock(&root->fs_info->alloc_mutex);
 	return ret;
 }
 
@@ -2180,6 +2185,8 @@ static int noinline drop_leaf_ref(struct btrfs_trans_handle *trans,
 	leaf_owner = btrfs_header_owner(leaf);
 	leaf_generation = btrfs_header_generation(leaf);
 
+	mutex_unlock(&root->fs_info->alloc_mutex);
+
 	for (i = 0; i < nritems; i++) {
 		u64 disk_bytenr;
 
@@ -2197,12 +2204,17 @@ static int noinline drop_leaf_ref(struct btrfs_trans_handle *trans,
 		disk_bytenr = btrfs_file_extent_disk_bytenr(leaf, fi);
 		if (disk_bytenr == 0)
 			continue;
+
+		mutex_lock(&root->fs_info->alloc_mutex);
 		ret = __btrfs_free_extent(trans, root, disk_bytenr,
 				btrfs_file_extent_disk_num_bytes(leaf, fi),
 				leaf_owner, leaf_generation,
 				key.objectid, key.offset, 0);
+		mutex_unlock(&root->fs_info->alloc_mutex);
 		BUG_ON(ret);
 	}
+
+	mutex_lock(&root->fs_info->alloc_mutex);
 	return 0;
 }
 
diff --git a/fs/btrfs/file.c b/fs/btrfs/file.c
index eccdb95..591a302 100644
--- a/fs/btrfs/file.c
+++ b/fs/btrfs/file.c
@@ -75,6 +75,7 @@ static void btrfs_drop_pages(struct page **pages, size_t num_pages)
 	for (i = 0; i < num_pages; i++) {
 		if (!pages[i])
 			break;
+		ClearPageChecked(pages[i]);
 		unlock_page(pages[i]);
 		mark_page_accessed(pages[i]);
 		page_cache_release(pages[i]);
diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c
index 50ee4be..8fb6dc2 100644
--- a/fs/btrfs/inode.c
+++ b/fs/btrfs/inode.c
@@ -418,7 +418,7 @@ void btrfs_writepage_fixup_worker(struct btrfs_work *work)
 
 	fixup = container_of(work, struct btrfs_writepage_fixup, work);
 	page = fixup->page;
-
+again:
 	lock_page(page);
 	if (!page->mapping || !PageDirty(page) || !PageChecked(page)) {
 		ClearPageChecked(page);
@@ -430,9 +430,21 @@ void btrfs_writepage_fixup_worker(struct btrfs_work *work)
 	page_end = page_offset(page) + PAGE_CACHE_SIZE - 1;
 
 	lock_extent(&BTRFS_I(inode)->io_tree, page_start, page_end, GFP_NOFS);
-	ordered = btrfs_lookup_ordered_extent(inode, page_start);
-	if (ordered)
+
+	/* already ordered? We're done */
+	if (test_range_bit(&BTRFS_I(inode)->io_tree, page_start, page_end,
+			     EXTENT_ORDERED, 0)) {
 		goto out;
+	}
+
+	ordered = btrfs_lookup_ordered_extent(inode, page_start);
+	if (ordered) {
+		unlock_extent(&BTRFS_I(inode)->io_tree, page_start,
+			      page_end, GFP_NOFS);
+		unlock_page(page);
+		btrfs_start_ordered_extent(inode, ordered, 1);
+		goto again;
+	}
 
 	set_extent_delalloc(&BTRFS_I(inode)->io_tree, page_start, page_end,
 			    GFP_NOFS);
@@ -1465,11 +1477,11 @@ void btrfs_delete_inode(struct inode *inode)
 	unsigned long nr;
 	int ret;
 
-	btrfs_wait_ordered_range(inode, 0, (u64)-1);
 	truncate_inode_pages(&inode->i_data, 0);
 	if (is_bad_inode(inode)) {
 		goto no_delete;
 	}
+	btrfs_wait_ordered_range(inode, 0, (u64)-1);
 
 	btrfs_i_size_write(inode, 0);
 	trans = btrfs_start_transaction(root, 1);
@@ -2707,6 +2719,7 @@ static void btrfs_invalidatepage(struct page *page, unsigned long offset)
 		 1, 1, GFP_NOFS);
 	__btrfs_releasepage(page, GFP_NOFS);
 
+	ClearPageChecked(page);
 	if (PagePrivate(page)) {
 		invalidate_extent_lru(tree, page_offset(page),
 				      PAGE_CACHE_SIZE);
@@ -2818,10 +2831,10 @@ static void btrfs_truncate(struct inode *inode)
 		return;
 
 	btrfs_truncate_page(inode->i_mapping, inode->i_size);
+	btrfs_wait_ordered_range(inode, inode->i_size & (~mask), (u64)-1);
 
 	trans = btrfs_start_transaction(root, 1);
 	btrfs_set_trans_block_group(trans, inode);
-	btrfs_wait_ordered_range(inode, inode->i_size & (~mask), (u64)-1);
 	btrfs_i_size_write(inode, inode->i_size);
 
 	/* FIXME, add redo link to tree so we don't leak on crash */
diff --git a/fs/btrfs/ordered-data.c b/fs/btrfs/ordered-data.c
index c2b4a9c..0d87795 100644
--- a/fs/btrfs/ordered-data.c
+++ b/fs/btrfs/ordered-data.c
@@ -336,7 +336,7 @@ void btrfs_wait_ordered_range(struct inode *inode, u64 start, u64 len)
 		orig_end = start + len - 1;
 		wait_end = orig_end;
 	}
-
+again:
 	/* start IO across the range first to instantiate any delalloc
 	 * extents
 	 */
@@ -369,6 +369,14 @@ void btrfs_wait_ordered_range(struct inode *inode, u64 start, u64 len)
 			break;
 		end--;
 	}
+	if (test_range_bit(&BTRFS_I(inode)->io_tree, start, orig_end,
+			   EXTENT_ORDERED | EXTENT_DELALLOC, 0)) {
+		printk("inode %lu still ordered or delalloc after wait "
+		       "%llu %llu\n", inode->i_ino,
+		       (unsigned long long)start,
+		       (unsigned long long)orig_end);
+		goto again;
+	}
 }
 
 /*
@@ -545,7 +553,6 @@ int btrfs_find_ordered_sum(struct inode *inode, u64 offset, u32 *sum)
 			sector_sums = &ordered_sum->sums;
 			for (i = 0; i < num_sectors; i++) {
 				if (sector_sums[i].offset == offset) {
-printk("find ordered sum inode %lu offset %Lu\n", inode->i_ino, offset);
 					*sum = sector_sums[i].sum;
 					ret = 0;
 					goto out;