btrfsck: add code to rebuild extent records

This also includes a new --repair btrfsck option.  For now it can
only fix errors in the extent allocation tree.

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

btrfsck.c

@@ -23,6 +23,7 @@
 #include <unistd.h>
 #include <fcntl.h>
 #include <sys/stat.h>
+#include <getopt.h>
 #include "kerncompat.h"
 #include "ctree.h"
 #include "disk-io.h"
@@ -74,9 +75,13 @@ struct extent_record {
 	struct cache_extent cache;
 	struct btrfs_disk_key parent_key;
 	u64 start;
+	u64 max_size;
 	u64 nr;
 	u64 refs;
 	u64 extent_item_refs;
+	u64 generation;
+	u64 info_objectid;
+	u8 info_level;
 	unsigned int content_checked:1;
 	unsigned int owner_ref_checked:1;
 	unsigned int is_root:1;
@@ -1082,7 +1087,9 @@ static int walk_down_tree(struct btrfs_root *root, struct btrfs_path *path,
 	ret = btrfs_lookup_extent_info(NULL, root,
 				       path->nodes[*level]->start,
 				       path->nodes[*level]->len, &refs, NULL);
-	BUG_ON(ret);
+	if (ret < 0)
+		goto out;
+
 	if (refs > 1) {
 		ret = enter_shared_node(root, path->nodes[*level]->start,
 					refs, wc, *level);
@@ -1109,7 +1116,8 @@ static int walk_down_tree(struct btrfs_root *root, struct btrfs_path *path,
 		blocksize = btrfs_level_size(root, *level - 1);
 		ret = btrfs_lookup_extent_info(NULL, root, bytenr, blocksize,
 					       &refs, NULL);
-		BUG_ON(ret);
+		if (ret < 0)
+			refs = 0;
 
 		if (refs > 1) {
 			ret = enter_shared_node(root, bytenr, refs,
@@ -1834,12 +1842,12 @@ static int all_backpointers_checked(struct extent_record *rec, int print_errs)
 			if (!print_errs)
 				goto out;
 			tback = (struct tree_backref *)back;
-			fprintf(stderr, "Backref %llu %s %llu not referenced\n",
+			fprintf(stderr, "Backref %llu %s %llu not referenced back %p\n",
 				(unsigned long long)rec->start,
 				back->full_backref ? "parent" : "root",
 				back->full_backref ?
 				(unsigned long long)tback->parent :
-				(unsigned long long)tback->root);
+				(unsigned long long)tback->root, back);
 		}
 		if (back->is_data) {
 			dback = (struct data_backref *)back;
@@ -1849,7 +1857,7 @@ static int all_backpointers_checked(struct extent_record *rec, int print_errs)
 					goto out;
 				fprintf(stderr, "Incorrect local backref count"
 					" on %llu %s %llu owner %llu"
-					" offset %llu found %u wanted %u\n",
+					" offset %llu found %u wanted %u back %p\n",
 					(unsigned long long)rec->start,
 					back->full_backref ?
 					"parent" : "root",
@@ -1858,7 +1866,7 @@ static int all_backpointers_checked(struct extent_record *rec, int print_errs)
 					(unsigned long long)dback->root,
 					(unsigned long long)dback->owner,
 					(unsigned long long)dback->offset,
-					dback->found_ref, dback->num_refs);
+					dback->found_ref, dback->num_refs, back);
 			}
 		}
 		if (!back->is_data) {
@@ -1965,12 +1973,28 @@ static int check_block(struct btrfs_root *root,
 {
 	struct extent_record *rec;
 	struct cache_extent *cache;
+	struct btrfs_key key;
 	int ret = 1;
+	int level;
 
 	cache = find_cache_extent(extent_cache, buf->start, buf->len);
 	if (!cache)
 		return 1;
 	rec = container_of(cache, struct extent_record, cache);
+	rec->generation = btrfs_header_generation(buf);
+
+	level = btrfs_header_level(buf);
+	if (btrfs_header_nritems(buf) > 0) {
+
+		if (level == 0)
+			btrfs_item_key_to_cpu(buf, &key, 0);
+		else
+			btrfs_node_key_to_cpu(buf, &key, 0);
+
+		rec->info_objectid = key.objectid;
+	}
+	rec->info_level = level;
+
 	if (btrfs_is_leaf(buf)) {
 		ret = check_leaf(root, &rec->parent_key, buf);
 	} else {
@@ -2035,6 +2059,7 @@ static struct tree_backref *alloc_tree_backref(struct extent_record *rec,
 		ref->node.full_backref = 0;
 	}
 	list_add_tail(&ref->node.list, &rec->backrefs);
+
 	return ref;
 }
 
@@ -2052,7 +2077,7 @@ static struct data_backref *find_data_backref(struct extent_record *rec,
 		if (!node->is_data)
 			continue;
 		back = (struct data_backref *)node;
-		if (parent > 0) { 
+		if (parent > 0) {
 			if (!node->full_backref)
 				continue;
 			if (parent == back->parent)
@@ -2070,11 +2095,13 @@ static struct data_backref *find_data_backref(struct extent_record *rec,
 
 static struct data_backref *alloc_data_backref(struct extent_record *rec,
 						u64 parent, u64 root,
-						u64 owner, u64 offset)
+						u64 owner, u64 offset,
+						u64 max_size)
 {
 	struct data_backref *ref = malloc(sizeof(*ref));
 	memset(&ref->node, 0, sizeof(ref->node));
 	ref->node.is_data = 1;
+
 	if (parent > 0) {
 		ref->parent = parent;
 		ref->owner = 0;
@@ -2089,13 +2116,16 @@ static struct data_backref *alloc_data_backref(struct extent_record *rec,
 	ref->found_ref = 0;
 	ref->num_refs = 0;
 	list_add_tail(&ref->node.list, &rec->backrefs);
+	if (max_size > rec->max_size)
+		rec->max_size = max_size;
 	return ref;
 }
 
 static int add_extent_rec(struct cache_tree *extent_cache,
 			  struct btrfs_key *parent_key,
 			  u64 start, u64 nr, u64 extent_item_refs,
-			  int is_root, int inc_ref, int set_checked)
+			  int is_root, int inc_ref, int set_checked,
+			  u64 max_size)
 {
 	struct extent_record *rec;
 	struct cache_extent *cache;
@@ -2136,11 +2166,15 @@ static int add_extent_rec(struct cache_tree *extent_cache,
 		if (parent_key)
 			btrfs_cpu_key_to_disk(&rec->parent_key, parent_key);
 
+		if (rec->max_size < max_size)
+			rec->max_size = max_size;
+
 		maybe_free_extent_rec(extent_cache, rec);
 		return ret;
 	}
 	rec = malloc(sizeof(*rec));
 	rec->start = start;
+	rec->max_size = max_size;
 	rec->nr = nr;
 	rec->content_checked = 0;
 	rec->owner_ref_checked = 0;
@@ -2187,7 +2221,7 @@ static int add_tree_backref(struct cache_tree *extent_cache, u64 bytenr,
 
 	cache = find_cache_extent(extent_cache, bytenr, 1);
 	if (!cache) {
-		add_extent_rec(extent_cache, NULL, bytenr, 1, 0, 0, 0, 0);
+		add_extent_rec(extent_cache, NULL, bytenr, 1, 0, 0, 0, 0, 0);
 		cache = find_cache_extent(extent_cache, bytenr, 1);
 		if (!cache)
 			abort();
@@ -2226,7 +2260,7 @@ static int add_tree_backref(struct cache_tree *extent_cache, u64 bytenr,
 
 static int add_data_backref(struct cache_tree *extent_cache, u64 bytenr,
 			    u64 parent, u64 root, u64 owner, u64 offset,
-			    u32 num_refs, int found_ref)
+			    u32 num_refs, int found_ref, u64 max_size)
 {
 	struct extent_record *rec;
 	struct data_backref *back;
@@ -2234,7 +2268,8 @@ static int add_data_backref(struct cache_tree *extent_cache, u64 bytenr,
 
 	cache = find_cache_extent(extent_cache, bytenr, 1);
 	if (!cache) {
-		add_extent_rec(extent_cache, NULL, bytenr, 1, 0, 0, 0, 0);
+		add_extent_rec(extent_cache, NULL, bytenr, 1, 0, 0, 0, 0,
+			       max_size);
 		cache = find_cache_extent(extent_cache, bytenr, 1);
 		if (!cache)
 			abort();
@@ -2244,9 +2279,13 @@ static int add_data_backref(struct cache_tree *extent_cache, u64 bytenr,
 	if (rec->start != bytenr) {
 		abort();
 	}
+	if (rec->max_size < max_size)
+		rec->max_size = max_size;
+
 	back = find_data_backref(rec, parent, root, owner, offset);
 	if (!back)
-		back = alloc_data_backref(rec, parent, root, owner, offset);
+		back = alloc_data_backref(rec, parent, root, owner, offset,
+					  max_size);
 
 	if (found_ref) {
 		BUG_ON(num_refs != 1);
@@ -2359,11 +2398,10 @@ static int process_extent_ref_v0(struct cache_tree *extent_cache,
 	btrfs_item_key_to_cpu(leaf, &key, slot);
 	ref0 = btrfs_item_ptr(leaf, slot, struct btrfs_extent_ref_v0);
 	if (btrfs_ref_objectid_v0(leaf, ref0) < BTRFS_FIRST_FREE_OBJECTID) {
-		add_tree_backref(extent_cache, key.objectid, key.offset,
+		add_tree_backref(extent_cache, key.objectid, key.offset, 0, 0);
-				 0, 0);
 	} else {
 		add_data_backref(extent_cache, key.objectid, key.offset, 0,
-				 0, 0, btrfs_ref_count_v0(leaf, ref0), 0);
+				 0, 0, btrfs_ref_count_v0(leaf, ref0), 0, 0);
 	}
 	return 0;
 }
@@ -2396,14 +2434,14 @@ static int process_extent_item(struct cache_tree *extent_cache,
 		BUG();
 #endif
 		return add_extent_rec(extent_cache, NULL, key.objectid,
-				      key.offset, refs, 0, 0, 0);
+				      key.offset, refs, 0, 0, 0, key.offset);
 	}
 
 	ei = btrfs_item_ptr(eb, slot, struct btrfs_extent_item);
 	refs = btrfs_extent_refs(eb, ei);
 
 	add_extent_rec(extent_cache, NULL, key.objectid, key.offset,
-		       refs, 0, 0, 0);
+		       refs, 0, 0, 0, key.offset);
 
 	ptr = (unsigned long)(ei + 1);
 	if (btrfs_extent_flags(eb, ei) & BTRFS_EXTENT_FLAG_TREE_BLOCK)
@@ -2431,14 +2469,14 @@ static int process_extent_item(struct cache_tree *extent_cache,
 								       dref),
 					btrfs_extent_data_ref_offset(eb, dref),
 					btrfs_extent_data_ref_count(eb, dref),
-					0);
+					0, key.offset);
 			break;
 		case BTRFS_SHARED_DATA_REF_KEY:
 			sref = (struct btrfs_shared_data_ref *)(iref + 1);
 			add_data_backref(extent_cache, key.objectid, offset,
 					0, 0, 0,
 					btrfs_shared_data_ref_count(eb, sref),
-					0);
+					0, key.offset);
 			break;
 		default:
 			fprintf(stderr, "corrupt extent record: key %Lu %u %Lu\n",
@@ -2515,6 +2553,8 @@ static int run_next_block(struct btrfs_root *root,
 	nritems = btrfs_header_nritems(buf);
 
 	ret = btrfs_lookup_extent_info(NULL, root, bytenr, size, NULL, &flags);
+	if (ret < 0)
+		flags = BTRFS_BLOCK_FLAG_FULL_BACKREF;
 
 	if (flags & BTRFS_BLOCK_FLAG_FULL_BACKREF) {
 		parent = bytenr;
@@ -2573,7 +2613,7 @@ static int run_next_block(struct btrfs_root *root,
 								       ref),
 					btrfs_extent_data_ref_offset(buf, ref),
 					btrfs_extent_data_ref_count(buf, ref),
-					0);
+					0, root->sectorsize);
 				continue;
 			}
 			if (key.type == BTRFS_SHARED_DATA_REF_KEY) {
@@ -2583,7 +2623,7 @@ static int run_next_block(struct btrfs_root *root,
 				add_data_backref(extent_cache,
 					key.objectid, key.offset, 0, 0, 0, 
 					btrfs_shared_data_ref_count(buf, ref),
-					0);
+					0, root->sectorsize);
 				continue;
 			}
 			if (key.type != BTRFS_EXTENT_DATA_KEY)
@@ -2606,26 +2646,33 @@ static int run_next_block(struct btrfs_root *root,
 			ret = add_extent_rec(extent_cache, NULL,
 				   btrfs_file_extent_disk_bytenr(buf, fi),
 				   btrfs_file_extent_disk_num_bytes(buf, fi),
-				   0, 0, 1, 1);
+				   0, 0, 1, 1,
+				   btrfs_file_extent_disk_num_bytes(buf, fi));
 			add_data_backref(extent_cache,
 				btrfs_file_extent_disk_bytenr(buf, fi),
 				parent, owner, key.objectid, key.offset -
-				btrfs_file_extent_offset(buf, fi), 1, 1);
+				btrfs_file_extent_offset(buf, fi), 1, 1,
+				btrfs_file_extent_disk_num_bytes(buf, fi));
 			BUG_ON(ret);
 		}
 	} else {
 		int level;
+		struct btrfs_key first_key;
+
+		first_key.objectid = 0;
+
+		if (nritems > 0)
+			btrfs_item_key_to_cpu(buf, &first_key, 0);
 		level = btrfs_header_level(buf);
 		for (i = 0; i < nritems; i++) {
 			u64 ptr = btrfs_node_blockptr(buf, i);
 			u32 size = btrfs_level_size(root, level - 1);
 			btrfs_node_key_to_cpu(buf, &key, i);
 			ret = add_extent_rec(extent_cache, &key,
-					     ptr, size, 0, 0, 1, 0);
+					     ptr, size, 0, 0, 1, 0, size);
 			BUG_ON(ret);
 
-			add_tree_backref(extent_cache, ptr, parent, 
+			add_tree_backref(extent_cache, ptr, parent, owner, 1);
-					 owner, 1);
 
 			if (level > 1) {
 				add_pending(nodes, seen, ptr, size);
@@ -2663,25 +2710,313 @@ static int add_root_to_pending(struct extent_buffer *buf,
 	else
 		add_pending(pending, seen, buf->start, buf->len);
 	add_extent_rec(extent_cache, NULL, buf->start, buf->len,
-		       0, 1, 1, 0);
+		       0, 1, 1, 0, buf->len);
 
 	if (root_key->objectid == BTRFS_TREE_RELOC_OBJECTID ||
 	    btrfs_header_backref_rev(buf) < BTRFS_MIXED_BACKREF_REV)
-		add_tree_backref(extent_cache, buf->start, buf->start, 0, 1);
+		add_tree_backref(extent_cache, buf->start, buf->start,
+				 0, 1);
 	else
 		add_tree_backref(extent_cache, buf->start, 0,
 				 root_key->objectid, 1);
 	return 0;
 }
 
+static int delete_extent_records(struct btrfs_trans_handle *trans,
+				 struct btrfs_root *root,
+				 struct btrfs_path *path,
+				 u64 bytenr)
+{
+	struct btrfs_key key;
+	struct btrfs_key found_key;
+	struct extent_buffer *leaf;
+	int ret;
+	int slot;
+
+
+	key.objectid = bytenr;
+	key.type = (u8)-1;
+	key.offset = (u64)-1;
+
+	while(1) {
+		ret = btrfs_search_slot(trans, root->fs_info->extent_root,
+					&key, path, 0, 1);
+		if (ret < 0)
+			break;
+
+		if (ret > 0) {
+			ret = 0;
+			if (path->slots[0] == 0)
+				break;
+			path->slots[0]--;
+		}
+		ret = 0;
+
+		leaf = path->nodes[0];
+		slot = path->slots[0];
+
+		btrfs_item_key_to_cpu(leaf, &found_key, slot);
+		if (found_key.objectid != bytenr)
+			break;
+
+		if (found_key.type != BTRFS_EXTENT_ITEM_KEY &&
+		    found_key.type != BTRFS_TREE_BLOCK_REF_KEY &&
+		    found_key.type != BTRFS_EXTENT_DATA_REF_KEY &&
+		    found_key.type != BTRFS_EXTENT_REF_V0_KEY &&
+		    found_key.type != BTRFS_SHARED_BLOCK_REF_KEY &&
+		    found_key.type != BTRFS_SHARED_DATA_REF_KEY) {
+			btrfs_release_path(NULL, path);
+			if (found_key.type == 0) {
+				if (found_key.offset == 0)
+					break;
+				key.offset = found_key.offset - 1;
+				key.type = found_key.type;
+			}
+			key.type = found_key.type - 1;
+			key.offset = (u64)-1;
+			continue;
+		}
+
+		fprintf(stderr, "repair deleting extent record: key %Lu %u %Lu\n",
+			found_key.objectid, found_key.type, found_key.offset);
+
+		ret = btrfs_del_item(trans, root->fs_info->extent_root, path);
+		if (ret)
+			break;
+		btrfs_release_path(NULL, path);
+
+		if (found_key.type == BTRFS_EXTENT_ITEM_KEY) {
+			ret = btrfs_update_block_group(trans, root, bytenr,
+						       found_key.offset, 0, 1);
+			if (ret)
+				break;
+		}
+	}
+
+	btrfs_release_path(NULL, path);
+	return ret;
+}
+
+/*
+ * for a single backref, this will allocate a new extent
+ * and add the backref to it.
+ */
+static int record_extent(struct btrfs_trans_handle *trans,
+			 struct btrfs_fs_info *info,
+			 struct btrfs_path *path,
+			 struct extent_record *rec,
+			 struct extent_backref *back,
+			 int allocated, u64 flags)
+{
+	int ret;
+	struct btrfs_root *extent_root = info->extent_root;
+	struct extent_buffer *leaf;
+	struct btrfs_key ins_key;
+	struct btrfs_extent_item *ei;
+	struct tree_backref *tback;
+	struct data_backref *dback;
+	struct btrfs_tree_block_info *bi;
+
+	if (!back->is_data)
+		rec->max_size = max_t(u64, rec->max_size,
+				    info->extent_root->leafsize);
+
+	if (!allocated) {
+		u32 item_size = sizeof(*ei);
+
+		if (!back->is_data)
+			item_size += sizeof(*bi);
+
+		ins_key.objectid = rec->start;
+		ins_key.offset = rec->max_size;
+		ins_key.type = BTRFS_EXTENT_ITEM_KEY;
+
+		ret = btrfs_insert_empty_item(trans, extent_root, path,
+					&ins_key, item_size);
+		if (ret)
+			goto fail;
+
+		leaf = path->nodes[0];
+		ei = btrfs_item_ptr(leaf, path->slots[0],
+				    struct btrfs_extent_item);
+
+		btrfs_set_extent_refs(leaf, ei, 0);
+		btrfs_set_extent_generation(leaf, ei, rec->generation);
+
+		if (back->is_data) {
+			btrfs_set_extent_flags(leaf, ei,
+					       BTRFS_EXTENT_FLAG_DATA);
+		} else {
+			struct btrfs_disk_key copy_key;;
+
+			tback = (struct tree_backref *)back;
+			bi = (struct btrfs_tree_block_info *)(ei + 1);
+			memset_extent_buffer(leaf, 0, (unsigned long)bi,
+					     sizeof(*bi));
+			memset(&copy_key, 0, sizeof(copy_key));
+
+			copy_key.objectid = le64_to_cpu(rec->info_objectid);
+			btrfs_set_tree_block_level(leaf, bi, rec->info_level);
+			btrfs_set_tree_block_key(leaf, bi, &copy_key);
+
+			btrfs_set_extent_flags(leaf, ei,
+					       BTRFS_EXTENT_FLAG_TREE_BLOCK | flags);
+		}
+
+		btrfs_mark_buffer_dirty(leaf);
+		ret = btrfs_update_block_group(trans, extent_root, rec->start,
+					       rec->max_size, 1, 0);
+		if (ret)
+			goto fail;
+		btrfs_release_path(NULL, path);
+	}
+
+	if (back->is_data) {
+		u64 parent;
+		int i;
+
+		dback = (struct data_backref *)back;
+		if (back->full_backref)
+			parent = dback->parent;
+		else
+			parent = 0;
+
+		for (i = 0; i < dback->found_ref; i++) {
+			/* if parent != 0, we're doing a full backref
+			 * passing BTRFS_FIRST_FREE_OBJECTID as the owner
+			 * just makes the backref allocator create a data
+			 * backref
+			 */
+			ret = btrfs_inc_extent_ref(trans, info->extent_root,
+						   rec->start, rec->max_size,
+						   parent,
+						   dback->root,
+						   parent ?
+						   BTRFS_FIRST_FREE_OBJECTID :
+						   dback->owner,
+						   dback->offset);
+			if (ret)
+				break;
+		}
+		fprintf(stderr, "adding new data backref"
+				" on %llu %s %llu owner %llu"
+				" offset %llu found %d\n",
+				(unsigned long long)rec->start,
+				back->full_backref ?
+				"parent" : "root",
+				back->full_backref ?
+				(unsigned long long)parent :
+				(unsigned long long)dback->root,
+				(unsigned long long)dback->owner,
+				(unsigned long long)dback->offset,
+				dback->found_ref);
+	} else {
+		u64 parent;
+
+		tback = (struct tree_backref *)back;
+		if (back->full_backref)
+			parent = tback->parent;
+		else
+			parent = 0;
+
+		ret = btrfs_inc_extent_ref(trans, info->extent_root,
+					   rec->start, rec->max_size,
+					   parent, tback->root, 0, 0);
+		fprintf(stderr, "adding new tree backref on "
+			"start %llu len %llu parent %llu root %llu\n",
+			rec->start, rec->max_size, tback->parent, tback->root);
+	}
+	if (ret)
+		goto fail;
+fail:
+	btrfs_release_path(NULL, path);
+	return ret;
+}
+
+/*
+ * when an incorrect extent item is found, this will delete
+ * all of the existing entries for it and recreate them
+ * based on what the tree scan found.
+ */
+static int fixup_extent_refs(struct btrfs_trans_handle *trans,
+			     struct btrfs_fs_info *info,
+			     struct extent_record *rec)
+{
+	int ret;
+	struct btrfs_path *path;
+	struct list_head *cur = rec->backrefs.next;
+	struct extent_backref *back;
+	int allocated = 0;
+	u64 flags = 0;
+
+	/* remember our flags for recreating the extent */
+	ret = btrfs_lookup_extent_info(NULL, info->extent_root, rec->start,
+				       rec->max_size, NULL, &flags);
+	if (ret < 0)
+		flags = BTRFS_BLOCK_FLAG_FULL_BACKREF;
+
+	path = btrfs_alloc_path();
+
+	/* step one, delete all the existing records */
+	ret = delete_extent_records(trans, info->extent_root, path,
+				    rec->start);
+
+	if (ret < 0)
+		goto out;
+
+	/* step two, recreate all the refs we did find */
+	while(cur != &rec->backrefs) {
+		back = list_entry(cur, struct extent_backref, list);
+		cur = cur->next;
+
+		/*
+		 * if we didn't find any references, don't create a
+		 * new extent record
+		 */
+		if (!back->found_ref)
+			continue;
+
+		ret = record_extent(trans, info, path, rec, back, allocated, flags);
+		allocated = 1;
+
+		if (ret)
+			goto out;
+	}
+out:
+	btrfs_free_path(path);
+	return ret;
+}
+
 static int check_extent_refs(struct btrfs_root *root,
-		      struct cache_tree *extent_cache)
+		      struct cache_tree *extent_cache, int repair)
 {
 	struct extent_record *rec;
 	struct cache_extent *cache;
+	struct btrfs_trans_handle *trans = NULL;
 	int err = 0;
+	int ret = 0;
+	int fixed = 0;
 
+	if (repair)
+		trans = btrfs_start_transaction(root, 1);
+
+	if (repair) {
+		/*
+		 * if we're doing a repair, we have to make sure
+		 * we don't allocate from the problem extents.
+		 * In the worst case, this will be all the
+		 * extents in the FS
+		 */
+		cache = find_first_cache_extent(extent_cache, 0);
+		while(cache) {
+			rec = container_of(cache, struct extent_record, cache);
+			btrfs_pin_extent(root->fs_info,
+					 rec->start, rec->nr);
+			cache = next_cache_extent(cache);
+		}
+	}
 	while(1) {
+		fixed = 0;
 		cache = find_first_cache_extent(extent_cache, 0);
 		if (!cache)
 			break;
@@ -2693,19 +3028,39 @@ static int check_extent_refs(struct btrfs_root *root,
 			fprintf(stderr, "extent item %llu, found %llu\n",
 				(unsigned long long)rec->extent_item_refs,
 				(unsigned long long)rec->refs);
+			if (!fixed && repair) {
+				ret = fixup_extent_refs(trans, root->fs_info, rec);
+				if (ret)
+					goto repair_abort;
+				fixed = 1;
+			}
 			err = 1;
+
 		}
 		if (all_backpointers_checked(rec, 1)) {
 			fprintf(stderr, "backpointer mismatch on [%llu %llu]\n",
 				(unsigned long long)rec->start,
 				(unsigned long long)rec->nr);
 
+			if (!fixed && repair) {
+				ret = fixup_extent_refs(trans, root->fs_info, rec);
+				if (ret)
+					goto repair_abort;
+				fixed = 1;
+			}
+
 			err = 1;
 		}
 		if (!rec->owner_ref_checked) {
 			fprintf(stderr, "owner ref check failed [%llu %llu]\n",
 				(unsigned long long)rec->start,
 				(unsigned long long)rec->nr);
+			if (!fixed && repair) {
+				ret = fixup_extent_refs(trans, root->fs_info, rec);
+				if (ret)
+					goto repair_abort;
+				fixed = 1;
+			}
 			err = 1;
 		}
 
@@ -2713,10 +3068,18 @@ static int check_extent_refs(struct btrfs_root *root,
 		free_all_extent_backrefs(rec);
 		free(rec);
 	}
+repair_abort:
+	if (repair) {
+		if (ret) {
+			fprintf(stderr, "failed to repair damaged filesystem, aborting\n");
+			exit(1);
+		}
+		btrfs_commit_transaction(trans, root);
+	}
 	return err;
 }
 
-static int check_extents(struct btrfs_root *root)
+static int check_extents(struct btrfs_root *root, int repair)
 {
 	struct cache_tree extent_cache;
 	struct cache_tree seen;
@@ -2797,7 +3160,7 @@ static int check_extents(struct btrfs_root *root)
 		if (ret != 0)
 			break;
 	}
-	ret = check_extent_refs(root, &extent_cache);
+	ret = check_extent_refs(root, &extent_cache, repair);
 	return ret;
 }
 
@@ -2808,6 +3171,12 @@ static void print_usage(void)
 	exit(1);
 }
 
+static struct option long_options[] = {
+	{ "super", 1, NULL, 's' },
+	{ "repair", 0, NULL, 0 },
+	{ 0, 0, 0, 0}
+};
+
 int main(int ac, char **av)
 {
 	struct cache_tree root_cache;
@@ -2816,10 +3185,13 @@ int main(int ac, char **av)
 	u64 bytenr = 0;
 	int ret;
 	int num;
+	int repair = 0;
+	int option_index = 0;
 
 	while(1) {
 		int c;
-		c = getopt(ac, av, "s:");
+		c = getopt_long(ac, av, "", long_options,
+				&option_index);
 		if (c < 0)
 			break;
 		switch(c) {
@@ -2829,9 +3201,14 @@ int main(int ac, char **av)
 				printf("using SB copy %d, bytenr %llu\n", num,
 				       (unsigned long long)bytenr);
 				break;
-			default:
+			case '?':
 				print_usage();
 		}
+		if (option_index == 1) {
+			printf("enabling repair mode\n");
+			repair = 1;
+		}
+
 	}
 	ac = ac - optind;
 
@@ -2849,15 +3226,23 @@ int main(int ac, char **av)
 		return -EBUSY;
 	}
 
-	info = open_ctree_fs_info(av[optind], bytenr, 0, 0);
+	info = open_ctree_fs_info(av[optind], bytenr, repair, 1);
 
 	if (info == NULL)
 		return 1;
 
+	if (!extent_buffer_uptodate(info->tree_root->node) ||
+	    !extent_buffer_uptodate(info->dev_root->node) ||
+	    !extent_buffer_uptodate(info->extent_root->node) ||
+	    !extent_buffer_uptodate(info->chunk_root->node)) {
+		fprintf(stderr, "Critical roots corrupted, unable to fsck the FS\n");
+		return -EIO;
+	}
+
 	root = info->fs_root;
 
 	fprintf(stderr, "checking extents\n");
-	ret = check_extents(root);
+	ret = check_extents(root, repair);
 	if (ret)
 		goto out;
 	fprintf(stderr, "checking fs roots\n");

ctree.h

@@ -1785,6 +1785,7 @@ static inline u32 btrfs_level_size(struct btrfs_root *root, int level) {
 	btrfs_item_offset_nr(leaf, slot)))
 
 /* extent-tree.c */
+void btrfs_pin_extent(struct btrfs_fs_info *fs_info, u64 bytenr, u64 num_bytes);
 int btrfs_extent_post_op(struct btrfs_trans_handle *trans,
 			 struct btrfs_root *root);
 int btrfs_copy_pinned(struct btrfs_root *root, struct extent_io_tree *copy);

extent-tree.c

@@ -1041,8 +1041,6 @@ static int lookup_inline_extent_backref(struct btrfs_trans_handle *trans,
 	}
 	if (ret) {
 		printf("Failed to find [%llu, %u, %llu]\n", key.objectid, key.type, key.offset);
-		btrfs_print_leaf(root, path->nodes[0]);
-		btrfs_free_path(path);
 		return -ENOENT;
 	}
 
@@ -1067,8 +1065,9 @@ static int lookup_inline_extent_backref(struct btrfs_trans_handle *trans,
 	}
 #endif
 	if (item_size < sizeof(*ei)) {
-		printf("Size is %u, needs to be %u, slot %d\n", item_size,
+		printf("Size is %u, needs to be %u, slot %d\n",
-		       sizeof(*ei), path->slots[0]);
+		       (unsigned)item_size,
+		       (unsigned)sizeof(*ei), path->slots[0]);
 		btrfs_print_leaf(root, leaf);
 		return -EINVAL;
 	}
@@ -1460,10 +1459,8 @@ int btrfs_lookup_extent_info(struct btrfs_trans_handle *trans,
 	if (ret < 0)
 		goto out;
 	if (ret != 0) {
-		btrfs_print_leaf(root, path->nodes[0]);
+		ret = -EIO;
-		printk("failed to find block number %Lu\n",
+		goto out;
-		       (unsigned long long)bytenr);
-		BUG();
 	}
 
 	l = path->nodes[0];
@@ -1484,9 +1481,8 @@ int btrfs_lookup_extent_info(struct btrfs_trans_handle *trans,
 			extent_flags = BTRFS_BLOCK_FLAG_FULL_BACKREF;
 #else
 			BUG();
-#endif		
+#endif
-		}
+	}
-		BUG_ON(num_refs == 0);
 	item = btrfs_item_ptr(l, path->slots[0], struct btrfs_extent_item);
 	if (refs)
 		*refs = num_refs;
@@ -2033,6 +2029,12 @@ pinit:
 	return 0;
 }
 
+void btrfs_pin_extent(struct btrfs_fs_info *fs_info,
+		       u64 bytenr, u64 num_bytes)
+{
+	update_pinned_extents(fs_info->extent_root, bytenr, num_bytes, 1);
+}
+
 /*
  * remove an extent from the root, returns 0 on success
  */

print-tree.c

@@ -239,7 +239,7 @@ static void print_extent_item(struct extent_buffer *eb, int slot)
 			       btrfs_shared_data_ref_count(eb, sref));
 			break;
 		default:
-			BUG();
+			return;
 		}
 		ptr += btrfs_extent_inline_ref_size(type);
 	}