1 files changed, 97 insertions, 37 deletions

diff --git a/fs/btrfs/relocation.c b/fs/btrfs/relocation.c
index 7f219851fa23..995d4b8b1cfd 100644
--- a/fs/btrfs/relocation.c
+++ b/fs/btrfs/relocation.c
@@ -21,6 +21,7 @@
 #include "qgroup.h"
 #include "print-tree.h"
 #include "delalloc-space.h"
+#include "block-group.h"
 
 /*
  * backref_node, mapping_node and tree_block start with this
@@ -146,7 +147,7 @@ struct file_extent_cluster {
 
 struct reloc_control {
 	/* block group to relocate */
-	struct btrfs_block_group_cache *block_group;
+	struct btrfs_block_group *block_group;
 	/* extent tree */
 	struct btrfs_root *extent_root;
 	/* inode for moving data */
@@ -516,6 +517,34 @@ static int update_backref_cache(struct btrfs_trans_handle *trans,
 	return 1;
 }
 
+static bool reloc_root_is_dead(struct btrfs_root *root)
+{
+	/*
+	 * Pair with set_bit/clear_bit in clean_dirty_subvols and
+	 * btrfs_update_reloc_root. We need to see the updated bit before
+	 * trying to access reloc_root
+	 */
+	smp_rmb();
+	if (test_bit(BTRFS_ROOT_DEAD_RELOC_TREE, &root->state))
+		return true;
+	return false;
+}
+
+/*
+ * Check if this subvolume tree has valid reloc tree.
+ *
+ * Reloc tree after swap is considered dead, thus not considered as valid.
+ * This is enough for most callers, as they don't distinguish dead reloc root
+ * from no reloc root.  But should_ignore_root() below is a special case.
+ */
+static bool have_reloc_root(struct btrfs_root *root)
+{
+	if (reloc_root_is_dead(root))
+		return false;
+	if (!root->reloc_root)
+		return false;
+	return true;
+}
 
 static int should_ignore_root(struct btrfs_root *root)
 {
@@ -524,6 +553,10 @@ static int should_ignore_root(struct btrfs_root *root)
 	if (!test_bit(BTRFS_ROOT_REF_COWS, &root->state))
 		return 0;
 
+	/* This root has been merged with its reloc tree, we can ignore it */
+	if (reloc_root_is_dead(root))
+		return 1;
+
 	reloc_root = root->reloc_root;
 	if (!reloc_root)
 		return 0;
@@ -1434,6 +1467,13 @@ int btrfs_init_reloc_root(struct btrfs_trans_handle *trans,
 	int clear_rsv = 0;
 	int ret;
 
+	/*
+	 * The subvolume has reloc tree but the swap is finished, no need to
+	 * create/update the dead reloc tree
+	 */
+	if (reloc_root_is_dead(root))
+		return 0;
+
 	if (root->reloc_root) {
 		reloc_root = root->reloc_root;
 		reloc_root->last_trans = trans->transid;
@@ -1470,8 +1510,7 @@ int btrfs_update_reloc_root(struct btrfs_trans_handle *trans,
 	struct btrfs_root_item *root_item;
 	int ret;
 
-	if (test_bit(BTRFS_ROOT_DEAD_RELOC_TREE, &root->state) ||
-	    !root->reloc_root)
+	if (!have_reloc_root(root))
 		goto out;
 
 	reloc_root = root->reloc_root;
@@ -1481,6 +1520,11 @@ int btrfs_update_reloc_root(struct btrfs_trans_handle *trans,
 	if (fs_info->reloc_ctl->merge_reloc_tree &&
 	    btrfs_root_refs(root_item) == 0) {
 		set_bit(BTRFS_ROOT_DEAD_RELOC_TREE, &root->state);
+		/*
+		 * Mark the tree as dead before we change reloc_root so
+		 * have_reloc_root will not touch it from now on.
+		 */
+		smp_wmb();
 		__del_reloc_root(reloc_root);
 	}
 
@@ -1552,11 +1596,10 @@ again:
 	return NULL;
 }
 
-static int in_block_group(u64 bytenr,
-			  struct btrfs_block_group_cache *block_group)
+static int in_block_group(u64 bytenr, struct btrfs_block_group *block_group)
 {
-	if (bytenr >= block_group->key.objectid &&
-	    bytenr < block_group->key.objectid + block_group->key.offset)
+	if (bytenr >= block_group->start &&
+	    bytenr < block_group->start + block_group->length)
 		return 1;
 	return 0;
 }
@@ -2186,7 +2229,6 @@ static int clean_dirty_subvols(struct reloc_control *rc)
 			/* Merged subvolume, cleanup its reloc root */
 			struct btrfs_root *reloc_root = root->reloc_root;
 
-			clear_bit(BTRFS_ROOT_DEAD_RELOC_TREE, &root->state);
 			list_del_init(&root->reloc_dirty_list);
 			root->reloc_root = NULL;
 			if (reloc_root) {
@@ -2195,6 +2237,12 @@ static int clean_dirty_subvols(struct reloc_control *rc)
 				if (ret2 < 0 && !ret)
 					ret = ret2;
 			}
+			/*
+			 * Need barrier to ensure clear_bit() only happens after
+			 * root->reloc_root = NULL. Pairs with have_reloc_root.
+			 */
+			smp_wmb();
+			clear_bit(BTRFS_ROOT_DEAD_RELOC_TREE, &root->state);
 			btrfs_put_fs_root(root);
 		} else {
 			/* Orphan reloc tree, just clean it up */
@@ -2238,7 +2286,7 @@ static noinline_for_stack int merge_reloc_root(struct reloc_control *rc,
 
 	if (btrfs_disk_key_objectid(&root_item->drop_progress) == 0) {
 		level = btrfs_root_level(root_item);
-		extent_buffer_get(reloc_root->node);
+		atomic_inc(&reloc_root->node->refs);
 		path->nodes[level] = reloc_root->node;
 		path->slots[level] = 0;
 	} else {
@@ -3187,7 +3235,6 @@ static noinline_for_stack
 int setup_extent_mapping(struct inode *inode, u64 start, u64 end,
 			 u64 block_start)
 {
-	struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
 	struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
 	struct extent_map *em;
 	int ret = 0;
@@ -3200,7 +3247,6 @@ int setup_extent_mapping(struct inode *inode, u64 start, u64 end,
 	em->len = end + 1 - start;
 	em->block_len = em->len;
 	em->block_start = block_start;
-	em->bdev = fs_info->fs_devices->latest_bdev;
 	set_bit(EXTENT_FLAG_PINNED, &em->flags);
 
 	lock_extent(&BTRFS_I(inode)->io_tree, start, end);
@@ -3269,6 +3315,8 @@ static int relocate_file_extent_cluster(struct inode *inode,
 			if (!page) {
 				btrfs_delalloc_release_metadata(BTRFS_I(inode),
 							PAGE_SIZE, true);
+				btrfs_delalloc_release_extents(BTRFS_I(inode),
+							PAGE_SIZE);
 				ret = -ENOMEM;
 				goto out;
 			}
@@ -3289,7 +3337,7 @@ static int relocate_file_extent_cluster(struct inode *inode,
 				btrfs_delalloc_release_metadata(BTRFS_I(inode),
 							PAGE_SIZE, true);
 				btrfs_delalloc_release_extents(BTRFS_I(inode),
-							       PAGE_SIZE, true);
+							       PAGE_SIZE);
 				ret = -EIO;
 				goto out;
 			}
@@ -3311,14 +3359,14 @@ static int relocate_file_extent_cluster(struct inode *inode,
 		}
 
 		ret = btrfs_set_extent_delalloc(inode, page_start, page_end, 0,
-						NULL, 0);
+						NULL);
 		if (ret) {
 			unlock_page(page);
 			put_page(page);
 			btrfs_delalloc_release_metadata(BTRFS_I(inode),
 							 PAGE_SIZE, true);
 			btrfs_delalloc_release_extents(BTRFS_I(inode),
-			                               PAGE_SIZE, true);
+			                               PAGE_SIZE);
 
 			clear_extent_bits(&BTRFS_I(inode)->io_tree,
 					  page_start, page_end,
@@ -3334,8 +3382,7 @@ static int relocate_file_extent_cluster(struct inode *inode,
 		put_page(page);
 
 		index++;
-		btrfs_delalloc_release_extents(BTRFS_I(inode), PAGE_SIZE,
-					       false);
+		btrfs_delalloc_release_extents(BTRFS_I(inode), PAGE_SIZE);
 		balance_dirty_pages_ratelimited(inode->i_mapping);
 		btrfs_throttle(fs_info);
 	}
@@ -3535,7 +3582,7 @@ static int block_use_full_backref(struct reloc_control *rc,
 }
 
 static int delete_block_group_cache(struct btrfs_fs_info *fs_info,
-				    struct btrfs_block_group_cache *block_group,
+				    struct btrfs_block_group *block_group,
 				    struct inode *inode,
 				    u64 ino)
 {
@@ -3551,7 +3598,7 @@ static int delete_block_group_cache(struct btrfs_fs_info *fs_info,
 	key.type = BTRFS_INODE_ITEM_KEY;
 	key.offset = 0;
 
-	inode = btrfs_iget(fs_info->sb, &key, root, NULL);
+	inode = btrfs_iget(fs_info->sb, &key, root);
 	if (IS_ERR(inode))
 		return -ENOENT;
 
@@ -3854,7 +3901,7 @@ int find_next_extent(struct reloc_control *rc, struct btrfs_path *path,
 	u64 start, end, last;
 	int ret;
 
-	last = rc->block_group->key.objectid + rc->block_group->key.offset;
+	last = rc->block_group->start + rc->block_group->length;
 	while (1) {
 		cond_resched();
 		if (rc->search_start >= last) {
@@ -3971,7 +4018,7 @@ int prepare_to_relocate(struct reloc_control *rc)
 		return -ENOMEM;
 
 	memset(&rc->cluster, 0, sizeof(rc->cluster));
-	rc->search_start = rc->block_group->key.objectid;
+	rc->search_start = rc->block_group->start;
 	rc->extents_found = 0;
 	rc->nodes_relocated = 0;
 	rc->merging_rsv_size = 0;
@@ -4210,7 +4257,7 @@ out:
  */
 static noinline_for_stack
 struct inode *create_reloc_inode(struct btrfs_fs_info *fs_info,
-				 struct btrfs_block_group_cache *group)
+				 struct btrfs_block_group *group)
 {
 	struct inode *inode = NULL;
 	struct btrfs_trans_handle *trans;
@@ -4237,9 +4284,9 @@ struct inode *create_reloc_inode(struct btrfs_fs_info *fs_info,
 	key.objectid = objectid;
 	key.type = BTRFS_INODE_ITEM_KEY;
 	key.offset = 0;
-	inode = btrfs_iget(fs_info->sb, &key, root, NULL);
+	inode = btrfs_iget(fs_info->sb, &key, root);
 	BUG_ON(IS_ERR(inode));
-	BTRFS_I(inode)->index_cnt = group->key.objectid;
+	BTRFS_I(inode)->index_cnt = group->start;
 
 	err = btrfs_orphan_add(trans, BTRFS_I(inode));
 out:
@@ -4274,7 +4321,7 @@ static struct reloc_control *alloc_reloc_control(struct btrfs_fs_info *fs_info)
  * Print the block group being relocated
  */
 static void describe_relocation(struct btrfs_fs_info *fs_info,
-				struct btrfs_block_group_cache *block_group)
+				struct btrfs_block_group *block_group)
 {
 	char buf[128] = {'\0'};
 
@@ -4282,7 +4329,16 @@ static void describe_relocation(struct btrfs_fs_info *fs_info,
 
 	btrfs_info(fs_info,
 		   "relocating block group %llu flags %s",
-		   block_group->key.objectid, buf);
+		   block_group->start, buf);
+}
+
+static const char *stage_to_string(int stage)
+{
+	if (stage == MOVE_DATA_EXTENTS)
+		return "move data extents";
+	if (stage == UPDATE_DATA_PTRS)
+		return "update data pointers";
+	return "unknown";
 }
 
 /*
@@ -4290,7 +4346,7 @@ static void describe_relocation(struct btrfs_fs_info *fs_info,
  */
 int btrfs_relocate_block_group(struct btrfs_fs_info *fs_info, u64 group_start)
 {
-	struct btrfs_block_group_cache *bg;
+	struct btrfs_block_group *bg;
 	struct btrfs_root *extent_root = fs_info->extent_root;
 	struct reloc_control *rc;
 	struct inode *inode;
@@ -4317,7 +4373,7 @@ int btrfs_relocate_block_group(struct btrfs_fs_info *fs_info, u64 group_start)
 	rc->extent_root = extent_root;
 	rc->block_group = bg;
 
-	ret = btrfs_inc_block_group_ro(rc->block_group);
+	ret = btrfs_inc_block_group_ro(rc->block_group, true);
 	if (ret) {
 		err = ret;
 		goto out;
@@ -4355,16 +4411,19 @@ int btrfs_relocate_block_group(struct btrfs_fs_info *fs_info, u64 group_start)
 	btrfs_wait_block_group_reservations(rc->block_group);
 	btrfs_wait_nocow_writers(rc->block_group);
 	btrfs_wait_ordered_roots(fs_info, U64_MAX,
-				 rc->block_group->key.objectid,
-				 rc->block_group->key.offset);
+				 rc->block_group->start,
+				 rc->block_group->length);
 
 	while (1) {
+		int finishes_stage;
+
 		mutex_lock(&fs_info->cleaner_mutex);
 		ret = relocate_block_group(rc);
 		mutex_unlock(&fs_info->cleaner_mutex);
 		if (ret < 0)
 			err = ret;
 
+		finishes_stage = rc->stage;
 		/*
 		 * We may have gotten ENOSPC after we already dirtied some
 		 * extents.  If writeout happens while we're relocating a
@@ -4390,13 +4449,13 @@ int btrfs_relocate_block_group(struct btrfs_fs_info *fs_info, u64 group_start)
 		if (rc->extents_found == 0)
 			break;
 
-		btrfs_info(fs_info, "found %llu extents", rc->extents_found);
-
+		btrfs_info(fs_info, "found %llu extents, stage: %s",
+			   rc->extents_found, stage_to_string(finishes_stage));
 	}
 
 	WARN_ON(rc->block_group->pinned > 0);
 	WARN_ON(rc->block_group->reserved > 0);
-	WARN_ON(btrfs_block_group_used(&rc->block_group->item) > 0);
+	WARN_ON(rc->block_group->used > 0);
 out:
 	if (err && rw)
 		btrfs_dec_block_group_ro(rc->block_group);
@@ -4546,6 +4605,7 @@ int btrfs_recover_relocation(struct btrfs_root *root)
 		fs_root = read_fs_root(fs_info, reloc_root->root_key.offset);
 		if (IS_ERR(fs_root)) {
 			err = PTR_ERR(fs_root);
+			list_add_tail(&reloc_root->root_list, &reloc_roots);
 			goto out_free;
 		}
 
@@ -4608,7 +4668,7 @@ int btrfs_reloc_clone_csums(struct inode *inode, u64 file_pos, u64 len)
 	LIST_HEAD(list);
 
 	ordered = btrfs_lookup_ordered_extent(inode, file_pos);
-	BUG_ON(ordered->file_offset != file_pos || ordered->len != len);
+	BUG_ON(ordered->file_offset != file_pos || ordered->num_bytes != len);
 
 	disk_bytenr = file_pos + BTRFS_I(inode)->index_cnt;
 	ret = btrfs_lookup_csums_range(fs_info->csum_root, disk_bytenr,
@@ -4632,7 +4692,7 @@ int btrfs_reloc_clone_csums(struct inode *inode, u64 file_pos, u64 len)
 		 * disk_len vs real len like with real inodes since it's all
 		 * disk length.
 		 */
-		new_bytenr = ordered->start + (sums->bytenr - disk_bytenr);
+		new_bytenr = ordered->disk_bytenr + sums->bytenr - disk_bytenr;
 		sums->bytenr = new_bytenr;
 
 		btrfs_add_ordered_sum(ordered, sums);
@@ -4679,7 +4739,7 @@ int btrfs_reloc_cow_block(struct btrfs_trans_handle *trans,
 		       node->new_bytenr != buf->start);
 
 		drop_node_buffer(node);
-		extent_buffer_get(cow);
+		atomic_inc(&cow->refs);
 		node->eb = cow;
 		node->new_bytenr = cow->start;
 
@@ -4711,7 +4771,7 @@ void btrfs_reloc_pre_snapshot(struct btrfs_pending_snapshot *pending,
 	struct btrfs_root *root = pending->root;
 	struct reloc_control *rc = root->fs_info->reloc_ctl;
 
-	if (!root->reloc_root || !rc)
+	if (!rc || !have_reloc_root(root))
 		return;
 
 	if (!rc->merge_reloc_tree)
@@ -4745,7 +4805,7 @@ int btrfs_reloc_post_snapshot(struct btrfs_trans_handle *trans,
 	struct reloc_control *rc = root->fs_info->reloc_ctl;
 	int ret;
 
-	if (!root->reloc_root || !rc)
+	if (!rc || !have_reloc_root(root))
 		return 0;
 
 	rc = root->fs_info->reloc_ctl;