1 files changed, 127 insertions, 81 deletions

diff --git a/fs/btrfs/relocation.c b/fs/btrfs/relocation.c
index c6d4bb8cbe29..f5d9e5f74a52 100644
--- a/fs/btrfs/relocation.c
+++ b/fs/btrfs/relocation.c
@@ -111,8 +111,8 @@ struct tree_block {
 	}; /* Use rb_simple_node for search/insert */
 	u64 owner;
 	struct btrfs_key key;
-	unsigned int level:8;
-	unsigned int key_ready:1;
+	u8 level;
+	bool key_ready;
 };
 
 #define MAX_EXTENTS 128
@@ -122,6 +122,13 @@ struct file_extent_cluster {
 	u64 end;
 	u64 boundary[MAX_EXTENTS];
 	unsigned int nr;
+	u64 owning_root;
+};
+
+/* Stages of data relocation. */
+enum reloc_stage {
+	MOVE_DATA_EXTENTS,
+	UPDATE_DATA_PTRS
 };
 
 struct reloc_control {
@@ -155,16 +162,12 @@ struct reloc_control {
 	u64 search_start;
 	u64 extents_found;
 
-	unsigned int stage:8;
-	unsigned int create_reloc_tree:1;
-	unsigned int merge_reloc_tree:1;
-	unsigned int found_file_extent:1;
+	enum reloc_stage stage;
+	bool create_reloc_tree;
+	bool merge_reloc_tree;
+	bool found_file_extent;
 };
 
-/* stages of data relocation */
-#define MOVE_DATA_EXTENTS	0
-#define UPDATE_DATA_PTRS	1
-
 static void mark_block_processed(struct reloc_control *rc,
 				 struct btrfs_backref_node *node)
 {
@@ -180,13 +183,6 @@ static void mark_block_processed(struct reloc_control *rc,
 	node->processed = 1;
 }
 
-
-static void mapping_tree_init(struct mapping_tree *tree)
-{
-	tree->rb_root = RB_ROOT;
-	spin_lock_init(&tree->lock);
-}
-
 /*
  * walk up backref nodes until reach node presents tree root
  */
@@ -299,7 +295,7 @@ static int update_backref_cache(struct btrfs_trans_handle *trans,
 	return 1;
 }
 
-static bool reloc_root_is_dead(struct btrfs_root *root)
+static bool reloc_root_is_dead(const struct btrfs_root *root)
 {
 	/*
 	 * Pair with set_bit/clear_bit in clean_dirty_subvols and
@@ -320,7 +316,7 @@ static bool reloc_root_is_dead(struct btrfs_root *root)
  * from no reloc root.  But btrfs_should_ignore_reloc_root() below is a
  * special case.
  */
-static bool have_reloc_root(struct btrfs_root *root)
+static bool have_reloc_root(const struct btrfs_root *root)
 {
 	if (reloc_root_is_dead(root))
 		return false;
@@ -329,31 +325,30 @@ static bool have_reloc_root(struct btrfs_root *root)
 	return true;
 }
 
-int btrfs_should_ignore_reloc_root(struct btrfs_root *root)
+bool btrfs_should_ignore_reloc_root(const struct btrfs_root *root)
 {
 	struct btrfs_root *reloc_root;
 
 	if (!test_bit(BTRFS_ROOT_SHAREABLE, &root->state))
-		return 0;
+		return false;
 
 	/* This root has been merged with its reloc tree, we can ignore it */
 	if (reloc_root_is_dead(root))
-		return 1;
+		return true;
 
 	reloc_root = root->reloc_root;
 	if (!reloc_root)
-		return 0;
+		return false;
 
 	if (btrfs_header_generation(reloc_root->commit_root) ==
 	    root->fs_info->running_transaction->transid)
-		return 0;
+		return false;
 	/*
-	 * if there is reloc tree and it was created in previous
-	 * transaction backref lookup can find the reloc tree,
-	 * so backref node for the fs tree root is useless for
-	 * relocation.
+	 * If there is reloc tree and it was created in previous transaction
+	 * backref lookup can find the reloc tree, so backref node for the fs
+	 * tree root is useless for relocation.
 	 */
-	return 1;
+	return true;
 }
 
 /*
@@ -547,7 +542,7 @@ out:
  */
 static int clone_backref_node(struct btrfs_trans_handle *trans,
 			      struct reloc_control *rc,
-			      struct btrfs_root *src,
+			      const struct btrfs_root *src,
 			      struct btrfs_root *dest)
 {
 	struct btrfs_root *reloc_root = src->reloc_root;
@@ -632,7 +627,7 @@ fail:
 /*
  * helper to add 'address of tree root -> reloc tree' mapping
  */
-static int __must_check __add_reloc_root(struct btrfs_root *root)
+static int __add_reloc_root(struct btrfs_root *root)
 {
 	struct btrfs_fs_info *fs_info = root->fs_info;
 	struct rb_node *rb_node;
@@ -1159,7 +1154,7 @@ int replace_file_extents(struct btrfs_trans_handle *trans,
 
 		key.offset -= btrfs_file_extent_offset(leaf, fi);
 		btrfs_init_generic_ref(&ref, BTRFS_ADD_DELAYED_REF, new_bytenr,
-				       num_bytes, parent);
+				       num_bytes, parent, root->root_key.objectid);
 		btrfs_init_data_ref(&ref, btrfs_header_owner(leaf),
 				    key.objectid, key.offset,
 				    root->root_key.objectid, false);
@@ -1170,7 +1165,7 @@ int replace_file_extents(struct btrfs_trans_handle *trans,
 		}
 
 		btrfs_init_generic_ref(&ref, BTRFS_DROP_DELAYED_REF, bytenr,
-				       num_bytes, parent);
+				       num_bytes, parent, root->root_key.objectid);
 		btrfs_init_data_ref(&ref, btrfs_header_owner(leaf),
 				    key.objectid, key.offset,
 				    root->root_key.objectid, false);
@@ -1181,15 +1176,15 @@ int replace_file_extents(struct btrfs_trans_handle *trans,
 		}
 	}
 	if (dirty)
-		btrfs_mark_buffer_dirty(leaf);
+		btrfs_mark_buffer_dirty(trans, leaf);
 	if (inode)
 		btrfs_add_delayed_iput(BTRFS_I(inode));
 	return ret;
 }
 
-static noinline_for_stack
-int memcmp_node_keys(struct extent_buffer *eb, int slot,
-		     struct btrfs_path *path, int level)
+static noinline_for_stack int memcmp_node_keys(const struct extent_buffer *eb,
+					       int slot, const struct btrfs_path *path,
+					       int level)
 {
 	struct btrfs_disk_key key1;
 	struct btrfs_disk_key key2;
@@ -1374,16 +1369,17 @@ again:
 		 */
 		btrfs_set_node_blockptr(parent, slot, new_bytenr);
 		btrfs_set_node_ptr_generation(parent, slot, new_ptr_gen);
-		btrfs_mark_buffer_dirty(parent);
+		btrfs_mark_buffer_dirty(trans, parent);
 
 		btrfs_set_node_blockptr(path->nodes[level],
 					path->slots[level], old_bytenr);
 		btrfs_set_node_ptr_generation(path->nodes[level],
 					      path->slots[level], old_ptr_gen);
-		btrfs_mark_buffer_dirty(path->nodes[level]);
+		btrfs_mark_buffer_dirty(trans, path->nodes[level]);
 
 		btrfs_init_generic_ref(&ref, BTRFS_ADD_DELAYED_REF, old_bytenr,
-				       blocksize, path->nodes[level]->start);
+				       blocksize, path->nodes[level]->start,
+				       src->root_key.objectid);
 		btrfs_init_tree_ref(&ref, level - 1, src->root_key.objectid,
 				    0, true);
 		ret = btrfs_inc_extent_ref(trans, &ref);
@@ -1392,7 +1388,7 @@ again:
 			break;
 		}
 		btrfs_init_generic_ref(&ref, BTRFS_ADD_DELAYED_REF, new_bytenr,
-				       blocksize, 0);
+				       blocksize, 0, dest->root_key.objectid);
 		btrfs_init_tree_ref(&ref, level - 1, dest->root_key.objectid, 0,
 				    true);
 		ret = btrfs_inc_extent_ref(trans, &ref);
@@ -1401,8 +1397,9 @@ again:
 			break;
 		}
 
+		/* We don't know the real owning_root, use 0. */
 		btrfs_init_generic_ref(&ref, BTRFS_DROP_DELAYED_REF, new_bytenr,
-				       blocksize, path->nodes[level]->start);
+				       blocksize, path->nodes[level]->start, 0);
 		btrfs_init_tree_ref(&ref, level - 1, src->root_key.objectid,
 				    0, true);
 		ret = btrfs_free_extent(trans, &ref);
@@ -1411,8 +1408,9 @@ again:
 			break;
 		}
 
+		/* We don't know the real owning_root, use 0. */
 		btrfs_init_generic_ref(&ref, BTRFS_DROP_DELAYED_REF, old_bytenr,
-				       blocksize, 0);
+				       blocksize, 0, 0);
 		btrfs_init_tree_ref(&ref, level - 1, dest->root_key.objectid,
 				    0, true);
 		ret = btrfs_free_extent(trans, &ref);
@@ -1518,8 +1516,8 @@ int walk_down_reloc_tree(struct btrfs_root *root, struct btrfs_path *path,
  * [min_key, max_key)
  */
 static int invalidate_extent_cache(struct btrfs_root *root,
-				   struct btrfs_key *min_key,
-				   struct btrfs_key *max_key)
+				   const struct btrfs_key *min_key,
+				   const struct btrfs_key *max_key)
 {
 	struct btrfs_fs_info *fs_info = root->fs_info;
 	struct inode *inode = NULL;
@@ -1897,7 +1895,7 @@ again:
 		}
 	}
 
-	rc->merge_reloc_tree = 1;
+	rc->merge_reloc_tree = true;
 
 	while (!list_empty(&rc->reloc_roots)) {
 		reloc_root = list_entry(rc->reloc_roots.next,
@@ -2517,11 +2515,12 @@ static int do_relocation(struct btrfs_trans_handle *trans,
 						node->eb->start);
 			btrfs_set_node_ptr_generation(upper->eb, slot,
 						      trans->transid);
-			btrfs_mark_buffer_dirty(upper->eb);
+			btrfs_mark_buffer_dirty(trans, upper->eb);
 
 			btrfs_init_generic_ref(&ref, BTRFS_ADD_DELAYED_REF,
 					       node->eb->start, blocksize,
-					       upper->eb->start);
+					       upper->eb->start,
+					       btrfs_header_owner(upper->eb));
 			btrfs_init_tree_ref(&ref, node->level,
 					    btrfs_header_owner(upper->eb),
 					    root->root_key.objectid, false);
@@ -2633,7 +2632,7 @@ static int tree_block_processed(u64 bytenr, struct reloc_control *rc)
 	u32 blocksize = rc->extent_root->fs_info->nodesize;
 
 	if (test_range_bit(&rc->processed_blocks, bytenr,
-			   bytenr + blocksize - 1, EXTENT_DIRTY, 1, NULL))
+			   bytenr + blocksize - 1, EXTENT_DIRTY, NULL))
 		return 1;
 	return 0;
 }
@@ -2660,7 +2659,7 @@ static int get_tree_block_key(struct btrfs_fs_info *fs_info,
 	else
 		btrfs_node_key_to_cpu(eb, &block->key, 0);
 	free_extent_buffer(eb);
-	block->key_ready = 1;
+	block->key_ready = true;
 	return 0;
 }
 
@@ -2830,7 +2829,7 @@ out_free_blocks:
 
 static noinline_for_stack int prealloc_file_extent_cluster(
 				struct btrfs_inode *inode,
-				struct file_extent_cluster *cluster)
+				const struct file_extent_cluster *cluster)
 {
 	u64 alloc_hint = 0;
 	u64 start;
@@ -2965,7 +2964,7 @@ static noinline_for_stack int setup_relocation_extent_mapping(struct inode *inod
 /*
  * Allow error injection to test balance/relocation cancellation
  */
-noinline int btrfs_should_cancel_balance(struct btrfs_fs_info *fs_info)
+noinline int btrfs_should_cancel_balance(const struct btrfs_fs_info *fs_info)
 {
 	return atomic_read(&fs_info->balance_cancel_req) ||
 		atomic_read(&fs_info->reloc_cancel_req) ||
@@ -2973,7 +2972,7 @@ noinline int btrfs_should_cancel_balance(struct btrfs_fs_info *fs_info)
 }
 ALLOW_ERROR_INJECTION(btrfs_should_cancel_balance, TRUE);
 
-static u64 get_cluster_boundary_end(struct file_extent_cluster *cluster,
+static u64 get_cluster_boundary_end(const struct file_extent_cluster *cluster,
 				    int cluster_nr)
 {
 	/* Last extent, use cluster end directly */
@@ -2985,7 +2984,7 @@ static u64 get_cluster_boundary_end(struct file_extent_cluster *cluster,
 }
 
 static int relocate_one_page(struct inode *inode, struct file_ra_state *ra,
-			     struct file_extent_cluster *cluster,
+			     const struct file_extent_cluster *cluster,
 			     int *cluster_nr, unsigned long page_index)
 {
 	struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
@@ -3120,7 +3119,7 @@ release_page:
 }
 
 static int relocate_file_extent_cluster(struct inode *inode,
-					struct file_extent_cluster *cluster)
+					const struct file_extent_cluster *cluster)
 {
 	u64 offset = BTRFS_I(inode)->index_cnt;
 	unsigned long index;
@@ -3158,11 +3157,12 @@ out:
 	return ret;
 }
 
-static noinline_for_stack
-int relocate_data_extent(struct inode *inode, struct btrfs_key *extent_key,
-			 struct file_extent_cluster *cluster)
+static noinline_for_stack int relocate_data_extent(struct inode *inode,
+				const struct btrfs_key *extent_key,
+				struct file_extent_cluster *cluster)
 {
 	int ret;
+	struct btrfs_root *root = BTRFS_I(inode)->root;
 
 	if (cluster->nr > 0 && extent_key->objectid != cluster->end + 1) {
 		ret = relocate_file_extent_cluster(inode, cluster);
@@ -3171,8 +3171,38 @@ int relocate_data_extent(struct inode *inode, struct btrfs_key *extent_key,
 		cluster->nr = 0;
 	}
 
-	if (!cluster->nr)
+	/*
+	 * Under simple quotas, we set root->relocation_src_root when we find
+	 * the extent. If adjacent extents have different owners, we can't merge
+	 * them while relocating. Handle this by storing the owning root that
+	 * started a cluster and if we see an extent from a different root break
+	 * cluster formation (just like the above case of non-adjacent extents).
+	 *
+	 * Without simple quotas, relocation_src_root is always 0, so we should
+	 * never see a mismatch, and it should have no effect on relocation
+	 * clusters.
+	 */
+	if (cluster->nr > 0 && cluster->owning_root != root->relocation_src_root) {
+		u64 tmp = root->relocation_src_root;
+
+		/*
+		 * root->relocation_src_root is the state that actually affects
+		 * the preallocation we do here, so set it to the root owning
+		 * the cluster we need to relocate.
+		 */
+		root->relocation_src_root = cluster->owning_root;
+		ret = relocate_file_extent_cluster(inode, cluster);
+		if (ret)
+			return ret;
+		cluster->nr = 0;
+		/* And reset it back for the current extent's owning root. */
+		root->relocation_src_root = tmp;
+	}
+
+	if (!cluster->nr) {
 		cluster->start = extent_key->objectid;
+		cluster->owning_root = root->relocation_src_root;
+	}
 	else
 		BUG_ON(cluster->nr >= MAX_EXTENTS);
 	cluster->end = extent_key->objectid + extent_key->offset - 1;
@@ -3193,7 +3223,7 @@ int relocate_data_extent(struct inode *inode, struct btrfs_key *extent_key,
  * the major work is getting the generation and level of the block
  */
 static int add_tree_block(struct reloc_control *rc,
-			  struct btrfs_key *extent_key,
+			  const struct btrfs_key *extent_key,
 			  struct btrfs_path *path,
 			  struct rb_root *blocks)
 {
@@ -3278,7 +3308,7 @@ static int add_tree_block(struct reloc_control *rc,
 	block->key.objectid = rc->extent_root->fs_info->nodesize;
 	block->key.offset = generation;
 	block->level = level;
-	block->key_ready = 0;
+	block->key_ready = false;
 	block->owner = owner;
 
 	rb_node = rb_simple_insert(blocks, block->bytenr, &block->rb_node);
@@ -3444,11 +3474,10 @@ static int delete_v1_space_cache(struct extent_buffer *leaf,
 /*
  * helper to find all tree blocks that reference a given data extent
  */
-static noinline_for_stack
-int add_data_references(struct reloc_control *rc,
-			struct btrfs_key *extent_key,
-			struct btrfs_path *path,
-			struct rb_root *blocks)
+static noinline_for_stack int add_data_references(struct reloc_control *rc,
+						  const struct btrfs_key *extent_key,
+						  struct btrfs_path *path,
+						  struct rb_root *blocks)
 {
 	struct btrfs_backref_walk_ctx ctx = { 0 };
 	struct ulist_iterator leaf_uiter;
@@ -3622,7 +3651,7 @@ int prepare_to_relocate(struct reloc_control *rc)
 	if (ret)
 		return ret;
 
-	rc->create_reloc_tree = 1;
+	rc->create_reloc_tree = true;
 	set_reloc_control(rc);
 
 	trans = btrfs_join_transaction(rc->extent_root);
@@ -3702,6 +3731,21 @@ restart:
 				    struct btrfs_extent_item);
 		flags = btrfs_extent_flags(path->nodes[0], ei);
 
+		/*
+		 * If we are relocating a simple quota owned extent item, we
+		 * need to note the owner on the reloc data root so that when
+		 * we allocate the replacement item, we can attribute it to the
+		 * correct eventual owner (rather than the reloc data root).
+		 */
+		if (btrfs_qgroup_mode(fs_info) == BTRFS_QGROUP_MODE_SIMPLE) {
+			struct btrfs_root *root = BTRFS_I(rc->data_inode)->root;
+			u64 owning_root_id = btrfs_get_extent_owner_root(fs_info,
+								 path->nodes[0],
+								 path->slots[0]);
+
+			root->relocation_src_root = owning_root_id;
+		}
+
 		if (flags & BTRFS_EXTENT_FLAG_TREE_BLOCK) {
 			ret = add_tree_block(rc, &key, path, &blocks);
 		} else if (rc->stage == UPDATE_DATA_PTRS &&
@@ -3734,7 +3778,7 @@ restart:
 
 		if (rc->stage == MOVE_DATA_EXTENTS &&
 		    (flags & BTRFS_EXTENT_FLAG_DATA)) {
-			rc->found_file_extent = 1;
+			rc->found_file_extent = true;
 			ret = relocate_data_extent(rc->data_inode,
 						   &key, &rc->cluster);
 			if (ret < 0) {
@@ -3771,7 +3815,7 @@ restart:
 			err = ret;
 	}
 
-	rc->create_reloc_tree = 0;
+	rc->create_reloc_tree = false;
 	set_reloc_control(rc);
 
 	btrfs_backref_release_cache(&rc->backref_cache);
@@ -3789,7 +3833,7 @@ restart:
 
 	merge_reloc_roots(rc);
 
-	rc->merge_reloc_tree = 0;
+	rc->merge_reloc_tree = false;
 	unset_reloc_control(rc);
 	btrfs_block_rsv_release(fs_info, rc->block_rsv, (u64)-1, NULL);
 
@@ -3835,7 +3879,7 @@ static int __insert_orphan_inode(struct btrfs_trans_handle *trans,
 	btrfs_set_inode_mode(leaf, item, S_IFREG | 0600);
 	btrfs_set_inode_flags(leaf, item, BTRFS_INODE_NOCOMPRESS |
 					  BTRFS_INODE_PREALLOC);
-	btrfs_mark_buffer_dirty(leaf);
+	btrfs_mark_buffer_dirty(trans, leaf);
 out:
 	btrfs_free_path(path);
 	return ret;
@@ -3874,9 +3918,9 @@ out:
  * helper to create inode for data relocation.
  * the inode is in data relocation tree and its link count is 0
  */
-static noinline_for_stack
-struct inode *create_reloc_inode(struct btrfs_fs_info *fs_info,
-				 struct btrfs_block_group *group)
+static noinline_for_stack struct inode *create_reloc_inode(
+					struct btrfs_fs_info *fs_info,
+					const struct btrfs_block_group *group)
 {
 	struct inode *inode = NULL;
 	struct btrfs_trans_handle *trans;
@@ -3971,8 +4015,9 @@ static struct reloc_control *alloc_reloc_control(struct btrfs_fs_info *fs_info)
 
 	INIT_LIST_HEAD(&rc->reloc_roots);
 	INIT_LIST_HEAD(&rc->dirty_subvol_roots);
-	btrfs_backref_init_cache(fs_info, &rc->backref_cache, 1);
-	mapping_tree_init(&rc->reloc_root_tree);
+	btrfs_backref_init_cache(fs_info, &rc->backref_cache, true);
+	rc->reloc_root_tree.rb_root = RB_ROOT;
+	spin_lock_init(&rc->reloc_root_tree.lock);
 	extent_io_tree_init(fs_info, &rc->processed_blocks, IO_TREE_RELOC_BLOCKS);
 	return rc;
 }
@@ -4004,7 +4049,7 @@ static void describe_relocation(struct btrfs_fs_info *fs_info,
 		   block_group->start, buf);
 }
 
-static const char *stage_to_string(int stage)
+static const char *stage_to_string(enum reloc_stage stage)
 {
 	if (stage == MOVE_DATA_EXTENTS)
 		return "move data extents";
@@ -4120,7 +4165,7 @@ int btrfs_relocate_block_group(struct btrfs_fs_info *fs_info, u64 group_start)
 	WARN_ON(ret && ret != -EAGAIN);
 
 	while (1) {
-		int finishes_stage;
+		enum reloc_stage finishes_stage;
 
 		mutex_lock(&fs_info->cleaner_mutex);
 		ret = relocate_block_group(rc);
@@ -4303,7 +4348,7 @@ int btrfs_recover_relocation(struct btrfs_fs_info *fs_info)
 		goto out_unset;
 	}
 
-	rc->merge_reloc_tree = 1;
+	rc->merge_reloc_tree = true;
 
 	while (!list_empty(&reloc_roots)) {
 		reloc_root = list_entry(reloc_roots.next,
@@ -4422,7 +4467,8 @@ int btrfs_reloc_clone_csums(struct btrfs_ordered_extent *ordered)
 }
 
 int btrfs_reloc_cow_block(struct btrfs_trans_handle *trans,
-			  struct btrfs_root *root, struct extent_buffer *buf,
+			  struct btrfs_root *root,
+			  const struct extent_buffer *buf,
 			  struct extent_buffer *cow)
 {
 	struct btrfs_fs_info *fs_info = root->fs_info;
@@ -4561,7 +4607,7 @@ int btrfs_reloc_post_snapshot(struct btrfs_trans_handle *trans,
  *
  * Return U64_MAX if no running relocation.
  */
-u64 btrfs_get_reloc_bg_bytenr(struct btrfs_fs_info *fs_info)
+u64 btrfs_get_reloc_bg_bytenr(const struct btrfs_fs_info *fs_info)
 {
 	u64 logical = U64_MAX;