btrfs: use dynamic allocation for root item in create_subvol
authorDavid Sterba <dsterba@suse.com>
Thu, 24 Mar 2016 16:49:22 +0000 (17:49 +0100)
committerDavid Sterba <dsterba@suse.com>
Fri, 6 May 2016 13:22:49 +0000 (15:22 +0200)
The size of root item is more than 400 bytes, which is quite a lot of
stack space. As we do IO from inside the subvolume ioctls, we should
keep the stack usage low in case the filesystem is on top of other
layers (NFS, device mapper, iscsi, etc).

Reviewed-by: Tsutomu Itoh <t-itoh@jp.fujitsu.com>
Signed-off-by: David Sterba <dsterba@suse.com>
fs/btrfs/ioctl.c

index cfa7d47b1ba048f3c0d9d24b536096e024d91f76..798f58e7338e8222ec6c99646db497fec48d6e3e 100644 (file)
@@ -439,7 +439,7 @@ static noinline int create_subvol(struct inode *dir,
 {
        struct btrfs_trans_handle *trans;
        struct btrfs_key key;
-       struct btrfs_root_item root_item;
+       struct btrfs_root_item *root_item;
        struct btrfs_inode_item *inode_item;
        struct extent_buffer *leaf;
        struct btrfs_root *root = BTRFS_I(dir)->root;
@@ -455,16 +455,22 @@ static noinline int create_subvol(struct inode *dir,
        u64 qgroup_reserved;
        uuid_le new_uuid;
 
+       root_item = kzalloc(sizeof(*root_item), GFP_KERNEL);
+       if (!root_item)
+               return -ENOMEM;
+
        ret = btrfs_find_free_objectid(root->fs_info->tree_root, &objectid);
        if (ret)
-               return ret;
+               goto fail_free;
 
        /*
         * Don't create subvolume whose level is not zero. Or qgroup will be
         * screwed up since it assume subvolme qgroup's level to be 0.
         */
-       if (btrfs_qgroup_level(objectid))
-               return -ENOSPC;
+       if (btrfs_qgroup_level(objectid)) {
+               ret = -ENOSPC;
+               goto fail_free;
+       }
 
        btrfs_init_block_rsv(&block_rsv, BTRFS_BLOCK_RSV_TEMP);
        /*
@@ -474,14 +480,14 @@ static noinline int create_subvol(struct inode *dir,
        ret = btrfs_subvolume_reserve_metadata(root, &block_rsv,
                                               8, &qgroup_reserved, false);
        if (ret)
-               return ret;
+               goto fail_free;
 
        trans = btrfs_start_transaction(root, 0);
        if (IS_ERR(trans)) {
                ret = PTR_ERR(trans);
                btrfs_subvolume_release_metadata(root, &block_rsv,
                                                 qgroup_reserved);
-               return ret;
+               goto fail_free;
        }
        trans->block_rsv = &block_rsv;
        trans->bytes_reserved = block_rsv.size;
@@ -509,47 +515,45 @@ static noinline int create_subvol(struct inode *dir,
                            BTRFS_UUID_SIZE);
        btrfs_mark_buffer_dirty(leaf);
 
-       memset(&root_item, 0, sizeof(root_item));
-
-       inode_item = &root_item.inode;
+       inode_item = &root_item->inode;
        btrfs_set_stack_inode_generation(inode_item, 1);
        btrfs_set_stack_inode_size(inode_item, 3);
        btrfs_set_stack_inode_nlink(inode_item, 1);
        btrfs_set_stack_inode_nbytes(inode_item, root->nodesize);
        btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755);
 
-       btrfs_set_root_flags(&root_item, 0);
-       btrfs_set_root_limit(&root_item, 0);
+       btrfs_set_root_flags(root_item, 0);
+       btrfs_set_root_limit(root_item, 0);
        btrfs_set_stack_inode_flags(inode_item, BTRFS_INODE_ROOT_ITEM_INIT);
 
-       btrfs_set_root_bytenr(&root_item, leaf->start);
-       btrfs_set_root_generation(&root_item, trans->transid);
-       btrfs_set_root_level(&root_item, 0);
-       btrfs_set_root_refs(&root_item, 1);
-       btrfs_set_root_used(&root_item, leaf->len);
-       btrfs_set_root_last_snapshot(&root_item, 0);
+       btrfs_set_root_bytenr(root_item, leaf->start);
+       btrfs_set_root_generation(root_item, trans->transid);
+       btrfs_set_root_level(root_item, 0);
+       btrfs_set_root_refs(root_item, 1);
+       btrfs_set_root_used(root_item, leaf->len);
+       btrfs_set_root_last_snapshot(root_item, 0);
 
-       btrfs_set_root_generation_v2(&root_item,
-                       btrfs_root_generation(&root_item));
+       btrfs_set_root_generation_v2(root_item,
+                       btrfs_root_generation(root_item));
        uuid_le_gen(&new_uuid);
-       memcpy(root_item.uuid, new_uuid.b, BTRFS_UUID_SIZE);
-       btrfs_set_stack_timespec_sec(&root_item.otime, cur_time.tv_sec);
-       btrfs_set_stack_timespec_nsec(&root_item.otime, cur_time.tv_nsec);
-       root_item.ctime = root_item.otime;
-       btrfs_set_root_ctransid(&root_item, trans->transid);
-       btrfs_set_root_otransid(&root_item, trans->transid);
+       memcpy(root_item->uuid, new_uuid.b, BTRFS_UUID_SIZE);
+       btrfs_set_stack_timespec_sec(&root_item->otime, cur_time.tv_sec);
+       btrfs_set_stack_timespec_nsec(&root_item->otime, cur_time.tv_nsec);
+       root_item->ctime = root_item->otime;
+       btrfs_set_root_ctransid(root_item, trans->transid);
+       btrfs_set_root_otransid(root_item, trans->transid);
 
        btrfs_tree_unlock(leaf);
        free_extent_buffer(leaf);
        leaf = NULL;
 
-       btrfs_set_root_dirid(&root_item, new_dirid);
+       btrfs_set_root_dirid(root_item, new_dirid);
 
        key.objectid = objectid;
        key.offset = 0;
        key.type = BTRFS_ROOT_ITEM_KEY;
        ret = btrfs_insert_root(trans, root->fs_info->tree_root, &key,
-                               &root_item);
+                               root_item);
        if (ret)
                goto fail;
 
@@ -601,12 +605,13 @@ static noinline int create_subvol(struct inode *dir,
        BUG_ON(ret);
 
        ret = btrfs_uuid_tree_add(trans, root->fs_info->uuid_root,
-                                 root_item.uuid, BTRFS_UUID_KEY_SUBVOL,
+                                 root_item->uuid, BTRFS_UUID_KEY_SUBVOL,
                                  objectid);
        if (ret)
                btrfs_abort_transaction(trans, root, ret);
 
 fail:
+       kfree(root_item);
        trans->block_rsv = NULL;
        trans->bytes_reserved = 0;
        btrfs_subvolume_release_metadata(root, &block_rsv, qgroup_reserved);
@@ -629,6 +634,10 @@ fail:
                d_instantiate(dentry, inode);
        }
        return ret;
+
+fail_free:
+       kfree(root_item);
+       return ret;
 }
 
 static void btrfs_wait_for_no_snapshoting_writes(struct btrfs_root *root)