Re: [PATCH 1/2] fuse: fix COPY_FILE_RANGE interface

[Chunsheng Luo <luochunsheng@ustc.edu>]

On Thu, Aug 07 2025, Chunsheng Luo wrote:

> On Tue, Aug 05 2025, Miklos Szeredi wrote:
>
> +	bytes_copied = fc->no_copy_file_range_64 ?
> +		outarg.size : outarg_64.bytes_copied;
> +
>  	truncate_inode_pages_range(inode_out->i_mapping,
>  				   ALIGN_DOWN(pos_out, PAGE_SIZE),
> -				   ALIGN(pos_out + outarg.size, PAGE_SIZE) - 1);
> +				   ALIGN(pos_out + bytes_copied, PAGE_SIZE) - 1);
>  
>  	file_update_time(file_out);
> -	fuse_write_update_attr(inode_out, pos_out + outarg.size, outarg.size);
> +	fuse_write_update_attr(inode_out, pos_out + bytes_copied, bytes_copied);

The copy_file_range syscall returns bytes_copied, a value provided by the userspace filesystem that the kernel cannot control. If bytes_copied > len, how should the application handle this? Similarly, if pos_out + bytes_copied < pos_outdue to integer overflow, could this cause any issues? Since vfs_copy_file_range->generic_copy_file_checks already check that pos_out + len does not overflow, so just need check bytes_copied > len.

Thanks
Chunsheng Luo