[PATCH 2/5] server: Add delayed processing for socket-specific ioctl() (resend).

[Alexandre Julliard]

"Erich E. Hoover" <erich.e.hoover at gmail.com> writes:

> @@ -523,17 +523,32 @@ obj_handle_t sock_ioctl( struct fd *fd, ioctl_code_t code, const async_data_t *a
>                           int blocking, const void *data, data_size_t size )
>  {
>      struct sock *sock = get_fd_user( fd );
> +    obj_handle_t wait_handle = 0;
> +    async_data_t new_data;
> +    int error;
>  
>      assert( sock->obj.ops == &sock_ops );
>  
> +    if (blocking)
> +    {
> +        if (!(wait_handle = alloc_wait_event( current->process ))) return 0;
> +        new_data = *async_data;
> +        new_data.event = wait_handle;
> +        async_data = &new_data;
> +    }
>      switch(code)
>      {
>      case WS_SIO_ADDRESS_LIST_CHANGE:
>          /* intentional fallthrough, not yet supported */
>      default:
> -        set_error( STATUS_NOT_SUPPORTED );
> -        return 0;
> +        error = STATUS_NOT_SUPPORTED;
> +        break;
>      }
> +    set_error( error );
> +    if (error == STATUS_PENDING)
> +        return wait_handle;
> +    close_handle( current->process, wait_handle );
> +    return 0;

You should be doing that only as needed. It doesn't make sense to
allocate a handle for unknown ioctls.

-- 
Alexandre Julliard
julliard at winehq.org