Re: [PATCH v3 0/1] MR498: server: Allow creating named pipes using \Device\NamedPipe\ as RootDirectory. (#52105)

[Zebediah Figura (＠zfigura)]

Zebediah Figura (@zfigura) commented about server/named_pipe.c:
>  if (!name) return NULL;  /* open the device itself */
>
> - if (name->str && !name->len)
> - {
> -
>    /* root directory: last component is empty ("\\Device\\NamedPipe\\") */
> -
>    struct named_pipe_device_file *dir = alloc_object( &named_pipe_device_file_ops );
> -
> -
>    if (!dir) return NULL;
> -
> -
>    dir->is_rootdir = 1;
> -
>    dir->fd = NULL;  /* defer alloc_pseudo_fd() until after we have options */
> -
>    dir->device = (struct named_pipe_device *)grab_object( obj );
> -
> -
>    return &dir->obj;

This on the other hand bothers me. Maybe my intuition as to what's sensible design is off, but creating an object from lookup_name() feels wrong, especially given that we have to defer creating the fd.

I guess the right solution here is to combine lookup_name() and open_file(), presumably using OBJ_OPENIF to distinguish the two.

Or possibly the right solution is to create (ahead of time) a separate named pipe "device" that represents the root directory, but I'm not thrilled about that idea; it feels less architecturally clear to me.