Re: init is pid 1 vs. initrd

H. Peter Anvin (hpa@transmeta.com)
Fri, 17 Apr 1998 18:55:00 -0700 (PDT)


> H. Peter Anvin <hpa@transmeta.com> wrote:
> > But you can't do that if the process you're running in is holding the
> > old root open. This is why I propose telling the kernel where the
> > pivot is, and then exiting (no processes running.)
>
> Why would it be holding the old root open?

Because it is the root directory. If you don't have a handle to the
old root, then you have already lost

> Just to re-iterate, I'm envisioning:
>
> ...
> mount new root
> cd /
> exec new init
> ...
> free old root
> ...
>
> True, you can't free up the ram disk if you have some prehistoric
> process hanging around. That's a feature, not a bug.
>
> I guess I am expecting to be able to do a little surgery to pull
> any references to the old root from any running processes. Basically,
> I'm envisioning that each such reference would be replaced with
> a reference to the new root, but I don't see why this is dangerous.
>

I just don't understand why you insist on doing it the hard way. All
of this is more complexity and doesn't add any generality, so why
bother?

-hpa

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu