Re: virtualized nfsd (Re: virtual kernels, syscall routing, etc.)

[Matthias Scheler <tron%zhadum.org.uk@localhost>]

On Fri, Oct 16, 2009 at 05:08:54AM -0400, der Mouse wrote:
> But...why?  I had an NFS daemon that ran in userland as an ordinary
> userland process, no "unholy cocktails" or "virtual kernels" or even
> hacked-up librpcsvc needed, decades ago. I can't have been the only
> one.

No, you weren't.

Linux used an userland NFS server for a long time, too. And AFAIK
someone ported that to Solaris to allow providing NFS servers from
Solaris containers (which didn't work initially).

> Is this just an exercise in recreating a kernel code environment
> in userspace, or is there something else I'm missing?

It is nice for debugging and improving the in-kernel NFS code:
1.) You can easily update code without reboots (kernel modules should
    be able to do this as well).
2.) If the NFS server in user land crashes you're machine doesn't
    crash as well.
This should allow rapid development.

In a production environment where speed matters you can run the code
previously developed in userland in the kernel.

        Kind regards

-- 
Matthias Scheler                                  http://zhadum.org.uk/