kernfs (was: Re: CVS commit: src/distrib/utils/sysinst)

["Christoph Egger" <Christoph_Egger%gmx.de@localhost>]

> Tom Spindler wrote:
> >>    > >- kernfs is hosed at the moment and nothing appears to rely on it
> any
> >>    > >  more so disable it. (IPsec did at one point, IIRC).
> >>    > 
> >>    > Huh? And what about /kern/xen? Did I miss something lately?
> >>    
> >>    Should the /kern/xen stuff be moved into /proc, then? (/kern/xen
> ain't
> >>    mentioned in mount_kernfs(8)....)
> >>
> >> exactly how is the /kern/xen stuff related to processes?
> > 
> > An excellent question. I have no idea what /kern/xen does.
> > 
> > However, given that the recent changes to /proc turn on the linux
> > stuff by default (which includes things like cpuinfo, mounts, and
> > other non strictly-process related information), I don't think it'd
> > be that much of a stretch to put whatever /kern/xen does in /proc.
> > (I could be completely misguided as to the nature of /kern/xen,
> > admittedly.)
> 
> It is used to provide kernel interfaces for dom0 to some Xen-specific 
> routines, like privileged commands to hypervisor (mapping domU's memory 
> inside dom0, altering domU's states, ...), and expose information 
> required by Xenstore to work properly.

[moving thread from source-changes to port-xen]

NetBSD/Xen can move away from kernfs. Use ioctl() for
privileged commands and device files for the other stuff.

Solaris goes that way at least.

Christoph