Re: [PATCH] improving JOB kernel/user interface

[Limin Gu <limin@dbear.engr.sgi.com>]

> 
> * John Hesterberg (jh@sgi.com) wrote:
> > The current job /proc ioctl interface is really a fake-syscall interface.
> > We only did that so that our product didn't have to lock into a syscall
> > number that would eventually be used by something else.
> > 
> > The easiest thing for us would probably be to turn it back into a system
> > call, if that would be acceptable for inclusion into the kernel.  We're
> > open to other job interfaces, such as a real /proc character interface,
> > or a new virtual filesystem, or a device driver using ioctls.
> 
> But that system call would still be a single mutliplexor for many calls, right?
> Not ideal.  Have you tried to map to an fs?  It's nice and contained, and may be a
> simple mapping.  Question comes with CKRM, and if they'll have similar needs.  If
> that's the case, first class syscalls (no multiplexor) may be way to go.

Hi Chris,

I don't have much experience on implementing virtual filesystem, 
but I am willing to try it if that is the right interface for job. 
However, I am not sure how to map all current job ioctls to a 
nice and simple filesystem, at the same time I would like to keep 
the user library interface the same so our applications will not 
break.

Would you mind giving me some help on the job ioctls and fs
mapping? Thanks in advance!

Limin

> 
> thanks,
> -chris
> -- 
> Linux Security Modules     http://lsm.immunix.org     http://lsm.bkbits.net
>