Re: Security hooks, "standard linux security" & embedded use

[Anton Altaparmakov]

At 19:47 12/07/2001, Greg KH wrote:
>On Thu, Jul 12, 2001 at 07:37:36PM +0100, Anton Altaparmakov wrote:
> >
> > This seems very good in view of implementing ACL support for NTFS, too. -
> > We have all the NTFS layout knowledge to do it now. We just lack the
> > kernel/user space infrastructure.
> >
> > When designing this modular security infrastructure it would be useful if
> > it is made generic enough to allow callbacks into user space for 
> permission
> > checking.
>
>The current model lets you do whatever you want in your kernel module.
>It imposes no policy, that's up to you.

Ok, that's fair enough. A wrapper module could always be written that then 
in turn invokes user space. That's good enough for me although it makes for 
additional overhead but I guess that is not too bad.

>All the better to keep userspace callbacks for security out of my
>kernels, for that way is ripe for problems (for specific examples why,
>see the linux-security-module mailing list archives.)

Oh, sure. There are problems. I don't deny that. But I am not too sure that 
those problems outweigh the problems created by putting in huge amounts of 
code into the kernel which could live outside it just as well. - IMHO the 
kernel should be as small as possible rather than contain everything under 
the sun just because it's easier to do that way...

Best regards,

Anton