> H. Peter Anvin wrote:
> > Richard Gooch wrote:
> > >
> > > >
> > > > Actually, it works as a true notifier if equipped with a buffer
> > > > (which only needs to be allocated when nonempty, which will not be
> > > > the steady-state configuration.) This would actually be my
> > > > preferred choice.
> > >
> > > OK, that's an improvement. But I still think it's a poor-man's
> > > substitute.
> >
> > I disagree. I think it gives you a better framework for doing the right
> > thing.
>
> What happens when the user-space daemon reading the notifier and maintaining
> the database of devices crashes or is accidentally killed? Would there be any
> way to get the device configuration back out of the kernel? Is this even
> something that's worth worrying about?
Write it as a program that can run once or as a daemon. When run as a
daemon, it checks /etc/devfs.conf to see how it should run, when run once,
it accepts command lines and can read other config files besides
/etc/devfs.conf. Further, put in a report only mode, so you can see how the
changes would affect the system before you do it. Add a log with rollback
if you want.
I.e. make it work perfectly in a maintenance mode before you let it
outta the bag as a daemon. For many large servers that see little or no
change, a daemon that could blow things up might not be the best choice, but
a program run when you need it program for adding a CPU card or something
would be a welcome change from rebooting.
For my workstation with a 6 disc CD changer and a zip drive, I'd rather have
the daemon version up and running.
Scott Marlowe
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/