Re: Building Big Ass Linux Machine, what are the limits?

Albert D. Cahalan (acahalan@cs.uml.edu)
Thu, 1 Oct 1998 18:21:02 -0400 (EDT)


Stephen C. Tweedie writes:
> R.E.Wolff@BitWizard.nl (Rogier Wolff) said:

>> I've been thinking about this "change disk now" questions
>> that the kernel might need to ask.
>>
>> I thought I'd make a character device that simply outputs the
>> question to be posed to the user.
>
> I'd love to know how this is supposed to work on a multi-user system
> that has users logging in from a dozen different modems and X
> terminals at the same time!

The easy answer is "don't do that". Remote users can't do a great
many things. They miss out on the PC speaker, keyboard lights,
floppy drive, power switch, etc.

There is no reason to deny features to the local user just because
remote users can't take advantage of them too. Some people use
Linux for desktop machines.

There is a better solution though...

> The kernel disk quota system also needs to be able to notify the current
> user when quota is exceeded, and currently (in common with other Unixen)
> we just output the message to the process's controlling tty device.
> You'd need much more than just a single character device to deal with
> this in any significantly better way.

Not really, as long as multiple processes can open the device.
Processes (say, part of GNOME) could listen for events from
some desired class of events:

1. all system events (privileged request)
2. all events for a process group
3. all events for a user
4. all events for a group
5. all events for processes on a particular tty
...
n. anything not otherwise caught

This is a rather important issue for a GUI, and I'm sure emacs would
have a use for the info. It's not nice to mess up the display and it's
even worse to have events end up only in the system log file.

It would be really nice to have ssh & sshd cooperate to ship remote
events back to a nice GNOME event manager on the user's display.

Quota violations could block with an interruptable sleep to let the
user delete some files. There could be a random cookie used to let
event managers respond to events that cause a sleep.

-
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/