Re: Migrating to larger numbers

Stephen Frost (sfrost@ns.snowman.net)
Tue, 8 Jun 1999 14:10:24 -0400 (EDT)


On Tue, 8 Jun 1999, Matt Aubury wrote:

> > Well, couple things, for one, what about those 'more important
> > drivers'? My biggest problem w/ linux these days is the scsi device
> > stuff. If this could be fixed w/ shell scripts and reading info out
> > of /proc, then that may work. But then, what about USB, PCMCIA, and
> > other dynamic pieces? Do you have modprobe or whatnot create and
> > delete the nodes? Do you just create the nodes w/ the assumption that
> > they'll be used?
>
> There are big bootstrapping issues with any device which may hold the
> root filesystem, so I doubt if it would be a good idea for SCSI.

I think perhaps I'm asking the wrong question. The problem I
have is the whole /dev/sd?? stuff. devfs allows for /dev/dsk/c0t0...
kind of addressing, which to me makes alot more sense and is easier
for me to work with. I don't see a way to do this (currently) in
user space.
Perhaps the problem is that SCSI major/minor numbers are
not addressed in a way that I would find useful. Or are there other
major/minor numbers addressed to scsi that are addressed wrt controller
and target? Of course, if you then have to create all of them before
booting because of bootstrapping issues, well, that's a heck of alot
of devices...

> As for dynamically pluggable devices, I guess the correct way to
> handle these is via a daemon which invokes the script when new devices
> are added or removed. That way we only have inodes for devices that
> are actually present, not for ones which might turn up later.

Hmm, I think maybe the question is, which way is it driven?
Does the kernel automagically load the driver when the device is
detected, or does the user attempt to access a device, which the kernel
then has to go out and see if it exists (And if it can load a module
for it). I've seen it both ways.
If the kernel detects it, then having a user-space daemon
running that handles the modprobe or whatnot and the creation of the
devices will work (pcmcia seems to work in this way). However, what
about devices that exist at all times in a system, but are compiled by
the user as modules to perhaps save memory? Or the issue with printers
and zip drives, where IIRC the user can rmmod and insmod the appropriate
modules depending on which device he wants to talk to.

I don't know if devfs handles all of these or not, I'm just
trying to bring up points where it would be nice to have something that
handles these different situations, and handles them in One way (if at
all possible) to make it easier for the user.

Stephen

-
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/