In message <199808210313.UAA03633@shell17.ba.best.com>  Todd Whitesel wrote:
> > Adding a DEV_DETACH/DEV_REATTACH and DEV_GONE would also allow the device
> > to react to insertion/removal, indication with the return code if it should
> > be deleted completly, considered dormant (entry functions switched to
> > error return), or handles the state by itself completly (this would allow
> > a network card to keep the interfaces up, and not drop TCP connections). 
> 
> Preserving TCP connections would be very cool, but you should _block_ user

It would be like a network outage, so if you have keepalives or something
in the sendbuffer, TCP will kill you after sometime, preventing this is
difficult.

> processes that attempt to call the device entry points, not return errors.
> 
> Either that or give me hard,intr and soft options for these devices. This
> exact issue comes up with NFS mounts all the time.

I think this depends on the device (and application??), if a modem is
gone from a pcmcia slot, blocking may be appropiate, esp. for already connected
processes, but it would get a hangup. If a tape is gone from the scsi-bus I 
would recommend an error, there is nothing to salvage there anymore.

> 
> > As I stated in a previous mail, I would prefer a solution where all the
> > device drivers are loaded dynamicaly, but thats sometime when the 
> > Y2K problem is history :-)) 
> 
> I like this. For a while I have wanted to figure out how I could burn a
> minimal kernel into the motherboard flash (kill all BIOS's!!) with just
> enough drivers to bring up the devices and LKM everything else from what
> used to be called my "boot" device.

Yes, just a minikernel with drivers to access the root-filesystem and the
console. You still need a hardware description somewhere on the root-filesystem,
which describes your configuration, but it'll make driver development and
installation stuff so much easier (take that install floppy and this floppy
with the device-driver for your non-standard hardware ... and lets you
change parameters before probing/attaching the device).

Again, if this can go in the main tree, I'm willing to create the framework
for it (kernel-version of ld.so etc. ).
I would make this an addtion to the current framework, so we don't hurt 
machines with can't do PIC to much.
The only necassary global change would be to make [cb]devsw an array of
pointers (else we would waste a lot of space).

Stefan
> 
> Todd Whitesel
> toddpw @ best.com

--
Stefan Grefen                                Tandem Computers Europe Inc.
grefen@hprc.tandem.com                       High Performance Research Center
 --- Hacking's just another word for nothing left to kludge. ---