Re: syslogd still crapping out
On Mon Dec 13, 1999 at 05:33:14PM -0600, Adam Di Carlo wrote:
> Eric Delaunay <firstname.lastname@example.org> writes:
> > hmm, how do you support multiple kernel revision then ?
> > Actually you can replace the kernel with another one (not necessarily of the
> > same revision) on the rescue disk assuming you also provide a drivers disk.
> > But if you hardcode unix.o in the root disk you also will need to provide new
> > root disk for an alternate kernel :(((
> Damn. That's a damn good point. I didn't think of that.
> What I was thinking of hacking up was
> (a) if modules.tgz *does* contain unix.o, then plop it on the
> rootdisk (of course, this scheme has the problems you mention,
> making the root disk also tied to the kernel, which is ok, but
> not a great idea)
> (b) after booting, in /etc/init.d/rcS, if
> /lib/modules/`uname -r`/misc/unix.o exists, load it
> > Is there no chance to integrate unix module to the standard kernel ?
> > Or is there an alternative to unix socket for mini syslogd ?
> I dunno. Herbert, any ideas?
How about this:
We could simply have a bootmods.tgz package on the boot disk that
contains any boot modules folks want loaded up on boot. When folks
drop in a new kernel, they also adjust the bootmods.tgz package. Then
we make sure bootmods.tgz has a README file in it that says to be sure
to include unix.o, and adjust the included /etc/modules.sh, which is a
script run on boot with lines like 'insmod unix'
On boot, /etc/init.d/rcS extracts bootmods.tgz from the floppy, and then
/etc/init.d/rcS runs /etc/modules.sh
Erik B. Andersen Web: http://www.xmission.com/~andersen/
--This message was written using 73% post-consumer electrons--