Subject: Re: compiling a new kernel?
To: Michael Graff <email@example.com>
From: Chris G Demetriou <Chris_G_Demetriou@ux2.sp.cs.cmu.edu>
Date: 08/18/1996 15:09:48
> der Mouse <mouse@Holo.Rodents.Montreal.QC.CA> writes:
> > And don't make the mistake I did, of doing "cp netbsd /netbsd+ && mv
> > /netbsd /netbsd- && mv /netbsd+ /netbsd && reboot". If the old kernel
> > is called netbsd- then you can't boot from it (!), at least on some
> > architectures. It turns out the second-stage boot program has a really
> > dumb parser that takes a - as a flag marker even when it's not preceded
> > by whitespace. (Nowadays I use /netbsdN and /netbsdX.)
> The reason I use netbsd and onetbsd is simple. The second-stage boot loader
> for i386 looks first at netbsd, then onetbsd, then something else which I
> have yet to need. :)
And, since i've jumped into this discussion:
Typically, i name kernels:
/netbsd.new testing kernel, likely to crash,
probably going to be used for one boot
/netbsd normal kernel, i.e. stable enough to
boot and mount single-user and frob
things and reboot.
/netbsd.bak last kernel that i was running that
was "normal," but not necessarily
stable enough to actually run for
/netbsd.old last stable kernel that i trust.
I use this as the replacement for
/netbsd, if it seems losing.
/onetbsd oldest stable kernel that should
run on my current configuration
(i.e. which supports my user-land
binaries and hardware, without
I try to replace /netbsd.old at least once a month.
I typically replace /onetbsd once every 2-4 months, but
sometimes more often depending on what i'm doing, in terms
The alpha boot blocks only boot the kernel specified at the
console prompt, so the names are arbitrary.
It's always bugged me that on the i386, if /netbsd is bad but is
"good enough" to be loaded (i.e. but doesn't boot successfully),
the machine will reboot forever. However, there's no way that I know
of to do anything about that. 8-(