Last minute errata:
o login as root produces "login_getclass: unknown class 'root'" on system console.
Fix: If you have the source distribution installed, simply
cp /usr/src/etc/login.conf /etc
otherwise, get it from the FreeBSD FTP site using this URL:
instead. Simply cd to /etc and then run fetch(1) with the provided URL.
o sysconfig scrambles rc.conf if run again.
Fix: Get updated /usr/src from RELENG_2_2 branch and build
/usr/src/release/sysinstall, copying the new binary to /stand.
If you do not have enough space for src then you could also
use the boot/fixit floppy combo from a later 2.2-YYMMDD-RELENG
release to simply mount your root partition (using the Fixit
option) and copy /stand/sysinstall from the floppy to /stand on
your root fs.
o Installation floppy does not boot at all - whereas the 2.2.1 floppy
worked fine. I get a "panic: double fault" right after it tries to
change the root device to fd0c.
Fix: The problem is that you have 48MB of RAM and something very
mysterious has happened to FreeBSD twixt 2.2.1 and 2.2.2 which makes
it fail with just that exact memory size. Given the popularity of
16MB simms, it also explains why none of us have seen it since we
typically have either 16MB, 32MB or 64MB of memory in our systems. :)
We're working on finding and fixing this problem, but until then
the following work-around is in effect for 48MB systems:
1. Boot the 2.2.2 boot floppy and when it comes to the first menu which
asks you whether or not you want to go into the kernel configuration
editor, choose the "experts only" CLI mode option. Now type:
iosize npx0 32768
< and do your visual kernel configuration as normal then exit>
If you can get through to the installation, go to step 3.
2. If the above does not work, physically remove all but 32MB of memory
from your machine and then boot the boot floppy. Unless your problem
is totally weird and something we've not seen at all before, you
should now be able to go on to step 3.
3. Complete the installation and then boot off your hard disk. This
boot should work fine, since you are no longer using the memory
filesystem that the installation uses and which seems to interact
badly with these memory size issues to create the failure you saw.
You will also want to boot with the -c flag at some point and
say "iosize npx0 0" to get the full use of all your memory back
since the old value of 32768 will have been saved to disk during
the initial installation. If you already plan on building a custom
kernel, you can skip this step since the value will be reset anyway.