I'm having trouble with coda installation - any thoughts?

I am trying a coda installation on a freshly created RedHat linux 5.1
system, attempting to put the server and client on the same machine.
The server installation went fine, as far as I can tell.  It involved
adding some partitions, installing some RPMs, answering some questions
from vice-setup; nothing seemed alarming, startserver seemed happy,
assorted Coda daemons are running. For all I know, I have a working
SCM Coda server.

The client, however, is a different story. I installed all the RPMs,
then grabbed the stock 2.1.120 kernel. I compiled the kernel with Coda
support enabled.  (make dep; make clean; make zImage; make modules;
make modules_install) and made a bootdisk (mkbootdisk 2.1.120) for
booting into the development kernel. Then I ran the venus-setup
(venus-setup localhost 40000)

One thing to note is that I only have the localhost interface running
(lo) as I am not sure how to get my eth:0 interface working under
2.1.120 quite yet. Still, pinging localhost works fine.

So far, so good. Until this:

% venus -init &
coda_psdev_write: downcall, no SB!
Just mapped in region 20000000,4096)
Just mapped in region 21000000,4481024)

% cat /usr/coda/etc/vstab
/coda:/dev/cfs0:localhost:/usr/coda/venus.cache:40000:1

% cat /usr/coda/etc/console
Coda Venus, version 4.6 (28)

Date: Wed 09/09/98

19:01:34 /usr/coda/LOG initialized at size 0x112f60
19:01:34 /usr/coda/DATA initialized at size 0x44bd80
19:01:34 brain-wiping recoverable store
19:01:35 loading recoverable store
19:01:35 fatal error -- CommInit RPC2_Init failed
19:01:35 Fatal Signal (11); pid 462 becoming zombie...
19:01:25 You may use gdb to attach to 462

% ls /coda
NOT_REALLY_CODA