[995] in NetBSD-Development
Updated install server & floppies
daemon@ATHENA.MIT.EDU (John Hawkinson)
Fri Sep 1 06:41:06 1995
Date: Fri, 1 Sep 1995 06:40:35 -0400
To: netbsd-dev@MIT.EDU
From: John Hawkinson <jhawk@MIT.EDU>
I've updated the install floppies so that:
On ep?* cards (like the 3c509) they check the case
of link0 link1 -link2 first, so that they will not be fooled
by false posivies. The number of false positives seems to be
unpredicatable, but in general it tends to be about half the
time on a flood ping of two packets. This should eliminate
a lot of strife.
The .shinit file contains an environment variable
with the location of msgbufp in the kernel so that the install
script on the server can run dmesg without needing the kernel
off the floppy. Sneaky, sneaky.
The version of the floppy is printed out at boot time. This
should probably be saved in an environment variable somewhere so
that the install script can include it in the mail it send (see below)
but for now it's better than what we had (no user-visable versioning).
The old images are in nfsaha.fs.old and nfsbt.fs.old.
I have also updated the install server on
sipb-nfs:/u2/lockers/netbsd/i386.current/install (the old one is in
install.old).
When the script finishes, it strongly suggests that the
user allow the script to shut down the machine, but otherwise strongly
cautions them to use halt. This is on the theory that <ilin@mit.edu>'s
ffs corruption (wherein he installed and then when he tried to edit
/etc/XF86Config his machine paniced0 was due to filesystem corruption
due to mounting it asynchronously and then not rebooting properly. I
guess we'll never know for sure.
Additionally, the script asks if it's ok for it to mail us
"configuration information" (the default being yes, it's ok). If so,
it sends a long message to netbsd-install, which goes to
bloom-picayune:/usr/spool/discuss/netbsd-install. That meeting is
currently acl'd to the members of the moira group netbsd-dev. Feel free
to take a look -- the last transaction reflects the current setup except
that the install script was not checked out so the $Revision$ tag
doesn't look very impressive. The intent of this is for debugging and
diagnostic purposes, so please keep the information confidential.
Yoav and Greg pointed out that they would like the script to:
1) Update /etc/athena/version with the installation time
like mainstream-platform athena installs.
2) Update /etc/rc.conf with real information
In order to do the former we need to:
1a remove /etc/athena/version from the syspacks and tarballs
1b decide on where we want to get the "Athena Workstation"
line and how it should be derived.
For the latter, we just need to
2a Change the rc.conf on the syspacks to have NONAME, NOADDR,
and NODEV (just like on the decstation syspacks).
2b Modify the athena-inst script to update those fields.
So if somenoe gets crashing on [12]a, then later someone can do [12]b.
That's about it. I need sleep.
--jhawk
