I'd still like to hear from anyone with insight into this (whether an actual
problem occurred or not), but here's what I've decided to do:

I know from the script output that the buildworld and buildkernel succeeded.
The only step that is suspect is the installkernel.  So:

In case something did screw up, I want to make sure I still have the known
good kernel available:

rm -rf /boot/kernel
mv /boot/kernel.old /boot/kernel

Then I booted into single-user mode and ran:

uname -a

It reported the unpatched kernel (good).  So:

cd /usr/src
script /root/mw/mw-200405111844.out
make installkernel KERNCONF=GOLLUM

This time I sat around and waited for it to finish.  Finally, it did, with
no errors or other suspect output.  Satisfied, I rebooted.  No problems, so
back into single-user-mode to installworld:

cd /usr/src
script /root/mw/mw-200405111852.out
make installworld && echo "sleeping now" && sleep 600 && reboot

I would have waited around for this to complete but I wanted to go home.  I
reasoned that 10 minutes should surely be enough for script to flush output
to disk (man 1 script says it flushes, by default, every 30 seconds).  By
rebooting I would then be able to continue setting up the server through an
SSH session from home (I established and tested SSH settings on the company
firewall before beginning the make world).  Sure enough, I just logged in

> tail mw-200405111852.out
===> etc/sendmail

>>> Rebuilding man page indices
cd /usr/src/share/man; make makedb
makewhatis /usr/share/man
makewhatis /usr/share/openssl/man
rm -rf /tmp/install.8m808LpI
sleeping now

So it looks like everything's good.

Is there any flaw in my reasoning?  In particular, even though it appeared
to work, is it a bad idea to delete the active kernel?  I've done it before
and I haven't had any problems, but I always feel like I'm playing Russian
Roulette when I do it.


-----Original Message-----
[mailto:[EMAIL PROTECTED] Behalf Of Dan MacMillan
Sent: May 11, 2004 18:11
Subject: make world : did it fail?


I just ran the make world procedure on a freshly installed FreeBSD system.
After dropping to single user mode, I ran:

cd /usr/src
script /root/mw/mw-200405111310.out
make buildworld && make buildkernel KERNCONF=GOLLUM && make installkernel

I do things this way because I don't want to babysit the machine waiting to
proceed to the next step.  When I heard the computer reboot (it's sitting
right next to me), I logged in and ran:

tail /root/mw.out

To see if anything untoward had happened.  The result looks like this:

> tail /root/mw/mw-200405111310.out
install -o root -g wheel -m 555   safe.ko /boot/kernel
===> sbni
install -o root -g wheel -m 555   if_sbni.ko /boot/kernel
===> scsi_low
install -o root -g wheel -m 555   scsi_low.ko /boot/kernel
===> smbfs
install -o root -g wheel -m 555   smbfs.ko /boot/kernel
===> sound
===> sound/pcm
install -o root -g wheel -m 555   snd_>

Doesn't look so good, does it?  Now I'm not sure if it rebooted because of
the '&& reboot' or because it ran into some terrible no good awful bad
problem.  I checked the timestamps on all the files in /boot/kernel and they
all look good (all .ko files dated today within 1 minute of each other, the
kernel a few minutes older).  The number of files in /boot/kernel matches
the number in /boot/kernel.old.  I ran (from my home directory):

ls /boot/kernel > kernel
ls /boot/kernel.old > kernel.old
diff kernel kernel.old

And there are no differences.

uname -a now reports 5.2.1-RELEASE-p6.

I'm thinking that probably what happened is that the machine rebooted
without the remaining script output being flushed to disk.  Is there any way
to tell for sure?  And should this even be possible?  I've followed this
technique on my home machines several times and never got the script output
truncated like this.

I'm now cursing myself for putting the && reboot on there.  I can re-run the
make world procedure, but unless I figure out for sure what happened here
I'm afraid it's indicative of a more significant problem.

Thanks in advance.

Danny MacMillan

[EMAIL PROTECTED] mailing list
To unsubscribe, send any mail to "[EMAIL PROTECTED]"

[EMAIL PROTECTED] mailing list
To unsubscribe, send any mail to "[EMAIL PROTECTED]"

Reply via email to