stuck /etc/rc autoboot processes

[Damien Fleuriot]

On 12/28/11 10:46 AM, Damien Fleuriot wrote:
> 
> 
> On 12/27/11 10:22 PM, Jeremy Chadwick wrote:
>> On Tue, Dec 27, 2011 at 07:46:52PM +0100, Damien Fleuriot wrote:
>>> Hello list,
>>>
>>> Yesterday and today, I've been busy either patching boxes for the BIND
>>> advisory that we received on the 23rd (when they were running 8.1 or
>>> 8.2-RELEASE), or upgrading them (when running 8.0-RELEASE).
>>>
>>>
>>> Today I've come across 2 boxes running 8.2-STABLE and of course, the
>>> BIND patch wouldn't apply correctly.
>>>
>>> I've decided to cvsup them to 8.2-RELEASE and "upgrade" them to it.
>>
>> Why cvsup and not csup?
>>
> 
> You're correct, and I indeed used csup from the base system.
> 
> 
>> Secondly, and more importantly, when upgrading to a different tag (e.g.
>> you were using RELENG_8_0 before, moving to RELENG_8 or RELENG_8_2 --
>> you don't really explain in a coherent way what you did, you've
>> mentioned 4 different FreeBSD versions above :-) ), I tend to do the
>> following:
>>
> 
> Indded I've mentionned quite a few versions, because I had to patch
> and/or upgrade many of them over the last few days.
> 
> They all went well, because most machines either didn't need upgrading
> (only BIND's patch), or were running 8.2-PRERELEASE or 8.0-RELEASE so I
> could upgrade them just fine.
> 
> /usr/src/UPDATING didn't yield any warning so I went ahead.
> 
> 
> 
> The "upgrade" that troubles me with these stuck processes is 8.2-STABLE
> down to 8.2-RELEASE-p5.
> 
> For this as well, I have not seen anything in UPDATING.
> 
> 
>> rm -fr /usr/obj/*
>> rm -fr /var/db/sup/src-all
>> rm -fr /usr/src/*
>> csup ...
>>
> 
> That's a goood idea, I'll note that down.
> 
> Are you sure about removing /usr/src/* ?
> 
> I usually cd there and make update, without the Makefile it's gonna be
> tricky.
> 
> 
>> The problem I've seen is that "some" source bits manage to figure out
>> that they need to be updated to a different version based on the release
>> tag in the cvsup/csup configuration file, but sometimes this doesn't
>> work quite right and the underlying source files in /usr/src end up
>> getting "mix-matched" between two versions.
>>
>> You may want to do this for ports as well, e.g.:
>>
>> rm -fr /var/db/sup/ports-all
>> rm -fr /usr/ports/*
>> csup ...
>>
>> It gets more tricky assuming during your original FreeBSD installation
>> you chose to install src and ports.  The below is the cvsup FAQ, but it
>> applies to csup too.  Read items 11, 12, and 13.
>>
>> http://www.cvsup.org/faq.html#caniadopt
>>
> 
> These boxes date back 1.5 year now and I don't remember how I installed
> them.
> 
> I think it was sysinstall so yeah, sources + ports at the time.
> 
> 
>> Welcome to why I never bother to install src or ports from CD/DVD, I
>> simply use csup once the system is up.  And for changing tags/releases,
>> I do what's described above.
>>
>>> I've gone through the following steps:
>>> - make buildworld
>>> - make buildkernel
>>> - make installkernel
>>> - nextboot -k my new kernel, to ensure it worked fine
>>> - rebooted again with the new kernel, this time correctly installed as
>>> /boot/kernel
>>> - installed the world
>>> - run mergemaster -FiPU
>>> - rebuild ports
>>
>> This doesn't look correct.  The process you should be following is
>> documented plainly in /usr/src/Makefile.  You're missing some steps.
>>
>> Try doing what I recommended above, and following what's in
>> /usr/src/Makefile, and then see if things improve.
>>
> 
> I'm reading through it now, the only steps I haven't run seem to be
> delete-old and delete-old-libs.
> 
> I'll try that removing all /usr/src/ , /usr/obj/ , csuping again and
> redoing the steps including the removals.



Ok so, for information.

I've tracked this down to "Local package initialization:" stuck
indefinitely at boot.

I just logged on the machine's remote console and ^C 'd it, and that
gave me a login prompt (although I could ssh just fine otherwise !) and
cleared the autoboot processes.

I'm now looking for the reason why this message was displayed and why it
was stuck.