RFC: support for re-sourcing /etc/rc.conf

[Colin Percival]

On 12/02/13 11:17, Teske, Devin wrote:
> On Dec 2, 2013, at 10:45 AM, Colin Percival wrote:
>> No, the other way around -- a different script (part of ec2-scripts) updates
>> /etc/rc.conf based on metadata provided at VM launch time.  One use of this
>> would be to set firstboot_pkgs_* variables to control the behaviour of the
>> firstboot-pkgs port moments later in the boot process.
> 
> sysrc(8) in 9.2+ base comes to mind.
> NB: Available as a port for older releases 4.8+ (sysutils/sysrc)

I should have been clearer, the VM launch user-data could specify any files to
modify; it just happens that rc.conf is a problem because it's over-aggressively
cached.

>> 3. Early in the boot process, ec2-scripts reads that user-data, updates
>>   /etc/rc.conf, and signals /etc/rc.
>> 4. /etc/rc re-sources /etc/rc.conf.
> 
> Cool. How's the signaling in #3 done to get #4?

My (one line plus comments) patch teaches rc to re-source rc.conf if it receives
a SIGALRM.

> I like the idea of firstboot_pkgs_*

Great.  It's in the ports tree, sysutils/firstboot-pkgs.

> I also like Hiroki's idea of stashing it all into rc.conf.d

As I mentioned in reply to him, forcing people to use rc.conf.d instead of
rc.conf seems like a usability challenge.

-- 
Colin Percival
Security Officer Emeritus, FreeBSD | The power to serve
Founder, Tarsnap | www.tarsnap.com | Online backups for the truly paranoid