Re: `root' on cgd with sysctl init.root

[Magnus Eriksson <magetoo%fastmail.fm@localhost>]

I was hoping that someone who actually knows cgd would weigh in, but since 
nobody did so far...


On Thu, 24 Feb 2011, Taylor R Campbell wrote:

> Now that I have set it up, it seems pretty kludgey,

I'm certainly no expert, but as I understand it cgd-on-root is pretty 
kludgy overall.


The potentially good news is that someone is actually hacking on this (Jan 
Danielsson, who explained the problem to me previously) and is, I think, 
aiming for a fairly significant rewrite; instead of a "handoff" you'd have 
a cgd device set up prior to anything important happening (such as 
mounting root) and for all intents and purposes the process would be more 
or less identical to a normal boot, with a root device that just happens 
to be backed by cgd.

As I understand it though, you run into some interesting[*] 
chicken-and-egg problems since the cgd code expects a lot of things to be 
already done by the time it sets up devices.  (Disclaimer: that's all from 
memory, and I'm not even certain I had it 100% correct to begin with.)


Of course, that's perhaps not relevant to you (I assume, at least for your 
point 1) but it seems less kludgy, and could perhaps even actually clean 
up some code too.

That was a while ago though, and I don't know how much time he's had to 
hack on it, but perhaps a friendly nudge and an indication that people 
actually use cgd on root could help move things along.  :-)


Sorry I couldn't be of more immediate help; I just use cgd for "data" 
disks and partitions myself.


[*]: interesting in the "ancient Chinese curse" sort of way.


MAgnus