RFC: enhancing the root mount logic

Bakul Shah bakul at bitblocks.com
Tue Aug 24 05:52:18 UTC 2010


On Mon, 23 Aug 2010 21:37:38 PDT Matthew Jacob <mj at feral.com>  wrote:
> Yes, this is the RedHat root pivot goop that's been around for ages.
> 
> It turns out to be a massive PITA, because the initrd image can get out 
> of sync with the kernel and hardware, and since some of the modules can 
> be loaded from there, but not from the root filesystem there is a 
> definite possibility (which has happened with more times than I care to 
> remember) that you'll get hosed and not be able to mount your root 
> filesystem.

To avoid getting out of sync is why I was advocating bundling
the ramfs root with the kernel. That too can have problems --
it is all matter of which compromise you can live with.

> This actually can happen so easily that when I install CentOS or Fedora, 
> I override the defaults and put the root filesystem on a plain 
> partition/filesystem rather than as part of an LVM2 volume.
> 
> > BTW, a friend tells me this is what Linux does (or more
> > likely, what they used in their server startup). Basically a
> > ramdisk with init + loadable drivers + tools needed to get
> > going.  Once the actual root fs device is found (even if
> > disks got switched around etc.) they switched to the actual
> > root.


More information about the freebsd-arch mailing list