[gentoo-user] Re: No /dev entries in recent stage3 snapshots?

[Grant Edwards]

On 2009-01-24, Neil Bothwick <neil@××××××××××.uk> wrote:
> On Fri, 23 Jan 2009 05:55:10 +0100, Volker Armin Hemmann wrote:
>
>> I have a server running that hets that null/console missing
>> message every boot - and it does not hurt it at any way.
>
> A missing /dev/console stops the boot process here. It boots
> without /dev/null, but only after udev spews out a load of
> messages.

Ah, not to worry: I've been assured in the gentoo forum thread
that the problems we see when the root filesystem doesn't have
proper /dev/null and /dev/console nodes aren't really
happening:

  Neither /dev/null nor /dev/console are needed at boot-time,
  therefore their absence doesn't cause problems.

[You must admit the argument is flawless -- though I still
question the premise.]

In order to get rid of my problems that weren't happening, I
initially tried the "mount -bind" and "cp -a" commands that
show up in /etc/issue when your /dev directory is hosed.  That
didn't help: after setting /etc/issue back to the default file
and rebooting all the same problems still weren't happening
(and /etc/issue was again modified to tell me to do mount -bind
and cp -a to fix them).

Then I tried booting with root in rw mode and init=/bin/bash
and then doing a MAKEDEV generic-i386.  (I found that recipe in
an old mailing list somewhere.) MAKEDEV complained a lot about
not being able to read /proc/devices. When I rebooted, I still
had the all same problems not happening as before.

I finally booted from a minimal install CD, mounted my root
partition, removed its /dev directory completely and then
re-created it by untaring ./dev from a good stage3 tarball. Now
the system boots up smoothly. I feel like a bit of a fool
expending so much effort getting rid of problems that weren't
happening -- but, now the problems that weren't happening are
gone, so I'm happy.

;)

-- 
Grant