On Sat, 2012-01-14 at 15:58 -0800, Bryan Kadzban wrote:
> Matt Burgess wrote:
> > This passes a boot test with no changes required to the bootscripts
> > or fstab.  Maybe I'm misunderstanding Bruce's and Bryan's comments in
> > the ticket, but this to me suggests that Udev >= 176 doesn't require
> > a devtmpfs mounted on /dev
> 
> That's very strange, given this commit included in -176:
> 
> http://git.kernel.org/?p=linux/hotplug/udev.git;a=commitdiff;h=220893b3cbdbf8932f95c44811b169a8f0d33939
> 
> (see udev-node.c), which removes all the mknod() calls.  If you don't
> have a devtmpfs mounted at /dev (but rather just a tmpfs), I don't see
> how anything will work anymore.  :-)
> 
> > (note that I do have devtmpfs support in the kernel, but I've added
> > nothing to fstab).
> 
> Did you edit the udev script to use devtmpfs instead of tmpfs, possibly?

You give me too much credit, Bryan :)

What I actually did was to include DEVTMPFS_MOUNT in my kernel .config
which automagically mounts devtmpfs to /dev.  I could try testing
without that option enabled and watch things break, but I'll trust you
and Kay when you say that is what will happen.  So, I think I'd rather
be explicit in our configuration in LFS, and put an entry
into /etc/fstab, even though it's not strictly required iff you have
DEVTMPFS_MOUNT selected in your kernel config.

As a possibly interesting aside, even though the DEVTMPFS_MOUNT
seemingly does the right thing here, it does not cause /dev to be listed
by either 'df' or 'mount'.  I'm assuming that the /etc/fstab entry will
result in the expected output, but will confirm after further testing
tomorrow.

Regards,

Matt.

-- 
http://linuxfromscratch.org/mailman/listinfo/lfs-dev
FAQ: http://www.linuxfromscratch.org/faq/
Unsubscribe: See the above information page

Reply via email to