[Clfs-support] Cannot login on serial port (embedded ARM, busybox)

[Lance Jump]

On Mon, Oct 10, 2011 at 2:09 PM, Andrew Bradford <bradfa at gmail.com> wrote:

> On Mon, Oct 10, 2011 at 11:39 AM, Lance Jump <lancej29 at gmail.com> wrote:
> >> If you follow the embedded book exactly for creating the user account
> >> to build an embedded system, nothing will leak into your build from
> >> the host.
> >
> > Yes, it is well thought out in that respect.  The problem is the "follow
> it
> > exactly" part. Experience has shown me that I make enough mistakes to
> want a
> > safety net.
>
> Understandable.
> If you can think of any way we could make the user creation less
> complex but still keep the nice separation between the host and target
> builds, please let me know.  I admit the current method is not as
> brain-dead-simple as it could be.  User creation is one of the spots
> where it's easy to make one tiny mistake that ruins and entire build
> by polluting the target environment.
>

I'd have to look at it, but I didn't think it was too bad the way it is now.
Actually, that part is not as compelling to me as being able to the build
relatively independent of distribution.  By chroot'ing and working in the
simple, but complete LFS-live system, I am able to work pretty independent
of the actual host distribution.  I did this a few years back for an LFS
build for an embedded X86 and it worked great.  It was repeatable and the
scripts worked without changes on several host distributions.


>
> > I agree that scripts are not running.  My earlier experiments of putting
> > messages (including echoing to files) appears to confirm this.  I seem to
> be
> > able to run various things by replacing sysinit and TTY entries in
> inittab.
> > So inittab is being read and busybox runs as do various applets.  But it
> > seems that sh (or ash or whatever) itself will not actually run so no
> shell
> > scripts (e.g. startup) can run.
>
> Stupid question, but you do have the symlink from
> ${CLFS}/etc/init.d/rcS to ../rc.d/startup, right? (section 10.2)
> That's required unless you want to change the config used for BusyBox
> before building it.  I don't remember (and not in front of a Linux box
> right now) but I think the bootscript location is somewhat hard coded
> (I think) into BusyBox which is why we have the symlink right now.
> Changing the bootscripts makefile and naming of the startup script
> could fix this and make it simpler.
>

Yes, the symlink is there.  As a test, I changed the script name in the
inittab file and and I get an error:

VFS: Mounted root (nfs filesystem) on device 0:13.
Freeing init memory: 156K
init started: BusyBox v1.18.4 (2011-10-07 23:48:57 GMT)
can't run '/etc/rc.d/startupx': No such file or directory

This suggests two things.  First, it is finding the correct file (without
the change) since I get no error and, second, that something is properly
processing inittab. What I can't tell is whether the message is coming from
the shell processor or something else (Busybox init?).


>
> Boot scripts haven't gotten much love yet, sorry.
>

No need to be sorry -- it's not like this is a paid, supported effort.  I
very much appreciate what you have done and how much work it must be to
create and maintain it. Besides, once a system has gotten to the point where
it can boot, creation/maintenance of boot scripts are something the end user
can, and probably should, do.

I prefer the approach of keeping it simple so that one can learn from it.
Learning has been my goal in this and, despite the issues (or, perhaps,
because of them) I have been learning.


>
> -Andrew
> _______________________________________________
> Clfs-support mailing list
> Clfs-support at lists.cross-lfs.org
> http://lists.cross-lfs.org/listinfo.cgi/clfs-support-cross-lfs.org
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.clfs.org/pipermail/clfs-support-clfs.org/attachments/20111010/6403bdfb/attachment-0001.htm>