Re: severe ld.so bug !!

Eric Horst (erich@cac.washington.edu)
Tue, 17 Dec 1996 23:10:13 -0800 (PST)


> ld.so (both 1.7.x and at least 1.8.3 using SuSE 4.4) has a SEVERE bug: if
>
> /etc/ld.so.cache
>
> is an empty file, every binary using shared libs will SEGV !

which brings up something i've complained about before in other forums. a
certain number of critical binaries need to be staticly linked. it seems
to me (and others) that you should be able to boot a system to some very
minimally usable state without requiring shared libs. traditionally the
linux community does this with floppies but is can be done a better way.

i recall working wonders with 'ln' and 'mv' on a sunos box with corrupted
libraries. looking at a sunos 4.1.3 box i see that init=/bin/sh would
have worked (wow! cuz /bin/sh is staticly linked as it should be). and a
handful of useful things in /bin and /etc are static (such as mount,
ifconfig, update, fsck, hostname, ldd, tar, sh, mv, ln). you can do lots
with these.

on my redhat system a brief look at /bin and /sbin reveals only a couple
static bins namely "ldconfig" and the e2fs tools (which are only static
cuz the author knows they should be). this situation should be improved.
the little bit of extra disk space used by staticly linking a few bins is
cheap.

you'll all think i'm crazy until the day it would have saved you some
precious minutes of downtime. take action! write your local governments
or at least the author of your favorite distribution!

--Eric

==========================================================================
Eric Horst University of Washington
Software Engineer Networks and Distributed Computing
erich@cac.washington.edu Voice:(206)543-8959 Fax:(206)685-4045
==========================================================================

On Tue, 17 Dec 1996, Harald Koenig wrote:

>
>
>
> this happend for unknown reasons when e2fsck decided to truncate this
> file and thus every reboot hung immedieately after the
>
> VFS: Mounted root (ext2 filesystem) readonly.
>
> message and things like booting "init=/bin/sh" didn't help either (obviously;)
> and locating this "reboot problem" was no fun (was a long distance call for some hours;)
>
>
> Harald
> --
> All SCSI disks will from now on ___ _____
> be required to send an email notice 0--,| /OOOOOOO\
> 24 hours prior to complete hardware failure! <_/ / /OOOOOOOOOOO\
> \ \/OOOOOOOOOOOOOOO\
> \ OOOOOOOOOOOOOOOOO|//
> Harald Koenig, \/\/\/\/\/\/\/\/\/
> Inst.f.Theoret.Astrophysik // / \\ \
> koenig@tat.physik.uni-tuebingen.de ^^^^^ ^^^^^
>
>