Re: Kernel panic. No init found
Tim wrote:
> At 01:17 20/07/02 +0100, Oliver Elphick wrote:
> >olly@linda$ ls -l /lib/ld-2.2.5.so /lib/libc-2.2.5.so
> >-rwxr-xr-x 1 root root 88578 Jul 17 04:52 /lib/ld-2.2.5.so
> >-rwxr-xr-x 1 root root 1149584 Jul 17 04:52
> >/lib/libc-2.2.5.so
>
> I've located the original files having seen your message; the problem was a
> name change in the symbolic link files (you may have noticed this-loss of
> one digit). I've corrected this by creating correctly named symbolic
> links, and the boot phase now goes past the 'No init found' message. So
> progress is being made!
>
> But. The boot process stops further on:
>
> /bin/sh: error while loading shared libraries: libncurses.so.5: cannot open
> shared object file: No such file or directory.
> INIT: Entering runlevel: 2
> /bin/sh: error while loading shared libraries: libncurses.so.5: cannot open
> shared object file: No such file or directory.
> (none) login: root
> /bin/login: error while loading shared libraries:libdl.so.2: cannot open
> shared object file: No such file or directory.
> (none) login: root
I'm using an older distro so I'm probably not totally fresh here, but of the
files that seems to be missing for you,
libdl is I believe the most crucial one, and it has to be
the right version as well.
It's not that large. (< 100kB)
libncurses (4.2) (< 300k
In my (very old) system libdl exists in /lib
while libncurses is installed in /usr/lib
( I'm still puzzled as to what program it is that uses ncurses while booting ?
I would actually consider it to be a bug.)
Other file that might be of some importance I think, is /lib/libm.so
Any way it seems like file corruption has struck you in more directories than
one.
A guess of mine is that You have downloaded Debian from a phone line?
This would make a complete reinstall pretty awkward.
So perhaps it's worth some effort to try to save the remains of your in-
stallation, when facing a near 50 hour download otherwise? second guess?
I'm not saying that it is though, Do read on to see what you have to deal
with.
You say that windows is bad for downloading large files? There are some tools
that you can download for downloading purposes. One is called Gozilla, the
other is called Getright. By ease of use I prefer Gozilla even tough it has
a funny gui. I believe they work in some similarity with the wget tool
for linux, or apt-get(without being package or domain specific)
Before I Continue Tim I want to say that I wrote this in order to clear
my own point of view regarding some of the Debian installation philosophy.
So All of this text is not really just for your eyes. mostly it's there for
my own an perhaps for a few others.
If you are just interested in specific info that might be of interest
to you regarding savouring your broken file system look a head for the word:
`sorry-bout-that-tim'
File system corruption really shouldn't mean that you are forced to do a
full re-install, especially not when it takes 50 hours to download the stuff
This is no better than f ex The Redhat or Suse philosophies.
Yet Debian developers say that they want to bring a distro with quality
assurance. I think that Debian developers should at least try to make their
system just a little bit more failsafe than the others, just because of
the network download and install issues.
I run on a tripple backuped system (yes I have a dos rescue partition
of 15MB with loadlin and root ramdisks (hda1), One 50 MB linux Resque
only partition (hda3), as well as a fairly resent bzipped image of my
install(hdc2), Yes forgot I also have some boot floppies that I will
probably never need ) I always use this install scheeme when I install
linux nowadays. And I need to have things this way, as I have noticed.
There fore anyway, I will probably never have to fear the trauma that
you are having right now. I have been close to it, in the beginning
before I built my resque scheeme but even then I always had an Installation
cd to use as a last resort.
This funny situation that you seem to be in is some what due to Debians
(lack of ?) philosophy. Every body seems to recommend every body else Not
to buy CD:s but instead download the needed software from the web. at the
delight of Many of debian unsupportive ISP:s Why?
sorry-bout-that-tim
I am only collecting mail from debian-user once a day, so I will not be
able to help in any step by step issue. You will need someone that runs
almost exactly the same kind of system that you had your self, since you
would want to compare shared library sizes, etc. I don't have.
I can point out a few interesting things:
If you have run e2fck, and it has made severe repairs to your file system,
you will probably find a whole bunch of numbers in the directory
/lost+found
Theese numbers represent (I believe) Inode numbers.
Every one of theese numbers is a file that once had a name and lived happily
in your file system. If you are in the least lucky, then most of theese
files are actually intact with size, owner rights ??? and all. Your main
problem will be to pair each one of these files with a file name and a
directory. This is not hopeless, but it will take time (lots), and effort.
Some of theese files might actually be directories. (I don't know the likely-
ness of this, but if I'm not right then some one will complain to this. )
Any way If you happen to find a directory file Then you have hit a small
jackpot.
A directory file: Basically Combines three things : Inode-number- the lengh
of a directory record, the lengt of a filename, and the filename it self
This is repeated for every single Inode - filename pair entry in it.
(This I know for sure) The first name record in a directory file is always
the directory files own inode, the name record is a simple ascii "." . The
next record is the inode of the parent directory the name record is in
ascii ".." theese two records are created when you mkdir. The following
records are the inodenumbers of files that you copy into the directory,
paired with filename lenghts and the filenames them selves (they are also
written in ascii)
So if you do cat one of these files you will see a lot of filenames.
You will only be able to see the Inode numbers, and lenghts that is in between
the file names in a hexeditor or similar. To determine what is an inode
number and what is a record lenght you will have to know that an Inode
takes up 32 bits of space (4 bytes) a name will only take up 8bits of
space (1 byte) while the directory record lenght is I believe: 16 bits
(2 bytes)
There are some tools for file system and file (directory) snooping and
editing in linux that I know of. One of theese tools is usually shipped
together with e2fck,
It's called debugfs. I don't really like it my self but I believe
that you can use it ( with some care though. Read manuals.
Theese tools are made for unmounted file systems, which you open read only
unless you really know what you are doing.
Another program is ext2ed. A problem with ext2ed is that it only works on
smaller partitions, (2GB) same here read manual carefully. ext2ed
has a very nice overview of the ext2fs storage structures.
Other files of interest:
A binary compiled file always begins with sequence ".ELF" .This makes it
possible to actually "grep" sort out all your lost libraries and
binaries. A few other useful tools:
useful commands
You have probably come across ldd all ready
the file command by which you can also determine filetypes.
the stat command will perhaps give you other file characteristics
such as file creation date file modification date, file size,
file type, etc.
nm tells you what variables a binary file such as a shared library contains.
This will give you a good fingerprint id on a binary.
file type, etc
/got to go now Good luck What ever you choose to do.
/Daniel Mose
--
To UNSUBSCRIBE, email to debian-user-request@lists.debian.org
with a subject of "unsubscribe". Trouble? Contact listmaster@lists.debian.org
Reply to: