• Please review our updated Terms and Rules here

486 Unix Server Help

Liebo

New Member
Joined
Dec 11, 2011
Messages
6
I've been given an old server from work, it's a 486 DX2 50 with 32 MB RAM. It has a 1 GB SCSI hard disc drive and a tape drive. It also had a floppy drive but it wouldn't boot from it. There is an IntelliPort II card, a Trident 1MB VESA VGA display card, the I/O card and a 3Com Etherlink NIC. I've bought an ISA Network card off of eBay with an RJ45 as it is currently only has a BNC connection on the current card for future use.

The server is running SCO OpenServer Release 5 and eventually gets to the login prompt. There are no known credentials or emergency boot floppies for the server.

As it has some old data on, I've promised that I will try and back up the data if I can and give work a copy then wipe the hard drive before I use it.

As the root password is unknown, I'm trying to mount the hard drive to clear the password or take a copy of it.

So far I've managed to get the floppy booting by disabling the floppy controller on the I/O card (jumper pin) but leaving it enabled in the BIOS and connecting the floppy drive to the Adaptec 1542 CP SCSI Card (floppy support DIP switch was already enabled).

I can boot Tom's boot floppy but I'm having difficulty mounting the hard drive.

I believe the file system to be HSFS (as it displays this when booting the hard drive) but "fdisk" then "p" shows the partition to be "GNU HURD or SysV" when booting from Tom's boot floppy.

Can anyone offer any advice as to how I can mount the hard drive and clear the root password?
 
Curious... unless there is some software and/or data on the machine that you're trying to access or save, wouldn't it be easier just to start fresh with a new installation?
 
Curious... unless there is some software and/or data on the machine that you're trying to access or save, wouldn't it be easier just to start fresh with a new installation?
There is.
Liebo: As it has some old data on, I've promised that I will try and back up the data if I can and give work a copy then wipe the hard drive before I use it.

Besides, he may enjoy the challenge. :)

I think you're on the right track. How about trying another boot floppy - say FreeBSD?
 
You'll probably need a larger bootable distro in order to read sys-v, Rootboot is really streamlined and I'm not sure it has support for that file system. You're going about it the right way though. I'm assuming you're familiar enough with mounting a device? mount -t hsfs /dev/whateveridandpartition /mnt/somewhere

Then you'd clear the password to nothing in /etc/shadow or wherever it is in SCO (or just in /etc/passwd if it's older and unencrypted passwords).

With systems that I was unable to get a working bootable and smart solution, I went ahead and just took a password hash from a system that I did know the password for, and overwrite the same length value/hash on the drive using a hex editor. Basically you can search the drive for the text "root:" and you'll end up finding the root:encryptedhash. Either way you should be able to overwrite that data as long as it's the same length of text without a problem. I always tend to back the device up first though if I have a system with enough space (use dd if=/dev/sourcedevice of=path_and_file_you_want_to_back_it_up_to) (ex. dd if=/dev/sda1 of=/tmp/drive_backup.dsk)

I'm not as familiar with SCO specifically but also if the system wasn't secured against it, there's often a -s argument you can give it at the boot loader/options to boot into single user mode which again if not secured/older would dump you right into a prompt without asking you for a password.
 
I think your best bet is to remove the hard drive + controller and try to get them working in a system that already has another hard disk. Install a year-appropriate FreeBSD (like 3.x or 4.x) or Linux distro like Slackware from circa 1997 and then try to mount the drive. Newer kernels in newer distros certainly have removed support for HTFS (what SCO uses IIRC) but the older ones should have the support there, or selectable when you rebuild the kernel and reboot.

Or, instead of disturbing the system by removing components, try to add an IDE hard disk and install the distro to that.

While the steps for doing this are relatively simple and foolproof, the (very) difficult part will be researching and finding an old Linux or FreeBSD distribution that supports HTFS. Good luck.
 
Found info on Expert Exchange and shamelessly cut and pasted from a query about accessing an Openserver drive via Ubuntu 6 something but would be applicable to any *nix I guess-

"Listed file systems for SCO Unix OpenSystem V are

File systems S51K, AFS, EAFS, HTFS, DTFS

and that the following applies:

Installation Must boot from S51K, AFS, EAFS file system types, which restricts the partition size to 2 GB maximum. HTFS and DTFS file systems support up to 1 TB in size, but cannot used to boot SCO Open Server."


Info could be usefull for future reference.


"To mount SCO file system you have to use the kernel with these enabled
- System V/Xenix/V7/Coherent file system support
- UFS file system support

You need to find out which version of SCO you have in the disk. You can plug the disk
to an old PC and boot up to single user mode, then run "divvy" command to find out.

HTFS and DTFS only available start from SCO 5.x, the older version of SCO are likely
use AFS or EAFS.

Also it appears the drive is split into slices-

"One has to remember that SCO in OpenServer5 further divides the fdisk partition into divisions where the filesystems would be created. From above (poster is refering to a basic install) it appears that the drive is a boot drive so the at a minimum you would have the following slices:

boot slice 0 fstype = EAFS
swap slice 1 fstype = Not-Used
root slice 2 fstype = HTFS
- slice 3 if used fstype = HTFS
- slice 4
- slice 5
recover slice 6 fstype = Not-used
wholedisk slice 7 whole disk so fstype is N/A

It is possible that slice 3 could be labled "u" if the whoever set it up specified a separate /u filesystem at install time. Of course slice 4 & 5 could have been set up as well. You would need some sort of docs from the original setup to know how the disk was configured. Another alternative would be if you have emergency boot & root floppies for this drive hook the hard disk up as a boot drive but boot from the emergency floppies then run the divvy command to see the slice table. If that works you may be able to mount the filesystems and retrieve your data to tape. You will have to check your ubuntu docs (or your choice of Linux) to see if it even supports the SCO Unix methods of drive slicing. If it does then there should be a unique device node for each of the slices mentioned above.

Another option is if you have the original OpenServer5 install media you may want to take a look at SCO TA #105094 and TA #105312. These TA's tell how to use the boot media as emergency recovery media."
 
Last edited:
I found this problem interesting so I went looking as well. As for old Linux distros, there's lots:
http://www.oldlinux.org/
http://www.oldlinux.org/Linux.old/
http://www.debian.org/distrib/archive

But when it comes to finding HTFS support, it looks like FreeBSD is a no go. I also found lots of dead end Linux threads where others had asked.

However, I did come up with two hopeful links. There's an "htfs" on Sourceforge, and there's a download on this page: Re: HTFS (SCO OpenServer 5) filesystem support

Anyway, I just wanted to look around to see what I could learn about this.
 
One would think there would be some live-cd specializing on forensics or data recovery of file systems. I'm not fresh in the loop anymore on that stuff, Backtrack was one I recall some folks working on, well ok I remember meeting a few different authors of different suites but not sure who's been playing with the most file systems being supported. Certainly seems like a desirable feature set. NetBSD loves it's claim to fame of being installable on the most architectures/systems, maybe it would have support on a live-cd?

I did find a list of bootable forensic type distros and google thinks one of them mentioned htfs however the current page doesn't reflect that though. I'll poke around, it's an interesting problem that I've also had to go through before hence the semi-stupid but working practice of replacing the text using an editor. That's what I ended up having to do on an older VMS system when I failed to find a file system driver.
 
But when it comes to finding HTFS support, it looks like FreeBSD is a no go. I also found lots of dead end Linux threads where others had asked.
Similar result apart from the above. The guy asking on Expert Exchange ended up just booting into single user mode and transfering the files, after splitting them, to floppy disks then reassembling them on a windows box.

I also found it quite enoying that google would automaticly put in RHEL when I was searching for anything related to earlier versions of RH Linux. Might see what Bing comes up with.

Better. Seems HTFS is also referred to as crosstor and the seems there is/was a commercial driver available that could read/write to these partitions under that name.
 
Last edited:
There is.


Besides, he may enjoy the challenge. :)

I think you're on the right track. How about trying another boot floppy - say FreeBSD?

Ah... not sure how I missed that!

I'll have to check what version of Linux we're running on an old-old server at work (I think it serves the GM's brother's personal webpage, as it supports Frontpage extensions). It's running on a P166 with 64mb ram and uses HTFS partitions. I should even have some old install CDs for it... unless someone's tossed them over the years. At this point, the only reason the server's still running is because it's been online all but 8 days out of the last 11 years and we're all curious to see how long the HW will keep going under that load (it was old/used hardware at the time it was setup, literally, as a test box!)
 
Hmm...these older servers\desktops converted to servers are unkillable. I have a Pentium 200 Dell that was used as an Active Directory server from 1997-June of this year. Still runs great with OS/2 Warp 4.

This thread makes me want to install Unix on my custom 486...the Compaq is reserved for Windows 3.1.
 
Ah... not sure how I missed that!

I'll have to check what version of Linux we're running on an old-old server at work (I think it serves the GM's brother's personal webpage, as it supports Frontpage extensions). It's running on a P166 with 64mb ram and uses HTFS partitions. . . .

From what I've read HTFS is exclusive to SCO.
Wikipedia: High Throughput File System (HTFS) is the journaling file system used by current versions of SCO OpenServer. It is the successor of EAFS.

The problem with SCO is that they are extremely litigious. In fact that's how they spend (not make) their money. Not to say it isn't possible, but it would be quite dangerous to write software concerning HTFS. This is why there isn't the support for HTFS we all expect to see in Linux.
 
The Openserver ftp supplement page is readily available ftp://ftp.sco.com/pub/openserver5/

Various ways to gain control from another source- http://www.unix.com/unix-dummies-questions-answers/20345-fortgot-root-password-sco-5-0-a.html Seems SCO boot and root floppies are a must have ideally or original install media.

The earlier EAFS file system seemed be able to be accessible in earlier Linux kernals via the sco_fs patch.

The link to htfs_fs-0.01alpha.tar.bz2 seems to be long dead Ole Juul. It must be archived somewhere surely.
 
Last edited:
Thanks for all yours posts guys, I didn't expect such a response. I'm definitely going to re-read them on my pc and investigate the suggestions :)
 
And goodness me not a library in sight ;) Good luck and don't forget to document your success or otherwise.
 
Last edited:
Using a new LAN with RJ-45 connection with Tom's Boot Floppy, and a version of Puppy Live CD on another, I am currently (I believe) making a copy of the entire partition in "img" format on to another computer.

This was done using a combination of commands sent to me by Tom himself (such a nice guy but I don't to take the micky by constant bombardment of questions!) something to the affect of :

Code:
nc -l -p 12345 > rawhtfs.img

on the "new" computer, then

Code:
dd if=/dev/sda1 | nc 192.168.1.5 12345

on the 486 server, finally

Code:
mount -o loop -t htfs rawhtfs.img /mtn/sda2

on the "new" computer

the file is growing, albeit slowly, but still cannot mount it in Puppy.

I assume that I could "restore" the img to another hard drive in newer hardware to easier detective work?
 
The instructions as far as I can tell are indicating you should boot off your installation media, when the installation media loads it will have that same "boot: " prompt (that's your bootloader checking if you want any custom options or drivers) and there they're saying "tools" to boot to the tools configuration of the media. Wow, yeah unfortunately those instructions are sorta hit and miss depending on the version of SCO installed.

I see your older videos, do you have one of it booting off your newly found media? Sounds like folks were breaking out of the installation or tools script with ctrl+a during any of the prompts where it's waiting for you to reply to a configuration option (like keyboard type).
 
Back
Top