Best of Technical Support

Various

Issue #93, January 2002

Our experts answer your technical questions.

I share two boxes, one Windows and one Linux, with one KVM. Both work when I start up the desktop, in this case GNOME/Enlightenment. When I switch from the Mandrake desktop to the other system, then back again, I lose mouse support entirely. I've checked cables, restarted the gpm dæmon and pressed Ctrl-Alt-Backspace to leave the desktop. When I restart the desktop, the mouse is detected again, but only for as long as I don't switch to Windows and then back again.

With the Windows side, I can switch to Linux and back with no problem, and every time I start the desktop on Linux it works—but only until I switch the screen away and back again. Any ideas as to what this needs to fix? Linux/Mandrake 7.1 is running on a Dell P90 (old) with PS/2 mouse, gpm runs with gpm -t ps/2. Could some other dæmon I'm not running because of security be what's causing the problem? I have amd, atd, innd, lpd and portmap disabled.

—Dave Dennis, dmd@speakeasy.org

This is most likely not a problem with the Linux setup. PS/2 mice have a configuration that is initialized during startup. A KVM is responsible for restoring this configuration on switch back to a machine; Linux is totally unaware of the switch.

—Christopher Wingert, cwingert@qualcomm.com

Didn't Run LILO

I recently upgraded my kernel and forgot to run LILO before rebooting. Now, if I start up from a different disk, that disk can be mounted and fsck shows it is clear. However, I can't boot from it.

—Willie Strickland, willie@istrick.com

Use the boot disk to boot from the hard drive with a command similar to this at the LILO prompt:

linux root=/dev/hda1

Then edit your lilo.conf and run LILO, as if you'd just installed the new kernel.

—Ben Ford, ben@kalifornia.com

I Have No “mail” and I Must Mail

I have networked Linux machines on my home network consisting of two desktop machines and a laptop. I would like to get my mail on any machine but can do so only on the older desktop machine. I have set up the Netscape preferences identically on all machines. On the newer machine and the laptop I get the message “Netscape unable to locate the server mail”, when I try to retrieve mail. The server “mail” is the name given by my ISP (Cox@home), which works perfectly fine on the older machine. On the other machines, when I try to get new messages, Netscape always asks for my password even though in the preferences I have explicitly selected the “remember password” button, so I'm wondering if NS is reading the wrong preferences file.

—Eric Smith, esmith289@home.com

Sounds like your one working machine has the Fully Qualified Domain Name (FQDN) for your ISP and the others do not. Try adding the rest of the hostname to the configuration for Netscape (i.e., mail.example.com, if example.com is your domain name). Alternately, you could update your /etc/resolv.conf “search” configuration line and add the correct domain name so that you don't have to type in all the time.

—Christopher Wingert, cwingert@qualcomm.com

To check that you have edited /etc/resolv.conf correctly, do a host mail from the shell to see what “mail” is resolving to.

—Don Marti, info@linuxjournal.com

audoupdate Can't Find Directory

I am trying to use autoupdate 3.1.5. When I type autoupdate, I get this error message:

CWD failed no such directory or file

When I run autoupdate --debug 2, it is able to log in as anonymous user, then it says:

CWD failed.
Error: Failed to check directory at ftp.redhat.com:
pub/redhat/linux/updates/7.1/en/os
no such file or directory.
I tried typing ftp ftp.redhat.com. I am able to change the directories to /pub/redhat/linux/updates/7.1/en/os as anonymous user.

—Adharsh Praveen R., adarsh@multitech.co.in

You should add a “/” in front of the directory passwd to autoupdate, i.e., /pub/redhat/linux/updates/7.1/en/os.

—Christopher Wingert, cwingert@qualcomm.com

Where's /dev/sdc?

I have two IDE devices, a 40GB WD HDD and a CD-RW drive. I also have a SIIG SCSI card. I believe that Linux is recognizing my SCSI card, but for some reason I cannot get access to my SCSI drives. When I try /dev/MAKEDEV sdc[0,1, ..., n], it tells me:

don't know how to make sdc[n].

My SCSI hdd is set to id 3. My SCSI CD-ROM is set to id 5 and my SCSI card is defaulted to id 7. Everything is terminated properly. What's the deal?

—Derrick Blackwell, db101055@hotmail.com

You seem to expect the HD with SCSI address 3 to be /dev/sdc. Linux doesn't work like that: the lowest-addressed SCSI HD is always /dev/sda, regardless of its SCSI address. The next-higher-addressed SCSI HD is always /dev/sdb, and so on. Your SCSI CD will be /dev/scd0 or /dev/sr0—both names work equally well.

—Scott Maxwell, maxwell@ScottMaxwell.org

What the fsck? The Superblock Could Not Be Read

When I e2fsck /dev/hda, I get the following message:

The superblock could not be read or does not describe
a correct ext2 filesystem. If the device is valid
and it really contains an ext2 filesystem (and not
swap or ufs or something else), then the superblock
is corrupt, and you might try running e2fsck with an
alternate superblock: e2fsck -b 8193 <device>

When I do as it says (<device> = dev/hda), I get the same message back. This is is the fourth time it has happened. The other times I just reloaded Red Hat, as I had not lost anything significant. This time I prefer to not lose programs I had working.

—Bob Wooden, bwooden@computelnet.com

/dev/hda is a drive. While it is possible to install a filesystem on a drive, this is most likely not the way it was installed. You should do a fdisk -l /dev/hda, which will show you all the partitions installed on the hda drive. Most likely, your partition is /dev/hda1 unless you are dual booting.

—Christopher Wingert, cwingert@qualcomm.com

/proc/kcore: 0wn3d!

I have two servers running Red Hat 7.1. I run Tripwire on both servers. Tripwire has twice reported that the file /proc/kcore has changed. Is this something that happens? No reboots between the occasions.

The file /lib/libc-2.2.2.so did change checksum, but nothing else was changed; dates, inode, etc., were the same. rpm -V glibc also revealed that the file MD5 checksum was altered. Since I am running a public web server on this machine, I reinstalled the glibc package. Then Tripwire complained about the dates and inodes; this is natural when you have reinstalled a package. I believe I have a well-setup firewall. How can the checksum of libc change? Why does the checksum of kcore change?

—Magnus Sundberg, Magnus.Sundberg@dican.se

I don't know why libc changed, but I know why /proc/kcore changed: /proc/kcore is a pseudo-file representing the system's physical memory. It's only natural that the contents of physical memory would vary with time. If it didn't, your system would be a lot less useful. So don't worry about that one.

—Scott Maxwell, maxwell@ScottMaxwell.org

If I Could Walk That Way, I Wouldn't Need Aftershave!

I am having trouble using two NICs in my Red Hat 7.2 (2.4.3-12) machines. Both NICs are recognized, and I can configure them. But I have two T1s from different providers, thus on different IP networks. I want to multihome these machines, but I cannot figure out how to add a default route for each NIC.

—Mike Kercher, mike@CamaroSS.net

You cannot add two default routes. You have to tell Linux which traffic should go to which network. Sounds like what you are trying to do is load balancing on the two T1s (or possibly redundancy). Check out the EQL or Bonding driver. You also could route via BGP. Check out the Advanced Routing HOWTO at www.linuxdoc.org/HOWTO/Adv-Routing-HOWTO.html.

—Christopher Wingert, cwingert@qualcomm.com

Yes, Sir, We Have Aftershave—Walk This Way Please

I have about 300 Linux servers and the floor space in the data center is at a premium. I need to access the servers as console for administration purposes, but they do not have monitors. Is there some way I could get a serial terminal connected as a console?

—Karthik, nkk@hotmail.com

Check out the Serial Console HOWTO at www.linuxdoc.org/HOWTO/Remote-Serial-Console-HOWTO.

—Christopher Wingert, cwingert@qualcomm.com

Backup Chokes

I use a DDS4 tape drive on a Compaq Alpha (Red Hat 7.1) to back up several filesystems on other Linux workstations (PCs with Red Hat 7.0 and 7.1). I've set up ssh so DSA authentication allows me to run a simple backup script at night via cron, without being prompted for a password. I use tar (gtar version 1.13.17) and dd (GNU fileutils 4.0x) to store the files on tape. Somewhere down the tar/dd combination I get the message:

select: Bad file descriptor

The backup script then stops tar/dd-ing and proceeds normally with the rest of the instructions. What am I missing?

—Martin Olivera, molivera@ucsd.edu

This error has shown up as a result of a bug in early versions of OpenSSH. The first thing to try is to upgrade ssh on all the systems to the latest stable version.

—Don Marti, info@linuxjournal.com