Best of Technical Support

by Various
Best of Technical Support

Modem Works with Old Kernel, Not New One

After an upgrade from Red Hat 7.1 to 7.2, my modem no longer works when I boot into the SMP kernel. It works fine when booting the non-SMP kernel. I have downloaded and compiled the latest stable kernel, and the problem still exists. Under 7.1 the same modem worked fine with the SMP kernel. I have a US Robotics 56K FaxModem (model 3CP5610A) with a Tyan motherboard and dual-Intel Pentium 133.

—Nathan Myers,

Check to see if the modem is detected by the kernel. You can see this with a grep ttyS /var/log/messages*. You will see a list of serial devices; note that some will be built-in on your motherboard.

—Christopher Wingert,

Thank you for including the model number of your modem; it is most useful. A quick search on Google shows that you have a PCI modem, which works slightly differently from old-style ISA modems. The good news is that it is a real modem and not a Winmodem. The best link I found for you is this one: USR also seems to provide an example script for Red Hat here:

—Marc Merlin,

NFS Living in the Past

I have one machine (Red Hat 7.2) that serves the user home directories, which is now an ext3 fs, to several other peripheral machines (all Red Hat 6.2). I've noticed that files updated on the peripheral machines don't get updated on the server, and the changes don't get reflected on the other peripheral machines.

The server exports options (rw, no_root_squash). The client invokes with defaults, nodev, rw. This seems to indicate that the clients are caching, but it never seems to flush (I have a file that was changed a day ago that is still unchanged on the server). I've searched the Web and have found nothing that would help.

—R. K. Owen,

Nothing in NFS should cache a file for a day. I would check the obvious and make extra sure that the clients are indeed writing in an NFS-mounted directory and accessing the NFS server. You also can check that when you modify a file on the server, the clients are seeing the new copy.

—Marc Merlin,

Can StarOffice Import EPS?

My system consists of Red Hat 7.1 with a 933MHz Pentium III. The problem, which occurs with both StarOffice 5.2 and 6.0 beta, is that encapsulated PostScript graphics files (.ps or .eps files) are read as text, not as graphics. This happens with any selection (e.g., Text Document, Presentation, Drawing or Chart). gv shows the correct graphics. So the question is: does StarOffice have the capability to read graphics files in PostScript format and display the graphics rather than the PostScript text? If so, how do you do it?

—John C. Burgess,

The menu item Insert-->Graphics-->From File does understand .eps files.

—Scott Maxwell,

PCI Modem Not Recognized

I recently purchased a US Robotics 56K PCI modem card. It is not a Winmodem, which is why I purchased it. I planned on using it on my dual booting (Windows 98/Linux) system. Windows 98 listed it as device COM5, not the usual COM2.

I went into the Linux side and created a /dev/ttyS4, using setserial to set the port and irq. I made a symbolic link to /dev/modem. When I run minicom, it does not complain that the device isn't there, it just does not seem to do anything. How do I get Linux to recognize my modem? I tried an echo ATH1>/dev/ttyS4, and there is no dial tone in the modem speaker, so I am pretty sure the command is not making it to the modem.

—Tony Preston,

This sounds like the IRQ is not set correctly. You should check the IRQ with lspci -vv. Look for your modem in the list and use setserial to set the IRQ.

—Christopher Wingert,

See for an example of how to set the IRQ with setserial.

—Marc Merlin,

SCSI Error Message

I recently upgraded from a SCSI TR4 Travan tape drive to a 24GB DAT drive on a Linux server, and every morning I have the following message on the console after an overnight backup:

st0: Error with sense data:
[valid=0] Info fld=0x0, Current st09:00:
sense key U
nit Attention
Additional sense indicates Not ready to ready
transition (medium may have changed)

Can anyone say why this is so?


Check to make sure your SCSI bus is correctly terminated.

—Christopher Wingert,

You do not say whether the backup actually occurs before you get this message, or whether you are able to write anything on tape. If you haven't yet been able to write anything on tape, make sure that you do not have SCSI connection or termination issues. If your backup does occur, you may have some mt command after the backup that does something that isn't supported by the tape drive, or the tape drive may require some attention, like a cleaning.

—Marc Merlin,

Where's /dev/fd0?

I try to mount my floppy disk (1.44MB) on Mandrake 8.1 with the command mount /dev/fd0 /mnt/floppy, and it says “unknown device”. It is connected correctly and it works perfectly, but Mandrake doesn't recognize it.


Check to make sure that /dev/fd0 exists. Also, check /var/log/messages after running the mount to see if there is any clue to the problem.

—Christopher Wingert,

Make sure the kernel sees your floppy. You should have something like this in /var/log/dmesg:

Floppy drive(s): fd0 is 1.44M
FDC 0 is a National Semiconductor PC87306

Then see what dmesg says after you try your mount command.

—Marc Merlin,

When you say that “it is connected correctly and it works perfectly”, I assume you dual boot your system and the floppy drive works in the other OS environment. Here are a few things you can check: make sure that the /dev/fd0 file exists, that it is indeed a device-special file and that it has the correct permissions on it. The floppy you are trying to mount has a filesystem on it, and the filesystem support is present in the kernel.

—Usman Ansari,

I Have No /dev/printer and I Must Print

I am using Red Hat 7.1 and can't seem to find /dev/printer. I need this socket for a Perl script. I can change the name in the script, but can someone tell me what to change it to? I do not know the name of the socket that the lpd dæmon uses in Red Hat 7.1.

Scott Statland,

Not to be obstructionist, but are you sure you need direct access to the device file? If you're just trying to print, you can do that in Perl with

open(FH, '| lpr')  ||  die $!;

and print your desired text to the FH filehandle. If your script is intended for broad distribution, bear in mind that UNIX printing is very flexible; a given print queue may be serviced by a printer attached to some other machine on the network, for example, so the local machine may not have any file under /dev that represents that printer. You might want to find a Perl module that helps you parse /etc/printcap and look for a print queue's lp resource to learn what device file, if any, is associated with that printer. (Type man printcap for a thorough description of that file.)

—Scott Maxwell,

The device nodes /dev/lp[0-2] give you direct access to your parallel printer (bypassing lpd altogether).

—Marc Merlin,

If you want to print using lpd, there's a Perl Net::Printer module downloadable from CPAN that lets you print to lpd and check the status of your print jobs from a Perl script. Read the man page on-line here:

—Don Marti,

Load Disqus comments