Red Hat Linux 7.1 Install on Dell Inspiron 8000

This document describes how I installed Red Hat Linux 7.1 on a Dell Inspiron 8000 laptop with an NVIDIA graphics card.
The order that you do things in does not really matter. My laptop came with Windows ME on it. I left ME there and installed RedHat 7.1, using Partition Magic to make room for it. Later I created the suspend to disk partition, and later still I replaced ME with Windows 2000. It is not critical that any one thing be done before another.
  Installing Red Hat 7.1


  • Create a suspend to disk floppy
    Download rh7s2d.exe from Dell's support site (just search for "rh7s2d.exe"). Unzip it (by executing it) and follow the directions to create the floppy.

  • Create a Windows ME boot disk
    In Windows ME, go to ControlPanel, Add/Remove Programs, Startup Disk.

  • Create a Partition Magic floppy
    I used Partition Magic 4.0. Don't install Partition Magic. Run it from CD and choose the "make rescue disk" option. Ignore errors about not working properly on Windows ME. While making the rescue floppy, Partition Magic will mess up when trying to write system files to the disk. Ignore this. The disk will not be bootable, but you can boot from the Windows ME boot disk and then run Partition Magic from this disk.

Partition the Disk

The laptop came with a small (290MB) "free" partition at the front of the disk and the rest of the disk allocated to Windows ME. The small partition is for suspend to disk use.

  • Create the suspend to disk partition
    Boot with the Windows ME boot disk, at the menu select Minimal Boot (option 4). At the DOS prompt A:> insert the suspend to disk floppy and run mks2d. (Note that there is another program called phdisk that should be used for other Inspiron models. The Dell knowledge base states that mks2d should be used for the Inspiron 8000.) (Note also that you should use rms2d to remove a previous suspend to disk partition if you've already created one and want to redo it.)

  • Make room for linux
    While still at the A:> prompt, insert the Partition Magic floppy and run pqmagic. Now you'll see that the small free partition has been mostly taken up by a Type 84 partition (275MB). The remaining space is still free. I later installed Windows 2000 into the Windows ME partition, and it gobbled up the free space. You could use Partition Magic to add it to Windows now. In any case, shrink the Windows partition from its current 17GB to about 4GB, leaving the freed space unallocated.

Install Linux

  • Allow booting from CD
    Enter the BIOS configuration by pressing Fn+F1 during boot. Make the CD boot before the hard drive.

  • Install linux from CD
    Boot from the first Red Hat 7.1 CD. Do the default graphical install (just press enter at the first menu). I did a "Laptop" install with manual partitioning. Leave the suspend to disk partition and the Windows partition alone. Give the free space to linux. I made a 4GB partition for / and a ~10GB (the rest of the disk) for /home. I told it to mount the Windows partition (listed as FAT32) on /win, which worked fine. Do not have it boot to X, it won't work until the NVIDIA driver is installed. Contrary to earlier versions of Red Hat, PCMCIA works fine. Lilo is put on the master boot record and has an entry to boot Windows. We'll have to move lilo in the next step.

  • Fix lilo and suspend to disk
    In order for suspend to disk to work, the suspend to disk partition must be a primary partition, and the master boot record must not contain any boot loaders. So, boot with the WinME boot disk and run fdisk /mbr to clear lilo out of the master boot record (N.B.: do not use the fdisk on the suspend to disk floppy for this, use the fdisk on your Windows boot floppy). Edit /etc/lilo.conf to put lilo on /dev/hda2 (see my /etc/lilo.conf), rerun lilo. Use fdisk to make the linux partition that contains /boot (the 4GB / partition for me) active. Now test suspend to disk (from the console, X isn't working yet, and suspend to disk from X does not work with the current NVIDIA driver) using Fn+A. After suspending if you boot from a floppy you'll see that the suspend to disk partition is now active. If you boot from the hard drive it should restore from disk and then restore the linux partition's activeness.

  • Set hard drive to use DMA
    I followed instructions from Joel Wijngaarde. Test your hard drive speed using

    hdparm -T -t /dev/hda
    It will take a few seconds. If you get a low speed (I got 4 MB/s), the drive is not using DMA. To enable DMA add this line to /etc/rc.d/rc.local:
    /sbin/hdparm -d 1 /dev/hda
    With this fix I now get 17 MB/s.

    Apparently there's an easier way to enable DMA: uncomment the USE_DMA=1 line in /etc/sysconfig/harddisks.

Set up X

  • Install NVIDIA driver
    First you must install kernel-source-2.4.2-2.i386.rpm from the second Red Hat CD. Then go to the NVIDIA linux driver web site and download the latest NVIDIA_kernel-x.x-xxx.src.rpm and NVIDIA_GLX-x.x-xxx.i386.rpm. I originally installed 0.9-769, then upgraded to 1.0-1251, which was not yet on the web page but was on the NVIDIA linux driver ftp site. The new driver fixed several problems but not all of them (and it added an NVIDIA splash screen...I've seen instructions elsewhere on how to get rid of it but it's not that annoying). Install the packages like this:

    rpm --rebuild NVIDIA_kernel-1.0-1251.src.rpm
    rpm -ivh /usr/src/redhat/RPMS/i386/NVIDIA_kernel-1.0-1251.i386.rpm
    rpm -ivh NVIDIA_GLX-1.0-1251.i386.rpm

  • Configure X
    I got an XF86Config-4 file for a 1600x1200 screen from Dell's web site by searching for "XF86Config" at Dell's support site. (I downloaded atimp316.rpm, which wants to install into /dell and directly to /etc/X11/XF86Config-4. You can use something like:

    rpm -i --relocate /dell=/tmp/dell --relocate /etc=/tmp/etc --badreloc
    to redirect the installation if you just want to inspect the files.) I modified the XF86Config-4 file as follows. In the Device section, make sure you have:
            Driver "nvidia"
    (I copied the Device section from the XF86Config file that came with the NVIDIA driver). In the Module section, make sure you have:
            Load   "glx"
    And remove these lines:
            Load  "dri"
            Load  "GLcore"
    In order to use other resolutions than 1600x1200, I had to add this to the Screen section (this tells the driver to ignore what the screen says its limitations are):
    	Option "ignoreEDID" "1"

    That's all I had to do to set up the 0.9-769 driver. For the 1.0-1251 driver you should take a look at /usr/share/doc/NVIDIA*/LAPTOP_README, which instructs you to add the line:

            options NVdriver NVreg_Mobile=1
    to the end of your /etc/modules.conf file. Without this line the driver will fail to initialize.

  • Problems with X
    NVIDIA is supposedly working on these problems, we have to wait for new drivers.

    • With the 0.9-769 driver I could only get 16BPP to work, and only at resolutions 1600x1200, 1280x1024, and smaller resolutions. With the 1.0-1251 driver I can run 24BPP or 32BPP but can only get 1600x1200 and 800x600 to work.
    • With the 0.9-679 driver, when switching to a virtual console from X the screen is blank. Usually bringing up a BIOS screen, for example Fn+F3, will show the BIOS screen and then escaping from it will reveal the virtual console (it may be dim though). The 1.0-1251 driver fixes this problem.
    • With the 0.9-769 driver, shutting down X produces the same problem as with the virtual console. The 1.0-1251 driver allows shutting down X just fine. However, the documentation explicitly says that restarting X will cause the machine to hang.
    • Don't suspend to disk while running X. After resuming from disk the machine will hang (it's not just blank like the virtual console problem). I couldn't suspend to RAM or to disk and resume properly with the 0.9-769 driver. With the 1.0-1251 driver I can sometimes switch to a virtual console while running X, suspend to RAM or disk, and resume properly, but not always and I haven't figured out exactly when it works. After successfully resuming the hard drive light is always on.
    • Using an external monitor does not work when in X, even when in 800x600 resolution. It works fine from a virtual console though. When in X the external monitor just shows colored bars.

  • Using fvwm2 instead of Gnome or KDE
    If you don't like Gnome or KDE and want to use fvwm2, you need to install fvwm2-2.2.4-9.i386.rpm, fvwm2-icons-2.2.4-9.i386.rpm, and wmconfig-0.9.10-3.i386.rpm from the second Red Hat CD. Copy /etc/X11/fvwm2/system.fvwm2rc to ~/.fvwm2rc.m4, then edit it as you wish. Change /etc/sysconfig/desktop to have the single line

    If you've installed KDE, you can add all the KDE menus to your fvwm menu with this Perl script.


  • Sound, ethernet, modem
    sndconfig identified the ESS Maestro 3i card and sound works fine. I also had no problems with the ethernet card. The modem is a WinModem for which there is no Linux driver (according to Rob Clark's site).

  • CD Writer
    The CD writer replaces the floppy drive. You must reboot after switching between them. You can read from the CD writer by mounting /mnt/cdrom1. To be able to record with the CD, add this line to the linux boot entry in /etc/lilo.conf:

    Now reboot (the ide-scsi module is automatically loaded by /etc/rc.d/rc.sysinit). I've been using xcdroast (as root, I had problems making it user-runnable), and it works great.

  • Toshiba PDR-M5 digital camera
    To use the usb connection with this camera, use the usb-storage driver (run insmod usb-storage). Now create an entry in /etc/fstab so you can mount the device:

    /dev/sda1               /mnt/usbcam             vfat    users,noauto,ro 0 0
    Now you can simply mount and umount it (of course you have to create the /mnt/usbcam directory first). It is read-only, so you can't delete images from linux (the camera manual recommends only deleting with the camera anyway).

  Installing Windows 2000

  Miscellaneous Information