RedHat Linux 7.2 on the Dell Inspiron 8100

(Update: Also Fedora Core 2!)

Henry Throop / SWRI



I bought this system to replace an Inspiron 5000e. Although the styling's improved marginally, both the 5000e and 8100 are big, heavy sytems in flimsy, cheap plastic cases. My main requirement was a 15" SXGA+ display. I first went with a Compaq 2700T, which was nicely made and worked almost perfectly under Linux. Unfortunately, APM was simply not supported in its BIOS, so back it went to Compaq. I then went with the i8100, in the following configuration:

Including a 3-year onsite warranty, it was just under $1400. Both the display and keyboard were replaced in the first year of the warranty on my 5000e, and I fully expect this one to fail too. For any number of reasons, I'll probably buy a Powerbook next time.

Update: 2-Apr-2002 I looked around for carrying cases for some time. I found a foam-padded sleeve made by Eagle Creek that fits perfectly, and then lets you stash the laptop in any old backpack. It's $25 from REI or EMS, called something like 'Computer Vault.'

Installing Linux

I installed dual-boot with Win2K should I ever need it. The process was pretty straightforward: First install Win2K, then use Partition Magic to shrink the NTFS partition and move it forward to allow room for /boot. Boot RH7.2 off CD, and set up the Linux partitions. ext3 works fine, and is apparently more reliable if the machine was shut down without unmounting all the disks (which does occasionally happen). The only disadvantage to using ext3 is that the useful noflushd package doesn't work. (noflushd buffers writes to the disk and allows it to spin down, meaning the drive isn't rattling away every few seconds when the system's completely idle.)


The RH7.2 graphical installer recognized the video card, but didn't configure it properly. This was expected, based on experiences of others on the Linux on laptops page. So, I set it up to boot into text mode at first, and then set up the video later. I went to the nVidia Linux Drivers page, and downloaded the 'RedHat 7.2 UP Kernel' and 'GLX RPM' files. For some reason, the former of these didn't install properly, giving me a large number of 'undefined symbol' references. (Strange, since I'm using a stock 7.2 install, and this is supposed to work.) I got the 'Kernel Source RPM' instead from the same page, and built it (using rpm --build, rather than rpm -ivh). Easy enough. Installed, and works fine. You must change the Driver "nv" entry in XF86Config-4 to Driver "nvidia". Here's my XF86Config-4.

The display is very nice, although the backlighting is noticeably non-uniform (top-bottom gradient, and some inch-sized hot-spots around the edges), and as a result it is not overall quite as good as that in the 5000e.

Update 20-Feb-2002: I called Dell and complained about the backlighting. A service rep came out and replaced the screen the next day. This fixed the blotchiness issue (perhaps the display panel was getting pinched around the edges?), though the top-bottom gradient remains. Still, it's much better than it was, and almost as good as the 5000e's. On Dell's support fora, there was discussion of the screen problems, and many people have reported it. Some said that it was due to IBM panels, and Dell had discontinued IBM in favor of Samsung. For what it's worth, the service rep told me he replaced my Hitachi with an IBM. Works for me.

Update 2-Apr-2002: Sure enough, display #2 just quit on me too. It had an entire column of pixels near the left hand edge which displayed always blue. Dell came out and gave me a new video controller and display, which fixed the problem. The new display has a definite pinkish tinge to it all over, but at this point I'm not going to worry about it too much. It is a Hitachi like the last one, and has the same top-bottom backlight gradient annoyance.

Update 16-Apr-2002: For whatever it's worth, I've now got a few red pixels that are always on, and a few black ones that are always off. Oh well.

Update 6-Aug-2002: I was having some graphics problems which turned out to be due to a different issue, but in the process of debugging things I installed the latest nVidia driver (using the same method as above -- source RPM). After doing this, the system was completely broken -- I couldn't even login in text mode as anyone other than root. This turned out to be a problem with the file /etc/security/console.perms. A few newsgroup searches led me to this thread ('Nvidia permissions', alt.os.linux.mandrake, 13-Oct-2001), which solved the problem.


At first, APM didn't work -- if I tried to suspend (apm -s) then I got a message saying that I could not do so. The official documentation from nVidia states that laptop suspension is not supported by this video card. What they apparently mean to say is that it is not supported if AGP is active (see message 6912 on linux-dell-laptops). AGP is some sort of graphics acceleration thingy; I'm a slow graphics type myself, so disabling it was not an issue for me. What must be done is: Suspend/resume now works. Occasionally it'll come up with a blank screen; sometimes this can be fixed by doing a suspend/resume cycle again. Occasionally I'll have to pull the plug on it. However, something strange happens here: when it first boots, the BIOS won't recognize either the hard drive or CD at all, and gives a text error message to this extent. I've found that it literally takes about a 10 minute wait after powering down for the system to recognize the drives. Odd.

Apparently BIOS A08 has somewhat broken APM features. I have A07, and several people have mentioned needing to go back back to that version. I also understand that the newer versions have some problems with time drift (minutes per day).

Interestingly, the 8100 apparently has a small fixed battery or capacitor inside, which maintains the memory long enough such that the battery can be swapped out while the computer's suspended. Nice. Then again, seeing that this computer can fit two batteries in it anyhow, the feature may be of little actual value.

Update 29-Apr-2002: APM works fine. Here are three tips that I've figured out to make it reliable:

Follow these, and it seems to be close to 100% reliable. I haven't had to reboot in a couple of weeks.


Pointers usually work pretty well under Linux. For whatever reason, I had a heck of a time getting the internal trackpad/stick to work in X. It looked like some sort of interaction between gpm (which controls the text-mode mouse) and X, that caused the mouse to get stuck in the upper right hand corner of the screen. External mice worked fine. Eventually I fixed this by telling X to read from gpm, rather than the mouse port itself. In general:

This seems to work, but I have no idea why it needs it, or why I had problems. I've heard of no other reports of this problem with the 8100. I haven't been able to get a wheel mouse working, unfortunately (since gpm doesn't support them -- but there are apparently patched versions that do, which I've not tried).

Update 16-Apr-2002: In fact, although most people seem to get the mouse working OK, there are others out there who have to do the gpm thing. One correspondent reported to me that this problem also popped up when going to RH7.3 beta2.

Update 19-Jun-2002: I've been told that my method for activating gpm will work fine, but fail if the OS is updated (e.g., if the whole system is updated to RH7.3 via an RPM). I understand this can be fixed by putting the line OPTIONS=-R in the /etc/sysconfig/mouse file. This forces the -R flag to be passed, and doesn't rely on it being in the /etc/init.d/gpm file, which could be overwritten. (Thanks to Alan R.)

Update 29-Sep-2002: Several people have told me that using Option "Protocol" "PS/2" in the X config file works for them. Doesn't work for me, for whatever reason. The only significant disadvantage of using the gpm mouse is that gpm doesn't support mouse scroll wheels. I tried a USB Logitech optical wheel mouse recently, and it was wonderful. ( mknod /dev/usbmouse c 13 32 (?? -- not sure on the arguments), and then put /dev/usbmouse in the Xconfig file.) This worked beautifully -- until I resumed from a suspend, and nothing short of rebooting would wake up the mouse again. The mouse was clearly alive, and /var/log/messages said it was claimed by a driver, but it wasn't talking to X. I tried rmmod usbmouse; insmod usbmouse and although usbmouse (or something else) was clearly there and working (since it up on boot), then it refused to find the module. Oh well -- I'm back to the gpm solution. If anyone knows how to do this, do let me know.

Ethernet & Modem

Several messages on linux-dell-laptops indicated that this system is available with mini-PCI cards made by either ActionTec or 3Com. People had suggested the ActionTec because the Lucent winmodem driver works with it. I was given no option of which one to get, however, and ended up with the 3Com. The ethernet works fine. I understand that the standard winmodem driver does not work; however, I haven't tried it out yet. I bought a PCMCIA modem off ebay for $40 which works fine (3Com 3CXM556). I use KPPP; note that you must explicity turn off the ethernet port (ifconfig eth0 down) before starting up PPP.

I'd read some reports that the ActionTec card does not re-initialize the ethernet port properly after a suspend. The 3Com one works fine, for whatever that's worth.

Update 16-Apr-2002: Apparently SuSE 7.2 does not properly set up the Ethernet card. According to /etc/modules.conf, RH 7.2 correctly uses the 3c95x driver and telling SuSE to use this makes it work too. (Thanks to Milos M.)


Seems to work fine. I didn't have to set anything up. It appears to be using the emu10k1 driver.

The electronics inside don't seem to be very well isolated, however, so there's a lot of electrical noise -- twiddles and bleeps and buzzes and so forth -- that emanate from the speakers in sync with what the CPU is doing at the time. It provides a sometimes-interesting soundtrack; turning down the volume makes most of it go away.


Haven't tried it. Works under Win2K.

Other things -- file syncing

One of the goals of having a laptop was to be able to work from home and in my office and have essentially the same files available on both my laptop and my desktop. (Desktop is an older Dell running RH7.1.) For a long time, I accomplished this by just carrying the laptop back and forth to my office, and essentially using it as a portable fileserver. (I mounted it with NFS, and stored everything on the laptop.) This worked fine, but I didn't want to continue carrying it back and forth every day, just to use it as a fileserver.

I looked at various options like Jaz disks and portable hard drives, but these looked fragile or awkward. What I ended up with is a 256MB Compact Flash card, with a pair of USB card readers (one each on my desktop and laptop). The USB readers mount under Linux (mkdir /mnt/flash; mount /dev/sda1 /mnt/flash) and appear as regular filesystems. I set up a script with rsync that automatically copies any files in a subset of my directory structure that have changed to & from the memory card. This way I just take the card back & forth every day, but keep an up-to-date filesystem on both machines. It takes only a few seconds to sync each machine up. The USB reader wouldn't work reliably under the RH7.1 kernel on my desktop (2.4.2-2), so I updated to 2.4.18. The only small difficulty is that the card doesn't store the permissions properly, so my script has to update these afterwards. Other than that, works great. As an extra bonus, the same card can be used (without even having to reformat!) in my camera.

I have both a SanDisk reader (SDDR-3101) and a SimpleTech one (STI-UCF100). Both work fine. I originally used a Viking PCMCIA adapter (CF-ADAPT) for the Flash card on the laptop end, but I started having system hangs afterwards, which seemed to be related. No worries -- the USB works fine. Total for the card and two readers was around $150.

NB: I'll note that I occasionally have problems dumping the entire contents of my 256MB Crucial CF card -- after doing so, the reader will lock up and not mount again. It seems to work fine on my 256MB Viking card, and for small transfers from either. It seems to be more than random; I assume this is the sort of thing that will get slowly better with new kernel versions.

Update 15-Oct-2002: The SimpleTech card reader quit working recently (seemed to be a hardware problem). I sent it back and the manufacturer quickly sent me a new one. However, this one always gives me a `No media present!' error when mounting. Perhaps there's an easy solution, but not finding one I gave up and bought another SanDisk reader instead, which works fine (SDDR-75, CompactFlash + SmartMedia, $30 from CompUSA or Amazon). Another correspondent (Donovan R.) was also unable to get the same SimpleTech reader to work, so it appears the manufacturer is using a somewhat different controller than they used to, even though the packaging and model number remains the same. Conclusion: get the SanDisk reader.

Update 4-Aug-2003: The problems with the Crucial CF card were apparently due to a defective card, not a driver issue. I sent it back -- a year after I bought it -- and it was replaced for free. The new one works fine. The place seems to be run by good people.

Update 25-Sep-2004: After a long and ultimately unsuccessful attempt at upgrading the kernel to support Firewire, I blew everything away and installed Fedora Core 2 (FC2). Everything worked easily, except for suspend/resume under APM, which took a bit more work but was straightforward. Basically, you have to

(I put a button that runs this script on the KDE panel. I also made the script automatically re-dial the modem when I resume, if it was connected when I suspended -- RH7 did this automatically, but FC2 doesn't seem to.) After doing this, suspend/resume work fine. The mouse problem from RH7.2 is fixed now too. No other issues that there's much to say about.


An adept soul on the Portland, OR light rail made off with my 5000e (service tag 8FKY201). E-mail me if you've seen it.

Henry Throop, Southwest Research Institute / throop (at) boulder swri edu

Last modified 25-Sep-2004