Hello,

I'm running into an annoying problem since I've bought a DELL XPS laptop two months ago and installed Linux on it (previously installed on an Acer 7720 and worked like a jewel). From time to time (usually within 30 minutes after booting), it'll freeze and display nothing but a black screen, with a small blinking 1-pixel width vertical line along the left border of the screen (see made-up picture : http://dark.magus.free.fr/example.png). I've been struggling for weeks without finding any solution. I'm not new to Linux, but I'm probably far from the knowledge of long-time Linux users.

I did not find anything useful in the system logs (at the time of the freeze, nothing special appears in the logs).

I tested it with the following distributions, resulting in the exact same problem :
- OpenSUSE 11.4 (Gnome) with Kernel 2.6.39
- ArchLinux Live CD Gnome with Kernel 2.6.37
- LinuxMint Debian Edition (Gnome) with Kernel 2.6.39
- LinuxMint Debian Edition (Xfce) with Kernel 2.6.39
- LinuxMint Debian Edition (Gnome) with Kernel 3.0.0

So, I guess this is not a distribution-specific problem. I can't tell if it's a kernel problem however, since the bug happens with different versions of the kernel.

Here are the things I checked :

* Memory : I ran memtest, it does not report any error, and besides, Windows does run fine.

* Hard disk : Live distributions also have the bug, and as they don't rely on the hard disk, I don't think the problem lies here.

* Intel SpeedStep : disabled it, problem still occurs

* Updated CPU microcode : problem still occurs.

* ACPI : I thought it could be an ACPI bug. I've tested sleep and hibernate several times and they work fine (at least with Kernel 3.0.0). Also, disabling ACPI on the kernel command line, the bug also occurs (though instead of a line on the left of the screen, there are random dots moving all over the screen). So it's probably not an ACPI-related issue.

* Graphics card : This laptop does have the so called Optimus technology. For testing purpose, I was able to get it working with Bumblebee. Anyways, I don't really need to use the Nvidia card under Linux. I tried blacklisting the nouveau module, but the bug still happens. Also tried blacklisting the i915 module : bug still happens (but in this case, the result is the same as when ACPI is disabled : instead of a line, it's dots all over the screen).

* X.Org : when X.Org is not started, the problem does not seem to occur. When it's started, the problem can occur as soon as the login screen is displayed (tested gdm and xdm, the bug occurs for both). Also, tried alternate window managers such as IceWM : bug still happens.

I'm running out of ideas. If only I could identify the piece of software causing the problem, or getting some kind of output in a log... But I don't know where to search. Any guess ?

Thanks in advance.

P.S : sorry if my english is not perfect, this is not my primary language

First, your English is great! Now about your problem. Since you have tried ArchLinux, OpenSUSE, and LinuxMint with different kernel versions, and Windows works fine, then it is most likely a problem with Xorg. Base on what you said, it sounds like the system has 2 graphics cards. If that's the case, then maybe you need special firmware to manage them, in addition to the graphics card drivers.

Greetings,

Have you tried installing the Nvidia driver? Maybe that will handle things better.

Also, have you looked in the bios on the laptop to see if anything looks like it might help there?

Thanks for the posting and thanks all.

Be real, be sober.

Hi,
Have you tried switching the graphics card to discrete mode in the bios. I had issues with Bumblebee and optimus, but once I switched it, it works fine. Also, check whether the option in bios - Allow the pc to check whether the OS is compatible with optimus - (or something like that, I don't remember the exact words) is turned off as well.

Hello,

Thanks for your replies.

To clarify, yes, the system has 2 graphics "cards", one is the Intel IGP integrated in the processor, the other one is an Nvidia GT425M, which can be optionally used when needed. That kind of mechanism isn't well supported on Linux yet, but this is not really my problem, since the system should run flawlessly even without using the Nvidia card at all. I've indeed read many posts about "how to make the Nvidia card work", but none mentionning problems with just using the Intel IGP (which is all I need).

@PhoenixAndThor : from what I understood, on my laptop, the Optimus technology is purely a software implementation (except for the internal "wireing" part) ; I did not hear about a firmware for this. Do you have any information about this ?

@WSmart : well, as I said, I don't really need the Nvidia card, but yeah, in the past, I did try the official Nvidia driver together with Bumblebee (experimental Optimus implementation) and it did work fine. Except for my current problem, which is that the laptom randomly freezes.

I did a bit more investigations, I tryed to remote-debug X.Org with SSH + GDB, but when the crash occurs, I'm getting disconnected, and in fact, the network does not even reply on the computer. So it's not just X.Org crashing, it's the whole computer freezing. Also tryed to use Netconsole to get some kind of output from the Kernel just before it crashes but... no output at all at the time of the crash. As it it were an hardware fault rather than a software issue... Although it's working fine under Windows (I let it ran for hours and not a single crash).

I was merely proposing the potential existence of a special firmware, as Linux does not (yet) have a reliable method for switching between different graphics cards, especially in a laptop or other mobile device. Maybe the two cards are in conflict with each other, perhaps they are both trying to use the same channels in memory or some other low-level stuff is happening. I can't help you in that area, so I'm out. I wish you the best of luck.

Perhaps you can login to your host according to this

http://="http://www.vanemery.com/Lin...-console.html"

@PhoenixAndThor : well, thank you anyway for taking time to reply.

@amilo : I thought about that, but would that work with a serial-to-USB adapter ? (my laptop does not have a native serial port)
I did something similar with SSH over the network in order to track down what happens, but it's probably less efficient than a serial console for that kind of job.