1.0.4349 completely locks out debian laptop

[andyb@b5]

setup

Dell Inspiron 8200
GeForce 440 Go (64Mb) bios vers 04.17.00.22.f6
Intel P4 1.7G mobility.
I845 based mobo.
kernel 2.4.20 (tried with 2.4.19 with no differences)
Debian Sarge

No problems in 3123, installation of 4349 goes without any issues, compiling and installing nicely.

Problem
Starting X locks the machine down completely. All services stop, low level stuiff like icmp pings no longer respond. Has to be buttoned to restart.

I have tried with agpgart/nvagp/none and no difference. The lockdown is *very* early in the cycle, possibly close after scanpci finishes (see log).

I can run the server in probeonly mode but when I actually try to use it, everything dies horribly, After several attempts the X logfile of the crashing process does look roughly like it is shown.

I've now absolutely no idea on how to proceed and any advice would be very welcome

Regards
Andy


Contained in the zip
nvidia-installer.log
proc contents for working 3123
proc contents for failing 4349
XFree86.0.log in probeonly mode
XFree86.0.log in genuine run attempt.

[Dreoth]

If you want to regain control of your system, try hitting
Alt+SysRq(Print Screen)+E
Right before X starts in the startup script...

Hopefully it is enabled in your kernel for that key to work, if not, try using a woody cd and rescue.

[Andy Mecham]

Are there any messages in /var/log/messages that might indicate what's going wrong?

--andy

[andyb@b5]

Dreoth, there is a much easier way than that

when you boot your machine hit shift to get the lilo prompt and then do linux 1 (or whatever label you give your boot kernel) This takes you to single user mode where you can easily de-install the drivers.

Andy

There are no messages at all, I don't think it is because there are none but because the system comes down hard not allowing it to write out any log data. I ran a couple of processes with another machine to show the time and as far as I can tell part way through the X server initialisation it crashes big time. I'll think about ways to capture X output.

At the moment I'm stumped as it is really is a showstopper type of crash.

Regards
Andy

[Zedd]

Same problem here, but on a Gentoo system. Only thing I have to add is that the older drivers get me into X fine, but with no GLX support.

(yes, glx is uncommented, etc ;))

[andyb@b5]

Tried to get better logging details out which isn't easy when the trashing happens so severely but managed to pipe the output through the serial port and capture it at the other end.

The only problem with this is that I don't know if it captured it all.......

I've attached the 3123 and the 4349 startup logs

I've checked my kernel build and the agpgart support is permamently in the kernel, not a module, is this potentially an issue?

Regards
Andy

[andyb@b5]

answered my own question abount the agpgart as a module......

makes no difference at all, went through nvagp as 0 / 1 /2 /3 make sure agpgart not loaded at all for appropriate bits, no different. still dies a horrible death :-(

have to hope the logs hold a clue...

Regards
Andy

[andyb@b5]

I looked through a lot other recent posts, enabled the splash screen and found I had the same problem as some others had.

A pretty consistent element seemed to be SMP support in the kernel.

Went back and recompiled both the 2.4.19 and 2.4.20 kernels I had built as non SMP and both worked fine.

SMP fails, no SMP works.

This is the *only* difference between the two.

I suggest that if you have SMP and don't need it to try and take it out and see if you get better results.

Regards
Andy

[bwkaz]

Well... you are using an Inspiron. They don't even come in SMP configurations... right? I've never seen a dual-processor laptop...

This is probably the SMP+DVI bug, again; at least that's my guess. If you turn off your flatpanel and use an external analog connector (if you've got a spare monitor to test with, that is...), does it work?

[andyb@b5]

Found a monitor and sure enough it worked....

With the lcd screen being fully enclosed I wasn't even thinking dvi :-/ oh well live and learn.

As for the SMP, in the past its never hurt to have smp active and I've other dual proc machines. I tend to do one kernel build and have lots of modules defined. This time its turned around and bit me *sighs*

[Andy Mecham]

Sorry for this - it will be fixed in the next release.

--andy