kernel update + udev killed console display

[ThurstonX]

Hi all.
I've just installed FC 4 (for the 3rd time) in order to isolate a problem with the console display. Briefly, running up2date or yum breaks the monitor. The display is fine through the boot, even looks like 1280x1024, but when it hits the xdm (gdm?) the screen is whacked. A green line crawls across the top left to right, slowly spreading. The middle is white and the bottom fifth or so is blue.

During the post-install/initial boot setup my video card (3Dfx Voodoo 4500) and monitor (Acer AL1715) are both recognized. All is fine after the install.

On my first two attempts to build this box I grabbed everything in up2date. Result: same busted monitor symptom. The latest kernel (kernel-2.6.15-1.1830_FC4) was not out during the first two installs, so I was grabbing the one prior. This time I updated yum to use livna and installed yumex and launched the latter, forgoing up2date. The latest kernel and kernel-devel rpms were listed.

This time around I only selected the kernel and kernel-devel rpms. The dependency check required udev-071-0.FC4.2, so what choice did I have. I ran the update, rebooted, whammo! same prob. Since booting to the original kernel doesn't help, I gotta finger udev as the culprit. Erasing the udev rpm screws the system royally (no X) as one might suspect, so I added udev* back using yum. reboot. Jacked-up monitor returns. Familiarity begins to breed contempt.

I can VNC into it fine, so again, all signs point to udev (vs. some xorg thing; went down that road on install #2). Some Linux rag article on udev states that /etc/dev will be repopulated dynamically on every boot. Since I have no /etc/dev, perhaps he meant /etc/udev. Regardless, I'm clueless how to proceed. Another install seems ... stupid. Hoping one of you smart, not-pissed-off folks can lend a hand.

In the meantime, that PC's got some BOiNC # crunching to do. In WinXP. It's a sad day. I hope I awake to a nugget of truth under my virtual pillow.

TIA.

PS - Tried searching bugzilla.redhat.com but that site is about as whacked as my console.

PPS - under Desktop | System Settings | Display the video card and monitor are correctly identified (this gleaned from a VNC session)

[weinerdogus]

Have a look here. It may be the same problem. http://www.fedoraforum.org/forum/showthread.php?t=94183

[ThurstonX]

Oh, Canada!
thx for standing on guard for me.
One nugget I gleaned from that other post is the CTRL+ALT+BACKSPACE trick. That got me the xdm/gdm login screen I was expecting. Vertical line interference goes away, login to Gnome is fine.

I'm a bit of a Linux noob, esp. when it comes to X. Too many years of non-GUI Solaris administration messin' with my head. So, if "normal" operation is:
boot
stuff loads
xdm/gdm appears

what is CTRL+ALT+BACKSPACE doing? Going "backward"? If so, backward from what/where?

I'm content with this work around and will wait for some update to fix it (or well and truly break the thing ;-). Glad someone managed to post it on bugzilla.

Thanks again.

[Firewing1]

it kills the X server and restarts it.
Firewing1

[Questor]

This problem seems to be related to Red Hat Bugzilla #179041:
https://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=179041

It may be that the publishing today of this update (http://www.redhat.com/archives/fedora-announce-list/2006-February/msg00032.html) will help with the problem.
(http://www.redhat.com/archives/fedora-announce-list/2006-February/msg00032.html).

It seems that for some reason, some updates with the new kernel are turning on
certain framebuffer drivers that ought not be turned on.

Please write back in this thread if you try this and let us know if this works for you.

-David

[ThurstonX]

Thanks, David. I found a thread (http://forums.fedoraforum.org/forum/showthread.php?t=94183) in these forums that had a good fix for it. Guess the -testing pkg was released. I didn't try any of the other solutions. I ref'd this thread in the other. I ran a couple yum updates yesterday and got to a point where there were none, but it looks like I'm still running the old -testing version:
module-init-tools-3.2-0.pre9.0.FC4.1

Someone in the other thread posted his rpm output:
module-init-tools-3.2-0.pre9.2

so now I need to figure out how to update it. Maybe I didn't have the right repositor[y|ies] set. Guess I could erase and re-add.

EDIT: yep, wasn't looking in the Dev repos. So, it hasn't been released. Updating now....
...
and it just killed the OS. Yumex output showed some errors and at the end said, Too many values to unpack, or somesuch. The module-init-tools-3.2-0.pre9.2 pkgs required some glib stuff, which I added. Then I couldn't open a term, run yuemx (not there) run /sbin/shutdow (not there). Forced reboot and the kernel panics:
"exec of init (/sbin/init) failed!!!: 2"
Yeah, it's about a 3 '!' kinda problem. Welcome to the bleeding edge, eh.

Any suggestions?

[ThurstonX]

As some rock band once sang, I'm an impatient cat.
Since this Fedora box is purely for play, I went ahead and tried a couple reinstalls rather than try to troubleshoot the failed yum(ex) glib(c) kernel-panic-causing updates ... that apparently were never applied, but which just trashed the OS. I did try the Rescue Disc approach, but it proved fruitless.

First attempt was to preserve '/' but reformat /boot and swap. The install went OK, but X was hosed, so surely something wasn't clicking with the installed pkgs and what was leftover.

Second attempt was a full re-install. I did mange to save a few config files to a shared FAT32 drive. Install was fine, of course. Instead of updating just the kernel rpms, and by extension the required udev update, I went for everything BUT those. That just finished and I rebooted...

...and guess what was broken?! The "greeter" screen. So, this problem is not caused by kernel and udev but by .... the latest xorg rpms? Who knows. CTRL+ALT+BACKSPACE had a different result this time, not restoring the default blue greeter but throwing an error and "trying another," far simpler greeter. That got me in. Time to try the kernel updates again.

What worries me is the module-init stuff, which seemed to be the cause of this most recent fatality. Perhaps they'll play nicer this time. Perhaps I'll try the "vga" option in the other thread.

Perhaps I'll boot to WinXp..... NAAAAAAAAAAHHHHHHHH! :p