breakage when installing 2007.1 by update from rc3

[terry-s]

I tried to update from 2007.1-rc3 (was working ok) to 2007.1 (official release), and now I can't get X/KDE running any more.

Also, the path config info (at least for 'root') looks wiped during the update.

 

Symptoms:

 

(1) # urpmi now gives error "bash: urpmi: command not found"

but # /usr/sbin/urpmi --auto-update

works as intended (and it confirms all media and packages up-to-date).

 

(2) When I try to get X/KDE back with "$ startx" it fails with errors as follows:-

(EE) failed to load module "nvidia" (module does not exist, 0)

(EE) no device detected.

Fatal server error

no screens found

XIO: fatal IO error 104 (connection reset by peer) on X server (:0.0)

after o requests (0 known processes) with 0 events

(then back to the $ prompt).

 

My machine has an Asus A7N8X-X mbd with nVidia GeForce4 440MX AGP video card.

 

Background is that I first installed 2007.1-rc3 shortly before 2007.1 official was released.

Install medium was the 'One' live CD. Installation went ok (fixing glitches of hang on startup and failure to power-down).

After install, I used package download sites offered by the 'smart urpmi' website for 2007.1-rc3 cooker, and installed the media sites from a root prompt # with the smart urpmi site's own 'smarturpmi.sh' script. When installing software from these sites using the GUI Mandriva Control Center tool, I included downloads of what seemed appropriate nvidia (binary I believe) driver package. All seemed to go very well, video was responding very fast.

 

After release of the official 2007.1 distro, and after reading hints on these forums, I tried to update from 2007.1 rc3 (cooker) to 2007.1 official release by using again the smarturpmi.sh script provided by the smart urpmi site, this time 'tuned' to refer to official 2007.1 release-sites. The script wiped the 2007.1-rc3 media- site-info ok, and uploaded new media-site-info ok, and it went on to update about 38 packages -- including some nvidia-related ones. (I did this from a # console prompt without X running.) There were several 'additional info' messages during the update, they flew past too quickly for me to read them. urpmi now reports that the media and packages are all up-to-date. Then the main problem became clear -- no X/GUI, as mentioned above. Reboot doesn't enable getting back into X/GUI.

 

I wonder which config files are likely broken? is it more broken than that? how do I get back into the GUI?

 

Thanks for any hints.

Terry

[adamw]

can you do:

 

rpm -qa | grep nvidia

rpm -qa | grep kernel

 

and post the results? thanks.

[terry-s]

Thanks: here they are:

 

"# rpm -qa | grep nvidia" yields:---

nvidia71xx-kernel-2.6.17-13mdvlegacy-7184-1mdk

nvidia96xx-1.0-9631.6plf2007.1

dkms-nvidia71xx-1.0-7185.1plf2007.1

nvidia97xx-kernel-2.6.17-13mdvlegacy-9755-1mdk

dkms-nvidia97xx-1.0-9755.2plf2007.1

nvidia97xx-1.0-9755.2plf2007.1

nvidia96xx-kernel-2.6.17-13mdvlegacy-9631-1mdk

dkms-nvidia96xx-1.0-9631.6plf2007.1

nvidia71xx-1.0-7185.1plf2007.1

 

"rpm -qa | grep kernel" yields:--

nvidia71xx-kernel-2.6.17-13mdvlegacy-7184-1mdk

slmodem-kernel-2.6.17-13mdvlegacy-2.9.11-1mdk

ndiswrapper-kernel-2.6.17-13mdvlegacy-1.21-2mdv2007.1

ati-kernel-2.6.17-13mdvlegacy-8.34.8-1mdk

hsfmodem-kernel-2.6.17-13mdvlegacy-7.47.00.03full-1mdk

nvidia97xx-kernel-2.6.17-13mdvlegacy-9755-1mdk

kernel-legacy-latest-2.6.17-13mdv

nvidia96xx-kernel-2.6.17-13mdvlegacy-9631-1mdk

kernel-legacy-2.6.17.13mdv-1-1mdv2007.1

hcfpcimodem-kernel-2.6.17-13mdvlegacy-1.10full-1mdk

madwifi-kernel-2.6.17-13mdvlegacy-0.9.2-1mdk

kernel-source-stripped-2.6.17.13mdv-1-1mdv2007.1

 

Terry

[terry-s]

Additional info:

1: I have found backup files that I kept from before trying the update from 2007.1rc3 to 2007.1official.

They allowed me to address/fix problem number 1, about the bash configuration: I managed to restore the config files (in '/root/') for the 'root' user -- because comparison with backup showed that these had somehow been erased during operation of the smarturpmi.sh update script. This restoration now allows the root user to call 'urpmi' without giving the full path, like before the update. (Also, the original X11-related path entry is also included back in the root user path, and I wondered if that might help with the GUI problem and finding the nvidia driver, but it doesn't.) The problem with the GUI and startx is still there.

 

2: The backup also tells me a bit more about the package history: out of the nvidia packages listed in the previous message in answer to your 'rpm -qa' question, two were newly added during the update: dkms-nvidia-71xx and dkms-nvidia-97xx. Four of the other nvidia packages (the ones now with 'plf' in their names) were slightly later numerical versions of packages in rc3. Before the update, there had been earlier 'mdv' versions of these packages.

 

Maybe this offers a clue about what I could try next?

Thanks in advance for any hints.

Terry

[terry-s]

More diagnostics:

Comparing with backup, /etc/X11/xorg.conf was not altered at all by the update.

The module & module path specified in xorg.conf was for nvidia96xx, and both the path and the file at the end of the path had the same name before and after the update.

But curiously enough, although the nvidia96xx package version had been changed during the update from -1.0-9631.3mdv2007.1 to -1.0-9631.6plf2007.1, the date of the module file had gone backwards in the update, so--

 

:in (updated partition) video driver

ls -al /usr/lib/xorg/modules/extensions/nvidia96xx/*9631 :

-rwxr-xr-x 1 root root 671548 Mar 13 15:43

/usr/lib/xorg/modules/extensions/nvidia96xx/libglx.so.1.0.9631*

 

:(backup) video driver was

ls -al /usr/lib/xorg/modules/extensions/nvidia96xx/*9631 :

-rwxr-xr-x 1 root root 675436 Mar 26 17:51

/usr/lib/xorg/modules/extensions/nvidia96xx/libglx.so.1.0.9631*

 

Could this be a bad update of the nvidia96xx package? Could I revert this package?

 

On the other hand, other folks also are getting into trouble with starting up X/GIU, with the same kind of messages as mine (e.g. see thread https://mandrivausers.org/index.php?showtopic=38062 ), and they don't have an update history.

 

So what to do next to find the cause of the trouble?

Terry

[adamw]

ah, yeah, PLF and MDV NVIDIA packages do not seem to be playing nicely together.

 

What graphics card do you have? I am going to suggest you remove all nvidia-related packages and re-install just the appropriate one from either PLF or MDV repositories, then re-configure the card with XFdrake.

[terry-s]

Thanks, Adam. That worked ok.

 

The correct set of driver packages (for nvidia GeForce4 440MX) turns out to be the nvidia96xx series.

The working rc3 install additionally had nvidia and kernel packages for the not-needed nvidia71xx and nvidia97xx series as well, but these clearly did no harm at that stage.

 

The smarturpmi.sh update script (used for moving from rc3 to official release) for some reason unknown pulled in both of two (not-needed) dkms-nvidia??xx packages, for the 71xx series and 97xx series, on top of all the rest. That looks as if it did the damage.

 

In regard to your specific advice, I used urpme to try to remove all the nvidia-lreated packages. urpme refused to remove the nvidia??xx-kernel packages, which are still there, and clearly they are doing no harm. urpme did remove all of the others, and after that, the dkms-nvidia96xx and nvidia96xx packages reinstalled ok with urpmi. I also used XFdrake (it indicated that it 'thought' my card was GeForce3), but I'm not sure that it had any work to do here, because xorg.conf had not been changed in the update and was still pointing to the nvidia96xx driver.

 

Terry

[adamw]

Great, glad it works.

 

Are you sure XFdrake doesn't see your card as 'GeForce 3 - GeForce 4'? I reworked all the NVIDIA card detection stuff for 2007.1, including putting cards into categories ('TNT - GeForce 2', 'GeForce 3 - GeForce 4', 'GeForce FX - GeForce 7950' and 'GeForce 8300 - GeForce 8800') according to what driver(s) support them rather than attempting to have definitions for every card type, or generation, or whatever, as we had before (it was a mess). See http://wiki.mandriva.com/en/Releases/Mandr...ds_and_chipsets . If the card is really detected in the wrong group, please let me know and I'll fix it.

[terry-s]

Well I just tried it again, and it says GeForce3 - G... . There's a linebox that's not very big there, I can't get anything else to display with the keyboard keys, none of my keys will scroll anything and there's no mouse availability, but perhaps as you suggest there's a longer message somewhere in there, and if there is it could be mentioning GeForce4 as well, only I can't see it! :-)