[SOLVED] nvidia driver issues

[ahz]

Hello,

vga-detect does not work for me with any tested Nvidia gfx adapter. What I see is, that the Nvidia logo is visible for about 1-2 seconds. After this the blue porteus logo and the mouse pointer appaers. At point 2 (Initializing System Services) the symbol blinks continously and after about 30 a clear blue screen with moveable mousepointer is displayed. I can reproduce this behaviour 4 different Nvidia adapters. The ati and nvidia modules are located in the optional folder (if I the nvidia module into the modules folder the behaviour is the same). I could upload logfiles to pastebin.com if necessary.

Important: If I boot with an ATI based card everything ist OK.

Another interesting point: Last known nvidia driver that worked for me was nVidia-275.09.07-porteus-v1.0-i486-1beny.xzm. All release after this one failed (in a more or less same manner as described above).

[Hamza]

Mostly sure there is a trouble between hardware version and software version..so you will need to downgrade to 275.09.07 version..

Beny, where are you ?

[brokenman]

Did you try the legacy nvidia modules on the server too?

[Ahau]

I can confirm this bug, when using kde-trinity and the nvidia driver (290 series by beny for 1.1). The issue only occurs with trinity, it does not happen in lxde or xfce.

Logs:

end of /var/log/messages:

Code: Select all

Jan 13 14:23:33 (none) kernel: [    9.896223] firewire_core: created device fw0: GUID 394fc0001b3f2dc1, S400
Jan 13 14:23:33 (none) kernel: [    9.980364] input: DualPoint Stick as /devices/platform/i8042/serio1/input/input8
Jan 13 14:23:33 (none) kernel: [    9.982934] mousedev: PS/2 mouse device common for all mice
Jan 13 14:23:33 (none) kernel: [   10.000663] input: AlpsPS/2 ALPS DualPoint TouchPad as /devices/platform/i8042/serio1/input/input9
Jan 13 14:23:34 (none) acpid: starting up with proc fs
Jan 13 14:23:34 (none) acpid: 1 rule loaded
Jan 13 14:23:34 (none) acpid: waiting for events: event logging is off
Jan 13 14:23:34 (none) dhcpcd[2252]: wlan0: waiting for carrier
Jan 13 14:23:34 (none) dhcpcd[2251]: eth0: waiting for carrier
Jan 13 14:23:35 (none) polkitd[2456]: started daemon version 0.101 using authority implementation `local' version `0.101'
Jan 13 14:23:36 (none) acpid: client connected from 2512[82:82]
Jan 13 14:23:36 (none) acpid: 1 client rule loaded
Jan 13 14:23:37 (none) /usr/sbin/gpm[2549]: *** info [startup.c(95)]: 
Jan 13 14:23:37 (none) /usr/sbin/gpm[2549]: Started gpm successfully. Entered daemon mode.
Jan 13 14:23:37 (none) kernel: [   15.657826] NET: Registered protocol family 10
Jan 13 14:23:37 (none) kernel: [   15.658782] ADDRCONF(NETDEV_UP): eth0: link is not ready
Jan 13 14:23:37 (none) acpid: client connected from 2601[0:0]
Jan 13 14:23:37 (none) acpid: 1 client rule loaded
Jan 13 14:23:43 (none) acpid: client connected from 2601[0:0]
Jan 13 14:23:43 (none) acpid: 1 client rule loaded
Jan 13 14:25:00 (none) acpid: client 2601[0:0] has disconnected
Jan 13 14:25:00 (none) acpid: client 2601[0:0] has disconnected
Jan 13 14:25:00 (none) acpid: client connected from 2803[0:0]
Jan 13 14:25:00 (none) acpid: 1 client rule loaded
Jan 13 14:25:02 (none) acpid: client connected from 2803[0:0]
Jan 13 14:25:02 (none) acpid: 1 client rule loaded
Jan 13 14:25:16 (none) acpid: client 2803[0:0] has disconnected
Jan 13 14:25:16 (none) acpid: client 2803[0:0] has disconnected
Jan 13 14:25:16 (none) acpid: client connected from 2969[0:0]
Jan 13 14:25:16 (none) acpid: 1 client rule loaded
Jan 13 14:25:19 (none) acpid: client connected from 2969[0:0]
Jan 13 14:25:19 (none) acpid: 1 client rule loaded
Jan 13 14:27:04 (none) kernel: [  222.194733] pcmanfm[3092]: segfault at 0 ip b6b31412 sp bf9527b0 error 4 in libglib-2.0.so.0.2800.6[b6abc000+ff000]
Jan 13 14:27:47 (none) kernel: [  265.787559] pcmanfm[3307]: segfault at 0 ip b6c0b412 sp bfad92f0 error 4 in libglib-2.0.so.0.2800.6[b6b96000+ff000]
Jan 13 14:27:58 (none) kernel: [  276.547226] pcmanfm[3351]: segfault at 0 ip b6c74412 sp bfc51e40 error 4 in libglib-2.0.so.0.2800.6[b6bff000+ff000]

(note, the pcmanfm segfaults were in lxde after I failed to get into kde -- I've had this issue with pcmanfm before)

kdm.log:

Code: Select all

X.Org X Server 1.9.5
Release Date: 2011-03-17
X Protocol Version 11, Revision 0
Build Operating System: Slackware 13.37 Slackware Linux Project
Current Operating System: Linux porteus 3.1.8-porteus #1 SMP PREEMPT Sat Jan 7 11:08:29 UTC 2012 i686
Kernel command line: root=/dev/ram0 rootfstype=ext4 rw initrd=/32-11/boot/initrd.xz from_dev=UUID:b459f3f5-b0c4-4462-a638-889d03cd8c19 from_dir=/32-11/porteus/ vga=791 toroot vga-detect BOOT_IMAGE=/32-11/boot/vmlinuz 
Build Date: 18 March 2011  12:08:03AM
 
Current version of pixman: 0.20.2
	Before reporting problems, check http://wiki.x.org
	to make sure that you have the latest version.
Markers: (--) probed, (**) from config file, (==) default setting,
	(++) from command line, (!!) notice, (II) informational,
	(WW) warning, (EE) error, (NI) not implemented, (??) unknown.
(==) Log file: "/var/log/Xorg.0.log", Time: Fri Jan 13 14:25:16 2012
(==) Using config file: "/etc/X11/xorg.conf"
(==) Using system config directory "/usr/share/X11/xorg.conf.d"
QImage::convertDepth: Image is a null image
QImage::smoothScale: Image is a null image
sh: warning: setlocale: LC_ALL: cannot change locale (en_US)
Corrupt JPEG data: 3323 extraneous bytes before marker 0xda
Corrupt JPEG data: 3323 extraneous bytes before marker 0xda
sh: warning: setlocale: LC_ALL: cannot change locale (en_US)

Xorg.0.log.old (I think this one is from the failed KDE startup):
http://pastebin.com/b4MRsAgj

Xorg.0.log (I think this one is from successful startup to lxde):
http://pastebin.com/0diiBDZD

Posted after 8 hours 28 minutes:
I tried beny's 275 series driver, and that got me in to kde, but it didn't seem to work with the vga-detect cheatcode -- the module was not activated at bootup, so I placed a copy in /porteus/modules (still with it present in /porteus/optional). that started nvidia up and it logged into KDE, but with an awful resolution.

I also tried the 290 series without using the kdesplash and without the default wallpaper, in case those images were corrupted, and that didn't help either.

So, I'm not a lot of help here, other than some testing....

[fanthom]

let's sum it up.

is 290.10 driver working correctly in 64bit kde-4 and lxde?
is 290.10 driver working correctly in 32bit lxde and xfce?
(can you run glxgears, any opengl game, etc?)

is the Trinity DE only affected?

from the log files i see that Xorg started successfully so driver must be ok (ahz also confirmed nvidia logo during startup) . the only thing which could mess up Trinity are libs provided by nVidia module. it could also be a driver fault itself (aka recent fglrx issues with Gnome-3 desktop, nVidia problems with kde-4 in the past, etc) but i doubt....

Monday evening i should have access to nVdia based PC so if we wont figure it out earlier i'll try to do it then.
in the mean time beny is recompiling nVidia driver with new cleanup script which doesn't strip libs/binaries by default (i have seen fglrx issues caused by this). gonna edit this post when new driver is uploaded on the server.

Cheers

EDIT:\\
i have uploaded new nVidia driver compiled by beny and created with new script:
http://ponce.cc/porteus/i486/current/nV ... -2beny.xzm
please try it and report if helps with Trinity issue.

thanks

[Ahau]

I hate to say it, but the issue continues with the updated driver.

I did verify that the 64-bit driver works with KDE4. I tried a few more things to see if I could track anything down, but still not there yet. There are a few error messages about fonts and scripting errors if I boot to init 3, run startx, and then press ctr-alt-backspace to get back to the command line, but nothing that looks unique to the nvidia driver (to me, anyway)...

[ahz]

Here are the Log-Files:

Xorg.0.log: http://pastebin.com/756zBhmC
lspci: http://pastebin.com/S5KYcW4K
dmesg: http://pastebin.com/QVTDW6v5
/var/log/messages: http://pastebin.com/6fuWwhmB

They are from a system with a nVidia Corporation G84M [Quadro FX 570M] gfx adapter. When I look across these files I cannot see anything that gives a hint for a solution...

Posted after 1 hour 8 minutes 15 seconds:
OK, I have some more test results:

NVidia 275 and the updated 290 work for me. But only, when I use the first LILO entry "KDE" (and adding vga-detect manually). If I use "Always fresh" the NVidia initialization fails. So there seems to be something in my changes folder which makes the NVidia driver work when accessed via KDE. On the other hand it seems that "Always fresh" is missing something (a script or configration file)?

Test setup:

I tested 275 first with KDE selection and it was successfull. Then I exchanged it with 290-2beny, selected KDE again and it worked too. Afterwards I used "Always fresh" with 290-2beny and it didn't work. Maybe I save the complete content of the changes folder and start all over again...

Posted after 2 hours 26 minutes 5 seconds:
After testing with a plain porteus (no /changes-directory) and new nvidia-275 and -290 drivers I can confirm what Ahau described above: the NVidia drivers do not work. It ends up like on friday with one difference: A error message appears which reads "No write access to $HOME directory (/home/guest)".

[Ahau]

Thanks, ahz! We're almost there, I know it!

Following your comments above, I ran a series of tests -- if I run the nvidia driver with 'saved changes' on a clean directory, it fails. If I run with the nouveau driver with saved changes, then re-use those saved changes with the nvidia driver, it succeeds, so there is something in the saved changes file that is needed in order for nvidia to work. I have traced it this far (using root account to login): /root/.kde/share/config

It is a file in this directory that we need, and not something deeper in a subdirectory. There are 44 candidates-- I'll narrow the field with some more tests tomorrow, but have to go to bed tonight.

[fanthom]

i have taken a different path.
i got nVidia-legacy working with Trinity without problems (Geforce Go 7300 or something) so i thought it must be some extra lib in 290.10 causing failures. i have stripped 290.10 from all additional libs like vdpau, opencl, cuda, xvmc....
i basically left only a DDX driver and GL libraries - and still got a failure!

i'm out of clues guys.
If Ahau wont figure out which files from home folder are needed for 290.10 then we will have to go back to 275 series for 32bits.

btw: i have replaced /etc/X11/xorg.conf with more generic one which should fix 800x600 resolution issue. bumped all nVidia drivers to -2beny. please download again if you are affected with this bug.

[Ahau]

the file is: ~/.kde/share/config/kconf_updaterc.

Anyone affected by this bug can boot into Trinity using the nouveau driver, copy this file to their rootcopy directory (e.g. /mnt/sdb2/porteus/rootcopy/root/.kde/share/config/kconf_updaterc), then boot with the nVidia driver as normal.

This file is not included with the default ISO, so it is created by KDE on startup if not present. I'm going to keep digging into it to find out more.

Posted after 58 minutes 45 seconds:
So, it appears that when the nVidia proprietary driver is used, kconf_updaterc will not be created, nor will it be updated. For example, I could delete some of the elements from that file in my rootcopy, and get the system to continue starting Trinity properly, but several of the elements needed to remain. What I just realized is that if I use nouveau, it always replaces any data that I delete from that file, whereas when I use nVidia, it never replaces any data. So, for some reason, nVidia doesn't allow whatever is creating or updating kconf_updaterc to do it's job. I'll keep looking...

Posted after 31 minute 6 seconds:
The problem seems to be with kconf_update. It is the process that is hanging while kde is starting up. If I start with vga-detect and boot into text mode, kconf_update hangs. I was able to deactivate the nVidia module, run kconf_update, then reactivate the nVidia module and boot into kde with the nVidia driver.

[ahz]

Thanks Ahau. You were faster! Trying to isolate the file this afternoon leads to the exact same result as you posted: it all depends on kconf_updaterc. If it is there the driver works.

[Ahau]

it appears we aren't the first to experience this with Trinity:
http://permalink.gmane.org/gmane.comp.d ... devel/1823

Googling will continue...

[fanthom]

Thanks Ahau,

uploaded patched version. here is URL:
http://ponce.cc/porteus/i486/current/nV ... -3beny.xzm

please test and report success/fail.

[Ahau]

works for me, but that's no surprise

[Onraj]

Just reporting that I had this same issue but new version seems to work fine.