3.10 kernel and Nvidia

[towo]

Könnt ihr mir mal ein

Code: [Select]

ls -ld /var/lib/dkms/nvidia/304.88/nv-i2c.c
nach einem buildversuch gegen 3.10 zeigen?

@spacepenguin
Daß 319.17 nicht baut ist klar, in siduction-experimental liegt dar aktuellste Nvidia-Treiber.

BTW, für solche Probleme wäre der IRC besser geeignet, weils einfach schneller geht.

[gcp]

Patch for Nvidia and Kernel 3.11

If you compile the Kernel yourself
https://git.kernel.org/cgit/linux/kernel/git/torvalds/linux.git/
and want to use Nvidia proprietary drivers with the upcoming Kernel 3.11,
you need an additional patch.

See:
mm: use totalram_pages instead of num_physpages at runtime
https://git.kernel.org/cgit/linux/kernel/git/torvalds/linux.git/commit/?id=0ed5fd138539940a493dc69359cb2f49de70ad89

mm: kill global variable num_physpages
https://git.kernel.org/cgit/linux/kernel/git/torvalds/linux.git/commit/?id=1895418189e08c1d1eec4fbdb5fb41d793f57ba5

"Now all references to num_physpages have been removed, so kill it."
... but Nvidia.


Open /var/lib/dkms/nvidia/325.08/source/patches/linux-3.10.patch
and add this to the start of the patch:


--- a/nv-linux.h   2013-07-04 17:30:49.773211000 +0200
+++ b/nv-linux.h   2013-07-04 19:24:29.524124509 +0200
@@ -959,7 +959,7 @@
 #endif
 
 #if !defined(NV_VMWARE)
-#define NV_NUM_PHYSPAGES                num_physpages
+#define NV_NUM_PHYSPAGES                totalram_pages
 #define NV_GET_CURRENT_PROCESS()        current->tgid
 #define NV_IN_ATOMIC()                  in_atomic()
 #define NV_LOCAL_BH_DISABLE()           local_bh_disable()


This works for me.
If you want to install or upgrade Kernel 3.10, you must switch the patches manually. Of course.

I use Nvidia 325.08 from Siduction experimental.
Thanks for that.

[michaa7]

Könnt ihr mir mal ein

Code: [Select]

ls -ld /var/lib/dkms/nvidia/304.88/nv-i2c.c
nach einem buildversuch gegen 3.10 zeigen?

(da hat wohl .../build/... im pfad gefehlt)

-rw-r--r-- 1 root root 9682 Jul  4 22:30 /var/lib/dkms/nvidia/304.88/build/nv-i2c.c

@spacepenguin
Daß 319.17 nicht baut ist klar, in siduction-experimental liegt dar aktuellste Nvidia-Treiber.

BTW, für solche Probleme wäre der IRC besser geeignet, weils einfach schneller geht.

Da marschiere ich jetzt mal hin ...

BTW, hast du gesehen was gcp gepostet hat?

[michaa7]

solved here with the new fixes from siduction repo

[towo]

BTW, hast du gesehen was gcp gepostet hat?

Ja, aber das ist im Moment nicht von Belang, da Kernel 3.11 noch in relativ weiter Ferne liegt.

[spacepenguin]

Danke für 304.88-5.c0.siduction, damit geht es jetzt hier auch.

[orinoco]

What about legacy 173.x?

[vayu]

I can't get nvidia working with the 3.10 kernel. I think I have something wrong with my repository lists. apt-cache policy nvidia-kernel-dkms only shows 304.88-4 0

This is after apt-get update with the below sources:

Code: [Select]


$ apt-cache policy nvidia-glx
nvidia-glx:
  Installed: 304.88-4
  Candidate: 304.88-4
  Version table:
 *** 304.88-4 0
        500 http://http.debian.net/debian/ unstable/non-free amd64 Packages
        100 /var/lib/dpkg/status


Code: [Select]


/etc/apt/sources.list.d$ cat debian.list
# debian loadbalancer
deb http://http.debian.net/debian/ unstable main contrib non-free
#deb-src http://http.debian.net/debian/ unstable main contrib non-free
/etc/apt/sources.list.d$ cat siduction.list
# official siduction-lists
deb http://ftp.uni-stuttgart.de/siduction/base unstable main
#deb-src http://ftp.uni-stuttgart.de/siduction/base unstable main

deb http://ftp.uni-stuttgart.de/siduction/extra unstable main
#deb-src http://ftp.uni-stuttgart.de/siduction/extra unstable main

deb http://ftp.uni-stuttgart.de/siduction/fixes unstable main
#deb-src http://ftp.uni-stuttgart.de/siduction/fixes unstable main

/etc/apt/sources.list.d$ cat misc.list
# misc. siduction-lists
deb http://ftp.uni-stuttgart.de/siduction/experimental unstable main
#deb-src http://ftp.uni-stuttgart.de/siduction/experimental unstable main

#deb http://ftp.uni-stuttgart.de/siduction/user unstable main
#deb-src http://ftp.uni-stuttgart.de/siduction/user unstable main

deb http://ftp.uni-stuttgart.de/siduction/kdenext unstable main
#deb-src http://ftp.uni-stuttgart.de/siduction/kdenext unstable main


I'm unclear on what steps to take to get nvidia working with the latest du. I would try some of the nvidia drivers mentioned in this thread but I can't find them.  I'm also unclear on what package(s) would I actually install.  Am I looking to install nvidia-glx, nvidia-kernel-dkms, both or something else?

[dibl]

Hi vayu.

Unless you have an older Nvidia card, you should try driver ver. 325-08 -- it comes from the siduction repo:

Code: [Select]

root@imerabox:/# apt-cache policy nvidia-glx
nvidia-glx:
  Installed: 325.08-0.siduction.1
  Candidate: 325.08-0.siduction.1
  Version table:
 *** 325.08-0.siduction.1 0
        500 http://packages.siduction.org/experimental/ unstable/non-free amd64 Packages
        100 /var/lib/dpkg/status
     304.88-4 0
        500 http://ftp.us.debian.org/debian/ unstable/non-free amd64 Packages
        500 http://http.debian.net/debian/ unstable/non-free amd64 Packages


So, add a "siduction.list" file under /etc/apt/sources.d/ and it looks like this:

Code: [Select]

deb http://packages.siduction.org/base unstable main
#deb-src http://packages.siduction.org/base unstable main
deb http://packages.siduction.org/extra unstable main
#deb-src http://packages.siduction.org/extra unstable main
deb http://packages.siduction.org/fixes unstable main
#deb-src http://packages.siduction.org/fixes unstable main
deb http://packages.siduction.org/experimental unstable main contrib non-free

Then run an update and d-u and it should pull in the newer driver.  If it's all working, then you can comment out the "siduction" lines in your old source list file.

[vayu]

Code: [Select]

deb http://packages.siduction.org/base unstable main
#deb-src http://packages.siduction.org/base unstable main
deb http://packages.siduction.org/extra unstable main
#deb-src http://packages.siduction.org/extra unstable main
deb http://packages.siduction.org/fixes unstable main
#deb-src http://packages.siduction.org/fixes unstable main
deb http://packages.siduction.org/experimental unstable main contrib non-free

Thank you, those mirrors allowed the nvidia 325 package to show up and I was able to install without problem.  I still can't get into X with the new kernel though.  The old kernel works with the new driver (it also did with the old driver).

Code: [Select]



$ apt-cache policy nvidia-glx
nvidia-glx:
  Installed: 325.08-0.siduction.1
  Candidate: 325.08-0.siduction.1
  Version table:
 *** 325.08-0.siduction.1 0
        100 http://packages.siduction.org/experimental/ unstable/non-free amd64 Packages
        100 /var/lib/dpkg/status
     304.88-4 0
        500 http://http.debian.net/debian/ unstable/non-free amd64 Packages

$ uname -a
Linux prajipati 3.9-7.towo-siduction-amd64 #1 SMP PREEMPT Fri Jun 21 16:02:32 UTC 2013 x86_64 GNU/Linux




How can I get the 3.10 kernel working with the nvidia driver?

[vayu]

I don't need this kernel for any reason other than it was part of the dist-upgrade I just did.  How do I set the default to boot back to the last one?

[dibl]

Regarding the video driver on kernel 3.10, there should be error messages at /var/log/Xorg.0.log.

Regarding pinning the kernel at a given version, follow the guidance in the manual to use dpkg --set-selections to set the kernel image on hold. You will need to manually remove the newer kernel image first -- I don't think you can set an older version to boot by default.

[michaa7]

I don't need this kernel for any reason other than it was part of the dist-upgrade I just did.  

I can understand this. And as a temporary workaround I suggest to follow dibl's advice.
OTOH there is no reason the nvidia driver / one of the nvidia drivers shouldn't run even on *your* system.

How do I set the default to boot back to the last one?

I think it is possible even with grub2.

Have a look at /etc/default/grub (using an editor like nano, mcedit as root)
The line "GRUB_DEFAULT=0" is the one you want to edit.

GRUB_DEFAULT=0
 The number at the end of this line can be changed to make the highlight bar or selection bar in the GRUB Menu appear automatically on any GRUB Menu entry we like.
 0 (zero) tells GRUB we want the first or top GRUB entry automatically selected unless we over ride that selection with our up or down arrow key during boot-up.
 Any number greater than 0 (zero) will mean boot entries further down the GRUB Menu will be selected instead.

I am not quite sure how to select an older kernel which is somewhat hidden in a submenue (#1 would be the submenue with the newest kernel selected (?), I assume. Would #2 be the second entry in the submenue or the third entry in the main menue?). Maybe someone can give an additonal hint.

After editing this line you have to run

Code: [Select]

update-grub

Anyways, in the long run I'd recomment to solve the problem with nvidia and the newest kernel. Maybe the easiest way would be to provide better info about your system/grafics/exact errormessages and asking towo in IRC directly.

[dibl]

How do I set the default to boot back to the last one?

I think it is possible even with grub2.

Have a look at /etc/default/grub (using an editor like nano, mcedit as root)
The line "GRUB_DEFAULT=0" is the one you want to edit.

GRUB_DEFAULT=0
 The number at the end of this line can be changed to make the highlight bar or selection bar in the GRUB Menu appear automatically on any GRUB Menu entry we like.
 0 (zero) tells GRUB we want the first or top GRUB entry automatically selected unless we over ride that selection with our up or down arrow key during boot-up.
 Any number greater than 0 (zero) will mean boot entries further down the GRUB Menu will be selected instead.

I am not quite sure how to select an older kernel which is somewhat hidden in a submenue (#1 would be the submenue with the newest kernel selected (?), I assume. Would #2 be the second entry in the submenue or the third entry in the main menue?). Maybe someone can give an additonal hint.

After editing this line you have to run

Code: [Select]

update-grub

Yes, I had forgotten about that method -- that should work to set an older kernel as default.

HOWEVER ...

The next time towo releases a new kernel, your d-u will want to install it.  And if you don't want it, then you can't proceed with the d-u, and you can't update your packages.  This means your rolling distro will stop rolling, or else you'll be accumulating new kernels that you can't use.   :lol:

So, the advice is to first try to troubleshoot the error with the nvidia driver, and get it working on the new kernel, including using IRC to get help if needed.  If that fails, then you probably want to pin the kernel to a 3.9-er and keep that one going forward.

[DeepDayze]

How do I set the default to boot back to the last one?

I think it is possible even with grub2.

Have a look at /etc/default/grub (using an editor like nano, mcedit as root)
The line "GRUB_DEFAULT=0" is the one you want to edit.

GRUB_DEFAULT=0
 The number at the end of this line can be changed to make the highlight bar or selection bar in the GRUB Menu appear automatically on any GRUB Menu entry we like.
 0 (zero) tells GRUB we want the first or top GRUB entry automatically selected unless we over ride that selection with our up or down arrow key during boot-up.
 Any number greater than 0 (zero) will mean boot entries further down the GRUB Menu will be selected instead.

I am not quite sure how to select an older kernel which is somewhat hidden in a submenue (#1 would be the submenue with the newest kernel selected (?), I assume. Would #2 be the second entry in the submenue or the third entry in the main menue?). Maybe someone can give an additonal hint.

After editing this line you have to run

Code: [Select]

update-grub

Yes, I had forgotten about that method -- that should work to set an older kernel as default.

HOWEVER ...

The next time towo releases a new kernel, your d-u will want to install it.  And if you don't want it, then you can't proceed with the d-u, and you can't update your packages.  This means your rolling distro will stop rolling, or else you'll be accumulating new kernels that you can't use.   :lol:

So, the advice is to first try to troubleshoot the error with the nvidia driver, and get it working on the new kernel, including using IRC to get help if needed.  If that fails, then you probably want to pin the kernel to a 3.9-er and keep that one going forward.

If you don't want the kernel to be updated automatically with each d-u just purge the siduction kernel metapackage.