HOWTO use NVIDIA 1.0.6111 on recent kernels

[moocha]

Why should I use 1.0.6111 and not some fancy new version?
Many users complain about lower frame rates with the 1.0.7xxx series of drivers. Additionally, NVIDIA cards are running much hotter with the 1.0.7xxx series. As much as in 20-30 degrees hotter in some cases.

Then why shouldn't I use 1.0.6629?
1.0.6629 has a known bug that prevents cards using cores older than NV20 from working under X at all (affected are, for example, GeForce4 MX440 series cards - i.e., mine ). If your card ran fine under 1.0.6629 then by all means use that one. I don't know whether it compiles on recent kernels though, because I can't use it.

But I thought 1.0.6111 doesn't compile at all on recent kernels because of some AGP issue or the other?
Contrary to popular opinion, it works just fine if you pull in a slightly modified agp_backend.h from the 2.6.8.1 headers and don't let it compile against the installed agp_backend.h header.

Does this work on my system?
No idea. It works on my playground systems (one Athlon XP, one dual Pentium 3) with a variety of kernels (at least gentoo-sources, ck-sources, mm-sources, love-sources, nitro-sources, hardened-sources, on the stable versions 2.6.8.1, 2.6.9, 2.6.10, 2.6.11, and on the current 2.6.12 release candidates). If it doesn't work for you - ah well...

What about AMD64?
The ebuilds are based on the official nvidia-kernel and nvidia-glx ebuilds with the same version and revision numbers, so if those worked these should work too. If not - ah well... I keyworded it -* x86, just to annoy you. Ha ha, only serious. (Translation: You will have to keyword it to your architecture too, explicitely, to prevent you from accidentally doing stuff you don't want to do.)

Yeah, yeah, now cut the crap, how do I use this thing?
If you know how to use the portage overlay, here are the nvidia-kernel and nvidia-glx ebuilds: http://euroconsult.cluj.astral.ro/overl ... lay.tar.gz. Have fun.
If you don't know what I mean by overlay, do the following (as root of course):

• Edit /etc/make.conf and add this:

  Code: Select all

  PORTDIR_OVERLAY="/usr/local/portage"

• Make sure the overlay directory exists:

  Code: Select all

  mkdir -p /usr/local/portage

• Grab the patched ebuilds:

  Code: Select all

  cd /root
  wget http://euroconsult.cluj.astral.ro/overlays/nv6111overlay.tar.gz

• Untar it into the overlay:

  Code: Select all

  tar -xzpvf /root/nv6111overlay.tar.gz -C /usr/local/portage

• Mask out more recent versions of nvidia-kernel and nvidia-glx:

  Code: Select all

  echo ">media-video/nvidia-kernel-1.0.6111-r3" >> /etc/portage/package.mask
  echo ">media-video/nvidia-glx-1.0.6111-r1" >> /etc/portage/package.mask

• Emerge the patched driver:

  Code: Select all

  emerge nvidia-kernel nvidia-glx

Don't forget to edit modules.autoload.d and do the xorg.conf thing, just like you'd do normally.

Hope this helps.

[darklegion]

I tried patching in all of those patches in the ebuild manually(I don't use gentoo),but it complained on modprobe with "unknown symbol inter_module_get".I assume that means something changed in the 2.6.12 releases,but it doesn't seem to be a simple rename of inter_module_get and I don't have the knowledge to experiment any further than that.I am using 2.6.12-rc1-love1 btw.

[kevev]

Thank You!!!

Inspiron 2650 ,A13 ,Bios "LCD"

[moocha]

I tried patching in all of those patches in the ebuild manually(I don't use gentoo),but it complained on modprobe with "unknown symbol inter_module_get".I assume that means something changed in the 2.6.12 releases,but it doesn't seem to be a simple rename of inter_module_get and I don't have the knowledge to experiment any further than that.I am using 2.6.12-rc1-love1 btw.

That's odd, it worked for me perfectly on that kernel... AFAIK, the patch that solves the inter_module_get problem is included (it's one of Zander's patches for mm-based kernels). Perhaps fetching the 1.0.6629 patches from Gentoo's CVS might help? Don't forget that the order of patching is critically important! The patches are here: http://www.gentoo.org/cgi-bin/viewcvs.c ... ia-kernel/

[moocha]

Thank You!!!

Inspiron 2650 ,A13 ,Bios "LCD"

You're welcome, glad it helped you!

[baeksu]

I tried patching in all of those patches in the ebuild manually(I don't use gentoo),but it complained on modprobe with "unknown symbol inter_module_get".I assume that means something changed in the 2.6.12 releases,but it doesn't seem to be a simple rename of inter_module_get and I don't have the knowledge to experiment any further than that.I am using 2.6.12-rc1-love1 btw.

That's odd, it worked for me perfectly on that kernel... AFAIK, the patch that solves the inter_module_get problem is included (it's one of Zander's patches for mm-based kernels). Perhaps fetching the 1.0.6629 patches from Gentoo's CVS might help? Don't forget that the order of patching is critically important! The patches are here: http://www.gentoo.org/cgi-bin/viewcvs.c ... ia-kernel/

I'm having same problem with darklegion. This is with 2.6.12 kernel and 6111 nvidia (not 6629). I guess the new kernel broke something.

If anyone has solution with the same kernel and nvidia versions, please let us know...

[moocha]

Can't reproduce it, but then I've given up on -mm kernels a long time ago. They redefine the meaning of "experimental" .

[darklegion]

I think the problem is I can't get hold of zander's mm fixes as his site has gone and I can't use the ebuilds since I don't have gentoo,and am not willing to set up portage just for this.If anyone could attach a regular patch of zander's fixes,it would be much appreciated.

[moocha]

Zander's site hasn't gone. Zander is NVIDIA's lead Linux driver developer. You can find his patches on his usual hangout, the NVNews forum: http://www.nvnews.net/vbulletin/

[darklegion]

Sorry,got Zander mixed up with someone else on the nvnews forums,who was supplying patches for 2.6.11-mm* kernels(whose site is gone).I'll search around nvnews for the correct patch.
Thanks

EDIT : Figured it out,but I had to forcefully patch os-agp.c and AGP does not seem to be working as a result :

Code: Select all

NVRM: not using NVAGP, an AGPGART backend is loaded! 

That is with either NVAgp=1 or 2

This might just require a reboot to fix this though,but to be honest I'm just happy to have a more stable driver.

[JinxterX]

Great job, thanks

[wip]

hi!

i'm using mm-sources (2.6.12-rc3-mm3) and i tried your patch cause it's not working from the documentation (nvidia for gentoo). so. this is what i get after emerging nvidia-kernel, glx:

...
.../nvidia-kernel-1.0.6111-r3/work/NVIDIA-Linux-x86-1.0-6111-pkg1/usr/src/nv/
nv.c
In file included from include/linux/list.h:7,
from include/linux/wait.h:23,
from include/asm/semaphore.h:41,
from include/linux/sched.h:20,
from include/linux/module.h:10,
from /var/tmp/portage/nvidia-kernel-1.0.6111-r3/work/NVIDIA-Lin
ux-x86-1.0-6111-pkg1/usr/src/nv/nv-linux.h:52,
from /var/tmp/portage/nvidia-kernel-1.0.6111-r3/work/NVIDIA-Lin
ux-x86-1.0-6111-pkg1/usr/src/nv/nv.c:14:
include/linux/prefetch.h: In function `prefetch_range':
include/linux/prefetch.h:62: warning: pointer of type `void *' used in arithmeti
c
In file included from include/linux/dmapool.h:14,
from include/linux/pci.h:868,
from /var/tmp/portage/nvidia-kernel-1.0.6111-r3/work/NVIDIA-Lin
ux-x86-1.0-6111-pkg1/usr/src/nv/nv-linux.h:75,
from /var/tmp/portage/nvidia-kernel-1.0.6111-r3/work/NVIDIA-Lin
ux-x86-1.0-6111-pkg1/usr/src/nv/nv.c:14:
include/asm/io.h: In function `check_signature':
include/asm/io.h:253: warning: wrong type argument to increment
/var/tmp/portage/nvidia-kernel-1.0.6111-r3/work/NVIDIA-Linux-x86-1.0-6111-pkg1/u
sr/src/nv/nv.c: In function `_get_phys_address':
/var/tmp/portage/nvidia-kernel-1.0.6111-r3/work/NVIDIA-Linux-x86-1.0-6111-pkg1/u
sr/src/nv/nv.c warning: passing arg 1 of `pmd_offset' from incompatible po
inter type
/var/tmp/portage/nvidia-kernel-1.0.6111-r3/work/NVIDIA-Linux-x86-1.0-6111-pkg1/u
sr/src/nv/nv.c: In function `nv_agp_init':
/var/tmp/portage/nvidia-kernel-1.0.6111-r3/work/NVIDIA-Linux-x86-1.0-6111-pkg1/u
sr/src/nv/nv.c warning: implicit declaration of function `inter_module_get
'
/var/tmp/portage/nvidia-kernel-1.0.6111-r3/work/NVIDIA-Linux-x86-1.0-6111-pkg1/u
sr/src/nv/nv.c warning: `inter_module_put' is deprecated (declared at incl
ude/linux/module.h:568)
cc -Wp,-MD,/var/tmp/portage/nvidia-kernel-1.0.6111-r3/work/NVIDIA-Linux-x86-1.
0-6111-pkg1/usr/src/nv/.os-agp.o.d -nostdinc -isystem /usr/lib/gcc-lib/i686-pc-l
inux-gnu/3.3.5-20050130/include -D__KERNEL__ -Iinclude -Iinclude2 -I/usr/src/lin
ux-2.6.12-rc3-mm3/include -I/var/tmp/portage/nvidia-kernel-1.0.6111-r3/work/NVI
DIA-Linux-x86-1.0-6111-pkg1/usr/src/nv -Wall -Wstrict-prototypes -Wno-trigraphs
-fno-strict-aliasing -fno-common -ffreestanding -O2 -fomit-frame-pointer -pipe -
msoft-float -mpreferred-stack-boundary=2 -march=i686 -I/usr/src/linux-2.6.12-rc3
-mm3/include/asm-i386/mach-default -Iinclude/asm-i386/mach-default -I/var/tmp/p
ortage/nvidia-kernel-1.0.6111-r3/work/NVIDIA-Linux-x86-1.0-6111-pkg1/usr/src/nv
-Wall -Wimplicit -Wreturn-type -Wswitch -Wformat -Wchar-subscripts -Wparentheses
-Wpointer-arith -Wno-multichar -Werror -O -fno-common -MD -Wno-cast-qual -Wno-e
rror -D_LOOSE_KERNEL_NAMES -D__KERNEL__ -DMODULE -DNTRM -D_GNU_SOURCE -D_LOOSE_K
ERNEL_NAMES -D__KERNEL__ -DMODULE -DNV_MAJOR_VERSION=1 -DNV_MINOR_VERSION=0 -DNV
_PATCHLEVEL=6111 -DNV_UNIX -DNV_LINUX -DNV_INT64_OK -DNVCPU_X86 -UDEBUG -U_DEBUG
-DNDEBUG -DREMAP_PFN_RANGE -DNV_CHANGE_PAGE_ATTR_PRESENT -DHAVE_PCI_GET_CLASS -
DMODULE -DKBUILD_BASENAME=os_agp -DKBUILD_MODNAME=nvidia -c -o /var/tmp/portage/
nvidia-kernel-1.0.6111-r3/work/NVIDIA-Linux-x86-1.0-6111-pkg1/usr/src/nv/os-agp.
o /var/tmp/portage/nvidia-kernel-1.0.6111-r3/work/NVIDIA-Linux-x86-1.0-6111-pkg1
/usr/src/nv/os-agp.c
In file included from include/linux/list.h:7,
from include/linux/wait.h:23,
from include/asm/semaphore.h:41,
from include/linux/sched.h:20,
from include/linux/module.h:10,
from /var/tmp/portage/nvidia-kernel-1.0.6111-r3/work/NVIDIA-Lin
ux-x86-1.0-6111-pkg1/usr/src/nv/nv-linux.h:52,
from /var/tmp/portage/nvidia-kernel-1.0.6111-r3/work/NVIDIA-Lin
ux-x86-1.0-6111-pkg1/usr/src/nv/os-agp.c:24:
include/linux/prefetch.h: In function `prefetch_range':
include/linux/prefetch.h:62: warning: pointer of type `void *' used in arithmeti
c
In file included from include/linux/dmapool.h:14,
from include/linux/pci.h:868,
from /var/tmp/portage/nvidia-kernel-1.0.6111-r3/work/NVIDIA-Lin
ux-x86-1.0-6111-pkg1/usr/src/nv/nv-linux.h:75,
from /var/tmp/portage/nvidia-kernel-1.0.6111-r3/work/NVIDIA-Lin
ux-x86-1.0-6111-pkg1/usr/src/nv/os-agp.c:24:
include/asm/io.h: In function `check_signature':
include/asm/io.h:253: warning: wrong type argument to increment
/var/tmp/portage/nvidia-kernel-1.0.6111-r3/work/NVIDIA-Linux-x86-1.0-6111-pkg1/u
sr/src/nv/os-agp.c: At top level:
/var/tmp/portage/nvidia-kernel-1.0.6111-r3/work/NVIDIA-Linux-x86-1.0-6111-pkg1/u
sr/src/nv/os-agp.c:48: error: syntax error before '*' token
/var/tmp/portage/nvidia-kernel-1.0.6111-r3/work/NVIDIA-Linux-x86-1.0-6111-pkg1/u
sr/src/nv/os-agp.c:48: warning: type defaults to `int' in declaration of `drm_ag
p_p'
/var/tmp/portage/nvidia-kernel-1.0.6111-r3/work/NVIDIA-Linux-x86-1.0-6111-pkg1/u
sr/src/nv/os-agp.c:48: warning: data definition has no type or storage class
/var/tmp/portage/nvidia-kernel-1.0.6111-r3/work/NVIDIA-Linux-x86-1.0-6111-pkg1/u
sr/src/nv/os-agp.c: In function `KernInitAGP':
/var/tmp/portage/nvidia-kernel-1.0.6111-r3/work/NVIDIA-Linux-x86-1.0-6111-pkg1/u
sr/src/nv/os-agp.c:76: warning: assignment discards qualifiers from pointer targ
et type
/var/tmp/portage/nvidia-kernel-1.0.6111-r3/work/NVIDIA-Linux-x86-1.0-6111-pkg1/u
sr/src/nv/os-agp.c:85: error: request for member `acquire' in something not a st
ructure or union
/var/tmp/portage/nvidia-kernel-1.0.6111-r3/work/NVIDIA-Linux-x86-1.0-6111-pkg1/u
sr/src/nv/os-agp.c:88: warning: `inter_module_put' is deprecated (declared at in
clude/linux/module.h:568)
/var/tmp/portage/nvidia-kernel-1.0.6111-r3/work/NVIDIA-Linux-x86-1.0-6111-pkg1/u
sr/src/nv/os-agp.c:113: error: request for member `copy_info' in something not a
structure or union
/var/tmp/portage/nvidia-kernel-1.0.6111-r3/work/NVIDIA-Linux-x86-1.0-6111-pkg1/u
sr/src/nv/os-agp.c:173: error: request for member `enable' in something not a st
ructure or union
/var/tmp/portage/nvidia-kernel-1.0.6111-r3/work/NVIDIA-Linux-x86-1.0-6111-pkg1/u
sr/src/nv/os-agp.c:185: error: request for member `release' in something not a s
tructure or union
/var/tmp/portage/nvidia-kernel-1.0.6111-r3/work/NVIDIA-Linux-x86-1.0-6111-pkg1/u
sr/src/nv/os-agp.c:186: warning: `inter_module_put' is deprecated (declared at i
nclude/linux/module.h:568)
/var/tmp/portage/nvidia-kernel-1.0.6111-r3/work/NVIDIA-Linux-x86-1.0-6111-pkg1/u
sr/src/nv/os-agp.c: In function `KernTeardownAGP':
/var/tmp/portage/nvidia-kernel-1.0.6111-r3/work/NVIDIA-Linux-x86-1.0-6111-pkg1/u
sr/src/nv/os-agp.c:216: error: request for member `release' in something not a s
tructure or union
/var/tmp/portage/nvidia-kernel-1.0.6111-r3/work/NVIDIA-Linux-x86-1.0-6111-pkg1/u
sr/src/nv/os-agp.c:218: warning: `inter_module_put' is deprecated (declared at i
nclude/linux/module.h:568)
/var/tmp/portage/nvidia-kernel-1.0.6111-r3/work/NVIDIA-Linux-x86-1.0-6111-pkg1/u
sr/src/nv/os-agp.c: In function `KernAllocAGPPages':
/var/tmp/portage/nvidia-kernel-1.0.6111-r3/work/NVIDIA-Linux-x86-1.0-6111-pkg1/u
sr/src/nv/os-agp.c:265: error: request for member `allocate_memory' in something
not a structure or union
/var/tmp/portage/nvidia-kernel-1.0.6111-r3/work/NVIDIA-Linux-x86-1.0-6111-pkg1/u
sr/src/nv/os-agp.c:273: error: request for member `bind_memory' in something not
a structure or union
/var/tmp/portage/nvidia-kernel-1.0.6111-r3/work/NVIDIA-Linux-x86-1.0-6111-pkg1/u
sr/src/nv/os-agp.c:290: error: request for member `unbind_memory' in something n
ot a structure or union
/var/tmp/portage/nvidia-kernel-1.0.6111-r3/work/NVIDIA-Linux-x86-1.0-6111-pkg1/u
sr/src/nv/os-agp.c:305: error: request for member `free_memory' in something not
a structure or union
/var/tmp/portage/nvidia-kernel-1.0.6111-r3/work/NVIDIA-Linux-x86-1.0-6111-pkg1/u
sr/src/nv/os-agp.c: In function `KernMapAGPPages':
/var/tmp/portage/nvidia-kernel-1.0.6111-r3/work/NVIDIA-Linux-x86-1.0-6111-pkg1/u
sr/src/nv/os-agp.c:345: error: request for member `unbind_memory' in something n
ot a structure or union
/var/tmp/portage/nvidia-kernel-1.0.6111-r3/work/NVIDIA-Linux-x86-1.0-6111-pkg1/u
sr/src/nv/os-agp.c: In function `KernFreeAGPPages':
/var/tmp/portage/nvidia-kernel-1.0.6111-r3/work/NVIDIA-Linux-x86-1.0-6111-pkg1/u
sr/src/nv/os-agp.c:444: error: request for member `unbind_memory' in something n
ot a structure or union
/var/tmp/portage/nvidia-kernel-1.0.6111-r3/work/NVIDIA-Linux-x86-1.0-6111-pkg1/u
sr/src/nv/os-agp.c:445: error: request for member `free_memory' in something not
a structure or union
make[4]: *** [/var/tmp/portage/nvidia-kernel-1.0.6111-r3/work/NVIDIA-Linux-x86-1
.0-6111-pkg1/usr/src/nv/os-agp.o] Error 1
make[3]: *** [_module_/var/tmp/portage/nvidia-kernel-1.0.6111-r3/work/NVIDIA-Lin
ux-x86-1.0-6111-pkg1/usr/src/nv] Error 2
make[2]: *** [modules] Error 2
NVIDIA: left KBUILD.
nvidia.ko failed to build!
make[1]: *** [module] Error 1
make: *** [module] Error 2

!!! ERROR: media-video/nvidia-kernel-1.0.6111-r3 failed.
!!! Function linux-mod_src_compile, Line 473, Exitcode 2
!!! Unable to make IGNORE_CC_MISMATCH=yes V=1 S
YSSRC=/usr/src/linux SYSOUT=/usr/src/linux clean module



any idea?

[moocha]

Can't help you. Morton's patchset breaks NVIDIA at each and every release since it's beyond bleeding edge - Morton merges highly experimental stuff. I don't bother with mm-sources release candidates any more, sorry.

[IronTek]

I'm getting essentially the same error wip is.

In the meantime, I've had to sell my soul and use the nv drivers. Oh the pain!

[moocha]

Why on Earth are you guys using mm-sources (which is known to break NVIDIA's drivers very often), and on top of that a Release Candidate version?
2.6.11 has been a very good kernel for me and I don't plan to upgrade anytime soon. This patched ebuild has worked for me flawlessly on the 2.6.11 ck-, gentoo-, hardened-, love-, cko-, and vanilla-sources...

[voytas]

Then why shouldn't I use 1.0.6629?
1.0.6629 has a known bug that prevents cards using cores older than NV20 from working under X at all (affected are, for example, GeForce4 MX440 series cards - i.e., mine ). If your card ran fine under 1.0.6629 then by all means use that one. I don't know whether it compiles on recent kernels though, because I can't use it.

not true, i have NV17 card (GF4 MX 440) and it works with this driver...

i had a TNT2PRO cards ant it didn't work with noting newer than 6111

[moocha]

Then why shouldn't I use 1.0.6629?
1.0.6629 has a known bug that prevents cards using cores older than NV20 from working under X at all (affected are, for example, GeForce4 MX440 series cards - i.e., mine ). If your card ran fine under 1.0.6629 then by all means use that one. I don't know whether it compiles on recent kernels though, because I can't use it.

not true, i have NV17 card (GF4 MX 440) and it works with this driver...

i had a TNT2PRO cards ant it didn't work with noting newer than 6111

GF4MX440 working with 6629? Whoa. Never seen that before. My information came from the official NVIDIA forums - they stated that 6629 had a known bug that caused pre-NV20 cores not to work with 6629...

[voytas]

GF4MX440 working with 6629? Whoa. Never seen that before. My information came from the official NVIDIA forums - they stated that 6629 had a known bug that caused pre-NV20 cores not to work with 6629...

Code: Select all

bash-2.05b$ cat /proc/fb0/vbe_info 
Version:    3.0
Vendor:     NVIDIA Corporation
Product:    NV17 Board
OEM rev:    Chip Rev A5
OEM string: NVIDIA

Code: Select all

bash-2.05b$ qpkg -I nvidia -v
media-video/nvidia-glx-1.0.6629-r6 *
media-video/nvidia-kernel-1.0.6629-r4 *

Code: Select all

bash-2.05b$ cat /proc/driver/nvidia/cards/0 
Model:           GeForce4 MX 440
IRQ:             11
Video BIOS:      04.17.00.69.00
Card Type:       AGP

Code: Select all

bash-2.05b$ uname -a Linux misio 2.6.11-gentoo-r8 #1 Wed May 11 23:06:16 CEST 2005 i686 Celeron (Coppermine) GenuineIntel GNU/Linux

Code: Select all

bash-2.05b$ cat /proc/driver/nvidia/agp/status 
Status:          Enabled
Driver:          NVIDIA
AGP Rate:        2x
Fast Writes:     Enabled
SBA:             Disabled

[moocha]

Lucky bastard . *glares at his own card*

[redhead]

Splendid stuff! moocha,

Using a laptop NX9105 with nVidia Corporation NV17 [GeForce4 420 Go 32M] (rev a3) chip (lspci says that, then it must be true ). This worked perfectly until > 6111 then all dropt. (black background, white/grey blocks, yuck!)

Followed your guide and running kernel 2.6.11-r4 with nvidia-kernel-1.0.6111-r3, and are happy again.

Great job, thanks

[moocha]

Glad to hear that .

[vipernicus]

I have the NV17 and it works for me too.

[battersea]

Additionally, NVIDIA cards are running much hotter with the 1.0.7xxx series. As much as in 20-30 degrees hotter in some cases.

How was this measured or what source(s) does this comes from?

[moocha]

Additionally, NVIDIA cards are running much hotter with the 1.0.7xxx series. As much as in 20-30 degrees hotter in some cases.

How was this measured or what source(s) does this comes from?

Comes from NVIDIA's official statements Linux forums at www.nvnews.net.

[Phlogiston]

on my machine it does not work: I can compile the driver but when I start X, the screen stays black and nothing happens. After this I get an unknown error from X. I'm using nitro sources thats probably the reason why it does not work. Any hints?