Nvidia Kernel 1.0.6111 Fix for mm-sources thread

Message

danone · Post by **danone** » Sun Oct 03, 2004 12:22 pm

All patches doesnt work for me X.Org started and black screen and system freezes imho. I tried DaMouse's 1.0.6111-r3 it gives me that device is missing or driver is missing although its loaded..*sight*

And the latest one from geki wont work X.Org started and black screen and system hung-hard reset

sLiCeR · Post by **sLiCeR** » Sun Oct 03, 2004 12:48 pm

So I gave up, switched back to 2.6.9_rc2, lets hope nvidia-kernel is working again in final 2.6.9

firephoto · Post by **firephoto** » Sun Oct 03, 2004 3:25 pm

Nvidia-kernel will only work with 2.6.9 if Nvidia releases new drivers. The kernel change is permanent it seems so nvidia needs to change their source for the drivers.

Insanity5902 · Post by **Insanity5902** » Sun Oct 03, 2004 5:46 pm

Just throwing in my experience.

I just updated to the mm-sources, and had problems with the nvidia kernel, found this thread, tried the first fix and that didn't work then skipped to the end to DaMouse's -r3 ebuld and it works fine.

R-Type · Post by **R-Type** » Sun Oct 03, 2004 11:30 pm

I tried damouse's r3 build.
The module loads alright, but I get "(EE) NVIDIA(0): Failed to allocate config DMA context" when I try to start Xorg. I spent some time with the source but wasn't able to get real far. Out of my league I guess. Any ideas?

Earlier kernels have been working fine.

p3 850
440BX
geforce 2 ultra (bladerunner)
2.6.9-rc3-mm1 http://home.comcast.net/~orez-bus/rc3-mm1.txt
xorg 6.8.0
gcc-3.4.1

Insanity5902 · Post by **Insanity5902** » Mon Oct 04, 2004 2:31 am

R-Type - try updated xorg, the latest is 6.8.0-r1, don't know but it my be worth a shot.

R-Type · Post by **R-Type** » Mon Oct 04, 2004 3:08 am

Insanity5902 wrote:R-Type - try updated xorg, the latest is 6.8.0-r1, don't know but it my be worth a shot.

Well I suspect its kernel<-->nvidia related. the same xorg install works fine with older kernels. I am already running 6.8.0-r1.

newblar · Post by **newblar** » Mon Oct 04, 2004 6:32 am

How many of you have update nvidia-glx ?

Insanity5902 · Post by **Insanity5902** » Mon Oct 04, 2004 12:44 pm

i am funning nvidia-glx-1.0.6111 . Which is the standard one in portage.

geki · Post by **geki** » Mon Oct 04, 2004 2:42 pm

there is a new nvidia-kernel update in portage anyone may try that...

R-Type · Post by **R-Type** » Tue Oct 05, 2004 1:07 am

tried that new r2 in portage..now I get
"Failed to allocate a DMA context
pushbuffer DMA allocation failed
***aborting***
failed to allocate DMA pushbuffer"
etc, when I start X. The module loads fine.

kuba · Post by **kuba** » Tue Oct 05, 2004 7:02 am

http://kernel.damouse.co.uk/nvidia-kern ... r3.tar.bz2

works nice with 2.6.9-rc3-mm2
allthough (not nvidia related) i had to patch kernel with:
linux.kernel

Code: Select all

--- 25/include/linux/netfilter_ipv4/ip_conntrack.h~conntrack-preempt-safety-fix Mon Oct  4 14:36:19 2004
+++ 25-akpm/include/linux/netfilter_ipv4/ip_conntrack.h Mon Oct  4 14:37:02 2004
@@ -311,10 +311,11 @@ struct ip_conntrack_stat
  unsigned int expect_delete;
 };
 
-#define CONNTRACK_STAT_INC(count)    \
- do {       \
-  per_cpu(ip_conntrack_stat, get_cpu()).count++; \
-  put_cpu();     \
+#define CONNTRACK_STAT_INC(count)     \
+ do {        \
+  preempt_disable();     \
+  per_cpu(ip_conntrack_stat, smp_processor_id()).count++; \
+  preempt_disable();     \
  } while (0)
 
 /* eg. PROVIDES_CONNTRACK(ftp); */
_
--- 25/include/linux/netfilter_ipv4/ip_conntrack.h~conntrack-preempt-safety-fix Mon Oct  4 14:36:19 2004
+++ 25-akpm/include/linux/netfilter_ipv4/ip_conntrack.h Mon Oct  4 14:37:02 2004
@@ -311,10 +311,11 @@ struct ip_conntrack_stat
  unsigned int expect_delete;
 };
 
-#define CONNTRACK_STAT_INC(count)    \
- do {       \
-  per_cpu(ip_conntrack_stat, get_cpu()).count++; \
-  put_cpu();     \
+#define CONNTRACK_STAT_INC(count)     \
+ do {        \
+  preempt_disable();     \
+  per_cpu(ip_conntrack_stat, smp_processor_id()).count++; \
+  preempt_disable();     \
  } while (0)
 
 /* eg. PROVIDES_CONNTRACK(ftp); */
_

fallow · Post by **fallow** » Tue Oct 05, 2004 12:31 pm

black screen on rc3-mm2 with http://kernel.damouse.co.uk/nvidia-kern ... r3.tar.bz2 here ....

greetings

Insanity5902 · Post by **Insanity5902** » Tue Oct 05, 2004 1:20 pm

fallow wrote:black screen on rc3-mm2 with http://kernel.damouse.co.uk/nvidia-kern ... r3.tar.bz2 here ....

greetings

I know it is dumb, but I did it:
Did you startx has root? I did, not thinking, after putting in rc3-mm1 and thought it was broke, then I realized what I did.

fallow · Post by **fallow** » Tue Oct 05, 2004 4:44 pm

hehe, You`re right Insanity5902 .
thx:)

starting as root - black screen
non-root - ok

but i`m used to / I often starting x`s as root
maybe some idea to fix that ?

greetings

DaMouse · Post by **DaMouse** » Wed Oct 06, 2004 10:03 pm

fix: run X as a user, really, its better for your health

Ok, can people try the patch from kuba to fix their problems, sorry i went away for a bit because i got my forum account all messed

-DaMouse

oggialli · Post by **oggialli** » Thu Oct 07, 2004 9:05 am

Why don't you use xdm/kdm/gdm/entrance?

Insanity5902 · Post by **Insanity5902** » Thu Oct 07, 2004 1:47 pm

oggialli wrote:Why don't you use xdm/kdm/gdm/entrance?

is that a question or a suggestion?

fallow · Post by **fallow** » Thu Oct 07, 2004 2:07 pm

oggialli wrote:Why don't you use xdm/kdm/gdm/entrance?

I don`t like them all

DaMouse wrote:fix: run X as a user, really, its better for your health

hehe , exactly , that is the gold fix

but at home I want to have capability of run X`s as root . however it`s work with mm3

greetings

gun26 · Post by **gun26** » Wed Oct 27, 2004 10:53 pm

I just compiled the new 2.6.10-rc1-mm1 kernel, and I can get the nvidia-kernel to build but not load. Whether I use the Gentoo nvidia-kernel-1.0.6111-r2 ebuild or Damouse's -r3 ebuild referenced in this thread, the module refuses to load, with this appearing in dmesg:

Code: Select all

nvidia: Unknown symbol class_simple_device_add
nvidia: Unknown symbol class_simple_destroy
nvidia: Unknown symbol class_simple_device_remove
nvidia: Unknown symbol class_simple_create

Do we need a new mm fix for the ebuild? I took a look at both Nvidia's forum and lkml but I didn't spot any posts about this.

potenzial · Post by **potenzial** » Wed Oct 27, 2004 11:48 pm

There is a disscussion on the lkm...
http://marc.theaimsgroup.com/?l=linux-k ... 300351&w=2

gun26 · Post by **gun26** » Thu Oct 28, 2004 2:05 am

Thanks, potenzial. It looks like this is due to a change, in the form of EXPORT_SYMBOL_GPL(), that Nvidia and other providers of non-GPL modules will have to deal with. I didn't see the discussion because I look in on lkml via the lkml.org web site and none of today's messages seem to be showing up for some reason.

Oh well, no biggie - I have two very good functioning Reiser4-enabled kernels to use in the absence of mm - vivid-sources and nitro-sources, both of which are working fine here and both including Spock's gensplash eye candy, which won't work on mm sources (or not yet, anyway.)

peabody124 · Post by **peabody124** » Fri Oct 29, 2004 12:08 am

i really hope they are working with nvidia to get this fixed as soon as possible. looks like the kernel maintainers have changed the permissions of some functions nvidia uses to only GPL'd software. considering most of the linux world only has the option of nvidia for a decent gfx card, i hope this is addressed.

firephoto · Post by **firephoto** » Fri Oct 29, 2004 12:37 am

I really don't get what the point of this latest change is. It's obviously coming from the top this time (Linus) and even though nvidia was correct in the way they were using what was there in the kernel the way it is handled was changed breaking some non-GPL'd software. Are they trying to push nvidia into opening up their drivers? If they are should they be doing this even? Is this something that is happening so these apparent bad changes won't matter in the near future? I really like how well my nvidia drivers work and I honestly don't give a crap what license they fall under if they work and they give them too me to use like a friend hands me an ice cold beer for me to drink...... I don't see nvidia calling me and saying "haha, you have our proprietary drivers, your box is ours!".

Anyone have an 8x agp graphics card with 256mb memory, opengl capable, with pure open source gpl'd till the cows come home, and can produce 100fps @ 1024x768+ with any of the games that are popular for linux? Oh it has to be less than..... $350 too.

I'm glad nvidia is working good with the patches and 2.6.9 kernels now.

djcapelis · Post by **djcapelis** » Sat Oct 30, 2004 8:10 am

Yes, the dev responsible for the change is trying to push nvidia towards a GPL license for their drivers. It's not coming from the top and Linus actually seems to have a policy about reclassifying things into GPL-only that weren't. That however, is exactly what has happened here.

As for just changing the symbols from EXPORT_SYMBOL_GPL to EXPORT_SYMBOL... well, I tried that and blank screen bad things seemed to happen... so I did it some more and it still didn't work. *shrug*

Although... we might try changing the module_license parameter for Nvidia's source code... they certainly can't do that to fix it... but as we don't have any rights to it, we probably can. Although it might be copyright infringement... hmmm, I wonder.

Nvidia Kernel 1.0.6111 Fix for mm-sources thread

Okay

this seems like a shitty change