Can't build or load nvidia video driver under 667
used an nvidia patch from http://ngc891.blogdns.net/index.php?2004/09/21/3-patched-nvidia-drivers
but still can't load the driver
errors: # dmesg nvidia: module license 'NVIDIA' taints kernel. nvidia: Unknown symbol __VMALLOC_RESERVE nvidia: Unknown symbol __VMALLOC_RESERVE
works fine under 2.6.8-541
BC
I'd like to know too.
There was a note about this on kerneltrap.org a few weeks back. I think a patch was needed.
BTW: Where did you get kernel source from ?
I've been running in VESA mode for 3 weeks now.
On Fri, 2004-11-05 at 14:59 -0700, Brian Callahan wrote:
Can't build or load nvidia video driver under 667
used an nvidia patch from http://ngc891.blogdns.net/index.php?2004/09/21/3-patched-nvidia-drivers
but still can't load the driver
errors: # dmesg nvidia: module license 'NVIDIA' taints kernel. nvidia: Unknown symbol __VMALLOC_RESERVE nvidia: Unknown symbol __VMALLOC_RESERVE
works fine under 2.6.8-541
BC
On Fri, 2004-11-05 at 14:01, Kim Lux wrote:
BTW: Where did you get kernel source from ?
As per the usual answer here, you really shouldn't need it just to build drivers. I think that Nvidia have picked up that clue by now, all even mildly sane distros have something useful under /lib/modules/<kernelversion>/build (be it a symlink or a real directory with the needed files).
If you actually want the source for building a custom kernel, that's a legitimate use of course, and then there's instructions for how to get the Fedora-patched source tree from the kernel SRPM in the distribution release notes.
I've been running in VESA mode for 3 weeks now.
The nv driver, while not providing any 3D acceleration, should be _much_ better than using a vesa driver in terms of usable video modes etc... You might want to use that as a backup instead.
Best regards, Per
I also want to enable ntfs so that I can move data between my Windows partition and Linux. I'll need source for that unless they've enabled ntfs in the kernel as default. I don't think they have:
mount -tntfs /dev/hda1 tmp mount: fs type ntfs not supported by kernel
On Fri, 2004-11-05 at 15:21 -0800, Per Bjornsson wrote:
On Fri, 2004-11-05 at 14:01, Kim Lux wrote:
BTW: Where did you get kernel source from ?
As per the usual answer here, you really shouldn't need it just to build drivers. I think that Nvidia have picked up that clue by now, all even mildly sane distros have something useful under /lib/modules/<kernelversion>/build (be it a symlink or a real directory with the needed files).
If you actually want the source for building a custom kernel, that's a legitimate use of course, and then there's instructions for how to get the Fedora-patched source tree from the kernel SRPM in the distribution release notes.
I've been running in VESA mode for 3 weeks now.
The nv driver, while not providing any 3D acceleration, should be _much_ better than using a vesa driver in terms of usable video modes etc... You might want to use that as a backup instead.
Best regards, Per
-- Per Bjornsson perbj@stanford.edu Ph.D. Candidate, Department of Applied Physics, Stanford University
If you are implying that the Nvidia driver installs without source, you are partially wrong in the general and totally wrong in this case. The Nvidia driver DOES install without kernel source code on common kernels, ie ones that the Nvidia team has compiled a kernel for and included in the install pack.
However, if you read the driver fine print, you will find that if your kernel version isn't supported, you need to have kernel source installed in order to build the kernel. That is precisely what the Nvidia "install" tool does.
Thus, I DO need kernel source to get an Nvidia driver running with 2.6.9-1.667 as I highly doubt Nvidia had included a driver for this kernel in their install pack.
On Fri, 2004-11-05 at 15:21 -0800, Per Bjornsson wrote:
On Fri, 2004-11-05 at 14:01, Kim Lux wrote:
BTW: Where did you get kernel source from ?
As per the usual answer here, you really shouldn't need it just to build drivers. I think that Nvidia have picked up that clue by now, all even mildly sane distros have something useful under /lib/modules/<kernelversion>/build (be it a symlink or a real directory with the needed files).
If you actually want the source for building a custom kernel, that's a legitimate use of course, and then there's instructions for how to get the Fedora-patched source tree from the kernel SRPM in the distribution release notes.
I've been running in VESA mode for 3 weeks now.
The nv driver, while not providing any 3D acceleration, should be _much_ better than using a vesa driver in terms of usable video modes etc... You might want to use that as a backup instead.
Best regards, Per
-- Per Bjornsson perbj@stanford.edu Ph.D. Candidate, Department of Applied Physics, Stanford University
On Fri, 2004-11-05 at 18:36 -0700, Kim Lux wrote:
If you are implying that the Nvidia driver installs without source, you are partially wrong in the general and totally wrong in this case. The Nvidia driver DOES install without kernel source code on common kernels, ie ones that the Nvidia team has compiled a kernel for and included in the install pack.
The nvidia installer includes pre-compiled modules for the kernels shipped with most distributions, that is correct. But the nvidia installer can compile the module without the kernel source installed.
However, if you read the driver fine print, you will find that if your kernel version isn't supported, you need to have kernel source installed in order to build the kernel. That is precisely what the Nvidia "install" tool does.
Did you try? I used to regularly compile the nvidia modules (before the 2.6.9 kernel) without the kernel-source RPM installed (which I've never installed in my system).
Regards,
On Sat November 6 2004 5:51 am, Ricardo Veguilla wrote:
On Fri, 2004-11-05 at 18:36 -0700, Kim Lux wrote:
If you are implying that the Nvidia driver installs without source, you are partially wrong in the general and totally wrong in this case. The Nvidia driver DOES install without kernel source code on common kernels, ie ones that the Nvidia team has compiled a kernel for and included in the install pack.
The nvidia installer includes pre-compiled modules for the kernels shipped with most distributions, that is correct. But the nvidia installer can compile the module without the kernel source installed.
However, if you read the driver fine print, you will find that if your kernel version isn't supported, you need to have kernel source installed in order to build the kernel. That is precisely what the Nvidia "install" tool does.
Did you try? I used to regularly compile the nvidia modules (before the 2.6.9 kernel) without the kernel-source RPM installed (which I've never installed in my system).
True, i never download the pre-compiled kernel modules, i always let the installer do the compilation.
Manu
True, the NVidia script will check if pre-built modules exist. But if they don't, it only needs the kernel source *headers* provided with each kernel, as previously noted. From NVIDIA-Linux-x86-1.0-6111-pkg1/usr/src/nv/makefile: KERNEL_SOURCES := /lib/modules/$(shell uname -r)/build KERNEL_HEADERS := $(KERNEL_SOURCES)/include
You can unpack the nvidia source using sh NVIDIA-Linux-x86-1.0-6111-pkg1.run -a -x (note: -a accepts the license) -Bob Arendt
Kim Lux wrote:
If you are implying that the Nvidia driver installs without source, you are partially wrong in the general and totally wrong in this case. The Nvidia driver DOES install without kernel source code on common kernels, ie ones that the Nvidia team has compiled a kernel for and included in the install pack.
However, if you read the driver fine print, you will find that if your kernel version isn't supported, you need to have kernel source installed in order to build the kernel. That is precisely what the Nvidia "install" tool does.
Thus, I DO need kernel source to get an Nvidia driver running with 2.6.9-1.667 as I highly doubt Nvidia had included a driver for this kernel in their install pack.
On Fri, 2004-11-05 at 15:21 -0800, Per Bjornsson wrote:
On Fri, 2004-11-05 at 14:01, Kim Lux wrote:
BTW: Where did you get kernel source from ?
As per the usual answer here, you really shouldn't need it just to build drivers. I think that Nvidia have picked up that clue by now, all even mildly sane distros have something useful under /lib/modules/<kernelversion>/build (be it a symlink or a real directory with the needed files).
If you actually want the source for building a custom kernel, that's a legitimate use of course, and then there's instructions for how to get the Fedora-patched source tree from the kernel SRPM in the distribution release notes.
I've been running in VESA mode for 3 weeks now.
The nv driver, while not providing any 3D acceleration, should be _much_ better than using a vesa driver in terms of usable video modes etc... You might want to use that as a backup instead.
Best regards, Per
-- Per Bjornsson perbj@stanford.edu Ph.D. Candidate, Department of Applied Physics, Stanford University
On Fri, 2004-11-05 at 14:59 -0700, Brian Callahan wrote:
Can't build or load nvidia video driver under 667
used an nvidia patch from http://ngc891.blogdns.net/index.php?2004/09/21/3-patched-nvidia-drivers
but still can't load the driver
errors: # dmesg nvidia: module license 'NVIDIA' taints kernel. nvidia: Unknown symbol __VMALLOC_RESERVE nvidia: Unknown symbol __VMALLOC_RESERVE
works fine under 2.6.8-541
BC
not sure about that patch but all I did was add
unsigned int __VMALLOC_RESERVE = 128 << 20;
right after:
#ifdef MODULE_ALIAS_CHARDEV_MAJOR MODULE_ALIAS_CHARDEV_MAJOR(NV_MAJOR_DEVICE_NUMBER); #endif
in the nv.c file and ran the installer. Worked perfectly fine after that. I am currently running the *-667 kernel. Are you making sure your xorg.conf file is setup properly?
Thanks.
I'll test this on the weekend.
Where did you get source from ?
On Fri, 2004-11-05 at 17:06 -0500, Patrick Ryan Vinson wrote:
On Fri, 2004-11-05 at 14:59 -0700, Brian Callahan wrote:
Can't build or load nvidia video driver under 667
used an nvidia patch from http://ngc891.blogdns.net/index.php?2004/09/21/3-patched-nvidia-drivers
but still can't load the driver
errors: # dmesg nvidia: module license 'NVIDIA' taints kernel. nvidia: Unknown symbol __VMALLOC_RESERVE nvidia: Unknown symbol __VMALLOC_RESERVE
works fine under 2.6.8-541
BC
not sure about that patch but all I did was add
unsigned int __VMALLOC_RESERVE = 128 << 20;
right after:
#ifdef MODULE_ALIAS_CHARDEV_MAJOR MODULE_ALIAS_CHARDEV_MAJOR(NV_MAJOR_DEVICE_NUMBER); #endif
in the nv.c file and ran the installer. Worked perfectly fine after that. I am currently running the *-667 kernel. Are you making sure your xorg.conf file is setup properly?
On Fri, 2004-11-05 at 15:23 -0700, Kim Lux wrote:
Thanks.
I'll test this on the weekend.
Where did you get source from ?
On Fri, 2004-11-05 at 17:06 -0500, Patrick Ryan Vinson wrote:
On Fri, 2004-11-05 at 14:59 -0700, Brian Callahan wrote:
Can't build or load nvidia video driver under 667
used an nvidia patch from http://ngc891.blogdns.net/index.php?2004/09/21/3-patched-nvidia-drivers
but still can't load the driver
errors: # dmesg nvidia: module license 'NVIDIA' taints kernel. nvidia: Unknown symbol __VMALLOC_RESERVE nvidia: Unknown symbol __VMALLOC_RESERVE
works fine under 2.6.8-541
BC
not sure about that patch but all I did was add
unsigned int __VMALLOC_RESERVE = 128 << 20;
right after:
#ifdef MODULE_ALIAS_CHARDEV_MAJOR MODULE_ALIAS_CHARDEV_MAJOR(NV_MAJOR_DEVICE_NUMBER); #endif
in the nv.c file and ran the installer. Worked perfectly fine after that. I am currently running the *-667 kernel. Are you making sure your xorg.conf file is setup properly?
you can extract the NVidia kernel source by passing -x if I remember right. It will extract the files rather than running the installer.
From there you can edit the source.
-
Bob Arendt -
Brian Callahan -
Kim Lux -
Manu Abraham -
Patrick Ryan Vinson -
Per Bjornsson -
Ricardo Veguilla