|
Home > Archive > Cheap Linux Hardware > August 2005 > nv driver support Asus N6600 video card?
You are viewing an archived Text-only version of the thread.
To view this thread in it's original format and/or if you want to reply to
this thread please [click here]
| Author |
nv driver support Asus N6600 video card?
|
|
| Haines Brown 2005-08-14, 5:53 pm |
| I am running debian with kernel 2.6.8-2-686 on an Asus A8N-SLI
motherboard and an Asus Extreme N6600 video card. This is a new set
up.
With # startx, the x server fails to load: Fatal server error; Fatal
IO error 104.
The xfree86 log looks OK up to the point it loads the driver.
(II) Primary nv driver for ...
This entry leads to a long list of supported chipsets. The Asus card
uses the GeForce 6600 GPU, but the list ends just before it reaches
that GPU number. Apparently my card is too new.
Then I get:
(II) Primary device is: PCI 01:00:0
(EE) No devices detected
The motherboard has a PCI Express x16 slot, and that is the card I'm
using rather than AGP. The Asus video card directions has a drawing
that suggests that a cable connects the card to the motherboard, but
I don't see the point of it. There is no connector on the card that I
can see.
--
Haines Brown
KB1GRM
| |
| Dances With Crows 2005-08-14, 5:53 pm |
| On Sun, 14 Aug 2005 19:43:13 GMT, Haines Brown staggered into the Black
Sun and said:
> I am running debian with kernel 2.6.8-2-686 on an Asus A8N-SLI
> motherboard and an Asus Extreme N6600 video card.
>
> The xfree86 log
Hang on. XFree86 or Xorg? Which version? You should be using Xorg
(current version 6.8.2); nobody's using XFree86 now. Also, the 2.6.8
and 2.6.9 kernels had a series of stupid problems with burning CDs with
cdrecord, so if you need to do that, you need 2.6.10 at least.
> (II) Primary nv driver for ...
> This entry leads to a long list of supported chipsets. The Asus card
> uses the GeForce 6600 GPU, but the list ends just before it reaches
> that GPU number. Apparently my card is too new.
What happens if you use the evil binary-only nVidia X module and kernel
module? You'll need that anyway for 3D.
> (II) Primary device is: PCI 01:00:0
> (EE) No devices detected
> The motherboard has a PCI Express x16 slot, and that is the card I'm
> using rather than AGP.
"man nv" for xorg 6.8.2 says that the nv X module supports PCI and AGP
nVidia cards. No mention of PCI-E. The man page may be out of date;
the Xorg people hate writing documentation.
--
Matt G|There is no Darkness in Eternity/But only Light too dim for us to see
Brainbench MVP for Linux Admin / mail: TRAP + SPAN don't belong
http://www.brainbench.com / "He is a rhythmic movement of the
-----------------------------/ penguins, is Tux." --MegaHAL
| |
| Jacob Larsen 2005-08-14, 5:53 pm |
| -----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1
Dances With Crows wrote:
> Hang on. XFree86 or Xorg? Which version? You should be using Xorg
> (current version 6.8.2); nobody's using XFree86 now. Also, the 2.6.8
Debian does. Unless you use unstable that is.
> What happens if you use the evil binary-only nVidia X module and kernel
> module? You'll need that anyway for 3D.
This was what was needed for me to get GF 6800GT AGP and GF 6200 TC
PCI-E working. So this is probably a safe bet.
> "man nv" for xorg 6.8.2 says that the nv X module supports PCI and AGP
> nVidia cards. No mention of PCI-E. The man page may be out of date;
> the Xorg people hate writing documentation.
xorg hasn't hit testing yet, so I can't say anything about this, but
XFree86 works fine with PCI-E (and the binary nvidia driver).
/Jacob
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.1 (GNU/Linux)
Comment: Using GnuPG with Thunderbird - http://enigmail.mozdev.org
iD8DBQFC/ 6LMiAWIAI3xXVYRAlpBAKCLLph5r52IdUiIzkNrF
Ki6hBZ4mwCfaG5Z
d7voPfzmLSJjyESasuFFAIo=
=fy6m
-----END PGP SIGNATURE-----
| |
| John-Paul Stewart 2005-08-14, 5:53 pm |
| Dances With Crows wrote:
> On Sun, 14 Aug 2005 19:43:13 GMT, Haines Brown staggered into the Black
> Sun and said:
>
>
>
> Hang on. XFree86 or Xorg? Which version? You should be using Xorg
> (current version 6.8.2); nobody's using XFree86 now.
Debian is still using XFree86 in Sarge and Etch. That kernel version
sounds like Sarge (currently "stable"). The transition to Xorg is
slowly starting. (I hear it is in "unstable".) It doesn't surprise me
that the rather old XFree86 4.3.0 in Sarge doesn't support the newest
graphics cards.
There might be "unofficial" packages of Xorg available for Sarge that
the OP could use, but who knows what else could break (anything
depending on the X libraries is a candidate for trouble).
| |
| Dances With Crows 2005-08-14, 5:53 pm |
| On Sun, 14 Aug 2005 16:10:15 -0400, John-Paul Stewart staggered into the
Black Sun and said:
> Dances With Crows wrote:
> Debian is still using XFree86 in Sarge and Etch.
....Jacob L. said the same thing in the other branch of this thread.
It's things like this that make me keep saying "if you have new
hardware, Debian stable is *not* for you".
> The transition to Xorg is slowly starting. (I hear it is in
> "unstable".)
Well, it's about freakin' time, eh? (I wonder when we'll see the first
X client that refuses to run on XFree86 because Damage and/or Composite
are missing.)
> There might be "unofficial" packages of Xorg available for Sarge that
> the OP could use, but who knows what else could break (anything
> depending on the X libraries is a candidate for trouble).
When I did "emerge -C xfree86 && emerge xorg-x11", none of the KDE/GTK+
apps that had been compiled against XFree86 4.3 broke. No, I didn't
recompile everything; that would've taken 36 hours. The Xorg people
were *serious* about backwards combatability. Heck, the config file
format's the same, the X modules follow the same ABI, the same core
functions are available in Xlib, and all the new functions got put in
extensions. Oh well, Haines can probably use the evil binary-only
nVidia module if he doesn't want to move to Xorg just yet....
--
Matt G|There is no Darkness in Eternity/But only Light too dim for us to see
Brainbench MVP for Linux Admin / mail: TRAP + SPAN don't belong
http://www.brainbench.com / "He is a rhythmic movement of the
-----------------------------/ penguins, is Tux." --MegaHAL
| |
| John-Paul Stewart 2005-08-14, 5:53 pm |
| Dances With Crows wrote:
> On Sun, 14 Aug 2005 16:10:15 -0400, John-Paul Stewart staggered into the
> Black Sun and said:
>
>
>
> ...Jacob L. said the same thing in the other branch of this thread.
> It's things like this that make me keep saying "if you have new
> hardware, Debian stable is *not* for you".
I love Debian/stable on my servers because things don't change. But for
a desktop, I agree that it can get behind the times.
The OP could simply upgrade from Sarge to Etch (which is *easy* to do)
and wait patiently for Xorg to make the move to testing---which could be
as early as a couple of weeks from now or could be much longer.
>
>
> Well, it's about freakin' time, eh? (I wonder when we'll see the first
> X client that refuses to run on XFree86 because Damage and/or Composite
> are missing.)
>
>
>
>
> When I did "emerge -C xfree86 && emerge xorg-x11", none of the KDE/GTK+
> apps that had been compiled against XFree86 4.3 broke. No, I didn't
> recompile everything; that would've taken 36 hours. The Xorg people
> were *serious* about backwards combatability. Heck, the config file
> format's the same, the X modules follow the same ABI, the same core
> functions are available in Xlib, and all the new functions got put in
> extensions.
If it really is that backwards-compatible, it makes me wonder why the
Debian transition is taking so long. (I suppose I could find reasons in
the Debian mailing lists, but for now I'll just wonder....)
At least from what you're saying it should be a fairly smooth
transition. I've heard great things about X.org but I've been reluctant
to poke around for fear of breaking everything. Maybe I don't need to
worry so much. I might have to try compiling it locally and see what
happens, or install the packages from "unstable" on my "testing"
workstation.
| |
| Haines Brown 2005-08-14, 8:47 pm |
| I infer from the thread:
1. Xorg apparently is the ideal way to go, but I'm running debian
sarge (XFree86). I find that there's an unstable deb package for
Xorg at http://debian.linux-systems.com which apparently can be
easily installed (along with xcompmgr and transset and with some
minor tinkering). Would that be the ideal solution?
2. #lsmod shows that my nv driver not even loaded. Does this make
sense?
3. The XFree86 log reported that no symbols found for
libGLcore.a:m_debut-norm.o and clip.o and xfom. and vertex.o The
GLCore was loaded, so I didn't worry about it.
4. An obvious approach short of going for Xorg would be to install the
nVidia X driver from the nVidia site. Is it likely it would work
better than the nv kernel driver? The 64-1.0-7676 driver release
date is 9 August 2005, but couldn't find any specific statement
about whether it supported GeForce 6600.
A conservative approach would be at least to try the nVidia driver,
and only if that didn't work, take the more adventurous step of trying
to insall Xorg. Does this make sense?
--
Haines Brown
KB1GRM
| |
| Bill Marcum 2005-08-14, 8:47 pm |
| On Sun, 14 Aug 2005 16:57:00 -0500, Dances With Crows
<danSPANceswitTRAPhcrows@gmail.com> wrote:
>
> ...Jacob L. said the same thing in the other branch of this thread.
> It's things like this that make me keep saying "if you have new
> hardware, Debian stable is *not* for you".
>
>
> Well, it's about freakin' time, eh? (I wonder when we'll see the first
> X client that refuses to run on XFree86 because Damage and/or Composite
> are missing.)
>
The way Debian works, Xorg won't go into stable until the next major
release, and it was three years between the current stable release
(sarge) and the previous one (woody).
Ubuntu has Xorg.
--
BOFH excuse #426:
internet is needed to catch the etherbunny
| |
| Haines Brown 2005-08-15, 7:47 am |
| I installed the nVidia driver, and that driver allowed me to start
X. But it left some questions.
I went to the nVidia website and downloaded the newest deriver for my
kernel. It is packaged with a self-running installation script. I
started the installation, and as expected, it offered to go to the
nVidia FTP site to download a precompiled kernel module (kernel
interface) if available for my kernel.
Because I am not yet on line with this new machine and knew they
wouldn't have the kernel module for my kernel anyway, I said "No",
intending just to install the nvidia driver and then go through the
long process of getting the kernel module source and compiling it
(long because it means moving a hard disk from one machine to
another).
There is a debian "nvidia-kernel-source" package (there are also
precompiled "nvidia-kernel..." packages, but not for my kernel). I
resigned myself to installing and compiling it, and so said "No" to
the driver installer's offer to get a precompiled kernel module on
line. Because I thought I was now finished installing the driver, I
was not paying much attention, but thought I saw something flash by
quickly about compiling.
Compiling what? So just in case, I tried to start X, and lo and
behold, it started! I never took any explicit steps to install a
kernel module.
Would someone explain this? The nVidia directions seem to assume that
a kernel driver must be installed one way or the other, but I did not
explicitly do it. The computer is not connected on line; I do not have
a nvidia-kernel-source package installed; the nvidia (and agpgart for
some reason) driver is definitely loaded.
Why the agpgart module? On my old machine, where the video card is
AGP, no such module loads; on the new machine, where the video card is
PCI Extra x 16, the module is loaded.
--
Haines Brown
KB1GRM
| |
|
| On Mon, 15 Aug 2005 12:01:58 GMT, a posting issued forth from Haines Brown...
> I installed the nVidia driver, and that driver allowed me to start
> X. But it left some questions.
>
> I went to the nVidia website and downloaded the newest deriver for my
> kernel. It is packaged with a self-running installation script. I
> started the installation, and as expected, it offered to go to the
> nVidia FTP site to download a precompiled kernel module (kernel
> interface) if available for my kernel.
>
> Because I am not yet on line with this new machine and knew they
> wouldn't have the kernel module for my kernel anyway, I said "No",
> intending just to install the nvidia driver and then go through the
> long process of getting the kernel module source and compiling it
> (long because it means moving a hard disk from one machine to
> another).
>
> There is a debian "nvidia-kernel-source" package (there are also
> precompiled "nvidia-kernel..." packages, but not for my kernel). I
> resigned myself to installing and compiling it, and so said "No" to
> the driver installer's offer to get a precompiled kernel module on
> line. Because I thought I was now finished installing the driver, I
> was not paying much attention, but thought I saw something flash by
> quickly about compiling.
>
> Compiling what? So just in case, I tried to start X, and lo and
> behold, it started! I never took any explicit steps to install a
> kernel module.
>
It compiles the kernel module. Wouldn't be much good if it didn't.
> Would someone explain this? The nVidia directions seem to assume that
> a kernel driver must be installed one way or the other, but I did not
> explicitly do it. The computer is not connected on line; I do not have
> a nvidia-kernel-source package installed; the nvidia (and agpgart for
> some reason) driver is definitely loaded.
>
That's what the installer is for.
--
Jacob
mailto:`echo wnpbo@urvqre.ubzryvahk.arg | tr [a-z] [n-za-m]`
| |
| Haines Brown 2005-08-15, 5:49 pm |
| I see the error of my ways: assuming that the nvidia module and the
kernel module were two different modules ;-(
Startx now works with root, but fails with users: The x server says it
is writing the .Xauthority file, and the nVidia splash screen shows up
briefly, but I'm kicked back to the console. ~/.xsession-errors says:
open: Permission denied. So no .Xauthority file is created.
--
Haines Brown
KB1GRM
| |
| Dances With Crows 2005-08-15, 5:49 pm |
| On Mon, 15 Aug 2005 17:34:02 GMT, Haines Brown staggered into the Black
Sun and said:
> I see the error of my ways: assuming that the nvidia module and the
> kernel module were two different modules ;-(
One's an X module and the other is a kernel module. The X module
requires that the kernel module be present for everything to work. This
is all described in the README.
> Startx now works with root, but fails with users: ~/.xsession-errors
> says: open: Permission denied.
The README that you should've gotten with the evil binary-only nVidia
modules says that this is probably a permissions problem with
/dev/nvidia* . Check them and make sure they're 0666 or root:video
0660 and all local users are in the video group.
--
Matt G|There is no Darkness in Eternity/But only Light too dim for us to see
Brainbench MVP for Linux Admin / mail: TRAP + SPAN don't belong
http://www.brainbench.com / "He is a rhythmic movement of the
-----------------------------/ penguins, is Tux." --MegaHAL
| |
| Haines Brown 2005-08-15, 5:49 pm |
| Dances With Crows <danSPANceswitTRAPhcrows@gmail.com> writes:
> On Mon, 15 Aug 2005 17:34:02 GMT, Haines Brown staggered into the Black
> Sun and said:
>
> The README that you should've gotten with the evil binary-only
> nVidia modules says that this is probably a permissions problem with
> /dev/nvidia* . Check them and make sure they're 0666 or root:video
> 0660 and all local users are in the video group.
The /dev/nvidia* files are in fact 666. They are owned by
root:root. Seems an obvious solution to make the owner root:video,
with 660 permissions, and add users to video group. Wonder why that
was not done automatically. Couldn't find this in the README.
--
Haines Brown
KB1GRM
| |
| Dances With Crows 2005-08-15, 5:49 pm |
| On Mon, 15 Aug 2005 18:48:04 GMT, Haines Brown staggered into the Black
Sun and said:
> Dances With Crows <danSPANceswitTRAPhcrows@gmail.com> writes:
> The /dev/nvidia* files are in fact 666. They are owned by
> root:root.
Obviously not the problem, then. Time to check /var/log/Xorg.?.log ,
I think.
> Seems an obvious solution to make the owner root:video, with 660
> permissions, and add users to video group. Wonder why that was not
> done automatically?
Group video may not exist. Adding users to groups requires the users to
log out and log back in again before they are added to those groups.
> Couldn't find this in the README.
# in xorg.conf
Section "DRI"
Group "video"
Mode 0666
EndSection
# end xorg.conf fragment
....you may wish to check and make sure something like that is in there,
then restart X if it wasn't. If DRI permissions aren't set so that
users logging in to X can read and write, then the users won't be able
to use accelerated OpenGL, which sort of defeats the purpose of the evil
binary modules. Also, if your kdm/gdm runs as user "blah" , then make
sure that blah is in the video group too. HTH,
--
Matt G|There is no Darkness in Eternity/But only Light too dim for us to see
Brainbench MVP for Linux Admin / mail: TRAP + SPAN don't belong
http://www.brainbench.com / "He is a rhythmic movement of the
-----------------------------/ penguins, is Tux." --MegaHAL
| |
| Tatu Portin 2005-08-15, 8:47 pm |
| Dances With Crows wrote:
> On Sun, 14 Aug 2005 19:43:13 GMT, Haines Brown staggered into the Black
> Sun and said:
>
>
>
> Hang on. XFree86 or Xorg? Which version? You should be using Xorg
> (current version 6.8.2); nobody's using XFree86 now. Also, the 2.6.8
> and 2.6.9 kernels had a series of stupid problems with burning CDs with
> cdrecord, so if you need to do that, you need 2.6.10 at least.
I'm very sorry if this discussion has already taken place, but why are
people migrating from XFree86 to X.org? I have this picture that the
license change has created some kind of vortex-of-panic that pushes
people away from XFree86. But I don't see anything crucial in the change
they made, and no reason to change from XFree to X.org. So, have all
core XFree programmers gone to X.org camp, or is there another valid
reason to promote X.org?
| |
| Dances With Crows 2005-08-15, 8:47 pm |
| On Mon, 15 Aug 2005 23:32:23 GMT, Tatu Portin staggered into the Black
Sun and said:
> Dances With Crows wrote:
> I'm very sorry if this discussion has already taken place, but why are
> people migrating from XFree86 to X.org?
The license changed, and almost all the developers migrated, and Xorg
introduced a couple of extensions that have very good potential. That's
about it--but that's all you really need.
> license change has created some kind of vortex-of-panic that pushes
> people away from XFree86. But I don't see anything crucial in the
> change they made, and no reason to change from XFree to X.org.
Damage, Composite, Xorg supports some of the very latest hardware that
XFree86 doesn't. That "hardware support" thing was Haines's original
problem, and it'll only get better for Xorg as new hardware gets
introduced and Xorg supports it while XFree... doesn't.
> have all core XFree programmers gone to [the] X.org camp?
Pretty much.
> Or is there another valid reason to promote X.org?
The user base of Xorg is, AFAICT, quite a bit larger than the user base
for XFree86. That means that Xorg (6.8.0 and 6.8.2) is better-tested
than XFree86 4.4. HTH,
--
Matt G|There is no Darkness in Eternity/But only Light too dim for us to see
Brainbench MVP for Linux Admin / mail: TRAP + SPAN don't belong
http://www.brainbench.com / "He is a rhythmic movement of the
-----------------------------/ penguins, is Tux." --MegaHAL
| |
| Haines Brown 2005-08-16, 5:55 pm |
| Dances With Crows <danSPANceswitTRAPhcrows@gmail.com> writes:
> On Mon, 15 Aug 2005 18:48:04 GMT, Haines Brown staggered into the Black
> Sun and said:
[vbcol=seagreen]
>
> Obviously not the problem, then. Time to check /var/log/Xorg.?.log ,
> I think.
Checking the XFree86.0.log was the first thing I did, and the x server
seems to have completed its job successfully. The log for a successful
startx for root is identical to the unsuccessful startx by user.
Since my user of 666 permission for /dev/nvidia* clearly lets anyone
access it, then my impression the "permission" problem meant that
..Xauthority could not be created. The startX messages say it is
being created, but none shows up, and its absense might be why there
is no permission. And, yes, user's home directory permits writing to
it.
>
> Group video may not exist. Adding users to groups requires the users to
> log out and log back in again before they are added to those groups.
Group video does exist. Users were not in the group, but with the
/dev being accessible to the world, I should not need it. Tried it
anyway without success.
>
> # in xorg.conf
> Section "DRI"
> Group "video"
> Mode 0666
> EndSection
> # end xorg.conf fragment
The nVidia READ me says nothing about this. My DRI section also has
"Mode 0666".
Apparently, something goes amiss with the creation of .Xauthority
file. The xserver says it is creating the .Xauthority file, but the
file fails to appear. That there are no indications in the XFree86 log
and that the nVidia spash screen appears very briefly suggest that the
problem is external to the xserver.
I suspect it is xauth that is having the permission problem, and that
the xsession-errors report of a permission problem is the result of
there being no .Xauthority file. It could be encountering locks caused
by xdm or another xauth session. I don't know where to look for such a
lock that I might try to delete a stale lock.
I'd like to create an .Xauthority file with something like:
# xauth -i [displayname protocolname hexkey]
but I don't understand the syntax concerning what should be used for
the options. How is the target user specified?
--
Haines Brown
KB1GRM
| |
| Haines Brown 2005-08-16, 5:55 pm |
| Sorry to append this note.
I find that the command $ xauth list for my running machine and the
machine having the problem has different results.
On my functional machine, I have magic-cookie and hex data for:
teufel.hartford-hwp.com/unix:0
localhost.localdomain:0
hartford-hwp.com/unix:0
teufel/unix:0
teufel.hartford-hwp.com:0
127.0.1.1:0
On my disfunctional machine I have just:
teufel.hartford-hwp.com:0
teufel/unix:0
hostname and domain names are the same on both machines and are
reported by various $ hostname options.
--
Haines Brown
KB1GRM
|
|
|
|
|