Video drivers on a MacBook Pro 5,2

[Henrik Frisk]

Hi,

First of all I wish to thank all the people involved in the release of F13.
I just upgraded from F11 and I really like it. The installation was
effortless and the system works and looks great!

But I have a question WRT the video ddriver. I have successfully installed
the nvidia drivers but for some reason, the resolution (1920x1200) gives me
a 'smaller' screen then what I have under F11 (also nvidia), i.e. I can fit
less things. As if the resolution was lower. BTW, this was the case with the
nouveau drivers too. The video card is a nVidia Corporation G96 [GeForce
9600M GT] [10de:0647]

Any ideas?

/h

[Tom H]

On Wed, Oct 20, 2010 at 3:28 AM, Henrik Frisk  wrote:
> First of all I wish to thank all the people involved in the release of F13.
> I just upgraded from F11 and I really like it. The installation was
> effortless and the system works and looks great!
>
> But I have a question WRT the video ddriver. I have successfully installed
> the nvidia drivers but for some reason, the resolution (1920x1200) gives me
> a 'smaller' screen then what I have under F11 (also nvidia), i.e. I can fit
> less things. As if the resolution was lower. BTW, this was the case with the
> nouveau drivers too. The video card is a nVidia Corporation G96 [GeForce
> 9600M GT] [10de:0647]

This has come up on another list:

"Xorg forces DPI 96 irrespective of the correct DPI detected
by/calculated from EDID."

https://bugs.freedesktop.org/show_bug.cgi?id=23705

[Chris Smart]

On Wed, Oct 20, 2010 at 6:28 PM, Henrik Frisk  wrote:
> Any ideas?

Tom's link is good, but in case it helps, you can check your current
resolution and DPI with:

xdpyinfo |grep -A1 dim

If you want to manually set the DPI to something else (say, 96x96),
edit the driver section of your /etc/X11/xorg.conf, adding:

Option      "UseEdidDpi" "false"
        Option      "DPI" "96 x 96"

If the DPI is fine, and it's the resolution, you should make sure that
you're definitely running the NVIDIA driver, and try using the
graphical tool to set specific resolutions and save as new xorg.conf
file (or add modes yourself).

-c

[Henrik Frisk]

On Wed, Oct 20, 2010 at 11:33 PM, Chris Smart wrote:
> On Wed, Oct 20, 2010 at 6:28 PM, Henrik Frisk  wrote:
> > Any ideas?
>
> Tom's link is good, but in case it helps, you can check your current
> resolution and DPI with:
>
> xdpyinfo |grep -A1 dim
>
> If you want to manually set the DPI to something else (say, 96x96),
> edit the driver section of your /etc/X11/xorg.conf, adding:
>    Option      "UseEdidDpi" "false"
>    Option      "DPI" "96 x 96"
>
> If the DPI is fine, and it's the resolution, you should make sure that
> you're definitely running the NVIDIA driver, and try using the
> graphical tool to set specific resolutions and save as new xorg.conf
> file (or add modes yourself).
>
> Thanks. I'll look into this once I've gotten my nvidia drivers to work.

/h

[Chris Smart]

On Thu, Oct 21, 2010 at 6:21 PM, Henrik Frisk  wrote:
> Thanks. I'll look into this once I've gotten my nvidia drivers to work.

This might help:
http://oliver.net.au/?p=70
http://oliver.net.au/?p=160

I install the NVIDIA drivers using RPMFusion's repo, using the akmod
(so that it's built automatically, unlike kmod).

Remove current xorg.conf file, else the auto-enabling feature will fail.

Something simple like this (for 64bit) should do, as root (from memory):
yum install @development-tools
yum install akmod-nvidia xorg-x11-drv-nvidia-libs.i686
xorg-x11-drv-nvidia-libs.x86_64 time dkms
dracut -f /boot/initramfs-$(uname -r).img $(uname -r)

This will blacklist nouveau, which is why we then re-build your initramfs.

-c

[Henrik Frisk]

On Thu, Oct 21, 2010 at 9:39 AM, Chris Smart wrote:
> On Thu, Oct 21, 2010 at 6:21 PM, Henrik Frisk  wrote:
> > Thanks. I'll look into this once I've gotten my nvidia drivers to work.
>
> This might help:
> http://oliver.net.au/?p=70
> http://oliver.net.au/?p=160
>
> I install the NVIDIA drivers using RPMFusion's repo, using the akmod
> (so that it's built automatically, unlike kmod).
>
> Remove current xorg.conf file, else the auto-enabling feature will fail.
>
> Something simple like this (for 64bit) should do, as root (from memory):
> yum install @development-tools
> yum install akmod-nvidia xorg-x11-drv-nvidia-libs.i686
> xorg-x11-drv-nvidia-libs.x86_64 time dkms
> dracut -f /boot/initramfs-$(uname -r).img $(uname -r)
>
> This will blacklist nouveau, which is why we then re-build your initramfs.
>
> Thanks Chris, this helped! Using the akmod-nvidia worked! The only other

difference (compared to how I did it before) was that I ran nvidia-xconfig
before restarting X.

Thanks again for all the help!

/h

[Henrik Frisk]

On Thu, Oct 21, 2010 at 1:33 PM, Henrik Frisk  wrote:
> On Thu, Oct 21, 2010 at 9:39 AM, Chris Smart 
> wrote:
> > On Thu, Oct 21, 2010 at 6:21 PM, Henrik Frisk  wrote:
> > > Thanks. I'll look into this once I've gotten my nvidia drivers to work.
> >
> > This might help:
> > http://oliver.net.au/?p=70
> > http://oliver.net.au/?p=160
> >
> > I install the NVIDIA drivers using RPMFusion's repo, using the akmod
> > (so that it's built automatically, unlike kmod).
> >
> > Remove current xorg.conf file, else the auto-enabling feature will fail.
> >
> > Something simple like this (for 64bit) should do, as root (from memory):
> > yum install @development-tools
> > yum install akmod-nvidia xorg-x11-drv-nvidia-libs.i686
> > xorg-x11-drv-nvidia-libs.x86_64 time dkms
> > dracut -f /boot/initramfs-$(uname -r).img $(uname -r)
> >
> > This will blacklist nouveau, which is why we then re-build your initramfs.
>
> Thanks Chris, this helped! Using the akmod-nvidia worked! The only other
> difference (compared to how I did it before) was that I ran nvidia-xconfig
> before restarting X.

It turns out my xorg.conf gets corrupt for some reason. Today when I
rebooted the computer wouldn't start (same flickering white bar).
Running nvidia-xconfig solved the problem. I will now make a copy of
my current xorg.conf and compare it to one that is corrupt. Hopefully
that will tell me what the problem is and maybe what causes it.

Should anyone have a better suggestion, please let me know.

/h

[Chris Smart]

On Fri, Oct 22, 2010 at 7:37 PM, Henrik Frisk  wrote:
> Should anyone have a better suggestion, please let me know.

Not sure, but maybe nouveau is also still being loaded? Do you still
get the full resolution, pretty slash on boot up?

You shouldn't, once you install nvidia, blacklist nouveau (done
automatically) and rebuild your initramfs. In fact, I think installing
the nvidia driver now also adds a blacklist option to the kernel line
in grub, as well as adding /etc/modprobe.d/blacklist-nouveau.conf

Everytime you boot, it starts the "nvidia" service, which enables
nvidia by checking/creating a new xorg.conf for you.

If you install the nvidia driver when you already have an xorg.conf in
place, this fails. That's why I suggested you move it out of the way
first.

Maybe you can also check to see that you have 3D working?

glxinfo |grep -i nvidia

-c

[Henrik Frisk]

On Sat, Oct 23, 2010 at 12:54 AM, Chris Smart  wrote:
> On Fri, Oct 22, 2010 at 7:37 PM, Henrik Frisk  wrote:
> > Should anyone have a better suggestion, please let me know.
>
> Not sure, but maybe nouveau is also still being loaded? Do you still
> get the full resolution, pretty slash on boot up?

Yes I do.
> You shouldn't, once you install nvidia, blacklist nouveau (done
> automatically) and rebuild your initramfs. In fact, I think installing
> the nvidia driver now also adds a blacklist option to the kernel line
> in grub, as well as adding /etc/modprobe.d/blacklist-nouveau.conf

It's blacklisted there as well as in the grub.conf, for the right kernel.
> Everytime you boot, it starts the "nvidia" service, which enables
> nvidia by checking/creating a new xorg.conf for you.
>
> If you install the nvidia driver when you already have an xorg.conf in
> place, this fails. That's why I suggested you move it out of the way
> first.

Yes, I removed it. This is not the problem, the problem now is that
something gets corrupted when I'm using nvidia so that upon restart
I'm back to the original problem (X never starts).

This time I made a copy of my xorg.conf right after I had a working
setup with nvidia drivers and a copy of it right before I shut the
computer down. There is no difference between the two files so it
isn't the xorg.conf which is the problem. Also, this time the computer
wouldn't start until I had completely removed (yum remove) all nvidia
drivers and manually removed the blacklistings.

This is really frustrating because I need the nvidia drivers to work.
> Maybe you can also check to see that you have 3D working?

I should obviuosly have checked this before I removed everything but
then again, I thought problem was solved...

Thanks to anyone that can help.

best,

/Henrik

[Chris Smart]

On Mon, Oct 25, 2010 at 9:26 AM, Henrik Frisk  wrote:
> >
>
> Yes I do.

Then I guess either your initramfs still has the nouveau driver (which
will cause problems with NVIDIA), or it doesn't but you have the rhgb
kernel option (which is fine).

Unless I'm mistaken, plymouth will use KMS by default if it has
support for it in the nouveau driver. If it doesn't have that, then it
will fall back to text mode, however you can still get a pretty splash
if you pass the rhgb option to the kernel.

This is what my kernel line has, which gives me a lower-res but still
pretty splash screen without KMS or nouveau:
nomodeset rhgb quiet vga=0x369 nouveau.modeset=0 rdblacklist=nouveau

So, if you're getting a pretty splash but you aren't passing similar
options as above, then your initramfs probably still has nouveau.
> Yes, I removed it. This is not the problem, the problem now is that
> something gets corrupted when I'm using nvidia so that upon restart
> I'm back to the original problem (X never starts).

Well all I can suggest is that you boot to say, run level 3, check
that nvidia is loaded, check that nouveau is NOT loaded, check that
your xorg.conf file is correctly set to use the nvidia driver.

If that's all good but X still doesn't load, read your
/var/log/Xorg.0.log log file to see what X is actually trying to do.
> From my experience, if you remove your xorg.conf file, install the

akmod-nvidia package and libraries (as per previous post), then NVIDIA
is pretty solid.

-c

[Henrik Frisk]

On Mon, Oct 25, 2010 at 2:16 AM, Chris Smart  wrote:
> On Mon, Oct 25, 2010 at 9:26 AM, Henrik Frisk  wrote:
> > >
> >
> > Yes I do.
>
> Then I guess either your initramfs still has the nouveau driver (which
> will cause problems with NVIDIA), or it doesn't but you have the rhgb
> kerning theel option (which is fine).

Maybe I misunderstood your question. I'm not getting a pretty splash
during startup, only the three horizontal bars showing the startup
progress. The few times I've managed to boot with X, with nvidia, I've
been getting the nvidia splash but always only the bars before that.
> Unless I'm mistaken, plymouth will use KMS by default if it has
> support for it in the nouveau driver. If it doesn't have that, then it
> will fall back to text mode, however you can still get a pretty splash
> if you pass the rhgb option to the kernel.
>
> This is what my kernel line has, which gives me a lower-res but still
> pretty splash screen without KMS or nouveau:
> nomodeset rhgb quiet vga=0x369 nouveau.modeset=0 rdblacklist=nouveau
>
> So, if you're getting a pretty splash but you aren't passing similar
> options as above, then your initramfs probably still has nouveau.

dracut is supposed to create a new initramfs for me, right? So if this
is the problem there is something going wrong when I create the new
initramfs, like nouveau hasn't been correctly blacklisted? What can I
check before I run dracut?
> > Yes, I removed it. This is not the problem, the problem now is that
> > something gets corrupted when I'm using nvidia so that upon restart
> > I'm back to the original problem (X never starts).
>
> Well all I can suggest is that you boot to say, run level 3, check
> that nvidia is loaded, check that nouveau is NOT loaded, check that
> your xorg.conf file is correctly set to use the nvidia driver.
>
> If that's all good but X still doesn't load, read your
> /var/log/Xorg.0.log log file to see what X is actually trying to do.

Ok, I'll try to reinstall everything again and see if I can get it right.

Thanks for the help!

/h

[Henrik Frisk]

> > If that's all good but X still doesn't load, read your
> > /var/log/Xorg.0.log log file to see what X is actually trying to do.
>
> Ok, I'll try to reinstall everything again and see if I can get it right.

Ok, so now I've reinstalled the drivers once again. Upon my first
reboot the startup would halt with the flickering white bar. Rebooting
mystically fixed it and I'm definitely running nvidia withour nouveau:

$ lsmod | grep nouveau
$ lsmod | grep nvidia
nvidia               9318413  40
mbp_nvidia_bl           1755  0
i2c_core               20553  3 nvidia,videodev,i2c_nforce2

Looking at xorg.0.log however I find the following errors at the end
(everything else in the log looks alright to me):

[    29.046] RANDR failure: 8 (extension base 174)
[    29.047] 088ae690 088ae694 088ae698 088ae69c
[    29.047] 088ae6a0 088ae6a4 088ae6a8 088ae6ac
[    29.047] 088ae6b0 088ae6b4 088ae6b8 088ae6bc
[    29.048] 088ae6c0 088ae6c4 088ae6c8 088ae6cc
... <loads of ouput similar to the above> ...
[    29.052] 088aec90 088aec94 088aec98
[    29.052] RANDR failure: 8 (extension base 174)
[    29.052] 088b0ac8 088b0acc 088b0ad0 088b0ad4
[    29.052] 088b0ad8 088b0adc 088b0ae0 088b0ae4
[    29.052] 088b0ae8 088b0aec 088b0af0 088b0af4
[    29.052] 088b0af8 088b0afc 088b0b00 088b0b04

When googling the RANDR failure I found this:
http://www.linuxquestions.org/questions/linux-hardware-18/nvidia-video-drive...
made me think that my problems may be related to ABI as well. Hard to
tell now since I don't have the problem now... My graphics card is a
dual head; could that be causing me trouble?

I'm curious to see how long I can run this system now before it fails again...

best,

/Henrik

[Chris Smart]

On Tue, Oct 26, 2010 at 9:47 PM, Henrik Frisk  wrote:
> Ok, so now I've reinstalled the drivers once again. Upon my first
> reboot the startup would halt with the flickering white bar. Rebooting
> mystically fixed it and I'm definitely running nvidia withour nouveau:

I don't know why you're getting that... seems strange. Can you post
the kernel line of your /boot/grub/menu.lst?
> $ lsmod | grep nouveau
> $ lsmod | grep nvidia
> nvidia               9318413  40
> mbp_nvidia_bl           1755  0
> i2c_core               20553  3 nvidia,videodev,i2c_nforce2

OK, so that means that your system does not have the nouveau driver
(your initramfs still might).
> Looking at xorg.0.log however I find the following errors at the end
> (everything else in the log looks alright to me):

Hmm.. I doubt that it's dualhead, you would have to configure that in
your xorg.conf before it worked with the NVIDIA driver.

Can you post this?
cat /etc/X11/xorg.conf |grep -i driver
> I'm curious to see how long I can run this system now before it fails again...

Yeah, well let us know :-) What I've mentioned works for me on many
systems... so I would have thought it would work for you too..

-c

[Henrik Frisk]

On Tue, Oct 26, 2010 at 1:41 PM, Chris Smart  wrote:
> On Tue, Oct 26, 2010 at 9:47 PM, Henrik Frisk  wrote:
> > Ok, so now I've reinstalled the drivers once again. Upon my first
> > reboot the startup would halt with the flickering white bar. Rebooting
> > mystically fixed it and I'm definitely running nvidia withour nouveau:
>
> I don't know why you're getting that... seems strange. Can you post
> the kernel line of your /boot/grub/menu.lst?

Well, see below. Now I've already uninstalled nvidia...
> > $ lsmod | grep nouveau
> > $ lsmod | grep nvidia
> > nvidia               9318413  40
> > mbp_nvidia_bl           1755  0
> > i2c_core               20553  3 nvidia,videodev,i2c_nforce2
>
> OK, so that means that your system does not have the nouveau driver
> (your initramfs still might).
> > Looking at xorg.0.log however I find the following errors at the end
> > (everything else in the log looks alright to me):
>
> Hmm.. I doubt that it's dualhead, you would have to configure that in
> your xorg.conf before it worked with the NVIDIA driver.
>
> Can you post this?
> cat /etc/X11/xorg.conf |grep -i driver

I'm sorry I didn't see your post before I uninstalled nvidia. But I
had created a fresh xorg.conf via nvidia-xconfig and I'm certain it
used nvidia.

/h

[Henrik Frisk]

> I'm curious to see how long I can run this system now before it fails again...

That didn't take so long :(

In Xorg.0.log it says, after all modules have been loaded, "Failed to
initilalize the NVIDIA graphics device PCI:2:0:0"

In /var/log/messages I find: "kernel: nvidia: probe of 000:00:03.5
failed with error -1"

Now I need to ditch the nvidia drivers for now because I desperately
need to get into my system...

/h

[Henrik Frisk]

On Tue, Oct 26, 2010 at 2:11 PM, Henrik Frisk  wrote:
> > I'm curious to see how long I can run this system now before it fails again...
>
> That didn't take so long :(
>
> In Xorg.0.log it says, after all modules have been loaded, "Failed to
> initilalize the NVIDIA graphics device PCI:2:0:0"
>
> In /var/log/messages I find: "kernel: nvidia: probe of 000:00:03.5
> failed with error -1"

Searching for the error I came across the nvidia README [1] which,
under chapter 9 "Known Issues", in the section "Kernel virtual address
space exhaustion on the X86 platform" describes the error I'm
experiencing. The solution is to pass vmalloc=192MB (or 256MB) to the
kernel boot options. If i get the time tomorrow I will try it and
report if it works.

[1] http://us.download.nvidia.com/XFree86/Linux-x86/169.12/README/chapter-09.htm...

/h

[Chris Smart]

On Wed, Oct 27, 2010 at 6:26 AM, Henrik Frisk  wrote:
> Searching for the error I came across the nvidia README [1] which,
> under chapter 9 "Known Issues", in the section "Kernel virtual address
> space exhaustion on the X86 platform" describes the error I'm
> experiencing. The solution is to pass vmalloc=192MB (or 256MB) to the
> kernel boot options. If i get the time tomorrow I will try it and
> report if it works.

OK, good luck! It's out of my depth now :-) The only other thing I was
going to suggest was to make sure that version of the driver supports
your card.

Damn proprietary drivers..

-c

[Henrik Frisk]

On Wed, Oct 27, 2010 at 12:06 AM, Chris Smart  wrote:
> On Wed, Oct 27, 2010 at 6:26 AM, Henrik Frisk  wrote:
> > Searching for the error I came across the nvidia README [1] which,
> > under chapter 9 "Known Issues", in the section "Kernel virtual address
> > space exhaustion on the X86 platform" describes the error I'm
> > experiencing. The solution is to pass vmalloc=192MB (or 256MB) to the
> > kernel boot options. If i get the time tomorrow I will try it and
> > report if it works.
>
> OK, good luck! It's out of my depth now :-) The only other thing I was
> going to suggest was to make sure that version of the driver supports
> your card.
>
> Damn proprietary drivers..

Indeed. However, the solution described in the nvidia readme appears
to haves solved my problem. I simply added vmalloc=192MB to the kernel
arguments and I've been able to rebbot the machine, with nvidia
drivers enabled, several time now. Keeping my fingers crossed...

If it stays good I will put up some instructions on my blog about the
procedure to make it work.

Thanks for all the help so far!

best,

/Henrik

[Joe Zeff]

On 10/27/2010 03:18 AM, Henrik Frisk wrote:
> If it stays good I will put up some instructions on my blog about the
> procedure to make it work.

Even better, put your instructions on fedoraforum.org so that even more 
fedora users will see them.

[Henrik Frisk]

On Wed, Oct 27, 2010 at 6:59 PM, Joe Zeff  wrote:
> On 10/27/2010 03:18 AM, Henrik Frisk wrote:
> > If it stays good I will put up some instructions on my blog about the
> > procedure to make it work.
>
> Even better, put your instructions on fedoraforum.org so that even more
> fedora users will see them.

Of course, will do!

/h

[Chris Smart]

On Wed, Oct 27, 2010 at 9:18 PM, Henrik Frisk  wrote:
> If it stays good I will put up some instructions on my blog about the
> procedure to make it work.

Great, good to know it's working for you :-)
> Thanks for all the help so far!

No worries. Sorry I didn't know about that particular issue ;-)

-c

[Henrik Frisk]

On Wed, Oct 20, 2010 at 11:33 PM, Chris Smart  wrote:
> On Wed, Oct 20, 2010 at 6:28 PM, Henrik Frisk  wrote:
> > Any ideas?
>
> Tom's link is good, but in case it helps, you can check your current
> resolution and DPI with:
>
> xdpyinfo |grep -A1 dim
>
> If you want to manually set the DPI to something else (say, 96x96),
> edit the driver section of your /etc/X11/xorg.conf, adding:
>
> Option      "UseEdidDpi" "false"
>        Option      "DPI" "96 x 96"

This is just to confirm that this did the trick. Thanks for the tip!

/h