Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]

Groups > linux.debian.user > #256262 > unrolled thread

nvidia package wrong documentation?

Back to article view | Back to linux.debian.user

Contents

  nvidia package wrong documentation? Hans <hans.ullrich@loop.de> - 2023-03-22 11:00 +0100
    Re: nvidia package wrong documentation? David Wright <deblis@lionunicorn.co.uk> - 2023-03-22 15:50 +0100
    Re: nvidia package wrong documentation? Vincent Lefevre <vincent@vinc17.net> - 2023-03-23 14:50 +0100
      Re: ID problem: "the nouveau driver" (was: nvidia package wrong  documentation?) Felix Miata <mrmazda@earthlink.net> - 2023-03-23 18:40 +0100
        Re: ID problem: "the nouveau driver" (was: nvidia package wrong  documentation?) Vincent Lefevre <vincent@vinc17.net> - 2023-03-24 04:00 +0100
          Re: ID problem: "the nouveau driver" Felix Miata <mrmazda@earthlink.net> - 2023-03-24 04:50 +0100
            Re: ID problem: "the nouveau driver" debian-user@howorth.org.uk - 2023-03-24 12:10 +0100
              Re: ID problem: "the nouveau driver" Felix Miata <mrmazda@earthlink.net> - 2023-03-24 17:20 +0100
                Re: ID problem: "the nouveau driver" debian-user@howorth.org.uk - 2023-03-24 21:30 +0100

#256262 — nvidia package wrong documentation?

[Multipart message — attachments visible in raw view] — view raw

Hi all,

I am running a Lenovo T520 with two graphic cards on board: Intel an Nvidia.

As you see here it is an GF119M [Quadro NVS 4200M] (rev a1)

--- snip ---
lspci | grep VGA 
00:02.0 VGA compatible controller: Intel Corporation 2nd Generation Core Processor Family 
Integrated Gra
phics Controller (rev 09) 
01:00.0 VGA compatible controller: NVIDIA Corporation GF119M [Quadro NVS 4200M] (rev a1)
--- snap ---

The documentation requires the nvidia-legacy-*390xx* driver to use, but as hard as I try, this 
will not work.

However, what is working is the driver nvidia-legacy-*340xx.*

(I am using bumblebee)

This one is working like a charm, but can not be build with kernels newer than 5.10-18.*. So I 
have to stay with this kernel.

What I believe is, that the documentation from NVidia might be wrong and Nvidia told wrong, to 
say 390xx is for NVS4200. 

On the other hand, it might be a bug, that the 340xx can not be build with newer kernels (I 
discovered the compilation is crashing due to some missing files. I filed a bugreport of this 
almost a year ago), which inhibits users, to use newer kernel-versions.

This is a pity especially for notebooks, where you can not exchange the GPU.

I just wanted to inform you, if someone got into the same trouble as me, wondering, why it does 
not run.

Thanks for reading this.

Oh, before someone tells: I do NOT want to use nouveau!

Best regards

Hans
 

 

[toc] | [next] | [standalone]

#256267

On Wed 22 Mar 2023 at 10:50:30 (+0100), Hans wrote:
> 
> I am running a Lenovo T520 with two graphic cards on board: Intel an Nvidia.
> 
> As you see here it is an GF119M [Quadro NVS 4200M] (rev a1)
> 
> --- snip ---
> lspci | grep VGA 
> 00:02.0 VGA compatible controller: Intel Corporation 2nd Generation Core Processor Family 
> Integrated Gra
> phics Controller (rev 09) 
> 01:00.0 VGA compatible controller: NVIDIA Corporation GF119M [Quadro NVS 4200M] (rev a1)
> --- snap ---
> 
> The documentation requires the nvidia-legacy-*390xx* driver to use, but as hard as I try, this 
> will not work.
> 
> However, what is working is the driver nvidia-legacy-*340xx.*
> 
> (I am using bumblebee)

Presumably not a reference to the malware of that name.

> This one is working like a charm, but can not be build with kernels newer than 5.10-18.*. So I 
> have to stay with this kernel.
> 
> What I believe is, that the documentation from NVidia might be wrong and Nvidia told wrong, to 
> say 390xx is for NVS4200. 
> 
> On the other hand, it might be a bug, that the 340xx can not be build with newer kernels (I 
> discovered the compilation is crashing due to some missing files. I filed a bugreport of this 
> almost a year ago), which inhibits users, to use newer kernel-versions.
> 
> This is a pity especially for notebooks, where you can not exchange the GPU.
> 
> I just wanted to inform you, if someone got into the same trouble as me, wondering, why it does 
> not run.
> 
> Thanks for reading this.
> 
> Oh, before someone tells: I do NOT want to use nouveau!

In view of your earlier statements "And yes, tell me paranoid - I am!",
perhaps you should read:

https://wiki.debian.org/NvidiaGraphicsDrivers#line-58

Cheers,
David.

[toc] | [prev] | [next] | [standalone]

#256303

On 2023-03-22 10:50:30 +0100, Hans wrote:
> What I believe is, that the documentation from NVidia might be wrong
> and Nvidia told wrong, to say 390xx is for NVS4200.

I don't know for your case, but there are inconsistencies in the
Nvidia pages. For instance, for Quadro K610M, the 470.161.03 page
https://www.nvidia.com/Download/driverResults.aspx/194637/en-us/
does not list is as supported by 470.xx (and it is not listed at
all on
https://us.download.nvidia.com/XFree86/Linux-x86_64/470.161.03/README/supportedchips.html),
but both
https://us.download.nvidia.com/XFree86/Linux-x86_64/520.56.06/README/supportedchips.html
and
https://us.download.nvidia.com/XFree86/Linux-x86_64/525.89.02/README/supportedchips.html
list it as supported by 470.xx. I'm not sure what to do, but
anyway, Debian considered that 470.xx did not support it.

> On the other hand, it might be a bug, that the 340xx can not be build with newer kernels (I 
> discovered the compilation is crashing due to some missing files. I filed a bugreport of this 
> almost a year ago), which inhibits users, to use newer kernel-versions.

New kernels (and new versions of the X server) introduce
incompatibilities, and then Nvidia corrects their drivers.
Unfortunately for you, 340.xx is no longer supported by
Nvidia, and 390.xx isn't supported either.

> This is a pity especially for notebooks, where you can not exchange
> the GPU.

Yes, both my laptop and my desktop machine (at my lab) from 2015
are concerned (even for the desktop machine, it will be easier
to change it, though I could still use the old machine remotely
for various kinds of work).

> Oh, before someone tells: I do NOT want to use nouveau!

Though I would prefer free software, nouveau is not usable for me,
in particular with my laptop. Remember...

  https://lists.debian.org/debian-user/2020/05/msg00464.html

The developers did not care to look at my bug reports (this one
and older ones).

-- 
Vincent Lefèvre <vincent@vinc17.net> - Web: <https://www.vinc17.net/>
100% accessible validated (X)HTML - Blog: <https://www.vinc17.net/blog/>
Work: CR INRIA - computer arithmetic / AriC project (LIP, ENS-Lyon)

[toc] | [prev] | [next] | [standalone]

#256310 — Re: ID problem: "the nouveau driver" (was: nvidia package wrong documentation?)

Vincent Lefevre composed on 2023-03-23 14:48 (UTC+0100):

> On 2023-03-22 10:50:30 +0100, Hans wrote:

>> Oh, before someone tells: I do NOT want to use nouveau!

> Though I would prefer free software, nouveau is not usable for me,
> in particular with my laptop. Remember...

>   https://lists.debian.org/debian-user/2020/05/msg00464.html

> The developers did not care to look at my bug reports (this one
> and older ones).

Quoting from the above URI:

[quote]
I use Debian/unstable with the nouveau driver
[/quote]

It could be the exhibition of ignorance represented by the above quote leads
triagers and developers to ignore many problem reports. There is no such thing as
"the nouveau driver" in Debian:

1-Each kernel provides *a* unique nouveau device module (driver) specifically for
that kernel version to provide modesetting services (KMS) and other DRM services.

2-X requires *a* display driver, which in the case of NVidia GPUs depends on KMS
functionality provided by a nouveau kernel device driver for optimal performance.
One such display driver is named nouveau, and is provided by
xserver-xorg-video-nouveau. The Xorg server .deb provides an alternative driver
named modesetting. The modesetting driver:

A-is a newer display driver technology;
B-does not require reverse-engineering;
C-is the upstream default display driver;
D-supports all GPUs for which a kernel device driver providing KMS is available;
E-is automatically employed when applicable xserver-xorg-video-* is not installed.

3-libdrm-nouveau2 provides a userspace interface to nouveau-specific kernel DRM
services: /usr/lib/x86_64-linux-gnu/libdrm_nouveau.so.2

4-libgl1-mesa-dri provides
  /usr/lib/x86_64-linux-gnu/dri/nouveau_dri.so
  /usr/lib/x86_64-linux-gnu/dri/nouveau_vieux_dri.so

5-When xserver-xorg-video-nouveau is not installed, NVidia users may see the
following:
# grep veau /var/log/Xorg.0.log
[    39.971] (II) modeset(0): [DRI2]   DRI driver: nouveau
[    39.971] (II) modeset(0): [DRI2]   VDPAU driver: nouveau
[    39.989] (II) AIGLX: Loaded and initialized nouveau
# grep odese /var/log/Xorg.0.log | grep -v 'set(0)'
[    34.877] (==) Matched modesetting as autoconfigured driver 0
[    34.877] (II) LoadModule: "modesetting"
[    34.904] (II) Loading /usr/lib/xorg/modules/drivers/modesetting_drv.so
[    34.936] (II) Module modesetting: vendor="X.Org Foundation"
[    34.966] (II) modesetting: Driver for Modesetting Kernel Drivers: kms
# inxi -S --vs
inxi 3.3.25-00 (2023-02-07)
System:
  Host: hp945 Kernel: 5.10.0-21-amd64 arch: x86_64 bits: 64 Desktop: Trinity
    v: R14.0.13 Distro: Debian GNU/Linux 11 (bullseye)
# inxi -Gaz
Graphics:
  Device-1: NVIDIA GT218 [GeForce 210] vendor: eVga.com. driver: nouveau
    v: kernel non-free: series: 340.xx status: legacy (EOL) last:
    release: 340.108 kernel: 5.4 xorg: 1.20 arch: Tesla process: 40-80nm
    built: 2006-13 pcie: gen: 1 speed: 2.5 GT/s lanes: 16 ports:
    active: DVI-I-1,HDMI-A-1 empty: VGA-1 bus-ID: 01:00.0 chip-ID: 10de:0a65
    class-ID: 0300 temp: 67.0 C
  Display: x11 server: X.Org v: 1.20.11 driver: X: loaded: modesetting
    unloaded: fbdev,vesa dri: nouveau gpu: nouveau display-ID: :0 screens: 1
  Screen-1: 0 s-res: 3600x1200 s-dpi: 120 s-size: 762x254mm (30.00x10.00")
    s-diag: 803mm (31.62")
  Monitor-1: DVI-I-1 pos: right model: Dell P2213 serial: <filter>
    built: 2012 res: 1680x1050 hz: 60 dpi: 90 gamma: 1.2
    size: 473x296mm (18.62x11.65") diag: 558mm (22") ratio: 16:10 modes:
    max: 1680x1050 min: 720x400
  Monitor-2: HDMI-A-1 mapped: HDMI-1 pos: primary,left model: NEC EA243WM
    serial: <filter> built: 2011 res: 1920x1200 hz: 60 dpi: 94 gamma: 1.2
    size: 519x324mm (20.43x12.76") diag: 612mm (24.1") ratio: 16:10 modes:
    max: 1920x1200 min: 640x480
  API: OpenGL v: 3.3 Mesa 20.3.5 renderer: NVA8 direct-render: Yes

Which nouveau is "the" nouveau???

Has use of the modesetting display driver been tested by the user with a "nouveau"
problem?

PS: Q: Why does a nouveau display driver exist when the default modesetting
display driver exists?
A: Because being newer, it doesn't support ancient hardware that may be supported
by other drivers, and determination of the optimal display driver to install would
be a highly complicated process that would severely bloat distribution
installation programs and yet be unreliable. Therefore, inclusion of
xserver-xorg-video-* is part of most installations. The option remains to use
/etc/X11/xorg.con* to specify a particular display driver rather than depending on
whether or not a particular display driver .deb is installed.
-- 
Evolution as taught in public schools is, like religion,
	based on faith, not based on science.

 Team OS/2 ** Reg. Linux User #211409 ** a11y rocks!

Felix Miata

[toc] | [prev] | [next] | [standalone]

#256317 — Re: ID problem: "the nouveau driver" (was: nvidia package wrong documentation?)

On 2023-03-23 13:38:05 -0400, Felix Miata wrote:
> It could be the exhibition of ignorance represented by the above
> quote leads triagers and developers to ignore many problem reports.
> There is no such thing as "the nouveau driver" in Debian:

This is not true. From old Xorg logs:

[194415.365] (II) NOUVEAU driver Date:   Fri Apr 21 14:41:17 2017 -0400
[194415.365] (II) NOUVEAU driver for NVIDIA chipset families :

> 2-X requires *a* display driver, which in the case of NVidia GPUs depends on KMS
> functionality provided by a nouveau kernel device driver for optimal performance.
> One such display driver is named nouveau, and is provided by
> xserver-xorg-video-nouveau. The Xorg server .deb provides an alternative driver
> named modesetting. The modesetting driver:
> 
> A-is a newer display driver technology;
> B-does not require reverse-engineering;
> C-is the upstream default display driver;
> D-supports all GPUs for which a kernel device driver providing KMS is available;
> E-is automatically employed when applicable xserver-xorg-video-* is not installed.

https://nouveau.freedesktop.org/DebianPackages.html recommends
xserver-xorg-video-nouveau.

However, it seems that a "modesetting" module is loaded by default:

[685925.554] (II) LoadModule: "modesetting"
[685925.554] (II) Loading /usr/lib/xorg/modules/drivers/modesetting_drv.so
[685925.554] (II) Module modesetting: vendor="X.Org Foundation"
[685925.554]    compiled for 1.18.1, module version = 1.18.1
[685925.554]    Module class: X.Org Video Driver
[685925.554]    ABI class: X.Org Video Driver, version 20.0
[...]
[685925.555] (II) modesetting: Driver for Modesetting Kernel Drivers: kms

This is from

  https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=814902

(a different machine, but basically with the same config).

It seems that I haven't kept any Xorg log for the concerned machine.

-- 
Vincent Lefèvre <vincent@vinc17.net> - Web: <https://www.vinc17.net/>
100% accessible validated (X)HTML - Blog: <https://www.vinc17.net/blog/>
Work: CR INRIA - computer arithmetic / AriC project (LIP, ENS-Lyon)

[toc] | [prev] | [next] | [standalone]

#256318 — Re: ID problem: "the nouveau driver"

Vincent Lefevre composed on 2023-03-24 03:57 (UTC+0100):

> On 2023-03-23 13:38:05 -0400, Felix Miata wrote:

>> It could be the exhibition of ignorance represented by the above
>> quote leads triagers and developers to ignore many problem reports.
>> There is no such thing as "the nouveau driver" in Debian:

> This is not true. From old Xorg logs:

> [194415.365] (II) NOUVEAU driver Date:   Fri Apr 21 14:41:17 2017 -0400
> [194415.365] (II) NOUVEAU driver for NVIDIA chipset families :

That is _A_ nouveau driver, one of several, among which:
/lib/modules/<version>/kernel/drivers/gpu/drm/nouveau/nouveau.ko
/usr/lib/xorg/modules/drivers/nouveau_drv.so
/usr/lib/x86_64-linux-gnu/libdrm_nouveau.so.2
/usr/lib/x86_64-linux-gnu/dri/nouveau_dri.so
/usr/lib/x86_64-linux-gnu/dri/nouveau_vieux_dri.so

Thus there can be no "the" nouveau driver without more information than usually
presented along with string "the".

>> 2-X requires *a* display driver, which in the case of NVidia GPUs depends on KMS
>> functionality provided by a nouveau kernel device driver for optimal performance.

> https://nouveau.freedesktop.org/DebianPackages.html recommends
> xserver-xorg-video-nouveau.

Of course they do. N.I.H. Check the history on the page and you will probably find
that language dates back to before the modesetting DIX driver was born, and
probably before KMS too. I think I already did that check last week. The
modesetting DIX and KMS changed a lot in the many years since.

> However, it seems that a "modesetting" module is loaded by default:

If no appropriate DDX driver is available, absent specific configuration to do
otherwise, the modesetting DIX loads and stays loaded, as long as an appropriate
KMS module is loaded and functional, which *.modeset=0 and nomodeset disable.

<https://www.linuxquestions.org/questions/blog/mrmazda-1035595/amd-intel-and-nvidia-x-graphics-driver-primer-38306/>
-- 
Evolution as taught in public schools is, like religion,
	based on faith, not based on science.

 Team OS/2 ** Reg. Linux User #211409 ** a11y rocks!

Felix Miata

[toc] | [prev] | [next] | [standalone]

#256319 — Re: ID problem: "the nouveau driver"

Felix Miata <mrmazda@earthlink.net> wrote:

> <https://www.linuxquestions.org/questions/blog/mrmazda-1035595/amd-intel-and-nvidia-x-graphics-driver-primer-38306/>

"You are not logged in or you do not have permission to access this
page."

?? why should I need an account to see this page?

[toc] | [prev] | [next] | [standalone]

#256327 — Re: ID problem: "the nouveau driver"

debian-user@howorth.org.uk composed on 2023-03-24 11:02 (UTC):

> Felix Miata <mrmazda@earthlink.net> wrote:

>> <https://www.linuxquestions.org/questions/blog/mrmazda-1035595/amd-intel-and-nvidia-x-graphics-driver-primer-38306/>

> "You are not logged in or you do not have permission to access this
> page."

> ?? why should I need an account to see this page?

You would have to login and ask the webmaster. :( I've submitted your question to
one of its admins.

This is its content, last updated 2021-09-14
[quote]
"Drivers" for the subject brands of GPUs for X, whether Wayland or Xorg, consist
of multiple components or layers. This primer discusses only the first two layers,
and only for the FOSS types included in the installation media for most GNU Linux
distributions and their standard FOSS repositories.

First layer is the kernel. Each kernel comes with one or more of its own modules
for each brand of GPU. For AMD, there is amdgpu for current and recent
GPUs/APUs/IGPs, and radeon for older. For Intel there is gma500 for many SOC
configurations, and i915 for the rest. For NVidia there is only nouveau. These
kernel modules provide Kernel ModeSetting (KMS) functionality that the
foundational X driver layer unconditionally depends upon.

These foundational X drivers are traditionally hardware device dependent, and thus
are called DDX, for device dependent X (driver). Again, for AMD there is amdgpu
for current and radeon for older; for Intel there is only intel; and for NVidia
there is only nouveau. Each of these is provided by an upstream optional
xf86-video-* source package. However, there is a newer technology device
independent X driver, now more than six years old, that functions as if it was
device dependent, but which is in fact created for supporting AMD, Intel, NVidia
(and others), not surprisingly, called modesetting. This modesetting DIX is
provided by the X server source package rather than a separate one, so is not
optional, unlike the xf86-video-* packages. Because it is relatively new, older
GPUs (roughly prior to Gen4 for Intel, or 2007 or 2008 for others) are not
supported by it.

Even though upstream modesetting is the default X driver, and the xf86-video-*
packages are technically optional, most Linux installations include a metapackage
that installs every one of the xf86-video-* packages. This metapackage causes
older GPUs unsupported by the modesetting DIX to be supported by a default
installation, which in effect in many cases preempts the upstream default
modesetting DIX that supports most non-ancient GPUs that are not too new for
standard support.

Because of the prevalence of this preemptive override, most people speaking of
video "drivers" don't even know the modesetting DIX exists. And because of the
similar naming for kernel modules and DDXes, discussion of graphics "drivers" is
often lacking clarity. This is particularly true for those for NVidia GPUs.
Discussion of kernel driver is often intended when technically speaking of a DDX
or DIX, and vice versa.

Those who wish to employ the modesetting DIX can most easily do so by uninstalling
the video DDX metapackage and whichever discrete video package is applicable to
the installed GPU(s). Whether the modesetting DIX will perform "better" than a
GPU-specific video package can only be determined by trying both and comparing
behavior. It is not unusual that tearing or other buggy behavior can be eliminated
or reduced by a switch between DDX and DIX or vice versa.

For X there is additionally a pair of limited functionality drivers that do not
depend on KMS that support AMD, Intel, NVidia and more. These are FBDEV and VESA.
These are unaccelerated, and support few, if any, modern widescreen resolutions,
typically limited to 4:3 modes 1280x1024, 1600x1200, or less, often only 1024x768.
Most find these acceptable only for rescue operations, or for servers where X is
in use little to none. Not infrequently it is one of these two that are required
for installation to proceed. When this happens, usually the bootloader on the
installed system is configured to include nomodeset on each of the linux lines.
The result of nomodeset in the bootloader is the competent DDX's and DIX's support
is precluded, making the user unhappy with graphics performance, or confronting
her with a black screen where a login greeter is expected until it is discovered
why and how to remove nomodeset from the boot process.

In summary, "drivers" required for competent X operation are:

    amdgpu (kernel) plus amdgpu (DDX) or modesetting (DIX) for current and
non-ancient AMD APUs/GPUs/IGPs
    radeon (kernel) plus modesetting (DIX) or radeon (DDX) for ancient AMD/ATI
GPUs/IGPs
    gma500 (kernel) or i915 (kernel) plus intel (DDX) or modesetting (DIX) for
Intel GPUs/IGPs
    nouveau (kernel) plus modesetting (DIX) or nouveau (DDX) for NVidia GPUs
[/quote]
-- 
Evolution as taught in public schools is, like religion,
	based on faith, not based on science.

 Team OS/2 ** Reg. Linux User #211409 ** a11y rocks!

Felix Miata

[toc] | [prev] | [next] | [standalone]

#256332 — Re: ID problem: "the nouveau driver"

Felix Miata <mrmazda@earthlink.net> wrote:
> debian-user@howorth.org.uk composed on 2023-03-24 11:02 (UTC):
> 
> > Felix Miata <mrmazda@earthlink.net> wrote:  
> 
> >> <https://www.linuxquestions.org/questions/blog/mrmazda-1035595/amd-intel-and-nvidia-x-graphics-driver-primer-38306/>  
> 
> > "You are not logged in or you do not have permission to access this
> > page."  
> 
> > ?? why should I need an account to see this page?  
> 
> You would have to login and ask the webmaster. :( I've submitted your
> question to one of its admins.
> 
> This is its content, last updated 2021-09-14
> [quote]
[snip]
 [/quote]

Thanks, Felix :)

[toc] | [prev] | [standalone]

Back to top | Article view | linux.debian.user

csiph-web