The Latest NVIDIA Linux Driver Release.

abga · 01-09-2018, 09:24 PM

Nvidia has just released patched drivers against Spectre related attacks:
https://nvidia.custhelp.com/app/answ...or-speculative

mats_b_tegner · 01-18-2018, 04:52 PM

The nVidia legacy driver has been updated to version 340.106. The driver needs the following patch for kernel 4.11 and above:

Code:

--- kernel/uvm/nvidia_uvm_lite.c	2017-09-27 13:50:46.334075042 +0200
+++ kernel/uvm/nvidia_uvm_lite.c	2017-09-27 13:56:06.358041280 +0200
@@ -818,7 +818,11 @@
 }
 
 #if defined(NV_VM_OPERATIONS_STRUCT_HAS_FAULT)
+#if LINUX_VERSION_CODE < KERNEL_VERSION(4, 11, 0)
 int _fault(struct vm_area_struct *vma, struct vm_fault *vmf)
+#else 	
+int _fault(struct vm_fault *vmf) 	
+#endif
 {
 #if defined(NV_VM_FAULT_HAS_ADDRESS)
     unsigned long vaddr = vmf->address;
@@ -828,7 +832,11 @@
     struct page *page = NULL;
     int retval;
 
+#if LINUX_VERSION_CODE < KERNEL_VERSION(4, 11, 0)
     retval = _fault_common(vma, vaddr, &page, vmf->flags);
+#else
+    retval = _fault_common(NULL, vaddr, &page, vmf->flags);
+#endif
 
     vmf->page = page;
 
@@ -866,7 +874,11 @@
 // it's dealing with anonymous mapping (see handle_pte_fault).
 //
 #if defined(NV_VM_OPERATIONS_STRUCT_HAS_FAULT)
+#if LINUX_VERSION_CODE < KERNEL_VERSION(4, 11, 0)
 int _sigbus_fault(struct vm_area_struct *vma, struct vm_fault *vmf)
+#else
+int _sigbus_fault(struct vm_fault *vmf)
+#endif
 {
     vmf->page = NULL;
     return VM_FAULT_SIGBUS;

https://www.nvidia.com/Download/driv...x/130041/en-us
https://www.nvidia.com/Download/driv...x/130042/en-us

Edit:
The patch works on either the .run-file or it can be added to the nvidia-legacy340-kernel SlackBuild-script:

Code:

sh $CWD/$SRCNAM.run --extract-only || exit 1
cp $CWD/kernel-4.11.patch $SRCNAM
cd $SRCNAM || exit 1
patch -p0 < kernel-4.11.patch

Mats

cwizardone · 01-29-2018, 11:31 AM

The latest "Long Lived Branch" NVidia driver, version 390.25, is now available,

http://www.nvidia.com/download/drive...x/130646/en-us

Release highlights,

Quote:

Added support for the following GPUs:
GeForce GTX 1060 5GB
Quadro P620
Fixed a regression introduced in 390.12 that prevented displays from working normally when running multiple X screens with emulated overlays.
Fixed a regression introduced in 390.12 that caused occasional hangs and hard lockup messages in the system log when screen transformations are in use.
Added new application profile settings, "EGLVisibleDGPUDevices" and "EGLVisibleTegraDevices", to control which discrete and Tegra GPU devices, respectively, may be enumerated by EGL. See the "Application Profiles" appendix of the driver README for more details.
Corrected the SONAME of the copy of the libnvidia-egl-wayland library included in the .run installer package to libnvidia-egl-wayland.so.1. The SONAME had previously been versioned incorrectly with the full version number of the library.
Updated nvidia.ko to veto the ACPI_VIDEO_NOTIFY_PROBE event on kernels that allow the handler for this event to be overridden, to improve interaction between the NVIDIA driver and acpi_video on display hotplug events.
Updated the SLI Mosaic layout page in the nvidia-settings control panel to support topologies with up to 32 displays.
Fixed a bug that prevented Xinerama Info from being handled properly in SLI or Base Mosaic layouts with more than 24 displays.
Updated the X driver's composition pipeline (used for rotation, warp and blend, transformation matrices, etc) to also support stereo.
Added an OpenGL stereo preview feature to the screen page in nvidia-settings.
Fixed a bug where GetTexSubImage() would read incorrect data into a pixel buffer object when supplied with a target of GL_TEXTURE_1D_ARRAY and a non-zero yoffset value.
Added support for generic active stereo with in-band DisplayPort signaling. The X configuration option "InbandStereoSignaling" is deprecated in favor of this stereo mode. See "Appendix B. X Config Options" in the README for more information.
Modified the driver to avoid restoring framebuffer console modes on virtual reality head-mounted displays.
Fixed a bug which could cause X servers that export a Video Driver ABI earlier than 0.8 to crash when running X11 applications which call XRenderAddTraps().

Paulo2 · 01-29-2018, 12:31 PM

Quote:

Originally Posted by cwizardone

The latest "Long Lived Branch" NVidia driver, version 390.25, is now available,

http://www.nvidia.com/download/drive...x/130646/en-us

Release highlights,

Thanks for the post, I will give it a try. I'm using right now 387.34 and the system is rock solid for a month now (GTX 960).

I had some problems with the NVidia driver in -stable since August last year,
some daily freezes with 'GPU has fallen off the bus' messages.
I changed to -current but didn't fix the problem.
I thought that could be hardware problem.

Only with the kernel 4.14 I could install NVidia driver with no problem so far.

edit-
Running 390.25 one day uptime and all is ok, no freezes or similar.

There is one thing that appears to be a bug, if /etc/X11/xorg.conf has the line

Code:

    Option         "metamodes" "1680x1050_60_0 +0+0 {ForceCompositionPipeline=On, ForceFullCompositionPipeline=On}"

X will start with a black screen. If I change to TTY1 and back to TTY7, desktop is displayed.
If that line isn't present or none xorg.conf at all, the problem doesn't happen.

It seems a global issue, there is one report at NVidia Linux forum.
https://devtalk.nvidia.com/default/t...-750-ti-390-25
(I tried to register to post a report but I couldn't, don't know why. It seems to be a corporate only forum,
not for domestic users!)

cwizardone · 01-29-2018, 01:53 PM

My GeForce 700 series is a much more modest card than yours, but it has ran very well with Slackware64-current and recent Nvidia drivers. At the moment the latest long term driver, 390.25, is running perfectly, as did 390.12.

dugan · 01-31-2018, 08:58 AM

I just have this (and then I turn VSync off in games that offer me that option):

Code:

Option "metamodes" "nvidia-auto-select +0+0 {ForceCompositionPipeline=On}"

Are you sure you need ForceFullCompositionPipeline=On in addition to that? I certainly don't.

Paulo2 · 01-31-2018, 09:50 AM

Quote:

Originally Posted by dugan

I just have this (and then I turn VSync off in games that offer me that option):

Code:

Option "metamodes" "nvidia-auto-select +0+0 {ForceCompositionPipeline=On}"

Are you sure you need ForceFullCompositionPipeline=On in addition to that? I certainly don't.

Just tested it, only with 'ForceCompositionPipeline=On' (without full composition) gives the black screen too.

The black screen seems to be related with compositor.
Xfce without any compositor, or with it's native compositor = black screen.
Xfce with compton = no black screen.
Kde (Kwin) = no black screen.
Xfce with kwin = no black screen.

I will keep running 390.25 since this is the only problem.

dugan · 02-07-2018, 01:53 AM

Hmm... I noticed that with 390.25 on -current, the ideal settings seem to be:

"Force Composition Pipeline": On
"Sync to VBlank": Off
"Allow Flipping": Off

With "Allow Flipping" turned on, I was getting tearing in YouTube movies in Firefox, and skipping in the same spots in the same movies in Chromium (so: same problem, really). Whether or not I had a compositor active didn't seem to make a difference.

elcore · 02-09-2018, 08:39 PM

driver 340.106 (x86) on new kernel 4.4.115 and new gcc 5.5.0

tested, passed, works fine here

dugan · 02-09-2018, 09:45 PM

I went back to 384.111. Having strange video playback issues (like screen tearing in movies in Firefox) with 390.25.

cwizardone · 02-09-2018, 11:23 PM

I've not had any problems with 390.25 long term driver and the 4.14.7, 4.14.8, 4.15.0, 4.15.1 or 4.15.2 kernels.

Coastal Disturbance · 02-10-2018, 05:48 AM

I know this is a bit off topic and I apologize.

But I just have to chime in and say I've used NVIDIA in the past but recently got myself Radeon RX 560. What a relief!

If you have some spare chips, I highly recommend the switch to AMD GPUs (especially from Polaris 11 and later).

No more recompiling shady drivers and ugly blobs after each kernel update, no more problems with kernel <-> driver incompatibilities, no more butchering the system provider libraries etc.

It is truly plug and play. AMDGPU on Slackware -current is a dream. You just plug the card in and that's the end of it.

DRM console works, VDPAU works, HW acceleration in Firefox works, VDPAU H.264 and H.265 (HEVC) works. Everything works and it works flawlessly.

Kernel updates? No problem, you just go ahead and update, without fear. The only thing that can happen with kernel upgrades and userland updates is that it might work better than before.

I even ripped of the fans and plastic shroud from the card and run it completely fanless, with a slight porformance downgrade (not really in practice):

Code:

echo low >/sys/class/drm/card0/device/power_dpm_force_performance_level

How cool is modifying the card parameters with a proper kernel interface? Even the sensors work:

Code:

% sensors                                                                      
amdgpu-pci-0200
Adapter: PCI adapter
fan1:             N/A
temp1:        +37.0°C  (crit =  +0.0°C, hyst =  +0.0°C)

coretemp-isa-0000
Adapter: ISA adapter
Package id 0:  +27.0°C  (high = +80.0°C, crit = +100.0°C)
Core 0:        +25.0°C  (high = +80.0°C, crit = +100.0°C)
Core 1:        +24.0°C  (high = +80.0°C, crit = +100.0°C)
Core 2:        +23.0°C  (high = +80.0°C, crit = +100.0°C)
Core 3:        +24.0°C  (high = +80.0°C, crit = +100.0°C)

asus-isa-0000
Adapter: ISA adapter
cpu_fan:        0 RPM

RX 560 should be more or less in the GTX 960 range (my previous card). RX 570 and 580 are *much more* powerful, but I don't play games.

Like Linus said, *fuck Nvidia*!

Give it a go, if you can.

cwizardone · 02-10-2018, 10:02 AM

Quote:

Originally Posted by Coastal Disturbance

I know this is a bit off topic and I apologize.

But I just have to chime in and say I've used NVIDIA in the past but recently got myself Radeon RX 560. What a relief!

If you have some spare chips, I highly recommend the switch to AMD GPUs (especially from Polaris 11 and later).

No more recompiling shady drivers and ugly blobs after each kernel update, no more problems with kernel <-> driver incompatibilities, no more butchering the system provider libraries etc.

It is truly plug and play. AMDGPU on Slackware -current is a dream. You just plug the card in and that's the end of it......

If that is true, then AMD has pulled off a miracle. As I've said before, I used ATi/AMD cards for over twenty years and they were,
with Linux, always more difficult to use (a royal PITA) than the Nivida based cards I've used since.
Sure you don't work for AMD?

<--big grin

Coastal Disturbance · 02-10-2018, 10:35 AM

Quote:

Originally Posted by cwizardone

If that is true, then AMD has pulled off a miracle. As I've said before, I used ATi/AMD cards for over twenty years and they were,
with Linux, always more difficult to use (a royal PITA) than the Nivida based cards I've used since.

Yes, that was the reason my previous card was Nvidia. But in the last few years things completely changed. AMD GPU drivers are now mainlined, directly in the kernel. The rest of the stack is mesa, kernel-firmware, vulkan-sdk, xf86-video-amdgpu, all available and recent in Slackware -current. Like I said, you just plug the card in and that's the end of it.

dugan · 02-10-2018, 10:59 AM

Quote:

Originally Posted by cwizardone

If that is true, then AMD has pulled off a miracle.

When companies open source their drivers, miracles follow soon after.