Gfilui

I got a new machine without any OS in it. The very first thing I did was to install Ubuntu 17.10.

Here are some specs;

i7 7700HQ

GTX 1060 6 GB

Samsung EVO 960 PRO m2 + 2 TB Seagate HDD

The install went pretty smooth. When it finished, the restart button hang quite a while just to fall to a black screen with some errors;

ACPI Error: [_UPC] Namespace lookup failure, AE_ALREADY_EXISTS(2170531/dswload-378)
0.031925 ACPI Exception: AE_ALREADY_EXISTS, During name lookup/catalog (2070531/psobject-252)
0.031958 ACPI Exception: AE_ALREADY_EXISTS, (SSDT:xh_rvp11) while loading...
0.033744 ACPI Error: 1 table load failures, 12 successful (20170531/tbxflo....)

[6.745929] Couldn'tget size: 0x8000000000e
[6.745943] MODSIGN: Couldn't get UEFI db list
[8.908673] sd 0:0:0:0: [sdb]] No Caching mode page found
[8.908674] sf 0:0:0:0: [sdb] Assuming drive cache: write through
[8.922136] print_req_error: critical medium error, dev sr0, sector 16368
[8.928612] print_req_error: critical medium error, dev sr0, sector 16368
[8.928612] Buffer I/O error on dev sr0, logical block 2046, async page read
[8.950494] print_req_error: critical medium error, dev sr0, sector 16360
[8.956755] print_req_error: critical medium error, dev sr0, sector 16360
[8.956756] Buffer I/O error on dev sr0, logical block 2045, async page read
[8.969426] print_req_error: critical medium error, dev sr0, sector 16344
[8.975715] print_req_error: critical medium error, dev sr0, sector 16344
[8.975716] Buffer I/O error on dev sr0, logical block 2043, async page read
[8.995095] print_req_error: critical medium error, dev sr0, sector 16336
[9.001279] print_req_error: critical medium error, dev sr0, sector 16336
[9.001280] Buffer I/O error on dev sr0, logical block 2042, async page read
[10.812098] nouveau 0000:01:00.0: secboot: error during falcon reset: -110
[10.812118] nouveau 0000:01:00.0: gr: init failed, -110

Of course went on hard reset and did some more digging and with some additional help and this link, I was able to start and login to my Ubuntu 17.10.

However the moment I chose Nvidia drivers instead of X.org X Server, I couldn't even get to login screen. There was just a black screen and flashing some texts every few seconds;

[0.000000] ACPI Error: [_UPC] Namespace lookup failure, AE_ALREADY_EXISTS(20170831/dswload-378/dswload-378)
[0.000000] ACPI Exception: AE_ALREADY_EXISTS, During name lookup/catalog (20170831//psobject-252)
[0.000000] ACPI Exception: AE_ALREADY_EXISTS, (SSDT:xh_rvp11) while loading table (20170831/tbxfload-228)
[0.000000] ACPI Error: 1 table load failures, 12 successful (20170831/tbxfload-246)
/dev/nvme0n1p5: clean, 199274/15269888 files, 3415948/61047808 blocks

[ OK ] Started Hostname Service.
[ OK ] Started Network Name Resolution.
[ OK ] Reached target Host and Network Name Lookups.
[ OK ] Created slice User Slice of gdm.
 Starting User Manager for UID 121...
[ OK ] Started Session c1 of user gdm.
[ OK ] Started User Manager for UID 121.
 Starting WPA supplicant...
[ OK ] Started WPA supplicant.
 Starting Manage, Install and Generate Color Profiles...
[ OK ] Started Manage, Install and Generate Color Profiles.

There are no more texts, screen just goes black and flashes back for a second with those texts. Left it over 10 - 15 minutes to see if anything would change but no luck.

I had to go into TTY and purge nvidia stuff to get it working again. I have tried all available drivers I could think of but none worked. I even tried bumblebee and directly downloading shell script from Nvidia's driver page. None worked.

I even installed ukuu and updated my kernel to the latest hoping the latest kernel would get this issue fixed. No luck.

Generally I'd leave it with Intel instead of Nvidia but 1 x mHDMI and HDMI seems to be dedicated to Nvidia. There is only 1 mHDMI for Intel (I have dual boot so not having Nvidia on Linux is all good) which I can't change the resolution to so on `21:9 Ultrawide 4K screen, 1080p resolution is unthinkable to work on it 8h. min on daily basis.

Of course I thought about adding resolution and fooling around with xrandr & cvt but they didn't play nicely either.

I decided to try my chances with 16.04 LTS. Downloaded it and installed it, after jumping through small hoops, I get it working;

• Advanced mode, get the lower kernel version.


• in TTY ran $ ubuntu-drivers autoinstall


• Restart the computer and the computer is working with Nvidia.

Now I'm hoping 18.04 LTS will not have this issue. My fear is, it might have the same issue. However I must say I'm kinda afraid it will not get this fixed so I'm looking for a way to solve this issue and I have already spent 3 days in and out with this thus felt back to AskUbuntu.

If I'm not mistaken, I'm seeing 2 issues here, one with disk related and second one with the graphics. I'd really appreciate any help on this one.

PS. I mentioned dual boot earlier, Windows 10 works flawlessly.

FYI Windows 10 installed as being necessary evil!

Which version of the nVidia drivers are you using ?
Did you visit one of the following Launchpad PPAs:

https://launchpad.net/~graphics-drivers/+archive/ubuntu/ppa

https://launchpad.net/~mamarley/+archive/ubuntu/staging

https://launchpad.net/~ricotz/+archive/ubuntu/red

Often even the Graphics Drivers published in these PPAs will lack the required patches for the latest Linux Kernel ...

I am using right now nvidia-390 drivers version 390.25 without problems on Xenial 16.04 running Kernel 4.15.0, only because I recompiled the nVidia drivers with the required patches for Kernel 4.15 and 4.16.

Here is the link to my PPA which contains nvidia-390 version 390.25 supporting Linux kernel 4.15 and 4.16, because I did apply the specific patches for these new Linux kernels:

https://launchpad.net/~anton+/+archive/ubuntu/miscellaneous

[NOTE]: The nVidia Graphics Drivers from the pseudo official PPA Proprietary GPU Drivers, still lacks support for Linux kernel 4.15 in releases other than Bionic 18.04 ... And their debian version is higher than mine, meaning they will install as an update if you subscribe to both PPA, and kill the nVidia drivers for your system. One way to avoid tis is to lock the update for these packages:

Here is a bash script to block update of nvidia-390 drivers:

#!/bin/sh
#
# Add nvidia-390 package to update blacklist
# Run script as super user i.e. sudo ./block-upd-nvidia-390.sh
#

if [ $(id -u) != 0 ]; then
 echo "This script requires root permissions"
 sudo "$0"
 exit
fi
echo "nvidia-390 hold" | dpkg --set-selections
echo "nvidia-390-dev hold" | dpkg --set-selections
echo "libcuda1-390 hold" | dpkg --set-selections
echo "nvidia-opencl-icd-390 hold" | dpkg --set-selections

To make it executable:

$ chmod a+x block-upd-nvidia-390

Then simply run (it will ask your password as it needs sudo power):

$ ./block-upd-nvidia-390

To unblock from update use this bash script:

#!/bin/sh
#
# Remove nvidia-390 package from update blacklist
# Run script as super user i.e. sudo ./block-upd-nvidia-390.sh
#

 if [ $(id -u) != 0 ]; then
 echo "This script requires root permissions"
 sudo "$0"
 exit
 fi
 echo "nvidia-390 install" | dpkg --set-selections
 echo "nvidia-390-dev install" | dpkg --set-selections
 echo "libcuda1-390 install" | dpkg --set-selections
 echo "nvidia-opencl-icd-390 install" | dpkg --set-selections

Installing the latest kernel may expose you to this kind of quirks ...

Still a few days ago after one full month of Kernel 4.15 out here, there were no working nVidia Graphics Drivers available in any of the Ubuntu Launchpad PPAs ...

OK. Now I will provide the way to update the nVidia Graphics Drivers in Ubuntu.

The nVidia Graphics Drivers are installed for each Linux kernel via dkms. When installing an nVidia driver, it will install in the currently used Linux kernel and the latest (more up to date) kernel installed. Right now it's Linux kernel 4.15 for Ubuntu. If you are installing the nVidia Graphics Drivers within the most up to date of the Linux kernels installed in your system, they will be installed only in this Linux kernel. Ubuntu 17.10 is using Linux kernel 4.13 by default.

On an Ubuntu system you can install more recent Linux kernels like 4.15 which was out last February. Just have a look at the PPA for Canonical Kernel Team. You can install hwe kernel which is now 4.13 in any Ubuntu supported release, or hwe edge which is now 4.15.

For example to install Linux kernel 4.15 use this command line in a Terminal:

$ sudo apt install linux-headers-4.15.0-13 
linux-headers-4.15.0-13-generic 
linux-image-4.15.0-13-generic 
linux-image-extra-4.15.0-13-generic 
linux-signed-generic-hwe-16.04-edge 
linux-signed-image-4.15.0-13-generic 
linux-signed-image-generic-hwe-16.04-edge

If you install a new Linux kernel you have to update all dkms module drivers for your system to work properly. This may be nVidia Graphics Drivers, and Broadcom Network Adapters.

First get the nvidia version install then remove and purge the existing nvidia drivers:

$ dpkg -l | grep nvidia

Assuming you have nvidia-384 installed:

$ sudo apt purge nvidia-384*

Then install the new driver from the PPA you selected:

$ sudo apt install nvidia-390

You will see this message about the dkms install while the packages are installed (or when you install a new kernel) like:

nvidia_390:
Running module version sanity check.
 - Original module
 - No original module exists within this kernel
 - Installation
 - Installing to /lib/modules/4.13.0-22-generic/updates/dkms/

nvidia_390_modeset.ko:
Running module version sanity check.
 - Original module
 - No original module exists within this kernel
 - Installation
 - Installing to /lib/modules/4.13.0-22-generic/updates/dkms/

nvidia_390_drm.ko:
Running module version sanity check.
 - Original module
 - No original module exists within this kernel
 - Installation
 - Installing to /lib/modules/4.13.0-22-generic/updates/dkms/

nvidia_390_uvm.ko:
Running module version sanity check.
 - Original module
 - No original module exists within this kernel
 - Installation
 - Installing to /lib/modules/4.13.0-22-generic/updates/dkms/

depmod....

DKMS: install completed.
Building initial module for 4.13.0-23-generic
Done.

nvidia_390:
Running module version sanity check.
 - Original module
 - No original module exists within this kernel
 - Installation
 - Installing to /lib/modules/4.13.0-23-generic/updates/dkms/

nvidia_390_modeset.ko:
Running module version sanity check.
 - Original module
 - No original module exists within this kernel
 - Installation
 - Installing to /lib/modules/4.13.0-23-generic/updates/dkms/

nvidia_390_drm.ko:
Running module version sanity check.
 - Original module
 - No original module exists within this kernel
 - Installation
 - Installing to /lib/modules/4.13.0-23-generic/updates/dkms/

nvidia_390_uvm.ko:
Running module version sanity check.
 - Original module
 - No original module exists within this kernel
 - Installation
 - Installing to /lib/modules/4.13.0-23-generic/updates/dkms/

depmod....

DKMS: install completed.

Or if the driver has no support for the target Linux kernel (specific patches for this kernel not applied), you will see errors like:

run-parts: executing /etc/kernel/postinst.d/apt-auto-removal 4.13.0-23-generic /boot/vmlinuz-4.13.0-23-generic
run-parts: executing /etc/kernel/postinst.d/dkms 4.13.0-23-generic /boot/vmlinuz-4.13.0-23-generic
ERROR (dkms apport): kernel package linux-headers-4.13.0-23-generic is not supported
Error! Bad return status for module build on kernel: 4.13.0-23-generic (x86_64)
Consult /var/lib/dkms/nvidia-390/390.25/build/make.log for more information.

Did you successfully install Artful 17.10 then ?

If you believe this answer responded to your question and solved your issues please select the V on the left side.

Strange I have reverse issue in 17.10: nVidia 390.12 works but Intel Graphics microcode isn't listed on additional drivers list to choose, but is there installed:

sudo dmesg | grep microcode 
[ 0.000000] microcode: microcode updated early to revision 0x22, date = 2017-01-27
[ 0.943427] microcode: sig=0x306c3, pf=0x20, revision=0x22
[ 0.943551] microcode: Microcode Update Driver: v2.2.

so I can't switch to Intel via nVidia-prime...and about ukuu kernels the 4.15.11 is the latest where on my laptop in Blender all Cuda cycles compute devices are found (nVidia and cpu) and in newer from 4.15.12 to 4.16 they aren't found so there's an issue with these newer kernels probably.