I’ve installed the nvidia driver the recommend way using the rpm-howTo https://rpmfusion.org/Howto/NVIDIA.
The last 10 times I resumed from Hibernate a lot of characters and symbols aren’t rendered, right at the lock screen and my gnome-shell crashes as well.
Setup:
fedora 30
gnome 3.32.2
GeForce GTX 1060 6GB/PCIe/SSE2
kernel: 5.1.15-300.fc30.x86_64
nvidia driver: 430.26
Steps to reproduce:
go into hibernate (I’ve configured that in my power settings to happen when I press the power button on my tower)
wake up from hibernate by pressing the power button again.
experience a lock screen which has missing characters and buttons (i guess the gnome-shell has already crashed here)
unlock as normal
get notified that gnome-shell crashes
experience even more rendering problems as seen here
restart gnome-shell to go back to normal.
After unlock and before gnome-shell restart (bad):
After gnome-shell restart (normal):
I’ve also tested this with nouveau drivers. Here the screen remains black when I initialize the wake up and I can only hard restart.
PS: The “_old” bug report was done right after the crash, but then I’ve noticed that one should run
startx -- -logverbose 6
. So the one without “old” in the filename was done after running the refered cmd.
My SWAP is now 1,5x of my RAM. Before it was 0,5x of my RAM. I got 16GB installed.
The problem still remains, but the rendering artifacts are a little different now.
It is the right place to ask but as to get support, YMMV.
This is partially a known problem with the proprietary nvidia driver. The known part is that on context switches like VT switch or suspend/resume FBOs like background images etc. get destroyed so the WM has to take care to rebuild them. Which was working pretty fine with Gnome before but isn’t working anymore telling from my own experience but I didn’t care enough so far to check if this is a regression in Gnome 3.32 or in the nvidia driver. The gnome-shell crashing sure is a gnome bug.
Ok, thanks for your reply.
My observation is that the gnome-shell doesn’t crash anymore now in this situations.
However the rendering artifacts are remaining.
So restarting the gnome-shell is the one thing I can do for now until someone from nvidia is keen enough to further investigate this…
Download the latest 430 .run installer:
[url]https://http.download.nvidia.com/XFree86/Linux-x86_64/[/url]
and just uncompress it using the -x option. In the created directory, there are the needed systemd units and suspend-script:
nvidia-suspend.service
nvidia-resume.service
nvidia-hibernate.service
nvidia-sleep.sh
Also report to your distro or the driver repo maintainer to add it.
I exctracted .run file, then I copied nvidia-sleep.sh to /usr/bin directory and registered system-d services.
Well:
My machine can’t go to sleep when “nvidia-suspend.service” is enabled
enabled “nvidia-resume.service” really helps to fix artefacts of shortcuts on desktop, but doesn’t help with applications like viber or latte-dock (etc.)
currently I’m on KDE, as I noted in topic by link, same behavior I was watching in Linux MInt 18.3 Mate (based on gtk) *
(Also, I returned back to driver 390 from official repo, because with driver v430 from run file I’ve got a black screen instead of Login Window and I can’t fix it)
Same here, regarding your point 1. my machine gets stuck in the process to enter hibernate.
Screen turns off, but my machine stays awake…that’s it
Did you also pass the mentioned kernel module parameters to the nvidia module?
I created an extra conf for this.
However I’m still using /tmp which hasn’t the recommended filesystem, but there is enough free space on it. So this might be related to the situation, but it’s too much of a hack for me now to create an extra partion with a different filesystem…
If there is no other solution I’m going to stay with it as it is for now.
No, I didn’t it manually. However, if you look at the .service files, you can see that some parameters are passing by systemd service.
So, I had got updates from KDE neon repo and NVIDIA Settings App was updated to version 418, after that afrtefacs on the desktop returned back. Even nvidia-susped.service is enabled.
Ok, I completely updated nvidia subsystem from ppa (ubuntu-drivers) to version 430.
$ nvidia-smi
Sat Jul 13 10:04:24 2019
+-----------------------------------------------------------------------------+
| NVIDIA-SMI 430.26 Driver Version: 430.26 CUDA Version: 10.2 |
|-------------------------------+----------------------+----------------------+
| GPU Name Persistence-M| Bus-Id Disp.A | Volatile Uncorr. ECC |
| Fan Temp Perf Pwr:Usage/Cap| Memory-Usage | GPU-Util Compute M. |
|===============================+======================+======================|
| 0 GeForce 930MX Off | 00000000:01:00.0 Off | N/A |
| N/A 48C P0 N/A / N/A | 310MiB / 2004MiB | 15% Default |
+-------------------------------+----------------------+----------------------+
+-----------------------------------------------------------------------------+
| Processes: GPU Memory |
| GPU PID Type Process name Usage |
|=============================================================================|
| 0 1063 G /usr/lib/xorg/Xorg 194MiB |
| 0 1500 G /usr/bin/kwin_x11 29MiB |
| 0 1507 G /usr/bin/krunner 1MiB |
| 0 1510 G /usr/bin/plasmashell 53MiB |
| 0 1551 G /usr/bin/latte-dock 21MiB |
| 0 1557 G ...Installed/WizNote-2.7.5-x86_64.AppImage 5MiB |
+-----------------------------------------------------------------------------+
I extracted .run-file (also v430) and registered needed system.d services by my little script:
#!/bin/bash
NVDIR="/home/r3d9u11/Downloads/NVIDIA-Linux-x86_64-430.34"
SDDIR="/lib/systemd/system"
function reg_service() {
if [ ! -f "$SDDIR/$1.service" ] ; then
if sudo cp "$NVDIR/$1.service" "$SDDIR/$1.service" ; then
if sudo systemctl enable "$1.service" ; then
return 0;
fi
fi
exit 1
else
if sudo systemctl enable "$1.service" ; then
return 0;
fi
fi
}
#--- --- --- --- --- --- --- --- --- --- --- --- --- --- --- --- --- --- --- --- --- --- --- --- ---
reg_service nvidia-hibernate
reg_service nvidia-suspend
reg_service nvidia-resume
#--- --- --- --- --- --- --- --- --- --- --- --- --- --- --- --- --- --- --- --- --- --- --- --- ---
if [ ! -f "/usr/bin/nvidia-sleep.sh" ] ; then
cp "$NVDIR/nvidia-sleep.sh" "/usr/bin/"
fi
#--- --- --- --- --- --- --- --- --- --- --- --- --- --- --- --- --- --- --- --- --- --- --- --- ---
echo ""
echo " AVAILABLE NVIDIA SERVICES:"
echo ""
ls /lib/systemd/system/ | grep nvidia
echo ""
echo " ENABLED NVIDIA SERVICES:"
echo ""
systemctl list-unit-files | grep enable | grep nvidia
echo ""
echo ""
All done without error/exceptions, all services was successfully registered:
And now, with driver v430, my machine is going asleep normally.
However there is afterfacts on my desktop and applications still yet after suspending.
Looks like nvidia services do nothing ;-(
At that time I see only one solution to avoid artefscts on the dekstop: reload all needed application after susped (Plasma, Viber and etc). In that case waking up process requires more time. Or just use nouveau, but I afraid to lose performance.
I also tested this shortly, with success. For my system, I had to change the path of pidof in the script. Nevertheless, on suspend it would just freeze, REISUB needed. This seems to be due to the fact that (like said in the doc) it’s needed to switch to VT, which doesn’t work. Switching manually to VT and then suspend, suspend/resume works fine and screen corruption is gone when switching back to gnome.
Without properly setting the module parameter, it doesn’t work, corruption still there.
So in my case, it’s left to modify the script so switching to VT and back works.
Note: I’m also using tmpfs.
Edit: use ‘cat /proc/driver/nvidia/params’ to check if the parameter is correctly set.
Intermediate result: I’ve taken a look at how the script determines the VT at which Xorg is currently running and it’s a quite naive approach, probably not working at most modern systems. So there’s still some brain work to do.
I now changed the nvidia-sleep.sh script to simply use fgconsole to determine the currently used VT for switching back and forth, changed the path of chvt to match my system and now it works for me:
#!/bin/bash
RUN_DIR="/var/run/nvidia-sleep"
XORG_VT_FILE="${RUN_DIR}"/Xorg.vt_number
case "$1" in
suspend|hibernate)
/bin/mkdir -p "${RUN_DIR}"
/usr/bin/fgconsole > "${XORG_VT_FILE}"
/usr/bin/chvt 63
if [[ $? -ne 0 ]]; then
exit $?
fi
/bin/echo "$1" > /proc/driver/nvidia/suspend
exit $?
;;
resume)
/bin/echo "$1" > /proc/driver/nvidia/suspend
#
# Check if Xorg was determined to be running at the time
# of suspend, and whether its VT was recorded. If so,
# attempt to switch back to this VT.
#
if [[ -f "${XORG_VT_FILE}" ]]; then
XORG_PID=$(cat "${XORG_VT_FILE}")
/bin/rm "${XORG_VT_FILE}"
/usr/bin/chvt "${XORG_PID}"
fi
exit 0
;;
*)
exit 1
esac
Thanks for your help and hint about *.service files in .run archive!
Seems like I’ve got success with proprietary nvidia-drivers and KDE Plasma (X11 backend)
