So the long standing issue with mouse becoming janky in parallel login sessions
drove me to desperation. Having posted online on
kreddit,
askubuntu and forum.kde.org and
getting zilch, I finally decided to install another distribution in parallel
and see if the issue persists. Decided to give PopOS (ubuntu derivative by
System76 folks) a shot but the
installation
didn’t go too well. To boot, I also managed to screw up my current installation by picking the same /boot
while
trying to install Pop OS. That adventure ended happily once I chrooted
, fixed up boot and got back to kde neon and
noticed a sluggish mouse on the first session itself. This was new and interesting and dmesg was being spammed with
nvidia driver mismatch errors (I was using a kernel installed by Pop with KDE neon). That indicated a nvidia issue
and sure enough, here
was the issue that had troubled me no end!
Once that got sorted out I was on a high and decided to clean up a bit and also fix boot animation that got screwed. In roughly sequential order
-
Clean up residual config
dpkg -l |grep ^rc| awk '{print $2}'| sudo xargs dpkg --purge
Resulted in some errors (I still have
mintsystem
) which will not go - but overall just housecleaning -
Set boot splash theme in KDE Settings Manager
-
Boot splash comes up, but very late. Another Askubuntu page later and framebuffer does better.
-
-
Now I shouldn’t need the above if I have KMS (which I should have) - but nevertheless, here’s what I did for KMS
-
Grub config
# /etc/default/grub GRUB_CMDLINE_LINUX_DEFAULT="quiet splash nvidia-drm.modeset=1" GRUB_GFXMODE="1920x1080x32" GRUB_GFXPAYLOAD_LINUX="keep" > sudo update-grub2
-
Initramfs
# /etc/initramfs-tools/modules nvidia_uvm nvidia_drm nvidia_modeset nvidia nvidia-drm modeset=1 NVreg_RegisterForACPIEvents=1 NVReg_EnableMSI=1 NVreg_RegistryDwords=RMUseSwI2c=0x01;RMI2cSpeed=100 > sudo update-initramfs -u -k `uname -r` # confirm if modules are included > lsinitramfs /boot/initrd.img-4.18.0-22-generic |less
-
-
Now from eye-candy to boot performance. I was thinking that I’d have to profile with bootchart etc that I’d come across but never used… Turns out that
systemd
has a few tricks up it’s hand# boot time summary systemd-analyze # longest pole on critical path ordered by time systemd-analyze blame # Similar - but in a chain form systemd-analyze critical-chain
So my boot was taking upwards of 40s to 1min and that shouldn’t be with an SSD.
-
Culprits
-
docker.service - 15s - replace with docker.socket
-
snapd - 8s - I wasn’t using any snap packages, so -
sudo apt autoremove --purge snapd
-
systemd-journal-flush - 5s - vacuum as advised
-
-
I think that’s about it - but now here’s what systemd-analyze
now says:
raghu@home-desktop ~/blog master systemd-analyze
Startup finished in 5.315s (kernel) + 10.297s (userspace) = 15.612s
graphical.target reached after 10.287s in userspace