Although I remain skeptical about AI, it often helps when you're quite clueless and have nothing to lose: even hallucinated lies can be better than nothing and might inspire creative solutions beyond. After AI helped me find an overlooked aspect of my recurring network connectivity issues with public Wi-Fi in German trains, I was hopeful to resolve my erratic black screen issues as well.
Erratic Black Screen Issues on Linux
I chose Linux Xfce on purpose for its classic UI and conservative lightweight energy usage. However, not choosing mainstream means not getting mainstream support, unless you hire an expert or take your time to fiddle with arcane system settings hoping to learn anything and feel proud of your nerdy half-knowledge.
- What I Tried and What Others Documented
- The Issue
- Workarounds and Xfce Config Fixes
- Other Desktops and Display Managers
- The PSR Hypothesis (falsified)
- Userspace Hang or Firmware?
What I already tried and What Others Documented
I had already examined several logfiles and configurations, as detailed in one of the numerous forum threads about erratic black screen issues on Xfce:
https://forum.xfce.org/viewtopic.php?id=18738
https://forums.linuxmint.com/viewtopic.php?t=450484
Half a year later, I lazily prompted Perplexity, which readily spat out a too-long-don't-read style mix of information that might or might not be helpful in my situation. Leveraging large language model technology, I then asked it to compare its answer with my forum thread on summarize only what I hadn't tried before.
https://www.perplexity.ai/search/i-have-occasional-problems-tha-qusAirrqRtKaXXbujzdYzA#1
The link points to the AI conversation at the time of writing. The paragraphs below are based on that answer and helpful comments in the forum thread. That's a community effort after all, as AI wouldn't find any useful suggestions without prior work of real people making an effort to understand and explain. There are numerous other posts and threads about similar black screen issues, including @an3223 Ethan's XFCE blank/black screen with cursor after login (solution) on DEV.to. His solution was to delete ~/.config/xfce4/xfconf/xfce-perchannel-xml/displays.xml as suggested on the Arch wiki and on this thread from the Nvidia forums.
- https://dev.to/an3223/xfce-blankblack-screen-with-cursor-after-login-solution-5fce
- https://wiki.archlinux.org/index.php/Xfce#Black_screens_at_boot_with_NVIDIA_and_multiple_monitors
- https://devtalk.nvidia.com/default/topic/1029381/linux/black-screen-at-desktop-login-gtx-750-ti-390-25-/3
The Issue
When I log into an Xfce4 session, the screen stays black, there is only a movable mouse pointer, but no menus, no right click, nothing interactive. Xfce, xfwm and compositor still work, as I can see when I either force another X session from a text cosole, or when I am lucky that one of the workarounds leads to a fully functional working Xfce desktop session.
Xfce4 is running on Linux Mint 22. However, similar Xfce issues have been reported on Arch Linux and other Mint versions as well.
Diagnostics and Observations
System information from lsb_release -a && uname -a && inxi && sudo lshw -C display && echo $XDG_SESSION_TYPE && xfwm4 --version and diagnosis tool, contents of ~/.xsession-errors and /var/log/Xorg.0.log after a failed attempt: details can be found in the linked forum threads.
The logs reflecting a defunct black screen session aren't showing any issues. No underrun log entries or any other logs that look related to the blackscreen issue.
Workarounds and Failed Fixes
I had the xserver-xorg-video-intel package installed, but removing the obsolete package was not the solution yet.
- The blackscreen issue used to happen erratically first, so a reboot "solved" the problem. One day, rebooting was not enough anymore.
- All other workarounds start with switching to a console session using Ctrl+Alt+F1
- if there is no cursor, type any key,
- log in at the login prompt,
- do one of the following:
- remove blackbox screensaver:
sudo apt remove blackbox - get all updates: `sudo apt update && sudo apt
- find, kill, and replace the running desktop session:
ps -aux | grep xfce-
kill -9(+ the process id of the xfce session) -
startxorstartxfce4(without lightdm)
- remove blackbox screensaver:
- switch back to the graphical desktop session using Ctrl+Alt+F7
The system keeps behaving erratically. The black screen session that generated the logfiles above, later turned into a functional session when I switched back to it using Ctrl+Alt+F7, but most sessions don't. In rare occasions, after a reboot, everything works well without the black screen issue.
Untested Fixes
The hardware manufacturer, Tuxedo, and a Linux Mint forum user both suggested updating my BIOS, but I fear that I might risk getting even more problems, as I still hope this is just a software/configuration issue.
Alternatively, the hardware manufacturer suggested to switch from their kernel to a default Linux kernel, but I fear I might lose working hardware support, especially for my laptop touchscreen.
I could try enabling the experimental Intel Xe driver for my specific chip ID.
Alternative Approaches
Common troubleshooting suggestions cover session restarts, log checks, and kernel/BIOS notes similar to the forum thread, which remains unsolved with no replies beyond requesting logs.
Diagnostics
Check xfce4-session logs: examine ~/.cache/sessions/xfce4-session-*:0 for session restore errors, absent from forum post.
LightDM and systemd journals: Run journalctl -b -u lightdm and journalctl -b --user -u xfce4-session.service for DM-specific failures.
more untried fixing suggestions
Reset displays.xml: mv ~/.config/xfce4/xfconf/xfce-perchannel-xml/displays.xml ~/.config/xfce4/xfconf/xfce-perchannel-xml/displays.xml.bak then restart LightDM; targets multi-monitor config corruption.
Clear session cache systematically: mv ~/.cache/sessions/* ~/.cache/sessions_backup/ then test startxfce4; prevents broken session reloads.
Test minimal session: Create temp user or disable autostart (mv ~/.config/autostart/*.desktop ~/.config/autostart_disabled/); isolates user config vs. system issue.
Manual component start: From black screen TTY, export DISPLAY=:0; xfce4-panel & xfdesktop & to verify if panel/desktop hangs.
Disable compositor: xfconf-query -c xfwm4 -p /general/use_compositing -s false; rules out xfwm4 rendering stalls on Iris Xe.
Prevention: disable automatic session saving in Xfce's "Session and Startup" settings to avoid recurring cache corruption.
Only after further research (see conclusion and comments below) I found yet another alternative solution approach switching the display manager (or the entire desktop environment):
Alternative Desktops and Display Managers
Asking narrow questions attracts narrow answers, both in specialized user forums and from AI chats that rarely question the question and its implicit assumptions.
On DEV, @leob suggested, I might switch to another desktop environment. After further research, I tried and switched my display manager and switch the desktop from Xfce to Cinnamon (or Mate, but not Gnome or KDE, as the latter two are not officially supported for Linux Mint 22) and see if and when something breaks.
Display Managers: LightDM, LXDM, SDDM
Install and use an alternative display manager instead of LightDM is a good idea in theory, but what would be the alternative that's still supported for Linux Mint?
It's not SDDM, which failed to start, hanging after login, no matter if I used the default (Wayland?) or switched back explicitly using x11. SDDM KDE stack dependencies don't make it an ideal replacement either.
Alternative LXDM: minimal and traditional X11, no Wayland, no login screen customization. LXDM can be use as a drop-in replacement for LightDM on Linux Mint 22 XFCE. However, LXDM hangs at startup, just an unresponsive Linux Mint logo, not even an option to switch to a text console with Ctrl + Alt + F... so I had to reboot into recovery mode and I'm back at LightDM. Back on square one.
I also tried another desktop environment:
sudo apt install cinnamon-desktop-environment
Then log out and select Cinnamon in the session settings.
Cinnamon desktop took a long time to set up so I first thought the black screen issue was back, but that's a one time issue - or at least, the second start was smooth.
I'm still not a big Cinnamon fan, but that wouldn't matter much, if it solved the problem. But the black screen issue still occurs, erratically, not matter if I boot into Cinnamon or Xfce. So the desktop environment is not the root cause. Maybe it's the display manager. We'll never know.
I will never feeld sure again that the black screen issue won't come back eventually, unless someone points to a convincing bug report that matches my problem exactly.
Conclusion? Workaround!
One good thing about Linux is that you can still switch to a text console using Ctrl + Alt + F1 / F2 / F3 in most situations or hit Esc once during boot to enter recovery mode to reconfigure the system. And we have mobile browsers on our phones these days to facilitate research.
Still, it can be hard to solve erratic issues and it's worth following different strategies in parallel, like asking on forums, researching risk-free configuration alternatives, and leveraging both AI and the human developer community to find a solution beyond the obvious suggestions.
Follow-Up: The PSR Hypothesis
Picking up the issue months later after several software updates have not magically solved the issue, gathering additional system information using lsb_release -a and neofetch, the GPU might be the crucial detail overlooked before:
GPU: Intel TigerLake-LP GT2 [Iris Xe Graphics]
A possible explanation sounds convincing and hopeful, unless Claude just made it up. Before proceeding, I should verify the solutions mentioned in the linked bug reports. Interestingly, none of those had turned up in my earlier search and research in January.
The hardware and symptom pattern points at a known, well-documented cause: Panel Self Refresh (PSR) on Intel's i915 driver, which is notorious for exactly this "movable cursor, black/frozen screen, works after a session restart" behavior on Iris Xe / Tiger Lake and related chips. It also explains why the desktop-environment swap (Xfce → Cinnamon) didn't help: PSR operates at the kernel/driver level below the desktop environment.
Fix for Intel i915 GPU Freeze on Recent Linux Kernels: an Ubuntu bug report on nearly identical hardware was resolved by adding
i915.enable_psr=0to grub.Same fix suggested in NVIDIA forums: Black Screen and Screen Flickering in Ubuntu 20.04 with 470 drivers
The classic diagnostic signature for this bug is a CPU pipe B FIFO underrun error from the i915 driver in dmesg/Xorg logs right around the time of the freeze: i915: black screen after boot on 915GM (Linux >= 3.4-rc1)
Confirming the PSR Hypothesis in Journal Logs
Before changing anything, I have to confirm the PSR signature in logs: after a black-screen incident, switch to a TTY (Ctrl+Alt+F1) and run
journalctl -b | grep -i -E 'i915|drm|fifo underrun|psr'
Look specifically for lines like 'CPU pipe B FIFO underrun' or PSR state changes. Done. Found several such signs. ✅
Activating Configuration Workarounds
Next, edit /etc/default/grub and change the line
GRUB_CMDLINE_LINUX_DEFAULT="quiet splash" to
GRUB_CMDLINE_LINUX_DEFAULT="quiet splash i915.enable_psr=0". ✅
After changing the grub configuration, run sudo update-grub and reboot. ✅
Another black screen incident immediately after rebooting. ⬛
If that is not enough, try adding
i915.enable_fbc=0and/ori915.enable_dc=0.
Linux Intel Graphics Flags
vaguely "explained" in layperson's terms, quoting Claude's and Mistral's Le Chat's research summaries:
The flag
i915.enable_psr=0is supposed to fix flickering/black screens, cons: slightly higher power usage. Hardware risks: none.i915.enable_dc=0fixes sleep/wake issues, hardware risks: none, cons: higher power usage.i915.enable_fbc=0should "fix graphical glitches", hardware risks: none, cons: increased memory bandwidth.
Memory bandwidth is the rate at which data can be read from or written to the GPU’s memory (VRAM or system RAM, in the case of integrated graphics like Intel i915). It’s measured in gigabytes per second (GB/s) and determines how quickly the GPU can process and display images, textures, and other graphical data.
For integrated Intel GPUs (like those using the i915 driver), the GPU shares system RAM (not dedicated VRAM), so memory bandwidth is a critical resource.
The GPU compresses the framebuffer (the portion of memory storing the pixels to be displayed). Compression reduces the amount of data that needs to be transferred between the GPU and memory. Result: Lower memory bandwidth usage, which saves power and can improve performance in memory-constrained systems.
Compressing and decompressing the framebuffer requires CPU/GPU cycles thus consumes more power, but usually less than the power saved by reducing memory bandwidth.
PSR Fix Attempt and Falsification
-
GRUB_CMDLINE_LINUX_DEFAULT="quiet splash i915.enable_psr=0 i915.enable_fbc=0"✅
Applying this fix solved my issue–in 50% of all two reboots tested so far. This took me only about half an hour and an undisclosed amount of tokens an energy wasted by the praised Claude AI. So, another hopeful dead end ruled out, but still no solution yet.
The PSR hypothesis turned out to be a wrong track in practice. Or, as the know-it-all AI, Claude, detailed in hindsight:
- A moving cursor means the display is scanning out. A stuck PSR would freeze the cursor too. A black desktop with a live cursor usually means the panel and desktop never drew.
- The session recovered on its own after ~10 minutes in the console. A crashed GPU doesn't do that. A waiting session does.
- The targeted FIFO underrun check was negative. Grepping
i915|drm|psrmatches normal boot messages, so it isn't evidence. A 1-of-2 success rate is consistent with no effect at all. - Cinnamon fails too. Both share LightDM, light-locker, systemd user services, autostart entries, the kernel and the firmware.
- In the bad session,
.xsession-errorsstops right before the desktop session starts. - Cold boots fail, reboots work, which points to firmware.
My latest follow-up, motivated by a severe case that made me reboot five times, and an updated Claude AI model motivated me to yet another go at fixing the exsting setup, asking what we overlooked last time, and how to prevent further wrong tracks.
Next steps: Userspace Hang or Firmware?
Short version: check that the kernel flags are actually active, capture the session state at the next black screen, and wait it out once. Those three steps show whether this is a userspace hang (fixable in minutes) or a firmware/driver problem (then update the BIOS).
1. Check whether the flags are active. Files in /etc/default/grub.d/ are read after /etc/default/grub and can silently override it.
cat /proc/cmdline
cat /sys/module/i915/parameters/enable_psr /sys/module/i915/parameters/enable_fbc
ls /etc/default/grub.d/
plaintext
2. Capture the hung state at the next black screen, from Ctrl+Alt+F1, before killing anything:
ps -u $USER -o pid,stat,etime,wchan:25,cmd | grep -E 'xfce|xfwm|xfdesktop|panel|cinnamon|locker|portal|at-spi|keyring'
systemctl --user list-jobs
journalctl -b --user --no-pager | tail -50
journalctl -b -p warning --no-pager | tail -50
DISPLAY=:0 xwininfo -root -children | head -30
cp ~/.xsession-errors ~/xsession-errors.bad
- Panel/xfdesktop missing or sleeping (
S), pending user jobs → the session is waiting on a service. Userspace problem. - Processes in state
D→ stuck in the kernel. Driver/firmware problem. - Windows exist but the screen stays black → rendering problem. PSR becomes plausible again.
3. Wait it out once. Leave the black screen alone for 3 minutes. Failed D-Bus services time out after 25 s each (xdg-desktop-portal is a known example, and Flatpak is installed).
Waiting it out helps sometimes, but not always.
Diagnostics: D-Bus Hang Collision? 😈💥🚌
Based on the diagnostic data captured (-xsession-errors, xwininfo, process list, and journal logs), the mystery of this erratic black screen is still not solved yet, but narrowed down.
The followning, AI-assisted analysis needs to be verified:
What the data shows
-
PSR and FBC are ruled out.
/proc/cmdlinecontains both flags,enable_fbcreads0, and/etc/default/grub.d/holds only Mint's own file. The black screen happened anyway. The flags can be reverted. -
It's not a kernel or GPU hang.
xfwm4,xfce4-panel,xfdesktopand the other session processes are all alive and idle (stateS). Nothing is in stateDandsystemctl --user list-jobsis empty. -
The session starts, but its components don't (at first).
xfce4-sessionandgnome-keyring-daemonwere running for 04:24, whilexfconfd, at-spi, the portals andxfwm4had only been running for 02:09. The panel followed after about 16 s andxfdesktopafter another 16 s. -
The journal is silent in between. There were no warnings between the login at 09:37:09 and 09:39:18, apart from
gkr-pam: unable to locate daemon control file(keyring/PAM) and some wireplumber messages.
Next steps (cheap, reversible, in this order)
-
A clean waiting test. At the next black screen, stay on VT7. Don't switch to a console and don't run
startxfce4. Wait 5 minutes and write down whether and when the desktop appears. -
Look at the silent gap afterwards. Take the login time from
head -1 ~/.xsession-errors, then:
journalctl -b --no-pager -o short-precise --since “HH:MM:SS” --until “HH:MM:SS” > ~/gap-all.log
Include all priorities, not just warnings. The last entries before the silence are the suspects: keyring/PAM, ibus (im-launch), the user D-Bus.
-
If it's still black after 5 minutes, check what
xfce4-sessionis blocked on, from a console:
sudo strace -p $(pgrep -x xfce4-session) 2>&1 | head -20
busctl --user list --no-pager | head -30
(install strace first if needed)
- Isolate user config from system. Log in as a fresh test user, 8–10 times. If it never stalls, the cause is in my user config. If it stalls too, it's LightDM/PAM/D-Bus.
-
Cheap experiments, one at a time (log 8–10 boots each):
mv ~/.config/xfce4/xfconf/xfce-perchannel-xml/displays.xml{,.bak}mv ~/.cache/sessions ~/.cache/sessions.bak- Disable autologin in LightDM and enter the password manually. The
gkr-pammessage suggests the keyring is unlocked at login, and autologin skips that.
- Still unexplained after that? Update the BIOS (see below).
Kernel update:
- Different kernel: the claimed i915 handshake race has no evidence in my logs. Cheap to test by installing the generic HWE kernel next to the Tuxedo one and choosing it in GRUB, but expect the touchscreen to be missing. Don't remove the Tuxedo kernel.
The kernel is shared by Xfce and Cinnamon, so it can’t be ruled out. A kernel can slow a userspace start indirectly, for example through a driver blocking in udev. Tuxedo itself suggested it.
- Install the generic kernel alongside the Tuxedo one (Mint’s Update Manager has a kernel option, or sudo apt install linux-generic-hwe-24.04).
- Pick it in GRUB’s “Advanced options”. Your Tuxedo kernel stays as the default.
- Expect the touchscreen and some Tuxedo hotkeys to be missing during the test, because they come from tuxedo-drivers.
- Don’t remove the Tuxedo kernel.
Irreversible BIOS update:
-
BIOS/UEFI update: plausible (Tuxedo suggested it, and cold boots fail more often than reboots), but the claim that it targets display initialization is unverified. Check the release notes and compare with
sudo dmidecode -s bios-version -s bios-release-date.
Cost and risk: about 30 minutes, USB-stick procedure, original power supply. Flashing failures are rare but not reversible, so it belongs after the free tests.
irst step: compare your version with Tuxedo’s download page:
sudo dmidecode -s bios-version -s bios-release-date
Alternatives if the capture doesn't settle it
- Log in as a fresh test user to separate user config from system problems.
-
Remove light-locker (inxi shows it running). It's a long-standing source of black-screen-with-cursor reports with LightDM. Replace it with
xfce4-screensaver. - Update the BIOS. Mine is INSYDE 1.07.03RTR from 09/2020, the launch firmware. It supplies the panel config that i915 uses, and it fits the cold-boot pattern. Use Tuxedo's official USB procedure with the original power supply.
Method: the issue is erratic (~50%), so change one thing at a time and log 8–10 boots, noting cold boot vs. reboot.
Deprioritized
- Experimental Xe driver: force-probe only on Tiger Lake/6.11, high effort, and it targets the layer the evidence argues against.
- Non-Tuxedo kernel: risks touchscreen/hotkey support for a weak hypothesis.
- More display managers: SDDM and LXDM already failed.
- Another desktop: Cinnamon already ruled this out.
-
displays.xml/ session cache: Xfce- and multi-monitor-specific, ruled out by Cinnamon. -
i915.enable_dc=0: more PSR-family guessing before the driver is implicated. -
Removing
blackbox: it's a window manager, not a screensaver, so it was never relevant.
...
(to be continued)
Top comments (13)
But it's still not solved? Maybe just dump Xfce and switch to a "mainstream" display manager (KDE or Gnome, and then preferably the "classic" variants)?
Yes, that's still less risk than changing drivers or configuration. No, not solved, as it's only erratic and I don't need to shutdown every day, I probably got lazy. Thanks for reminding me of the simple alternative!
I had other issues after an Ubuntu distribution upgrade in 2024, and cinnamon seemed too restricted so I felt returning to Xfce also visually going back to the good old days. Maybe I took one step too far backwards though.
I used to use XFCE4 in the past and used to love it. 1 year ago I tried to change taste and tried KDE and honestly it's good (apart from some nvidia-wayland issues). I definitely recommend KDE. I used KDE with x11 because the waylabd version crashes always with nvidia GPU
Interesting. Last time I used KDE it was version 0.x of an obscure alternative desktop system while I struggled to get X11 working at all on my 486 PC with SuSE 2.x or something, haha. I should really try and see how it evolved. Is KDE an option for Mint though?
I think so. Although it is not officially supported by mint, it should work without problems.
Here is a not too old tutorial for your reference (in case you need something) - linuxconfig.org/how-to-install-kde...
Gnome classic works for me - not Gnome "Unity", with its huge showy sidebar, which (for me at least) just clutters and gets in the way, and doesn't seem to do anything that I need - but Gnome "classic", with its simple Windows 95/XP style taskbar at the bottom, showing my active/running applications so that I can easily and simply switch between them - and a way to launch apps (via icons or a menu) - and that's it really, nothing more !
That plain, simple, and lovely taskbar is Microsoft's best invention ever - until they messed it up in Windows Vista and later by adding tons of useless features to it ;-)
Unity still exists? like Ubuntu One or what was it called? Last time I had Ubuntu Budgie with a MacOS-inspired Dock panel that looked really nice.
Gnome and KDE don't seem to be officially supported mainstream Xfce alternative for Linux Mint 22 though, but Cinnamon and Mate are.
I used Ubuntu back when I was a (satisfied) Linux user, later on I switched to MacOS - but Linux wasn't bad at all ...
I think I will first try and switch the display manager from lightdm to sddm, maybe I can even keep Xfce, otherwise I'd probaboly switch to Cinnamon. MacOS is nice, its UI much better than Windows, but still too much Apple vendor lock in.
Yeah well you're tied to their hardware, which is obviously pretty expensive, but I'm now used to it and pretty satisfied, so I'm not going to switch back ... but, Linux (Ubuntu in my case) was nice, nothing wrong with it!
Updated this post, updated my configuration ... (to be continued)
nice
Some comments may only be visible to logged-in visitors. Sign in to view all comments.