I am seeing a suspend/resume issue on my laptop. When the system goes into suspend, it later comes back to a black screen. I cannot
recover the display normally and usually need to force reboot.
System information:
Laptop: Lenovo ThinkPad T1g Gen 8
Machine type: 21TD000MUS
Ubuntu: 26.04 (Resolute Raccoon, development branch)
Kernel: 7.0.0-10-generic
CPU: Intel Core Ultra 9 285H
Graphics:
Intel integrated GPU: Intel Arrow Lake-P [Arc Pro 130T/140T]
NVIDIA discrete GPU: GeForce RTX 5070 Max-Q / Mobile
NVIDIA driver: 595.58.03
Suspend-related details:
/sys/power/mem_sleep shows only: [s2idle]
Kernel log says: “Low-power S0 idle used by default for system suspend”
Relevant observations:
I can see suspend handling starting in the logs, for example:
“rtkit-daemon: Handling system-suspend using logind.”
But I do not see a clean resume sequence afterward.
This matches the actual behavior: after suspend, the machine returns to a black screen and I often need to hard reboot.
There are also NVIDIA-related messages in the logs, including:
“nvidia-modeset: WARNING: GPU:0: Correcting number of heads for current head configuration”
“nvidia … [drm] No compatible format found”
“nvidia … [drm] Cannot find any crtc or sizes”
uname -a:
Linux yangshirong 7.0.0-10-generic #10-Ubuntu SMP PREEMPT_DYNAMIC Thu Mar 19 10:24:42 UTC 2026 x86_64 GNU/Linux
Reproducing this on different hardware (Framework Laptop 16, AMD Ryzen 7940HS with RTX 5070 Max-Q) on Ubuntu 26.04 kernel 7.0.0-14-generic. Same symptoms — lid-close s2idle suspend never resumes, hard reset required.
I isolated kernel vs driver independently:
Kernel 6.17.0-22 + nvidia loaded → works
Kernel 6.17.0-22 + nvidia blacklisted → works
Kernel 7.0.0-14 + nvidia loaded → hangs
Kernel 7.0.0-14 + nvidia blacklisted (module_blacklist=nvidia_drm,nvidia_modeset,nvidia_uvm,nvidia) → works
Same nvidia-open driver (tested both 580.142 and 595.58.03), only the kernel version matters. On kernel 7.0 with nvidia blacklisted, s2idle completes a clean `PM: suspend entry → PM: resume from suspend-to-idle → PM: suspend exit` cycle — so the regression is in the nvidia-open suspend path, triggered by kernel-PM changes between 6.17 and 7.0.
Failure signature on the hung boot: `journalctl -b -1 -k` ends at `nvidia 0000:01:00.0: Enabling HDA controller` with no subsequent `PM: suspend entry (s2idle)` line — which the kernel emits only after the device suspend cascade returns. So the nvidia PCI `.suspend` callback does not return on 7.0.
Your hardware is Intel + Lenovo, mine is AMD Phoenix + Framework — different platform, different iGPU, different CPU vendor. The only common factor is kernel 7.0 + RTX 50-series Max-Q + nvidia-open, which rules out platform/firmware.
Happens to me too!!
Nvidia drivers 595 (rtx 3060), ubuntu 26.04.
It was working previously with 24.04 LTS, after a modification suggested by Gemini. Now I tried 3 different things by AI with no luck.
nvidia-bug-report.log.gz (351.8 KB)
I can reproduce a suspend/resume black-screen hang on Ubuntu 26.04 with Linux 7.0.0-15-generic and NVIDIA proprietary driver 595.58.03.
The system goes black and never returns to the desktop. A hard power-button reset is required.
Important isolation:
PRIME intel mode works: suspend/resume succeeds when NVIDIA modules are not loaded.
PRIME on-demand fails: suspend/resume hangs when NVIDIA modules are loaded.
Switching s2idle/deep did not solve it.
This looks related to the existing Ubuntu 26.04 + kernel 7.0 + NVIDIA 595 suspend/resume regression, but my system uses the proprietary NVIDIA module and an RTX 4060 Laptop GPU.
Attempted the workaround mentioned in nvidia-modeset: suspend crash. Issue #1095 . Unfortunately, the issue continues in my system after applying veldenb solution. I even clean reinstalled NVIDIA drivers to specifically use the dkms package.
Investigation showed that suspend failures on Ubuntu 26.04 with the open NVIDIA kernel driver on Linux 7.0 were tied to how the Ubuntu-packaged DKMS driver was being built. A distro-specific packaging change disabled objtool validation for parts of the driver build. That allowed the package to compile, but the resulting nvidia-modeset module could hit a fatal kernel bug during suspend (jump_label patching in nvkms_kthread_q_callback), which matches the hang/black-screen behavior many of you reported.
This is not a generic kernel 7.0 or hardware-platform issue by itself — the same driver and hardware suspend/resume cleanly on kernel 6.17, and suspend works on 7.0 when the NVIDIA kernel modules are not loaded. The problem is specific to the Ubuntu DKMS build of the open driver on 7.0.
Canonical has informed us that updated respin driver packages released on May 19, 2026 address this packaging issue.
While I can’t be certain it’s the exact same root cause, I have a black screen as well when waking from sleep on a Lenovo LOQ 17IRX10 even on the newest packages. In my case, this only happens when the BIOS is set to use discrete GPU only. In hybrid mode it works correctly, but of course this lowers performance.