Broken Sleep on NVIDIA due to Commit #201 on NVIDIA.conf (https://github.com/CachyOS/CachyOS-Settings/pull/201)

[BombasticBooger]

#201.

Discord Support link that shows other's having similar problems:
https://discord.com/channels/862292009423470592/1471327087611740376
CachyOS Forums:
https://discuss.cachyos.org/t/nvidia-black-screen-after-waking-up-from-sleep/22952
CachyOS Distribution Github issue:
CachyOS/distribution#356

The PR mentions a specific issue, and not to be rude, but I find it really weird to link only one issue, and in the end, OP of the issue said that it didn't fix their suspension issues anyways.

Steps:

1. Editing /usr/lib/modprobe.d/nvidia.conf
2. Turning NVreg_EnableS0ixPowerManagement to 0
3. Executing sudo mkinitcpio -P in terminal, but I dont know if this is necessary.

Journald Logs:
S0ix Enabled (default right now)
S0ix Disabled

[BombasticBooger]

wouldn't it be better to follow nvidia's official doc and use deep sleep?

[BombasticBooger]

found out my laptop doesnt support deep: moot comment.

[satoko89]

I encountered the same hard freeze during suspend on a Lenovo LOQ (AMD Ryzen + RTX 4060 Mobile) using CachyOS with KDE Plasma 6.

While the discussion here focuses on NVreg_EnableS0ixPowerManagement, my logs revealed a deeper conflict during the transition to s2idle. Even with NVIDIA power management services enabled, the system would hang indefinitely at the very last stage:

kernel: PM: suspend entry (s2idle)
org_kde_powerdevil[1453]: [ 1525] Udev event detected
kwin_wayland[1299]: Failed to delay sleep: The operation inhibition has been requested for is already running

The Workaround / Solution:
In my case, neither the NVIDIA S0ix settings nor the usual processor.max_cstate fixes were enough. The only thing that resolved the hard freeze was disabling the PCI Express Active State Power Management:

• Kernel Parameter: Added pcie_aspm=off to the cmdline.
• Result: Standby and resume now work perfectly.

Suggestion:
It seems that CachyOS's current default power settings might be too aggressive for Lenovo’s PCIe implementation. If NVreg_EnableS0ixPowerManagement=1 is kept as default, it might be worth investigating if pcie_aspm=off (or a more granular ASPM setting) should be recommended or automated for Lenovo LOQ/Legion users to prevent these unrecoverable freezes.

If you are experiencing hard freezes on a Lenovo LOQ or similar hardware, this configuration solves the deadlock between KDE PowerDevil and the PCIe bus.

Step 1: Edit /etc/default/limine (or your bootloader config)
Add pcie_aspm=off to your kernel parameters:
KERNEL_CMDLINE[default]+="pcie_aspm=off nowatchdog nvidia.NVreg_PreserveVideoMemoryAllocations=1"

Step 2: Edit /etc/modprobe.d/nvidia.conf
Ensure these options are set (required for stable VRAM wake-up):

options nvidia NVreg_PreserveVideoMemoryAllocations=1
options nvidia NVreg_TemporaryFilePath=/var/tmp
options nvidia NVreg_EnableS0ixPowerManagement=1

Step 3: Apply changes
Run the following in your terminal:

sudo mkinitcpio -P
sudo systemctl enable --now nvidia-suspend.service nvidia-resume.service