[Intel Ice Lake] Hard lockups avoided by intel_idle.max_cstate=1

Pablo De Napoli <[email protected]>
Newsgroups org.kernel.vger.linux-pm
Message-ID <CAHoaYyeZXZ3A0KmkV-SuwQj79w1iVaFuNEib-upkryZSzEdCvg@mail.gmail.com>
Dear Linux Kernel Maintainers,

## Summary

I am experiencing intermittent complete system freezes on a BMAX B6
Plus Mini PC with an Intel Ice Lake CPU and integrated Intel graphics.

The lockup is not deterministic. I have not found a specific
application or user action that triggers it. It occurs during ordinary
desktop use, apparently after several hours.

The keyboard becomes completely unresponsive and the machine appears
to be hard-locked. I have not been able to recover it through the
usual userspace mechanisms.

## Hardware

* BMAX B6 Plus Mini PC
* Intel Core i3-1000NG4 (Ice Lake)
* Intel Iris Plus Graphics G4
* 32 GB RAM
* Integrated Intel GPU
* X11 desktop environment (MATE)

## Software

* Manjaro Linux
* Linux 6.18-x86_64
* i915 DRM driver

The GPU is using the i915 driver and DRI acceleration works normally.
MATE, Firefox and Zed work normally, including hardware-accelerated
graphics.

## Kernel command line

The configuration with which I experienced the lockups was:

```
i915.enable_psr=0
i915.enable_fbc=0
i915.enable_dc=0
video=HDMI-A-1:1920x1080@60e
```

I also tested the system with:

```
intel_idle.max_cstate=1
```

## Important finding

Adding:

```
intel_idle.max_cstate=1
```

appears to prevent the lockups.

With this parameter enabled, the system has remained stable during
normal desktop use. Without it, the system eventually hard-locks
again.

With the workaround enabled, the available CPU idle states are:

```
POLL
C1_ACPI
```

The active cpuidle driver is `intel_idle`.

This makes me suspect an interaction involving CPU
idle/power-management states, ACPI/firmware, and possibly the graphics
stack. I cannot determine whether the underlying problem is in i915,
CPU idle, ACPI, or an interaction between these components.

## Reproducer

Unfortunately, I do not currently have a deterministic reproducer.

The lockup happens intermittently during normal desktop use,
apparently after several hours. I have not found a particular
application, operation, or sequence of actions that reliably triggers
it.

The most useful observation I can provide is that the lockup can be
avoided by limiting the CPU to C1 with:

```
intel_idle.max_cstate=1
```

## Expected behavior

The system should remain responsive indefinitely during normal desktop
use with the default CPU idle states and the i915 driver enabled.

## Additional observations

I initially suspected i915 because the system only exhibited the
problem when using the Intel graphics driver. However, disabling
`i915` avoids the graphics path entirely, while limiting CPU C-states
also appears to eliminate the problem while leaving i915 and DRI fully
functional.

I have therefore not concluded that i915 itself is responsible. The
`intel_idle.max_cstate=1` result makes a power-management/idle-state
interaction seem worth investigating.

I am reporting this because the workaround has a significant
functional impact: it requires disabling deeper CPU idle states,
presumably increasing idle power consumption, in order to obtain a
stable system.

I would be happy to provide additional logs or hardware information if
there are specific diagnostics that would help identify the cause.

best regards
Pablo De Nápoli
lmpx.com only provides a reader for public news (NNTP) servers. It is not affiliated with the servers or forums shown here and is not responsible for the content of articles, which is written by their respective authors.