Re: ASUS GU605CX: diagnostic guidance for hard freezes and USB-C power-state changes

[Heikki Krogerus <heikki.krogerus@linux.intel.com>]

On Sat, Sep 26, 2026 at 02:45:15PM +0300, Leo Korotkevich wrote:
> Hello,
> 
> Could you recommend low-overhead USB-C/UCSI diagnostics for intermittent
> hard freezes on an ASUS GU605CX, Core Ultra 9 285H, BIOS 311?
> 
> The latest event was on CachyOS 7.2.7-1-cachyos with out-of-tree NVIDIA
> open 615.71.09. The screen froze and local input/VT switching stopped
> responding. The journal ends without a nearby USB error, OOM or panic. I
> have no evidence that UCSI caused it.

You should be able to disable the driver:

        echo "blacklist ucsi_acpi" > /etc/modprobe.d/ucsi.conf

In your case it may be enought to just unload it (modprobe -r
ucsi_acpi). Then just see can you reproduce the issue.

Thanks,

> Freezes recur across Ubuntu/Fedora/CachyOS releases, BIOS/kernel
> versions and performance profiles, Xorg/Wayland and
> dGPU-only/Dynamic-Hybrid modes. I also tried modprobe.blacklist=i915
> through GRUB; freezes continued. They also happen overnight on a
> stationary desk without cable movement. Some coincided with movement of
> the charging plug, while deliberate unplug/replug usually succeeds. They
> have occurred with the OEM ASUS charger and Lenovo/Dell adapters with
> different voltage/power ratings; exact specifications were not recorded.
> The current Apple charger reports a 20 V / 4.7 A PD contract.
> 
> Historical Fedora logs contain 1747 software-reported AC/DC alternations
> on March 9, not verified physical disconnects or a trace of the latest
> freeze. During successful recent reconnections, asusd changed profiles
> and restarted nvidia-powerd. UCSI_GET_PDOS failed (-95) appeared twice,
> but connection recovery succeeded. Brief profile/EPP changes also
> completed.
> 
> ASUS PD_FW 0003 declares firmware version 39 for the resource currently
> reporting ESRT version 1. It remains unapplied; its release note claims
> no freeze fix.
> 
> A bounded collector records power-supply uevents and contract fields. A
> temporary UCSI trace saw no events and has not been validated against a
> physical reconnection.
> 
> Which tracepoints best distinguish PD/PPM failure from normal contract
> changes and the platform's AC/profile reaction, without active
> controller polling? Sanitized excerpts are available.
> 
> Thank you.

-- 
heikki