Quantcast
Channel: Raspberry Pi Forums
Viewing all articles
Browse latest Browse all 8597

Troubleshooting • Persistent System Freeze/I/O Stall on RPi CM5 eMMC (Ubuntu 24.04+) after Long Uptime

$
0
0
1. Environment & Symptoms
I am experiencing a critical stability issue involving the Raspberry Pi Compute Module 5 (CM5) eMMC variant running Ubuntu.

* Hardware: Raspberry Pi Compute Module 5 (CM5)

* eMMC Size: Reproduced on both 4GB and 8GB eMMC boards.

* Operating System: Ubuntu 24.04 LTS and subsequent versions (currently testing on 24.10).

* Kernel Versions: The freeze occurs on multiple versions, including 6.11.0-1009-raspi and 6.14.0-1012-raspi.

* Issue: The system randomly freezes (I/O Stall) after 3.5 or more days of continuous uptime.

Observed Symptoms during Freeze:
  • SSH access is impossible (system is unresponsive).

    Ping remains functional (lower network layers are alive).

    Network communication slows down significantly or completely fails (DB connections, network file access fail).

    Resource Usage: Monitoring confirms that CPU, memory, log capacity, and network load were all normal before the failure.
2. Key Logs and Errors
Two primary and distinct log patterns have been observed prior to the system becoming completely unresponsive:

1) Critical File System Error (Leading to Freeze)
The most severe event is the file system spontaneously switching to a read-only state, which immediately precedes the system freeze:

Code:

kernel: fallocate[323]: fallocate: cannot open /swapfile: Read-only file system
2) Recurring USB I/O Errors
In some freeze instances, a stream of USB errors is observed in the logs just before the system becomes unresponsive.

Code:

kernel: usb 2-1.4.2: device descriptor read/64, error -71kernel: usb 2-1.4.2: device descriptor read/64, error -71kernel: usb 2-1.4.2: new high-speed USB device number 37 using xhci-hcdkernel: usb 2-1.4.2: device descriptor read/64, error -71kernel: usb 2-1.4-port2: attempt power cycle
(Note: Errors related to USB disconnection or current have also been observed.) USB power is confirmed to be stable and supplied by an external, dedicated source.

3. Diagnosis and Action Plan
  • Diagnosis: Based on the CM5 eMMC hardware, the I/O stall, and the Read-Only file system error, we highly suspect a deadlock caused by the eMMC Controller's CQE (Command Queue Engine) algorithm, similar to issues tracked in GitHub Issue #6512.(https://github.com/raspberrypi/linux/issues/6512)

    Concern: While disabling CQE via a boot parameter (sdhci.cqe=0) is a known workaround, we are concerned about the inevitable degradation of random I/O performance and increased latency. We seek a fix that maintains optimal eMMC performance.
4. Questions for the Community/Developers
  • Does the error sequence (USB Errors → Read-Only File System Error → Freeze) definitively point to the CQE Deadlock issue, or should the USB errors be investigated as an independent, primary trigger?

    Given that the problem persists on recent kernels (up to 6.14.0-1012), is the official fix (Device Tree modification/kernel patch) for the CM5 CQE issue included in the public Ubuntu/Raspberry Pi kernel releases yet?

    We want to avoid performance compromise. What is the current, official stable kernel version that we can upgrade to that reliably resolves this eMMC deadlock without requiring boot parameters like sdhci.cqe=0?
Thank you in advance for your assistance.

Statistics: Posted by JinsukChoi — Fri Dec 05, 2025 2:28 am



Viewing all articles
Browse latest Browse all 8597

Trending Articles