Intel Xe Driver Adds Cold Reset to Recover GPU Without Reboot

Intel Xe graphics driver updates for Linux 7.4 aim to reduce downtime for users experiencing GPU crashes. The changes allow systems to recover from persistent hardware errors without requiring a full PC reboot. This capability matters because it saves time for users who would otherwise need to restart their machines to restore display functionality.

New Linux 7.4 driver update allows systems to recover from persistent GPU hardware errors without requiring a full PC reboot

Intel submitted the first pull requests for these Xe kernel graphics driver features to DRM-Next after the Linux 7.3 merge window closed. The updates target Panther Lake platforms and integrate with new GuC firmware to improve power management under high load. This release represents the initial batch of features expected to merge into the mainline Linux kernel during the Linux 7.4 merge window in late October.

• Cold Reset Recovery Mechanism: Allows GPU recovery from persistent hardware errors without full system reboot by triggering a device-level power cycle via DRM_WEDGE_RECOVERY_COLD_RESET uevent and udev.
• Intelligent Bias Control v3 (IBC v3): Enabled for Panther Lake Xe3 integrated graphics, integrated into new GuC firmware for improved power management under high load.
• DRM_RAS Netlink Error Events: Support added for Reliability, Availability, and Serviceability error event notifications.
• Virtualization VF Multi-Queue Migration: Added capability for migrating virtual function multi-queues.
• Parallel Page Fault Handling: Mechanism for handling page faults in parallel.

The most significant addition is a cold reset recovery mechanism that triggers a device-level power cycle via DRM_WEDGE_RECOVERY_COLD_RESET uevent and udev. This function specifically addresses power management unit (PUNIT) errors by allowing the system to execute a complete device restart without a full system reboot. The update also adds support for DRM_RAS Netlink error events to improve reliability, availability, and serviceability notifications.

Additional technical improvements include virtualization VF multi-queue migration capability and a mechanism for handling page faults in parallel. Intelligent Bias Control v3 (IBC v3) is enabled for Panther Lake Xe3 integrated graphics to optimize power consumption. These changes provide a more stable experience for Linux users running Intel integrated graphics on next-generation hardware.