// ITS FOSS — LINUX & OPEN SOURCE
Linux 7.2 Arrives With Cache Aware Scheduling After More Than a Year of Work
The newest point release for Linux is here a few months after the last one, and the time in-between was not spent idly. Scheduler work that took over a year finally came together, several pieces of hardware got proper support, and a series of 2023 MacBooks can finally boot Linux.
During all this, Linus Torvalds' transition from treating AI as noise for years to slowly warming up to it has been something of a wildcard entry.
Anyhow, let's see what this point release has to offer.
AMD's platform driver changes lean heavily on Zen 6 prep. The Hardware Feedback Interface (HFI) driver adds v3 ranking table support for dynamic core rankings, and the power management controller driver adds support for Family 1Ah Model 80H SoCs (a Zen 6-focused change, basically).
What Ubuntu's OEM kernel had been carrying out of tree is now in. The long-awaited AMD ISP4 driver completes webcam support for the HP ZBook Ultra G1a and other upcoming Ryzen laptops.
Similarly, AMDGPU has gained initial HDMI 2.1 Fixed Rate Link support (FRL), years after the HDMI Forum blocked the feature from the open source driver.
For Intel, most of this round's EDAC work targets next-gen Xeon Diamond Rapids chips, adding sub-memory channel awareness and Rank Retry Logic to preserve data integrity during memory errors. Nova Lake H and Panther Lake H gain in-band ECC error reporting through the IGEN6 driver.
Linux 7.2 also continues clearing out old i486 chunks that started disappearing in 7.1 and makes TSC and CX8 support unconditional on x86. TDX modules can now update at runtime, similar to CPU microcode, so security fixes no longer force a full reboot on TDX-enabled systems.
Another highlight of this kernel release has been Cache Aware Scheduling, which was merged after more than a year of work, landing behind the CONFIG_SCHED_CACHE option.
Without cache awareness, the scheduler can end up scattering related tasks across zones on chips with multiple last-level caches (LLCs). This feature groups those tasks onto cores within the same LLC instead.