Arm Preparing The Linux Kernel For 128-Bit Page Table Entries "FEAT_D128"

Key Takeaways

• FEAT_D128 adds optional 128‑bit page table entries to Armv9.3.
• Supports larger physical and virtual memory ranges for ARM servers.
• Provides extra bits for future MMU management features.
• Current Linux patches lack KVM and KASAN support.

Pulse Analysis

The shift to 128‑bit page table entries marks a pivotal upgrade for ARM’s server‑grade silicon. Traditional 64‑bit entries cap addressable memory at 256 TB, a ceiling increasingly strained by AI models and large‑scale analytics. By introducing VMSAv9‑128, Arm’s FEAT_D128 expands the addressable space dramatically, allowing terabyte‑scale memory pools while preserving compatibility with existing software stacks. This architectural headroom also reserves bit patterns for forthcoming MMU enhancements, future‑proofing the platform.

Linux kernel developers have responded with a dedicated RFC patch series that integrates FEAT_D128 as an optional kernel feature. The patches expose new kernel configuration flags, enabling system builders to test the capability on Armv9.3 silicon. However, critical components such as KVM virtualization and Kernel Address Sanitizer remain unsupported, highlighting a phased rollout. The community’s collaborative review process aims to resolve these gaps before the feature graduates to mainline status, likely within the next kernel release cycle.

For cloud providers and hyperscale operators, the ability to address larger memory regions directly translates into cost efficiencies and performance gains. ARM‑based instances can now compete more aggressively with x86 offerings for memory‑intensive workloads, reducing the need for complex memory tiering. As the ecosystem matures, software vendors will gain the flexibility to design applications that leverage the expanded address space without resorting to cumbersome workarounds, accelerating ARM’s adoption in enterprise data centers.