Kernel build speedups and BPF verifier work

Kernel developers focused on large kbuild redesigns that claim major reductions in compile time, alongside several BPF series tightening verifier rules and enabling richer control flow. Memory management and Rust kernel library work also advanced through multi-patch review.

kbuild series targets large build-time gains

Two related patch series on linux-kernel propose substantial speedups for kernel builds, with claims ranging from roughly 36 percent to 90 percent on noop rebuilds. The work centers on parallel compression with gzip and zstd, lower .mod.c overhead, faster dependency checking, ELF parsing changes, and link-time adjustments. Linus Torvalds and other maintainers are actively reviewing the sets, which matter because faster iteration cycles directly affect day-to-day kernel development productivity.

BPF gains verified indirect calls via callx

Alexei Starovoitov posted a 17-patch series adding a callx instruction that lets the verifier accept indirect calls to static BPF subprograms. The patches include libbpf and rodata support intended for vtables and Rust use cases. The change expands what safe BPF programs can express without weakening the verifier’s static guarantees.

Verifier scopes callback arguments to their frame

A six-patch series introduces REF_TYPE_FRAME so the BPF verifier can ensure callback arguments do not outlive the frame that owns them. The fix addresses potential kernel memory exposure through ringbuf and map helpers. Readers tracking BPF security will care because the change closes a class of lifetime bugs that could otherwise leak or corrupt kernel state.

bpf_throw() gains exception cleanup landing pads

Version 5 of a 21-patch series adds .bpf_cleanup tables and matching verifier support so bpf_throw() can run landing pads for resources owned by the unwinding frame. The design is aimed at enabling Rust panic paths inside BPF programs. It matters for anyone building higher-level languages on top of the BPF runtime that need reliable resource release on exceptional control flow.

PMD-level swap entries for anonymous THPs

A resend of the v7 29-patch set adds PMD-level swap entries so anonymous transparent huge pages can be swapped without first splitting them into PTEs. The series continues under active discussion among mm developers. Keeping THPs intact across swap reduces fragmentation and page-table overhead for large anonymous mappings.

Rust SPDM requester library reaches v3

A 21-part patchset posts a Rust implementation of the SPDM requester for in-kernel device authentication and attestation, including over PCIe. The library is intended to give drivers a memory-safe path for security-protocol exchanges with devices. Kernel developers working on confidential computing or hardware root-of-trust flows have a concrete Rust component to review.

Proxy execution made compatible with sched_ext

Version 14 of a sched_ext patchset adds hooks and admission rules so proxy execution and sched_ext can be built together, removing their previous mutual exclusion under CONFIG options. The work is aimed at the 7.4 merge window. It lets scheduler developers combine the two extensibility mechanisms instead of choosing one or the other.