Kernel build speedups, BPF callx and stack limits

Kernel and BPF developers advanced large multi-patch series on build performance, program stack limits, indirect calls, and memory management. Work spanned kbuild caching, VMA flag cleanup, THP swap handling, BTF inline encoding, and ext4 iomap conversion.

kbuild series targets large build speedups

A 22-patch v4 series proposes significant reductions in kernel build times through toolchain caching, modpost and objtool changes, parallel Rust and objtool work, and pigz. Claimed gains range from 36 to 90 percent depending on configuration. Frequent kernel builders would see direct wall-clock improvements if the series lands.

BPF callx enables indirect subprogram calls

A 17-patch series adding BPF callx, indirect calls of BPF subprograms via function pointers, landed on bpf-next. The work updates the JIT along with the verifier, liveness analysis, and call-graph handling. The change expands BPF control flow for more complex program structures.

BPF stack limit raised toward 2 KiB

Successive v3 and v4 patch series propose raising the BPF program stack limit from 512 bytes to 2 KiB on x86-64 and arm64. Verifier changes use dynamic sizing and new support hooks to avoid proportional memory growth. Larger stacks reduce artificial constraints on complex BPF programs.

VMA flags made explicit, VM_SPECIAL removed

A 40-patch v3 series refactors Linux mm VMA flags to eliminate VM_SPECIAL and clarify flag semantics. The work is under review by mm maintainers across a large discussion thread. Explicit semantics reduce long-standing ambiguity in core memory management.

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 survive swap without splitting to PTEs. The approach preserves THP benefits across reclaim. Memory-intensive workloads retain better performance under pressure.

BTF inline function support merged

An 11-patch series adding BTF kinds (LOC_PARAM, LOC_PROTO, LOCSEC) to encode inline function sites for kprobe tracing was reviewed and merged into bpf-next. The encoding improves observability of inlined code paths. Tracer authors gain better visibility into optimized kernels.

ext4 buffered I/O moves to iomap

Version 6 of a 31-patch series converts ext4 regular-file buffered I/O from buffer_head to iomap. The series addresses ordering, i_disksize, and data=ordered issues. The conversion aligns ext4 with modern I/O paths used elsewhere in the kernel.