Kernel: Apple USB4, qdisc DoS clamps, VMA locks

Kernel traffic centered on first USB4 support for Apple silicon, security clamps and BPF fixes, and memory-management locking cleanups. A sysctl refactoring drew a direct NAK while Rust interrupt guards saw a revert after syzbot findings.

Initial USB4/Thunderbolt support for Apple M1/M2/M3

A 19-patch series adds the first ACIO/NHI driver and device-tree nodes for USB4/Thunderbolt on Apple M1, M2, and M3 SoCs. The proposal lands on linux-kernel as the starting point for native support of these interfaces. Developers tracking Apple silicon upstreaming gain a concrete path for high-speed I/O beyond existing drivers.

Clamps on quantum and psched_mtu in qdisc change paths

An eight-patch netdev series adds missing clamps on quantum and psched_mtu in the change paths of eight qdiscs. The work blocks crafted size-table denial-of-service under the qdisc lock. Traffic-control maintainers and operators gain protection against a practical lock-held DoS vector.

Unconditional per-VMA locks and related cleanups

Version 6 of a five-patch mm series makes per-VMA locks unconditional on all configs. The change enables simplifications for binder and networking code and has been picked up by Andrew Morton. Subsystems that previously carried conditional locking paths can now rely on a uniform model.

Stack out-of-bounds write fix in BPF cgroup link update

A bpf-next patch validates replacement programs against attach type to stop a stack out-of-bounds write during cgroup link update. The correction removes a memory-corruption path in the link-update logic. BPF and cgroup users obtain a targeted safety fix before wider exposure.

Linus NAK of sysctl_field SCTP conversion

An RFC converting SCTP sysctls to sysctl_field was NAKed by Linus Torvalds for illegible macros and indentation. The author acknowledged the criticism and is revising the series. The exchange underscores readability requirements that still bind core kernel interfaces.

Race handling in flush_sigqueue via list_del_init_careful

A patch switches flush_sigqueue to list_del_init_careful after discussion of a race between lockless flush and POSIX timer delivery after non-leader execve. Multiple developers examined the interaction on linux-kernel. Correct signal-queue teardown under exec remains material for process reliability.

Revert of Rust SpinLockIrq scoped-guard change

Peter Zijlstra reverted a locking/irq guard change after Syzbot reported unsafe nested lock patterns inside scoped_guard(spinlock_irq). The broader v4 series on refcounted interrupt disable and SpinLockIrq for Rust continues under review. Solid interrupt primitives remain a gating item for Rust kernel code.

Restored u32 index bound check in BPF queue/stack maps

A two-patch bpf series restores the u32 index bound check removed in an earlier commit, preventing wrap when max_entries times value_size reaches or exceeds U32_MAX. The fix closes an overflow that could produce incorrect map access. Queue and stack map users regain the original safety boundary.