Re: [PATCH 00/17] arm64: Support FEAT_NMI and Rework Exception Masking

Hi Jinjie
On Fri, Jul 03, 2026 at 06:01:18PM +0800, Jinjie Ruan wrote:
> Hi all,

Vladimir (cc'd) has also been working on NMI support, but hadn't posted
to the list yet (due to some delays with testing).

I haven't had the chance to look at this yet, but at a high-level we'll
need to figure out what to do going forwards.

Mark.

> This patch series implements support for the ARMv8.8-A/v9.3-A
> hardware NMI extension (FEAT_NMI), introducing the use of PSTATE.ALLINT
> to manage superpriority interrupts on arm64.
> 
> Background and Current Status
> -----------------------------
> 
> The hardware ecosystem is already mature and fully ready for FEAT_NMI:
> 
> - Production Hardware: Huawei's Kunpeng servers (such as the HIP12 based
>   platforms) already feature native hardware support for FEAT_NMI.
> 
> - Emulation: QEMU has integrated robust FEAT_NMI emulation[1] support
>   since 2024.
> 
> Despite the hardware readiness, upstream support for FEAT_NMI has been
> delayed. Previous patch[2] attempts tried to bundle ALLINT management
> directly into the existing DAIF abstraction layers. As Mark Rutland
> pointed out[3], pretending ALLINT or pseudo-NMI (GIC PMR) is part of
> DAIF creates convoluted, unmaintainable hacks that frequently fail to
> handle complex edge cases correctly (such as state escape during
> context switching).
> 
> Reworking Exception Masking (Mark's Feedback)
> ---------------------------------------------
> 
> Following Mark Rutland's strong recommendation ("We must clean up
> the existing approach before we add the real NMI support"), this series
> does not simply stack FEAT_NMI on top of the old framework. Instead, it
> completely reworks how the arm64 kernel manages abstract and logical
> exception masks.
> 
> Per Mark's guidelines, this series achieves the following architectural
> improvements:
> 
> 1. Entry/Exit Specific Helpers (a):
> 
>    Introduces abstract exception mask helpers specifically for exception
>    boundaries. They handle unified unmask-at-entry and mask-at-exit
>    behaviors. This decouples the entry/exit paths from raw DAIF
>    manipulation. In this series, these helpers are first refactored to
>    manage DAIF + PMR cleanly, preparing the ground before any FEAT_NMI
>    code is introduced.
> 
> 2. Logical Exception Mask Separation (b):
> 
>    Introduces a decoupled logical mask tracking mechanism that treats DAIF,
>    PMR, and ALLINT as separate, distinct elements. This enables accurate
>    irqflag tracking and debug assertions to save, restore, and validate all
>    elements without forcing them to fake or pollute a traditional DAIF
>    layout.
> 
> Production Bug Fixes & Integration
> -----------------------------------
> 
> On top of this solid architectural foundation, this series adds the actual
> support for FEAT_NMI (ALLINT management). Crucially, during baisc testing
> and validation on production Kunpeng (HIP12) servers, we identified and
> resolved several critical bugs.
> 
> The series is structured as follows:
> 
> - Patches 1-5: Clean up and rework the existing DAIF/PMR masking into
>   separate logical exception helpers (Pre-requisite refactoring).
> 
> - Patches 6-17: Add FEAT_NMI support for ARM64, including specific
>   stability fixes found on Kunpeng hardware and QEmu.
> 
> Any feedback, testing, or review, especially regarding the exception
> masking refactoring, is highly appreciated.
> 
> [1]: https://lore.kernel.org/all/[email protected]/
> [2]: https://lore.kernel.org/linux-arm-kernel/[email protected]/
> [3]: https://lore.kernel.org/linux-arm-kernel/Y5c9SLeJacLYHmP7@FVFF77S0Q05N/
> 
> Jinjie Ruan (5):
>   arm64: Move DAIF macros to ptrace.h and use them centrally
>   arm64: Rework exception masking into abstract logical mask
>   arm64: entry: arm64: entry: Move DAIF masking for EL1 exit to C code
>   arm64: entry: Add entry-specific helpers
>   arm64: Introduce helpers for restoring standard exception masks
> 
> Lorenzo Pieralisi (1):
>   irqchip/gic-v3: Implement FEAT_GICv3_NMI support
> 
> Mark Brown (11):
>   arm64/booting: Document boot requirements for FEAT_NMI
>   arm64/sysreg: Add definitions for immediate versions of MSR ALLINT
>   arm64/hyp-stub: Enable access to ALLINT
>   arm64/idreg: Add an override for FEAT_NMI
>   arm64/cpufeature: Detect PE support for FEAT_NMI
>   KVM: arm64: Hide FEAT_NMI from guests
>   arm64/nmi: Manage masking for superpriority interrupts along with DAIF
>   arm64/entry: Don't call preempt_schedule_irq() with NMIs masked
>   arm64/irq: Document handling of FEAT_NMI in irqflags.h
>   arm64/nmi: Add handling of superpriority interrupts as NMIs
>   arm64/nmi: Add Kconfig for NMI
> 
>  Documentation/arch/arm64/booting.rst     |   6 +
>  arch/arm64/Kconfig                       |  17 ++
>  arch/arm64/include/asm/arch_gicv3.h      |   7 +-
>  arch/arm64/include/asm/assembler.h       |  17 +-
>  arch/arm64/include/asm/cpufeature.h      |   5 +
>  arch/arm64/include/asm/cpuidle.h         |  30 ++-
>  arch/arm64/include/asm/daifflags.h       | 144 --------------
>  arch/arm64/include/asm/efi.h             |  22 ++-
>  arch/arm64/include/asm/el2_setup.h       |  13 ++
>  arch/arm64/include/asm/entry-common.h    |  11 +-
>  arch/arm64/include/asm/exception_masks.h | 232 +++++++++++++++++++++++
>  arch/arm64/include/asm/irq.h             |   2 +
>  arch/arm64/include/asm/irqflags.h        |  10 +
>  arch/arm64/include/asm/kvm_host.h        |   2 +-
>  arch/arm64/include/asm/mmu_context.h     |   2 +-
>  arch/arm64/include/asm/nmi.h             |  23 +++
>  arch/arm64/include/asm/ptrace.h          |  11 +-
>  arch/arm64/include/asm/sysreg.h          |   2 +
>  arch/arm64/include/uapi/asm/ptrace.h     |   1 +
>  arch/arm64/kernel/acpi.c                 |  14 +-
>  arch/arm64/kernel/cpufeature.c           |  58 +++++-
>  arch/arm64/kernel/debug-monitors.c       |   9 +-
>  arch/arm64/kernel/entry-common.c         | 167 +++++++++++-----
>  arch/arm64/kernel/entry.S                |   4 -
>  arch/arm64/kernel/hibernate.c            |  10 +-
>  arch/arm64/kernel/idle.c                 |   7 +-
>  arch/arm64/kernel/irq.c                  |  36 +++-
>  arch/arm64/kernel/machine_kexec.c        |   4 +-
>  arch/arm64/kernel/pi/idreg-override.c    |   1 +
>  arch/arm64/kernel/probes/kprobes.c       |   9 +-
>  arch/arm64/kernel/process.c              |   7 +-
>  arch/arm64/kernel/setup.c                |   4 +-
>  arch/arm64/kernel/signal.c               |   2 +-
>  arch/arm64/kernel/smp.c                  |  22 +--
>  arch/arm64/kernel/suspend.c              |  15 +-
>  arch/arm64/kernel/traps.c                |   2 +-
>  arch/arm64/kvm/emulate-nested.c          |   6 +-
>  arch/arm64/kvm/hyp/include/hyp/switch.h  |   6 +
>  arch/arm64/kvm/hyp/nvhe/host.S           |   4 +-
>  arch/arm64/kvm/hyp/nvhe/hyp-init.S       |   3 +-
>  arch/arm64/kvm/hyp/nvhe/hyp-main.c       |   4 +-
>  arch/arm64/kvm/hyp/vgic-v3-sr.c          |   6 +-
>  arch/arm64/kvm/hyp/vhe/switch.c          |   6 +-
>  arch/arm64/kvm/reset.c                   |   6 +-
>  arch/arm64/mm/fault.c                    |   2 +-
>  arch/arm64/mm/mmu.c                      |   6 +-
>  arch/arm64/tools/cpucaps                 |   2 +
>  drivers/firmware/psci/psci.c             |   7 +-
>  drivers/irqchip/irq-gic-v3.c             | 152 +++++++++++++--
>  include/linux/irqchip/arm-gic-v3.h       |   4 +
>  50 files changed, 822 insertions(+), 320 deletions(-)
>  delete mode 100644 arch/arm64/include/asm/daifflags.h
>  create mode 100644 arch/arm64/include/asm/exception_masks.h
>  create mode 100644 arch/arm64/include/asm/nmi.h
> 
> -- 
> 2.34.1
>