Re: [PATCH v2 05/15] gpu: nova-core: add the per-architecture GIN CPU interrupt HAL

["Alexandre Courbot" <acourbot@nvidia.com>]

On Sat Aug 29, 2026 at 10:22 AM JST, John Hubbard wrote:
> GIN, the GPU Interrupt and Notification unit, is the GPU's interrupt
> controller. Each PCIe function has its own tree, whose leaf count
> depends on the GPU family.
>
> Message-signaled delivery stops after each edge until the CPU rearms it,
> and the rearm write differs by family and interrupt type:
>
>     * Pre-Hopper MSI writes an EOI through the BAR0 PCI configuration
>     space mirror.
>
>     * MSI for Hopper and later cycles the TOP enable bits of every
>     serviced subtree.
>
>     * MSI-X on any family cycles the bits of the handler's own subtree.

The indentation of this bullet list is a bit unconventional for a kernel
git log (`*` typically starts at column 1).

<...>
> +impl PciIrqRearmMethod {
> +    /// Performs this method's register write.
> +    ///
> +    /// `serviced` holds every subtree the driver services, and `subtree` is the one subtree the
> +    /// calling handler serves. Each method uses whichever of the two its interrupt type delivers
> +    /// on, so both are required.
> +    pub(super) fn rearm(self, bar: Bar0<'_>, serviced: SubtreeSet, subtree: Subtree) {
> +        let subtrees = match self {
> +            // The written value is ignored, so any write rearms delivery.
> +            Self::ConfigMirrorEoi => {
> +                bar.write(regs::NV_XVE_CYA_2, 0u32.into());
> +                return;
> +            }
> +            Self::TopEnableCycleServiced => serviced,
> +            Self::TopEnableCycleSubtree => SubtreeSet::from(subtree),
> +        };
> +
> +        bar.write(
> +            regs::NV_VIRTUAL_FUNCTION_PRIV_CPU_INTR_TOP_EN_CLEAR,
> +            subtrees.into_raw().into(),
> +        );
> +        bar.write(
> +            regs::NV_VIRTUAL_FUNCTION_PRIV_CPU_INTR_TOP_EN_SET,
> +            subtrees.into_raw().into(),
> +        );

With the typed register fields in patch 3, these can become:

    bar.write_reg(
        regs::NV_VIRTUAL_FUNCTION_PRIV_CPU_INTR_TOP_EN_CLEAR::zeroed().with_subtrees(subtrees),
    );
    bar.write_reg(
        regs::NV_VIRTUAL_FUNCTION_PRIV_CPU_INTR_TOP_EN_SET::zeroed().with_subtrees(subtrees),
    );

> +    }
> +}
> +
> +/// Per-architecture properties of the GIN CPU interrupt tree.
> +///
> +/// The tree size and the method that rearms PCI interrupt delivery differ by family. The tree
> +/// walk, the vector encoding, and the read-and-clear sequence do not, and are in generic code.

The last sentence sounds a bit self-evident and unneeded.

> +///
> +/// See `Documentation/gpu/nova/core/interrupts.rst`.
> +pub(super) trait CpuInterruptHal {
> +    /// Returns the number of leaves the CPU tree implements.
> +    ///
> +    /// [`LeafCount::subtree_set`] gives the subtrees behind them, and
> +    /// [`LeafCount::vector_count`] the vectors they carry.
> +    fn leaf_count(&self) -> LeafCount;
> +
> +    /// Returns the method that rearms PCI interrupt delivery for `irq_type`.
> +    ///
> +    /// `None` means that `irq_type` needs no rearm write. That is the case for `INTx`, which is
> +    /// level-triggered, and which nova-core does not allocate.
> +    fn pci_irq_rearm_method(&self, irq_type: IrqType) -> Option<PciIrqRearmMethod>;
> +}
> +
> +/// Returns the [`CpuInterruptHal`] for `chipset`.
> +pub(super) fn cpu_interrupt_hal(chipset: Chipset) -> &'static dyn CpuInterruptHal {
> +    match chipset.arch() {
> +        Architecture::Turing | Architecture::Ampere | Architecture::Ada => tu102::TU102_HAL,
> +        Architecture::Hopper | Architecture::BlackwellGB10x | Architecture::BlackwellGB20x => {
> +            gh100::GH100_HAL
> +        }
> +    }
> +}
> diff --git a/drivers/gpu/nova-core/irq/hal/gh100.rs b/drivers/gpu/nova-core/irq/hal/gh100.rs
> new file mode 100644
> index 000000000000..32b9b4a01adb
> --- /dev/null
> +++ b/drivers/gpu/nova-core/irq/hal/gh100.rs
> @@ -0,0 +1,31 @@
> +// SPDX-License-Identifier: GPL-2.0
> +// SPDX-FileCopyrightText: Copyright (c) 2026 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
> +
> +use kernel::pci::IrqType;
> +
> +use super::{
> +    CpuInterruptHal,
> +    LeafCount,
> +    PciIrqRearmMethod, //
> +};
> +
> +/// GIN parameters for Hopper and Blackwell, which implement a 16-leaf CPU tree. Only 12 leaves
> +/// carry sources.
> +struct Gh100;
> +
> +impl CpuInterruptHal for Gh100 {
> +    fn leaf_count(&self) -> LeafCount {
> +        LeafCount::Sixteen
> +    }
> +
> +    fn pci_irq_rearm_method(&self, irq_type: IrqType) -> Option<PciIrqRearmMethod> {
> +        match irq_type {
> +            IrqType::Intx => None,
> +            IrqType::Msi => Some(PciIrqRearmMethod::TopEnableCycleServiced),
> +            IrqType::MsiX => Some(PciIrqRearmMethod::TopEnableCycleSubtree),
> +        }
> +    }

AFAIU we do not support INTx at all, right? In this case it just
shouldn't be handled here, and we could simplify this method to just
return `PciIrqRearmMethod`.

I suspect we are considering INTx because we work with the kernel's
`IrqType`, which includes it, and must make our match arms exhaustive.
Let's just use our own IRQ type for Nova:

    pub(crate) enum MsiType {
        Msi,
        MsiX,
    }

Then every instance of `IrqType` in the `irq` module can be replaced by
this one, and we don't need to care about INTx anymore. This method can
also just return a `PciIrqRearmMethod`.

You will probably need to include a `MsiType` into `SubtreeVectors` to
make its `irq_type` method work, but that's still better than having
code for managing an INTx variant that is dead code since it would have
failed at probe time anyway.