dri-devel.lists.freedesktop.org archive mirror
 help / color / mirror / Atom feed
From: "Alexandre Courbot" <acourbot@nvidia.com>
To: "Joel Fernandes" <joelagnelf@nvidia.com>,
	linux-kernel@vger.kernel.org,
	"Danilo Krummrich" <dakr@kernel.org>,
	"David Airlie" <airlied@gmail.com>,
	"Simona Vetter" <simona@ffwll.ch>,
	"Miguel Ojeda" <ojeda@kernel.org>,
	"Alex Gaynor" <alex.gaynor@gmail.com>,
	"Boqun Feng" <boqun.feng@gmail.com>,
	"Gary Guo" <gary@garyguo.net>,
	"Björn Roy Baron" <bjorn3_gh@protonmail.com>,
	"Benno Lossin" <lossin@kernel.org>,
	"Andreas Hindborg" <a.hindborg@kernel.org>,
	"Alice Ryhl" <aliceryhl@google.com>,
	"Trevor Gross" <tmgross@umich.edu>
Cc: "John Hubbard" <jhubbard@nvidia.com>,
	"Alistair Popple" <apopple@nvidia.com>,
	<nouveau@lists.freedesktop.org>,
	<dri-devel@lists.freedesktop.org>,
	<rust-for-linux@vger.kernel.org>
Subject: Re: [PATCH 1/2] nova-core: Add a library for bitfields in Rust structs
Date: Mon, 25 Aug 2025 20:07:01 +0900	[thread overview]
Message-ID: <DCBGLCQVD1RF.6V5UT0NQ4GLB@nvidia.com> (raw)
In-Reply-To: <20250824135954.2243774-1-joelagnelf@nvidia.com>

Hi Joel,

On Sun Aug 24, 2025 at 10:59 PM JST, Joel Fernandes wrote:
> Add a minimal bitfield library for defining in Rust structures (called
> bitstruct), similar in concept to bit fields in C structs. This will be used
> for defining page table entries and other structures in nova-core.

This is basically a rewrite (with some improvements, and some missing
features) of the part of the `register!` macro that deals with
bitfields. We planned to extract it into its own macro, and the split is
already effective in its internal rules, so I'd suggest to just move the
relevant rules into a new macro (as it will give you a couple useful
features, like automatic conversion to enum types), and then apply your
improvements on top of it. Otherwise we will end up with two
implementations of the same thing, for no good justification IMHO.

We were also planning to move the `register!` macro into the kernel
crate this cycle so Tyr can use it, but this changes the plan a bit.
Actually it is helpful, since your version implements one thing that we
needed in the `register!` macro before moving it: visibility specifiers.
So I would do things in this order:

1. Extract the bitfield-related code from the `register!` macro into its
   own macro (in nova-core), and make `register!` call into it,
2. Add support for visibility specifiers and non-u32 types in your macro and
   `register!`,
3. Move both macros to the kernel crate (hopefully in time for the next
   merge window so Tyr can make use of them).

A few more comments inline.

>
> Signed-off-by: Joel Fernandes <joelagnelf@nvidia.com>
> ---
>  drivers/gpu/nova-core/bitstruct.rs | 149 +++++++++++++++++++++++++++++
>  drivers/gpu/nova-core/nova_core.rs |   1 +
>  2 files changed, 150 insertions(+)
>  create mode 100644 drivers/gpu/nova-core/bitstruct.rs
>
> diff --git a/drivers/gpu/nova-core/bitstruct.rs b/drivers/gpu/nova-core/bitstruct.rs
> new file mode 100644
> index 000000000000..661a75da0a9c
> --- /dev/null
> +++ b/drivers/gpu/nova-core/bitstruct.rs

I wonder whether this should go under the existing `bits.rs`, or be its
own module?

> @@ -0,0 +1,149 @@
> +// SPDX-License-Identifier: GPL-2.0
> +//
> +// bitstruct.rs — C-style library for bitfield-packed Rust structures
> +//
> +// A library that provides support for defining bit fields in Rust
> +// structures to circumvent lack of native language support for this.
> +//
> +// Similar usage syntax to the register! macro.

Eventually the `register!` macro is the one that should reference this
(simpler) one, so let's make it the reference. :)

> +
> +use kernel::prelude::*;
> +
> +/// Macro for defining bitfield-packed structures in Rust.
> +/// The size of the underlying storage type is specified with #[repr(TYPE)].
> +///
> +/// # Example (just for illustration)
> +/// ```rust
> +/// bitstruct! {
> +///     #[repr(u64)]
> +///     pub struct PageTableEntry {
> +///         0:0       present     as bool,
> +///         1:1       writable    as bool,
> +///         11:9      available   as u8,
> +///         51:12     pfn         as u64,
> +///         62:52     available2  as u16,
> +///         63:63     nx          as bool,

A note on syntax: for nova-core, we may want to use the `H:L` notation,
as this is what OpenRM uses, but in the larger kernel we might want to
use inclusive ranges (`L..=H`) as it will look more natural in Rust
code (and is the notation the `bits` module already uses).

> +///     }
> +/// }
> +/// ```
> +///
> +/// This generates a struct with methods:
> +/// - Constructor: `default()` sets all bits to zero.
> +/// - Field accessors: `present()`, `pfn()`, etc.
> +/// - Field setters: `set_present()`, `set_pfn()`, etc.
> +/// - Builder methods: `with_present()`, `with_pfn()`, etc.
> +/// - Raw conversion: `from_raw()`, `into_raw()`
> +#[allow(unused_macros)]
> +macro_rules! bitstruct {
> +    (
> +        #[repr($storage:ty)]
> +        $vis:vis struct $name:ident {
> +            $(
> +                $hi:literal : $lo:literal $field:ident as $field_type:tt
> +            ),* $(,)?
> +        }
> +    ) => {
> +        #[repr(transparent)]
> +        #[derive(Copy, Clone, Default)]
> +        $vis struct $name($storage);
> +
> +        impl $name {
> +            /// Create from raw value
> +            #[inline(always)]
> +            $vis const fn from_raw(val: $storage) -> Self {
> +                Self(val)
> +            }
> +
> +            /// Get raw value
> +            #[inline(always)]
> +            $vis const fn into_raw(self) -> $storage {
> +                self.0
> +            }
> +        }
> +
> +        impl core::fmt::Debug for $name {
> +            fn fmt(&self, f: &mut core::fmt::Formatter<'_>) -> core::fmt::Result {
> +                write!(f, "{}({:#x})", stringify!($name), self.0)
> +            }
> +        }
> +
> +        // Generate all field methods
> +        $(
> +            bitstruct_field_impl!($vis, $name, $storage, $hi, $lo, $field as $field_type);

Maybe use internal rules [1] instead of a private macro (that cannot be so
private :))

[1] https://lukaswirth.dev/tlborm/decl-macros/patterns/internal-rules.html

> +        )*
> +    };
> +}
> +
> +/// Helper to calculate mask for bit fields
> +#[allow(unused_macros)]
> +macro_rules! bitstruct_mask {
> +    ($hi:literal, $lo:literal, $storage:ty) => {{
> +        let width = ($hi - $lo + 1) as usize;
> +        let storage_bits = 8 * core::mem::size_of::<$storage>();
> +        if width >= storage_bits {
> +            <$storage>::MAX
> +        } else {
> +            ((1 as $storage) << width) - 1
> +        }
> +    }};
> +}

Is there a way to leverage the `genmask_*` functions of the `bits` module?

> +
> +#[allow(unused_macros)]
> +macro_rules! bitstruct_field_impl {
> +    ($vis:vis, $struct_name:ident, $storage:ty, $hi:literal, $lo:literal, $field:ident as $field_type:tt) => {
> +        impl $struct_name {
> +            #[inline(always)]
> +            $vis const fn $field(&self) -> $field_type {
> +                let field_val = (self.0 >> $lo) & bitstruct_mask!($hi, $lo, $storage);
> +                bitstruct_cast_value!(field_val, $field_type)
> +            }
> +        }
> +        bitstruct_make_setters!($vis, $struct_name, $storage, $hi, $lo, $field, $field_type);
> +    };
> +}
> +
> +/// Helper macro to convert extracted value to target type
> +///
> +/// Special handling for bool types is required because the `as` keyword
> +/// cannot be used to convert to bool in Rust. For bool fields, we check
> +/// if the extracted value is non-zero. For all other types, we use the
> +/// standard `as` conversion.
> +#[allow(unused_macros)]
> +macro_rules! bitstruct_cast_value {
> +    ($field_val:expr, bool) => {
> +        $field_val != 0
> +    };
> +    ($field_val:expr, $field_type:tt) => {
> +        $field_val as $field_type
> +    };
> +}
> +
> +#[allow(unused_macros)]
> +macro_rules! bitstruct_write_bits {
> +    ($raw:expr, $hi:literal, $lo:literal, $val:expr, $storage:ty) => {{
> +        let mask = bitstruct_mask!($hi, $lo, $storage);
> +        ($raw & !(mask << $lo)) | ((($val as $storage) & mask) << $lo)
> +    }};
> +}
> +
> +#[allow(unused_macros)]
> +macro_rules! bitstruct_make_setters {
> +    ($vis:vis, $struct_name:ident, $storage:ty, $hi:literal, $lo:literal, $field:ident, $field_type:tt) => {
> +        ::kernel::macros::paste! {
> +            impl $struct_name {
> +                #[inline(always)]
> +                #[allow(dead_code)]
> +                $vis fn [<set_ $field>](&mut self, val: $field_type) {
> +                    self.0 = bitstruct_write_bits!(self.0, $hi, $lo, val, $storage);
> +                }
> +
> +                #[inline(always)]
> +                #[allow(dead_code)]
> +                $vis const fn [<with_ $field>](mut self, val: $field_type) -> Self {
> +                    self.0 = bitstruct_write_bits!(self.0, $hi, $lo, val, $storage);
> +                    self
> +                }
> +            }
> +        }
> +    };
> +}

Yep, I think you definitely want to put these into internal rules - see
how `register!` does it, actually it should be easy to extract these
rules only and implement your improvements on top.

  parent reply	other threads:[~2025-08-25 11:07 UTC|newest]

Thread overview: 23+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2025-08-24 13:59 [PATCH 1/2] nova-core: Add a library for bitfields in Rust structs Joel Fernandes
2025-08-24 13:59 ` [PATCH 2/2] nova-core: Add KUNIT tests for bitstruct Joel Fernandes
2025-08-24 17:37 ` [PATCH 1/2] nova-core: Add a library for bitfields in Rust structs John Hubbard
2025-08-25  4:14 ` Boqun Feng
2025-08-25  4:16   ` Joel Fernandes
2025-08-25 10:42     ` Alexandre Courbot
2025-08-25 10:46 ` Danilo Krummrich
2025-08-25 11:07 ` Alexandre Courbot [this message]
2025-09-03 15:15   ` Joel Fernandes
2025-09-04  3:16     ` Alexandre Courbot
2025-09-04  7:16       ` Danilo Krummrich
2025-09-04 11:06         ` Alexandre Courbot
2025-09-05 21:29           ` John Hubbard
2025-09-06  1:58             ` Alexandre Courbot
2025-09-04 11:33         ` Joel Fernandes
2025-09-04 11:02       ` Daniel Almeida
2025-09-04 11:32       ` Joel Fernandes
2025-08-25 23:20 ` Elle Rhumsaa
2025-09-03 21:52   ` Joel Fernandes
2025-09-03 13:29 ` Daniel Almeida
2025-09-03 17:54   ` Joel Fernandes
2025-09-04 21:35 ` Yury Norov
2025-09-05 18:45   ` Joel Fernandes

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=DCBGLCQVD1RF.6V5UT0NQ4GLB@nvidia.com \
    --to=acourbot@nvidia.com \
    --cc=a.hindborg@kernel.org \
    --cc=airlied@gmail.com \
    --cc=alex.gaynor@gmail.com \
    --cc=aliceryhl@google.com \
    --cc=apopple@nvidia.com \
    --cc=bjorn3_gh@protonmail.com \
    --cc=boqun.feng@gmail.com \
    --cc=dakr@kernel.org \
    --cc=dri-devel@lists.freedesktop.org \
    --cc=gary@garyguo.net \
    --cc=jhubbard@nvidia.com \
    --cc=joelagnelf@nvidia.com \
    --cc=linux-kernel@vger.kernel.org \
    --cc=lossin@kernel.org \
    --cc=nouveau@lists.freedesktop.org \
    --cc=ojeda@kernel.org \
    --cc=rust-for-linux@vger.kernel.org \
    --cc=simona@ffwll.ch \
    --cc=tmgross@umich.edu \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).