Rust for Linux List
 help / color / mirror / Atom feed
From: Vasileios Almpanis <vasilisalmpanis@gmail.com>
To: "Danilo Krummrich" <dakr@kernel.org>,
	"Abdiel Janulgue" <abdiel.janulgue@gmail.com>,
	"Daniel Almeida" <daniel.almeida@collabora.com>,
	"Robin Murphy" <robin.murphy@arm.com>,
	"Andreas Hindborg" <a.hindborg@kernel.org>,
	"Miguel Ojeda" <ojeda@kernel.org>,
	"Boqun Feng" <boqun@kernel.org>, "Gary Guo" <gary@garyguo.net>,
	"Björn Roy Baron" <bjorn3_gh@protonmail.com>,
	"Benno Lossin" <lossin@kernel.org>,
	"Alice Ryhl" <aliceryhl@google.com>,
	"Trevor Gross" <tmgross@umich.edu>,
	"Tamir Duberstein" <tamird@kernel.org>,
	"Alexandre Courbot" <acourbot@nvidia.com>,
	"Onur Özkan" <work@onurozkan.dev>
Cc: driver-core@lists.linux.dev, rust-for-linux@vger.kernel.org,
	 linux-kernel@vger.kernel.org,
	 Vasileios Almpanis <vasilisalmpanis@gmail.com>
Subject: [PATCH] rust: dma: add `Range` type
Date: Thu, 06 Aug 2026 09:46:14 +0200	[thread overview]
Message-ID: <20260806-dma-v1-1-e8a39327c7f5@gmail.com> (raw)

The base dma address of `Coherent` and `CoherentHandle` is a bare
`DmaAddress` integer so any arithmetic, a driver does on it is unchecked
and can potentially go past the end of the allocation or overflow.

Add a `dma::Range` type that couples a base `DmaAddress` with a length
and only hands out addresses and sub-ranges within `[start, start + len)`
with all arithmetics checked against both the length and the overflow of
of the underlying `dma_addr_t`. Let `Coherent` and `CoherentHandle`
provide the `Range` covering their allocation.

Suggested-by: Danilo Krummrich <dakr@kernel.org>
Link: https://github.com/Rust-for-Linux/linux/issues/1248
Signed-off-by: Vasileios Almpanis <vasilisalmpanis@gmail.com>
---
This is one of my first rust-for-linux patches, so any suggestions are
extremely welcome. Some words about the decisions taken:

- I chose lengths offsets to be `DmaAddress` instead of usize since the
  bus space can be 64-bit on 32-bit CPUS.

- The constructor returns EOVERFLOW for `dma_addr_t` overflow, while
  out-of-bounds requests return EINVAL; happy to use a single error
  code if preferred.

- `dma_range()` is added to both `Coherent` and `CoherentHandle` so the
  type has users from the start. I could split it into a follow-up if
  that is preferred.

Tested with `make rustdoc`, `make rustfmtcheck`, `CLIPPY=1`, and the
KUnit doctests `rust_doctests_kernel`.
---
 rust/kernel/dma.rs | 174 +++++++++++++++++++++++++++++++++++++++++++++++++++++
 1 file changed, 174 insertions(+)

diff --git a/rust/kernel/dma.rs b/rust/kernel/dma.rs
index 200def84fb69e006bca0b1c578bac9f1dc8da708..e65e99ae4d915ac2e385df8b72fc518c07e4c82d 100644
--- a/rust/kernel/dma.rs
+++ b/rust/kernel/dma.rs
@@ -41,6 +41,142 @@
 /// Note that this may be `u64` even on 32-bit architectures.
 pub type DmaAddress = bindings::dma_addr_t;
 
+/// A range of DMA addresses.
+///
+/// Couples a base [`DmaAddress`] with the length in bytes of the region it belongs to,
+/// representing the half-open range `[start, start + len)` of DMA addresses.
+///
+/// Unlike a bare [`DmaAddress`], a [`Range`] only hands out addresses and sub-ranges that are
+/// guaranteed to lie within `[start, start + len)`; all arithmetic is checked against both the
+/// length of the range and overflow of the underlying [`DmaAddress`].
+///
+/// # Invariants
+///
+/// `start + len` does not overflow [`DmaAddress`].
+#[derive(Clone, Copy, Debug, PartialEq, Eq)]
+pub struct Range {
+    start: DmaAddress,
+    len: DmaAddress,
+}
+
+impl Range {
+    /// Creates a new [`Range`] of `len` bytes, starting at `start`.
+    ///
+    /// Returns [`EOVERFLOW`] if `start + len` overflows [`DmaAddress`].
+    ///
+    /// # Examples
+    ///
+    /// ```
+    /// use kernel::dma::{DmaAddress, Range};
+    ///
+    /// let range = Range::new(0x1000, 0x200)?;
+    /// assert_eq!(range.start(), 0x1000);
+    /// assert_eq!(range.end(), 0x1200);
+    /// assert_eq!(range.len(), 0x200);
+    ///
+    /// assert!(Range::new(DmaAddress::MAX, 1).is_err());
+    /// # Ok::<(), Error>(())
+    /// ```
+    #[inline]
+    pub const fn new(start: DmaAddress, len: DmaAddress) -> Result<Self> {
+        if start.checked_add(len).is_none() {
+            return Err(EOVERFLOW);
+        }
+
+        // INVARIANT: We just checked that `start + len` does not overflow `DmaAddress`.
+        Ok(Self { start, len })
+    }
+
+    /// Returns the first address of the range.
+    #[inline]
+    pub const fn start(&self) -> DmaAddress {
+        self.start
+    }
+
+    /// Returns the first address after the end of the range.
+    #[inline]
+    pub const fn end(&self) -> DmaAddress {
+        // By the type invariant, `start + len` does not overflow `DmaAddress`.
+        self.start + self.len
+    }
+
+    /// Returns the length of the range in bytes.
+    #[inline]
+    pub const fn len(&self) -> DmaAddress {
+        self.len
+    }
+
+    /// Returns `true` if the range is empty.
+    #[inline]
+    pub const fn is_empty(&self) -> bool {
+        self.len == 0
+    }
+
+    /// Returns the address at `offset` bytes into the range.
+    ///
+    /// The returned address is guaranteed to lie within the range; returns [`EINVAL`] if `offset`
+    /// is not smaller than the length of the range.
+    ///
+    /// # Examples
+    ///
+    /// ```
+    /// use kernel::dma::Range;
+    ///
+    /// let range = Range::new(0x1000, 0x200)?;
+    ///
+    /// assert_eq!(range.address(0)?, 0x1000);
+    /// assert_eq!(range.address(0x1ff)?, 0x11ff);
+    /// assert!(range.address(0x200).is_err());
+    /// # Ok::<(), Error>(())
+    /// ```
+    #[inline]
+    pub const fn address(&self, offset: DmaAddress) -> Result<DmaAddress> {
+        if offset >= self.len {
+            return Err(EINVAL);
+        }
+
+        // By the type invariant, `start + offset < start + len` does not overflow `DmaAddress`.
+        Ok(self.start + offset)
+    }
+
+    /// Returns the sub-range of `len` bytes, starting `offset` bytes into the range.
+    ///
+    /// The returned range is guaranteed to lie within the range; returns [`EINVAL`] if
+    /// `offset + len` overflows [`DmaAddress`] or exceeds the length of the range.
+    ///
+    /// # Examples
+    ///
+    /// ```
+    /// use kernel::dma::Range;
+    ///
+    /// let range = Range::new(0x1000, 0x200)?;
+    ///
+    /// let sub = range.subrange(0x100, 0x80)?;
+    /// assert_eq!(sub.start(), 0x1100);
+    /// assert_eq!(sub.end(), 0x1180);
+    ///
+    /// assert!(range.subrange(0x100, 0x101).is_err());
+    /// # Ok::<(), Error>(())
+    /// ```
+    #[inline]
+    pub const fn subrange(&self, offset: DmaAddress, len: DmaAddress) -> Result<Self> {
+        let Some(end) = offset.checked_add(len) else {
+            return Err(EINVAL);
+        };
+
+        if end > self.len {
+            return Err(EINVAL);
+        }
+
+        // INVARIANT: `start + offset + len <= start + self.len`, which by the type invariant of
+        // `self` does not overflow `DmaAddress`.
+        Ok(Self {
+            start: self.start + offset,
+            len,
+        })
+    }
+}
+
 /// Trait to be implemented by DMA capable bus devices.
 ///
 /// The [`dma::Device`](Device) trait should be implemented by bus specific device representations,
@@ -626,6 +762,25 @@ pub fn dma_handle(&self) -> DmaAddress {
         self.dma_handle
     }
 
+    /// Returns the [`Range`] of DMA addresses covering this allocation.
+    ///
+    /// Unlike [`Self::dma_handle`], which hands out the base address as a bare integer, the
+    /// returned [`Range`] couples the base address with the size of the allocation, such that
+    /// any offset arithmetic performed on it is checked.
+    #[inline]
+    pub fn dma_range(&self) -> Range {
+        // INVARIANT: By the type invariants of `Self`, `dma_handle` is the DMA address base of an
+        // allocated region of `self.size()` bytes; the DMA API guarantees that a mapped region
+        // never wraps the DMA address space, hence `dma_handle + size` does not overflow
+        // `DmaAddress`.
+        Range {
+            start: self.dma_handle,
+            // CAST: `usize` always fits in `DmaAddress`, which is at least 32 bits wide and
+            // always 64 bits wide on 64-bit architectures.
+            len: self.size() as DmaAddress,
+        }
+    }
+
     /// Returns a reference to the data in the region.
     ///
     /// # Safety
@@ -1101,6 +1256,25 @@ pub fn dma_handle(&self) -> DmaAddress {
         self.dma_handle
     }
 
+    /// Returns the [`Range`] of DMA addresses covering this allocation.
+    ///
+    /// Unlike [`Self::dma_handle`], which hands out the base address as a bare integer, the
+    /// returned [`Range`] couples the base address with the size of the allocation, such that
+    /// any offset arithmetic performed on it is checked.
+    #[inline]
+    pub fn dma_range(&self) -> Range {
+        // INVARIANT: By the type invariants of `Self`, `dma_handle` is the DMA address base of an
+        // allocated region of `self.size` bytes; the DMA API guarantees that a mapped region
+        // never wraps the DMA address space, hence `dma_handle + size` does not overflow
+        // `DmaAddress`.
+        Range {
+            start: self.dma_handle,
+            // CAST: `usize` always fits in `DmaAddress`, which is at least 32 bits wide and
+            // always 64 bits wide on 64-bit architectures.
+            len: self.size as DmaAddress,
+        }
+    }
+
     /// Returns the size in bytes of this allocation.
     #[inline]
     pub fn size(&self) -> usize {

---
base-commit: dc01dfb37b34beeefcfe1c3055364d41a4070c7e
change-id: 20260805-dma-7c5330baadb6

Best regards,
-- 
Vasileios Almpanis <vasilisalmpanis@gmail.com>


             reply	other threads:[~2026-08-06  7:46 UTC|newest]

Thread overview: 3+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2026-08-06  7:46 Vasileios Almpanis [this message]
2026-08-06 11:23 ` [PATCH] rust: dma: add `Range` type Robin Murphy
2026-08-06 13:40   ` Vasileios Almpanis

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=20260806-dma-v1-1-e8a39327c7f5@gmail.com \
    --to=vasilisalmpanis@gmail.com \
    --cc=a.hindborg@kernel.org \
    --cc=abdiel.janulgue@gmail.com \
    --cc=acourbot@nvidia.com \
    --cc=aliceryhl@google.com \
    --cc=bjorn3_gh@protonmail.com \
    --cc=boqun@kernel.org \
    --cc=dakr@kernel.org \
    --cc=daniel.almeida@collabora.com \
    --cc=driver-core@lists.linux.dev \
    --cc=gary@garyguo.net \
    --cc=linux-kernel@vger.kernel.org \
    --cc=lossin@kernel.org \
    --cc=ojeda@kernel.org \
    --cc=robin.murphy@arm.com \
    --cc=rust-for-linux@vger.kernel.org \
    --cc=tamird@kernel.org \
    --cc=tmgross@umich.edu \
    --cc=work@onurozkan.dev \
    /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