Re: [PATCH v14 06/11] rust: dma: add dma addressing capabilities

[Alexandre Courbot]

Hi Abdiel,

On Wed Mar 12, 2025 at 2:48 AM JST, Abdiel Janulgue wrote:
> From: Danilo Krummrich <dakr@xxxxxxxxxx>
>
> Implement `dma_set_mask()` and `dma_set_mask_and_coherent()` in the
> `dma::Device` trait.
>
> Those methods are used to inform the kernel about the device's DMA
> addressing capabilities.
>
> Co-developed-by: Abdiel Janulgue <abdiel.janulgue@xxxxxxxxx>
> Signed-off-by: Abdiel Janulgue <abdiel.janulgue@xxxxxxxxx>
> Signed-off-by: Danilo Krummrich <dakr@xxxxxxxxxx>
> ---
>  rust/helpers/dma.c     |  8 ++++++++
>  rust/helpers/helpers.c |  1 +
>  rust/kernel/dma.rs     | 46 +++++++++++++++++++++++++++++++++++++++---
>  3 files changed, 52 insertions(+), 3 deletions(-)
>  create mode 100644 rust/helpers/dma.c
>
> diff --git a/rust/helpers/dma.c b/rust/helpers/dma.c
> new file mode 100644
> index 000000000000..8eb482386f93
> --- /dev/null
> +++ b/rust/helpers/dma.c
> @@ -0,0 +1,8 @@
> +// SPDX-License-Identifier: GPL-2.0
> +
> +#include <linux/dma-mapping.h>
> +
> +int rust_helper_dma_set_mask_and_coherent(struct device *dev, u64 mask)
> +{
> +	return dma_set_mask_and_coherent(dev, mask);
> +}
> diff --git a/rust/helpers/helpers.c b/rust/helpers/helpers.c
> index 0640b7e115be..8f3808c8b7fe 100644
> --- a/rust/helpers/helpers.c
> +++ b/rust/helpers/helpers.c
> @@ -13,6 +13,7 @@
>  #include "build_bug.c"
>  #include "cred.c"
>  #include "device.c"
> +#include "dma.c"
>  #include "err.c"
>  #include "fs.c"
>  #include "io.c"
> diff --git a/rust/kernel/dma.rs b/rust/kernel/dma.rs
> index 7ff797a7ad18..ac3ec0042327 100644
> --- a/rust/kernel/dma.rs
> +++ b/rust/kernel/dma.rs
> @@ -5,10 +5,10 @@
>  //! C header: [`include/linux/dma-mapping.h`](srctree/include/linux/dma-mapping.h)
>  
>  use crate::{
> -    bindings, build_assert,
> -    device,
> +    bindings, build_assert, device,
>      error::code::*,
>      error::Result,
> +    prelude::*,
>      transmute::{AsBytes, FromBytes},
>      types::ARef,
>  };
> @@ -18,7 +18,35 @@
>  /// The [`Device`] trait should be implemented by bus specific device representations, where the
>  /// underlying bus has potential support for DMA, such as [`crate::pci::Device`] or
>  /// [crate::platform::Device].
> -pub trait Device: AsRef<device::Device> {}
> +pub trait Device: AsRef<device::Device> {
> +    /// Inform the kernel about the device's DMA addressing capabilities.
> +    ///
> +    /// Set both the DMA mask and the coherent DMA mask to the same value.
> +    ///
> +    /// Note that we don't check the return value from the C `dma_set_coherent_mask` as the DMA API
> +    /// guarantees that the coherent DMA mask can be set to the same or smaller than the streaming
> +    /// DMA mask.
> +    fn dma_set_mask_and_coherent(&mut self, mask: u64) -> Result {
> +        // SAFETY: By the type invariant of `device::Device`, `self.as_ref().as_raw()` is valid.
> +        let ret = unsafe { bindings::dma_set_mask_and_coherent(self.as_ref().as_raw(), mask) };
> +        if ret != 0 {
> +            Err(Error::from_errno(ret))
> +        } else {
> +            Ok(())
> +        }
> +    }
> +
> +    /// Same as [`Self::dma_set_mask_and_coherent`], but set the mask only for streaming mappings.
> +    fn dma_set_mask(&mut self, mask: u64) -> Result {
> +        // SAFETY: By the type invariant of `device::Device`, `self.as_ref().as_raw()` is valid.
> +        let ret = unsafe { bindings::dma_set_mask(self.as_ref().as_raw(), mask) };
> +        if ret != 0 {
> +            Err(Error::from_errno(ret))
> +        } else {
> +            Ok(())
> +        }
> +    }
> +}

I'm not quite sure why this trait is needed.

The default implementations of the methods are not overridden in this
patchset and I don't see any scenario where they would need to be. Why
not do it the simple way by just adding an implementation block for
device::Device?

This also introduces a `&dyn Device` trait object as parameter to the
coherent allocation methods - not a big deal, but still inelegant IMHO.