[RFC PATCH v1 0/2] Add NonEmptySlice

[RFC PATCH v1 0/2] Add NonEmptySlice

[Francis Laniel]

Hi,

Following the discussion in:
https://lore.kernel.org/rust-for-linux/DM0D9TERQAM3.2Y3HV6P68E2N8@garyguo.net/
I wrote some code to add a new NonEmptySlice type which ensures the underlying
type is no empty. This way, caller can use first()/last() without relying on
unsafe.

For now, only new_mut() and last_mut() are used in uaccess.rs. Please let me
know if I need to remove the other functions from slice.rs to only what is
needed.

Francis Laniel (2):
  rust: add NonEmptySlice
  rust: uaccess: Use slice::NonEmptySlice in strcpy_into_buf()

 rust/kernel/lib.rs     |  1 +
 rust/kernel/slice.rs   | 76 ++++++++++++++++++++++++++++++++++++++++++
 rust/kernel/uaccess.rs | 11 +++---
 3 files changed, 81 insertions(+), 7 deletions(-)
 create mode 100644 rust/kernel/slice.rs

Best regards,
-- 
2.47.3

[RFC PATCH v1 1/2] rust: add NonEmptySlice

[Francis Laniel]

Add NonEmptySlice<T>, a [T] wrapper that is guaranteed to contain at
least one element.
This lets methods such as last_mut() return a plain reference instead
of an Option, so callers that have already checked for emptiness no
longer need unwrap() or unwrap_unchecked().

Suggested-by: Gary Guo <gary@garyguo.net>
Signed-off-by: Francis Laniel <laniel_francis@npnm.eu>
---
 rust/kernel/lib.rs   |  1 +
 rust/kernel/slice.rs | 76 ++++++++++++++++++++++++++++++++++++++++++++
 2 files changed, 77 insertions(+)
 create mode 100644 rust/kernel/slice.rs

diff --git a/rust/kernel/lib.rs b/rust/kernel/lib.rs
index 7225abc64084..8cc01aceb3b2 100644
--- a/rust/kernel/lib.rs
+++ b/rust/kernel/lib.rs
@@ -130,6 +130,7 @@
 #[cfg(CONFIG_RUST_SERIAL_DEV_BUS_ABSTRACTIONS)]
 pub mod serdev;
 pub mod sizes;
+pub mod slice;
 #[cfg(CONFIG_SOC_BUS)]
 pub mod soc;
 #[doc(hidden)]
diff --git a/rust/kernel/slice.rs b/rust/kernel/slice.rs
new file mode 100644
index 000000000000..7e467cb952c3
--- /dev/null
+++ b/rust/kernel/slice.rs
@@ -0,0 +1,76 @@
+// SPDX-License-Identifier: GPL-2.0
+
+//! Slice abstractions to ease development.
+
+use core::ops::{Deref, DerefMut};
+
+/// A slice that is guaranteed to contain at least one element.
+///
+/// # Invariants
+///
+/// `self.0` is not empty.
+#[repr(transparent)]
+pub struct NonEmptySlice<T>([T]);
+
+impl<T> NonEmptySlice<T> {
+    /// Returns `None` if `s` is empty.
+    pub fn new(s: &mut [T]) -> Option<&Self> {
+        if s.is_empty() {
+            return None;
+        }
+        // SAFETY: `Self` is `repr(transparent)` over `[T]`, so the cast keeps the slice length.
+        // INVARIANT: `s` is not empty, checked above.
+        Some(unsafe { &*(core::ptr::from_mut(s) as *const Self) })
+    }
+
+    /// Returns `None` if `s` is empty.
+    pub fn new_mut(s: &mut [T]) -> Option<&mut Self> {
+        if s.is_empty() {
+            return None;
+        }
+        // SAFETY: `Self` is `repr(transparent)` over `[T]`, so the cast keeps the slice length.
+        // INVARIANT: `s` is not empty, checked above.
+        Some(unsafe { &mut *(core::ptr::from_mut(s) as *mut Self) })
+    }
+
+    /// Returns a reference to the first element.
+    pub fn first(&mut self) -> &T {
+        // SAFETY: By the type invariant, `len >= 1`, so `0` is in bounds.
+        unsafe { self.0.get_unchecked(0) }
+    }
+
+    /// Returns a mutable reference to the first element.
+    pub fn first_mut(&mut self) -> &mut T {
+        // SAFETY: By the type invariant, `len >= 1`, so `0` is in bounds.
+        unsafe { self.0.get_unchecked_mut(0) }
+    }
+
+    /// Returns a reference to the last element.
+    pub fn last(&mut self) -> &T {
+        let last = self.0.len() - 1;
+        // SAFETY: By the type invariant, `len >= 1`, so `last` is in bounds.
+        unsafe { self.0.get_unchecked(last) }
+    }
+
+    /// Returns a mutable reference to the last element.
+    pub fn last_mut(&mut self) -> &mut T {
+        let last = self.0.len() - 1;
+        // SAFETY: By the type invariant, `len >= 1`, so `last` is in bounds.
+        unsafe { self.0.get_unchecked_mut(last) }
+    }
+}
+
+impl<T> Deref for NonEmptySlice<T> {
+    type Target = [T];
+
+    fn deref(&self) -> &[T] {
+        &self.0
+    }
+}
+
+impl<T> DerefMut for NonEmptySlice<T> {
+    // Mutating elements through `&mut [T]` can't change the length, so the invariant holds.
+    fn deref_mut(&mut self) -> &mut [T] {
+        &mut self.0
+    }
+}
-- 
2.47.3

[RFC PATCH v1 2/2] rust: uaccess: Use slice::NonEmptySlice in strcpy_into_buf()

[Francis Laniel]

This removes an unsafe block in strcpy_into_buf(), the unsafe
block now lies in last_mut() in NonEmptySlice.

Signed-off-by: Francis Laniel <laniel_francis@npnm.eu>
Suggested-by: Gary Guo <gary@garyguo.net>
---
 rust/kernel/uaccess.rs | 11 ++++-------
 1 file changed, 4 insertions(+), 7 deletions(-)

diff --git a/rust/kernel/uaccess.rs b/rust/kernel/uaccess.rs
index 5f6c4d7a1a51..b6bf9074d67a 100644
--- a/rust/kernel/uaccess.rs
+++ b/rust/kernel/uaccess.rs
@@ -13,6 +13,7 @@
     fs::file,
     prelude::*,
     ptr::KnownSize,
+    slice::NonEmptySlice,
     transmute::{AsBytes, FromBytes},
 };
 use core::mem::{size_of, MaybeUninit};
@@ -390,13 +391,11 @@ pub fn read_all<A: Allocator>(mut self, buf: &mut Vec<u8, A>, flags: Flags) -> R
     /// copied).
     #[doc(alias = "strncpy_from_user")]
     pub fn strcpy_into_buf<'buf>(self, buf: &'buf mut [u8]) -> Result<&'buf CStr> {
-        if buf.is_empty() {
-            return Err(EINVAL);
-        }
+        let buf = NonEmptySlice::new_mut(buf).ok_or(EINVAL)?;
 
         // SAFETY: The types are compatible and `strncpy_from_user` doesn't write uninitialized
         // bytes to `buf`.
-        let mut dst = unsafe { &mut *(core::ptr::from_mut(buf) as *mut [MaybeUninit<u8>]) };
+        let mut dst = unsafe { &mut *(core::ptr::from_mut(&mut **buf) as *mut [MaybeUninit<u8>]) };
 
         // We never read more than `self.length` bytes.
         if dst.len() > self.length {
@@ -422,9 +421,7 @@ pub fn strcpy_into_buf<'buf>(self, buf: &'buf mut [u8]) -> Result<&'buf CStr> {
             // This means that we filled the buffer exactly. In this case, we add a NUL-terminator
             // and return it. Unlike the `len < dst.len()` branch, don't modify `len` because it
             // already represents the length including the NUL-terminator.
-            //
-            // SAFETY: Due to the check at the beginning, the buffer is not empty.
-            unsafe { *buf.last_mut().unwrap_unchecked() = 0 };
+            *buf.last_mut() = 0;
         }
 
         // This method consumes `self`, so it can only be called once, thus we do not need to
-- 
2.47.3

Re: [RFC PATCH v1 0/2] Add NonEmptySlice

[Alexandre Courbot]

On Mon Oct 12, 2026 at 4:15 AM JST, Francis Laniel wrote:
> Hi,
>
> Following the discussion in:
> https://lore.kernel.org/rust-for-linux/DM0D9TERQAM3.2Y3HV6P68E2N8@garyguo.net/
> I wrote some code to add a new NonEmptySlice type which ensures the underlying
> type is no empty. This way, caller can use first()/last() without relying on
> unsafe.
>
> For now, only new_mut() and last_mut() are used in uaccess.rs. Please let me
> know if I need to remove the other functions from slice.rs to only what is
> needed.

Thanks for looking into this, but have you seen [1]? There is a Github
issue opened with some comments [2] and you might want to coordinate
with Lorenzo about who is going to work on this.

I'll try to comment on the code later today.

[1] https://lore.kernel.org/all/CANiq72=rwnogLqmpxnsvhsPWX+mBtH62=Yu_avQEYEmkWSVidQ@mail.gmail.com/
[2] https://github.com/Rust-for-Linux/linux/issues/1262

Re: [RFC PATCH v1 0/2] Add NonEmptySlice

[Francis Laniel]

Hi,

Le lundi 12 octobre 2026, 03:21:21 heure d’été d’Europe de l’Est Alexandre Courbot a écrit :
> On Mon Oct 12, 2026 at 4:15 AM JST, Francis Laniel wrote:
> > Hi,
> >
> > Following the discussion in:
> > https://lore.kernel.org/rust-for-linux/DM0D9TERQAM3.2Y3HV6P68E2N8@garyguo.net/
> > I wrote some code to add a new NonEmptySlice type which ensures the underlying
> > type is no empty. This way, caller can use first()/last() without relying on
> > unsafe.
> >
> > For now, only new_mut() and last_mut() are used in uaccess.rs. Please let me
> > know if I need to remove the other functions from slice.rs to only what is
> > needed.
> 
> Thanks for looking into this, but have you seen [1]? There is a Github
> issue opened with some comments [2] and you might want to coordinate
> with Lorenzo about who is going to work on this.

Sorry, I went faster than the music! Let me coordinate on GitHub with Lorenzo!

> I'll try to comment on the code later today.
> 
> [1] https://lore.kernel.org/all/CANiq72=rwnogLqmpxnsvhsPWX+mBtH62=Yu_avQEYEmkWSVidQ@mail.gmail.com/
> [2] https://github.com/Rust-for-Linux/linux/issues/1262
> 

Best regards,

Re: [RFC PATCH v1 1/2] rust: add NonEmptySlice

[Alexandre Courbot]

Looks like we are converging towards the series sent by Lorenzo, but
wanted to share a couple a quick comments nonetheless:

On Mon Oct 12, 2026 at 4:15 AM JST, Francis Laniel wrote:
> Add NonEmptySlice<T>, a [T] wrapper that is guaranteed to contain at
> least one element.
> This lets methods such as last_mut() return a plain reference instead
> of an Option, so callers that have already checked for emptiness no
> longer need unwrap() or unwrap_unchecked().
>
> Suggested-by: Gary Guo <gary@garyguo.net>
> Signed-off-by: Francis Laniel <laniel_francis@npnm.eu>
> ---
>  rust/kernel/lib.rs   |  1 +
>  rust/kernel/slice.rs | 76 ++++++++++++++++++++++++++++++++++++++++++++
>  2 files changed, 77 insertions(+)
>  create mode 100644 rust/kernel/slice.rs
>
> diff --git a/rust/kernel/lib.rs b/rust/kernel/lib.rs
> index 7225abc64084..8cc01aceb3b2 100644
> --- a/rust/kernel/lib.rs
> +++ b/rust/kernel/lib.rs
> @@ -130,6 +130,7 @@
>  #[cfg(CONFIG_RUST_SERIAL_DEV_BUS_ABSTRACTIONS)]
>  pub mod serdev;
>  pub mod sizes;
> +pub mod slice;
>  #[cfg(CONFIG_SOC_BUS)]
>  pub mod soc;
>  #[doc(hidden)]
> diff --git a/rust/kernel/slice.rs b/rust/kernel/slice.rs
> new file mode 100644
> index 000000000000..7e467cb952c3
> --- /dev/null
> +++ b/rust/kernel/slice.rs
> @@ -0,0 +1,76 @@
> +// SPDX-License-Identifier: GPL-2.0
> +
> +//! Slice abstractions to ease development.
> +
> +use core::ops::{Deref, DerefMut};
> +
> +/// A slice that is guaranteed to contain at least one element.
> +///
> +/// # Invariants
> +///
> +/// `self.0` is not empty.
> +#[repr(transparent)]
> +pub struct NonEmptySlice<T>([T]);
> +
> +impl<T> NonEmptySlice<T> {
> +    /// Returns `None` if `s` is empty.

This is not a standard doccomment. Doccomments first start by stating
what the function does before explaining the return value.

> +    pub fn new(s: &mut [T]) -> Option<&Self> {

It looks strange to take a `&mut` and return a `&Self`. There should be
no reason to take a mutable reference here, and you can use
`ptr::from_ref` to create the reference.

> +        if s.is_empty() {
> +            return None;
> +        }
> +        // SAFETY: `Self` is `repr(transparent)` over `[T]`, so the cast keeps the slice length.
> +        // INVARIANT: `s` is not empty, checked above.
> +        Some(unsafe { &*(core::ptr::from_mut(s) as *const Self) })
> +    }
> +
> +    /// Returns `None` if `s` is empty.
> +    pub fn new_mut(s: &mut [T]) -> Option<&mut Self> {
> +        if s.is_empty() {
> +            return None;
> +        }
> +        // SAFETY: `Self` is `repr(transparent)` over `[T]`, so the cast keeps the slice length.
> +        // INVARIANT: `s` is not empty, checked above.
> +        Some(unsafe { &mut *(core::ptr::from_mut(s) as *mut Self) })
> +    }
> +
> +    /// Returns a reference to the first element.
> +    pub fn first(&mut self) -> &T {

Same here, `first` and `last` should take a non-mutable reference as
their standard library counterparts do. They do not allow to mutate the
slice.

Re: [RFC PATCH v1 1/2] rust: add NonEmptySlice

[Francis Laniel]

Hi,

Le lundi 12 octobre 2026, 17:36:41 heure d’été d’Europe de l’Est Alexandre Courbot a écrit :
> Looks like we are converging towards the series sent by Lorenzo, but
> wanted to share a couple a quick comments nonetheless:

Thanks for the feedback!

> On Mon Oct 12, 2026 at 4:15 AM JST, Francis Laniel wrote:
> > Add NonEmptySlice<T>, a [T] wrapper that is guaranteed to contain at
> > least one element.
> > This lets methods such as last_mut() return a plain reference instead
> > of an Option, so callers that have already checked for emptiness no
> > longer need unwrap() or unwrap_unchecked().
> >
> > Suggested-by: Gary Guo <gary@garyguo.net>
> > Signed-off-by: Francis Laniel <laniel_francis@npnm.eu>
> > ---
> >  rust/kernel/lib.rs   |  1 +
> >  rust/kernel/slice.rs | 76 ++++++++++++++++++++++++++++++++++++++++++++
> >  2 files changed, 77 insertions(+)
> >  create mode 100644 rust/kernel/slice.rs
> >
> > diff --git a/rust/kernel/lib.rs b/rust/kernel/lib.rs
> > index 7225abc64084..8cc01aceb3b2 100644
> > --- a/rust/kernel/lib.rs
> > +++ b/rust/kernel/lib.rs
> > @@ -130,6 +130,7 @@
> >  #[cfg(CONFIG_RUST_SERIAL_DEV_BUS_ABSTRACTIONS)]
> >  pub mod serdev;
> >  pub mod sizes;
> > +pub mod slice;
> >  #[cfg(CONFIG_SOC_BUS)]
> >  pub mod soc;
> >  #[doc(hidden)]
> > diff --git a/rust/kernel/slice.rs b/rust/kernel/slice.rs
> > new file mode 100644
> > index 000000000000..7e467cb952c3
> > --- /dev/null
> > +++ b/rust/kernel/slice.rs
> > @@ -0,0 +1,76 @@
> > +// SPDX-License-Identifier: GPL-2.0
> > +
> > +//! Slice abstractions to ease development.
> > +
> > +use core::ops::{Deref, DerefMut};
> > +
> > +/// A slice that is guaranteed to contain at least one element.
> > +///
> > +/// # Invariants
> > +///
> > +/// `self.0` is not empty.
> > +#[repr(transparent)]
> > +pub struct NonEmptySlice<T>([T]);
> > +
> > +impl<T> NonEmptySlice<T> {
> > +    /// Returns `None` if `s` is empty.
> 
> This is not a standard doccomment. Doccomments first start by stating
> what the function does before explaining the return value.
> 
> > +    pub fn new(s: &mut [T]) -> Option<&Self> {
> 
> It looks strange to take a `&mut` and return a `&Self`. There should be
> no reason to take a mutable reference here, and you can use
> `ptr::from_ref` to create the reference.

Indeed, this sounds weird! I will keep it consistant for future contributions.

> > +        if s.is_empty() {
> > +            return None;
> > +        }
> > +        // SAFETY: `Self` is `repr(transparent)` over `[T]`, so the cast keeps the slice length.
> > +        // INVARIANT: `s` is not empty, checked above.
> > +        Some(unsafe { &*(core::ptr::from_mut(s) as *const Self) })
> > +    }
> > +
> > +    /// Returns `None` if `s` is empty.
> > +    pub fn new_mut(s: &mut [T]) -> Option<&mut Self> {
> > +        if s.is_empty() {
> > +            return None;
> > +        }
> > +        // SAFETY: `Self` is `repr(transparent)` over `[T]`, so the cast keeps the slice length.
> > +        // INVARIANT: `s` is not empty, checked above.
> > +        Some(unsafe { &mut *(core::ptr::from_mut(s) as *mut Self) })
> > +    }
> > +
> > +    /// Returns a reference to the first element.
> > +    pub fn first(&mut self) -> &T {
> 
> Same here, `first` and `last` should take a non-mutable reference as
> their standard library counterparts do. They do not allow to mutate the
> slice.
> 

Best regards,