linux.git/rust/kernel/sync, branch v6.18.21

rust: sync: refcount: always inline functions using build_assert with arguments

2026-02-06T15:57:41+00:00

commit d6ff6e870077ae0f01a6f860ca1e4a5a825dc032 upstream.

`build_assert` relies on the compiler to optimize out its error path.
Functions using it with its arguments must thus always be inlined,
otherwise the error path of `build_assert` might not be optimized out,
triggering a build error.

Cc: stable@vger.kernel.org
Fixes: bb38f35b35f9 ("rust: implement `kernel::sync::Refcount`")
Reviewed-by: Daniel Almeida 
Signed-off-by: Alexandre Courbot 
Acked-by: Boqun Feng 
Link: https://patch.msgid.link/20251208-io-build-assert-v3-5-98aded02c1ea@nvidia.com
Signed-off-by: Miguel Ojeda 
Signed-off-by: Greg Kroah-Hartman

rust: sync: atomic: Provide stub for `rusttest` 32-bit hosts

2026-02-06T15:57:41+00:00

commit bd36f6e2abf7f85644f7ea8deb1de4040b03bbc1 upstream.

For arm32, on a x86_64 builder, running the `rusttest` target yields:

    error[E0080]: evaluation of constant value failed
      --> rust/kernel/static_assert.rs:37:23
       |
    37 |         const _: () = ::core::assert!($condition $(,$arg)?);
       |                       ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ the evaluated program panicked at 'assertion failed: size_of::() == size_of::()', rust/kernel/sync/atomic/predefine.rs:68:1
       |
      ::: rust/kernel/sync/atomic/predefine.rs:68:1
       |
    68 | static_assert!(size_of::() == size_of::());
       | -------------------------------------------------------------------- in this macro invocation
       |
       = note: this error originates in the macro `::core::assert` which comes from the expansion of the macro `static_assert` (in Nightly builds, run with -Z macro-backtrace for more info)

The reason is that `rusttest` runs on the host, so for e.g. a x86_64
builder `isize` is 64 bits but it is not a `CONFIG_64BIT` build.

Fix it by providing a stub for `rusttest` as usual.

Fixes: 84c6d36bcaf9 ("rust: sync: atomic: Add Atomic<{usize,isize}>")
Cc: stable@vger.kernel.org
Reviewed-by: Onur Özkan 
Acked-by: Boqun Feng 
Link: https://patch.msgid.link/20260123233432.22703-1-ojeda@kernel.org
Signed-off-by: Miguel Ojeda 
Signed-off-by: Greg Kroah-Hartman

rust: condvar: fix broken intra-doc link

2025-11-02T22:18:42+00:00

The future move of pin-init to `syn` uncovers the following broken
intra-doc link:

    error: unresolved link to `crate::pin_init`
      --> rust/kernel/sync/condvar.rs:39:40
       |
    39 | /// instances is with the [`pin_init`](crate::pin_init!) and [`new_condvar`] macros.
       |                                        ^^^^^^^^^^^^^^^^ no item named `pin_init` in module `kernel`
       |
       = note: `-D rustdoc::broken-intra-doc-links` implied by `-D warnings`
       = help: to override `-D warnings` add `#[allow(rustdoc::broken_intra_doc_links)]`

Currently, when rendered, the link points to a literal `crate::pin_init!`
URL.

Thus fix it.

Cc: stable@vger.kernel.org
Fixes: 129e97be8e28 ("rust: pin-init: fix documentation links")
Reviewed-by: Alice Ryhl 
Link: https://patch.msgid.link/20251029073344.349341-1-ojeda@kernel.org
Signed-off-by: Miguel Ojeda

Merge tag 'rust-6.18' of git://git.kernel.org/pub/scm/linux/kernel/git/ojeda/linux

2025-10-01T02:12:49+00:00

Pull rust updates from Miguel Ojeda:
 "Toolchain and infrastructure:

   - Derive 'Zeroable' for all structs and unions generated by 'bindgen'
     where possible and corresponding cleanups. To do so, add the
     'pin-init' crate as a dependency to 'bindings' and 'uapi'.

     It also includes its first use in the 'cpufreq' module, with more
     to come in the next cycle.

   - Add warning to the 'rustdoc' target to detect broken 'srctree/'
     links and fix existing cases.

   - Remove support for unused (since v6.16) host '#[test]'s,
     simplifying the 'rusttest' target. Tests should generally run
     within KUnit.

  'kernel' crate:

   - Add 'ptr' module with a new 'Alignment' type, which is always a
     power of two and is used to validate that a given value is a valid
     alignment and to perform masking and alignment operations:

         // Checked at build time.
         assert_eq!(Alignment::new::<16>().as_usize(), 16);

         // Checked at runtime.
         assert_eq!(Alignment::new_checked(15), None);

         assert_eq!(Alignment::of::().log2(), 0);

         assert_eq!(0x25u8.align_down(Alignment::new::<0x10>()), 0x20);
         assert_eq!(0x5u8.align_up(Alignment::new::<0x10>()), Some(0x10));
         assert_eq!(u8::MAX.align_up(Alignment::new::<0x10>()), None);

     It also includes its first use in Nova.

   - Add 'core::mem::{align,size}_of{,_val}' to the prelude, matching
     Rust 1.80.0.

   - Keep going with the steps on our migration to the standard library
     'core::ffi::CStr' type (use 'kernel::{fmt, prelude::fmt!}' and use
     upstream method names).

   - 'error' module: improve 'Error::from_errno' and 'to_result'
     documentation, including examples/tests.

   - 'sync' module: extend 'aref' submodule documentation now that it
     exists, and more updates to complete the ongoing move of 'ARef' and
     'AlwaysRefCounted' to 'sync::aref'.

   - 'list' module: add an example/test for 'ListLinksSelfPtr' usage.

   - 'alloc' module:

      - Implement 'Box::pin_slice()', which constructs a pinned slice of
        elements.

      - Provide information about the minimum alignment guarantees of
        'Kmalloc', 'Vmalloc' and 'KVmalloc'.

      - Take minimum alignment guarantees of allocators for
        'ForeignOwnable' into account.

      - Remove the 'allocator_test' (including 'Cmalloc').

      - Add doctest for 'Vec::as_slice()'.

      - Constify various methods.

   - 'time' module:

      - Add methods on 'HrTimer' that can only be called with exclusive
        access to an unarmed timer, or from timer callback context.

      - Add arithmetic operations to 'Instant' and 'Delta'.

      - Add a few convenience and access methods to 'HrTimer' and
        'Instant'.

  'macros' crate:

   - Reduce collections in 'quote!' macro.

  And a few other cleanups and improvements"

* tag 'rust-6.18' of git://git.kernel.org/pub/scm/linux/kernel/git/ojeda/linux: (58 commits)
  gpu: nova-core: use Alignment for alignment-related operations
  rust: add `Alignment` type
  rust: macros: reduce collections in `quote!` macro
  rust: acpi: use `core::ffi::CStr` method names
  rust: of: use `core::ffi::CStr` method names
  rust: net: use `core::ffi::CStr` method names
  rust: miscdevice: use `core::ffi::CStr` method names
  rust: kunit: use `core::ffi::CStr` method names
  rust: firmware: use `core::ffi::CStr` method names
  rust: drm: use `core::ffi::CStr` method names
  rust: cpufreq: use `core::ffi::CStr` method names
  rust: configfs: use `core::ffi::CStr` method names
  rust: auxiliary: use `core::ffi::CStr` method names
  drm/panic: use `core::ffi::CStr` method names
  rust: device: use `kernel::{fmt,prelude::fmt!}`
  rust: sync: use `kernel::{fmt,prelude::fmt!}`
  rust: seq_file: use `kernel::{fmt,prelude::fmt!}`
  rust: kunit: use `kernel::{fmt,prelude::fmt!}`
  rust: file: use `kernel::{fmt,prelude::fmt!}`
  rust: device: use `kernel::{fmt,prelude::fmt!}`
  ...

rust: sync: use `kernel::{fmt,prelude::fmt!}`

2025-09-16T07:26:59+00:00

Reduce coupling to implementation details of the formatting machinery by
avoiding direct use for `core`'s formatting traits and macros.

Acked-by: Greg Kroah-Hartman 
Reviewed-by: Alice Ryhl 
Reviewed-by: Benno Lossin 
Signed-off-by: Tamir Duberstein 
Signed-off-by: Miguel Ojeda

rust: block: convert `block::mq` to use `Refcount`

2025-09-15T07:38:36+00:00

Currently there's a custom reference counting in `block::mq`, which uses
`AtomicU64` Rust atomics, and this type doesn't exist on some 32-bit
architectures. We cannot just change it to use 32-bit atomics, because
doing so will make it vulnerable to refcount overflow. So switch it to
use the kernel refcount `kernel::sync::Refcount` instead.

There is an operation needed by `block::mq`, atomically decreasing
refcount from 2 to 0, which is not available through refcount.h, so
I exposed `Refcount::as_atomic` which allows accessing the refcount
directly.

[boqun: Adopt the LKMM atomic API]
Signed-off-by: Gary Guo 
Signed-off-by: Boqun Feng 
Signed-off-by: Peter Zijlstra (Intel) 
Reviewed-by: Benno Lossin 
Reviewed-by: Elle Rhumsaa 
Acked-by: Andreas Hindborg 
Tested-by: David Gow 
Link: https://lore.kernel.org/r/20250723233312.3304339-5-gary@kernel.org

rust: convert `Arc` to use `Refcount`

2025-09-15T07:38:35+00:00

With `Refcount` type created, `Arc` can use `Refcount` instead of
calling into FFI directly.

Signed-off-by: Gary Guo 
Signed-off-by: Boqun Feng 
Signed-off-by: Peter Zijlstra (Intel) 
Reviewed-by: Alexandre Courbot 
Reviewed-by: Benno Lossin 
Reviewed-by: Elle Rhumsaa 
Link: https://lore.kernel.org/r/20250723233312.3304339-4-gary@kernel.org

rust: make `Arc::into_unique_or_drop` associated function

2025-09-15T07:38:35+00:00

Make `Arc::into_unique_or_drop` to become a mere associated function
instead of a method (i.e. removing the `self` receiver).

It's a general convention for Rust smart pointers to avoid having
methods defined on them, because if the pointee type has a method of the
same name, then it is shadowed. This is normally for avoiding semver
breakage, which isn't an issue for kernel codebase, but it's still
generally a good practice to follow this rule, so that `ptr.foo()` would
always be calling a method on the pointee type.

Signed-off-by: Gary Guo 
Signed-off-by: Boqun Feng 
Signed-off-by: Peter Zijlstra (Intel) 
Reviewed-by: Benno Lossin 
Reviewed-by: Alexandre Courbot 
Reviewed-by: Elle Rhumsaa 
Link: https://lore.kernel.org/r/20250723233312.3304339-3-gary@kernel.org

rust: implement `kernel::sync::Refcount`

2025-09-15T07:38:35+00:00

This is a wrapping layer of `include/linux/refcount.h`. Currently the
kernel refcount has already been used in `Arc`, however it calls into
FFI directly.

[boqun: Add the missing <> for the link in comment]
Signed-off-by: Gary Guo 
Signed-off-by: Boqun Feng 
Signed-off-by: Peter Zijlstra (Intel) 
Reviewed-by: Alice Ryhl 
Reviewed-by: Boqun Feng 
Reviewed-by: Fiona Behrens 
Reviewed-by: Benno Lossin 
Reviewed-by: Elle Rhumsaa 
Link: https://lore.kernel.org/r/20250723233312.3304339-2-gary@kernel.org

rust: sync: Add memory barriers

2025-09-15T07:38:34+00:00

Memory barriers are building blocks for concurrent code, hence provide
a minimal set of them.

The compiler barrier, barrier(), is implemented in inline asm instead of
using core::sync::atomic::compiler_fence() because memory models are
different: kernel's atomics are implemented in inline asm therefore the
compiler barrier should be implemented in inline asm as well. Also it's
currently only public to the kernel crate until there's a reasonable
driver usage.

Signed-off-by: Boqun Feng 
Signed-off-by: Peter Zijlstra (Intel) 
Reviewed-by: Alice Ryhl 
Reviewed-by: Elle Rhumsaa 
Link: https://lore.kernel.org/all/20250719030827.61357-10-boqun.feng@gmail.com/