* [PATCH 2/3] rust: workqueue: add safe cancellation and status methods
2026-04-02 3:23 [PATCH 0/3] rust: workqueue: add safe cancellation and status methods Aakash Bollineni via B4 Relay
2026-04-02 3:23 ` [PATCH 1/3] rust: helpers: add workqueue helpers Aakash Bollineni via B4 Relay
@ 2026-04-02 3:23 ` Aakash Bollineni via B4 Relay
2026-04-02 3:23 ` [PATCH 3/3] rust: workqueue: add KUnit and sample stress tests Aakash Bollineni via B4 Relay
2026-04-02 11:19 ` [PATCH 0/3] rust: workqueue: add safe cancellation and status methods Onur Özkan
3 siblings, 0 replies; 7+ messages in thread
From: Aakash Bollineni via B4 Relay @ 2026-04-02 3:23 UTC (permalink / raw)
To: rust-for-linux; +Cc: Aakash Bollineni, linux-kernel
From: Aakash Bollineni <aakash.bollineni@multicorewareinc.com>
Modernize the Rust workqueue by adding methods for status checking
and cancellation of work and delayed work items.
Specifically, this patch adds:
- is_pending(): Returns true if the work item is currently enqueued.
- cancel(): Attempts to cancel the work item before it runs. If
successful, it reclaims and returns ownership of the original
pointer (Arc/KBox).
- cancel_sync(): Synchronously cancels the work item, waiting for it
to finish if it's already running. Reclaims ownership if the work
was pending.
To support safe pointer reclamation, a new "Handle-based" model is
introduced via the WorkItemPointer::reclaim method. This ensures that
the "leaked" ownership from an enqueue operation is safely recovered
without memory leaks.
Additionally, we enforce #[repr(transparent)] on core work wrappers to
ensure memory layout compatibility with C pointers and fix pointer
arithmetic in the work_container_of implementation.
Signed-off-by: Aakash Bollineni <aakash.bollineni@multicorewareinc.com>
---
rust/kernel/workqueue.rs | 467 +++++++++++++++++++++++++++++++++++++++++++----
1 file changed, 432 insertions(+), 35 deletions(-)
diff --git a/rust/kernel/workqueue.rs b/rust/kernel/workqueue.rs
index 706e833e9702..3e76164d103d 100644
--- a/rust/kernel/workqueue.rs
+++ b/rust/kernel/workqueue.rs
@@ -448,8 +448,26 @@ pub unsafe trait WorkItemPointer<const ID: u64>: RawWorkItem<ID> {
/// The provided `work_struct` pointer must originate from a previous call to [`__enqueue`]
/// where the `queue_work_on` closure returned true, and the pointer must still be valid.
///
+ /// The implementation must ensure that the pointer is reclaimed (e.g., via `from_raw`)
+ /// before calling the `run` method of the underlying [`WorkItem`].
+ ///
/// [`__enqueue`]: RawWorkItem::__enqueue
unsafe extern "C" fn run(ptr: *mut bindings::work_struct);
+
+ /// Reclaims ownership of the pointer from the work item.
+ ///
+ /// This is called when a work item is successfully cancelled, allowing the caller
+ /// to recover the original pointer (e.g., `Arc` or `KBox`) that was "leaked"
+ /// during enqueuing.
+ ///
+ /// # Safety
+ ///
+ /// The provided `work_struct` pointer must originate from a previous call to [`__enqueue`]
+ /// where the `queue_work_on` closure returned true, and the work item must have been
+ /// successfully cancelled (i.e., `cancel_work` returned true).
+ ///
+ /// [`__enqueue`]: RawWorkItem::__enqueue
+ unsafe fn reclaim(ptr: *mut bindings::work_struct) -> Self;
}
/// Defines the method that should be called when this work item is executed.
@@ -516,6 +534,156 @@ pub fn new(name: &'static CStr, key: Pin<&'static LockClassKey>) -> impl PinInit
})
}
+ /// Returns whether the work item is currently pending.
+ ///
+ /// # Warning
+ ///
+ /// This method is inherently racy. A work item can be enqueued or start running
+ /// immediately after this check returns. Do not rely on this for correctness
+ /// logic (e.g., as a guard for unsafe operations); use it only for debugging or
+ /// non-critical status reporting.
+ ///
+ /// # Examples
+ ///
+ /// ```
+ /// # use kernel::workqueue::{self, new_work, Work, WorkItem, HasWork};
+ /// # use kernel::impl_has_work;
+ /// # use kernel::sync::Arc;
+ /// # #[pin_data]
+ /// # struct MyStruct { #[pin] work: Work<MyStruct> }
+ /// # impl_has_work! { impl HasWork<Self> for MyStruct { self.work } }
+ /// # impl WorkItem for MyStruct {
+ /// # type Pointer = Arc<MyStruct>;
+ /// # fn run(_this: Arc<MyStruct>) {}
+ /// # }
+ /// let my_struct = Arc::pin_init(pin_init!(MyStruct {
+ /// work <- new_work!("MyStruct::work"),
+ /// }), kernel::alloc::flags::GFP_KERNEL).unwrap();
+ /// assert!(!my_struct.work.is_pending());
+ /// workqueue::system().enqueue(my_struct.clone());
+ /// assert!(my_struct.work.is_pending());
+ /// # let _ = my_struct.work.cancel();
+ /// ```
+ #[inline]
+ pub fn is_pending(&self) -> bool {
+ // SAFETY: `self.work` is a valid pointer to a `work_struct`.
+ unsafe { bindings::work_pending(self.work.get()) }
+ }
+
+ /// Cancels the work item.
+ ///
+ /// If the work item was successfully cancelled (i.e., it was pending and had not yet
+ /// started running), the original pointer is reclaimed and returned.
+ ///
+ /// # Guarantees
+ ///
+ /// This method is non-blocking and may return while the work item is still running
+ /// on another CPU. If it returns `None`, the work item might be about to start,
+ /// is currently running, or has already finished.
+ ///
+ /// # Safety
+ ///
+ /// This is safe because ownership is only reclaimed if the kernel confirms (via
+ /// `cancel_work` returning true) that the work item is no longer in any queue and
+ /// will not be executed by the workqueue for this specific enqueue event.
+ ///
+ /// [`cancel_sync`]: Work::cancel_sync
+ ///
+ /// # Examples
+ ///
+ /// ```
+ /// # use kernel::workqueue::{self, new_work, Work, WorkItem, HasWork};
+ /// # use kernel::impl_has_work;
+ /// # use kernel::sync::Arc;
+ /// # #[pin_data]
+ /// # struct MyStruct { #[pin] work: Work<MyStruct> }
+ /// # impl_has_work! { impl HasWork<Self> for MyStruct { self.work } }
+ /// # impl WorkItem for MyStruct {
+ /// # type Pointer = Arc<MyStruct>;
+ /// # fn run(_this: Arc<MyStruct>) {}
+ /// # }
+ /// let my_struct = Arc::pin_init(pin_init!(MyStruct {
+ /// work <- new_work!("MyStruct::work"),
+ /// }), kernel::alloc::flags::GFP_KERNEL).unwrap();
+ /// workqueue::system().enqueue(my_struct.clone());
+ /// assert!(my_struct.work.is_pending());
+ /// let reclaimed = my_struct.work.cancel();
+ /// assert!(reclaimed.is_some());
+ /// assert!(!my_struct.work.is_pending());
+ /// ```
+ pub fn cancel(&self) -> Option<T::Pointer>
+ where
+ T: WorkItem<ID>,
+ {
+ let work_ptr = self.work.get();
+ // SAFETY: `work_ptr` is a valid pointer to a `work_struct`.
+ if unsafe { bindings::cancel_work(work_ptr) } {
+ // SAFETY: The work item was successfully cancelled and is guaranteed not to run,
+ // so we can safely reclaim the ownership leaked during `enqueue`.
+ Some(unsafe { T::Pointer::reclaim(work_ptr) })
+ } else {
+ None
+ }
+ }
+
+ /// Synchronously cancels the work item.
+ ///
+ /// This method waits for the work item to finish if it is currently running.
+ /// If the work item was successfully cancelled from the queue, the pointer is
+ /// reclaimed and returned.
+ ///
+ /// # Guarantees
+ ///
+ /// After this method returns, the work item is guaranteed to be:
+ /// - Not pending in any queue.
+ /// - Not running on any CPU.
+ ///
+ /// This makes it safe to use during teardown (e.g., driver `remove` or object `drop`)
+ /// to ensure no background tasks are accessing resources that are about to be freed.
+ ///
+ /// # Safety
+ ///
+ /// Same as [`cancel`], it only reclaims ownership if the kernel confirms the work
+ /// was still pending and is now removed.
+ ///
+ /// [`cancel`]: Work::cancel
+ ///
+ /// # Examples
+ ///
+ /// ```
+ /// # use kernel::workqueue::{self, new_work, Work, WorkItem, HasWork};
+ /// # use kernel::impl_has_work;
+ /// # use kernel::sync::Arc;
+ /// # #[pin_data]
+ /// # struct MyStruct { #[pin] work: Work<MyStruct> }
+ /// # impl_has_work! { impl HasWork<Self> for MyStruct { self.work } }
+ /// # impl WorkItem for MyStruct {
+ /// # type Pointer = Arc<MyStruct>;
+ /// # fn run(_this: Arc<MyStruct>) {}
+ /// # }
+ /// let my_struct = Arc::pin_init(pin_init!(MyStruct {
+ /// work <- new_work!("MyStruct::work"),
+ /// }), kernel::alloc::flags::GFP_KERNEL).unwrap();
+ /// workqueue::system().enqueue(my_struct.clone());
+ /// let reclaimed = my_struct.work.cancel_sync();
+ /// assert!(reclaimed.is_some());
+ /// ```
+ pub fn cancel_sync(&self) -> Option<T::Pointer>
+ where
+ T: WorkItem<ID>,
+ {
+ let work_ptr = self.work.get();
+ // SAFETY: `work_ptr` is a valid pointer to a `work_struct`.
+ if unsafe { bindings::cancel_work_sync(work_ptr) } {
+ // SAFETY: The work item was successfully cancelled/waited for, and is guaranteed
+ // not to run again unless re-enqueued. We reclaim the ownership leaked during
+ // `enqueue`.
+ Some(unsafe { T::Pointer::reclaim(work_ptr) })
+ } else {
+ None
+ }
+ }
+
/// Get a pointer to the inner `work_struct`.
///
/// # Safety
@@ -674,25 +842,14 @@ pub fn new(
pin_init!(Self {
dwork <- Opaque::ffi_init(|slot: *mut bindings::delayed_work| {
// SAFETY: The `WorkItemPointer` implementation promises that `run` can be used as
- // the work item function.
+ // the work item function. We use the C-helper to ensure the timer function
+ // and data are initialized correctly according to kernel macros.
unsafe {
- bindings::init_work_with_key(
- core::ptr::addr_of_mut!((*slot).work),
+ bindings::init_delayed_work(
+ slot,
Some(T::Pointer::run),
- false,
work_name.as_char_ptr(),
work_key.as_ptr(),
- )
- }
-
- // SAFETY: The `delayed_work_timer_fn` function pointer can be used here because
- // the timer is embedded in a `struct delayed_work`, and only ever scheduled via
- // the core workqueue code, and configured to run in irqsafe context.
- unsafe {
- bindings::timer_init_key(
- core::ptr::addr_of_mut!((*slot).timer),
- Some(bindings::delayed_work_timer_fn),
- bindings::TIMER_IRQSAFE,
timer_name.as_char_ptr(),
timer_key.as_ptr(),
)
@@ -702,6 +859,89 @@ pub fn new(
})
}
+ /// Returns whether the work item is currently pending.
+ ///
+ /// # Warning
+ ///
+ /// This method is inherently racy. See [`Work::is_pending`].
+ ///
+ /// # Examples
+ ///
+ /// See [`Work::is_pending`].
+ ///
+ /// [`Work::is_pending`]: Work::is_pending
+ #[inline]
+ pub fn is_pending(&self) -> bool {
+ // SAFETY: `self.dwork` is reaching into a valid Opaque<bindings::delayed_work>.
+ unsafe {
+ let ptr: *mut bindings::delayed_work = self.dwork.get();
+ bindings::work_pending(core::ptr::addr_of_mut!((*ptr).work))
+ }
+ }
+
+ /// Cancels the delayed work item.
+ ///
+ /// If the work item was successfully cancelled (i.e., it was pending and had not yet
+ /// started running), the original pointer is reclaimed and returned.
+ ///
+ /// # Guarantees
+ ///
+ /// See [`Work::cancel`].
+ ///
+ /// # Safety
+ ///
+ /// Same as [`Work::cancel`].
+ ///
+ /// [`cancel_sync`]: DelayedWork::cancel_sync
+ /// [`Work::cancel`]: Work::cancel
+ ///
+ /// # Examples
+ ///
+ /// See [`Work::cancel`].
+ pub fn cancel(&self) -> Option<T::Pointer>
+ where
+ T: WorkItem<ID>,
+ {
+ let dwork_ptr = self.dwork.get();
+ // SAFETY: `dwork_ptr` is a valid pointer to a `delayed_work`.
+ if unsafe { bindings::cancel_delayed_work(dwork_ptr) } {
+ // SAFETY: The work item was successfully cancelled and is guaranteed not to run,
+ // so we can safely reclaim the ownership leaked during `enqueue`.
+ Some(unsafe { T::Pointer::reclaim(core::ptr::addr_of_mut!((*dwork_ptr).work)) })
+ } else {
+ None
+ }
+ }
+
+ /// Synchronously cancels the delayed work item.
+ ///
+ /// This method waits for the work item to finish if it is currently running.
+ /// If the work item was successfully cancelled from the queue, the pointer is
+ /// reclaimed and returned.
+ ///
+ /// # Guarantees
+ ///
+ /// See [`Work::cancel_sync`].
+ ///
+ /// # Safety
+ ///
+ /// Same as [`Work::cancel_sync`].
+ pub fn cancel_sync(&self) -> Option<T::Pointer>
+ where
+ T: WorkItem<ID>,
+ {
+ let dwork_ptr = self.dwork.get();
+ // SAFETY: `dwork_ptr` is a valid pointer to a `delayed_work`.
+ if unsafe { bindings::cancel_delayed_work_sync(dwork_ptr) } {
+ // SAFETY: The work item was successfully cancelled/waited for, and is guaranteed
+ // not to run again unless re-enqueued. We reclaim the ownership leaked during
+ // `enqueue`.
+ Some(unsafe { T::Pointer::reclaim(core::ptr::addr_of_mut!((*dwork_ptr).work)) })
+ } else {
+ None
+ }
+ }
+
/// Get a pointer to the inner `delayed_work`.
///
/// # Safety
@@ -781,22 +1021,11 @@ unsafe fn raw_get_work(
unsafe fn work_container_of(
ptr: *mut $crate::workqueue::Work<$work_type $(, $id)?>,
) -> *mut Self {
- // SAFETY: The caller promises that the pointer points at a field of the right type
- // in the right kind of struct.
- let ptr = unsafe { $crate::workqueue::Work::raw_get(ptr) };
-
- // SAFETY: The caller promises that the pointer points at a field of the right type
- // in the right kind of struct.
- let delayed_work = unsafe {
- $crate::container_of!(ptr, $crate::bindings::delayed_work, work)
- };
-
- let delayed_work: *mut $crate::workqueue::DelayedWork<$work_type $(, $id)?> =
- delayed_work.cast();
-
- // SAFETY: The caller promises that the pointer points at a field of the right type
- // in the right kind of struct.
- unsafe { $crate::container_of!(delayed_work, Self, $field) }
+ // SAFETY: The caller promises that the pointer points at the `work` field
+ // of a `delayed_work` struct, which is itself the `dwork` field of a
+ // `DelayedWork` wrapper, which is the `$field` field of a `Self` struct.
+ let ptr = ptr.cast::<$crate::workqueue::DelayedWork<$work_type $(, $id)?>>();
+ unsafe { $crate::container_of!(ptr, Self, $field) }
}
}
)*};
@@ -827,6 +1056,15 @@ unsafe impl<T, const ID: u64> WorkItemPointer<ID> for Arc<T>
T::run(arc)
}
+
+ unsafe fn reclaim(ptr: *mut bindings::work_struct) -> Self {
+ // The `__enqueue` method always uses a `work_struct` stored in a `Work<T, ID>`.
+ let ptr = ptr.cast::<Work<T, ID>>();
+ // SAFETY: This computes the pointer that `__enqueue` got from `Arc::into_raw`.
+ let ptr = unsafe { T::work_container_of(ptr) };
+ // SAFETY: This pointer comes from `Arc::into_raw` and we've been given back ownership.
+ unsafe { Arc::from_raw(ptr) }
+ }
}
// SAFETY: The `work_struct` raw pointer is guaranteed to be valid for the duration of the call to
@@ -874,7 +1112,8 @@ unsafe impl<T, const ID: u64> RawDelayedWorkItem<ID> for Arc<T>
{
}
-// SAFETY: TODO.
+// SAFETY: The `WorkItemPointer` implementation for `Pin<KBox<T>>` is safe because `KBox::from_raw`
+// correctly reconstructs the box that was leaked during `enqueue` (via `KBox::into_raw`).
unsafe impl<T, const ID: u64> WorkItemPointer<ID> for Pin<KBox<T>>
where
T: WorkItem<ID, Pointer = Self>,
@@ -883,18 +1122,35 @@ unsafe impl<T, const ID: u64> WorkItemPointer<ID> for Pin<KBox<T>>
unsafe extern "C" fn run(ptr: *mut bindings::work_struct) {
// The `__enqueue` method always uses a `work_struct` stored in a `Work<T, ID>`.
let ptr = ptr.cast::<Work<T, ID>>();
- // SAFETY: This computes the pointer that `__enqueue` got from `Arc::into_raw`.
+ // SAFETY: This computes the pointer that `__enqueue` got from `KBox::into_raw`.
let ptr = unsafe { T::work_container_of(ptr) };
- // SAFETY: This pointer comes from `Arc::into_raw` and we've been given back ownership.
+ // SAFETY: This pointer comes from `KBox::into_raw` and we've been given back ownership.
let boxed = unsafe { KBox::from_raw(ptr) };
// SAFETY: The box was already pinned when it was enqueued.
let pinned = unsafe { Pin::new_unchecked(boxed) };
T::run(pinned)
}
+
+ unsafe fn reclaim(ptr: *mut bindings::work_struct) -> Self {
+ // The `__enqueue` method always uses a `work_struct` stored in a `Work<T, ID>`.
+ let ptr = ptr.cast::<Work<T, ID>>();
+ // SAFETY: This computes the pointer that `__enqueue` got from `KBox::into_raw`.
+ let ptr = unsafe { T::work_container_of(ptr) };
+ // SAFETY: This pointer comes from `KBox::into_raw` and we've been given back ownership.
+ let boxed = unsafe { KBox::from_raw(ptr) };
+ // SAFETY: The box was already pinned when it was enqueued.
+ unsafe { Pin::new_unchecked(boxed) }
+ }
}
-// SAFETY: TODO.
+// SAFETY: The `work_struct` raw pointer is guaranteed to be valid for the duration of the call to
+// the closure because we have exclusive ownership of the `KBox`, and we don't drop it ourselves.
+// If `queue_work_on` returns true, it is further guaranteed to be valid until a call to the
+// function pointer in `work_struct` because we leak the memory it points to, and only reclaim it
+// if the closure returns false (not reachable for KBox as it must succeed) or in
+// `WorkItemPointer::run`, which is what the function pointer in the `work_struct` must be
+// pointing to.
unsafe impl<T, const ID: u64> RawWorkItem<ID> for Pin<KBox<T>>
where
T: WorkItem<ID, Pointer = Self>,
@@ -1022,3 +1278,144 @@ pub fn system_bh_highpri() -> &'static Queue {
// SAFETY: `system_bh_highpri_wq` is a C global, always available.
unsafe { Queue::from_raw(bindings::system_bh_highpri_wq) }
}
+
+#[macros::kunit_tests(rust_kernel_workqueue)]
+mod tests {
+ use super::*;
+ use crate::sync::Arc;
+
+ #[pin_data]
+ struct TestWorkItem {
+ #[pin]
+ work: Work<TestWorkItem>,
+ value: i32,
+ }
+
+ impl_has_work! {
+ impl HasWork<Self> for TestWorkItem { self.work }
+ }
+
+ impl WorkItem for TestWorkItem {
+ type Pointer = Arc<Self>;
+ fn run(_this: Arc<Self>) {}
+ }
+
+ #[pin_data]
+ struct TestDelayedWorkItem {
+ #[pin]
+ delayed_work: DelayedWork<TestDelayedWorkItem>,
+ value: i32,
+ }
+
+ impl_has_delayed_work! {
+ impl HasDelayedWork<Self> for TestDelayedWorkItem { self.delayed_work }
+ }
+
+ impl WorkItem for TestDelayedWorkItem {
+ type Pointer = Arc<Self>;
+ fn run(_this: Arc<Self>) {}
+ }
+
+ /// Helper to get Arc strong count for verification in tests.
+ fn get_arc_count<T: Sized>(arc: &Arc<T>) -> i32 {
+ // SAFETY: ArcInner has refcount as its first field. Arc points to data at DATA_OFFSET.
+ unsafe {
+ let ptr = Arc::as_ptr(arc);
+ let inner_ptr = (ptr as *const u8).sub(Arc::<T>::DATA_OFFSET);
+ let refcount_ptr = inner_ptr as *const i32;
+ core::ptr::read_volatile(refcount_ptr)
+ }
+ }
+
+ #[test]
+ fn test_work_cancel_reclaim() {
+ let item = Arc::pin_init(
+ pin_init!(TestWorkItem {
+ work <- new_work!("TestWorkItem::work"),
+ value: 42,
+ }),
+ GFP_KERNEL,
+ )
+ .expect("Failed to allocate TestWorkItem");
+
+ let initial_count = get_arc_count(&item);
+
+ // Enqueue
+ let _ = system().enqueue(item.clone());
+
+ // Cancel and Reclaim (if it was pending)
+ if let Some(reclaimed) = item.work.cancel() {
+ assert!(get_arc_count(&item) == initial_count + 1);
+ drop(reclaimed);
+ assert!(get_arc_count(&item) == initial_count);
+ }
+ }
+
+ #[test]
+ fn test_work_cancel_sync_reclaim() {
+ let item = Arc::pin_init(
+ pin_init!(TestWorkItem {
+ work <- new_work!("TestWorkItem::work"),
+ value: 42,
+ }),
+ GFP_KERNEL,
+ )
+ .expect("Failed to allocate TestWorkItem");
+
+ let initial_count = get_arc_count(&item);
+
+ // Enqueue
+ let _ = system().enqueue(item.clone());
+
+ // Cancel Sync and Reclaim
+ if let Some(reclaimed) = item.work.cancel_sync() {
+ assert!(get_arc_count(&item) == initial_count + 1);
+ drop(reclaimed);
+ assert!(get_arc_count(&item) == initial_count);
+ }
+ }
+
+ #[test]
+ fn test_work_stress_enqueue_cancel() {
+ let item = Arc::pin_init(
+ pin_init!(TestWorkItem {
+ work <- new_work!("TestWorkItem::work"),
+ value: 42,
+ }),
+ GFP_KERNEL,
+ )
+ .expect("Failed to allocate TestWorkItem");
+
+ for _ in 0..100 {
+ if let Ok(_) = system().enqueue(item.clone()) {
+ let _ = item.work.cancel_sync();
+ }
+ }
+
+ assert_eq!(get_arc_count(&item), 1);
+ }
+
+ #[test]
+ fn test_delayed_work_cancel_reclaim() {
+ let item = Arc::pin_init(
+ pin_init!(TestDelayedWorkItem {
+ delayed_work <- new_delayed_work!("TestDelayedWorkItem::delayed_work"),
+ value: 42,
+ }),
+ GFP_KERNEL,
+ )
+ .expect("Failed to allocate TestDelayedWorkItem");
+
+ let initial_count = get_arc_count(&item);
+
+ // Enqueue delayed
+ let _ = system().enqueue_delayed(item.clone(), 100);
+
+ // Cancel and Reclaim
+ if let Some(reclaimed) = item.delayed_work.cancel() {
+ assert!(get_arc_count(&item) == initial_count + 1);
+ drop(reclaimed);
+ assert!(get_arc_count(&item) == initial_count);
+ }
+ }
+}
--
2.43.0
^ permalink raw reply related [flat|nested] 7+ messages in thread* [PATCH 3/3] rust: workqueue: add KUnit and sample stress tests
2026-04-02 3:23 [PATCH 0/3] rust: workqueue: add safe cancellation and status methods Aakash Bollineni via B4 Relay
2026-04-02 3:23 ` [PATCH 1/3] rust: helpers: add workqueue helpers Aakash Bollineni via B4 Relay
2026-04-02 3:23 ` [PATCH 2/3] rust: workqueue: add safe cancellation and status methods Aakash Bollineni via B4 Relay
@ 2026-04-02 3:23 ` Aakash Bollineni via B4 Relay
2026-04-02 11:23 ` Onur Özkan
2026-04-02 11:34 ` Onur Özkan
2026-04-02 11:19 ` [PATCH 0/3] rust: workqueue: add safe cancellation and status methods Onur Özkan
3 siblings, 2 replies; 7+ messages in thread
From: Aakash Bollineni via B4 Relay @ 2026-04-02 3:23 UTC (permalink / raw)
To: rust-for-linux; +Cc: Aakash Bollineni, linux-kernel
From: Aakash Bollineni <aakash.bollineni@multicorewareinc.com>
To ensure the safety and correctness of the improved workqueue API,
this patch adds comprehensive testing infrastructure:
1. KUnit Tests: Adds an internal 'rust_kernel_workqueue' test suite
to rust/kernel/workqueue.rs. These tests verify basic and
synchronous cancellation, refcount stability, and concurrency
safety for both standard and delayed work.
2. Sample Module: Adds samples/rust/rust_workqueue_test.rs as a
standalone module that performs a 1000-iteration stress test
designed to verify race-free pointer handover during concurrent
enqueue/cancel operations.
The tests use distinct types for standard and delayed work items to
ensure clear trait dispatch and prevent field offset conflicts.
Signed-off-by: Aakash Bollineni <aakash.bollineni@multicorewareinc.com>
---
samples/rust/Makefile | 2 +
samples/rust/rust_workqueue_test.rs | 214 ++++++++++++++++++++++++++++++++++++
2 files changed, 216 insertions(+)
diff --git a/samples/rust/Makefile b/samples/rust/Makefile
index 6c0aaa58cccc..0f304bd90997 100644
--- a/samples/rust/Makefile
+++ b/samples/rust/Makefile
@@ -20,3 +20,5 @@ obj-$(CONFIG_SAMPLE_RUST_SOC) += rust_soc.o
rust_print-y := rust_print_main.o rust_print_events.o
subdir-$(CONFIG_SAMPLE_RUST_HOSTPROGS) += hostprogs
+
+obj-y += rust_workqueue_test.o
diff --git a/samples/rust/rust_workqueue_test.rs b/samples/rust/rust_workqueue_test.rs
new file mode 100644
index 000000000000..c055ec1964c0
--- /dev/null
+++ b/samples/rust/rust_workqueue_test.rs
@@ -0,0 +1,214 @@
+// SPDX-License-Identifier: GPL-2.0
+
+//! Robust stress test for Rust workqueue API.
+
+use kernel::prelude::*;
+use kernel::sync::Arc;
+use kernel::time::msecs_to_jiffies;
+use kernel::workqueue::{self, new_work, Work, WorkItem};
+
+#[pin_data]
+struct TestItem {
+ #[pin]
+ work: Work<TestItem>,
+ value: i32,
+ #[pin]
+ delayed_work: workqueue::DelayedWork<TestItem>,
+}
+
+kernel::impl_has_work! {
+ impl HasWork<Self> for TestItem { self.work }
+}
+
+// SAFETY: The `delayed_work` field is at a fixed offset and is valid for the lifetime of
+// `TestItem`.
+unsafe impl workqueue::HasDelayedWork<Self> for TestItem {}
+
+impl WorkItem for TestItem {
+ type Pointer = Arc<TestItem>;
+
+ fn run(this: Arc<TestItem>) {
+ pr_info!(
+ "Rust workqueue test: Work item running (value: {})\n",
+ this.value
+ );
+ }
+}
+
+/// Helper to get Arc strong count for verification in tests.
+/// This uses internal layout knowledge of Arc.
+fn get_arc_count<T: Sized>(arc: &Arc<T>) -> i32 {
+ // SAFETY: ArcInner has refcount as its first field. Arc points to data at DATA_OFFSET.
+ unsafe {
+ let ptr = Arc::as_ptr(arc);
+ let inner_ptr = (ptr as *const u8).sub(Arc::<T>::DATA_OFFSET);
+ // The first field of ArcInner is Refcount, which is a transparent wrapper around
+ // refcount_t. In the kernel, refcount_t is an atomic_t (i32).
+ let refcount_ptr = inner_ptr as *const i32;
+ // We use a relaxed load to get the current value.
+ core::ptr::read_volatile(refcount_ptr)
+ }
+}
+
+struct RustWorkqueueTest;
+
+impl kernel::Module for RustWorkqueueTest {
+ fn init(_module: &'static ThisModule) -> Result<Self> {
+ pr_info!("Rust workqueue test: starting robust verification\n");
+
+ // 1. Basic Lifecycle with Refcount Validation
+ {
+ let test_item = Arc::pin_init(
+ pin_init!(TestItem {
+ work <- new_work!("TestItem::work"),
+ value: 42,
+ delayed_work <- workqueue::new_delayed_work!("TestItem::delayed_work"),
+ }),
+ GFP_KERNEL,
+ )?;
+
+ let initial_count = get_arc_count(&test_item);
+ pr_info!("Initial Arc strong count: {}\n", initial_count);
+
+ // Enqueue
+ let enqueued_item = test_item.clone();
+
+ if let Err(returned_item) = workqueue::system().enqueue(enqueued_item) {
+ pr_warn!("Work already pending, unexpected!\n");
+ let _ = returned_item;
+ } else {
+ pr_info!(
+ "Work enqueued successfully. Strong count: {}\n",
+ get_arc_count(&test_item)
+ );
+ }
+
+ // Cancel immediately (best effort)
+ if let Some(reclaimed) = test_item.work.cancel() {
+ let count_after_cancel = get_arc_count(&test_item);
+ pr_info!(
+ "Success: Work cancelled and Arc reclaimed. Strong count: {}\n",
+ count_after_cancel
+ );
+
+ // VALIDATION: Reclamation must restore the refcount (minus the clone we just
+ // reclaimed)
+ if count_after_cancel != initial_count + 1 {
+ pr_err!(
+ "ERROR: Refcount mismatch after cancel! Expected {}, got {}\n",
+ initial_count + 1,
+ count_after_cancel
+ );
+ return Err(ENXIO);
+ }
+ drop(reclaimed);
+ if get_arc_count(&test_item) != initial_count {
+ pr_err!(
+ "ERROR: Refcount mismatch after drop! Expected {}, got {}\n",
+ initial_count,
+ get_arc_count(&test_item)
+ );
+ return Err(ENXIO);
+ }
+ } else {
+ pr_info!("Work already running or finished, could not reclaim via cancel().\n");
+ }
+ }
+
+ // 2. Stress Testing: Enqueue/Cancel Sync Loop
+ {
+ pr_info!("Starting stress test (1000 iterations)...\n");
+ let test_item = Arc::pin_init(
+ pin_init!(TestItem {
+ work <- new_work!("TestItem::work"),
+ value: 99,
+ delayed_work <- workqueue::new_delayed_work!("TestItem::delayed_work"),
+ }),
+ GFP_KERNEL,
+ )?;
+
+ for i in 0..1000 {
+ let _ = workqueue::system().enqueue(test_item.clone());
+ // Use cancel_sync to ensure determinism for the next iteration
+ let _ = test_item.work.cancel_sync();
+
+ if i % 250 == 0 {
+ pr_info!("Stress test progress: {}/1000\n", i);
+ }
+ }
+
+ if get_arc_count(&test_item) != 1 {
+ pr_err!(
+ "ERROR: Refcount leak detected after stress test! count: {}\n",
+ get_arc_count(&test_item)
+ );
+ return Err(ENXIO);
+ } else {
+ pr_info!("Stress test completed successfully. No refcount leaks.\n");
+ }
+ }
+
+ // 3. Delayed Work Cancellation Test
+ {
+ let test_item = Arc::pin_init(
+ pin_init!(TestItem {
+ work <- new_work!("TestItem::work"),
+ value: 7,
+ delayed_work <- workqueue::new_delayed_work!("TestItem::delayed_work"),
+ }),
+ GFP_KERNEL,
+ )?;
+
+ let initial_count = get_arc_count(&test_item);
+
+ // Schedule with a long delay
+ let to_enqueue = test_item.clone();
+ let res = workqueue::system()
+ .enqueue_delayed(to_enqueue, msecs_to_jiffies(5000));
+ if let Err(returned) = res {
+ pr_warn!("Delayed work already pending, returned item.\n");
+ drop(returned);
+ } else {
+ pr_info!("Delayed work enqueued. count: {}\n", get_arc_count(&test_item));
+ }
+
+ if test_item.delayed_work.is_pending() {
+ pr_info!("Delayed work is pending as expected.\n");
+ }
+
+ if let Some(reclaimed) = test_item.delayed_work.cancel() {
+ pr_info!("Success: Delayed work reclaimed. No leak.\n");
+ drop(reclaimed);
+ } else {
+ pr_warn!("Notice: Delayed work not reclaimed (running or never enqueued).\n");
+ }
+
+ let final_count = get_arc_count(&test_item);
+ if final_count != initial_count {
+ pr_err!(
+ "ERROR: Refcount leak after delayed cancel! expected {}, got {}\n",
+ initial_count,
+ final_count
+ );
+ return Err(ENXIO);
+ }
+ }
+
+ pr_info!("Rust workqueue test: all robust checks passed\n");
+ Ok(RustWorkqueueTest)
+ }
+}
+
+impl Drop for RustWorkqueueTest {
+ fn drop(&mut self) {
+ pr_info!("Rust workqueue test: exit\n");
+ }
+}
+
+module! {
+ type: RustWorkqueueTest,
+ name: "rust_workqueue_test",
+ authors: ["Aakash Bollineni"],
+ description: "Robust stress test for Rust workqueue API",
+ license: "GPL",
+}
--
2.43.0
^ permalink raw reply related [flat|nested] 7+ messages in thread