[PATCH v2] rust: DmaFence: Remove static lifetime

[PATCH v2] rust: DmaFence: Remove static lifetime

[Philipp Stanner]

A FenceCallbackRegistration can stem from another party than the one
that has created a Fence. Should that party forget the registration
object (for example through a refcount cycle) and then unload the
module, a fence signaling would run into the unloaded module, causing
UAF bugs.

So far, this has been solved by demanding that the payload data of the
registration object demanding static lifetime.

It turns out, however, that this is harmful because the static lifetime
bubbles up to all users, ultimately potentially causing a large amount
of driver data to be static, which renders the lifetime obsolete.

Solve this issue instead through an unsafe requirement which demands
that the user does not forget the registration object. This is also the
solution chosen by ScopedWork.

Suggested-by: Danilo Krummrich <dakr@kernel.org>
Signed-off-by: Philipp Stanner <phasta@kernel.org>
Reviewed-by: Onur Özkan <work@onurozkan.dev>
---
Changes in v2:
  - Safety Docu: Be more precise about what must not be forgotten.
    (Gary)
---
 rust/kernel/dma_buf/dma_fence.rs | 16 ++++++++++++----
 1 file changed, 12 insertions(+), 4 deletions(-)

diff --git a/rust/kernel/dma_buf/dma_fence.rs b/rust/kernel/dma_buf/dma_fence.rs
index 18a43e1bb442..a1d851656562 100644
--- a/rust/kernel/dma_buf/dma_fence.rs
+++ b/rust/kernel/dma_buf/dma_fence.rs
@@ -291,7 +291,7 @@ fn from(e: AllocError) -> Self {
 ///     }
 /// }
 /// ```
-pub trait FenceCallback: Send + 'static {
+pub trait FenceCallback: Send {
     /// Called when the fence is signaled.
     ///
     /// This is called from the fence signaling path, which may be in interrupt
@@ -310,7 +310,7 @@ pub trait FenceCallback: Send + 'static {
 /// When this object is dropped, the callback is automatically removed if it
 /// hasn't been called yet.
 #[pin_data(PinnedDrop)]
-pub struct FenceCallbackRegistration<T: FenceCallback + 'static> {
+pub struct FenceCallbackRegistration<T: FenceCallback> {
     #[pin]
     callback_foreign: Opaque<bindings::dma_fence_cb>,
     callback: ManuallyDrop<T>,
@@ -326,7 +326,14 @@ impl<T: FenceCallback> FenceCallbackRegistration<T> {
     /// On success the callback is pinned in place and will fire when the fence
     /// signals. On `AlreadySignaled` the callback is returned to the caller so
     /// that owned resources can be reclaimed.
-    pub fn new<'a>(fence: &'a Fence, callback: T) -> impl PinInit<Self, CallbackError<T>> + 'a
+    ///
+    /// # Safety
+    ///
+    /// The callback registration must not be forgotten.
+    pub unsafe fn new<'a>(
+        fence: &'a Fence,
+        callback: T,
+    ) -> impl PinInit<Self, CallbackError<T>> + 'a
     where
         T: 'a,
     {
@@ -693,7 +700,8 @@ struct DriverFenceData<'a, T: Send + Sync + FenceContextOps> {
 ///
 /// let cb_data = CallbackData { };
 /// let waiting_fence = ARef::from(fence.as_fence());
-/// let cb_reg = FenceCallbackRegistration::new(&waiting_fence, cb_data);
+/// // SAFETY: `cb_data` is not forgotten.
+/// let cb_reg = unsafe { FenceCallbackRegistration::new(&waiting_fence, cb_data) };
 /// let cb_reg = KBox::pin_init(cb_reg, GFP_KERNEL)?;
 ///
 /// // TODO signalling guards

base-commit: 6bd5eaeb9827ca7ba953b68fb27948fdb3e40630
-- 
2.55.0

Re: [PATCH v2] rust: DmaFence: Remove static lifetime

[Philipp Stanner]

On Wed, 2026-09-30 at 14:27 +0200, Philipp Stanner wrote:
> A FenceCallbackRegistration can stem from another party than the one
> that has created a Fence. Should that party forget the registration
> object (for example through a refcount cycle) and then unload the
> module, a fence signaling would run into the unloaded module, causing
> UAF bugs.
> 
> So far, this has been solved by demanding that the payload data of the
> registration object demanding static lifetime.
> 
> It turns out, however, that this is harmful because the static lifetime
> bubbles up to all users, ultimately potentially causing a large amount
> of driver data to be static, which renders the lifetime obsolete.
> 
> Solve this issue instead through an unsafe requirement which demands
> that the user does not forget the registration object. This is also the
> solution chosen by ScopedWork.
> 
> Suggested-by: Danilo Krummrich <dakr@kernel.org>
> Signed-off-by: Philipp Stanner <phasta@kernel.org>
> Reviewed-by: Onur Özkan <work@onurozkan.dev> 

Gentle ping about the merge since we need this to upstream
drm::JobQueue


P.

Re: [PATCH v2] rust: DmaFence: Remove static lifetime

[Alice Ryhl]

On Wed, Sep 30, 2026 at 02:27:56PM +0200, Philipp Stanner wrote:
> A FenceCallbackRegistration can stem from another party than the one
> that has created a Fence. Should that party forget the registration
> object (for example through a refcount cycle) and then unload the
> module, a fence signaling would run into the unloaded module, causing
> UAF bugs.
> 
> So far, this has been solved by demanding that the payload data of the
> registration object demanding static lifetime.
> 
> It turns out, however, that this is harmful because the static lifetime
> bubbles up to all users, ultimately potentially causing a large amount
> of driver data to be static, which renders the lifetime obsolete.
> 
> Solve this issue instead through an unsafe requirement which demands
> that the user does not forget the registration object. This is also the
> solution chosen by ScopedWork.
> 
> Suggested-by: Danilo Krummrich <dakr@kernel.org>
> Signed-off-by: Philipp Stanner <phasta@kernel.org>
> Reviewed-by: Onur Özkan <work@onurozkan.dev>

Reviewed-by: Alice Ryhl <aliceryhl@google.com>

Applied to drm-rust-next. Thanks!