While rebasing rvkms I noticed that timers I was setting seemed to have
pretty random timer values that amounted slightly over 2x the time value I
set each time. After a lot of debugging, I finally managed to figure out
why: it seems that since we moved to Instant and Delta, we mistakenly
began passing the clocksource ID to hrtimer_start_range_ns, when we should
be passing the timer mode instead. Presumably, this works fine for simple
relative timers - but immediately breaks on other types of timers.
So, fix this by passing the ID for the timer mode instead.
Signed-off-by: Lyude Paul <lyude@redhat.com>
Acked-by: Andreas Hindborg <a.hindborg@kernel.org>
Reviewed-by: FUJITA Tomonori <fujita.tomonori@gmail.com>
Fixes: e0c0ab04f6 ("rust: time: Make HasHrTimer generic over HrTimerMode")
Link: https://lore.kernel.org/r/20250710225129.670051-1-lyude@redhat.com
[ Removed cast, applied `rustfmt`, fixed `Fixes:` tag. - Miguel ]
Signed-off-by: Miguel Ojeda <ojeda@kernel.org>
Pull timekeeping updates from Andreas Hindborg:
- Make 'Instant' generic over clock source. This allows the compiler to
assert that arithmetic expressions involving the 'Instant' use
'Instants' based on the same clock source.
- Make 'HrTimer' generic over the timer mode. 'HrTimer' timers take a
'Duration' or an 'Instant' when setting the expiry time, depending on
the timer mode. With this change, the compiler can check the type
matches the timer mode.
- Add an abstraction for 'fsleep'. 'fsleep' is a flexible sleep
function that will select an appropriate sleep method depending on
the requested sleep time.
- Avoid 64-bit divisions on 32-bit hardware when calculating
timestamps.
- Seal the 'HrTimerMode' trait. This prevents users of the
'HrTimerMode' from implementing the trait on their own types.
* tag 'rust-timekeeping-for-v6.17' of https://github.com/Rust-for-Linux/linux:
rust: time: Add wrapper for fsleep() function
rust: time: Seal the HrTimerMode trait
rust: time: Remove Ktime in hrtimer
rust: time: Make HasHrTimer generic over HrTimerMode
rust: time: Add HrTimerExpires trait
rust: time: Replace HrTimerMode enum with trait-based mode types
rust: time: Add ktime_get() to ClockSource trait
rust: time: Make Instant generic over ClockSource
rust: time: Replace ClockId enum with ClockSource trait
rust: time: Avoid 64-bit integer division on 32-bit architectures
Pull alloc and DMA updates from Danilo Krummrich:
Box:
- Implement Borrow / BorrowMut for Box<T, A>.
Vec:
- Implement Default for Vec<T, A>.
- Implement Borrow / BorrowMut for Vec<T, A>.
DMA:
- Clarify wording and be consistent in 'coherent' nomenclature.
- Convert the read!() / write!() macros to return a Result.
- Add as_slice() / write() methods in CoherentAllocation.
- Fix doc-comment of dma_handle().
- Expose count() and size() in CoherentAllocation and add the
corresponding type invariants.
- Implement CoherentAllocation::dma_handle_with_offset().
- Require mutable reference for as_slice_mut() and write().
MAINTAINERS:
- Add Vlastimil Babka, Liam R. Howlett, Uladzislau Rezki and Lorenzo
Stoakes as reviewers (thanks everyone).
* tag 'alloc-next-v6.17-2025-07-15' of https://github.com/Rust-for-Linux/linux:
MAINTAINERS: add mm folks as reviewers to rust alloc
rust: dma: require mutable reference for as_slice_mut() and write()
rust: dma: add dma_handle_with_offset method to CoherentAllocation
rust: dma: expose the count and size of CoherentAllocation
rust: dma: fix doc-comment of dma_handle()
rust: dma: add as_slice/write functions for CoherentAllocation
rust: dma: convert the read/write macros to return Result
rust: dma: clarify wording and be consistent in `coherent` nomenclature
rust: alloc: implement `Borrow` and `BorrowMut` for `KBox`
rust: alloc: implement `Borrow` and `BorrowMut` for `Vec`
rust: vec: impl Default for Vec with any allocator
This patch is being sent for use in the various Rust GPU drivers that
are under development. It provides the additional feature of work items
that are executed after a delay.
The design of the existing workqueue is rather extensible, as most of
the logic is reused for delayed work items even though a different work
item type is required. The new logic consists of:
* A new DelayedWork struct that wraps struct delayed_work.
* A new impl_has_delayed_work! macro that provides adjusted versions of
the container_of logic, that is suitable with delayed work items.
* A `enqueue_delayed` method that can enqueue a delayed work item.
This patch does *not* rely on the fact that `struct delayed_work`
contains `struct work_struct` at offset zero. It will continue to work
even if the layout is changed to hold the `work` field at a different
offset.
Please see the example introduced at the top of the file for example
usage of delayed work items.
Acked-by: Tejun Heo <tj@kernel.org>
Reviewed-by: Boqun Feng <boqun.feng@gmail.com>
Signed-off-by: Alice Ryhl <aliceryhl@google.com>
Link: https://lore.kernel.org/r/20250711-workqueue-delay-v3-1-3fe17b18b9d1@google.com
[ Replaced `as _` with `as ffi::c_int` to clean warning. - Miguel ]
Signed-off-by: Miguel Ojeda <ojeda@kernel.org>
In the previous patch we added Opaque::cast_from() that performs the
opposite operation to Opaque::raw_get(). For consistency with this
naming, rename raw_get() to cast_from().
There are a few other options such as calling cast_from() something
closer to raw_get() rather than renaming this method. However, I could
not find a great naming scheme that works with raw_get(). The previous
version of this patch used from_raw(), but functions of that name
typically have a different signature, so that's not a great option.
Suggested-by: Danilo Krummrich <dakr@kernel.org>
Signed-off-by: Alice Ryhl <aliceryhl@google.com>
Acked-by: Benno Lossin <lossin@kernel.org>
Acked-by: Andreas Hindborg <a.hindborg@kernel.org>
Acked-by: Boqun Feng <boqun.feng@gmail.com>
Reviewed-by: Danilo Krummrich <dakr@kernel.org>
Acked-by: Danilo Krummrich <dakr@kernel.org>
Link: https://lore.kernel.org/r/20250624-opaque-from-raw-v2-2-e4da40bdc59c@google.com
[ Removed `HrTimer::raw_get` change. - Miguel ]
Signed-off-by: Miguel Ojeda <ojeda@kernel.org>
Since commit b20fbbc08a ("rust: check type of `$ptr` in
`container_of!`") we have enforced that the field pointer passed to
container_of! must match the declared field. This caused mismatches when
using a pointer to bindings::x for fields of type Opaque<bindings::x>.
This situation encourages the user to simply pass field.cast() to the
container_of! macro, but this is not great because you might
accidentally pass a *mut bindings::y when the field type is
Opaque<bindings::x>, which would be wrong.
To help catch this kind of mistake, add a new Opaque::cast_from that
wraps a raw pointer in Opaque without changing the inner type. Also
update the docs to reflect this as well as some existing users.
Signed-off-by: Alice Ryhl <aliceryhl@google.com>
Acked-by: Andreas Hindborg <a.hindborg@kernel.org>
Acked-by: Boqun Feng <boqun.feng@gmail.com>
Reviewed-by: Danilo Krummrich <dakr@kernel.org>
Acked-by: Danilo Krummrich <dakr@kernel.org>
Link: https://lore.kernel.org/r/20250624-opaque-from-raw-v2-1-e4da40bdc59c@google.com
Signed-off-by: Miguel Ojeda <ojeda@kernel.org>
Currently, Rust code uses a typedef for unsigned long to represent
userspace addresses. This is unfortunate because it means that userspace
addresses could accidentally be mixed up with other integers. To
alleviate that, we introduce a new UserPtr struct that wraps a raw
pointer to represent a userspace address. By using a struct, type
checking enforces that userspace addresses cannot be mixed up with
anything else.
This is similar to the __user annotation in C that detects cases where
user pointers are mixed with non-user pointers.
Note that unlike __user pointers in C, this type is just a pointer
without a target type. This means that it can't detect cases such as
mixing up which struct this user pointer references. However, that is
okay due to the way this is intended to be used - generally, you create
a UserPtr in your ioctl callback from the provided usize *before*
dispatching on which ioctl is in use, and then after dispatching on the
ioctl you pass the UserPtr into a UserSliceReader or UserSliceWriter;
selecting the target type does not happen until you have obtained the
UserSliceReader/Writer.
The UserPtr type is not marked with #[derive(Debug)], which means that
it's not possible to print values of this type. This avoids ASLR
leakage.
The type is added to the prelude as it is a fairly fundamental type
similar to c_int. The wrapping_add() method is renamed to
wrapping_byte_add() for consistency with the method name found on raw
pointers.
Reviewed-by: Benno Lossin <lossin@kernel.org>
Reviewed-by: Danilo Krummrich <dakr@kernel.org>
Reviewed-by: Christian Schrefl <chrisi.schrefl@gmail.com>
Reviewed-by: Boqun Feng <boqun.feng@gmail.com>
Reviewed-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Signed-off-by: Alice Ryhl <aliceryhl@google.com>
Link: https://lore.kernel.org/r/20250616-userptr-newtype-v3-1-5ff7b2d18d9e@google.com
[ Reworded title. - Miguel ]
Signed-off-by: Miguel Ojeda <ojeda@kernel.org>
This patch adds a more convenient method for reading C strings from
userspace. Logic is added to NUL-terminate the buffer when necessary so
that a &CStr can be returned.
Note that we treat attempts to read past `self.length` as a fault, so
this returns EFAULT if that limit is exceeded before `buf.len()` is
reached.
Reviewed-by: Danilo Krummrich <dakr@kernel.org>
Signed-off-by: Alice Ryhl <aliceryhl@google.com>
Reviewed-by: Benno Lossin <lossin@kernel.org>
Link: https://lore.kernel.org/r/20250616-strncpy-from-user-v5-2-2d3fb0e1f5af@google.com
[ Use `from_mut` to clean `clippy::ref_as_ptr` lint. Reworded
title. - Miguel ]
Signed-off-by: Miguel Ojeda <ojeda@kernel.org>
This patch adds a direct wrapper around the C function of the same name.
It's not really intended for direct use by Rust code since
strncpy_from_user has a somewhat unfortunate API where it only
nul-terminates the buffer if there's space for the nul-terminator. This
means that a direct Rust wrapper around it could not return a &CStr
since the buffer may not be a cstring. However, we still add the method
to build more convenient APIs on top of it, which will happen in
subsequent patches.
Reviewed-by: Danilo Krummrich <dakr@kernel.org>
Reviewed-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Reviewed-by: Boqun Feng <boqun.feng@gmail.com>
Reviewed-by: Benno Lossin <lossin@kernel.org>
Signed-off-by: Alice Ryhl <aliceryhl@google.com>
Link: https://lore.kernel.org/r/20250616-strncpy-from-user-v5-1-2d3fb0e1f5af@google.com
[ Reworded title. - Miguel ]
Signed-off-by: Miguel Ojeda <ojeda@kernel.org>
Pull pin-init updates from Benno Lossin:
"Added:
- 'impl<T, E> [Pin]Init<T, E> for Result<T, E>', so results are now
(pin-)initializers.
- 'Zeroable::init_zeroed()' delegating to 'init_zeroed()'.
- New 'zeroed()', a safe version of 'mem::zeroed()' and also provide
it via 'Zeroable::zeroed()'.
- Implement 'Zeroable' for 'Option<&T>' and 'Option<&mut T>'.
- Implement 'Zeroable' for 'Option<[unsafe] [extern "abi"]
fn(...args...) -> ret>' for '"Rust"' and '"C"' ABIs and up to 20
arguments.
Changed:
- Blanket impls of 'Init' and 'PinInit' from 'impl<T, E>
[Pin]Init<T, E> for T' to 'impl<T> [Pin]Init<T> for T'.
- Renamed 'zeroed()' to 'init_zeroed()'.
Upstream dev news:
- More CI improvements to deny warnings, use '--all-targets'. Also
check the synchronization status of the two '-next' branches in
upstream and the kernel."
Acked-by: Andreas Hindborg <a.hindborg@kernel.org>
* tag 'pin-init-v6.17' of https://github.com/Rust-for-Linux/linux:
rust: pin-init: examples, tests: use `ignore` instead of conditionally compiling tests
rust: init: remove doctest's `Error::from_errno` workaround
rust: init: re-enable doctests
rust: pin-init: implement `ZeroableOption` for function pointers with up to 20 arguments
rust: pin-init: change `impl Zeroable for Option<NonNull<T>>` to `ZeroableOption for NonNull<T>`
rust: pin-init: implement `ZeroableOption` for `&T` and `&mut T`
rust: pin-init: add `zeroed()` & `Zeroable::zeroed()` functions
rust: pin-init: add `Zeroable::init_zeroed`
rust: pin-init: rename `zeroed` to `init_zeroed`
rust: pin-init: feature-gate the `stack_init_reuse` test on the `std` feature
rust: pin-init: examples: pthread_mutex: disable the main test for miri
rust: pin-init: examples, tests: add conditional compilation in order to compile under any feature combination
rust: pin-init: change blanket impls for `[Pin]Init` and add one for `Result<T, E>`
rust: pin-init: improve safety documentation for `impl<T> [Pin]Init<T> for T`
Implement `Borrow<T>` and `BorrowMut<T>` for `UniqueArc<T>`, and
`Borrow<T>` for `Arc<T>`. This allows these containers to be used in
generic APIs asking for types implementing those traits. `T` and `&mut
T` also implement those traits allowing users to use either owned,
shared or borrowed values.
`ForeignOwnable` makes a call to its own `borrow` method which must be
disambiguated.
Reviewed-by: Alice Ryhl <aliceryhl@google.com>
Reviewed-by: Benno Lossin <lossin@kernel.org>
Signed-off-by: Alexandre Courbot <acourbot@nvidia.com>
Link: https://lore.kernel.org/r/20250616-borrow_impls-v4-2-36f9beb3fe6a@nvidia.com
Signed-off-by: Miguel Ojeda <ojeda@kernel.org>
Given the safety requirements of as_slice_mut() and write() taking an
immutable reference is technically not incorrect.
However, let's leverage the compiler's capabilities and require a
mutable reference to ensure exclusive access.
This also fixes a clippy warning introduced with 1.88:
warning: mutable borrow from immutable input(s)
--> rust/kernel/dma.rs:297:78
|
297 | pub unsafe fn as_slice_mut(&self, offset: usize, count: usize) -> Result<&mut [T]> {
| ^^^^^^^^
Fixes: d37a39f607 ("rust: dma: add as_slice/write functions for CoherentAllocation")
Reviewed-by: Alice Ryhl <aliceryhl@google.com>
Reviewed-by: Andreas Hindborg <a.hindborg@kernel.org>
Reviewed-by: Abdiel Janulgue <abdiel.janulgue@gmail.com>
Reviewed-by: Alexandre Courbot <acourbot@nvidia.com>
Link: https://lore.kernel.org/r/20250628165120.90149-1-dakr@kernel.org
Signed-off-by: Danilo Krummrich <dakr@kernel.org>
In Rust Binder I need to be able to determine whether a red/black tree
is empty. Thus, add a method for that operation to replace
rbtree.iter().next().is_none()
This is terrible, so add a method for this purpose. We do not add a
RBTree::len method because computing the number of elements requires
iterating the entire tree, but checking whether it is empty can be done
cheaply.
Signed-off-by: Alice Ryhl <aliceryhl@google.com>
Reviewed-by: Benno Lossin <lossin@kernel.org>
Link: https://lore.kernel.org/r/20250616-rbtree-is-empty-v1-1-61f7cfb012e3@google.com
[ Adjusted title. - Miguel ]
Signed-off-by: Miguel Ojeda <ojeda@kernel.org>
Add a `TimerMode` associated type to the `HasHrTimer` trait to
represent the operational mode of the timer, such as absolute or
relative expiration. This new type must implement the `HrTimerMode`
trait, which defines how expiration values are interpreted.
Update the `start()` method to accept an `expires` parameter of type
`<Self::TimerMode as HrTimerMode>::Expires` instead of the fixed `Ktime`.
This enables different timer modes to provide strongly typed expiration
values, such as `Instant<C>` or `Delta`.
The `impl_has_hr_timer` macro is also extended to allow specifying the
`HrTimerMode`. In the following example, it guarantees that the
`start()` method for `Foo` only accepts `Instant<Monotonic>`. Using a
`Delta` or an `Instant` with a different clock source will result in a
compile-time error:
struct Foo {
#[pin]
timer: HrTimer<Self>,
}
impl_has_hr_timer! {
impl HasHrTimer<Self> for Foo {
mode : AbsoluteMode<Monotonic>,
field : self.timer
}
}
This design eliminates runtime mismatches between expires types and
clock sources, and enables stronger type-level guarantees throughout
hrtimer.
Signed-off-by: FUJITA Tomonori <fujita.tomonori@gmail.com>
Reviewed-by: Andreas Hindborg <a.hindborg@kernel.org>
Link: https://lore.kernel.org/r/20250610132823.3457263-5-fujita.tomonori@gmail.com
[ changed conversion method names to `as_*` - Andreas ]
Signed-off-by: Andreas Hindborg <a.hindborg@kernel.org>
Introduce the `HrTimerExpires` trait to represent types that can be
used as expiration values for high-resolution timers. Define a
required method, `into_nanos()`, which returns the expiration time as a
raw nanosecond value suitable for use with C's hrtimer APIs.
Also extend the `HrTimerMode` to use the `HrTimerExpires` trait.
This refactoring is a preparation for enabling hrtimer code to work
uniformly with both absolute and relative expiration modes.
Reviewed-by: Andreas Hindborg <a.hindborg@kernel.org>
Signed-off-by: FUJITA Tomonori <fujita.tomonori@gmail.com>
Link: https://lore.kernel.org/r/20250610132823.3457263-4-fujita.tomonori@gmail.com
[ changed conversion method names to `as_*` - Andreas ]
Signed-off-by: Andreas Hindborg <a.hindborg@kernel.org>
Replace the `HrTimerMode` enum with a trait-based approach that uses
zero-sized types to represent each mode of operation. Each mode now
implements the `HrTimerMode` trait.
This refactoring is a preparation for replacing raw `Ktime` in HrTimer
with the `Instant` and `Delta` types, and for making `HrTimer` generic
over a `ClockSource`.
Reviewed-by: Andreas Hindborg <a.hindborg@kernel.org>
Signed-off-by: FUJITA Tomonori <fujita.tomonori@gmail.com>
Link: https://lore.kernel.org/r/20250610132823.3457263-3-fujita.tomonori@gmail.com
Signed-off-by: Andreas Hindborg <a.hindborg@kernel.org>
Lyude Paul
Miguel Ojeda
Alice Ryhl
Andreas Hindborg
Onur Özkan
Krishna Ketan Rai
Lorenzo Stoakes
Alexandre Courbot
Danilo Krummrich
FUJITA Tomonori