diff options
Diffstat (limited to 'Documentation/rust')
| -rw-r--r-- | Documentation/rust/arch-support.rst | 1 | ||||
| -rw-r--r-- | Documentation/rust/coding-guidelines.rst | 258 | ||||
| -rw-r--r-- | Documentation/rust/quick-start.rst | 61 | ||||
| -rw-r--r-- | Documentation/rust/testing.rst | 89 |
4 files changed, 401 insertions, 8 deletions
diff --git a/Documentation/rust/arch-support.rst b/Documentation/rust/arch-support.rst index 54be7ddf3e57..6e6a515d0899 100644 --- a/Documentation/rust/arch-support.rst +++ b/Documentation/rust/arch-support.rst @@ -15,6 +15,7 @@ support corresponds to ``S`` values in the ``MAINTAINERS`` file. ============= ================ ============================================== Architecture Level of support Constraints ============= ================ ============================================== +``arm`` Maintained ARMv7 Little Endian only. ``arm64`` Maintained Little Endian only. ``loongarch`` Maintained \- ``riscv`` Maintained ``riscv64`` and LLVM/Clang only. diff --git a/Documentation/rust/coding-guidelines.rst b/Documentation/rust/coding-guidelines.rst index 329b070a1d47..3198be3a6d63 100644 --- a/Documentation/rust/coding-guidelines.rst +++ b/Documentation/rust/coding-guidelines.rst @@ -38,6 +38,81 @@ Like ``clang-format`` for the rest of the kernel, ``rustfmt`` works on individual files, and does not require a kernel configuration. Sometimes it may even work with broken code. +Imports +~~~~~~~ + +``rustfmt``, by default, formats imports in a way that is prone to conflicts +while merging and rebasing, since in some cases it condenses several items into +the same line. For instance: + +.. code-block:: rust + + // Do not use this style. + use crate::{ + example1, + example2::{example3, example4, example5}, + example6, example7, + example8::example9, + }; + +Instead, the kernel uses a vertical layout that looks like this: + +.. code-block:: rust + + use crate::{ + example1, + example2::{ + example3, + example4, + example5, // + }, + example6, + example7, + example8::example9, // + }; + +That is, each item goes into its own line, and braces are used as soon as there +is more than one item in a list. + +The trailing empty comment allows to preserve this formatting. Not only that, +``rustfmt`` will actually reformat imports vertically when the empty comment is +added. That is, it is possible to easily reformat the original example into the +expected style by running ``rustfmt`` on an input like: + +.. code-block:: rust + + // Do not use this style. + use crate::{ + example1, + example2::{example3, example4, example5, // + }, + example6, example7, + example8::example9, // + }; + +The trailing empty comment works for nested imports, as shown above, as well as +for single item imports -- this can be useful to minimize diffs within patch +series: + +.. code-block:: rust + + use crate::{ + example1, // + }; + +The trailing empty comment works in any of the lines within the braces, but it +is preferred to keep it in the last item, since it is reminiscent of the +trailing comma in other formatters. Sometimes it may be simpler to avoid moving +the comment several times within a patch series due to changes in the list. + +There may be cases where exceptions may need to be made, i.e. none of this is +a hard rule. There is also code that is not migrated to this style yet, but +please do not introduce code in other styles. + +Eventually, the goal is to get ``rustfmt`` to support this formatting style (or +a similar one) automatically in a stable release without requiring the trailing +empty comment. Thus, at some point, the goal is to remove those comments. + Comments -------- @@ -85,6 +160,18 @@ written after the documentation, e.g.: // ... } +This applies to both public and private items. This increases consistency with +public items, allows changes to visibility with less changes involved and will +allow us to potentially generate the documentation for private items as well. +In other words, if documentation is written for a private item, then ``///`` +should still be used. For instance: + +.. code-block:: rust + + /// My private function. + // TODO: ... + fn f() {} + One special kind of comments are the ``// SAFETY:`` comments. These must appear before every ``unsafe`` block, and they explain why the code inside the block is correct/sound, i.e. why it cannot trigger undefined behavior in any case, e.g.: @@ -191,6 +278,23 @@ or: /// [`struct mutex`]: srctree/include/linux/mutex.h +C FFI types +----------- + +Rust kernel code refers to C types, such as ``int``, using type aliases such as +``c_int``, which are readily available from the ``kernel`` prelude. Please do +not use the aliases from ``core::ffi`` -- they may not map to the correct types. + +These aliases should generally be referred directly by their identifier, i.e. +as a single segment path. For instance: + +.. code-block:: rust + + fn f(p: *const c_char) -> c_int { + // ... + } + + Naming ------ @@ -227,3 +331,157 @@ The equivalent in Rust may look like (ignoring documentation): That is, the equivalent of ``GPIO_LINE_DIRECTION_IN`` would be referred to as ``gpio::LineDirection::In``. In particular, it should not be named ``gpio::gpio_line_direction::GPIO_LINE_DIRECTION_IN``. + + +Lints +----- + +In Rust, it is possible to ``allow`` particular warnings (diagnostics, lints) +locally, making the compiler ignore instances of a given warning within a given +function, module, block, etc. + +It is similar to ``#pragma GCC diagnostic push`` + ``ignored`` + ``pop`` in C +[#]_: + +.. code-block:: c + + #pragma GCC diagnostic push + #pragma GCC diagnostic ignored "-Wunused-function" + static void f(void) {} + #pragma GCC diagnostic pop + +.. [#] In this particular case, the kernel's ``__{always,maybe}_unused`` + attributes (C23's ``[[maybe_unused]]``) may be used; however, the example + is meant to reflect the equivalent lint in Rust discussed afterwards. + +But way less verbose: + +.. code-block:: rust + + #[allow(dead_code)] + fn f() {} + +By that virtue, it makes it possible to comfortably enable more diagnostics by +default (i.e. outside ``W=`` levels). In particular, those that may have some +false positives but that are otherwise quite useful to keep enabled to catch +potential mistakes. + +On top of that, Rust provides the ``expect`` attribute which takes this further. +It makes the compiler warn if the warning was not produced. For instance, the +following will ensure that, when ``f()`` is called somewhere, we will have to +remove the attribute: + +.. code-block:: rust + + #[expect(dead_code)] + fn f() {} + +If we do not, we get a warning from the compiler:: + + warning: this lint expectation is unfulfilled + --> x.rs:3:10 + | + 3 | #[expect(dead_code)] + | ^^^^^^^^^ + | + = note: `#[warn(unfulfilled_lint_expectations)]` on by default + +This means that ``expect``\ s do not get forgotten when they are not needed, which +may happen in several situations, e.g.: + +- Temporary attributes added while developing. + +- Improvements in lints in the compiler, Clippy or custom tools which may + remove a false positive. + +- When the lint is not needed anymore because it was expected that it would be + removed at some point, such as the ``dead_code`` example above. + +It also increases the visibility of the remaining ``allow``\ s and reduces the +chance of misapplying one. + +Thus prefer ``expect`` over ``allow`` unless: + +- Conditional compilation triggers the warning in some cases but not others. + + If there are only a few cases where the warning triggers (or does not + trigger) compared to the total number of cases, then one may consider using + a conditional ``expect`` (i.e. ``cfg_attr(..., expect(...))``). Otherwise, + it is likely simpler to just use ``allow``. + +- Inside macros, when the different invocations may create expanded code that + triggers the warning in some cases but not in others. + +- When code may trigger a warning for some architectures but not others, such + as an ``as`` cast to a C FFI type. + +As a more developed example, consider for instance this program: + +.. code-block:: rust + + fn g() {} + + fn main() { + #[cfg(CONFIG_X)] + g(); + } + +Here, function ``g()`` is dead code if ``CONFIG_X`` is not set. Can we use +``expect`` here? + +.. code-block:: rust + + #[expect(dead_code)] + fn g() {} + + fn main() { + #[cfg(CONFIG_X)] + g(); + } + +This would emit a lint if ``CONFIG_X`` is set, since it is not dead code in that +configuration. Therefore, in cases like this, we cannot use ``expect`` as-is. + +A simple possibility is using ``allow``: + +.. code-block:: rust + + #[allow(dead_code)] + fn g() {} + + fn main() { + #[cfg(CONFIG_X)] + g(); + } + +An alternative would be using a conditional ``expect``: + +.. code-block:: rust + + #[cfg_attr(not(CONFIG_X), expect(dead_code))] + fn g() {} + + fn main() { + #[cfg(CONFIG_X)] + g(); + } + +This would ensure that, if someone introduces another call to ``g()`` somewhere +(e.g. unconditionally), then it would be spotted that it is not dead code +anymore. However, the ``cfg_attr`` is more complex than a simple ``allow``. + +Therefore, it is likely that it is not worth using conditional ``expect``\ s when +more than one or two configurations are involved or when the lint may be +triggered due to non-local changes (such as ``dead_code``). + +For more information about diagnostics in Rust, please see: + + https://doc.rust-lang.org/stable/reference/attributes/diagnostics.html + +Error handling +-------------- + +For some background and guidelines about Rust for Linux specific error handling, +please see: + + https://rust.docs.kernel.org/kernel/error/type.Result.html#error-codes-in-c-and-rust diff --git a/Documentation/rust/quick-start.rst b/Documentation/rust/quick-start.rst index 2d107982c87b..152289f0bed2 100644 --- a/Documentation/rust/quick-start.rst +++ b/Documentation/rust/quick-start.rst @@ -39,8 +39,8 @@ of the box, e.g.:: Debian ****** -Debian Testing and Debian Unstable (Sid), outside of the freeze period, provide -recent Rust releases and thus they should generally work out of the box, e.g.:: +Debian 13 (Trixie), as well as Testing and Debian Unstable (Sid) provide recent +Rust releases and thus they should generally work out of the box, e.g.:: apt install rustc rust-src bindgen rustfmt rust-clippy @@ -87,6 +87,61 @@ they should generally work out of the box, e.g.:: zypper install rust rust1.79-src rust-bindgen clang +Ubuntu +****** + +25.04 +~~~~~ + +The latest Ubuntu releases provide recent Rust releases and thus they should +generally work out of the box, e.g.:: + + apt install rustc rust-src bindgen rustfmt rust-clippy + +In addition, ``RUST_LIB_SRC`` needs to be set, e.g.:: + + RUST_LIB_SRC=/usr/src/rustc-$(rustc --version | cut -d' ' -f2)/library + +For convenience, ``RUST_LIB_SRC`` can be exported to the global environment. + + +24.04 LTS and older +~~~~~~~~~~~~~~~~~~~ + +Though Ubuntu 24.04 LTS and older versions still provide recent Rust +releases, they require some additional configuration to be set, using +the versioned packages, e.g.:: + + apt install rustc-1.80 rust-1.80-src bindgen-0.65 rustfmt-1.80 \ + rust-1.80-clippy + ln -s /usr/lib/rust-1.80/bin/rustfmt /usr/bin/rustfmt-1.80 + ln -s /usr/lib/rust-1.80/bin/clippy-driver /usr/bin/clippy-driver-1.80 + +None of these packages set their tools as defaults; therefore they should be +specified explicitly, e.g.:: + + make LLVM=1 RUSTC=rustc-1.80 RUSTDOC=rustdoc-1.80 RUSTFMT=rustfmt-1.80 \ + CLIPPY_DRIVER=clippy-driver-1.80 BINDGEN=bindgen-0.65 + +Alternatively, modify the ``PATH`` variable to place the Rust 1.80 binaries +first and set ``bindgen`` as the default, e.g.:: + + PATH=/usr/lib/rust-1.80/bin:$PATH + update-alternatives --install /usr/bin/bindgen bindgen \ + /usr/bin/bindgen-0.65 100 + update-alternatives --set bindgen /usr/bin/bindgen-0.65 + +``RUST_LIB_SRC`` needs to be set when using the versioned packages, e.g.:: + + RUST_LIB_SRC=/usr/src/rustc-$(rustc-1.80 --version | cut -d' ' -f2)/library + +For convenience, ``RUST_LIB_SRC`` can be exported to the global environment. + +In addition, ``bindgen-0.65`` is available in newer releases (24.04 LTS and +24.10), but it may not be available in older ones (20.04 LTS and 22.04 LTS), +thus ``bindgen`` may need to be built manually (please see below). + + Requirements: Building ---------------------- @@ -128,7 +183,7 @@ Rust standard library source **************************** The Rust standard library source is required because the build system will -cross-compile ``core`` and ``alloc``. +cross-compile ``core``. If ``rustup`` is being used, run:: diff --git a/Documentation/rust/testing.rst b/Documentation/rust/testing.rst index 568b71b415a4..f43cb77bcc69 100644 --- a/Documentation/rust/testing.rst +++ b/Documentation/rust/testing.rst @@ -97,7 +97,7 @@ operator are also supported as usual, e.g.: /// ``` /// # use kernel::{spawn_work_item, workqueue}; - /// spawn_work_item!(workqueue::system(), || pr_info!("x"))?; + /// spawn_work_item!(workqueue::system(), || pr_info!("x\n"))?; /// # Ok::<(), Error>(()) /// ``` @@ -123,16 +123,95 @@ A current limitation is that KUnit does not support assertions in other tasks. Thus, we presently simply print an error to the kernel log if an assertion actually failed. Additionally, doctests are not run for nonpublic functions. +Since these tests are examples, i.e. they are part of the documentation, they +should generally be written like "real code". Thus, for example, instead of +using ``unwrap()`` or ``expect()``, use the ``?`` operator. For more background, +please see: + + https://rust.docs.kernel.org/kernel/error/type.Result.html#error-codes-in-c-and-rust + The ``#[test]`` tests --------------------- -Additionally, there are the ``#[test]`` tests. These can be run using the -``rusttest`` Make target:: +Additionally, there are the ``#[test]`` tests. Like for documentation tests, +these are also fairly similar to what you would expect from userspace, and they +are also mapped to KUnit. + +These tests are introduced by the ``kunit_tests`` procedural macro, which takes +the name of the test suite as an argument. + +For instance, assume we want to test the function ``f`` from the documentation +tests section. We could write, in the same file where we have our function: + +.. code-block:: rust + + #[kunit_tests(rust_kernel_mymod)] + mod tests { + use super::*; + + #[test] + fn test_f() { + assert_eq!(f(10, 20), 30); + } + } + +And if we run it, the kernel log would look like:: + + KTAP version 1 + # Subtest: rust_kernel_mymod + # speed: normal + 1..1 + # test_f.speed: normal + ok 1 test_f + ok 1 rust_kernel_mymod + +Like documentation tests, the ``assert!`` and ``assert_eq!`` macros are mapped +back to KUnit and do not panic. Similarly, the +`? <https://doc.rust-lang.org/reference/expressions/operator-expr.html#the-question-mark-operator>`_ +operator is supported, i.e. the test functions may return either nothing (i.e. +the unit type ``()``) or ``Result`` (i.e. any ``Result<T, E>``). For instance: + +.. code-block:: rust + + #[kunit_tests(rust_kernel_mymod)] + mod tests { + use super::*; + + #[test] + fn test_g() -> Result { + let x = g()?; + assert_eq!(x, 30); + Ok(()) + } + } + +If we run the test and the call to ``g`` fails, then the kernel log would show:: + + KTAP version 1 + # Subtest: rust_kernel_mymod + # speed: normal + 1..1 + # test_g: ASSERTION FAILED at rust/kernel/lib.rs:335 + Expected is_test_result_ok(test_g()) to be true, but is false + # test_g.speed: normal + not ok 1 test_g + not ok 1 rust_kernel_mymod + +If a ``#[test]`` test could be useful as an example for the user, then please +use a documentation test instead. Even edge cases of an API, e.g. error or +boundary cases, can be interesting to show in examples. + +The ``rusttest`` host tests +--------------------------- + +These are userspace tests that can be built and run in the host (i.e. the one +that performs the kernel build) using the ``rusttest`` Make target:: make LLVM=1 rusttest -This requires the kernel ``.config``. It runs the ``#[test]`` tests on the host -(currently) and thus is fairly limited in what these tests can test. +This requires the kernel ``.config``. + +Currently, they are mostly used for testing the ``macros`` crate's examples. The Kselftests -------------- |