lists.openwall.net   lists  /  announce  owl-users  owl-dev  john-users  john-dev  passwdqc-users  yescrypt  popa3d-users  /  oss-security  kernel-hardening  musl  sabotage  tlsify  passwords  /  crypt-dev  xvendor  /  Bugtraq  Full-Disclosure  linux-kernel  linux-netdev  linux-ext4  linux-hardening  linux-cve-announce  PHC 
Open Source and information security mailing list archives
 
Hash Suite: Windows password security audit tool. GUI, reports in PDF.
[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <CABVgOS=2f3Yg8Wb7qxneRC_+s-W_TQey083niujpZD3fYcfL_w@mail.gmail.com>
Date: Sat, 15 Feb 2025 17:03:38 +0800
From: David Gow <davidgow@...gle.com>
To: Tamir Duberstein <tamird@...il.com>
Cc: Miguel Ojeda <ojeda@...nel.org>, José Expósito <jose.exposito89@...il.com>, 
	Rae Moar <rmoar@...gle.com>, Boqun Feng <boqun.feng@...il.com>, 
	Alex Gaynor <alex.gaynor@...il.com>, Gary Guo <gary@...yguo.net>, 
	Benno Lossin <benno.lossin@...ton.me>, Björn Roy Baron <bjorn3_gh@...tonmail.com>, 
	Alice Ryhl <aliceryhl@...gle.com>, Matt Gilbride <mattgilbride@...gle.com>, 
	Brendan Higgins <brendan.higgins@...ux.dev>, kunit-dev@...glegroups.com, 
	linux-kselftest@...r.kernel.org, rust-for-linux@...r.kernel.org, 
	linux-kernel@...r.kernel.org
Subject: Re: [PATCH v6 3/3] rust: kunit: allow to know if we are in a test

On Fri, 14 Feb 2025 at 22:41, Tamir Duberstein <tamird@...il.com> wrote:
>
> On Fri, Feb 14, 2025 at 2:42 AM David Gow <davidgow@...gle.com> wrote:
> >
> > From: José Expósito <jose.exposito89@...il.com>
> >
> > In some cases, we need to call test-only code from outside the test
> > case, for example, to mock a function or a module.
> >
> > In order to check whether we are in a test or not, we need to test if
> > `CONFIG_KUNIT` is set.
> > Unfortunately, we cannot rely only on this condition because:
> > - a test could be running in another thread,
> > - some distros compile KUnit in production kernels, so checking at runtime
> >   that `current->kunit_test != NULL` is required.
> >
> > Forturately, KUnit provides an optimised check in
> > `kunit_get_current_test()`, which checks CONFIG_KUNIT, a global static
> > key, and then the current thread's running KUnit test.
> >
> > Add a safe wrapper function around this to know whether or not we are in
> > a KUnit test and examples showing how to mock a function and a module.
> >
> > Signed-off-by: José Expósito <jose.exposito89@...il.com>
> > Co-developed-by: David Gow <davidgow@...gle.com>
> > Signed-off-by: David Gow <davidgow@...gle.com>
> > Co-developed-by: Miguel Ojeda <ojeda@...nel.org>
> > Signed-off-by: Miguel Ojeda <ojeda@...nel.org>
> > ---
> >
> > Changes since v5:
> > https://lore.kernel.org/all/20241213081035.2069066-4-davidgow@google.com/
> > - Greatly improved documentation, which is both clearer and better
> >   matches the rustdoc norm. (Thanks, Miguel)
> > - The examples and safety comments are also both more idiomatic an
> >   cleaner. (Thanks, Miguel)
> > - More things sit appropriately behind CONFIG_KUNIT (Thanks, Miguel)
> >
> > Changes since v4:
> > https://lore.kernel.org/linux-kselftest/20241101064505.3820737-4-davidgow@google.com/
> > - Rebased against 6.13-rc1
> > - Fix some missing safety comments, and remove some unneeded 'unsafe'
> >   blocks. (Thanks Boqun)
> >
> > Changes since v3:
> > https://lore.kernel.org/linux-kselftest/20241030045719.3085147-8-davidgow@google.com/
> > - The example test has been updated to no longer use assert_eq!() with
> >   a constant bool argument (fixes a clippy warning).
> >
> > No changes since v2:
> > https://lore.kernel.org/linux-kselftest/20241029092422.2884505-4-davidgow@google.com/
> >
> > Changes since v1:
> > https://lore.kernel.org/lkml/20230720-rustbind-v1-3-c80db349e3b5@google.com/
> > - Rebased on top of rust-next.
> > - Use the `kunit_get_current_test()` C function, which wasn't previously
> >   available, instead of rolling our own.
> > - (Thanks also to Boqun for suggesting a nicer way of implementing this,
> >   which I tried, but the `kunit_get_current_test()` version obsoleted.)
> > ---
> >  rust/kernel/kunit.rs | 66 ++++++++++++++++++++++++++++++++++++++++++++
> >  1 file changed, 66 insertions(+)
> >
> > diff --git a/rust/kernel/kunit.rs b/rust/kernel/kunit.rs
> > index 9e27b74a605b..3aad7a281b6d 100644
> > --- a/rust/kernel/kunit.rs
> > +++ b/rust/kernel/kunit.rs
> > @@ -286,11 +286,77 @@ macro_rules! kunit_unsafe_test_suite {
> >      };
> >  }
> >
> > +/// Returns whether we are currently running a KUnit test.
> > +///
> > +/// In some cases, you need to call test-only code from outside the test case, for example, to
> > +/// create a function mock. This function allows to change behavior depending on whether we are
> > +/// currently running a KUnit test or not.
> > +///
> > +/// # Examples
> > +///
> > +/// This example shows how a function can be mocked to return a well-known value while testing:
> > +///
> > +/// ```
> > +/// # use kernel::kunit::in_kunit_test;
> > +/// fn fn_mock_example(n: i32) -> i32 {
> > +///     if in_kunit_test() {
> > +///         return 100;
> > +///     }
> > +///
> > +///     n + 1
> > +/// }
> > +///
> > +/// let mock_res = fn_mock_example(5);
> > +/// assert_eq!(mock_res, 100);
> > +/// ```
> > +///
> > +/// Sometimes, you don't control the code that needs to be mocked. This example shows how the
> > +/// `bindings` module can be mocked:
>
> [`bindings`] here, please. There are two more instances below but
> those aren't doc comments, so I don't think bracketing them will do
> anything.
>

Done in v7. Alas, I'll have to keep getting used to the differences
between kerneldoc and rustdoc...

> > +///
> > +/// ```
> > +/// // Import our mock naming it as the real module.
> > +/// #[cfg(CONFIG_KUNIT)]
> > +/// use bindings_mock_example as bindings;
> > +/// #[cfg(not(CONFIG_KUNIT))]
> > +/// use kernel::bindings;
> > +///
> > +/// // This module mocks `bindings`.
> > +/// #[cfg(CONFIG_KUNIT)]
> > +/// mod bindings_mock_example {
> > +///     /// Mock `ktime_get_boot_fast_ns` to return a well-known value when running a KUnit test.
> > +///     pub(crate) fn ktime_get_boot_fast_ns() -> u64 {
> > +///         1234
> > +///     }
> > +/// }
> > +///
> > +/// // This is the function we want to test. Since `bindings` has been mocked, we can use its
> > +/// // functions seamlessly.
> > +/// fn get_boot_ns() -> u64 {
> > +///     // SAFETY: `ktime_get_boot_fast_ns()` is always safe to call.
> > +///     unsafe { bindings::ktime_get_boot_fast_ns() }
> > +/// }
> > +///
> > +/// let time = get_boot_ns();
> > +/// assert_eq!(time, 1234);
> > +/// ```
>
> Isn't this swapping out the bindings module at compile time, and for
> the whole build? In other words cfg(CONFIG_KUNIT) will apply to all
> code, both test and non-test.
>

I believe so, so this is probably something best done only in test files.

Ideally, we'd have support for something like the KUnit function
mocking features here, but that's horribly C-specific at the moment.

> > +pub fn in_kunit_test() -> bool {
> > +    // SAFETY: `kunit_get_current_test()` is always safe to call (it has fallbacks for
> > +    // when KUnit is not enabled).
> > +    unsafe { !bindings::kunit_get_current_test().is_null() }
>
> Nit if you care about reducing unsafe blocks:
>
> !unsafe { bindings::kunit_get_current_test() }.is_null()
>
>

Huh, I thought this wouldn't work, but it's working fine for me here,
so I've made the change.

Thanks!

> > +}
> > +
> >  #[kunit_tests(rust_kernel_kunit)]
> >  mod tests {
> > +    use super::*;
> > +
> >      #[test]
> >      fn rust_test_kunit_example_test() {
> >          #![expect(clippy::eq_op)]
> >          assert_eq!(1 + 1, 2);
> >      }
> > +
> > +    #[test]
> > +    fn rust_test_kunit_in_kunit_test() {
> > +        assert!(in_kunit_test());
> > +    }
> >  }
> > --
> > 2.48.1.601.g30ceb7b040-goog
> >
> >

Thanks a lot, these should be fixed in v7.

Cheers,
-- David

Download attachment "smime.p7s" of type "application/pkcs7-signature" (5294 bytes)

Powered by blists - more mailing lists

Powered by Openwall GNU/*/Linux Powered by OpenVZ